Systemic treatments for eczema: a network meta‐analysis

Abstract

Background

Eczema is a common and chronic, relapsing, inflammatory skin disorder. It seriously impacts quality of life and economic outcomes, especially for those with moderate to severe eczema. Various treatments allow sustained control of the disease; however, their relative benefit remains unclear due to the limited number of trials directly comparing treatments.

Objectives

To assess the comparative efficacy and safety of different types of systemic immunosuppressive treatments for moderate to severe eczema using NMA and to generate rankings of available systemic immunosuppressive treatments for eczema according to their efficacy and safety.

Search methods

We searched the following databases up to August 2019: the Cochrane Skin Specialised Register, CENTRAL, MEDLINE, and Embase.

Selection criteria

All randomised controlled trials (RCTs) of systemic immunosuppressive agents for moderate to severe atopic eczema when compared against placebo or any other eligible eczema treatment.

Data collection and analysis

We synthesised data using pair‐wise analysis and NMA to compare treatments and rank them according to their effectiveness.

Effectiveness was assessed primarily by determining the proportion of participants who achieved at least 75% improvement in the Eczema Area and Severity Index (EASI75) and improvement in the Patient‐Oriented Eczema Measure (POEM). Safety was evaluated primarily by considering the proportion of participants with serious adverse events (SAEs) and infection.

We deemed short‐term follow‐up as ≤ 16 weeks and long‐term follow‐up as > 16 weeks.

We assessed the certainty of the body of evidence from the NMA for these primary outcomes using six domains of CiNEMA grading.

Main results

We included a total of 74 studies, with 8177 randomised participants. Approximately 55% of participants were male, with average age of 32 years (range 2 to 84 years), although age and gender were unreported for 419 and 902 participants, respectively. Most of the included trials were placebo controlled (65%), 34% were head‐to‐head studies (15% assessed the effects of different doses of the same drug), and 1% were multi‐armed studies with both an active comparator and a placebo.

All trials included participants with moderate to severe eczema, but 62% of studies did not separate data by severity; 38% of studies assessed only severe eczema. The total duration of included trials ranged from 2 weeks to 60 months, whereas treatment duration varied from a single dose (CIM331, KPL‐716) to 60 months (methotrexate (MTX)).

Seventy studies were available for quantitative synthesis; this review assessed 29 immunosuppressive agents from three classes of interventions. These included (1) conventional treatments, with ciclosporin assessed most commonly; (2) small molecule treatments, including phosphodiesterase (PDE)‐4 inhibitors, tyrosine kinase inhibitors, and Janus kinase (JAK) inhibitors; and (3) biological treatments, including anti‐CD31 receptors, anti‐interleukin (IL)‐22, anti‐IL‐31, anti‐IL‐13, anti‐IL‐12/23p40, anti‐OX40, anti‐TSLP, anti‐CRTH2, and anti‐immunoglobulin E (IgE) monoclonal antibodies, but most commonly dupilumab.

Most trials (73) assessed outcomes at a short‐term duration ranging from 2 to 16 weeks, whereas 33 trials assessed long‐term outcomes, with duration ranging from 5 to 60 months. All participants were from a hospital setting. Fifty‐two studies declared a source of funding, and of these, pharmaceutical companies funded 88%. We rated 37 studies as high risk; 21, unclear risk, and 16, low risk of bias, with studies most commonly at high risk of attrition bias.

Network meta‐analysis suggests that dupilumab ranks first for effectiveness when compared with other biological treatments. Dupilumab is more effective than placebo in achieving EASI75 (risk ratio (RR) 3.04, 95% confidence interval (CI) 2.51 to 3.69) and improvement in POEM score (mean difference 7.30, 95% CI 6.61 to 8.00) at short‐term follow‐up (high‐certainty evidence).

Very low‐certainty evidence means we are uncertain of the effects of dupilumab when compared with placebo, in terms of the proportion of participants who achieve EASI75 (RR 2.59, 95% CI 1.87 to 3.60) at longer‐term follow‐up.

Low‐certainty evidence indicates that tralokinumab may be more effective than placebo in achieving short‐term EASI75 (RR 2.54, 95% CI 1.21 to 5.34), but there was no evidence for tralokinumab to allow us to assess short‐term follow‐up of POEM or long‐term follow‐up of EASI75.

We are uncertain of the effect of ustekinumab compared with placebo in achieving EASI75 (long‐term follow‐up: RR 1.17, 95% CI 0.40 to 3.45; short‐term follow‐up: RR 0.91, 95% CI 0.28 to 2.97; both very low certainty). We found no evidence on ustekinumab for the POEM outcome.

We are uncertain whether other immunosuppressive agents that targeted our key outcomes influence the achievement of short‐term EASI75 compared with placebo due to low‐ or very low‐certainty evidence.

Dupilumab and ustekinumab were the only immunosuppressive agents evaluated for longer‐term EASI75. Dupilumab was the only agent evaluated for improvement in POEM during short‐term follow‐up.

Low‐ to moderate‐certainty evidence indicates a lower proportion of participants with SAEs after treatment with QAW039 and dupilumab compared to placebo during short‐term follow‐up, but low‐ to very low‐certainty evidence suggests no difference in SAEs during short‐term follow‐up of other immunosuppressive agents compared to placebo.

Evidence for effects of immunosuppressive agents on risk of any infection during short‐term follow‐up and SAEs during long‐term follow‐up compared with placebo was of low or very low certainty but did not indicate a difference.

We did not identify differences in other adverse events (AEs), but dupilumab is associated with specific AEs, including eye inflammation and eosinophilia.

Authors' conclusions

Our findings indicate that dupilumab is the most effective biological treatment for eczema. Compared to placebo, dupilumab reduces eczema signs and symptoms in the short term for people with moderate to severe atopic eczema. Short‐term safety outcomes from clinical trials did not reveal new safety concerns with dupilumab. Overall, evidence for the efficacy of most other immunosuppressive treatments for moderate to severe atopic eczema is of low or very low certainty.

Given the lack of data comparing conventional with newer biological treatments for the primary outcomes, there remains high uncertainty for ranking the efficacy and safety of conventional treatments such as ciclosporin and biological treatments such as dupilumab.

Most studies were placebo‐controlled and assessed only short‐term efficacy of immunosuppressive agents. Further adequately powered head‐to‐head RCTs should evaluate comparative long‐term efficacy and safety of available treatments for moderate to severe eczema.

Plain language summary

Which oral or injected medicines work best to treat moderate to severe eczema?

Why this question is important

Eczema is a persistent condition that causes dry, cracked, and itchy skin. People with mild eczema have small patches of dry skin, and people with moderate eczema have larger, redder, or swollen areas of skin. People with severe eczema have red crusts and broken skin (which may ooze fluid) that develop all over the body.

Although there is currently no cure for eczema, treatments that aim to relieve symptoms are available. Usually, the first treatment option is to apply creams, ointments, or liquids to the affected skin. If this does not work, it is possible to take oral or injected (systemic) medicines that work throughout the body.

Many systemic medicines are available for eczema. To help people decide which one is most appropriate for managing their symptoms, we reviewed the evidence on benefits and risks of different systemic medicines for people with moderate or severe eczema. We particularly wanted to find out:

• whether some medicines were more likely than others to have an important positive effect on symptoms (defined as either at least a 75% improvement on the Eczema Area and Severity Index (EASI), or improvement on the Patient‐Oriented Eczema Measure (POEM)—two scales that are used by clinicians to evaluate changes in eczema symptoms); and

• whether some medicines are associated with more serious unwanted events, including infection, than others.

How we identified and assessed the evidence

First, we searched the medical literature for randomised controlled studies (studies where people are randomly divided into different treatment groups) because these studies provide the most robust evidence about the effects of a treatment. We then compared study results and summarised the evidence from all studies. Finally, we assessed how certain the evidence was. To do this, we considered factors such as the way studies were conducted, the size of studies, and the consistency of findings across studies. Based on our assessments, we categorised the evidence as being of very low, low, moderate, or high certainty.

What we found

We found 74 studies that involved a total of 8177 people with moderate to severe eczema. Studies lasted between 2 weeks and 60 months. Treatments received varied from a single, one‐off dose to weekly doses for 60 months. These studies evaluated 29 different medicines by comparing them to a placebo (fake treatment) or to another medicine, or by comparing different doses of the same medicine. The medicine that was studied most frequently was dupilumab (12 studies), a laboratory‐made version of a protein, which blocks parts of the immune system involved in causing eczema.

Dupilumab versus placebo

High‐certainty evidence shows that, compared to a placebo or other laboratory‐made proteins, dupilumab improves the symptoms of people with moderate to severe eczema in the short term (within 16 weeks of treatment). It is unclear whether this improvement is sustained after 16 weeks because no studies have looked at changes in POEM scores beyond that time, and because the evidence from studies that measured EASI scores was of very low certainty. Dupilumab may be associated with fewer serious unwanted events than placebo (low‐certainty evidence).

Other systemic medicines versus placebo

Evidence on the benefits and risks of other systemic medicines compared to placebo is limited because no studies have measured their effects on symptom improvement or serious unwanted effects, or because the certainty of the evidence is low or very low.

Comparing different systemic medicines with one another

Reviewers found too few studies that compared different systemic medicines against one another to determine which worked best for people with moderate to severe eczema.

What this means

Evidence shows that, compared to placebo, dupilumab improves the symptoms of people with moderate to severe eczema within 16 weeks of treatment and may be associated with fewer serious unwanted events.

Reviewers found too few robust studies to conclude whether dupilumab improves symptoms beyond 16 weeks, or whether this medicine works better than older systemic medicines. Future studies need to compare different systemic treatments beyond 16 weeks in people with moderate to severe eczema.

How up‐to‐date is this review?

The evidence in this Cochrane Review is current to August 2019.

Summary of findings

Summary of findings 1. Summary of findings for EASI75 during short‐term follow‐up.

Estimates of effects, confidence intervals, and certainty of evidence for theproportion of participants who achieved EASI75 with any systemic intervention compared with placebo in the short term (≤ 16 weeks)
Patient or population: patients with moderate to severe eczema Intervention: dupilumab, tralokinumab, tezepelumab, GBR830, lebrikizumab, ustekinumab, ASN002 Comparison: placebo Outcome: achieving 75% improvement in Eczema Area and Severity Index (EASI75); range of follow‐up between 4 weeks and 16 weeks Settings: all participants were recruited from a hospital setting Network geometry plots:Figure 1
Total studies: 14 RCTs Total participants: 3851	Relative effect (95% CI)	Anticipated absolute effect (95% CI)			Certainty of evidence (CiNEMA)	SUCRA
		Without intervention	With intervention	Difference
Dupilumab (8 RCTs; 1978 participants)	RR 3.04 (2.51 to 3.69) Network estimate	184 per 1000	560 per 1000	376 fewer per 1000 (278 fewer to 496 fewer)	High	92.7
Tralokinumab (1 RCT; 153 participants)	RR 2.54 (1.21 to 5.34) Network estimate	184 per 1000	468 per 1000	284 fewer per 1000 (39 fewer to 800 fewer)	Low confidence in estimate due to major concern of within‐study bias	78.2
Tezepilumab (1 RCT; 153 participants)	RR 1.70 (0.85 to 3.40) Network estimate	184 per 1000	313 per 1000	129 fewer per 1000 (442 fewer to 28 more)	Low confidence in estimate due to major concern of imprecision	57.3
GBR830 (1 RCT; 55 participants)	RR 1.91 (0.46 to 8.02) Network estimate	184 per 1000	352 per 1000	168 fewer per 1000 (1293 fewer to 99 more)	Low confidence in estimate due to major concern of imprecision	48.6
Lebrikizumab (1 RCT; 46 participants)	RR 1.40 (0.83 to 2.36) Network estimate	184 per 1000	258 per 1000	74 fewer per 1000 (251 fewer to 31 more)	Very low confidence in estimate due to some concern of within‐study bias and major concern of imprecision	45
ASN002 (1 RCT; 27 participants)	RR 1.50 (0.38 to 5.92) Network estimate	184 per 1000	276 per 1000	92 fewer per 1000 (907 fewer to 114 more)	Low confidence in estimate due to major concern of imprecision	37.5
Ustekinumab (1 RCT; 52 participants)	RR 0.91 (0.28 to 2.97) Network estimate	184 per 1000	168 per 1000	17 more per 1000 (363 fewer to 133 more)	Very low confidence in estimate due to some concern of within‐study bias and major concern of imprecision	19.6
Placebo	Reference comparator	Reference comparator	Not estimable	Not estimable	Reference comparator	21
CI: confidence interval; EASI: Eczema Area and Severity Index (EASI75 = proportion of participants who achieved 75% improvement in EASI score); RR: risk ratio; SUCRA: surface under the cumulative ranking (SUCRA was expressed as a percentage between 0 (when a treatment is certain to be the worst) and 100% (when a treatment is certain to be the best)).
GRADE Working Group grades of evidence (or certainty of evidence). High quality: further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: we are very uncertain about the estimate.

4.

Network plot for all primary outcomes. The size of the nodes is proportionate to the total number of participants allocated to each intervention, and the thickness of the lines is proportionate to the number of studies evaluating each direct comparison. Colours of nodes are proportionate to the number of studies with average risk of bias. Red, yellow, and green colours represent high, unclear, and low risk of bias. AZA: azathioprine; CsA: cyclosporin A; CsAI: cyclosporin A increased regimen (initial dose 3 mg/kg/d, then increased to 5 mg/kg/d); EASI75: improvement of 75% in Eczema Area and Severity Index; MTX: methotrexate; PF: PF‐04965842 (Abrocitinib); POEM: Patient‐Oriented Eczema Measure; SAEs: serious adverse events; steroids: corticosteroids.

Summary of findings 2. Summary of findings for EASI75 during long‐term follow‐up.

Estimates of effects, confidence intervals, and certainty of evidence for theproportion of participants who achievedEASI75 with any systemic intervention compared with placebo in the long term (> 16 weeks)
Patient or population: patients with moderate to severe eczema Intervention: dupilumab and ustekinumab Comparison: placebo Outcome: achieving 75% improvement in Eczema Area and Severity Index (EASI75); range of follow‐up between 6 months and 13 months Settings: all participants were recruited from a hospital setting Network geometry plots:Figure 1
Total studies: 3 RCTs Total participants: 1241	Relative effect (95% CI)	Anticipated absolute effect (95% CI)			Certainty of evidence (CiNEMA)	SUCRA
		Without intervention	With intervention	Difference
Dupilumab (2 RCTs; 764 participants)	RR 2.59 (1.87 to 3.60) Pair‐wise estimate	200 per 1000	518 per 1000	318 fewer per 1000 (174 fewer to 520 fewer)	Very low confidence in estimate due to some concern of within‐study bias and major concern of heterogeneity	N/A
Ustekinumab (1 RCT; 52 participants)	RR 1.17 (0.4 to 3.45) Pair‐wise estimate	200 per 1000	234 per 1000	34 fewer per 1000 (490 fewer to 120 more)	Very low confidence in estimate due to some concern of within‐study bias and major concern of imprecision	N/A
Placebo	Reference comparator	Reference comparator	Not estimable	Not estimable	Reference comparator	N/A
CI: confidence interval; EASI: Eczema Area and Severity Index ((EASI75 = proportion of participants who achieved 75% improvement in EASI score); N/A: data not available because network meta‐analysis cannot be performed; RCT: randomised controlled trial; RR: risk ratio; SUCRA: surface under the cumulative ranking (SUCRA was expressed as a percentage between 0 (when a treatment is certain to be the worst) and 100% (when a treatment is certain to be the best)).
GRADE Working Group grades of evidence. High quality: further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: we are very uncertain about the estimate.

Summary of findings 3. Summary of findings for POEM scores during short‐term follow‐up.

Estimates of effects, confidence intervals, and certainty of evidence forPatient‐Oriented Eczema Measure (POEM) scores with any systemic intervention compared with placebo in the short term (≤ 16 weeks)
Patient or population: patients with moderate to severe eczema Intervention: dupilumab Comparison: placebo Outcome: change in POEM scores; time of follow‐up 16 weeks Settings: all participants were recruited from a hospital setting Network geometry plots:Figure 1
Total studies: 6 RCTs Total participants: 2680	Relative effect (95% CI)	Anticipated absolute effect (95% CI)			Certainty of evidence (CiNEMA)	SUCRA
		Without intervention	With intervention	Difference
Dupilumab (5 RCTs; 1997 participants)	‐	Mean of improving score was 5.18	Mean of improving score was 12.48 (11.79 to 13.18)	Mean difference in improving POEM score was 7.3 higher (6.61 higher to 8.00 higher)	High	N/A
Placebo	Reference comparator	Not estimable	Not estimable	Not estimable	Reference comparator	N/A
CI: confidence interval; N/A: data not available because network meta‐analysis cannot perform; POEM: Patient‐Oriented Eczema Measure; RCT: randomised controlled trial; SUCRA: surface under the cumulative ranking (SUCRA was expressed as a percentage between 0 (when a treatment is certain to be the worst) to 100% (when a treatment is certain to be the best)).
GRADE Working Group grades of evidence (or certainty in the evidence). High quality: further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: we are very uncertain about the estimate.

Summary of findings 4. Summary of findings for patients with SAEs during short‐term follow‐up.

Estimates of effects, confidence intervals, and certainty of evidence for serious adverse events (SAEs) with any systemic intervention compared with placebo in the short term (≤ 16 weeks)
Patient or population: patients with moderate to severe eczema Intervention: dupilumab, tralokinumab, tezepelumab, apremilast, baricitinib, lebrikizumab, PF‐04965842, QAW039, Timapiprant Comparison: placebo Outcome: serious adverse events (SAEs); range of follow‐up between 1 month and 16 weeks Settings: all participants were recruited from a hospital setting Network geometry plots:Figure 1
Total studies: 17 RCTs Total participants: 3972	Relative effect (95% CI)	Anticipated absolute effect (95% CI)			Certainty of evidence (CiNEMA)	SUCRA
		Without intervention	With intervention	Difference
QAW039 (1 RCT; 76 participants)	RR 0.09 (0.01 to 0.76) Network estimate	54 per 1000	5 per 1000	49 more per 1000 (13 more to 53 more)	Moderate confidence in estimate due to some concern of within‐study bias	94.2
Dupilumab (9 RCTs; 1663 participants)	RR 0.37 (0.23 to 0.59) Network estimate	54 per 1000	20 per 1000	34 more per 1000 (22 more to 44 more)	Low confidence in estimate due to major concern of within‐study bias	75.5
Timapiprant (1 RCT; 70 participants)	RR 0.34 (0.07 to 1.62) Network estimate	54 per 1000	18 per 1000	36 more per 1000 (33 fewer to 50 more)	Low confidence in estimate due to major concern of imprecision	74
Tezepelumab (1 RCT; 56 participants)	RR 0.65 (0.11 to 3.77) Network estimate	54 per 1000	35 per 1000	19 more per 1000 (149 fewer to 48 more)	Low confidence in estimate due to major concern of imprecision	54.9
Lebrikizumab (1 RCT; 156 participants)	RR 0.85 (0.17 to 4.25) Network estimate	54 per 1000	46 per 1000	8 more per 1000 (175 fewer to 45 more)	Very low confidence in estimate due to some concern of within‐study bias and major concern of imprecision	47.7
PF‐04965842 (1 RCT; 211 participants)	RR 0.93 (0.20 to 4.35) Network estimate	54 per 1000	50 per 1000	4 more per 1000 (181 fewer to 43 more)	Very low confidence in estimate due to some concern of within‐study bias and major concern of imprecision	45.5
Tralokinumab (1 RCT; 153 participants)	RR 1.67 (0.20 to 13.93) Network estimate	54 per 1000	90 per 1000	36 fewer per 1000 (697 fewer to 43 more)	Very low confidence in estimate due to major concern of within‐study bias and imprecision	31.1
Apremilast (1 RCT; 121 participants)	RR 3.73 (0.20 to 71.1) Network estimate	54 per 1000	201 per 1000	147 fewer per 1000 (3,780 fewer to 43 more)	Low confidence in estimate due to major concern of imprecision	20
Baricitinib (1 RCT; 75 participants)	RR 4.61 (0.24 to 87.25) Network estimate	54 per 1000	249 per 1000	195 fewer per 1000 (4650 fewer to 41 more)	Very low confidence in estimate due to some concern of within‐study bias and major concern of imprecision	16.5
Placebo	Reference comparator	Reference comparator	Not estimable	Not estimable	Reference comparator	40.5
CI: confidence interval; RR: risk ratio; SAE: serious adverse event; SUCRA: surface under the cumulative ranking (SUCRA was expressed as a percentage between 0 (when a treatment is certain to be the worst) and 100% (when a treatment is certain to be the best)).
GRADE Working Group grades of evidence. High quality: further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: we are very uncertain about the estimate.

Summary of findings 5. Summary of findings for patients with SAEs during long‐term follow‐up.

Estimates of effects, confidence intervals, and certainty of evidence for serious adverse events (SAEs) with any systemic intervention compared with placebo in the long term (> 16 weeks)
Patient or population: patients with moderate to severe eczema Intervention: dupilumab, fezakinumab, methotrexate (MTX), QGE031, cyclosporin A Comparison: placebo Outcome: serious adverse events (SAEs); range of follow‐up between 5 months and 16 months Settings: all participants were recruited from a hospital setting Network geometry plots:Figure 1
Total studies: 6 RCTs Total participants: 1720	Relative effect (95% CI)	Anticipated absolute effect (95% CI)			Certainty of evidence (CiNEMA)	SUCRA
		Without intervention	With intervention	Difference
Dupilumab (3 RCTs; 1082 participants)	RR 0.68 (0.38 to 1.21) Network estimate	10 per 1000	7 per 1000	3 more per 1000 (2 fewer to 6 more)	Low confidence in estimate due to major concern of imprecision	78.7
Methotrexate (1 RCT; 50 participants)	RR 1.15 (0.01 to 151.54) Network estimate	10 per 1000	12 per 1000	2 fewer per 1000 (1,539 fewer to 10 more)	Very low confidence in estimate due to some concern of within‐study bias and major concern of imprecision	58.5
Fezakinumab (1 RCT; 40 participants)	RR 2.56 (0.13 to 50.95) Network estimate	10 per 1000	26 per 1000	16 fewer per 1000 (511 fewer to 9 more)	Very low confidence in estimate due to major concern of within‐study bias and imprecision	38.8
QGE031 (1 RCT; 10 participants)	RR 3.00 (0.14 to 65.90) Network estimate	10 per 1000	31 per 1000	20 fewer per 1000 (664 fewer to 9 more)	Very low confidence in estimate due to major concern of within‐study bias and imprecision	35.3
Cyclosporin A (2 RCTs; 49 participants)	RR 3.67 (0.09 to 149.20) Network estimate	10 per 1000	38 per 1000	27 fewer per 1000 (1515 fewer to 9 more)	Very low confidence in estimate due to major concern of within‐study bias and imprecision	31.3
Placebo	Reference comparator	Reference comparator	Not estimable	Not estimable	Reference comparator	57.5
CI: confidence interval; RCT: randomised controlled trial; RR: risk ratio; SAE: serious adverse event; SUCRA: surface under the cumulative ranking (SUCRA was expressed as a percentage between 0 (when a treatment is certain to be the worst) and 100% (when a treatment is certain to be the best)).
GRADE Working Group grades of evidence. High quality: further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: we are very uncertain about the estimate.

Summary of findings 6. Summary of findings for patients with infections during short‐term follow‐up.

Estimates of effects, confidence intervals, and certainty of evidence forinfections with any systemic intervention compared with placebo in the short term (≤ 16 weeks)
Patient or population: patients with moderate to severe eczema Intervention: dupilumab, GBR830, azathiopine (AZA), cyclosporin A (CsA), CIM331, lebrikizumab, QAW039, steroid, timapiprant, cyclosporin A (CsAI) increased regimen Comparison: placebo Outcome: infection; range of follow‐up between 1 month and 4 months Settings: all participants were recruited from a hospital setting Network geometry plots:Figure 1
Total studies: 13 RCTs Total participants: 1658	Relative effect (95% CI)	Effects and confidence in the estimate of effects (95% CI)			Certainty of evidence (CiNEMA)	SUCRA
		Without intervention	With intervention	Difference
Corticosteroid (1 RCT; 21 participants)	RR 0.15 (0.01 to 4.10) Network estimate	279 per 1000	42 per 1000	237 more per 1000 (864 fewer to 276 more)	Very low confidence in estimate due to some concern of within‐study bias and major concern of imprecision	84.7
Dupilumab (4 RCTs; 990 participants)	RR 0.43 (0.18 to 1.06) Network estimate	279 per 1000	120 per 1000	159 more per 1000 (17 fewer to 228 more)	Very low confidence in estimate due to some concern of within‐study report, imprecision, and heterogeneity	75.4
Cyclosporin A (3 RCTs; 74 participants)	RR 0.75 (0.10 to 5.75) Network estimate	279 per 1000	209 per 1000	70 more per 1000 (1,323 fewer to 251 more)	Very low confidence in estimate due to some concern of within‐study report and major concern of imprecision	54.4
Lebrikizumab (1 RCT; 156 participants)	RR 0.79 (0.17 to 3.79) Network estimate	279 per 1000	220 per 1000	59 more per 1000 (777 fewer to 231 more)	Very low confidence in estimate due to some concerns of within‐study report and major concern of imprecision	53.1
Timapiprant (1 RCT; 70 participants)	RR 0.82 (0.16 to 4.14) Network estimate	279 per 1000	228 per 1000	50 more per 1000 (875 fewer to 234 more)	Low confidence in estimate due to major concern of imprecision	52.5
CIM331 (1 RCT; 27 participants)	RR 0.87 (0.16 to 4.63) Network estimate	279 per 1000	242 per 1000	36 more per 1000 (1,011 fewer to 234 more)	Low confidence in estimate due to major concern of imprecision	50.1
GBR830 (1 RCT; 46 participants)	RR 1.04 (0.12 to 8.81) Network estimate	279 per 1000	290 per 1000	11 fewer per 1000 (2,176 fewer to 245 more)	Low confidence in estimate due to major concern of imprecision	45.6
Azathiopine (AZA) (1 RCT; 41 participants)	RR 1.14 (0.16 to 8.11) Network estimate	279 per 1000	318 per 1000	39 fewer per 1000 (1,981 fewer to 234 more)	Low confidence in estimate due to major concern of imprecision	43.2
QAW039 (1 RCT; 76 participants)	RR 2.66 (0.34 to 21.18) Network estimate	279 per 1000	741 per 1000	462 fewer per 1000 (5,622 fewer to 184 more)	Very low confidence in estimate due to major concern of imprecision and some concern of within‐study report	22
CsAI (1 RCT; 40 participants)	RR 3.57 (0.07 to 182.64) Network estimate	279 per 1000	995 per 1000	716 fewer per 1000 (50,605 fewer to 184 more)	Very low confidence in estimate due to major concern of imprecision and some concerns of within‐study report	25
Placebo	Reference comparator	Reference comparator	Not estimable	Not estimable	Reference comparator	44.1
CI: confidence interval; CsAI: cyclosporin A increased regimen (initial dose 3 mg/kg/d, then titrate dose to 5 mg/kg/d); RCT: randomised controlled trial; RR: risk ratio; SUCRA: surface under the cumulative ranking (SUCRA was expressed as a percentage between 0 (when a treatment is certain to be the worst) and 100% (when a treatment is certain to be the best)).
GRADE Working Group grades of evidence. High quality: further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: we are very uncertain about the estimate.

Summary of findings 7. Summary of findings for IGA 0/1 during short‐term follow‐up.

Estimates of effects, confidence intervals, and certainty of evidence for the proportion of participants who achieved an Investigators' Global Assessment (IGA 0/1) of clear or almost clear with any systemic intervention compared with placebo in the short term (≤ 16 weeks)
Patient or population: patients with moderate to severe eczema. Intervention: ASN002, azathiopine (AZA), cyclosporin A (CsA), dupilumab, fezakinumab, GBR830, lebrikizumab, methotrexate (MTX), mepolizumab, PF‐04965842 (PF), QGE031, secukinumab, corticosteroid, tralokinumab, and ustekinumab Comparison: placebo Outcome: IGA 0/1; range of follow‐up between 1 month and 4 months Settings: all participants were recruited from a hospital setting Network geometry plots:Figure 1
Total studies: 24 RCTs Total participants: 4823	Relative effect (95% CI)	Anticipated absolute effect (95% CI)			Certainty of evidence (CiNEMA)	SUCRA
		Without intervention	With intervention	Difference
Dupilumab (10 RCTs; 2347 participants)	RR 3.58 (3.0 to 4.26) Network estimate	98 per 1000	352 per 1000	253 fewer per 1000 (320 fewer to 196 fewer)	Moderate confidence in estimate due to some concern of within‐study bias	84.0
PF‐04965842 (1 RCT; 211 participants)	RR 3.32 (1.25 to 8.79) Network estimate	98 per 1000	326 per 1000	228 fewer per 1000 (765 fewer to 25 fewer)	Moderate confidence in estimate due to some concern of within‐study bias	74.6
CsA (2 RCTs; 19 participants)	RR 3.67 (0.09 to 149.1) Network estimate	98 per 1000	361 per 1000	262 fewer per 1000 (14,551 fewer to 89 more)	Very low confidence in estimate due to major concern of within‐study bias and imprecision	74.4
Mepolizumab (2 RCTs; 38 participants)	RR 4.56 (0.82 to 25.4) Network estimate	98 per 1000	448 per 1000	350 fewer per 1000 (2397 fewer to 18 more)	Very low confidence in estimate due to some concern of within‐study bias and major concern of imprecision	72.7
Fezakinumab (1 RCT; 40 participants)	RR 3.00 (0.39 to 23.25) Network estimate	98 per 1000	295 per 1000	196 fewer per 1000 (2186 fewer to 60 more)	Very low confidence in estimate due to major concern of within‐study bias and imprecision	56.5
Steroid (1 RCT; 21 participants)	RR 1.00 (0.01 to 128.59) Network estimate	98 per 1000	98 per 1000	0 fewer per 1000 (12,535 fewer to 97 more)	Very low confidence in estimate due to some concern of within‐study bias and major concern of imprecision	50.7
Tralokinumab (1 RCT; 153 participants)	RR 1.61 (0.71 to 3.66) Network estimate	98 per 1000	158 per 1000	60 fewer per 1000 (261 fewer to 28 more)	Very low confidence in estimate due to major concern of within‐study bias and imprecision	49.8
Lebrikizumab (1 RCT; 156 participants)	RR 1.46 (0.79 to 2.7) Network estimate	98 per 1000	143 per 1000	45 fewer per 1000 (167 fewer to 21 more)	Very low confidence in estimate due to some concern of within‐study bias and major concern of imprecision	47.5
GBR830 (1 RCT; 46 participants)	RR 2.09 (0.27 to 16.04) Network estimate	98 per 1000	205 per 1000	107 fewer per 1000 (1,478 fewer to 72 more)	Very low confidence in estimate due to major concern of within‐study bias and imprecision	45.8
ASN002 (1 RCT; 27 participants)	RR 2.00 (0.28 to 14.46) Network estimate	98 per 1000	196 per 1000	98 fewer per 1000 (1322 fewer to 71 more)	Low Confidence in estimate due to major concern of imprecision	44.6
MTX (1 RCT; 20 participants)	RR 1.65 (0.07 to 39.17) Network estimate	98 per 1000	162 per 1000	64 fewer per 1000 (3750 fewer to 91 more)	Very low confidence in estimate due to major concern of within‐study bias and some concern of imprecision	43.9
QGE031 (1 RCT; 10 participants)	RR 1.00 (0.02 to 46.05) Network estimate	98 per 1000	98 per 1000	0 fewer per 1000 (4426 fewer to 96 more)	Very low confidence in estimate due to major concern of within‐study bias and imprecision	40.8
AZA (2 RCTs; 63 participants)	RR 1.50 (0.06 to 34.81) Network estimate	98 per 1000	147 per 1000	49 fewer per 1000 (3322 fewer to 92 more)	Low Confidence in estimate due to major concern of imprecision	39.1
Ustekinumab (1 RCT; 52 participants)	RR 1.04 (0.48 to 2.26) Network estimate	98 per 1000	102 per 1000	4 fewer per 1000 (124 fewer to 51 more)	Very low confidence in estimate due to some concern of within‐study bias and major concern of imprecision	33.4
Secukinumab (1 RCT; 22 participants)	RR 0.16 (0.01 to 3.61) Network estimate	98 per 1000	16 per 1000	83 more per 1000 (256 fewer to 96 more)	Very low confidence in estimate due to some concern of within‐study bias and major concern of imprecision	10.5
Placebo (22 RCTs; 1598 participants)	Reference comparator	Reference comparator	Not estimable	Not estimable	Reference comparator	31.8
CI: confidence interval; IGA 0/1: proportion of participants who achieve 0 or 1 (clear or almost clear) value on Investigators' Global Assessment; RCT: randomised controlled trial; RR: risk ratio; SUCRA: surface under the cumulative ranking (SUCRA was expressed as a percentage between 0 (when a treatment is certain to be the worst) and 100% (when a treatment is certain to be the best)).
GRADE Working Group grades of evidence. High quality: further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: we are very uncertain about the estimate.

Background

Relevant terms used in this review are explained in the Glossary (Table 8).

1. Glossary of terms.

Term	Definition
Atrophy	Thinning of overlying skin due to long‐term or excessive usage of steroids
Cyclic adenosine monophosphate (cAMP)	A chemical compound used in intracellular signal communications in various biological processes and pathways, including the pathway dealing with inflammation
Cytokine	A chemical secreted by certain cells of the immune system; it has an effect on other cells
Erythema	Redness of the skin
Excoriations	Linear breaks in the skin surface due to scratching
First‐line	Treatment regimen accepted by the medical establishment for initial treatment
Fissured	Cracking of the superficial layer of the skin
Fitzpatrick skin type	Skin type classification that was developed by Thomas B. Fitzpatrick, based on a person’s skin colour and responses to sun exposure in terms of degree of burning and tanning. Six different skin types were classified from very fair (skin type I) to very dark (skin type VI) depending upon whether the patient burns at the first average sun exposure or tans at the first average sun exposure
Glucocorticosteroid	A type of corticosteroid hormone with anti‐inflammatory and immunosuppressive effects
Immunomodulatory	Regulating the immune system via auto‐regulatory processes (homeostasis) or therapeutically
Immunosuppressive	Reducing the activation or efficacy of the immune system
Lesion (skin)	A region of the skin that has suffered damage through injury or disease of the skin
Lichenification	Thickening of the skin
Lichenified	Thickened first layer of the skin, with the skin lines more obvious
Oedema	Swelling
Palmoplantar	Palms and soles
Papules	Small bump‐like swellings of the skin, forming part of the rash
Pruritus	Severe itching
Second‐line	Treatment regimen that follows if there is failure of response to standard or first‐line therapy
T cell	Also known as the T lymphocyte; a type of lymphocyte or white blood cell that carries T cells; a T‐cell receptor on the cell surface
Xerotic	Dry skin

Description of the condition

Eczema (also known as atopic eczema or atopic dermatitis) is a common and chronic, relapsing inflammatory skin disorder, characterised by intense pruritus and excoriation, with erythematous, xerotic, lichenified, fissured skin, and increased risk of skin infection (Eichenfield 2014; Hanifin 1980; McCollum 2010; Weidinger 2016).

Eczema lesions vary in appearance; papules, vesicles, scaling, fissuring, excoriations, crusting, oedema, and lichenification may be seen. Dry skin resulting from an impaired barrier function is also a key feature of eczema (Wollenberg 2016). Acute lesions typically comprise ill‐defined red scaly patches, often with oedema and vesicle formation, while lichenification and pigmentation are more typical of chronic lesions. Excoriations due to intense pruritus may be seen at any stage. Although eczema can develop on any area of skin, different distribution patterns are often observed at different stages of life. In children younger than two years of age, eczema typically arises on the face, trunk, and limbs including the extensor surfaces. In older children and adults, involvement of the neck and flexural aspects of the limbs (on the inside of joints, such as behind the knees and in the elbow creases) is common, as is involvement of the hands (Akdis 2006; Bos 2010).

Criteria commonly used to diagnose eczema include the Hanifin and Rajka diagnostic criteria, the UK Working Party diagnostic criteria, the Japanese Dermatological Association criteria, and the American Academy of Dermatology criteria (Brenninkmeijer 2008; Vakharia 2018). The severity and extent of eczema are extremely variable, ranging from mild eczema, with localised, occasionally dry, mildly scaly patches; to moderate eczema, with slightly more redness and swelling, with little or no oozing or crusting; to severe, generalised involvement of the whole body, resulting in acute skin failure with widespread, red, oozing, secondarily infected lesions. Both objective signs of eczema and subjective symptoms, such as itch and sleeplessness, contribute to the assessment of clinical severity (Schmitt 2014). Disease severity is routinely assessed during a patient's clinical consultation to track the progress of the disease and to judge the efficacy of therapy.

The main objective physician assessment tools used are the SCORAD (severity SCORing of Atopic Dermatitis) Index (this also includes a patient assessment component) (mild eczema corresponds to SCORAD levels below 25, and severe eczema to SCORAD levels above 50 (Kunz 1997)), the EASI (Eczema Area and Severity Index) score (Ricci 2009), and the Six Area Six Sign Atopic Dermatitis Severity (SASSAD) score (Charman 2002). The principal patient self‐assessment tools are the POEM (Patient‐Oriented Eczema Measure) Scale (Spuls 2017), the SA‐EASI (Self‐Administered Eczema Area and Severity Index) Rating Scale (Housman 2002), and the ADQ (Atopic Dermatitis Quickscore (Carel 2008)). The Harmonising Outcome Measures for Eczema (HOME) initiative reached consensus that EASI should be the core instrument used for clinician‐reported signs, and POEM should be used for patient‐reported symptoms (Schmitt 2014; Spuls 2017).

Eczema affects 5% to 20% of children and 2% to 5% of adults worldwide, posing a significant burden for affected patients, their families, and society (Johansson 2004; Odhiambo 2009). Eczema mainly affects infants and young children, but it can persist, relapse, or first develop in adulthood (Ellis 2012). About 80% of cases develop before the age of five years (Williams 2000). Although it was previously estimated that early‐onset eczema progresses to persistent eczema in adulthood in about 25% of patients (Williams 2005), the Odense Adolescence Cohort Study (TOACS) showed that up to 50% of patients had persistent eczema in adulthood (Mortz 2015). Similarly, another study found that 50% of patients were still affected by age 20 (Margolis 2014). The clinical presentation of eczema is similar across different populations.

The International Study of Asthma and Allergies in Childhood (ISAAC) phase 3, conducted in children six to seven years old and 13 to 14 years old, found a decreased prevalence of eczema in some formerly high‐prevalence countries in the developed world, especially in Northwest Europe, but prevalence was increased in many formerly low‐prevalence developing countries, particularly for the younger age group (Williams 2008). Latin America emerged as a region of comparatively high prevalence of eczema symptoms, and a new area of high prevalence was seen in Southeast Asia (Odhiambo 2009). A UK‐based cross‐sectional survey of 1760 children with eczema from one to five years of age classified 84% as having mild disease, 14% moderate disease, and 2% severe disease. Referral to a dermatologist was correlated with disease severity; 43% of severe cases had been seen by a dermatologist over the preceding year (Emerson 1998).

Eczema is a complex condition that is caused by a combination of genetic and environmental influences and is characterised by cutaneous inflammation, immune dysregulation with a T‐helper 2 cell‐biased response, and epidermal barrier dysfunction. It is strongly associated with genetic factors, in particular loss‐of‐function mutations in filaggrin, a key protein involved in formation of the skin barrier, making a primary skin barrier defect the likely primary trigger of eczematous skin inflammation (Flohr 2014; McAleer 2013).

Eczema often occurs in families with atopic diseases, including asthma, allergic rhinitis, hay fever (and food allergy), and atopic eczema. These diseases share a common pathogenesis and frequently are present together in the same individual and family. The word 'atopy' refers to the genetic tendency to produce immunoglobulin E (IgE) antibodies in response to small quantities of common environmental proteins, such as pollen, house dust mites, and food allergens (Stone 2002; Thomsen 2015). Around 30% of people with eczema develop asthma, and 35% develop allergic rhinitis (Luoma 1983). However, it is known that atopy does not occur concurrently in all patients with atopic eczema. In view of this, recent proposals have suggested that the term 'eczema' should be used to define patients both with and without atopy. Therefore, in agreement with the 'Revised nomenclature for allergy for global use' (Johansson 2004), and similar to other Cochrane Reviews evaluating eczema therapies (van Zuuren 2017; Yew 2018), we will use the term 'eczema' throughout this review.

Several environmental factors such as hard water, hygiene practices, and use of antibiotics early in life have been associated with eczematous skin inflammation (Flohr 2014). Patients' skin may be prone to inflammation in the presence of environmental insults such as soaps and detergents, washing with hard water, and exposure to house dust mites (Cork 2009).

Many have assessed the ways in which eczema can affect quality of life. This condition can have a profound impact on the social, emotional, and physical health of an affected individual. Symptoms and visible lesions can cause behavioural problems, dependency, irritability, sleep loss, pain, itch, physical fatigue, shame, low self‐esteem, anxiety, problems with relationships, and emotional distress (Maksimović 2012). The important economic impact due to frequent visits to physicians, frequent treatments, and days lost at work may also lead to fewer opportunities (Brenninkmeijer 2009; Chamlin 2004). Eczema is most commonly treated with topical medications. Patients often need to alter their daily routine to incorporate regular use of emollients and other topical treatments. Topical treatments can be messy and can cause staining of bed sheets and attire. Many individuals make changes to the style of their attire to hide their rashes when in public. The severity of the condition bears a close relation to the degree of impact on an affected individual’s quality of life. The component of out‐of‐pocket direct expenses for treatments is significant. A systematic review estimated annual direct and indirect costs of eczema in the USA to be USD 364 million to USD 3.8 billion (Mancini 2008).

Description of the intervention

Standard initial management of eczema includes the use of emollients and topical corticosteroids of appropriate potency to treat affected sites. Topical calcineurin inhibitors (including tacrolimus and pimecrolimus) are licensed for use in adults and children two years of age and older as second‐line treatments for moderate to severe eczema that has not been controlled by topical corticosteroids, when there is serious risk of important adverse effects from further topical corticosteroid use (particularly irreversible skin atrophy). Systemic immunosuppressive treatments are reserved for the treatment of more severe cases of eczema that have been inadequately controlled with topical treatments (such as topical corticosteroids and topical calcineurin inhibitors). Systemic treatments available for eczema in adults and children may vary (Sidbury 2014; Wollenberg 2016). For adults with severe eczema, therapeutic options are oral glucocorticosteroids (given for short courses only), cyclosporin A (ciclosporin), methotrexate, azathioprine, mycophenolate mofetil, and psoralen‐ultraviolet A (PUVA). For children with severe eczema, therapeutic options include cyclosporin A (ciclosporin), methotrexate, azathioprine, and mycophenolate mofetil. Based on expert opinion, the threshold for using systemic treatment for children with eczema is typically higher than for adults (Flohr 2013). The choice of systemic agent and treatment duration have not been standardised, and decisions are made on an individual patient basis (Wollenberg 2016).

All of the interventions of interest in this systematic review are systemic (oral, subcutaneous, intravenous, sublingual, or inhalation administration) immunosuppressive treatments that are used to treat moderate to severe eczema. In this review, 'immunosuppressive treatment' is used synonymously with 'anti‐inflammatory treatment' and 'immune‐modulatory treatment', as we consider immunosuppression or anti‐inflammatory effects to be the main mechanism of action of the interventions investigated. We are aware that immunosuppression usually relates to modulation of the immune response.

In this review, we included conventional systemic treatments such as systemic corticosteroids, cyclosporin A (ciclosporin), methotrexate, azathioprine, mycophenolate mofetil, interferon gamma, intravenous immunoglobulin (IVIG), and psoralen‐ultraviolet A (PUVA), as well as small molecule inhibitors such as apremilast, tofacitinib, and baricitinib, and biological therapies such as dupilumab, mepolizumab, nemolizumab, omalizumab, timapiprant (OC000459), and fevipiprant (QAW039) (Newsom 2020; Nygaard 2017; Sidbury 2014; Wollenberg 2016; Zirwas 2018). These treatments cover all systemic immunomodulatory agents included in the recent European Task Force of the Academy of Dermatology/European Academy of Dermatology and Venereology (ETFAD/EADV ‐ Wollenberg 2016) and the guidelines of the American Academy of Dermatology (AAD) (Sidbury 2014).

Conventional systemic treatments

Conventional systemic treatments comprise a heterogeneous group of treatments that are the oldest drugs given to manage eczema.

Systemic corticosteroids

Systemic corticosteroids are used in the short term to manage significant flares in eczema, rather than as long‐term treatment. Anecdotally, they rapidly improve clinical symptoms of eczema, but their side effect profile (particularly when used over the longer term) is unfavourable, and a flare of eczema commonly develops once the corticosteroid dose is weaned (Schmitt 2010). For this reason, and because of their largely unfavourable risk‐to‐benefit ratio, only short‐term use (for a few weeks) to treat severe acute exacerbations of eczema is advised (Wollenberg 2016; Yu 2018). Different formulations of systemic corticosteroids are used to treat eczema, including prednisone, prednisolone, and triamcinolone acetonide. Prednisone and prednisolone are available as an oral tablet or an oral solution; triamcinolone is available as a suspension, which is administered intramuscularly. Dosing is based on the patient’s body weight; typically, doses of 0.5 mg/kg/d to 1 mg/kg/d are administered. Adverse effects of corticosteroids are numerous and include increased risk of infection, hypertension, glucose intolerance, weight gain, gastritis, reduced bone mineral density, adrenal suppression, ophthalmological complications (including cataracts and glaucoma), sleep disturbance, and mood lability. Systemic corticosteroids are often prescribed in combination with proton pump inhibitors to protect against gastric irritation, along with calcium and vitamin D supplements to protect against bone density loss. Use of systemic corticosteroids in children may result in stunted growth (Daley‐Yates 2004).

Cyclosporin A (ciclosporin)

Cyclosporin A (ciclosporin) is an effective systemic treatment option that is recommended for patients with eczema that is refractory to conventional topical treatment (AAD Guideline Strength of Recommendation: B, Level of Evidence: I, II (Sidbury 2014)), usually in doses ranging from 2 mg/kg/d to 5 mg/kg/d, taken in two divided doses (Schmitt 2007). The lowest dose that controls eczema effectively should be given. Cyclosporin A may be used long term (up to 12 months) or for shorter‐term courses (e.g. three to six months). Adverse effects associated with the use of cyclosporin A include increased risk of infection, nephrotoxicity, hypertension, hypertrichosis, gingival hyperplasia, tremor, and increased risk of skin malignancy and lymphoma. Drug interactions with cyclosporin A are common. Caution should be exercised when individuals are taking anti‐fungals (e.g. fluconazole), antibiotics (e.g. macrolides, fluoroquinolones, rifampicin), amiodarone, diuretics (e.g. furosemide), calcium channel antagonists (e.g. diltiazem), statins (e.g. atorvastatin, simvastatin), anti‐epileptics (e.g. carbamazepine, phenytoin), serotonin reuptake inhibitors (e.g. fluoxetine), warfarin, or anti‐HIV drugs (e.g. ritonavir).

Methotrexate

Methotrexate is recommended as a systemic agent for the treatment of refractory eczema (AAD Guideline Strength of Recommendation: B, Level of Evidence: I, II) in adults and children (El‐Khalawany 2013; Schram 2011; Sidbury 2014; Weatherhead 2007). Folate supplementation is recommended during treatment with methotrexate, as methotrexate is an anti‐folate metabolite (Sidbury 2014). It can be administered in oral tablet form or as an oral solution, auto‐injected subcutaneously, which results in improved bioavailability. It is used at low doses (in contrast to its use in cancer therapy) ranging from 7.5 mg to 25 mg, usually as a single weekly dose. Potential side effects include nausea and gastrointestinal upset (which usually can be avoided through subcutaneous administration), oral and mucosal ulceration, bone marrow suppression, increased risk of infection, hepatotoxicity, and pulmonary fibrosis. Methotrexate is teratogenic and must be avoided during pregnancy and breastfeeding. Drug interactions occur with folic acid antagonists (e.g. antibiotics containing trimethoprim, sulphonamides such as co‐trimoxazole), which increase the risk of toxicity, drugs that interfere with renal excretion of methotrexate such as non‐steroid anti‐inflammatory drugs (e.g. ibuprofen) and penicillins, and hepatotoxic drugs (e.g. barbiturates), which can increase the risk of hepatotoxicity.

Azathioprine

Azathioprine is recommended as a systemic agent for the treatment of refractory eczema (AAD Guideline Strength of Recommendation: B, Level of Evidence: I, II), usually at doses ranging from 1 mg/kg/d to 3 mg/kg/d (Berth‐Jones 2002; Meggitt 2006; Sidbury 2014). Its metabolism is dependent on thiopurine methyltransferase (TPMT), a key enzyme in the thiopurine pathway. Genetic polymorphisms in TPMT may result in variable enzyme activity. Therefore, a baseline TPMT level should be checked before azathioprine is initiated; it should be avoided in those with low or absent enzyme activity, who are at greater risk of developing azathioprine toxicity, particularly bone marrow suppression. Common side effects include nausea, vomiting, and other gastrointestinal upset. Other potential side effects include hypersensitivity reactions, hepatotoxicity, pancreatitis, increased risk of infection and skin cancer, and bone marrow suppression. Co‐administration with allopurinol increases the risk of bone marrow suppression and ideally should be avoided.

Mycophenolate

Mycophenolate may be considered an alternative therapy for refractory eczema (AAD Guideline Strength of Recommendation: B, Level of Evidence: II). It shows variable effectiveness and typically is tried after other systemic immunosuppressive treatments such as ciclosporin, methotrexate, or azathioprine have failed or are contraindicated (Haeck 2011; Heller 2007; Sidbury 2014). Mycophenolate can be administered by oral suspension, capsules, or tablets and is given in doses ranging from 0.5 mg/d to 3 mg/d, in two divided doses. The most common side effect is gastrointestinal upset. Less common side effects include headaches, bone marrow suppression (particularly leukopaenia), and increased risk of infection. Several drugs, including rifampicin, cholestyramine, and antacids, reduce blood levels of mycophenolate, and co‐administration with acyclovir increases blood acyclovir levels.

Interferon gamma

Interferon gamma is moderately and variably effective, and may be considered as an alternative therapy for refractory eczema in adults and children who have not responded to, or have contraindications to, the use of other systemic therapies or phototherapy (AAD Guideline Strength of Recommendation: C, Level of Evidence: III (Hanifin 1993; Jang 2000; Sidbury 2014)). It is administered as a subcutaneous injection, but there is no recommended optimal dose or established regimen for the treatment of eczema. Side effects include fatigue, fever, nausea, vomiting, and myalgia.

Intravenous immunoglobulin (IVIG)

Intravenous immunoglobulin (IVIG) has rarely been used; it has been tested with variable success only in small, uncontrolled studies to treat severe refractory eczema in adults and children (Jee 2011; Paul 2002). No regimen has been established for the treatment of eczema.

Psoralen‐ultraviolet A (PUVA)

Psoralen‐ultraviolet A (PUVA) is a type of ultraviolet radiation treatment that starts with a photosensitising medicine, called a psoralen, and is followed by exposure of the affected skin to ultraviolet A (UVA). For oral PUVA, oral psoralens (e.g. 8‐methoxypsoralen, 5‐methoxypsoralen) are taken two to three hours before exposure to UVA in a phototherapy cabinet. Phototherapy is often used as a second‐line option for treating eczema that has failed to respond to topical corticosteroids and calcineurin inhibitors, but it requires regular attendance at the hospital to receive treatment, which may not be logistically feasible for all patients. Narrowband ultraviolet B (UVB) phototherapy is preferred over broadband UV phototherapy, including PUVA, as the latter is associated with increased risk of non‐melanoma skin cancer, despite its effectiveness with refractory eczema (Atherton 1988; Sidbury 2014; Uetsu 2003). The dosing and scheduling of phototherapy are usually based on the patients' response to sun exposure according to Fitzpatrick skin type (Sachdeva 2009), their measured minimal erythema dose, or both. Typically, twice‐weekly PUVA treatments are given for eczema, and the dose of UVA radiation is gradually increased over the course of treatment. The total number of PUVA treatments will depend on individual factors. It is relatively common for patients to experience some skin redness or itching after treatments. Overdose of PUVA may result in skin burning; ideally, the patient should wear the same cut of clothes or underwear during the course of treatment. Male patients are advised to cover their genitals with a close‐fitting pouch or jock strap to prevent burning. Eye protection must be worn to protect against damage to the eyes, which could result in cataracts. Psoralen tablets may result in nausea. Patients with a history of cold sores triggered by sunlight are advised to wear sunblock while receiving phototherapy, to help prevent recurrence. PUVA is not advised during pregnancy or breastfeeding due to the potential risk of foetal damage secondary to the use of psoralen. Although limited courses of PUVA may be used, long‐term continuous treatment is not advised due to increased risk of developing skin cancers. PUVA can be used as a monotherapy or in combination with emollients and topical corticosteroids. Topical calcineurin inhibitors should be avoided on days of PUVA treatment, as these are photosensitising and could increase the risk of burning. Other oral or topical photosensitising treatments (e.g. tetracyclines) should be avoided during PUVA treatment.

Small molecule agents

Small molecule agents affect molecules inside immune cells. Several small molecule agents have been developed and show potential to treat moderate to severe eczema in patients not responding to conventional treatments. Examples of small molecule drugs used in eczema are phosphodiesterase (PDE)‐4 (apremilast) inhibitors, Janus kinase (JAK) inhibitors (tofacitinib, baricitinib, PF‐04965842 (abrocitinib)), and JAK/spleen tyrosine kinase inhibitors (ASN002) (Nygaard 2017).

Apremilast

Apremilast is a selective inhibitor of phosphodiesterase‐4 (PDE‐4) that is administered orally at a dose of 30 mg twice daily. It is licensed in Europe for treatment of moderate to severe psoriasis and psoriatic arthritis. Although apremilast showed promising results in an early pilot study and in several uncontrolled case reports and case series of severe eczema, it did not meet its primary endpoint in a double‐blind placebo‐controlled trial; therefore, apremilast has not been approved for treatment of eczema (Samrao 2012). Of interest, a topical PDE‐4 inhibitor, crisaborole, is approved in the USA for treatment of mild to moderate eczema in children younger than two years (Paller 2016).

Biological therapies

Biological therapies use substances made from living organisms or synthetic versions to target the immune system. Several biological agents such as dupilumab, mepolizumab, omalizumab, and timapiprant have been developed for treatment of moderate to severe eczema (Nygaard 2017).

Dupilumab

Dupilumab is a humanised monoclonal antibody targeting the alpha subunit of the interleukin‐4 receptor (IL‐4Rα). By blocking this receptor, dupilumab reduces signalling through both interleukin‐4 and interleukin‐13 pathways. Randomised controlled trial evidence indicates that dupilumab is an effective treatment for eczema (Beck 2014; Wang 2018). It is licensed in the USA and in Europe at a dose of 300 mg every two weeks, administered subcutaneously following a 600‐mg loading dose. The main side effects of dupilumab are injection site reactions, conjunctivitis, and cold sores.

Mepolizumab

Mepolizumab is a humanised monoclonal antibody administered by subcutaneous injection that prevents interleukin‐5 from binding to its receptor. It is licensed for treatment of severe asthma with an eosinophilic phenotype in the USA and in Europe. However, it has not appeared to be more effective than placebo for treatment of severe eczema in adults; therefore, it is not a licensed treatment for eczema (Oldhoff 2005).

Omalizumab

Omalizumab is a humanised monoclonal antibody that specifically binds to free human immunoglobulin E (IgE) in the blood and interstitial fluid, and to membrane‐bound forms of IgE on the surface of IgE‐expressing B lymphocytes. It is administered by subcutaneous injection and is approved for treatment of moderate to severe allergic asthma and chronic spontaneous urticaria. Results of trials using omalizumab to treat eczema are conflicting, and this is not a licensed treatment (Belloni 2007; Heil 2010; Iyengar 2013).

How the intervention might work

The different systemic immunosuppressive treatments used for eczema suppress skin and systemic inflammation to control disease signs and symptoms (Roekevisch 2014; Schmitt 2007; Simon 2014; Slater 2015). They may also help to reduce the need for regular use of topical corticosteroid therapy (Wollenberg 2016). Some have been shown to improve quality of life (Roekevisch 2014). Interruption of the inflammatory pathways implicated in eczema is the general mode of action of all systemic immunosuppressive treatments for eczema. Potential targeted interruption of the inflammatory pathways for each systemic treatment of interest for eczema include the following.

Conventional systemic treatments

Systemic corticosteroids

Oral corticosteroids such as prednisolone have a broad anti‐inflammatory effect. They act in a non‐specific manner and suppress several immunological functions such as interfering with antigen presentation to T lymphocytes, inhibiting prostaglandin and leukotriene synthesis, inhibiting neutrophil and monocyte superoxide radical generation, impairing cell migration, and causing redistribution of monocytes, lymphocytes, and neutrophils, thus blunting inflammatory and autoimmune responses. Therefore, they are used widely for treatment of various inflammatory diseases, including eczema (Schmitt 2009; Schmitt 2010).

Cyclosporin A

Cyclosporin A (CyA, or ciclosporin) is a calcineurin inhibitor. The immunosuppressive role of cyclosporin A can be explained by inhibition of the transcription of T‐cell cytokine genes; this is the mechanism by which it is thought to treat various inflammatory skin conditions, including eczema. It has been shown to inhibit interleukin‐2 (IL‐2), IL‐3, IL‐3 granulocyte macrophage colony‐stimulating factor, tumour necrosis factor‐alpha (TNF‐alpha), and expression of interferon gamma (Faulds 1993).

Methotrexate

Methotrexate is a folic acid antagonist that has previously proved to be effective for chronic inflammatory diseases such as psoriasis and rheumatoid arthritis. The mechanism(s) by which methotrexate treats eczema is unknown, although it is thought that its anti‐inflammatory effects may be mediated through adenosine pathways, and its immunomodulatory effects may be due to alteration of signalling and trafficking of immune cells, including T cells (Braun 2009).

Azathioprine

Azathioprine is a purine analogue that inhibits de novo purine biosynthesis to produce broad immunosuppressive effects (Dutz 1998).

Mycophenolate mofetil

Mycophenolate mofetil, or mycophenolic acid, is a purine biosynthesis inhibitor. It inhibits the synthesis of deoxyribonucleic acid (DNA) and ribonucleic acid (RNA) by interfering with de novo purine biosynthesis. Its mechanism of action in treating eczema is unknown, but it is thought to involve inhibition of the proliferative response of both B and T lymphocytes, reducing cutaneous inflammation (Abd Rahman 2013).

Interferon gamma

Interferon gamma is a cytokine that plays a key role in the innate and adaptive immune system. It has been used rarely in cases of severe eczema unresponsive to other systemic immunosuppressive treatments. Its mechanism of action for the treatment of eczema is unclear. It is postulated to be able to correct immunological imbalances in patients with eczema by decreasing serum IgE levels and IL‐4 levels and restoring immune balance, thereby leading to clinical improvement (Chang 2002).

Intravenous immunoglobulins

Intravenous immunoglobulins (IVIGs) are blood products derived from the pooled plasma of many individuals. A variety of anti‐inflammatory and immunomodulatory effects have been proposed to explain the effectiveness of IVIGs in treating a wide variety of autoimmune and inflammatory conditions, including immune thrombocytopaenia, myasthenia gravis, and dermatomyositis (Ballow 2014; Gelfand 2012).

Psoralen‐ultraviolet A

Psoralen‐ultraviolet A (PUVA) is thought to exert its therapeutic effect through multiple mechanisms, primarily through its immunomodulatory properties, including altering cytokine and cytokine receptor expression and lymphocyte apoptosis, reducing adhesion molecule expression, and affecting the function of antigen‐presenting cells (Johnson 1996; Laing 1995; Liszewski 2017; Singh 2010).

Small molecule treatments

Apremilast

Apremilast is a small molecule inhibitor of phosphodiesterase‐4 (PDE‐4). Leukocytes from patients with eczema display elevated phosphodiesterase (PDE) activity compared to normal controls, which leads to leukocyte hyperactivity and inflammation (Hanifin 1996). Inhibition of PDE‐4 results in increased intracellular cyclic adenosine monophosphate, which activates protein kinase A and other intermediate mediators, resulting in inhibition of pro‐inflammatory cytokine production and reduced inflammation (Souness 2000).

Baricitinib

Baricitinib is an oral, small molecule, selective inhibitor of Janus kinase 1 (JAK1) and Janus kinase 2 (JAK2) (Guttman‐Yassky 2019a). Generally, patients with atopic dermatitis (AD) show elevated levels of inflammatory cytokines, including Th2 (IL‐4, IL‐13, IL‐31), Th22 (IL‐22), and Th1 (IFN‐γ), many of which depend on Janus kinase (JAK)/signal transducer and activator of transcription (STAT) intracellular signalling. In a phase 2 clinical trial in adults with AD, results at week 16 indicate that a higher percentage of participants receiving baricitinib 4 mg with a topical corticosteroid (TCS) achieved an EASI50 compared to participants given placebo with TCS (P = 0.027) (Guttman‐Yassky 2019a).

Biological treatments

Dupilumab

Dupilumab is a monoclonal antibody that targets the alpha subunit of the interleukin‐4 receptor (IL‐4 alpha). In doing so, it reduces signalling through interleukin‐4 and interleukin‐13 pathways, which are known to drive cutaneous inflammation in eczema (Kraft 2017).

Mepolizumab

Mepolizumab is an anti‐IL‐5 monoclonal antibody that leads to reduced serum eosinophil counts. The inflammation of eczema is characterised by eosinophil infiltration; therefore, depletion of eosinophils by this drug would appear to be a promising approach for eczema treatment (Leiferman 2001).

Nemolizumab

The monoclonal antibody (mAb) nemolizumab is directed against the IL‐31 receptor A, which, together with the oncostatin M receptor, constitutes the functional IL‐31 receptor. Results of a phase 1 study indicate that there is a statistically significant 50% reduction in itch when assessed by the pruritus visual analogue scale at four weeks after initiation of treatment compared to placebo (Nemoto 2016). A phase 2 study of nemolizumab in 264 patients with moderate to severe AD for which topical therapy cannot adequately control disease activity shows significantly decreased pruritus (Ruzicka 2017).

Omalizumab

Omalizumab is an anti‐IgE monoclonal antibody. As an atopic condition, eczema is characterised by a genetic tendency to produce IgE antibodies in response to common environmental proteins, and high levels of serum IgE are commonly seen, making IgE inhibitors a potential therapeutic target in patients with eczema (Liu 2011).

Timapiprant

Timapiprant is an oral anti‐CRTH2 mAb. The phase 2 trial included patients with moderate to severe eczema who were randomised 1:1 to 50 mg daily of timapiprant or placebo. The effect on the primary endpoint of change from baseline EASI was a reduction of –3.8 points (standard error 1.72) in the treatment group and –6.1 points (standard error 1.71) in the placebo group (NCT02002208).

Why it is important to do this review

Moderate to severe eczema is often difficult to treat (Ring 2012). Current clinical practice frequently is based not on a systematic critical appraisal of the published evidence, but rather on expert opinion, without clear evidence for the given condition (Darsow 2010; Sidbury 2014; Wollenberg 2016). A survey among 700 dermatologists and paediatricians from eight European countries confirmed wide variation in systemic treatment approaches for children and adolescents with moderate to severe atopic eczema (Proudfoot 2013). First‐line systemic agents were cyclosporin A (ciclosporin), oral corticosteroids, and azathioprine. Systemic immunosuppressive treatments are normally used in severe eczema, but their use in moderate eczema is controversial due to safety concerns. In general, expensive systemic immunosuppressive treatments with potentially serious adverse effects should not be used for moderate cases, which usually respond well to patient education and to adequate and safe use of topical corticosteroids, including proactive control (weekend therapy), and treatment combined with topical calcineurin inhibitors (TCIs) (Sidbury 2014; Wollenberg 2016). Current clinical practice guidelines do not provide explicit guidance on systemic treatment strategies for adults and children with eczema, and their comparative effectiveness and role in treating different patient subgroups remain unclear (Ring 2012; Saeki 2016; Sidbury 2014; Wollenberg 2016). Limited randomised studies have directly compared different systemic agents for eczema. Although dupilumab has become the first approved biological treatment for moderate to severe eczema that is not adequately controlled with topical therapies, its efficacy relative to other systemic immunosuppressive treatments for eczema has not been established (Boguniewicz 2017).

Network meta‐analysis (NMA) refers to a meta‐analysis in which multiple treatments (i.e. three or more treatment options) are compared through both direct comparison of interventions within studies and indirect comparison across studies, based on a common comparator. Available studies comparing different systemic immunosuppressive treatments are limited; therefore, NMA allows a combination of direct and indirect evidence and can be used to rank different systemic immunosuppressive treatments (Salanti 2011; Salanti 2012). To date, one NMA ‐ Drucker 2020 ‐ and several pair‐wise systematic reviews on systemic immunosuppressive treatments for eczema have been published (Roekevisch 2014; Schmitt 2007). This review is also timely, as previous pair‐wise systematic reviews did not include new treatments such as biological therapies, and they focused mainly on clinical signs rather than addressing the core outcome domains identified by the Harmonising Outcome Measures for Eczema (HOME) initiative: clinical signs, patient symptoms, long‐term control, and quality of life (Chalmers 2016; Chalmers 2018; Schmitt 2014). Although the NMA of systemic treatment was published in 2020 (Drucker 2020), several limitations are noted. This review focused only on short‐term outcomes (8 to 16 weeks) and limited analyses to only continuous outcomes. Given greater availability of randomised controlled trials (RCTs) on systemic treatments for eczema and lack of long‐term outcomes analysis, there remains a gap, prompting us to perform an NMA. Moreover, the Cochrane Skin Group undertook an extensive prioritisation exercise, alongside the Global Burden of Disease and the World Health Organization, to identify a core portfolio of the most clinically important titles. The topic of systemic immunosuppressive treatments for eczema was identified as one of the top five titles for clinically important research priorities (Crowe 2012).

Objectives

To assess the comparative efficacy and safety of different types of systemic immunosuppressive treatments for moderate to severe eczema using network meta‐analysis and to generate rankings of available systemic immunosuppressive treatments for eczema according to their efficacy and safety.

Methods

Criteria for considering studies for this review

Types of studies

We included only randomised controlled trials (RCTs) of one or more systemic immunosuppressive treatments for eczema. We included placebo‐controlled, head‐to‐head trials and multi‐arm studies. We included both parallel and cross‐over design trials, but we excluded cluster‐RCTs. We included only data from the first part of any included study (i.e. before cross‐over) because we were not certain about the period effect. We believe it is possible that the condition can change after the intervention during the first period of the study, thus modifying response to an alternative intervention during the second phase of the study. In other words, effects of interventions may differ if they are used at different stages of the disease and after preceding treatments.

Types of participants

Studies conducted in participants with eczema of similar type and severity were included to ensure that our analysis met the transitivity assumption of network meta‐analysis. Briefly, we pre‐defined the diagnostic criteria for atopic eczema and excluded other types of eczema. We assessed the distribution of all relevant patient characteristics, including severity of eczema, concomitant medications, and age, and we determined whether they are different across pairs. In addition, we assessed differences between study design and settings across studies. Details of inclusion and exclusion criteria for types of participants are shown below.

We considered participants of all ages with a clinical diagnosis of moderate to severe eczema. They could have fulfilled diagnostic criteria such as the Hanifin and Rajka definition ‐ Hanifin 1980 (Appendix 1) ‐ or the UK modification (Williams 1994; Appendix 2), or they could have been diagnosed clinically by a clinical healthcare professional, using the term 'atopic eczema' or 'atopic dermatitis'.

We posed no restrictions on age, sex, or ethnicity of participants.

We excluded studies that included participants with other types of eczema such as contact dermatitis, seborrhoeic eczema (seborrhoeic dermatitis), varicose eczema, discoid eczema, and hand eczema.

Types of interventions

We considered studies to be eligible if they investigated the efficacy or safety of at least one systemic immunosuppressive or immunomodulatory therapy for eczema, or a combination of treatments from the following: systemic corticosteroids, cyclosporin A (ciclosporin), methotrexate, azathioprine, mycophenolate mofetil, interferon gamma, intravenous immunoglobulin (IVIG), psoralen‐ultraviolet A (PUVA), apremilast, dupilumab, mepolizumab, omalizumab, and others, including new immunosuppressive or immunomodulatory agents whose first trials were published between publication of our protocol ‐ Sawangjit 2018 ‐ and our final literature search. We defined nodes in NMA based on interventions regardless of dose, while splitting interventions with different doses for subgroup analysis. We assume that the common comparator is the placebo arm, which is very similar across studies.

Types of outcome measures

Outcomes of interest in this review include recommended core outcome domains for eczema trials following the global Harmonizing Outcome Measures for Eczema (HOME) initiative, including clinical signs (measured by a physician‐assessed instrument), symptoms (measured by a patient‐assessed instrument), health‐related quality of life, and long‐term control of eczema (Chalmers 2016; Chalmers 2018; Schmitt 2014; Spuls 2017).

Several clinical severity of eczema measures such as the Eczema Area and Severity Index (EASI; Ricci 2009), Scoring Atopic Dermatitis (SCORAD; Kunz 1997), Six Area, Six Sign Atopic Dermatitis (SASSAD) severity score (Charman 2002), Investigators' Global Assessment (IGA), and affected body surface area (BSA) were used and reported in eczema trials. Participant‐reported symptom measures such as the Patient‐Oriented Eczema Measure (POEM; Spuls 2017), Self‐Administered Eczema Area and Severity Index (SA‐EASI; Housman 2002), the Itch Severity Scale (ISS) (Majeski 2007), or other less validated measures such as the Atopic Dermatitis Assessment Measure (ADAM; Charman 1999), Patient‐Oriented SCORAD (PO‐SCORAD; Stalder 2011), and the Leuven Itch Scale (LIS; Haest 2011), were used and reported in eczema trials. We chose the primary outcomes according to consensus agreement by the HOME (Schmitt 2014).

Timing of outcomes

We defined outcomes as short‐term (≤ 16 weeks of follow‐up time), long‐term, or maintenance (> 16 weeks of follow‐up time; Blauvelt 2017). For multiple times of outcome measurement, we used data at the end of the study (end of follow‐up) and classified them as short‐term outcomes if the study period was 16 weeks or shorter. If the study period exceeded 16 weeks, we used data from the end of the study and classified them as long‐term outcomes. We classified outcomes measured closest to 16 weeks as short‐term outcomes.

Primary outcomes

• Proportions of participants who achieved EASI75 (achieved 75% improvement in EASI score) at short‐term (≤ 16 weeks) and long‐term (> 16 weeks) durations were the primary outcomes for assessing clinical severity of eczema. This outcome was selected because the HOME initiative had reached consensus that this should be a core instrument for clinician‐reported signs (Schmitt 2014), and dichotomous outcomes could reflect clinical differences in clinical severity better than continuous outcomes

• Proportions of participants who achieved POEM50 (achieved 50% improvement in POEM score) at short‐term and long‐term durations were planned to be the primary outcome for participant‐reported symptoms because the HOME Initiative consensus indicates that this should be a core instrument for patient‐reported symptoms (Spuls 2017). However, only a few studies reported POEM50; therefore, POEM in continuous scores at short‐ and long‐term durations was used instead of POEM50

• Proportions of participants with serious adverse effects (SAEs) and proportions of participants with infection at short‐term and long‐term durations were used to evaluate the safety of each systemic treatment

Secondary outcomes

• Proportions of participants who achieved an Investigators' Global Assessment or Physicians' Global Assessment value of 0 or 1 (clear or almost clear) (IGA 0/1) at short‐term and long‐term durations

Search methods for identification of studies

We aimed to identify all relevant RCTs, regardless of language or publication status (published, unpublished, in press, or in progress).

Electronic searches

The Cochrane Skin Information Specialist searched the following databases up to 25 August 2019, using the following strategies based on the draft strategy for MEDLINE in our published protocol (Sawangjit 2018).

• Cochrane Skin Group Specialised Register using the search strategy in Appendix 3.

• Cochrane Central Register of Controlled Trials (CENTRAL; 2019, Issue 8), in the Cochrane Library, using the strategy in Appendix 4.

• MEDLINE via Ovid (from 1946) using the strategy in Appendix 5.

• Embase via Ovid (from 1974) using the strategy in Appendix 6.

• GREAT database (Global Resource of EczemA Trials), Centre of Evidence‐Based Dermatology (accessed at www.greatdatabase.org.uk on 27 June 2018) using the browse filter 'Systemic immunomodulatory agents'. This was not searched again, as the content has since migrated to CENTRAL and is available there.

Trial registers

Two review authors (RS, PD) used the following terms adapted from the search strategy for MEDLINE (Appendix 5).

• For the population: eczema or dermatitis or atopic dermatitis.

• For interventions: immunosuppressive agents, immunomodulatory agents, name of each systemic agent used for eczema, including cyclosporin A (ciclosporin), azathioprine, methotrexate, mycophenolate mofetil or mycophenolic acid, oral glucocorticosteroids, oral prednisolone, psoralen‐ultraviolet A, tacrolimus, interferon‐gamma, intravenous immunoglobulin (IVIG), alitretinoin, apremilast, and biological therapies including infliximab, rituximab, tocilizumab, dupilumab, mepolizumab, omalizumab, or their synonyms.

These review authors searched the trials registers below.

• International Standard Randomized Controlled Trials Number (ISRCTN) registry (www.isrctn.com).

• ClinicalTrials.gov (www.clinicaltrials.gov).

• Australian New Zealand Clinical Trials Registry (www.anzctr.org.au).

• World Health Organization International Clinical Trials Registry Platform (ICTRP) (apps.who.int/trialsearch/).

• EU Clinical Trials Register (https:// www.clinicaltrialsregister.eu/).

Searching other resources

References from published studies

We checked the bibliographies of included studies and related systematic reviews to identify further references to relevant trials.

Adverse effects

We did not perform a separate search for adverse effects of the target interventions. We considered only adverse effects described in included studies.

Data collection and analysis

Selection of studies

We used Covidence software to screen our search results (Covidence). At least two review authors (of RS, PD, ALL, and NC) independently screened the title and abstract of all reports identified by searching electronic databases and conducting other search strategies, against the aforementioned criteria, to identify studies that appear relevant. Subsequently, we retrieved the full text of all potentially relevant studies and independently compared them against the pre‐defined inclusion criteria. We provided reasons for exclusion of these short‐listed papers in a Characteristics of excluded studies table. During the selection process, the two review authors discussed any disagreements within the reviewing team until they reached consensus. We intended to follow the PRISMA statement and created a flowchart for selection of studies (Hutton 2015).

Data extraction and management

Two review authors (RS, PD) independently extracted data from all eligible studies, using a structured data collection form to ensure consistency of appraisal. We resolved disagreements about data extraction by discussion. For each of the included studies, we extracted the following information and transferred it into a Characteristics of included studies table.

• General information: publication status, title, authors, source, country, language of publication, year of publication, study design, study location, study setting (single centre or multi‐centre), information on training methods, and alignment of assessment methods.

• Participant characteristics: number of people with eczema included (sample size, total, and by intervention), age in years (mean ± standard deviation (SD)), sex, severity of eczema, previous treatments, atopic comorbidities, inclusion and exclusion criteria, and baseline characteristics of intervention and comparison groups.

• Interventions: description of intervention (treatment and comparison), mode of administration, dose, frequency, duration of active treatment and follow‐up, and adjuvant treatments.

• Outcomes: primary and secondary outcomes, total serious adverse events, total adverse events, the most common worst side effect of each systemic treatment such as infection with all agents, conjunctivitis with the new biological agents, renal function impairment and hypertension with cyclosporin A, stomach upset with mycophenolate mofetil, and other defined outcomes and reported time points. In addition, we extracted data at the group level such as effect sizes and standard error, when a pair of interventions were compared (not summary effects), if possible.

• Note: sponsorship or funding for studies and notable conflicts of interest of trial authors.

Potential effect modifiers in this review included age and severity of eczema. These variables were extracted as specified in the protocol. We determined the distribution of these potential effect modifiers across available pair‐wise comparisons in the network.

Assessment of risk of bias in included studies

Two review authors (RS, PD) independently assessed each study for risk of bias, using the Cochrane tool (Higgins 2011a; Higgins 2011b). This tool examines seven major sources of bias.

• Random sequence generation (selection bias).

• Allocation concealment (selection bias).

• Participant and personnel blinding (performance bias).

• Outcome assessment blinding (detection bias).

• Incomplete outcome data (attrition bias).

• Selective reporting (reporting bias).

• Other sources of bias, including design‐specific risks of bias and baseline imbalance.

We assessed the risk of bias for cross‐over studies using the first period according to the Cochrane Handbook for Systematic Reviews of Interventions.

We considered a study to be at low risk of attrition bias if the dropout rate was less than 20%. However, if an RCT had dropouts and used an intention‐to‐treat (ITT) analysis but gave no information on handling of missing data, we judged the risk of bias as unclear due to incomplete outcome information. We judged the risk of bias in each study against each source as low, high, or unclear (due to lack of information or uncertainty over the potential of bias). We considered the relative importance of each item to be equal.

We determined the overall risk of bias for each outcome (across domains) within studies as follows (Higgins 2011a; Higgins 2011b).

• Low risk of bias, when all domains are assessed as being at low risk, or plausible bias is unlikely to seriously alter the results.

• Unclear risk of bias, when at least one domain is classified as being at unclear risk, or plausible bias raises some doubt about the results.

• High risk of bias, when at least one domain is judged as being at high risk, or when plausible bias seriously weakens confidence in the results.

Measures of treatment effect

For dichotomous outcome data (i.e. proportion of participants who achieved EASI75, proportion of participants who achieved IGA 0/1, proportion of participants with SAE or infection), we used risk ratios (RRs), along with their respective 95% confidence intervals (CIs) (Higgins 2011a). We used risk ratios to compare adverse events, as all included studies were RCTs. RR is an appropriate outcome measure for comparison of adverse events in this situation.

Clinical severity of eczema, as measured by SCORAD, is classified into three categories: mild disease (SCORAD 0 to 24), moderate disease (SCORAD ≥ 25 to < 50), and severe disease (SCORAD ≥ 50). The POEM score is classified into five categories: clear or almost clear (score = 0 to 2), mild (score = 3 to 7), moderate (score = 8 to 16), severe (score = 17 to 24), and very severe eczema (score = 25 to 28). The Investigator’s Global Assessment Scale (IGA) is a five‐point scale that provides global clinical assessment of eczema severity ranging from 0 to 4, where 0 indicates clear, 1 is almost clear, 2 is mild, 3 is moderate, and 4 indicates severe eczema. A decrease in score is associated with improvement in signs and symptoms.

For categorical outcomes such as POEM and IGA, data were dichotomised to the proportion of participants achieving clear or almost clear eczema (POEM score 0 to 2, IGA score 0/1) or mild disease (SCORAD 0 to 24). RR was then calculated to analyse the data as for dichotomous outcome data. In addition, the proportion of participants achieving 50% improvement in SCORAD or POEM was recorded and analysed as RR.

For continuous outcomes, such as changes in or post‐treatment values of clinical severity of the disease, clinical signs and symptoms, quality of life, and flares, we calculated mean differences (MDs) with 95% CIs. We used standardised mean differences (SMDs) with 95% CIs for continuous outcomes if the included studies used different scales for the same outcome such as pruritus score or itching score. The SMD expresses the intervention effect in standard units rather than in the original units of measurement. When SMDs were estimated, the absolute values of intervention and comparison groups were not useful because studies used different measurement instruments with different units. Therefore, we interpreted the SMDs as follows: 0.2 represents a small effect, 0.5 a moderate effect, and 0.8 a large effect. Variations were represented as small, moderate, and large when values were lower than 0.40, ranged from 0.40 to 0.70, and were greater than 0.70, respectively (Cohen 1988).

Ranking probabilities

From the results of the NMA for each comparison, we estimated for each treatment the probability of being at each possible rank. Using these results, we determined a treatment hierarchy, using the surface under the cumulative ranking curve (SUCRA) (Chaimani 2017; Salanti 2011). The SUCRA could also be re‐expressed as a percentage, interpreted as the percentage of effectiveness or acceptability of an intervention that would be ranked first, with no uncertainty (Salanti 2012). The SUCRA was used to estimate relative rankings for all four primary outcomes (EASI75, POEM, SAE, and infection) and for one secondary outcome (IGA 0/1) at week 16 of follow‐up time or nearest (short‐term duration), and at time point over 16 weeks of follow‐up (long‐term duration). We also developed and used rankograms to reflect the uncertainty of ranking probabilities visually and graphically (Chaimani 2017).

Unit of analysis issues

The unit of analysis was the participant. For studies with multiple intervention groups (e.g. three‐arm design), we performed separate pair‐wise comparisons for each pair without splitting the control group. This means that we included data from treatment number one compared to control in a pair‐wise meta‐analysis, and we included data from treatment number two compared to control in another pair‐wise meta‐analysis. The network meta‐analysis method accounts for relative effect estimates arising from multiple‐arm studies, so there is no need to split the control group evenly to avoid a unit of analysis error (Higgins 2012).

Dealing with missing data

We performed an ITT analysis whenever possible. In cases of missing data about key study characteristics or outcomes, we contacted the study investigators for additional information. However, for this review, we did not contact any trial authors.

For continuous outcomes, we imputed the standard deviation from P values, standard errors, and confidence intervals, according to guidance given in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011a). If the data are likely to be normally distributed, we used the median for meta‐analysis when the mean was not available. If it is not possible to calculate the standard deviation from available information, we imputed the standard deviation by using the largest standard deviation in other trials for that outcome. This form of imputation may decrease the weight of the study when mean differences are calculated, and it may bias the effect estimate towards the line of no effect when standardised mean differences are calculated (Higgins 2011a).

Assessment of heterogeneity

Our review presented both pair‐wise and network comparisons. For pair‐wise meta‐analysis, we assessed the treatment effects of individual trials and heterogeneity between trial results by visually inspecting the forest plots. We explored clinical heterogeneity by assessing clinical and methodological characteristics of the included studies (e.g. differences in study risk of bias, participants, interventions, or outcome assessments). We attempted to pool data in a meta‐analysis only if the clinical heterogeneity amongst the selected studies is negligible. If we found major discrepancies in clinical or methodological characteristics, we determined whether to exclude some studies from the meta‐analysis altogether, or to include them and perform a sensitivity analysis of the main outcome.

We used the Q‐test and the I² statistic to measure heterogeneity in study results. We interpreted the I² statistic according to the following thresholds (Section 9.5.2 of the Cochrane Handbook for Systematic Reviews of Interventions; Deeks 2017): 0% to 40% might not be important; 30% to 60% may represent moderate heterogeneity; 50% to 90% may represent substantial heterogeneity; 75% to 100% represents considerable heterogeneity. We considered the range of I² of 30% to 60% and a P value < 0.05 as indicating substantial statistical heterogeneity. If we found significant statistical heterogeneity, we explored the potential of heterogeneity. When we considered that studies were suitable to combine in a meta‐analysis based on clinical and methodological characteristics, we pooled the effect estimates provided using a random‐effects model.

Assessment of transitivity and consistency across treatment comparisons

The validity of network meta‐analysis relied on the transitivity assumption. We assessed the assumption of transitivity by comparing the distribution of potential effect modifiers across different pair‐wise comparisons. Potential effect modifiers included age, baseline severity of eczema, co‐intervention, or dosage regimen. We compared the distribution of these variables across direct comparisons available in the network (Salanti 2012).

We used the design‐by‐treatment model to evaluate inconsistency in the network as a whole. In addition, we used loop‐specific and node‐splitting methods to identify which piece of evidence is responsible for the inconsistency (Cipriani 2013; Donegan 2013; Salanti 2012).

Assessment of reporting biases

Reporting biases, such as selective outcome reporting bias, could occur in systematic reviews even when a well‐designed and comprehensive search strategy is applied. We constructed and visually assessed asymmetry of comparison‐adjusted funnel plots to investigate possible small‐study effects for any outcome of the network (Chaimani 2013). We included all comparisons in the plot.

Data synthesis

We started by applying a random‐effects model to the pair‐wise meta‐analyses to calculate pooled RRs, MDs, or SMDs with corresponding 95% CIs for each comparison. We performed pair‐wise meta‐analyses for all outcomes and comparisons.

Next, we performed the network meta‐analysis by using White’s frequentist framework when consistency and inconsistency models were formulated as multi‐variate random‐effects meta‐analyses (White 2012). We combined direct and indirect evidence from any pair of interventions to generate mixed treatment effect sizes as pooled RRs, pooled MDs, or pooled SMDs, with corresponding 95% CIs, and we presented them in league tables. To assess whether direct and indirect estimates are consistent, which is the key assumption of multiple‐treatment meta‐analysis, we tested for inconsistency by using a design‐by‐treatment interaction model; if the two‐sided P of the test = 0.05, this indicates significant global inconsistency (Cipriani 2013; Donegan 2013).

We performed NMA on all outcomes including both efficacy and safety outcomes at week 16 of follow‐up time or nearest (short‐term duration) and at a time point longer than 16 weeks (long‐term duration). Our primary analysis comprised RCTs that included participants with moderate to severe eczema who used systemic immunosuppressive treatments administered by any route. In network meta‐analysis, our main comparisons included dupilumab, mepolizumab, tralokinumab, omalizumab, or other biological agents, along with systemic conventional treatments such as cyclosporin A (ciclosporin), methotrexate, azathioprine, mycophenolate, and systemic corticosteroid compared with placebo (plus standard topical therapies) or other active agents.

We used Stata version 15 to perform standard pair‐wise and network meta‐analysis (Stata 2017), using the self‐programmed Stata routines described elsewhere for network meta‐analysis (Chaimani 2013; Salanti 2018).

Subgroup analysis and investigation of heterogeneity

We planned to perform the following subgroup analyses of participant characteristics, which may result in different treatment responses if there are sufficient studies to examine different subgroups for comparison. We formally evaluated differences in treatment effects between subgroups using Stata to test these differences (Stata 2017).

• Age of participants: for example, children (age ≤ 12 years, age > 12 years) versus adults.

• Duration: short‐term treatment (≤ 16 weeks) versus long‐term or maintenance treatment (> 16 weeks).

• Severity of eczema at baseline (moderate versus severe).

• Dosage schedule.

Given that outcome data were not separately available for those ≤ 12 and >12 years of age nor for those with moderate and severe eczema in the included studies, we were not able to perform subgroup analyses by age of participant and severity of eczema at baseline. Therefore, we performed subgroup analysis based on dosage schedule and duration of follow‐up for both pair‐wise and network meta‐analyses of EASI75, POEM score, SAEs, and infection outcomes.

Sensitivity analysis

To investigate the robustness of our findings, we performed a series of sensitivity analyses using Stata for both pair‐wise and NMAs of EASI75 outcomes at short‐term and long‐term durations as follows (Stata 2017).

• Inclusion of studies with low overall risk of bias.

• Exclusion of studies with small sample size (i.e. < 25th percentile for total number of participants).

Summary of findings and assessment of the certainty of the evidence

We included seven 'Summary of findings' (SoF) tables in our review reporting short‐term outcomes of EASI75, POEM score, IGA 0/1, SAEs, and infection, and long‐term outcomes of EASI75 and SAEs, and we presented the overall grading of evidence for these seven main outcomes in each SoF table. The elements reported in each SoF table were network geometry, relative effect estimates of all interventions of each outcome compared with placebo, certainty of evidence for NMA estimates with judgements for downgrading the body of evidence, and ranking of treatment and its uncertainty (Yepes‐Nuñez 2019). We chose the comparisons for the ‘Summary of findings’ tables based on their importance for decision makers, rather than the availability of evidence.

For each outcome, we estimated overall risk in the placebo group of all included studies using a meta‐analysis. We presented the quality of evidence in all SoF tables, using the Confidence in Network Meta‐Analysis (CINeMA) software, online version (CINeMA 2017). This software was developed from the GRADE framework (Salanti 2014). It is based on a methodological framework that considers six domains: within‐study bias (risk of bias), across‐studies bias, indirectness, imprecision, heterogeneity, and incoherence (Salanti 2014).

For within‐study bias, we assessed the risk of bias of each direct estimate and then integrated these judgements with the contribution of each direct estimate to the network estimates. For across‐studies bias, we considered publication bias. We assessed publication bias by considering the comprehensive search strategy that we performed and used comparison‐adjusted funnel plots that test the presence of small‐study effects in the network to guide our judgement. We assessed indirectness by evaluating the distribution of potential effect modifiers (baseline demographic and clinical characteristics of participants) and the relevance of each study to the research question. We assessed imprecision by focusing on the CIs of network meta‐analysis treatment effect estimates, then comparing the range of treatment effects included in the 95% CI with the range of equivalence. For heterogeneity, network meta‐analysis estimates of between‐study variance and prediction intervals were considered. We assessed incoherence when estimates from direct and indirect evidence conflict (inconsistency test), using global inconsistency in the networks. We summarised judgements across the six domains to obtain four levels of confidence for each relative treatment effect, corresponding to the usual GRADE approach: very low, low, moderate, or high. At least two review authors (of RS, PD, ALL, and NC) independently graded the quality of evidence for each treatment effect compared with placebo. If they disagreed, the two review authors discussed the decision within the review team until consensus was reached.

Results

Description of studies

Results of the search

Searches of five databases yielded 5377 records (see Electronic searches). We identified an additional 12 studies by handsearching and by contacting clinical experts. Our searches of the trials registers revealed 1336 further studies. Our screening of the reference lists of included trial publications did not reveal additional RCTs. We therefore had a total of 6725 records.

Once duplicates had been removed, we had a total of 6242 records. We excluded 6054 records based on review of titles and abstracts. We obtained the full text of the remaining 188 records. We excluded 20 studies (see Characteristics of excluded studies). We found no studies awaiting classification. We identified 28 ongoing studies (see Characteristics of ongoing studies). We included 74 studies reported in 140 references. For further details of our screening process, see the study flow diagram (Figure 2).

1.

Study flow diagram.

Included studies

This review included 74 trials. A total of 8177 participants were randomised to different interventions.

Trial design

Both parallel and cross‐over trial designs were included, but most trials used a parallel‐group design (n = 60; 81%). The remaining 14 trials were designed as cross‐over trials (Allen 1991; Berth‐Jones 1991; Berth‐Jones 2002; Heddle 1984; Khattri 2017; Koppelhus 2014; Kwon 2013; La Rosa 1995; Munro 1994; Salek 1993; Sowden 1991; Tzaneva 2010; Wahlgren 1990; Zurbriggen 1999). Of these 14, only eight (57%) reported washout periods, which ranged from one to six weeks depending on the half‐life of each systemic treatment (Berth‐Jones 2002; Heddle 1984; Khattri 2017; Koppelhus 2014; La Rosa 1995; Munro 1994; Tzaneva 2010; Wahlgren 1990). We believe the washout period of these trials is sufficient as it is longer than 5 times the half‐life of the interventions. However, five trials evaluating the effects of cyclosporin A in severe AD included no washout period between treatment schedules (Allen 1991; Kwon 2013; Salek 1993; Sowden 1991; Zurbriggen 1999), and one trial did not provide information on the washout period (Berth‐Jones 1991).

In total, 17 trials were multi‐arm studies; 16 multi‐arm trials assessed the same experimental drug at multiple dose levels (Bissonnette 2019; Blauvelt 2017; de Bruin‐Weller 2018; Guttman‐Yassky 2019a; Jang 2000; NCT01945086; NCT02395133; NCT02780167; NCT03054428; Nemoto 2016; Simpson 2016; Simpson 2018; Simpson 2019a; Thaci 2016; Wollenberg 2019; Neoral group 2008). Only 13 trials (17.6%) were carried out before the year 2000 (Allen 1991; Berth‐Jones 1991; Cordero 1999; Hanifin 1993; Heddle 1984; La Rosa 1995; Munro 1994; Salek 1993; Sowden 1991; van Joost 1994; Wahlgren 1990; Zonneveld 1996; Zurbriggen 1999).

Setting

In all, 46 trials were multi‐centre trials and 11 were single‐centre trials (Der‐Petrossian 2000; Gerbens 2018; Han 2007; Jin 2015; Khattri 2017; Kwon 2013; NCT02002208; NCT02594098; Tzaneva 2010; Wahlgren 1990; Zurbriggen 1999); for 17 trials, single‐centre or multi‐centre status was not clear (Bangert 2016; Berth‐Jones 2002; Cordero 1999; Haeck 2011; Hamilton 2014a; Iyengar 2013; Jang 2000; Jee 2011; Kim 2016; Munro 1994; Nemoto 2016; Pacor 2004; Paul 2002; Price 2019; Schram 2011; Sowden 1991; Zonneveld 1996. All included trials recruited participants from a hospital setting. These trials took place in Europe (n = 34; 45.9%), were worldwide (n = 15; 20.3%), or were conducted in North America (n = 11; 14.9%), in Asia (n = 10; 13.5%), and in the Middle East (n = 1; 1.4%) (Bemanian 2005). The location was not stated for three trials (Bangert 2016; Berth‐Jones 1991; Han 2007).

Sample size

The mean sample size was 109 participants (range 12 to 1379 participants).

Follow‐up

Most trials (98.6%; 73/74) assessed outcomes in the short term, with follow‐up duration ranging from 2 to 16 weeks, with 4 to 16 weeks (1 to 4 months) most common. Thirty‐three trials assessed outcomes in the long term, with follow‐up duration ranging from 5 to 60 months, with 5 to 12 months most common. However, only three trials had follow‐up longer than one year (Blauvelt 2017; Gerbens 2018; Goujon 2018). Follow‐up occurred from the start of treatment.

Characteristics of participants

Age and gender

We summarised the characteristics of participants in the Characteristics of included studies tables. The mean or median age in included trials ranged from 2 to 84 years, with an overall mean or median age of 32 years. Seven of 74 trials determined the effects of systemic treatment in children with reported overall mean or median age ranging from 3.6 to 14.5 years (Bemanian 2005; El‐Khalawany 2013; Harper 2000; Heddle 1984; Jee 2011; La Rosa 1995; NCT03054428). Trials included more men (54.7%; 3824 participants) than women. Age and gender were unreported for 419 and 902 participants (10 and 20 studies), respectively.

Severity of eczema

All trials included participants with moderate to severe eczema. However, most of the studies (46/74; 62%) included participant with moderate to severe eczema without separately reporting outcomes for moderate or severe disease. Twenty‐eight trials (28/74; 38%) included only participants with severe eczema. Only 30 studies (40%) provided information on the duration of the participants’ condition. Among those reported, the average duration of disease was 23 years (SD 8.4 years), with a range of 1 to 37 years.

Characteristics of interventions

Co‐interventions

Of all the included trials, 60 trials provided a co‐intervention, mainly consisting of emollients or topical corticosteroids, or both (81.1%).

Trial and treatment duration

The total duration of included trials ranged from 2 weeks for prednisolone to 60 months for methotrexate (MTX), whereas treatment duration varied from a single dose (CIM331, KPL‐716) to 60 months of treatment (MTX).

Comparisons

Of the included studies, 70 studies were available for quantitative synthesis, and 29 immunosuppressive agents from three classes of interventions were assessed in this review: (1) conventional immunosuppressive agents; (2) small molecule treatments including PDE‐4 inhibitors, tyrosine kinase inhibitors, and JAK inhibitors; and (3) biological treatments including anti‐CD31 receptors, anti‐interleukin (IL)‐22, anti‐IL‐31, anti‐IL‐13, anti‐IL‐12/23p40, anti‐OX40, anti‐TSLP, anti‐CRTH2, and anti‐IgE monoclonal antibodies.

When different doses of the same intervention were grouped together in one "arm", 62 trials included two parallel arms and one trial included three parallel arms (NCT01552629); 11 trials included one study arm because they compared the same intervention in different doses or dosage forms. The study with three arms compared a biological treatment (QGE031, or ligelizumab) with a conventional systemic treatment (cyclosporin A, or ciclosporin) and placebo.

Most of the included trials were placebo‐controlled (48/74; 65%), 34% were head‐to‐head studies (15% assessed effects of different doses of the same drug), and 1% were multi‐armed studies with both an active comparator and placebo.

Conventional systemic treatments compared with placebo

Fifteen trials compared conventional systemic treatments with placebo.

• Azathioprine (AZA) (n = 2) (Berth‐Jones 2002; Meggitt 2006).

• Cyclosporin A (ciclosporin) (n = 7) (Allen 1991; Berth‐Jones 1991; Munro 1994; Salek 1993; Sowden 1991; van Joost 1994; Wahlgren 1990).

• Intravenous immunoglobulin (IVIG) (n = 2) (Hanifin 1993; Jee 2011).

• Interferon gamma (IFG) (n = 1) (Jang 2000).

• Systemic corticosteroids (n = 3) (Heddle 1984; La Rosa 1995; Price 2019).

Small molecule inhibitors compared with placebo

Four trials compared small molecule inhibitors with placebo.

• Apremilast (n = 1) (Simpson 2019a).

• ASN002 (n = 1) (Bissonnette 2019).

• Baricitinib (n = 1) (Guttman‐Yassky 2019a).

• PF‐04965842 (abrocitinib) (n = 1) (NCT02780167).

Biological treatments compared with placebo

Twenty‐nine trials compared a biological treatment with placebo.

• CIM331 (n = 1) (Nemoto 2016).

• Dupilumab (n = 11) (Beck 2014; Bieber 2014; Blauvelt 2017; Blauvelt 2018; Guttman‐Yassky 2019b; Hamilton 2014a; NCT02395133; NCT03054428; Simpson 2016; Thaci 2016; Tsianakas 2018).

• Fezakinumab (n = 1) (Guttman‐Yassky 2018).

• GBR830 (n = 1) (Guttman‐Yassky 2019c).

• KPL‐716 (n = 1) (Mikhak 2019).

• Lebrikizumab (n = 1) (Simpson 2018).

• Ligelizumab (n = 1) (Bangert 2016).

• Mepolizumab (n = 2) (NCT03055195; Oldhoff 2005).

• Nemolizumab (n = 1) (Ruzicka 2017).

• Omalizumab (n = 2) (Heil 2010; Iyengar 2013).

• OC000459 (timapiprant, anti‐CRTH2 monoclonal antibody; n = 1) (NCT02002208).

• QAW039 (n = 1) (NCT01785602).

• Secukinumab (n = 1) (NCT02594098).

• Tezepelumab (n = 1) (Simpson 2019b).

• Tralokinumab (n = 1) (Wollenberg 2019).

• Ustekinumab (n = 2) (Khattri 2017; NCT01945086).

Head‐to‐head trials

Twenty‐five trials are head‐to‐head trials (34%); 14 of 25 trials compared systemic treatment with other comparators, and 11 trials compared the same intervention in different doses or dosage forms.

• Nine of 25 trials compared CsA with different active interventions including IVIG (Bemanian 2005), transfer factor (Cordero 1999), UVB (Granlund 2001), methotrexate (MTX) (El‐Khalawany 2013; Goujon 2018), extracorporeal photopheresis (Koppelhus 2014), topical tacrolimus (Pacor 2004), mycophenolate sodium (MPS) (Haeck 2011), and systemic corticosteroid (Schmitt 2010).

• Other comparisons include AZA versus MTX (Gerbens 2018; Schram 2011), PUVA versus UVB (Der‐Petrossian 2000), PUVA versus UVA (Tzaneva 2010), and systemic corticosteroid versus cyproheptadine (Han 2007).

• The remaining 11 trials evaluated the same intervention in different doses or dosage forms. Most interventions are CsA (n = 8) (Czech 2000; Harper 2000; Jin 2015; Kim 2016; Kwon 2013; Neoral group 2008; Zonneveld 1996; Zurbriggen 1999); the other interventions are IVIG (Paul 2002), MOR106 (Thaci 2018), and dupilumab (de Bruin‐Weller 2018).

Characteristics of outcomes

A total of 7955 participants from 70 trials provided data on one or more outcomes for data analysis. Four trials that evaluated the effects of dupilumab, ligelizumab, MOR106, and systemic corticosteroids were excluded from any analysis because data from published reports were insufficient (only an abstract was available) (Hamilton 2014a; Bangert 2016; Thaci 2018; Price 2019, respectively). These four trials involved 222 participants (3% of participants included in this review).

One trial evaluated effects of systemic corticosteroids on total clinical severity score at two weeks after randomisation (La Rosa 1995), whereas two trials evaluated effects of systemic treatments at longer than 12 months (Blauvelt 2017; Gerbens 2018). One trial evaluated effects of dupilumab on efficacy outcomes such as EASI75 at 13 months and evaluated safety outcomes such as SAEs and infection rates at 16 months (Blauvelt 2017); another trial evaluated effects of methotrexate compared with azathioprine on SCORAD50 at 60 months after randomisation (Gerbens 2018).

Out of 74 trials, 14 reported the proportion of participants who achieved 75% improvement in EASI75, 21 reported EASI50, 13 reported the proportion of participants who achieved 50% improvement in SCORAD (SCORAD50), one reported the proportion of participants who achieved 50% improvement in affected body surface area (BSA50) or in Six Area, Six Sign Atopic Dermatitis (SASSAD50); 26 reported the proportion of participants who achieved the Investigators' Global Assessment value of clear (0) or almost clear (1) (IGA 0/1), one reported the proportion of participants who achieved IGA50, and eight reported the proportion of participants who achieved 50% or three to four levels of improvement in pruritus or itching score (Pruritus 3/4).

Regarding safety outcomes, 44, 24, and 15 trials reported the proportion of participants with any AEs, SAEs, and infections at the maximal time of follow‐up, respectively.

No trials reported Patient‐Oriented Eczema Measure (POEM)50, Self‐Administered EASI (SA‐EASI)50, Patient‐oriented SCORAD (PO‐SCORAD)50, and long‐term control of flares as measured by the number of relapses that occurred over a period of up to one year.

Funding sources

In all, 52 studies declared a source of funding: 46 studies declared pharmaceutical company funding, six declared unique institutional funding (Beck 2014; Bieber 2014; Der‐Petrossian 2000; Granlund 2001; Mikhak 2019; Simpson 2019a), and 22 did not report a funding source.

Excluded studies

After review of the full text, we excluded 20 studies from this review. The main reason for exclusion was that the trial was a non‐RCT (n = 8); other reasons were that the record was a review article (n = 4), the study was not related to an intervention of interest (n = 4), or the study was not related to moderate to severe AD (n = 3). The remaining study was an animal study (n = 1). Additional details of reasons for exclusion are provided in the Characteristics of excluded studies tables.

Studies awaiting classification

There were no studies awaiting classification.

Ongoing studies

We classified 28 trials as ongoing studies. More details are available in the Characteristics of ongoing studies tables. Most of them compare biological treatments or conventional systemic treatments against placebo (n = 20 and n = 2, respectively). Six ongoing studies are head‐to‐head trials assessing biological treatments versus biological treatments (n = 2) or conventional treatments (n = 2), and conventional treatments versus conventional treatments (n = 2).

Risk of bias in included studies

Figure 3 and Figure 4 present a summary of risk of bias assessments. Briefly, we judged a total of 16 articles (22%) to be at low risk of bias and 21 articles (28%) and 37 articles (50%) to be at unclear and high risk of bias, respectively. Incomplete outcome data was the domain with the largest number of articles at high risk of bias (18 articles; 24%), followed by the blinding of outcome assessor domain (9 articles; 12%). It should be noted that we judged studies available only in abstract form (9 articles; 12%) to be at unclear risk of bias if insufficient information was provided.

2.

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

3.

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

Allocation

Thirty‐eight trials (51.4%) did not provide sufficient information on sequence generation and therefore were judged as having unclear risk of bias; 35 trials (47.2%) provided clear information on sequence generation and were judged as having low risk of bias; we judged the single remaining trial to be at high risk of bias due to use of a quasi‐random method (1.4%). However, only 29 trials (39%) provided sufficient information on allocation concealment and were judged as having low risk of bias; the remainder (45 trials; 61%) were judged as having unclear risk of bias due to lack of information on allocation concealment.

Blinding

For assessing risk of performance bias, we judged a total of 49 studies (66%) as having low risk of bias due to blinding of participants. Some studies, which we judged as having low risk of bias, were not blinding participants, but this practice might not affect the results if the primary outcome was assessed primarily by clinical assessors rather than by participants themselves. We judged 19 studies (26%) as having unclear risk of performance bias because they provided little clear information on blinding of participants.

For assessing risk of detection bias, we judged 45 studies (61%) as having a low risk of bias, 21 studies (28%) as having unclear risk of bias, and 8 studies (11%) as having high risk of bias. Studies with high risk of bias used an open‐label design or reported use of unblinded clinical assessors to perform subjective assessment of the primary outcome.

Incomplete outcome data

A total of 45 studies (61%) reported sufficient information on participants' inclusion and analyses along with the < 20% withdrawal rate of included participants. Therefore, we judged them as having low risk of bias. We judged 11 studies (15%) as having unclear risk of bias due to lack of important information on participants’ inclusion and analyses, and we judged 18 studies (24%) as having high risk of bias because most had a > 20% withdrawal rate.

Selective reporting

We included most included studies (65 trials; 88%) as having low risk of bias in this domain because they reported all pre‐specified outcomes with no evidence of reporting bias. We judged only eight trials (11%) as having unclear risk of bias due to lack of information, and one study (1%) as having high risk of bias due to multiple outcome measures and reporting of unclear information on the measures.

Other potential sources of bias

We considered baseline imbalance as another potential source of bias in studies. We found that three trials (4%) had baseline imbalance and judged them as having a high risk of bias. One was imbalanced in baseline characteristics between study arms, one was imbalanced in the history of previous treatments, and another was imbalanced regarding participant age, duration of disease, and baseline characteristics. We judged 52 trials (70%) as having low risk of bias and 19 trials (26%) as having unclear risk of bias.

Effects of interventions

See: Table 1; Table 2; Table 3; Table 4; Table 5; Table 6; Table 7

See Table 1; Estimates of effects, confidence intervals, and certainty of evidence for the proportion of participants who achieved EASI75 with any systemic intervention compared with placebo in the short term (≤ 16 weeks).

See Table 2; Estimates of effects, confidence intervals, and certainty of evidence for the proportion of participants who achieved EASI75 with any systemic intervention compared with placebo in the long term (> 16 weeks).

See Table 3; Estimates of effects, confidence intervals, and certainty of evidence for Patient‐Oriented Eczema Measure (POEM) scores with any systemic intervention compared with placebo in the short term (≤ 16 weeks).

See Table 4; Estimates of effects, confidence intervals, and certainty of evidence for serious adverse events (SAEs) with any systemic intervention compared with placebo in the short term (≤ 16 weeks).

See Table 5; Estimates of effects, confidence intervals, and certainty of evidence for serious adverse events (SAEs) with any systemic intervention compared with placebo in the long term (> 16 weeks).

See Table 6; Estimates of effects, confidence intervals, and certainty of evidence for infections with any systemic intervention compared with placebo in the short term (≤ 16 weeks).

See Table 7; Estimates of effects, confidence intervals, and certainty of evidence for the proportion of participants who achieved an Investigators’ Global Assessment (IGA 0/1) of clear or almost clear with any systemic intervention compared with placebo in the short term (≤ 16 weeks).

Our summary of findings for the main comparisons provides overall estimates of treatment effects compared with placebo and the certainty of available evidence for the six primary outcomes including EASI75, POEM score, SAEs, and infections at a short‐term duration of follow‐up (≤ 16 weeks), and EASI75 and SAEs at a long‐term duration of follow‐up (> 16 weeks). In addition, we included "IGA 0/1" as a secondary outcome in the SoF table because it shows the comparison effect between biological treatments and conventional treatments such as azathioprine (AZA), cyclosporin A (CsA, ciclosporin), and methotrexate (MTX). All outcomes except EASI75 assessed at a long‐term duration of follow‐up were obtained through network meta‐analysis. Most of the patient population included in this review was recruited from trials investigating dupilumab, whereas AZA and CsA were the most conventional systemic treatments investigated in included trials. All included participants had moderate to severe eczema.

Figure 1 shows network diagrams for the seven main outcomes presented in the review. The size of the nodes is proportionate to the total number of participants allocated to each intervention, and the thickness of the lines is proportionate to the number of studies evaluating each direct comparison. The colours of nodes indicate the number of studies with each risk of bias. Red, yellow, and green colours represent high, unclear, and low risk of bias, respectively.

We included a total of 70 trials, involving 7955 participants (97% of included participants in this review) in the classical or network meta‐analysis for at least one of the outcomes. We excluded from any analysis four trials that evaluated the effects of dupilumab, ligelizumab, MOR106, and systemic corticosteroids because data were insufficient (only abstracts were available) (Hamilton 2014a; Bangert 2016; Thaci 2018; Price 2019, respectively. These four trials involved 222 participants (3% of participants included in this review). Most times to follow‐up for analyses of short‐term outcomes were 1 to 4 months, and for analysis of long‐term outcomes 5 to 12 months. One trial evaluated the effects of systemic corticosteroid at two weeks after randomisation (La Rosa 1995), whereas two trials evaluated the effects of systemic treatments beyond 12 months (Blauvelt 2017; Gerbens 2018). Of these two trials, one evaluated the effects of dupilumab compared with placebo at 13 and 16 months (Blauvelt 2017), and the other evaluated the effects of MTX compared with AZA at 60 months after randomisation (Gerbens 2018).

Primary outcomes

We present treatment estimates for pair‐wise meta‐analyses or direct evidence of all primary outcomes in Figure 5Figure 6Figure 7Figure 8Figure 9Figure 10Figure 11 and Figure 12. Treatment effect estimates for network meta‐analysis and ranking estimated effect for each treatment for all primary outcomes are presented in Figure 13Figure 14Figure 15Figure 16Figure 17Figure 18Figure 19Figure 20Figure 21 and Figure 22. Assessment of study and clinical characteristics of included studies revealed no clinically significant differences among studies included in the network meta‐analysis. We found no inconsistency and no evidence of violation of transitivity assumptions in all networks.

5.

Direct effects of main analysis of short‐term EASI75. Main analysis of direct summary effects of all comparisons for achieving improvement of 75% on the Eczema Area and Severity Index (EASI75) during short‐term follow‐up.

6.

Sensitivity analysis including only studies with low risk of bias for direct effects of EASI75. Sensitivity analysis including only studies with low risk of bias of direct summary effects of all comparisons for achieving improvement of 75% on the Eczema Area and Severity Index (EASI75) during short‐term follow‐up.

7.

Sensitivity analysis excluding studies with small sample size for direct effects of EASI75. Sensitivity analysis including only studies with low risk of bias of direct summary effects of all comparisons for achieving improvement of 75% on the Eczema Area and Severity Index (EASI75) during short‐term follow‐up.

8.

Direct effects of long‐term EASI75. Direct summary effects of all comparisons for achieving improvement of 75% on the Eczema Area and Severity Index (EASI75) during long‐term follow‐up.

9.

Direct effects of short‐term POEM. Direct summary effects of all comparisons for Patient‐Oriented Eczema Measure score (POEM) during short‐term follow‐up. AZA: azathioprine; MTX: methotrexate.

10.

Direct effects of short‐term SAEs. Direct summary effects of all comparisons for the number of serious adverse events (SAEs) during short‐term follow‐up. CsA: cyclosporin A; CsAI: cyclosporin A increased regimen (initial dose 3 mg/kg/d, then increased to 5 mg/kg/d); PF: PF‐04965842 (abrocitinib); steroids: corticosteroids.

11.

Direct effects of long‐term SAEs. Direct summary effects of all comparisons for the number of serious adverse events (SAEs) during long‐term follow‐up.CsA: cyclosporin A; MTX: methotrexate.

12.

Direct effects of short‐term infections. Direct summary effects of all comparisons for the number of infections during short‐term follow‐up. AZA: azathioprine; CsA: cyclosporin A; CsAI: cyclosporin A increased regimen (initial dose 3 mg/kg/d, then increased to 5 mg/kg/d); MTX: methotrexate; steroids: corticosteroids.

13.

League tablespresent main and sensitivity analyses for EASI75 outcomes during short‐term follow‐up. Relative effects of systemic interventions as estimated from the network meta‐analysis for achieving improvement of 75% on the Eczema Area and Severity Index (EASI75) during short‐term follow‐up; a. Main analysis, b. Sensitivity analysis including only studies with low risk of bias, c. Sensitivity analysis excluding studies with small sample size. Interventions are ordered by the names of interventions. Results are the risk ratios (RRs) (95% confidence intervals (CIs)) from the network meta‐analysis between the column‐defining intervention and the row‐defining intervention. Comparisons should be read from left to right. Numbers in bold and underlined represent statistically significant results.

14.

Interval plot for main analysis of short‐term EASI75. Network meta‐analysis estimates of systemic interventions versus placebo for achieving improvement of 75% on the Eczema Area and Severity Index (EASI75) during short‐term follow‐up. Red lines represent prediction intervals.

15.

Interval plot for sensitivity analysis of short‐term EASI75. Network meta‐analysis estimates of systemic interventions versus placebo for achieving improvement of 75% on the Eczema Area and Severity Index (EASI75) during short‐term follow‐up. Sensitivity analysis including only studies with low risk of bias and excluding studies with small sample size. Red lines represent prediction intervals.

16.

Ranking plot for main analysis of short‐term EASI75 and sensitivity analysis. Network meta‐analysis estimates of ranking to be the best among all systemic interventions for achieving improvement of 75% on the Eczema Area and Severity Index (EASI75) during short‐term follow‐up. a. Main analysis. b. Sensitivity analysis including only studies with low risk of bias. c. Sensitivity analysis excluding studies with small sample size.

17.

Cluster rank plot between SUCRA for effectiveness and safety. Ranking plot representing simultaneously the efficacy (x‐axis, EASI75 for short‐term measurement) and safety (y‐axis, for short‐term measurement) of all interventions in patients with moderate to severe eczema. Optimal treatment should be characterised by high efficacy and safety and should be seen in the right upper corner of this graph. The different colours represent different groups of interventions showing their performance on both outcomes simultaneously. Interventions belonging to the same group are assumed to have similar performance when the two primary outcomes are considered jointly. SAEs: serious adverse events.

18.

League tables present primary safety outcomes including SAEs and infections. Relative effects of systemic interventions as estimated from the network meta‐analysis for the number of serious adverse events (SAEs) during short‐term and long‐term follow‐up, and for the number of infections during short‐term follow‐up. Interventions are ordered by the name of interventions. Results are the risk ratios (RRs) (95% confidence intervals; 95% CIs) from the network meta‐analysis between the column‐defining intervention and row‐defining intervention. Comparisons should be read from left to right. Numbers in bold and underlined represent statistically significant results. PF: PF‐04965842 (Abrocitinib); CsA: cyclosporine A; CsAI: cyclosporine A increased regimen (initial dose 3 mg/kg/day then increase to 5 mg/kg/day); MTX: methotrexate; AZA: azathioprine; steroids: corticosteroids.

19.

Interval plot for short‐term and long‐term SAEs. Network meta‐analysis estimates of systemic interventions versus placebo for the number of serious adverse events (SAEs) during short‐term and long‐term follow‐up. PF: PF‐04965842 (abrocitinib); CsA: cyclosporin A; MTX: methotrexate. Red lines represent prediction intervals.

20.

Ranking plot for short‐term and long‐term SAEs. Network meta‐analysis estimates of ranking to be the best among all systemic interventions for the number of serious adverse events (SAEs) during short‐term and long‐term follow‐up. CsA: cyclosporin; MTX: methotrexate; PF: PF‐04965842 (abrocitinib).

21.

Interval plot for short‐term infections. Network meta‐analysis estimates for systemic interventions versus placebo for the number of infections during short‐term and long‐term follow‐up. AZA: azathioprine; CsA: cyclosporin A; CsAI: cyclosporin A increased regimen (initial dose 3 mg/kg/d, then increased to 5 mg/kg/d); MTX: methotrexate; steroids: corticosteroids.

22.

Ranking plot for short‐term infections. Network meta‐analysis estimates of ranking to be the best among all systemic interventions for infections during short‐term follow‐up. PF: PF‐04965842 (abrocitinib).

1. Clinical severity of eczema as measured by a validated or objective measure

1.1 Proportion of participants who achieved 75% improvement in EASI (EASI75) during short‐term follow‐up (< 16 weeks)

Thirteen trials (3851 participants) reported EASI75 at short‐term (≤ 16 weeks) follow‐up (Beck 2014; Bissonnette 2019; Blauvelt 2017; Blauvelt 2018; de Bruin‐Weller 2018; Guttman‐Yassky 2018; Guttman‐Yassky 2019c; NCT03054428; NCT01945086; Simpson 2016; Simpson 2018; Simpson 2019b; Wollenberg 2019). A total of eight interventions including placebo, ASN002, dupilumab, GBR830, lebrikizumab, tezepelumab, tralokinumab, and ustekinumab were compared with each other. All were connected to the network (Figure 1). Time to follow‐up for trials reporting this outcome ranged from one month to four months.

Direct evidence

Treatment estimates of achieving EASI75 for pair‐wise meta‐analyses are presented in Figure 5. A total of eight trials reported the effect of dupilumab on EASI75, whereas only one trial investigated another intervention. Among the eight interventions, dupilumab and tralokinumab were more effective than placebo (risk ratio (RR) 3.04, 95% confidence interval (CI) 2.53 to 3.65; RR 2.54, 95% CI 1.31 to 4.94, respectively). Six trials including four interventions (ASN002, dupilumab, GBR830, and tezepelumab) had low risk of bias. Analysis of these alone showed that dupilumab is still more effective for achieving EASI75 compared to placebo, with a reduction in the degree of heterogeneity (I²) from 40.9% to 28.3% (Figure 6).

Network meta‐analysis

The summary relative effects for the proportion of participants who achieve EASI75 from the network meta‐analysis are presented in league tables (Figure 13), as well as in an interval plot (Figure 14). We found no evidence of inconsistency for the main analysis of this outcome (Figure 13, and Table 9). Assessment of study and clinical characteristics of included studies revealed no clinically significant differences among studies included in the network meta‐analysis. In terms of achieving EASI75, dupilumab and tralokinumab were superior to placebo (RR 3.04, 95% CI, 2.51 to 3.69; RR 2.54, 95% CI 1.21 to 5.34, respectively). These results supported the finding from direct evidence. Dupilumab was probably associated with a higher likelihood of achieving EASI75 compared to lebrikizumab (RR 2.18, 95% CI 1.25 to 3.81) and ustekinumab (RR 3.35, 95% CI 1.01 to 11.10). When only trials with low risk of bias were included, only dupilumab was still more effective than placebo (RR 2.53, 95% CI, 2.04 to 3.15) for this outcome. The main and sensitivity analyses of results of comparisons between systemic treatments are presented in Figure 13Figure 14 and Figure 15. Results show no clear differences between the other interventions.

2. Inconsistency tests for main and subgroup analyses by dosing regimen.

Outcomes	Inconsistency test (P value)
	Main analysis	Subgroup analysis
1. EASI75 at short‐term measurement	0.8739	0.4321
2. EASI75 at long‐term measurement	Data insufficiency	Data insufficiency
3. POEM score at short‐term measurement	Data insufficiency	0.2783
4. POEM score at long‐term measurement	Data insufficiency	Data insufficiency
5. Infections at short‐term measurement	0.2266	0.1385
6. Infections at long‐term measurement	Data insufficiency	Data insufficiency
7. SAEs at short‐term measurement	0.6373	0.4021
8. SAEs at long‐term measurement	0.4744	0.5327

EASI75:

POEM:

SAEs:

Ranking analysis

Ranking analysis for short‐term EASI75 outcomes performed with SUCRA strongly suggest that dupilumab was the most effective treatment among all systemic treatments in the network (versus placebo: 3.04, 95% CI 2.51 to 3.69; SUCRA = 92.7; high‐certainty evidence), followed by tralokinumab (versus placebo: RR 2.54, 95% CI 1.21 to 5.34; SUCRA = 72; low‐certainty evidence) and tezepelumab (versus placebo: RR 2.54, 95% CI 1.21 to 5.34; SUCRA = 49.6; low‐certainty evidence). The ranking effects of other systemic treatments are presented in Figure 16 and in Table 1.

Relationship between EASI75 and serious adverse events during short‐term follow‐up

A cluster rank plot between SUCRA for effectiveness (EASI75) and safety (SAEs) is presented in Figure 17. A ranking plot simultaneously represents efficacy (x‐axis, EASI75 at short‐term follow‐up) and safety (y‐axis, SAEs at short‐term follow‐up) of all interventions for patients with moderate to severe eczema. The optimal treatment should be characterised by high efficacy and safety and therefore should be presented in the right upper corner of this graph. The different colours represent different groups of interventions when their performance on both outcomes are considered simultaneously. Interventions belonging to the same group are assumed to have a similar performance when the two primary outcomes are considered jointly. According to the results presented in the cluster rank plot, dupilumab was the systemic treatment option that demonstrated a better compromise between efficacy and safety outcomes (Figure 17).

1.2 Proportion of participants who achieve 75% improvement in EASI (EASI75) during long‐term follow‐up

Three trials with 1241 participants reported EASI75 outcomes at long‐term (>16 weeks) follow‐up (Blauvelt 2017; NCT01945086; NCT02395133). A total of three interventions including placebo, dupilumab, and ustekinumab were compared with each other. All were connected to the network (Figure 1). Time to follow‐up for trials reporting this outcome ranged from 6 months to 13 months. Due to insufficiency of outcome data, network meta‐analysis and ranking of treatment effect for each treatment for this outcome were not performed.

Direct evidence

Treatment estimates for achieving EASI75 at long‐term follow‐up for pair‐wise meta‐analyses are reported in Figure 8. Two trials reported the effect of dupilumab on this outcome. We are uncertain of the effect of dupilumab on this outcome when compared with placebo (RR 2.59, 95% CI 1.87 to 3.60) due to very low certainty of the estimate. This is related to concerns about within‐study bias and major concerns about heterogeneity. Treatment estimates with grading of certainty of evidence for achieving EASI75 at long‐term follow‐up for each systemic treatment compared with placebo are presented in Table 2.

2. Patient‐reported symptoms as measured by a validated measure

2.1 Patient‐Oriented Eczema Measure (POEM) scores during short‐term follow‐up (< 16 weeks)

Six trials reported in five papers and involving 2680 participants reported POEM scores at short‐term follow‐up (Blauvelt 2017; Blauvelt 2018; de Bruin‐Weller 2018; Schram 2011; Simpson 2016). A total of four interventions including placebo, dupilumab, azathioprine (AZA), and methotrexate (MTX) were compared with each other. Because only one trial evaluated the effect of AZA compared with MTX, we noted no connection to dupilumab between them (Figure 1). Therefore, only direct evidence for each treatment for this outcome was estimated. Time of follow‐up of trials reporting this outcome ranged from one month to four months.

Direct evidence

Treatment estimates of POEM scores at short‐term follow‐up for pair‐wise meta‐analyses are reported in Figure 9. Five trials reported the effect of dupilumab compared with placebo, and one trial reported the effect of AZA compared with MTX for this outcome. The results of direct evidence strongly suggest that dupilumab was more effective than placebo (mean difference (MD) ‐7.3, 95% CI ‐8 to ‐6.61) with high certainty of the estimate. The treatment estimate with grading of certainty of evidence for POEM score at short‐term follow‐up for dupilumab compared with placebo is available in Table 3.

3. Safety outcomes

3.1 Proportion of participants experiencing serious adverse events (SAEs) during short‐term follow‐up (< 16 weeks)

A total of 19 trials presented in 18 papers and involving 4088 participants reported outcomes of SAEs at short‐term follow‐up. A total of 13 interventions including placebo, apremilast, baricitinib, CsA (ciclosporin), and CsAI underwent an increased regimen: CsA with initial dose of 3 mg/kg/d was increased to a maximum dose of 5 mg/kg/d; dupilumab, lebrikizumab, PF‐04965842 (abrocitinib), QAW039, corticosteroid, tezepelumab, timapiprant, and tralokinumab were compared with each other. However, only 10 interventions from 17 trials (including 3972 participants) were connected to the network and were used for network meta‐analysis (Figure 1). Three interventions (CsA, CsAI, and corticosteroids) from two trials were not included in the network meta‐analysis because of lack of connection (Schmitt 2010; Zonneveld 1996). The results of unconnected interventions are presented in direct evidence in Figure 10. Time to follow‐up for trials reporting this outcome ranged from one month to four months.

Direct evidence

Treatment estimates for the proportion of participants with SAEs at short‐term follow‐up for pair‐wise meta‐analyses are reported in Figure 10. A total of nine trials from eight papers provided data on participants with SAEs for dupilumab, whereas only one trial reported each of the other interventions. Among the 10 included interventions, dupilumab was safer than placebo (RR 0.35, 95% CI 0.19 to 0.64). There was no clear difference between the other interventions and placebo.

Network meta‐analysis

Summary relative effects for participants with SAEs from the network meta‐analysis are presented in league tables (Figure 18), as well as in an interval plot (Figure 19). We found no evidence of inconsistency for this outcome (Figure 18 and Table 9). QAW039 and dupilumab appeared safer than placebo in terms of having a lower proportion of participants with SAEs at short‐term follow‐up. Among the active treatments, apremilast and baricitinib appeared to be associated with a higher rate of SAEs compared to QAW039 (RR 41.99, 95% CI 1.09 to 1610.39; RR 51.85, 95% CI 1.36 to 1978.53). There was no difference between other active treatments for this outcome.

Ranking analysis

Ranking analysis for short‐term SAEs performed with SUCRA suggests that QAW039 was the best treatment (versus placebo: RR 0.09, 95% CI 0.01 to 0.76; SUCRA = 94.2; moderate‐certainty evidence), followed by dupilumab (versus placebo: RR 0.37, 95% CI 0.23 to 0.59; SUCRA = 75.5; low‐certainty evidence) and timapiprant (versus placebo: RR 0.34, 95% CI 0.07 to 1.62; SUCRA = 74.0; low‐certainty evidence). The ranking effects of other systemic treatments are shown in Figure 20 and in Table 4. However, these results should be interpreted with caution as the results of only one trial were available for most of the treatments analysed (low‐ to moderate‐certainty evidence).

3.2 Proportion of participants experiencing serious adverse events (SAEs) during long‐term (> 16 weeks) follow‐up

A total of six trials including 1720 participants reported SAEs during long‐term follow‐up. Six interventions including placebo, CsA, dupilumab, fezakinumab, MTX, and QGE031 were compared with each other. All interventions were connected to the network (Figure 1). Length of follow‐up for five of these trials ranged from five months to nine months, and one trial reported SAEs at 16 months (Blauvelt 2017).

Direct evidence

Treatment estimates for the proportion of participants with SAEs during long‐term follow‐up for pair‐wise meta‐analyses are reported in Figure 11. A total of three trials reported SAEs for dupilumab, whereas only one trial reported each of the other interventions. There was no clear difference between interventions for this outcome.

Network meta‐analysis

Summary relative effects for participants experiencing SAEs during long‐term follow‐up from the network meta‐analysis are presented in league tables (Figure 18 ), as well as in an interval plot (Figure 19). We found no evidence of inconsistency for this outcome (Figure 18 and Table 9). Treatment estimates for SAEs developing during long‐term follow‐up for each active treatment were no different than those of placebo or the other active treatments.

Ranking analysis

Ranking analysis for long‐term SAEs performed with SUCRA indicated that dupilumab may be the best treatment regarding this outcome (versus placebo: RR 0.68, 95% CI 0.38 to 1.21; SUCRA = 78.8; low‐certainty evidence), followed by MTX (versus placebo: RR 1.15, 95% CI 0.01 to 151.54; SUCRA = 58.5; very low‐certainty evidence). The ranking effects of other systemic treatments are presented in Figure 20 and in Table 5. However, these results should be interpreted with caution for all systemic treatments, as imprecision of estimate effects was a major concern (very low‐ to low‐certainty evidence).

3.3 Proportion of participants developing any infection during short‐term follow‐up

A total of 13 trials including 1658 participants reported the rate of infection during short‐term follow‐up (Beck 2014; Bieber 2014; de Bruin‐Weller 2018; Guttman‐Yassky 2019a; Meggitt 2006; Munro 1994; NCT01785602; NCT02002208; NCT03054428; Nemoto 2016; Schmitt 2010; Simpson 2018; Zonneveld 1996). These studies investigated an increased regimen of 11 interventions including placebo, azathioprine (AZA), CIM331, cyclosporin A (CsA), and CsAI (ciclosporin): CsA initial dose of 3 mg/kg/d was increased to a maximum dose of 5 mg/kg/d); dupilumab, GBR830, lebrikizumab, QAW039, corticosteroids, and timapiprant were compared with each other. All were connected to the network (Figure 1). Length of follow‐up in these trials ranged from one month to four months.

Direct evidence

Treatment estimates for participants developing any infection for pair‐wise meta‐analyses are reported in Figure 12. We noted no difference between the rate of infection seen with dupilumab versus placebo during one to four months of follow‐up. There were no differences in rate of infection between the other active treatments analysed.

Network meta‐analysis

Summary relative effects of rates of infection during short‐term follow‐up from the network meta‐analysis are presented in league tables (Figure 18), as well as in an interval plot (Figure 21). We found no evidence of inconsistency for this outcome (Figure 18 and Table 9). Treatment estimates for rates of infection during short‐term follow‐up for each active treatment were not different from those for placebo or for other active treatments.

Ranking analysis

Ranking analysis for the rate of infection developing during short‐term follow‐up performed with SUCRA indicates that corticosteroids may be the best treatment regarding this outcome (versus placebo: RR 0.15, 95% CI 0.01 to 4.10; SUCRA = 84.7; very low‐certainty evidence), followed by dupilumab (versus placebo: RR 0.43, 95% CI 0.18 to 1.06; SUCRA = 75.4; very low‐certainty evidence). The ranking effects of other systemic treatments are presented in Figure 22 and in Table 6. However, these results should be interpreted with caution for all systemic treatments, as within‐study reporting and imprecision were the major reasons for downgrading of estimate effects to very low‐ or low‐certainty evidence.

Secondary outcomes

1. Proportion of participants who achieved an Investigators' Global Assessment or Physicians' Global Assessment value of 0 or 1 (clear or almost clear) (IGA 0/1) at short‐term and long‐term durations

1.1 Proportion of participants who achieve clear (0) or almost clear (1) on Investigators' Global Assessment (IGA 0/1) during short‐term (≤ 16 weeks) follow‐up

A total of 24 trials including 4823 participants reported an IGA 0/1 outcome during short‐term follow‐up. A total of 16 interventions including placebo, ASN002, AZA, CsA, dupilumab, fezakinumab, GBR830, lebrikizumab, MTX, mepolizumab, PF‐04965842 (PF), QGE031, secukinumab, corticosteroids, tralokinumab, and ustekinumab were compared with each other. All interventions were connected to the network (Figure 1). Length of follow‐up in these trials ranged from one month to four months.

Direct evidence

Treatment estimates for the proportion of participants achieving IGA 0/1 for pair‐wise meta‐analyses are reported in Figure 23. A total of 10 trials reported effects of dupilumab on IGA 0/1. Of the 16 interventions, dupilumab and PF‐04965842 were more effective than placebo (RR 3.58, 95% CI 3.00 to 4.26; RR 3.32, 95% CI, 1.25 to 8.79, respectively). No clear differences between other interventions were observed for this outcome.

23.

Direct effects of short‐term IGA 0/1. Direct summary effects of all comparisons for the proportions of participants who achieved IGA 0/1 during short‐term follow‐up. ASN: ASN002; AZA: azathioprine; CsA: cyclosporin A; MTX: methotrexate; PF: PF‐04965842 (abrocitinib); steroids: corticosteroids.

Network meta‐analysis

Summary relative effects of the proportion of participants achieving IGA 0/1 during short‐term follow‐up from the network meta‐analysis are presented in league tables with no evidence of inconsistency (Figure 24). In terms of achieving IGA 0/1, dupilumab and PF‐04965842 appeared superior to placebo. Compared with other active treatments, dupilumab was associated with a greater chance of achieving IGA 0/1 compared to lebrikizumab and ustekinumab (RR 2.45, 95% CI 1.29 to 4.63; RR 3.44, 95% CI 1.55 to 7.65, respectively). There were no clear differences between the other interventions for this outcome (Figure 24).

24.

League tablepresents short‐term IGA 0/1. Relative effects of systemic interventions as estimated from the network meta‐analysis for achieving Investigators’ Global Assessment value of 0 (clear) or 1 (almost clear) (IGA 0/1) during short‐term follow‐up. Interventions are ordered by the names of interventions. Results are risk ratios (RRs) (95% confidence intervals (CIs)) from the network meta‐analysis between the column‐defining intervention and the row‐defining intervention. Comparisons should be read from left to right. Numbers in bold and underlined represent statistically significant results. ASN: ASN002; AZA: azathioprine; CsA: cyclosporin A; CsAI: cyclosporin A increased regimen (initial dose 3 mg/kg/d, then increased to 5 mg/kg/d); MTX: methotrexate; PF: PF‐04965842 (abrocitinib); steroids: corticosteroids.

Ranking analysis

Ranking analysis for achieving IGA 0/1 performed with SUCRA suggests that dupilumab was the best treatment (versus placebo: RR 3.58, 95% CI 3.00 to 4.26; SUCRA = 84.0; moderate‐certainty evidence), followed by PF‐04965842 (versus placebo: RR 3.32, 95% CI 1.25 to 8.79; SUCRA = 74.6; moderate‐certainty evidence), then CsA (versus placebo: RR 3.67, 95% CI 0.09 to 149.16; SUCRA = 74.4; very low‐certainty evidence). The ranking effects of other systemic treatments are presented in Figure 25 and in Table 7. However, results for PF‐04965842 and CsA should be interpreted with caution, as only one trial evaluated the effect of PF‐04965842, and the trial investigating CsA was very small, including only 19 participants.

25.

Ranking plot for short‐term IGA 0/1. Network meta‐analysis estimates of ranking to be the best among all systemic interventions for IGA 0/1 during short‐term follow‐up. AZA: azathioprine; CsA: cyclosporin A; CsAI: cyclosporin A increased regimen (initial dose 3 mg/kg/d, then increased to 5 mg/kg/d); MTX: methotrexate; PF: PF‐04965842 (abrocitinib); steroids: corticosteroids.

Inconsistency tests

The results of Inconsistency tests indicate there was no evidence of inconsistency between direct and indirect effects for all primary outcomes evaluated in this review (Table 9).

Assessment of small‐study effects

Small‐study effects were assessed using adjusted funnel plots (Figure 26), which show no small‐study effect for all outcomes analysed.

26.

Funnel plot for all primary outcomes. Publication bias test using funnel plot for network meta‐analysis of all primary outcomes.

Sensitivity and subgroup analyses

Sensitivity analyses performed by including only studies with low risk of bias and excluding studies with small sample sizes (i.e. < 25th percentile for total number of participants) were performed for EASI75 during short‐term follow‐up. Results of sensitivity analyses regarding EASI75 outcomes are provided in Figure 6, Figure 7, Figure 13, Figure 15, and Figure 16. Most of the results of sensitivity analyses confirmed results of the main analyses. Dupilumab appeared to be the most effective treatment for achieving EASI75 during short‐term follow‐up.

Subgroup analyses performed by dosing and follow‐up time (short‐ versus long‐term follow‐up) were performed. Network plots for subgroup analyses of all primary outcomes are provided in Figure 27. Summary estimate effects from subgroup analyses are provided in Figure 28, Figure 29, Figure 30, Figure 31, and Figure 32. Most of the results from subgroup analyses indicate that dupilumab 300 mg given every week, dupilumab 200 mg given every week, and dupilumab 300 mg given every 2 weeks appear to be the best treatment options for all primary outcomes analysed.

27.

Network plot for subgroup analysis of all primary outcomes at short‐term and long‐term follow‐up. The size of the nodes is proportionate to the total number of participants allocated to each intervention, and the thickness of the lines is proportionate to the number of studies evaluating each direct comparison. CsA: cyclosporin A; EASI75: achieving improvement of 75% on the Eczema Area and Severity Index; MTX: methotrexate; POEM: Patient‐Oriented Eczema Measure score; SAEs: serious adverse events; a. Subgroup analysis of EASI75 during short‐term follow‐up. b. Subgroup analysis of EASI75 during long‐term follow‐up. c. Subgroup analysis of POEM scores during short‐term follow‐up. d. Subgroup analysis of SAEs during short‐term follow‐up. e. Subgroup analysis of infections during short‐term follow‐up. f. Subgroup analysis of SAEs during long‐term follow‐up.

28.

League tablespresent subgroup analysis of short‐term EASI75. Relative effects of systemic interventions as estimated from the network meta‐analysis for achieving improvement of 75% on the Eczema Area and Severity Index (EASI75) during short‐term follow‐up. Interventions are ordered by the names of interventions. Results are the risk ratios (RRs) (95% confidence intervals (CIs)) from the network meta‐analysis between the column‐defining intervention and the row‐defining intervention. Comparisons should be read from left to right.

29.

League tablefor subgroup analysis of short‐term SAEs. Relative effects of systemic interventions as estimated from the network meta‐analysis for the number of serious adverse events (SAEs) during short‐term and long‐term follow‐up, and for the number of infections during short‐term follow‐up. Interventions are ordered by the names of interventions. Results are the risk ratios (RRs) (95% confidence intervals (CIs)) from the network meta‐analysis between the column‐defining intervention and the row‐defining intervention. Comparisons should be read from left to right. PF: PF‐04965842 (abrocitinib).

30.

League table for subgroup analysis of short‐term infections. Relative effects of systemic interventions as estimated from the network meta‐analysis for the number of infections in short‐term measurements. Interventions are ordered by the names of interventions. Results are the risk ratios (RRs) (95% confidence intervals (CIs)) from the network meta‐analysis between the column‐defining intervention and the row‐defining intervention. Comparisons should be read from left to right.

31.

League tablefor subgroup analysis of long‐term SAEs. Relative effects of systemic interventions as estimated from the network meta‐analysis for the number of serious adverse events (SAEs) during long‐term follow‐up. Interventions are ordered by the names of the interventions. Results are the risk ratios (RRs) (95% confidence intervals (CIs)) from the network meta‐analysis between the column‐defining intervention and the row‐defining intervention. Comparisons should be read from left to right.

32.

SUCRA and ranking for subgroup analysis of all primary outcomes. Network meta‐analysis estimates of ranking to be the best among all systemic interventions for all primary outcomes.

Discussion

Summary of main results

Our study aimed to assess the efficacy and safety of different types of systemic immunosuppressive treatments for moderate to severe eczema. We analysed 74 trials including 8177 participants with eczema, comparing 29 systemic immunosuppressive treatments with placebo or other systemic immunosuppressive treatments. Our primary outcome measures were proportions of participants who achieved 75% improvement in Eczema Area and Severity Index scores (EASI75) and improvement in Patient‐Oriented Eczema Measure (POEM) scores; safety outcomes consisted of the proportions of serious adverse events (SAEs) and any infection; however, no more than 19 studies assessed any of the primary outcomes.

Our findings are presented separately for short‐term (≤ 16 weeks) and long‐term (> 16 weeks) follow‐up and pertain to moderate to severe atopic eczema. However, follow‐up was mainly short term, with only three studies following up with participants for longer than a year. Ciclosporin was the most investigated systemic treatment (24 trials), followed by dupilumab (12 studies).

With a high degree of certainty, network meta‐analysis (NMA) indicates that when compared to placebo, dupilumab is likely to be the more effective treatment for eczema and is ranked highest among the biological treatments in terms of achieving EASI75 and improving POEM scores during short‐term follow‐up (Table 1; Table 3). Dupilumab was the only immunosuppressive agent for which improvement in POEM in the short term was evaluated.

We are uncertain of the effect of dupilumab on achieving EASI75 in the long term when compared against placebo, as the certainty of this evidence is very low (Table 2). We are uncertain how conventional immunosuppressive treatments rank for our primary efficacy or safety outcomes compared with newer treatments such as the biological agent dupilumab due to lack of comparative data.

NMA suggests that tralokinumab may be more effective than placebo in achieving EASI75 in the short term (low‐certainty evidence; Table 1). None of the included studies assessing tralokinumab measured POEM in the short term or EASI75 in the long term.

Based on our NMA, we are uncertain of the effect of ustekinumab on achieving EASI75 in the short or long term when compared with placebo (very low‐certainty evidence; Table 1). None of the included studies assessing ustekinumab measured POEM.

Low‐ and very low‐certainty evidence means we are uncertain how the other immunosuppressive agents in Table 1 and Table 2 influence the achievement of short‐term EASI75 when compared with placebo. Dupilumab and ustekinumab were the only immunosuppressive agents for which achievement of long‐term EASI75 was evaluated.

Compared to placebo, QAW039 and dupilumab may be safer based on association of these treatments with fewer SAEs during short‐term follow‐up, with evidence judged to have a low to moderate degree of certainty. For the other immunosuppressive agents when compared to placebo, we found no difference in SAEs during short‐term follow‐up, but this finding is based on low‐ to very low‐certainty evidence (Table 4).

Evidence of a very low to low degree of certainty indicates there was no difference in the rate of any infection with systemic immunosuppressive treatments compared to placebo during short‐term follow‐up (Table 6).

When safety outcomes during long‐term follow‐up were assessed, evidence (which was of very low to low certainty) indicates there was no statistical difference in the proportions of participants with SAE when any immunosuppressive agent was compared to placebo (Table 5).

We did not identify differences in other adverse events (AEs), but dupilumab is associated with specific AEs, including eye inflammation and eosinophilia.

Overall completeness and applicability of evidence

Given that available studies comparing all systemic immunosuppressive treatments are limited, we employed NMA to enable combination of direct and indirect evidence and ranking of different systemic immunosuppressive treatments in a coherent and methodologically robust way for efficacy and safety outcomes. Our review included 74 trials (up to August 2019) assessing almost all currently used systemic immunosuppressive agents for treatment of eczema. Even though most of the population analysed in this review originated from recent trials investigating dupilumab (n = 4060), all other data obtained from this review covered various interventions and were considered comprehensive and relevant to assessment of treatment options in the management of people with moderate to severe eczema. We followed all key outcomes that met the criteria of the Harmonising Outcome Measures for Eczema initiative (HOME). Assessment of short‐term outcomes yielded an overall good summary of findings for application in clinical settings, but data on long‐term outcomes are currently inadequate.

In all included studies, participants were diagnosed with moderate to severe eczema and were eligible for systemic immunomodulatory therapy, which is reassuring regarding the transitivity assumption. Although non‐systemic treatment options are also widely investigated in trials of eczema, we did not include them in this review, as the population is different and typically includes people with less severe eczema as compared to trials investigating systemic immunosuppressive agents. Moreover, it is not possible to compare all systemic and non‐systemic treatment options in a single coherent analysis given the expected large heterogeneity of all trials.

Our study has some limitations. There is some heterogeneity in the design of the included trials, which led to widely varying placebo responses. This heterogeneity may relate to variable use of background or concomitant therapy such as topical steroids or topical calcineurin inhibitors (topical anti‐inflammatory agents) among studies, which could affect the transitivity assumption. The included population was quite homogeneous, as it was limited to people with moderate to severe eczema who were eligible for systemic immunomodulatory therapy. However, some variation in the use of topical anti‐inflammatory agents in these populations could affect the transitivity assumption. The pooled effect of placebo response actually reflected what would be expected in standard practice in the placebo group.

Evidence of some interventions was informed by only a single randomised controlled trial (RCT), usually with a small number of participants, which led to imprecision and low to very low certainty of evidence.

Although 74 trials were included, only a few compared conventional systemic treatment such as cyclosporin A (ciclosporin), methotrexate, and azathioprine to placebo, leading to lack of a connected network of these treatments for our primary outcomes.

Although eczema is common in children, evidence assessing systemic treatments for this population is limited and inconclusive. In this review, we found only seven trials, with a relatively small sample size, whose participants' ages ranged from 3.6 to 14.5 years. These trials reported the effects of cyclosporin A (Bemanian 2005; Harper 2000), intravenous immunoglobulin (IVIG) (Jee 2011), methotrexate (MTX) (El‐Khalawany 2013), systemic corticosteroids (Heddle 1984; La Rosa 1995), and dupilumab (NCT03054428), which involved various clinical outcomes. Thus, no pooled estimates could be drawn from these trials.

As a result, there is unfortunately an absence of evidence comparing effects of conventional systemic and biological agents or small molecule treatments our primary outcome.

Only 13 trials measured short‐term EASI75; three measured this outcome in the long term. Six trials reported short‐term POEM scores. SAEs were the most reported primary outcome, with 19 trials measuring these in the short term but over two‐thirds fewer studies measuring long‐term SAEs. Thirteen trials reported the rate of infection during short‐term follow‐up. Our secondary outcome 'proportions of participants who achieved a value of 0 or 1 (clear or almost clear) on the Investigators' Global Assessment (IGA 0/1) or on the Physicians' Global Assessment during short‐term and long‐term follow‐up' was the most assessed outcome ‐ evaluated by 24 studies.

Our current protocol does not include plans for regular updating (Sawangjit 2018); however, given that this area of research is rapidly evolving, with new trials being added to the literature regularly, continuous updating of current findings as a living systematic review and NMA should be performed.

Quality of the evidence

The overall degree of certainty of the evidence evaluated ranged from very low to high based on grading using CiNEMA assessment. Specifically, the degree of certainty of the evidence comparing dupilumab and placebo was assessed as high for short‐term EASI75 and improved POEM outcomes and was assessed as low for the incidence of SAEs and very low for the incidence of short‐term infection. Evidence of the effectiveness of other interventions such as ASN002, GBR830, lebrikizumab, tezepelumab, tralokinumab, and ustekinumab was assessed as being of very low to low confidence for all primary outcomes.

The major reason for downgrading the certainty of evidence was the imprecision of results with wide confidence intervals and within‐study bias. Most comparisons yielded very low‐certainty evidence due to these concerns.

Potential biases in the review process

The most important potential bias in the review process was the selection of reported outcomes. Although the protocol was published before this systematic review and network meta‐analysis were conducted, the outcomes of interest were not comprehensively defined. The protocol defined many measures for assessing the clinical severity of eczema, including Eczema Area and Severity Index (EASI), severity SCORing of Atopic Dermatitis Index (SCORAD), Six Area, Six Sign Atopic Dermatitis (SASSAD), and IGA, amongst others. However, we did not specify in the protocol whether these should be continuous or dichotomous outcome measurements. In our review process, we decided to use a dichotomous outcome for our outcomes of interest and a continuous outcome when dichotomous outcomes were not available. Based on this decision, EASI75 was selected as one of our primary outcomes, along with 50% improvement in Patient‐Oriented Eczema Measure score (POEM50). However, few studies reported POEM50 as an outcome; therefore, we selected POEM reported as a continuous score as a primary outcome for our review instead of POEM50.

Our subgroup analyses were well planned to include participant age, duration of treatment, severity of eczema, and dosage schedule. However, because of data availability, it was possible to perform only a subgroup analysis related to treatment duration and dosage schedule. Similar to subgroup analyses, our sensitivity analyses could not be performed as planned. Only inclusion of studies with low risk of bias and exclusion of small sample size could be used for our sensitivity analysis. We could not exclude unblinded studies or industry‐sponsored studies because of data availability.

We initially included some studies with hand eczema in our review, but after consultation with clinical and methodological experts, we decided to exclude studies with hand eczema because causes of this form of eczema are frequently multi‐factorial and are not specifically limited to atopic eczema.

Our main analysis results from combining various doses of medicines, which, when combined into a single mode, can bias treatment effects if less effective dosing regimens are combined with more effective regimens. Therefore, we performed subgroup analysis to assess the effects of interventions based on each dose of medication.

Agreements and disagreements with other studies or reviews

Our findings have demonstrated greater efficacy of dupilumab (for short‐term assessment of EASI75 and POEM) compared to placebo and are consistent with the findings of previous meta‐analyses (Drucker 2020; Snast 2018; Wang 2018).

Snast 2018 included 13 RCTs and 10 observational studies comparing biological agents for treating people with atopic dermatitis. Pair‐wise meta‐analysis revealed that dupilumab 300 mg given every week to every two weeks showed 3.3 times higher efficacy than placebo at 12 to 16 weeks as assessed by EASI75. Other biological agents, including nemolizumab, lebrikizumab, tralokinumab, omalizumab, and ustekinumab, showed similar efficacy or were borderline superior to placebo. However, this meta‐analysis had limitations. The study was performed via a pair‐wise meta‐analysis approach and no efficacy comparison of all biological agents was performed. In addition, a small number of studies were included in each pair‐wise meta‐analysis for each biological agent assessed. This study was also limited by lack of RCTs and variation in outcome measures.

Wang 2018 aimed to determine the efficacy and safety of dupilumab 300 mg given every week or every two weeks for people with atopic dermatitis. This study included six RCTs with 2447 participants. Review authors found that dupilumab was effective in treating atopic dermatitis as assessed by EASI score, percentages of body surface area involvement, pruritus score, and Dermatology Life Quality Index. However, this study lacked long‐term follow‐up results and included only a small number of RCTs. In addition, these review authors evaluated the efficacy of dupilumab but did not evaluate other biological agents.

Drucker 2020 compared the efficacy and safety of systemic immunomodulatory treatments for clinical signs of atopic dermatitis using a network meta‐analysis approach. This review included 39 RCTs with 6360 participants with atopic dermatitis only if they received follow‐up of eight or more weeks to 16 weeks. This is different from our review, which included a total of 74 RCTs with 7913 participants. In addition, we included all RCTs regardless of follow‐up periods. Our included studies had study periods ranging from 2 weeks to 60 months.

The overall finding of our review and of the Drucker review was consistent. Drucker reported that dupilumab 300 mg given every two weeks could improve EASI score by 11.3 points compared to placebo, with high‐certainty evidence as assessed by GRADE, and our review concurred that dupilumab was more effective than placebo with both short‐term follow‐up (≤ 16 weeks) (high‐certainty evidence) and long‐term follow‐up (> 16 weeks) (very low‐certainty evidence). Obviously, the key difference between our review and the Drucker study is that we primarily assessed the efficacy of immunomodulatory treatment using the EASI75, which is defined as the proportion of participants who achieve 75% improvement in EASI, but the Drucker study primarily assessed that effect based on EASI score on a continuous scale. Other differences are that our review is not limited to articles published in English and is not limited by duration of follow‐up, but the Drucker review is limited to English language articles and a follow‐up duration of eight or more weeks to 16 weeks, which might have led to omission of some studies published in other languages and reporting different follow‐up periods, and this could affect the observed findings.

Our review adds important information on the comparison of all available biological agents used to treat eczema via a network meta‐analysis technique and a larger number of RCTs than previous meta‐analyses performed on this topic. In addition, we evaluated both short‐term and long‐term follow‐up outcome data. Our review also highlights the lack of studies directly comparing conventional systemic immunosuppressive agents to newer biological agents or small molecule treatments and the preponderance of recent studies investigating biological agents such as dupilumab.

Authors' conclusions

Implications for practice.

With high certainty of available evidence, we conclude that dupilumab is the most effective of the biological treatments used to treat people with moderate to severe eczema, based on short‐term NMA of EASI75 and POEM. Dupilumab is safer than other agents based on short‐term safety data (≤ 16 weeks).

It is not currently possible to confidently rank the efficacy and safety of conventional immunosuppressive treatments for moderate to severe eczema compared with newer treatments such as biological agents for our primary efficacy and safety outcomes due to limited data.

Based on NMA, when compared to placebo, dupilumab increases the proportion of participants who achieve EASI75 and improves POEM score in the short term (high‐certainty evidence). We are uncertain of the effect of dupilumab on EASI75 in the long term due to very low‐certainty evidence. In addition, lack of long‐term outcome data after cessation of immunosuppressive treatment renders difficulty in drawing conclusion on the long‐term efficacy of any systemic treatment.

Based on NMA, when compared to placebo, tralokinumab may increase the proportion of patients who achieve EASI75 in the short term. Studies evaluating tralokinumab did not assess this outcome in the long term (low‐certainty evidence).

Due to very low‐certainty evidence, we are not certain of the effect of ustekinumab on the proportion of participants achieving EASI75 in the short or long term. This is based on NMA and comparison of ustekinumab to placebo.

Due to low‐ or very low‐certainty evidence, we cannot be sure how other immunosuppressive agents for which our key outcomes were assessed affect the proportion of patients achieving short‐term EASI75. These agents were compared against placebo.

The only immunosuppressive agent used to assess improvement in POEM score in the short term was dupilumab. Dupilumab and ustekinumab were the only immunosuppressive agents for which EASI75 was evaluated in the long term.

Based on low‐ to moderate‐certainty evidence, QAW039 and dupilumab show a lower proportion of participants with SAEs assessed in the short term when compared with placebo. However, no difference is seen in the proportion of participants with SAEs assessed in the short term when other immunosuppressive agents are compared to placebo (low‐ to very low‐certainty evidence).

Based on low‐ or very low‐certainty evidence, we found no evidence of a difference in risk of any infection (measured in the short or long term) or in the proportion of participants with SAEs assessed in the long term when immunosuppressive agents were compared with placebo.

We did not identify differences in other AEs, but dupilumab is associated with specific AEs, including eye inflammation and eosinophilia.

Implications for research.

Despite high certainty of evidence for the efficacy of dupilumab in treating atopic eczema, little is known about its comparative efficacy and safety relative to other biological therapies or conventional systemic immunosuppressive treatments. More primary studies of dupilumab reporting head‐to‐head comparisons with other systemic immunosuppressive treatments are required to evaluate comparative safety and efficacy. In addition, most of the trials included in our network meta‐analysis were placebo controlled, which limited our ability to accurately estimate all of the effect sizes, especially those without direct comparison.

More primary studies with head‐to‐head comparisons are required to evaluate the comparative efficacy and safety of these treatments. More important, comparative studies are needed that investigate long‐term safety and longer‐term efficacy after systemic immunosuppressive treatment has ceased. Studies in children, for whom the burden of eczema is high, are also needed to evaluate safety and efficacy in this patient group.

These should ideally be adequately powered, randomised controlled trials that report on recommended core outcome domains to facilitate treatment comparisons, following the global Harmonizing Outcome Measures for Eczema (HOME) initiative to improve the evidence base.

History

Protocol first published: Issue 11, 2018
Review first published: Issue 9, 2020

Appendices

Appendix 1. Hanifin and Rajka Diagnostic Criteria

Three or more basic features: Pruritus Typical morphology and distribution: Flexural lichenification or linearity in adults Facial and extensor involvement in infants and children Chronic or chronically relapsing dermatitis Personal or family history of atopy (asthma, allergic rhinitis, atopic dermatitis)
Plus 3 or more minor features:
Xerosis Ichthyosis, palmar hyperlinearity, keratosis pilaris Nipple eczema Cheilitis Recurrent conjunctivitis Dennie‐Morgan infraorbital fold Keratoconus Anterior subcapsular cataracts Orbital darkening Facial pallor, facial erythema Pityriasis alba Anterior neck folds White dermatographism, delayed blanc	Immediate (type 1) skin test reactivity Elevated serum IgE Early age of onset Tendency toward cutaneous infections (especially S aureus & HSV), impaired cell‐mediated immunity Tendency toward non‐specific hand or foot dermatitis Itch when sweating Intolerance to wool or lipid solvents Perifollicular accentuation Food intolerance Course influenced by environmental or emotional factors (or both)

Appendix 2. UK Working Party Diagnostic Criteria

Mandatory: An itchy skin condition (parental report of scratching or rubbing in a child)

Plus 3 or more of the following:

History of involvement of the skin creases, such as folds of elbow, behind the knees, fronts of ankles, or around the neck (including cheeks in children under 10) Personal history of asthma or hay fever (or history of AD in a first‐degree relative in children younger than 4) History of general dry skin in the last year Visible flexural eczema (or eczema involving the cheeks, forehead, and outer limbs in children younger than 4) Onset younger than the age of 2 (not used if child is younger than 4)

Appendix 3. Cochrane Skin Specialised Register (CRSW)

1. eczema or dermatitis AND INREGISTER
2. adalimumab or alefacept or alitretinoin or aminopterin or antileukotriene* or apremilast or avakine or azathioprine or beclomethasone or bioferon or biogen or bosatria or BRM* or cyclosporin* or ciclosporin* or dehydrocortisone or dexamethason* or dupilumab or efalizumab or elidel or etanercept or enbrel or flunisolide or HdIVIg or humira or idec or infliximab or immuneron or imukin or IVIG or keliximab or leukast* or mabthera or mepolizumab or mycophenol* or mofetil or MMF or melbex or myfortic or MTX or methylprednis* or methotrexate or omalizumab or olizumab or pimecrolimus or polyferon or prednison* or prednisolone or puva or quinolone* or raptiva or remicade or rituximab or rituxan or rituxin or tacrolimus or tocilizumab or trudexa or xanelim or xolair or zafirlukast* or zileuton* AND INREGISTER
3. leukotriene antagonist* AND INREGISTER
4. peptide fragment* AND INREGISTER
5. recombinant protein* AND INREGISTER
6. thymus hormone* AND INREGISTER
7. tnfr fc fusion protein AND INREGISTER
8. tumor necrosis factor AND INREGISTER
9. tumour necrosis factor AND INREGISTER
10. anti‐TNF AND INREGISTER
11. monoclonal antibod* AND INREGISTER
12. fluocinolone acetonide AND INREGISTER
13. interferon gamma AND INREGISTER
14. intravenous immunoglobulin* AND INREGISTER
15. adrenal cortex hormone* AND INREGISTER
16. biologic* response modifier* AND INREGISTER
17. targeted therap* AND INREGISTER
18. psoralen ultraviolet A therap* AND INREGISTER
19. systemic treatment* AND INREGISTER
20. immuno‐modulatory treatment* AND INREGISTER
21. anti inflammatory treatment* AND INREGISTER
22. Immunosuppressive Agent* AND INREGISTER
23. Anti‐Inflammatory Agent* AND INREGISTER
24. ((systemic or oral* or sublingual* or per os or inhal* or nasal* or parenteral*) and (steroid* or glucosteroid* or corticosteroid* or glucocorticosteroid*)) AND INREGISTER
25. ((biologic*) and (treatment* or therap* or medicine* or drug* or agent* or product*)) AND INREGISTER
26. #2 OR #3 OR #4 OR #5 OR #6 OR #7 OR #8 OR #9 OR #10 OR #11 OR #12 OR #13 OR #14 OR #15 OR #16 OR #17 OR #18 OR #19 OR #20 OR #21 OR #22 OR #23 OR #24 OR #25
27. #1 AND #26

Appendix 4. CENTRAL (Cochrane Library) search strategy

#1 MeSH descriptor: [Eczema] explode all trees
#2 MeSH descriptor: [Dermatitis, Atopic] explode all trees
#3 MeSH descriptor: [Dermatitis] explode all trees
#4 eczema* or dermatitis:ti,ab,kw
#5 #1 or #2 or #3 or #4
#6 MeSH descriptor: [Cyclosporine] explode all trees
#7 (c*closporin* or CyA or Cy‐A or CsA or Cs‐A or csaneoral or neoral or sandimmun*):ti,ab,kw
#8 MeSH descriptor: [Aminopterin] explode all trees
#9 (aminopterin or MTX or methotrexate):ti,ab,kw
#10 MeSH descriptor: [Azathioprine] explode all trees
#11 (az*thioprine or im*uran):ti,ab,kw
#12 MeSH descriptor: [Leukotriene Antagonists] explode all trees
#13 MeSH descriptor: [Quinolines] explode all trees
#14 (antileukotriene* or anti‐leukotriene* or (leukotriene near/3 (antagonist* or block* or inhibitor*))):ti,ab,kw
#15 (leukast* or zafirlukast* or zileuton* or quinoline*):ti,ab,kw
#16 MeSH descriptor: [Peptide Fragments] explode all trees
#17 MeSH descriptor: [Thymus Hormones] explode all trees
#18 MeSH descriptor: [Recombinant Proteins] explode all trees
#19 MeSH descriptor: [Interferon‐gamma] explode all trees
#20 (rIFN* or bioferon or biogen or immuneron or imukin or kw‐2202 or polyferon or ru‐42369 or ru42369 or s‐6810 or sch‐36850 or sun‐4800):ti,ab,kw
#21 ((r or recombinant) near/3 (interferon* or IFN or IFNg or IFNgamma)):ti,ab,kw
#22 ((Interferon* or IFN or IFNg or IFNgamma) near/3 (therap* or treat* or administ* or given or deliver* or systemic* or oral*)):ti,ab,kw
#23 (etanercept or enbrel or tnfr fc fusion protein):ti,ab,kw
#24 alefacept:ti,ab,kw
#25 (((tnf or tumor necrosis factor) near/2 receptor) or (tnf near/2 (fusion near protein*))):ti,ab,kw
#26 MeSH descriptor: [Tumor Necrosis Factor‐alpha] explode all trees
#27 (anti‐TNF* or anti tumo*r necrosis factor):ti,ab,kw
#28 MeSH descriptor: [Antibodies, Monoclonal] explode all trees
#29 ((humanized near/8 (monoclonal* or antibod* or MoAb* or mAb or mAbs or fab*1)) or rhuMAb*):ti,ab,kw
#30 (chim*eric near/3 (monoclonal* or antibod* or MoAb* or mAb or mAbs)):ti,ab,kw
#31 ((against or anti) near IgE near/2 (monoclonal* or antibod* or MoAb* or mAb or mAbs)):ti,ab,kw
#32 ((anti or against or block* or MoAb* or mAb or mAbs or antibod* or monoclonal*) near/3 (IL5 or IL‐5 or interleukin‐5 or LFA1 or LFA‐1 or CD11a)):ti,ab,kw
#33 (omalizumab or (olizumab or hu‐901 or hu901 or tnx‐901 or tnx901 or xolair)):ti,ab,kw
#34 (infliximab or avakine or remicade):ti,ab,kw
#35 (efalizumab or raptiva or xanelim or hu1124 or hu‐1124):ti,ab,kw
#36 (adalimumab or humira or D2E7 or trudexa):ti,ab,kw
#37 (rituximab or idec c2b8 or mabthera or rituxan or rituxin):ti,ab,kw
#38 (keliximab or sb 210396 or sb210396):ti,ab,kw
#39 (mepolizumab or bosatria or sb 240563 or sb240563):ti,ab,kw
#40 MeSH descriptor: [Mycophenolic Acid] explode all trees
#41 (mycophen*lat* or mycophenol* or mycofen*lat* or mycofenol* or mofetil* or MMF or erl‐080* or erl080* or melbex or myfortic or nsc‐129185 or nsc129185):ti,ab,kw
#42 MeSH descriptor: [Immunoglobulins, Intravenous] explode all trees
#43 (((intravenous or IV) near (immune globulin* or IG or immun*globulin* or antibod*)) or IVIG or HdIVIg):ti,ab,kw
#44 MeSH descriptor: [Prednisolone] explode all trees
#45 MeSH descriptor: [Beclomethasone] explode all trees
#46 MeSH descriptor: [Fluocinolone Acetonide] explode all trees
#47 MeSH descriptor: [Adrenal Cortex Hormones] explode all trees
#48 (prednison* or methylprednis* or dehydrocortisone or dexamethason* or beclomethasone or flunisolide):ti,ab,kw
#49 (pimecrolimus or asm981 or asm 981 or elidel):ti,ab,kw
#50 MeSH descriptor: [Tacrolimus] explode all trees
#51 (tacrolimus or fk506 or fk 506):ti,ab,kw
#52 ((systemic or oral* or sublingual* or per os or inhal* or nasal* or parenteral*) near (steroid* or glucosteroid* or corticosteroid* or glucocorticosteroid*)):ti,ab,kw
#53 ((biological* or biologic*) near (treatment* or therap* or medicine* or drug* or agent* or product*)):ti,ab,kw
#54 (biologic* response modifier* or BRM*):ti,ab,kw
#55 targeted therap*:ti,ab,kw
#56 ((systemic* or oral* or sublingual* or "per os" or inhal* or nasal* or parenteral*) near/3 (treat* or agent* or drug* or monotherap* or medication* or medicine* or administ* or given or deliver* or immunosuppres*)):ti,ab,kw
#57 dupilumab:ti,ab,kw
#58 (Puva or Psoralen ultraviolet A Therapy):ti,ab,kw
#59 MeSH descriptor: [PUVA Therapy] explode all trees
#60 alitretinoin:ti,ab,kw
#61 apremilast:ti,ab,kw
#62 tocilizumab:ti,ab,kw
#63 (systemic near immunosuppressive treatment*):ti,ab,kw
#64 immuno‐modulatory treatment*:ti,ab,kw
#65 anti inflammatory treatment*:ti,ab,kw
#66 MeSH descriptor: [Immunosuppressive Agents] this term only
#67 MeSH descriptor: [Anti‐Inflammatory Agents] this term only
#68 {or #6‐#67}
#69 #5 and #68

Appendix 5. MEDLINE (Ovid) search strategy

1. exp Eczema/ or eczema$.ti,ab.
2. exp Dermatitis, Atopic/
3. exp Dermatitis/ or dermatitis.ti,ab.
4. or/1‐3
5. Cyclosporine/
6. (c?closporin* or CyA or Cy‐A or CsA or Cs‐A or csaneoral or neoral or sandimmun*).tw.
7. exp Aminopterin/
8. (aminopterin or MTX or methotrexate).tw.
9. Azathioprine/
10. (az?thioprine or im?uran).tw.
11. Leukotriene Antagonists/
12. Quinolines/
13. (antileukotriene* or anti‐leukotriene* or (leukotriene adj3 (antagonist* or block* or inhibitor*))).tw.
14. (leukast* or zafirlukast* or zileuton* or quinoline*).tw.
15. Peptide Fragments/
16. exp Thymus Hormones/
17. exp Recombinant Proteins/
18. Interferon‐gamma/
19. (rIFN* or bioferon or biogen or immuneron or imukin or kw‐2202 or polyferon or ru‐42369 or ru42369 or s‐6810 or sch‐36850 or sun‐4800).tw.
20. ((r or recombinant) adj3 (interferon* or IFN or IFNg or IFNgamma)).tw.
21. ((Interferon* or IFN or IFNg or IFNgamma) adj3 (therap* or treat* or administ* or given or deliver* or systemic* or oral*)).tw.
22. (etanercept or enbrel or tnfr fc fusion protein).tw.
23. alefacept.tw.
24. (((tnf or tumor necrosis factor) adj2 receptor) or (tnf adj2 (fusion adj protein*))).tw.
25. Tumor Necrosis Factor‐alpha/
26. (anti‐TNF* or anti tumo?r necrosis factor).tw.
27. Antibodies, Monoclonal/
28. ((humanized adj8 (monoclonal* or antibod* or MoAb* or mAb or mAbs or fab*1)) or rhuMAb*).tw.
29. (chim?eric adj3 (monoclonal* or antibod* or MoAb* or mAb or mAbs)).tw.
30. ((against or anti) adj IgE adj2 (monoclonal* or antibod* or MoAb* or mAb or mAbs)).tw.
31. ((anti or against or block* or MoAb* or mAb or mAbs or antibod* or monoclonal*) adj3 (IL5 or IL‐5 or interleukin‐5 or LFA1 or LFA‐1 or CD11a)).tw.
32. (omalizumab or (olizumab or hu‐901 or hu901 or tnx‐901 or tnx901 or xolair)).tw.
33. (infliximab or avakine or remicade).tw.
34. (efalizumab or raptiva or xanelim or hu1124 or hu‐1124).tw.
35. (adalimumab or humira or D2E7 or trudexa).tw.
36. (rituximab or idec c2b8 or mabthera or rituxan or rituxin).tw.
37. (keliximab or sb 210396 or sb210396).tw.
38. (mepolizumab or bosatria or sb 240563 or sb240563).tw.
39. Mycophenolic Acid/
40. (mycophen?lat* or mycophenol* or mycofen?lat* or mycofenol* or mofetil* or MMF or erl‐080* or erl080* or melbex or myfortic or nsc‐129185 or nsc129185).tw.
41. Immunoglobulins, Intravenous/
42. (((intravenous or IV) adj (immune globulin* or IG or immun?globulin* or antibod*)) or IVIG or HdIVIg).tw.
43. exp Prednisolone/ or Beclomethasone/ or exp Fluocinolone Acetonide/ or Adrenal Cortex Hormones/
44. (prednison* or methylprednis* or dehydrocortisone or dexamethason* or beclomethasone or flunisolide).tw.
45. exp pimecrolimus/ or (asm981 or asm 981 or elidel).tw.
46. exp tacrolimus/ or (fk506 or fk 506).tw.
47. ((systemic or oral* or sublingual* or "per os" or inhal* or nasal* or parenteral*) adj3 (steroid* or glucosteroid* or corticosteroid* or glucocorticosteroid*)).tw.
48. ((biological*1 or biologic*1) adj (treatment* or therap* or medicine* or drug* or agent* or product*)).tw.
49. (biologic* response modifier* or BRM*).tw.
50. targeted therap*.tw.
51. dupilumab.tw.
52. Puva.tw. or Puva therapy/ or Psoralen ultraviolet A Therapy.tw.
53. alitretinoin.tw.
54. apremilast.tw.
55. tocilizumab.tw.
56. (systemic adj immunosuppressive treatment$).tw.
57. immuno‐modulatory treatment$.tw.
58. anti inflammatory treatment$.tw.
59. Immunosuppressive Agents/
60. Anti‐Inflammatory Agents/
61. or/5‐60
62. randomized controlled trial.pt.
63. controlled clinical trial.pt.
64. randomized.ab.
65. placebo.ab.
66. clinical trials as topic.sh.
67. randomly.ab.
68. trial.ti.
69. 62 or 63 or 64 or 65 or 66 or 67 or 68
70. exp animals/ not humans.sh.
71. 69 not 70
72. 4 and 61 and 71

[Lines 62‐71: Cochrane Highly Sensitive Search Strategy for identifying randomized trials in MEDLINE: sensitivity‐ and precision‐maximizing version (2008 revision); Ovid format, from section 3.6.1 in Lefebvre C, Glanville J, Briscoe S, Littlewood A, Marshall C, Metzendorf M‐I, et al. Technical Supplement to Chapter 4: Searching for and selecting studies. In: Higgins JPT, Thomas J, Chandler J, Cumpston MS, Li T, Page MJ, Welch VA (eds). Cochrane Handbook for Systematic Reviews of Interventions Version 6. Cochrane, 2019. Available from: www.training.cochrane.org/handbook]

Appendix 6. Embase (Ovid) search strategy

1. exp eczema/
2. eczema$.ti,ab.
3. exp atopic dermatitis/
4. exp dermatitis/
5. dermatitis.ti,ab.
6. or/1‐5
7. cyclosporin/
8. (c?closporin* or CyA or Cy‐A or CsA or Cs‐A or csaneoral or neoral or sandimmun*).tw.
9. exp aminopterin/
10. (aminopterin or MTX or methotrexate).tw.
11. exp methotrexate/
12. azathioprine/
13. (az?thioprine or im?uran).tw.
14. leukotriene receptor blocking agent/
15. quinoline derivative/
16. (antileukotriene* or anti‐leukotriene* or (leukotriene adj3 (antagonist* or block* or inhibitor*))).tw.
17. (leukast* or zafirlukast* or zileuton* or quinoline*).tw.
18. peptide fragment/
19. exp thymus hormone/
20. exp recombinant protein/
21. gamma interferon/
22. (rIFN* or bioferon or biogen or immuneron or imukin or kw‐2202 or polyferon or ru‐42369 or ru42369 or s‐6810 or sch‐36850 or sun‐4800).tw.
23. ((r or recombinant) adj3 (interferon* or IFN or IFNg or IFNgamma)).tw.
24. ((Interferon* or IFN or IFNg or IFNgamma) adj3 (therap* or treat* or administ* or given or deliver* or systemic* or oral*)).tw.
25. (etanercept or enbrel or tnfr fc fusion protein).tw.
26. alefacept.tw.
27. (((tnf or tumor necrosis factor) adj2 receptor) or (tnf adj2 (fusion adj protein*))).tw.
28. tumor necrosis factor/
29. (anti‐TNF* or anti tumo?r necrosis factor).tw.
30. monoclonal antibody/
31. ((humanized adj8 (monoclonal* or antibod* or MoAb* or mAb or mAbs or fab*1)) or rhuMAb*).tw.
32. (chim?eric adj3 (monoclonal* or antibod* or MoAb* or mAb or mAbs)).tw.
33. ((against or anti) adj IgE adj2 (monoclonal* or antibod* or MoAb* or mAb or mAbs)).tw.
34. ((anti or against or block* or MoAb* or mAb or mAbs or antibod* or monoclonal*) adj3 (IL5 or IL‐5 or interleukin‐5 or LFA1 or LFA‐1 or CD11a)).tw.
35. omalizumab/
36. (omalizumab or (olizumab or hu‐901 or hu901 or tnx‐901 or tnx901 or xolair)).tw.
37. infliximab/
38. (infliximab or avakine or remicade).tw.
39. efalizumab/
40. (efalizumab or raptiva or xanelim or hu1124 or hu‐1124).tw.
41. adalimumab/
42. (adalimumab or humira or D2E7 or trudexa).tw.
43. rituximab/
44. (rituximab or idec c2b8 or mabthera or rituxan or rituxin).tw.
45. keliximab/
46. (keliximab or sb 210396 or sb210396).tw.
47. mepolizumab/
48. (mepolizumab or bosatria or sb 240563 or sb240563).tw.
49. mycophenolic acid/
50. (mycophen?lat* or mycophenol* or mycofen?lat* or mycofenol* or mofetil* or MMF or erl‐080* or erl080* or melbex or myfortic or nsc‐129185 or nsc129185).tw.
51. (((intravenous or IV) adj (immune globulin* or IG or immun?globulin* or antibod*)) or IVIG or HdIVIg).tw.
52. exp prednisolone/
53. beclometasone/
54. exp fluocinolone acetonide/
55. corticosteroid/
56. (prednison* or methylprednis* or dehydrocortisone or dexamethason* or beclomethasone or flunisolide).tw.
57. exp pimecrolimus/
58. (asm981 or asm 981 or elidel).tw.
59. exp tacrolimus/
60. (fk506 or fk 506).tw.
61. ((systemic or oral* or sublingual* or "per os" or inhal* or nasal* or parenteral*) adj3 (steroid* or glucosteroid* or corticosteroid* or glucocorticosteroid*)).tw.
62. ((biological*1 or biologic*1) adj (treatment* or therap* or medicine* or drug* or agent* or product*)).tw.
63. (biologic* response modifier* or BRM*).tw.
64. targeted therap*.tw.
65. dupilumab/
66. dupilumab.tw.
67. PUVA/
68. Puva.tw.
69. Psoralen ultraviolet A Therapy.tw.
70. alitretinoin/
71. alitretinoin.tw.
72. apremilast/
73. apremilast.tw.
74. tocilizumab/
75. tocilizumab.tw.
76. (systemic adj immunosuppressive treatment$).tw.
77. immuno‐modulatory treatment$.tw.
78. anti inflammatory treatment$.tw.
79. immunosuppressive agent/
80. antiinflammatory agent/
81. systemic therapy/
82. or/7‐81
83. crossover procedure.sh.
84. double‐blind procedure.sh.
85. single‐blind procedure.sh.
86. (crossover$ or cross over$).tw.
87. placebo$.tw.
88. (doubl$ adj blind$).tw.
89. allocat$.tw.
90. trial.ti.
91. randomized controlled trial.sh.
92. random$.tw.
93. or/83‐92
94. exp animal/ or exp invertebrate/ or animal experiment/ or animal model/ or animal tissue/ or animal cell/ or nonhuman/
95. human/ or normal human/
96. 94 and 95
97. 94 not 96
98. 93 not 97
99. 6 and 82 and 98
100. remove duplicates from 99

[Lines 83‐92: based on terms suggested for identifying RCTs in Embase (Section 3.6.2) in Lefebvre C, Glanville J, Briscoe S, Littlewood A, Marshall C, Metzendorf M‐I, et al. Technical Supplement to Chapter 4. Searching for and selecting studies. In: Higgins JPT, Thomas J, Chandler J, Cumpston MS, Li T, Page MJ, Welch VA (editors). Cochrane Handbook for Systematic Reviews of Interventions Version 6. Cochrane, 2019. Available from www.training.cochrane.org/handbook.]

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Allen 1991.

Study characteristics
Methods	Cross‐over double‐blinded randomised controlled trial Dates of study: not stated Location: United Kingdom, multi‐centre
Participants	Randomised: 33 participants (age range 17 to 56 years; 11 male) Inclusion criteria: not stated Exclusion criteria: not stated Dropouts and withdrawals: not stated
Interventions	Intervention: CsA 5 mg/kg/d for 2 months Comparator: placebo
Outcomes	Assessment at 2, 4, 6, and 8 weeks Outcomes: SASSAD, sleep score, itch score, ADRs
Funding source	Sandoz Pharmaceutical
Declarations of interest	Not described
Notes	Reports by all period data
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Insufficient information
Allocation concealment (selection bias)	Unclear risk	Insufficient information
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Identical placebo
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Identical placebo
Incomplete outcome data (attrition bias) All outcomes	Low risk	No withdrawn participants
Selective reporting (reporting bias)	Low risk	No potential sources of reporting bias
Other bias	Low risk	Cross‐over design

Bangert 2016.

Study characteristics
Methods	Parallel double‐blinded randomised controlled trials Dates of study: not stated Location: not stated
Participants	Randomised: 20 participants Inclusion criteria: adult patients with moderate to severe AD Exclusion criteria: not stated Dropouts and withdrawals: not stated
Interventions	Intervention: subcutaneous ligelizumab 280 mg, every 2 weeks, for 12 weeks Comparator: placebo
Outcomes	Assessment every 2 weeks for 12 weeks Primary outcomes: IgE and FcERI levels on leukocytes Secondary outcomes: clinical disease scores (EASI50, pruritus), drug safety, IgE‐related biomarkers, pruritus, sleep disturbance
Funding source	Not stated
Declarations of interest	Not stated
Notes	Only abstract available
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Insufficient information
Allocation concealment (selection bias)	Unclear risk	Insufficient information
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Insufficient information
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Insufficient information
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Insufficient information
Selective reporting (reporting bias)	Unclear risk	Insufficient information
Other bias	Unclear risk	Only abstract available

Beck 2014.

Study characteristics
Methods	Parallel double‐blind placebo‐controlled RCT Dates of study: not stated Locations: United States and multi‐national Europe Europe
Participants	• Randomised: 67 participants (mean age 42.6 (dupilumab), 37.4 (placebo)); 39 male Inclusion criteria: 18 years old and older, moderate to severe atopic dermatitis, SCORAD score > 20, EASI score ≥ 12 Exclusion criteria: not stated Dropouts and withdrawals: not stated • Randomised: 109 participants (mean age 33.7 (dupilumab), 39.4 (placebo)); 58 male Inclusion criteria: 18 years old and older, moderate to severe atopic dermatitis, SCORAD score > 20, EASI score ≥ 12 Exclusion criteria: not stated Dropouts and withdrawals: not stated • Randomised: 31 participants (mean age 37.8 (placebo and topical glucocorticoids), 36.0 (dupilumab and topical glucocorticoids)); 13 male Exclusion criteria: not stated Dropouts and withdrawals: not stated
Interventions	• Intervention: dupilumab 75 to 300 mg, subcutaneously once a week, 4 weeks Comparator: placebo • Intervention: dupilumab 300 mg, subcutaneously once a week, ≥ 12 weeks Comparator: placebo • Intervention: dupilumab 300 mg, subcutaneously once a week, with topical corticosteroid, 4 weeks Comparator: placebo, with topical corticosteroid
Outcomes	Assessment at 4 weeks Primary outcomes: EASI50, EASI score Secondary outcomes: IGA score 0 to 1, % BSA reduction, 5‐D Pruritus Scale score, score on Pruritus Numerical Rating Scale Assessment every 4 weeks until week 16 Primary outcomes: EASI50, EASI score Secondary outcomes: IGA score 0 to 1, % BSA reduction, 5‐D Pruritus Scale score, score on Pruritus Numerical Rating Scale, SCORAD score Assessment at 4 weeks Primary outcome: severity of AE Secondary outcomes: EASI50 up (%) reduction, IGA score 0 to 1, % BSA reduction, EASI score (0 to 72), 5‐D Pruritus Scale score, score on Pruritus Numerical Rating Scale, SCORAD score
Funding source	Supported by Regeneron Pharmaceuticals and Sanofi
Declarations of interest	Not stated
Notes	Not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Central randomisation
Allocation concealment (selection bias)	Low risk	Central randomisation
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Study patients, principal investigators, and study site personnel (with the exception of the designated unblinded study pharmacist, or designee) will remain blinded to all randomisation assignments throughout the duration of the study
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Study patients, principal investigators, and study site personnel (with the exception of the designated unblinded study pharmacist, or designee) will remain blinded to all randomisation assignments throughout the duration of the study
Incomplete outcome data (attrition bias) All outcomes	High risk	Withdrawal from the study occurred approximately twice as frequently in the placebo groups as in the dupilumab groups and was due primarily to lack of efficacy
Selective reporting (reporting bias)	Low risk	No potential sources of reporting bias
Other bias	Low risk	Baseline balanced

Bemanian 2005.

Study characteristics
Methods	Parallel open‐label randomised controlled trial Dates of study: September 2003 to January 2005 Location: Children's Hospital Medical Center, Tehran University of Medical Sciences, Tehran, Iran
Participants	Randomised: 14 participants (median age 11.91 (4.29) 4 mg/kg ciclosporin, 6.44 (1.59) 2 g/kg intravenous immunoglobulin (IVIG)); 7 male Inclusion criteria: all patients with severe atopic dermatitis (SCORAD > 70) who failed to respond to the first and second lines of therapy Exclusion criteria: not stated Dropouts and withdrawals: 2 participants, due to poor co‐operation
Interventions	Intervention: oral ciclosporin 4 mg/kg (Neoral‐Novartis) daily for 3 months Comparator: single dose of 2 g/kg intravenous immunoglobulin (Sandglobulin‐Novartis) as slow IV infusion over 4 to 8 hours (1 drop/kg/min)
Outcomes	Assessment at 15, 30, 60, and 90 days Outcomes: median eosinophil count, reduction in SCORAD, adverse drug reaction
Funding source	Not stated
Declarations of interest	Not stated
Notes	All patients used daily emollient post bathing for skin hydration and oral hydroxyzine (0.5 mg/kg at bedtime) and topical corticosteroid in short courses; when there was flaring up of dermatitis, systemic antibiotic (Cephalexin) or topical antibiotics were used
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No method of randomisation provided
Allocation concealment (selection bias)	Unclear risk	No method of randomisation provided
Blinding of participants and personnel (performance bias) All outcomes	High risk	No blinding with SCORAD outcomes
Blinding of outcome assessment (detection bias) All outcomes	High risk	No blinding with SCORAD outcomes
Incomplete outcome data (attrition bias) All outcomes	Low risk	< 20% dropout
Selective reporting (reporting bias)	Unclear risk	No potential sources of reporting bias; the outcome measure of the study is confusing and might be at risk of reporting bias
Other bias	Low risk	Baseline balanced

Berth‐Jones 1991.

Study characteristics
Methods	Cross‐over double‐blinded randomised controlled trial. No details given about the washout period. Dates of study: not stated Location: not stated
Participants	Randomised: 33 participants Inclusion criteria: not stated Exclusion criteria: not stated Dropouts and withdrawals: not stated
Interventions	Intervention: cyclosporin 5 mg/kg/day for 8 weeks Comparator: matching placebo
Outcomes	Assessment at baseline, every 2 weeks during treatment periods, and at 4 weeks after completion of the second period Outcomes: % BSA, SASSAD score, safety
Funding source	Not stated
Declarations of interest	Not stated
Notes	Only abstract available
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No method of randomisation provided
Allocation concealment (selection bias)	Unclear risk	No method of randomisation provided
Blinding of participants and personnel (performance bias) All outcomes	Low risk	No blinding with some subjective measures for physician
Blinding of outcome assessment (detection bias) All outcomes	High risk	No blinding with some subjective measures for physician
Incomplete outcome data (attrition bias) All outcomes	Low risk	No withdrawn participants
Selective reporting (reporting bias)	Low risk	No potential sources of reporting bias
Other bias	Unclear risk	No baseline provided

Berth‐Jones 2002.

Study characteristics
Methods	Cross‐over double‐blinded randomised controlled trial Dates of study: not stated Location: United Kingdom
Participants	Randomised: 37 participants, mean age 38 years (17 to 73); 25 male Inclusion criteria: 16 years of age and older Exclusion criteria: pregnant and lactating females, impaired hepatic or renal function, myelosuppression, sensitivity to azathioprine, concurrent treatment with allopurinol or systemic drugs, history of serious malignancy, any active chronic infection Dropouts and withdrawals: 16 participants were withdrawn: 12 during azathioprine treatment and 4 during placebo treatment, due to lack of response, adverse events, non‐compliance
Interventions	Intervention: azathioprine 2.5 mg/kg/d, once daily in the morning, for 3 months Comparator: placebo
Outcomes	Assessment at 2, 4, 8, and 12 weeks of each period Primary outcome: SASSAD score Secondary outcomes: adverse events, severity of Pruritus Scales score
Funding source	This study was funded by the Leicester Dermatology Research Foundation and the Nuneaton Dermatology Research Fund
Declarations of interest	Not stated
Notes	Reports by first period Participants were allowed to continue all topical medication in use at baseline, including emollients and potent topical corticosteroids
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Treatment was randomised in balanced blocks of 4 using computer‐generated random numbers, and codes were kept in sealed envelopes until data collection was completed
Allocation concealment (selection bias)	Unclear risk	The trial was of double‐blind randomised placebo‐controlled cross‐over design. Each participant received active and placebo treatment in randomised sequence, each for a total duration of 12 weeks, with no interval between treatment periods
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Matching placebo
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Matching placebo
Incomplete outcome data (attrition bias) All outcomes	High risk	Per protocol with high degree of withdrawal and differential withdrawal
Selective reporting (reporting bias)	Low risk	No evidence of selective outcome reporting
Other bias	Unclear risk	No baseline provided

Bieber 2014.

Study characteristics
Methods	Parallel double‐blinded randomised controlled trial Dates of study: not stated Location: multi‐centre, Germany
Participants	Randomised: 109 participants Inclusion criteria: not stated Exclusion criteria: not stated Dropouts and withdrawals: not stated
Interventions	Intervention: dupilumab 300 mg, subcutaneously once a week, 12 weeks Comparator: placebo
Outcomes	Assessment at 12 weeks Primary outcome: EASI score Secondary outcomes: body surface area, SCORAD, Pruritus Numerical Rating Scale, 5‐D Pruritus Scale score
Funding source	Asana BioSciences
Declarations of interest	Not stated
Notes	Only abstract available
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Insufficient information
Allocation concealment (selection bias)	Unclear risk	Insufficient information
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Insufficient information
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Insufficient information
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Insufficient information
Selective reporting (reporting bias)	Unclear risk	Insufficient information
Other bias	Unclear risk	Insufficient information

Bissonnette 2019.

Study characteristics
Methods	Parallel double‐blinded randomised controlled trial Dates of study: April 2017 to November 2017 Location: Canada and USA (10 centres)
Participants	Randomised: 36 participants, mean age 38.2 (14.36) (ASN002 20 mg), 42.4 (13.88) (ASN002 40 mg), 33.1 (10.42) (ASN002 80 mg), 29.9 (9.33) (placebo); 18 male Inclusion criteria: 18 to 75 years of age, moderate to severe AD, EASI score ≥ 16, IGA score 3 to 4, BSA ≥ 10%, BMI < 35 at day 1 Exclusion criteria: Clinically infected atopic dermatitis Presence of any of the following laboratory abnormalities at the screening visit: haemoglobin < 11 g/dL, white blood cell (WBC) count < 3.0 × 103 /μL, platelet count < 125 × 103/μL, neutrophils < 1.75 × 103 /μL, aspartate aminotransferase (AST)/alanine aminotransferase (ALT) > 1.5 × upper limit of normal (ULN), total bilirubin > ULN, creatinine > ULN Serious uncontrolled condition, including hypertension, history of tuberculosis, hepatitis B or C infection, immune deficiency, heart disease, heart conduction disorder, diverticulitis, diabetes, reflux disease requiring protocol pump inhibitor therapy, malabsorption syndrome, or cancer Any condition requiring the use of anticoagulants History of hypertrophic scarring or keloid formation in scars or at suture sites Any medical or psychiatric condition that, in the opinion of the investigator or the sponsor's medical monitor, would place the patient at risk, interfere with participation in the study, or interfere with interpretation of study results Pregnant or breastfeeding woman Known hypersensitivity to ASN002 or its excipients Prior treatment with SYK or JAK inhibitors from which the individual received no clinical benefit, or the individual relapsed whilst on therapy Has used oral or intravenous treatments (other than biological agents) that could affect atopic dermatitis less than 4 weeks before day 1 Has received any marketed or investigational biological agent within 12 weeks or 5 half‐lives (whichever is longer) before day 1 Currently receiving a non‐biological investigational product or device, or has received 1 within 4 weeks of day 1 Excessive sun exposure, planning a trip to a sunny climate, has used tanning booths within 4 weeks before baseline (day 1), or not willing to minimise natural and artificial sunlight exposure during the study Has received or plans to receive a live attenuated vaccine within 4 weeks before day 1 throughout the follow‐up period Planned major surgical procedure during the length of the patient's participation in this study Dropouts and withdrawals: not stated
Interventions	Interventions: ASN002 (20 mg), ASN002 (40 mg), ASN002 (80 mg), oral once daily, 28 days Comparator: placebo
Outcomes	Assessment at 15, 29 days Primary outcome: TEAEs Secondary outcomes: EASI50, EASI75, IGA 0/1, Pruritus NRS, BSA
Funding source	Asana BioSciences
Declarations of interest	R.B. is an investigator, consultant, advisory board member, and speaker for and/or receives honoraria from Aquinox Pharma, AntibioTx, Asana BioSciences, Astellas, Brickell Biotech, Dermavant, Dermira, Dignity Sciences, Galderma, Glenmark, GSK Stiefel, Hoffman LaRoche Posay, Kiniksa, Leo Pharma, Neokera, Pfizer, Regeneron, Sienna, and Vitae. R.B. is also a shareholder at Innovaderm Research C.M. has received grants and research support or has received honoraria from Aquinox Pharma, Asana BioScience, Astellas, Brickell Biotech, Dermavant, Lilly Pharma, Galderma, Glenmark, GSK Stiefel, Hoffman LaRoche Posay, Leo Pharma, Pfizer, Regeneron‐Sanofi, Vitae, and Valeant S.F. has received grants and research support or has received honoraria from AbbVie, Janssen, Eli Lilly, Novartis, Pfizer, Xbiotech, Galderma, Asana BioSciences, and Regeneron N.B. has received grants and research support from Asana BioSciences M.L. has received grants and research support from AbbVie, Boehringer Ingelheim, Novartis, Lilly, Janssen, Leo Pharmaceuticals, Dermira, UCB, Aclaris, Valeant, and Asana BioSciences J.F. has received grants and research support from Asana BioSciences S.T. has received grants and research support from Asana BioSciences D.P. is an investigator, consultant, advisory board member, and speaker for and/or receives honoraria from Abbott Laboratories, Amgen, Asana BioSciences, Atacama Therapeutics, Bickel Biotechnology, Biofrontera AG, Celgene Corporation, Dermira, Dermavant Sciences, DUSA Pharmaceuticals, Inc., Eli Lilly, GSK Stiefel, Leo Pharma, Merck, Novartis Pharmaceuticals Corp., Novo Nordisk A/S, Ortho Dermatologics, Peplin Inc., Pfizer Inc., Photocure ASA, Promius Pharmaceuticals, Regeneron, Sanofi, TDM Surgitech, Inc., TheraVida, and Valeant H.S. has received grants and research support or has received honoraria from Dermira, Leo Pharma, Ralexar, Incyte, Dermavant, Lilly, Regeneron, Genentech, and Asana BioSciences S.D. has received grants and research support from AbbVie, Allergan, Asana BioSciences, Dermira, Galderma, GSK Stiefel, Leo Pharma, Revance, and Valeant M.Z. has received grants and research support from Pfizer, Novartis, Sienna, Janssen, Endo International, Lilly, Dermira, Moberg Pharma, Soligenix, Allergan, Asana BioSciences, Athenex, Foamix, Incyte, Sun Pharma, and Verrice D.J.Z., H.U., L.D., and N.R. are employees of Asana BioSciences E.G.‐Y. is an employee of Mount Sinai and has received research funds (grants paid to the institution) from AbbVie, Celgene, Eli Lilly, Novan, Relaxar, Janssen, Novartis, Pfizer, Regeneron, Glenmark, Dermavent, DBV, DS Biopharma, Concert, Galderma, Asana BioSciences, Innovaderm, and Dermira E.G.‐Y is also a consultant for Amgen, Sanofi Aventis, Escallier, Allergan, Regeneron, Celgene, Dermira, Galderma, Glenmark, Novartis, Pfizer, Leo Pharma, AbbVie, Eli Lilly, Kyowa, Mitsubishi Tanabe, and Asana BioSciences
Notes	Not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Via central interactive voice/web response
Allocation concealment (selection bias)	Low risk	Randomisation (Materials and methods, page 2, paragraph 2 "...12 patients were randomized in a 3:1 ratio to receive ASN002 or placebo according to a central randomization scheme provided by an interactive web response system")
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Masking: double‐blind (patient, caregiver, investigator, outcomes assessor)
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Masking: double‐blind (patient, caregiver, investigator, outcomes assessor)
Incomplete outcome data (attrition bias) All outcomes	High risk	A number of losses to follow‐up are known (Figure 1 = "The proportion of participants who completed the study in each group ranged from 67% to 89%, 78% to 89% of the participants were included in the efficacy outcome analysis, and 100% were included in safety analysis"). Overall we considered the study to be at high risk in incomplete outcome data due to the differential proportion of participants who completed the trial and were included in the efficacy analysis
Selective reporting (reporting bias)	Low risk	No potential sources of selective outcome reporting
Other bias	Low risk	No other potential sources of bias

Blauvelt 2017.

Study characteristics
Methods	Parallel double‐blinded randomised controlled trial Dates of study: 3 October 2014 to 31 July 2015 Location: multi‐centre (161 sites)
Participants	Randomised (3:1:3): 740 participants, median age 34.0 (25.0 to 45.0) (193 male) (placebo once weekly), 40.5 (28.0 to 49.0) (62 male) (dupilumab 300 mg every 2 weeks), 34.0 (26.0 to 45.0); 191 male (dupilumab 300 mg once weekly) Inclusion criteria: 18 years of age or older, atopic dermatitis (American Academy of Dermatology Consensus Criteria 3) present for 3 years or longer before screening, documented history within 6 months before screening of inadequate response to medium‐potency to high‐potency topical corticosteroids (with or without topical calcineurin inhibitors as appropriate), documented systemic treatment within the past 6 months or both, Investigators’ Global Assessment (IGA) score ≥ 3 (moderate to severe on a scale of 0 to 4,.Eczema Area and Severity Index (EASI) score ≥ 16 at screening and baseline Exclusion criteria: Participation in a prior dupilumab clinical trial Important side effects of topical medication (e.g. intolerance to treatment, hypersensitivity reactions, significant skin atrophy, systemic effects), as assessed by the investigator or treating physician Having used any of the following treatments within 4 weeks before the baseline visit, or any condition that, in the opinion of the investigator, was likely to require such treatment(s) during the first 2 weeks of study treatment: immunosuppressive/immunomodulating drugs (e.g. systemic steroids, ciclosporin, mycophenolate‐mofetil, Janus kinase inhibitors, interferon‐gamma (IFN‐γ), azathioprine, methotrexate) Phototherapy for AD Treatment with a live (attenuated) vaccine within 12 weeks before the baseline visit History of human immunodeficiency virus (HIV) infection or positive HIV serology at screening Positive hepatitis B surface antigen (HBsAg), hepatitis B core antibody (HBcAb), or hepatitis C antibody at the screening visit Active or acute infection requiring systemic treatment within 2 weeks before baseline visit Known or suspected history of immunosuppression Pregnant or breastfeeding women, or women planning to become pregnant or to breastfeed during participation in this study Dropouts and withdrawals: not stated
Interventions	Interventions: Dupilumab 300 mg every 2 weeks, 52 weeks Dupilumab 300 mg once weekly, 52 weeks Comparator: placebo once weekly, 52 weeks
Outcomes	Assessment at 1, 2, 4, 6, 8, 12, 16, 20, 24, 28, 32, 36, 40, 44, 48, 52 weeks Primary outcomes: proportion of patients with IGA 0/1, ≥ 2‐point reduction from baseline at week 16, EASI75 at week 16 Secondary outcomes: proportion of patients with IGA 0/1; ≥ 2‐point reduction from baseline at week 52; EASI75 at week 52; peak Pruritus Numerical Rating Scale (NRS) improvement (reduction) ≥ 4 points (baseline to weeks 2, 4, 16, 24, and 52), ≥ 3 points (baseline to weeks 16 and 52); peak Pruritus NRS percentage change (baseline to weeks 16 and 52). For peak Pruritus NRS, patients reported the intensity of their worst itch via an interactive voice/web response system on a scale of 0 to 10 (no itch to worst itch imaginable) during the previous 24 hours daily (weeks 0 to 16) or weekly (weeks 17 to 52). For weeks with daily peak Pruritus NRS, reported scores (weeks 0 to 16)
Funding source	This study was funded by Sanofi and Regeneron Pharmaceuticals, Inc.
Declarations of interest	Funders participated in the conception and design of the study, analysis and interpretation of data, and drafting and critical revision of the report, and gave approval to submit
Notes	‐
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Central randomisation
Allocation concealment (selection bias)	Low risk	Central randomisation
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Patients given dupilumab Q2W received matching placebo in the weeks when dupilumab was not given Blinded study drug kits with a medication numbering system were used. Placebo was provided in identical syringes. Study remained blinded to all individuals (including patients, investigators, and study personnel) until the pre‐specified time of unblinding
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Patients given dupilumab Q2W received matching placebo in the weeks when dupilumab was not given Blinded study drug kits with a medication numbering system were used. Placebo was provided in identical syringes. Study remained blinded to all individuals (including patients, investigators, and study personnel) until the pre‐specified time of unblinding
Incomplete outcome data (attrition bias) All outcomes	Low risk	Missing data were imputed appropriately
Selective reporting (reporting bias)	Low risk	No potential sources of reporting bias
Other bias	Low risk	Baseline balanced

Blauvelt 2018.

Study characteristics
Methods	Parallel double‐blinded randomised controlled trial Dates of study: August 2014 to September 2015 Location: USA
Participants	Randomised: 194 participants, mean age 39 (14); (49 male) (dupilumab); 40 (14); (46 male) (placebo) Inclusion criteria: patients with AD, 18 to 64 years of age, moderate to severe disease for ≥ 3 years, documented history (within 6 months of screening) of inadequate response to medium‐ to high‐potency topical corticosteroids (with/without topical calcineurin inhibitors) despite treatment for ≥ 28 days, patients for whom topical AD therapies were inadvisable, Eczema Area and Severity Index (EASI) score ≥ 16 (range 0 to 72), Investigator’s Global Assessment (IGA) score ≥ 3 (range 0 to 4), ≥ 10% BSA affected by AD at screening and at baseline. Use of emollients and TCS with/without TCI was permitted at entry if patients had been on a stable regimen for ≥ 14 days Exclusion criteria: not stated Dropouts and withdrawals: not stated
Interventions	Intervention: dupilumab (300 mg) was administered subcutaneously weekly, 16 weeks. A 600‐mg loading dose of matching placebo was administered on day 1 Comparator: placebo. A 600‐mg loading dose of matching placebo was administered on day 1
Outcomes	Assessment at 12, 16, and 32 weeks Primary outcomes: proportion of patients with a positive response to the tetanus component of Tdap at week 16, defined as ≥ 4‐fold increase from baseline anti‐tetanus IgG concentration in patients with pre‐vaccination tetanus antibody concentrations ≥ 0.1 or ≥ 0.2 IU/mL and in patients with pre‐vaccination concentrations < 0.1 IU/mL Secondary outcomes: Positive response to tetanus toxoid (≥ 2‐fold increase from baseline anti‐tetanus IgG concentration in patients with pre‐vaccination tetanus antibody concentrations ≥ 0.1 or ≥ 0.2 IU/mL in patients with pre‐vaccination concentrations < 0.1 IU/mL) Positive response to MPSV4 (meningococcal serogroup C serum bactericidal assay (SBA) titre ≥ 8) IGA score of 0 or 1 > 50% reduction in EASI (EASI50) > 75% reduction in EASI (EASI75) Changes from baseline at week 16 in: peak Pruritus Numerical Rating Scale (NRS); BSA; erythema; infiltration/papulation; excoriations; lichenification of Global Individual Sign Score (GISS); or Patient‐Oriented Eczema Measure (POEM) Total IgE levels and proportions of patients who were tetanus‐ and pertussis‐antigen‐specific IgE seropositive at weeks 12, 16, and 32 were analysed post hoc
Funding source	Regeneron Pharmaceuticals, Inc., and Sanofi
Declarations of interest	Not stated
Notes	Not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Patients and personnel involved were blinded to all randomisation assignments
Allocation concealment (selection bias)	Low risk	Patients and personnel involved were blinded to all randomisation assignments
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Patients and personnel involved were blinded to all randomisation assignments
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Patients and personnel involved were blinded to all randomisation assignments
Incomplete outcome data (attrition bias) All outcomes	Low risk	91.8% completed
Selective reporting (reporting bias)	Low risk	No potential sources of reporting bias
Other bias	Low risk	Baseline balanced

Cordero 1999.

Study characteristics
Methods	A prospective, experimental, longitudinal, comparative, open study Dates of study: September 1997 to June 1998 Location: Spain
Participants	Randomised: 18 participants, 7 to 43 years, with an average of 16 years Inclusion criteria: participants with a clinical diagnosis of severe refractory AD, according to the criteria established by Hanifin and Rajka, who entered the clinical immunology and allergy service of the Regional Hospital Lic. Adolfo López Mateos. Either sex, aged between 3 and 40 years, who did not suffer from previous kidney disease or hypertension, who had not been treated with steroids, systemic or cytotoxic agents, or phototherapy (PUVA UVB) during at least one month prior to the study, or that another disease coexisted (dermatological or systemic) that limited the use of the proposed therapeutic regimens. Therapy with antihistamines, topical emollients and antibiotics was allowed. All participants or guardians who agreed to participate in the study signed an informed consent form. Exclusion criteria: participants with mild or moderate AD, who had previous renal injury or hypertension, pregnant or breastfeeding women, participants who were under treatment with systemic steroids or were known to be hypersensitive to cyclosporin A. Participants who abandoned the treatment during the course of the study were eliminated from the study, as were those that did not want to continue with it, or developed diseases that limited the use of the chosen medication. Dropouts and withdrawals: 0
Interventions	Intervention: Ciclosporin A (Sandimum Neoral capsules with microemulsified solution of 25 and 100 mg of Novartis laboratory) at a dose of 4 mg / kg / day for 3 months (n = 6) Comparator: Transfer factor (developed and donated by the department of immunology of the National Polytechnic Institute) administered orally, as follows: one every third day in the first week; two per week to complete the month and finally one per month to complete six months in total (n = 12). The topical application of 1% hydrocortisone, diluted in equal parts with inert cream, was used as a rescue medication during the study.
Outcomes	Assessment at 12 and 24 weeks Global clinical assessment by physician, global clinical assessment by participant, and adverse events
Funding source	Not stated
Declarations of interest	Not stated
Notes	Not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No details provided
Allocation concealment (selection bias)	Unclear risk	No details provided
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	No details provided
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	No details provided
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	No details provided
Selective reporting (reporting bias)	Unclear risk	No details provided
Other bias	Unclear risk	No details provided

Czech 2000.

Study characteristics
Methods	Parallel double‐blinded randomised controlled trial Dates of study: not stated Locations: multi‐centre, Germany
Participants	Randomised: 106 participants Inclusion criteria: performed in men and women 18 years of age or older who had a clinical diagnosis of severe atopic dermatitis not adequately controlled by conventional therapy Exclusion criteria: creatinine levels > 10% above upper normal limit (UNL) of the central laboratory, uric acid > 30% above UNL; bilirubin > 50% over UNL; alkaline phosphatase and transaminase levels > 100% above UNL, hyperkalaemia and uncontrolled hypertension Dropouts and withdrawals: in the high‐dose group, 3 patients were withdrawn: 1 because of abdominal pain and nausea; another because of hair loss, gingival hyperplasia, and headache; and 1 because of paraesthesia and headache
Interventions	Intervention: cyclosporin A 150 mg/d for 8 weeks Comparator: cyclosporin A 300 mg/d for 8 weeks After 2 weeks, dose could be reduced by 50% if clinical symptom score was reduced by ≥ 50%. After 8 weeks, responders entered a 4‐week follow‐up phase and were randomised to stop treatment or to continue on their last effective dose every second day.
Outcomes	Asssessment at weeks 2, 8, and 12 Outcomes: TBSA score, BSA score, itching, DLQI, sleep loss
Funding source	Not stated
Declarations of interest	Not stated
Notes	Not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer generated
Allocation concealment (selection bias)	Unclear risk	No allocation concealment information provided
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Details of double dummy provided
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Details of double dummy provided
Incomplete outcome data (attrition bias) All outcomes	Low risk	LOCF appropriately
Selective reporting (reporting bias)	Low risk	No potential sources of reporting bias
Other bias	Low risk	Baseline balanced

de Bruin‐Weller 2018.

Study characteristics
Methods	Parallel double‐blinded randomised controlled trial Dates of study: January and December 2016 Location: multi‐centre, European countries
Participants	Randomised: 325 participants, mean age 37.5 (29 to 49) (68 male) (placebo), 38.0 (25 to 47) (65 male) (dupilumab every 2 weeks), 38.0 (29 to 48) (66 male) (dupilumab weekly) Inclusion criteria (main): individuals ≥ 18 years of age with chronic atopic dermatitis according to American Academy of Dermatology consensus criteria; treatment with a potent TCS is indicated; inadequate response to TCS (as defined by investigator) within 6 months before screening; history of (i) prior CsA exposure and inadequate response to CsA, requirement for CsA at doses or durations beyond those specified in prescribing information, or intolerance and/or unacceptable toxicity; or (ii) CsA naive and not eligible for CsA because of medical contraindications (e.g. uncontrolled hypertension on medication); use of prohibited concomitant medications; increased susceptibility to CsA‐induced renal damage (elevated creatinine) and/or liver damage (elevated liver function tests); increased risk of serious infection or hypersensitivity to CsA active substance or excipients; Eczema Area Severity Index (EASI) score ≥ 20 at screening and at baseline; Investigator’s Global Assessment (IGA) score ≥ 3 (scale 0 (clear) to 4 (severe)) at screening and at baseline; ≥ 10% BSA of atopic dermatitis involvement at screening and baseline Exclusion criteria: not stated Dropouts and withdrawals: 3 (3%) in placebo + TCS group withdrew from study treatment because of lack of efficacy, as did 2 (2%) each in the placebo + TCS and dupilumab weekly (qw) + TCS groups because of AEs
Interventions	Intervention: subcutaneous dupilumab 300 mg qw or every 2 weeks (q2w) Comparator: placebo
Outcomes	Assessment every 2 weeks, at 16 weeks of treatment, at 12 weeks' follow‐up Primary outcome: proportion of patients with ≥ 75% improvement from baseline in EASI score (EASI75) at week 16 Secondary outcomes (all at week 16, unless otherwise indicated): percentage change from baseline in EASI, SCORing Atopic Dermatitis (SCORAD), weekly average of peak daily pruritus numerical rating scale (NRS) score (weeks 2 and 16), and Global Individual Sign Score (GISS); change from baseline in percentage BSA affected by atopic dermatitis; Dermatology Life Quality Index (DLQI); Patient‐Oriented Eczema Measure (POEM); Hospital Anxiety and Depression Scale (HADS); mean weekly dose of TCS during treatment period; proportions of patients with ≥ 50% or ≥ 90% improvement from baseline in EASI score (EASI50 or EASI90); EASI75 (among patients with prior CsA exposure); ≥ 4‐point reduction in weekly average of peak daily pruritus NRS score (among patients with baseline pruritus NRS score ≥ 4); ≥ 50% improvement from baseline in SCORAD (SCORAD‐50) and both IGA 0 or 1 (clear or almost clear) and 2‐point reduction in IGA from baseline
Funding source	Research was sponsored by Sanofi and Regeneron Pharmaceuticals, Inc.
Declarations of interest	M.d.B.‐W. is principal investigator, advisory board member, and consultant for Regeneron Pharmaceuticals, Inc., and Sanofi Genzyme; and principal investigator and advisory board member for AbbVie D.T. has received honoraria for participation on advisory boards, as a speaker, and for consultancy from AbbVie, Almiral, Amgen, Biogen‐Idec, Bristol‐Myers Squibb, Celgene, Dignity, Dr. Reddy, Galapagos, Galderma, Janssen, Leo, Maruho, Mitsubishi, Lilly, Novartis, Pfizer, Sandoz Hexal, Regeneron Pharmaceuticals, Inc., Sanofi, UCB, and Xenoport; and received research grants from Celgene and Novartis C.H.S. is principal investigator for Regeneron Pharmaceuticals, Inc., and Sanofi Genzyme. K.R. is an advisor and/or paid speaker for and/or has participated in clinical trials for AbbVie, Amgen, Biogen, Boehringer Ingelheim, Celgene, Covagen, Forward Pharma, GlaxoSmithKline, Janssen‐Cilag, Leo, Lilly, Medac, Merck Sharp & Dohme Corp., Novartis, Ocean Pharma, Pfizer, Regeneron Pharmaceuticals, Inc., Sanofi, Takeda, UCB Pharma, and Xenoport M.C. is an investigator for/received honoraria from Astellas, Johnson & Johnson, Leo Pharmaceuticals, Novartis, Regeneron Pharmaceuticals, Inc., Sanofi, and Stiefel‐GSK; served on advisory boards for/received honoraria from Amgen, Astellas, Bayer, Johnson & Johnson, Merck Sharp & Dohme, Leo Pharmaceuticals, Novartis, Pfizer, Regeneron Pharmaceuticals, Inc., Sanofi, Stiefel‐GSK, and Unilever; served as consultant for/received honoraria from Amgen, Astellas, Johnson & Johnson, Leo Pharmaceuticals, Novartis, Pfizer, Regeneron Pharmaceuticals, Inc., Sanofi, Stiefel‐GSK, and Unilever; gave lectures for/received honoraria from Astellas, Johnson & Johnson, Leo Pharmaceuticals, Novartis, Regeneron Pharmaceuticals, Inc., Sanofi, and Stiefel‐GSK; and received research grants/honoraria from Bayer and Merck Sharp & Dohme B.A., Z.C., A.G., N.M.H.G., A.R., B.S., and Q.Z. are all employees and shareholders of Regeneron Pharmaceuticals, Inc. L.E., T.H., and G.P. are all employees of and may hold stock and/or stock options in Sanofi
Notes	Not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Central randomisation
Allocation concealment (selection bias)	Low risk	Central randomisation
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Details of double dummy provided
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Details of double dummy provided
Incomplete outcome data (attrition bias) All outcomes	Low risk	Balanced missing data
Selective reporting (reporting bias)	Low risk	No potential sources of reporting bias
Other bias	Low risk	Baseline balanced

Der‐Petrossian 2000.

Study characteristics
Methods	Parallel single‐blinded randomised controlled trial Dates of study: not stated Location: Australia
Participants	Randomised: 12 participants, mean age 27 (11.3) years Inclusion criteria: not stated Exclusion criteria: not stated Dropouts and withdrawals: 1 patient experienced an exacerbation of AD after 3 weeks of treatment and had started to take oral corticosteroids. In another patient, considerably fewer erythema reactions were recorded in response to bath‐PUVA as compared with narrowband UVB; thus the criterion of equi‐erythemogenic dosages was not fulfilled
Interventions	Intervention: 8‐methoxypsoralen bath‐PUVA performed 3 times weekly over a period of 6 weeks Comparator: narrowband UVB
Outcomes	Assessments at 2, 4, and 6 weeks Primary outcome: modified SCORAD score used to assess half‐side severity of AD Secondary outcome: adverse events
Funding source	Not stated
Declarations of interest	Not stated
Notes	Not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No randomisation details provided
Allocation concealment (selection bias)	Unclear risk	No randomisation details provided
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Open label with SCORAD
Blinding of outcome assessment (detection bias) All outcomes	High risk	Open label with SCORAD
Incomplete outcome data (attrition bias) All outcomes	Low risk	All completed
Selective reporting (reporting bias)	Low risk	No potential sources of reporting bias
Other bias	Low risk	Baseline balanced

El‐Khalawany 2013.

Study characteristics
Methods	Parallel open‐label randomised controlled trial Dates of study: not stated Location: multi‐centre, Germany
Participants	Randomised: 40 participants, mean age 11.16 (1.52) (12 male) in methotrexate group; 10.30 (2.82) (14 male) in ciclosporin group Inclusion criteria: 8 to 14 years of age; severe AD; failed to be treated with topical therapy; unfit, uncooperative, or poorly responsive to phototherapy Exclusion criteria: children with chronic or recurrent infection or with history of severe or uncontrolled systemic disease. In addition, patients with history of organ transplantation or history of cancer, or those who had herpes zoster infection within 2 months before the study, were excluded. Moreover, patients who were known to have hypersensitivity to methotrexate or ciclosporin were excluded from the study Dropouts and withdrawals: not stated
Interventions	Intervention: methotrexate (7.5 mg/week) Comparator: ciclosporin (2.5 mg/kg/d)
Outcomes	Assessment at weeks 4, 8, and 12 Primary outcome: SCORAD score Secondary outcome: adverse events
Funding source	Not stated
Declarations of interest	Not stated
Notes	Not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer generated
Allocation concealment (selection bias)	Unclear risk	No allocation concealment information provided
Blinding of participants and personnel (performance bias) All outcomes	High risk	Open label with SCORAD
Blinding of outcome assessment (detection bias) All outcomes	High risk	Open label with SCORAD
Incomplete outcome data (attrition bias) All outcomes	Low risk	All completed
Selective reporting (reporting bias)	Low risk	No potential sources of reporting bias
Other bias	Low risk	Baseline balanced

Gerbens 2018.

Study characteristics
Methods	Cross‐over double‐blinded randomised controlled trial Dates of study: September 2009 to May 2010 Location: The Netherlands
Participants	Randomised: 35 participants, mean age 43.3 (15.8) (10 male) in methotrexate group; 39.5 (14.4) (11 male) in azathioprine group Inclusion criteria: moderate to severe AD, unresponsive, contraindicated or intolerant to cyclosporin A Excluion criteria: not stated Dropouts and withdrawals: not stated
Interventions	Intervention: methotrexate dosed at maximum 22.5 mg per week (dose range 10 to 22.5 mg per week), in combination with 5 mg folic acid once a week Comparator: azathioprine at dosage restricted to maximum 2.5 mg/kg daily (dose range 25 to 275 mg daily)
Outcomes	Assessment every 3 months Primary outcome: SCORAD Secondary outcomes: IGA, POEM, EASI, pruritus VAS score
Funding source	Not stated
Declarations of interest	Not stated
Notes	Not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer generated
Allocation concealment (selection bias)	Low risk	Allocation concealment mentioned
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Single‐blind (assessor) with SCORAD
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Single‐blind (assessor) with SCORAD
Incomplete outcome data (attrition bias) All outcomes	High risk	Large number of dropouts
Selective reporting (reporting bias)	Low risk	No potential sources of reporting bias
Other bias	Low risk	Baseline balanced

Goujon 2018.

Study characteristics
Methods	Parallel open‐label randomised controlled trial
Participants	Randomised: 97 participants, mean age 33 (31 males) in cyclosporin A group, 32 (28 males) in methotrexate group Inclusion criteria: chronic moderate to severe AD as per diagnostic criteria of the UK Working Party with score > 15 on SCORAD; inadequate response to topical corticosteroids or tacrolimus Exclusion criteria: use of systemic corticosteroids and immunosuppressive treatment within 4 weeks before inclusion visit Dropouts and withdrawals: first 8 weeks of treatment, 2 patients in the cyclosporin A arm were lost to follow‐up and 2 were excluded for protocol deviation. After 8 weeks of treatment, 27 patients in the methotrexate arm were dropouts, 15 were non‐responders, 6 had adverse events, 3 protocol deviations occurred, and 3 withdrew consent. 10 patients in the cyclosporin A arm were dropouts, 2 protocol deviations occurred, 1 had adverse events, 1 withdrew consent, and 6 were non‐responders
Interventions	Methotrexate vs cyclosporin A
Outcomes	Assessment every 4 weeks for 24 weeks Primary outcome: SCORAD 50 at 8 weeks of treatment EASI, SCORAD, DLQI, Pruritus VAS score, BSA
Funding source	Not described
Declarations of interest	Not described
Notes	Not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated random numbers using random fixed permuted blocks with stratification at the centre and disease severity
Allocation concealment (selection bias)	Low risk	Computer‐generated random numbers using random fixed permuted blocks with stratification at the centre and disease severity
Blinding of participants and personnel (performance bias) All outcomes	High risk	Single‐blind (assessor) with SCORAD (physician‐patient assessment tool)
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Single‐blind (assessor) with SCORAD
Incomplete outcome data (attrition bias) All outcomes	Low risk	< 20% dropout at primary outcome assessment
Selective reporting (reporting bias)	Low risk	No potential sources of reporting bias
Other bias	Low risk	Baseline balanced

Granlund 2001.

Study characteristics
Methods	Parallel open‐label randomised controlled trial
Participants	Randomised: 72 participants, mean age 33.3 (21 males) in cyclosporin A group, 33.2 (14 males) in UVAB group Inclusion criteria: between 18 and 70 years of age; AD Exclusion criteria: treated with systemic corticosteroids, cyclosporin A, or UVAB within the 2 weeks before entry into the study Dropouts and withdrawals: 24 patients discontinued treatment prematurely: 4 in the cyclosporin A group and 20 in the UVAB treatment group. Treatment failure was the main reason for withdrawal; all 6 cases occurred in the UVAB group. Adverse events and withdrawal in 4 patients: 1 on cyclosporin A and 3 on UVAB. Most other reasons were protocol violations, which occurred in 3 patients on cyclosporin A and in 11 on UVAB. Major protocol deviations not resulting in premature withdrawal occurred in 13 patients (8 in the cyclosporin A group and 4 in the UVAB group)
Interventions	Cyclosporin A vs UVAB
Outcomes	Assessment time: every month for 12 months Primary outcome: The number of days in remission based on SCORAD values All outcomes measured: SCORAD, use of emollients and topical corticosteroids, overall assessment of efficacy by participant and investigator, EDI, AEs
Funding source	Novartis Finland
Declarations of interest	Not described
Notes	Not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No randomisation details provided
Allocation concealment (selection bias)	Unclear risk	No randomisation details provided
Blinding of participants and personnel (performance bias) All outcomes	High risk	No blinding provided with SCORAD
Blinding of outcome assessment (detection bias) All outcomes	High risk	No blinding provided with SCORAD
Incomplete outcome data (attrition bias) All outcomes	High risk	24 of 72 patients prematurely discontinued
Selective reporting (reporting bias)	Low risk	No potential sources of reporting bias
Other bias	Low risk	Baseline balanced

Guttman‐Yassky 2018.

Study characteristics
Methods	Parallel double‐blinded randomised controlled trial
Participants	Randomised: 60 participants in a 2:1 ratio vs placebo Inclusion criteria: moderate to severe AD (scoring of AD/SCORAD > 25) treated for 12 weeks with bi‐weekly intravenous administrations of fezakinumab Exclusion criteria: not stated Dropouts and withdrawals: not stated
Interventions	Fezakinumab vs placebo
Outcomes	Assessment at week 12 and 20 SCORAD, IGA, BSA
Funding source	Pfizer, Inc. (New York, NY, USA)
Declarations of interest	Not described
Notes	Assessor not blinded
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No details of randomisation
Allocation concealment (selection bias)	Unclear risk	No details of randomisation
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Triple (participant, care provider, investigator)
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Few losses to follow‐up ("All randomised patients were included in the intention‐to‐treat analysis" (Results, Patients, paragraph 2))
Incomplete outcome data (attrition bias) All outcomes	Low risk	No potential sources of selective outcome reporting
Selective reporting (reporting bias)	Low risk	Baseline balanced
Other bias	Low risk	Baseline balanced

Guttman‐Yassky 2019a.

Study characteristics
Methods	Parallel double‐blinded randomised controlled trial Dates of study: February 2016 to March 2017 Locations: United States and Japan
Participants	Randomised: 124 participants Inclusion criteria: age ≥ 18 years; moderate to severe atopic dermatitis (EASI ≥ 12 and > 10% BSA involvement at screening and randomisation; diagnosis ≥ 2 years before first visit) Exclusion criteria: not stated Dropouts and withdrawals: patients discontinued placebo + TCS (41%) and baricitinib + TCS (27% on 2 mg and 24% on 4 mg), including discontinuation due to lack of efficacy (placebo 18%; baricitinib 2 mg 11%, baricitinib 4 mg 0%)
Interventions	Intervention: baricitinib 2 mg QD + TCS (n = 37) or baricitinib 4 mg QD + TCS (n = 38) Comparator: placebo QD + TCS (n = 49)
Outcomes	Assessment at week 16 Primary outcome: EASI50 at week 16 Other outcomes: EASI, SCORAD, IGA, DLQI, POEM, SAEs
Funding source	Regeneron Pharmaceuticals, Inc.
Declarations of interest	Not described
Notes	Not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Patients were randomised via interactive response technology (IRT) system in a 4:3:3 ratio to once‐daily placebo or 2 mg or 4 mg baricitinib tablets. Blocked randomisation with stratification was used with 2 strata and 150 randomisation numbers per stratum in blocks of 10. Parexel (Waltham, MA, USA) generated the randomisation schedule and maintained the IRT
Allocation concealment (selection bias)	Low risk	Patients were randomised via interactive response technology (IRT) system in a 4:3:3 ratio to once‐daily placebo or 2 mg or 4 mg baricitinib tablets. Blocked randomisation with stratification was used with 2 strata and 150 randomisation numbers per stratum in blocks of 10. Parexel (Waltham, MA, USA) generated the randomisation schedule and maintained the IRT
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Clinical laboratory tests, evaluation of vital signs, and other safety assessments were performed by blinded investigators
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Clinical laboratory tests, evaluation of vital signs, and other safety assessments were performed by blinded investigators
Incomplete outcome data (attrition bias) All outcomes	High risk	> 20% dropout
Selective reporting (reporting bias)	Low risk	No evidence of selective outcome reporting
Other bias	Low risk	No evidence of other bias

Guttman‐Yassky 2019b.

Study characteristics
Methods	Parallel double‐blinded randomised controlled trial Dates of study: not stated Locations: USA and Canada
Participants	Randomised: 54 participants Inclusion criteria: age ≥ 18 years; moderate to severe atopic dermatitis not controlled by topical treatments within 6 months before screening Exclusion criteria: not stated Dropouts and withdrawals: not stated
Interventions	Intervention: dupilumab 200 mg every week (n = 27) Comparator: placebo every week (n = 27)
Outcomes	Assessment time: at week 4, 16, and 32 (primary outcome was assessed at week 16) EASI, SCORAD, DLQI, Pruritus NRS, BSA, POEM, IGA
Funding source	Sanofi
Declarations of interest	Not described
Notes	Only abstract available
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Central randomisation
Allocation concealment (selection bias)	Low risk	Central randomisation
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Double‐blind
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Double‐blind
Incomplete outcome data (attrition bias) All outcomes	Low risk	< 20% dropout
Selective reporting (reporting bias)	Low risk	No potential sources of reporting bias
Other bias	Low risk	Baseline balanced

Guttman‐Yassky 2019c.

Study characteristics
Methods	Parallel double‐blinded randomised controlled trial Dates of study: March 2016 to June 2017 Locations: USA and Canada
Participants	Randomised: 64 participants Inclusion criteria: age ≥ 18 years; moderate to severe atopic dermatitis not controlled by topical treatments Exclusion criteria: not stated Dropouts and withdrawals: 2 subjects assigned to GBR830 withdrew informed consent before dosing
Interventions	Intervention: GBR830 (10 mg/kg) (n = 46) Comparator: placebo (n = 16)
Outcomes	Assessment time: at week 4 and 12 EASI50, IGA, TEAEs
Funding source	Glenmark Pharmaceuticals, S.A.
Declarations of interest	Not described
Notes	Not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Eligible subjects were randomised 3:1 to receive intravenous GBR830 (10 mg/kg) or corresponding placebo via a computer‐generated scheme
Allocation concealment (selection bias)	Low risk	Randomised via a computer‐generated scheme developed independently by the statistician
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quadruple (participant, care provider, investigator, outcomes assessor)
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quadruple (participant, care provider, investigator, outcomes assessor)
Incomplete outcome data (attrition bias) All outcomes	Low risk	Few losses to follow‐up (despite only 34/62 (54.8%) participants having completed the study, all 62 participants were included in the ITT analysis (Figure 2))
Selective reporting (reporting bias)	Low risk	No potential sources of selective outcome reporting
Other bias	Low risk	No other source of bias

Haeck 2011.

Study characteristics
Methods	Parallel open‐label randomised controlled trial Dates of study: November 2005 to November 2007 Location: not stated
Participants	Randomised: 55 participants Inclusion criteria: age ≥ 18 years; moderate to severe atopic dermatitis not controlled by topical treatments Exclusion criteria: oral immunosuppressive treatment in the last 6 weeks; concomitant ultraviolet therapy; any known hypersensitivity to CsA or MPA or components of the formulations; thrombocytopaenia (75,000/mm³); absolute neutrophil count < 1500/mm³ and/or leukocytopaenia (2500/mm³), and/or haemoglobin < 6.0 g/dL; history of malignancy within the last 5 years; wish for pregnancy during the study, pregnancy, or lactation; infection requiring continued therapy; known positivity for HIV; drug or alcohol abuse Dropouts and withdrawals: 8 participants discontinued the study: 4 in intervention group (1 possible allergy, 1 non‐compliance, 2 side effects), 4 in comparator group (3 side effects, 1 non‐compliance)
Interventions	Intervention: mycophenolate sodium 1440 mg/d (n = 24) Comparator: ciclosporin (3 mg/kg/d) (n = 26)
Outcomes	Assessment time: at week 6, 18, 30, and 42 SCORAD final score, safety, common AEs (fatigue, hypertrichosis), infection
Funding source	Novartis Pharma
Declarations of interest	Not described
Notes	Non‐blinding of patients
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Random allocation sequence using blocks of random numbers
Allocation concealment (selection bias)	Unclear risk	Rrandom allocation sequence using blocks of random numbers
Blinding of participants and personnel (performance bias) All outcomes	High risk	No
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Assessor of the objective SCORAD was blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	ADR (all 50 participants initially randomised were included in the analysis)
Selective reporting (reporting bias)	Low risk	No potential sources of selective outcome reporting
Other bias	Low risk	Baseline balanced

Hamilton 2014a.

Study characteristics
Methods	Parallel double‐blinded randomised controlled trial Dates of study: not stated Location: not stated
Participants	Randomised: 109 participants Inclusion criteria: moderate to severe atopic dermatitis Exclusion criteria: not stated Dropouts and withdrawals: not stated
Interventions	Intervention: dupilumab 300 mg SC weekly Comparator: placebo
Outcomes	Assessment time: at week 12 EASI change, 5‐D Pruritus Scale score, AEs, SAEs
Funding source	Sanofi
Declarations of interest	Not described
Notes	Data from abstract.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Insufficient information
Allocation concealment (selection bias)	Unclear risk	Insufficient information
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Insufficient information
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Insufficient information
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Insufficient information
Selective reporting (reporting bias)	Unclear risk	Insufficient information
Other bias	Unclear risk	Insufficient information

Han 2007.

Study characteristics
Methods	Parallel open‐label randomised controlled trial Dates of study: not stated Location: not stated
Participants	Randomised: 169 participants Inclusion criteria: chronic atopic dermatitis Exclusion criteria: not stated Dropouts and withdrawals: not stated
Interventions	Interventions: triamcinolone acetonide acetate (5 mg/5 mL) and 2% lidocaine (n = 86) Comparators: cyproheptadine 2 mg, cimetidine 200 mg 3 times/d; external application of cyproheptadine cream, 3 times/d (n = 83)
Outcomes	Assessment time: at week 6 and 24 IGA
Funding source	Not described
Declarations of interest	Not described
Notes	Not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No information
Allocation concealment (selection bias)	Unclear risk	No information
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	No information
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	No information
Incomplete outcome data (attrition bias) All outcomes	Low risk	All patients were included in the final analysis with no dropouts
Selective reporting (reporting bias)	Low risk	No potential sources of reporting bias
Other bias	Low risk	Baseline balanced

Hanifin 1993.

Study characteristics
Methods	Parallel double‐blinded randomised controlled trial Dates of study: not stated Location: not stated
Participants	Randomised: 83 participants Inclusion criteria: 2 to 65 years of age, severe atopic dermatitis Exclusion criteria: not stated Dropouts and withdrawals: 2 interferon gamma‐treated patients had therapy discontinued early because of constitutional symptoms (1) and disease flare (1). Three patients in the placebo group were dropped, 1 each because of disease flare, patient request, and patient error
Interventions	Interventions: interferon gamma 50 mg/m² (n = 40) SC daily Comparator: placebo (n = 43) SC daily
Outcomes	Unclear
Funding source	Grant from Genentech, Inc.
Declarations of interest	Not described
Notes	Not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Adaptive randomisation by central randomisation
Allocation concealment (selection bias)	Low risk	Adaptive randomisation by central randomisation
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Double‐blind
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Double‐blind
Incomplete outcome data (attrition bias) All outcomes	Low risk	Few losses to follow‐up
Selective reporting (reporting bias)	Low risk	No potential sources of reporting bias
Other bias	Low risk	Baseline balanced

Harper 2000.

Study characteristics
Methods	Parallel open‐label randomised controlled trial Dates of study: June 1995 to December 1997 Location: UK
Participants	Randomised: 43 participants Inclusion criteria: 2 ± 16 years of age, severe atopic dermatitis refractory to topical steroid therapy, without contraindications to use of CyA Exclusion criteria: treatment with systemic corticosteroids, cytotoxic agents, or phototherapy within 2 weeks before entry; previous treatment with CyA; abnormal renal or liver function; hypertension Dropouts and withdrawals: 3 were excluded due to no or minimal post‐baseline assessments, 11 patients were withdrawn prematurely: 6 short‐course and 5 continuous‐therapy patients
Interventions	Intervention: cyclosporin A short‐course (n = 21); treatment was stopped at week 12 following a 4‐week dose‐tapering period Comparator: cyclosporin A continuous (n = 19); after week 4, patients were maintained on a dose judged as appropriate by the investigator
Outcomes	SASSAS score, AEs, QOL, BSA
Funding source	Novartis UK
Declarations of interest	Not described
Notes	Not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No randomisation details provided
Allocation concealment (selection bias)	Unclear risk	No randomisation details provided
Blinding of participants and personnel (performance bias) All outcomes	Low risk	No blinding provided with subjective assessment
Blinding of outcome assessment (detection bias) All outcomes	High risk	No blinding provided with subjective assessment
Incomplete outcome data (attrition bias) All outcomes	High risk	Loss to follow‐up occurred
Selective reporting (reporting bias)	Low risk	No potential sources of reporting bias
Other bias	Unclear risk	No baseline provided

Heddle 1984.

Study characteristics
Methods	Cross‐over double‐blinded randomised controlled trial (4‐week washout period) Dates of study: not stated Location: not stated
Participants	Randomised: 27 participants Inclusion criteria: moderate or severe atopic dermatitis for at least 3 months that had failed to respond adequately to conventional therapy with emollients, weak topical corticosteroids, and systemic antihistamines Exclusion criteria: not stated Dropouts and withdrawals: 1 child withdrew because he developed whooping cough
Interventions	Intervention: beclomethasone dipropionate 4 times daily for 4 weeks Comparator: placebo
Outcomes	Unclear
Funding source	Not described
Declarations of interest	Not described
Notes	Not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Random numbers
Allocation concealment (selection bias)	Unclear risk	No details provided
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Identical appearance
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Identical appearance
Incomplete outcome data (attrition bias) All outcomes	Low risk	Few losses to follow‐up
Selective reporting (reporting bias)	Low risk	No potential sources of reporting bias
Other bias	Unclear risk	No sufficient baseline provided

Heil 2010.

Study characteristics
Methods	Double‐blinded randomised controlled trial Dates of study: not stated Location: not stated
Participants	Randomised: 20 participants, median age 30.0 (18.0 to 47.0) (omalizumab) (5 male), 26.0 (18.0 to 43.0) (placebo) (1 male) Inclusion criteria: male and female patients; 12 to 60 years of age; clinical diagnosis of AD (criteria of Hanifin and Rajka [11]); serum IgE between 30 and 1300 IU/mL; ≥ 1 significantly positive CAP; positive skin prick test of the same specificity as CAP; Investigator`s Global Assessment score of 2 at randomisation; stable AD, defined as active AD (IGA 2) for > 9 months per year; signed informed consent Exclusion criteria: medications or any other circumstances (e.g. poor compliance, concurrent skin disease) that might interfere with the outcome of the study
Interventions	Intervention: omalizumab at 0.016 mg/kg/IgE (IU/mL) per 4 weeks subcutaneously Comparator: placebo
Outcomes	Assessment every 4 weeks, for 16 weeks Primary outcomes: changes in immunological disease parameters assessed throughout the study: flow cytometry, immunohistology, and measurement of serum IgE levels Secondary outcomes: changes in skin tests (skin prick test, titrated skin prick test, atopy patch test) and in clinical course of the disease as assessed by Investigator's Global Assessment (IGA); Eczema and Severity Index (EASI) score; Investigators' Pruritus Severity Assessment (IPSA)
Funding source	Study was sponsored by Novartis
Declarations of interest	Thomas Hultsch: former full‐time employee of Novartis, co‐promoter of Xolair® (omalizumab)
Notes	Patients were allowed to use emollients and hydrocortisone acetate 1% cream
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	At the randomisation visit, patients were randomised by allocation to the next blinded treatment kit in the sequence of their inclusion
Allocation concealment (selection bias)	Unclear risk	No allocation concealment process was provided
Blinding of participants and personnel (performance bias) All outcomes	Low risk	At the randomisation visit, patients were randomised by allocation to the next blinded treatment kit in the sequence of their inclusion
Blinding of outcome assessment (detection bias) All outcomes	Low risk	At the randomisation visit, patients were randomised by allocation to the next blinded treatment kit in the sequence of their inclusion
Incomplete outcome data (attrition bias) All outcomes	Low risk	< 20% dropout
Selective reporting (reporting bias)	Low risk	No potential sources of reporting bias
Other bias	Low risk	Baseline balanced

Iyengar 2013.

Study characteristics
Methods	Parallel double‐blinded randomised controlled trial Dates of study: not stated Location: not stated
Participants	Randomised: 8 participants Inclusion criteria: severe atopic dermatitis Exclusion criteria: not stated Dropouts and withdrawals: not stated
Interventions	Intervention: omalizumab 150 to 375 mg every 2 to 4 weeks SC (n = 4) Comparator: placebo (n = 4)
Outcomes	SCORAD
Funding source	Not described
Declarations of interest	Not described
Notes	Not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No details provided
Allocation concealment (selection bias)	Unclear risk	No details provided
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	No details provided
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	No details provided
Incomplete outcome data (attrition bias) All outcomes	Low risk	All completed
Selective reporting (reporting bias)	High risk	Repeated measures with no clear presentation
Other bias	Unclear risk	No baseline provided

Jang 2000.

Study characteristics
Methods	Parallel open‐label randomised controlled trial Dates of study: not stated Location: not stated
Participants	Randomised: 51 participants Inclusion criteria: severe atopic dermatitis, not responding to conventional therapy Exclusion criteria: not stated Dropouts and withdrawals: not stated
Interventions	Interventions: low‐dose group (n = 20): interferon gamma 0.5 × 106 IU/m³; high‐dose group (n = 21): interferon gamma 1.5 × 106 IU/m³ Comparator: placebo (n = 10)
Outcomes	BSA, pruritus
Funding source	Grant from LG Chemical Ltd., and Catholic Medical Center Research fund for special projects
Declarations of interest	Not described
Notes	Not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No details provided
Allocation concealment (selection bias)	Unclear risk	No details provided
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	No details provided
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	No details provided
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	No details provided
Selective reporting (reporting bias)	Low risk	No potential sources of reporting bias
Other bias	Low risk	Baseline balanced

Jee 2011.

Study characteristics
Methods	Parallel randomised controlled trial Dates of study: not stated Location: not stated
Participants	Randomised: 40 participants Inclusion criteria: age > 2 years, moderate to severe atopic dermatitis, no response to conventional therapy Exclusion criteria: not stated Dropouts and withdrawals: 7 patients did not complete the study: 5 IVIG patients had side effects, including headache, nausea, and abdominal pain; 2 control patients did not complete the trial for personal reasons
Interventions	Intervention: IVIG 2 g/kg body weight/mo at each monthly visit (n = 30) Comparator: placebo (n = 10)
Outcomes	SCORAD
Funding source	Greencross Pharm Co.
Declarations of interest	Not described
Notes	Assessment was conducted at the first injection, at the second injection, at the third injection, after 3 months, and after 6 months
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No details provided
Allocation concealment (selection bias)	Unclear risk	No details provided
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	No details provided
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	No details provided
Incomplete outcome data (attrition bias) All outcomes	Low risk	All completed
Selective reporting (reporting bias)	Low risk	No potential sources of reporting bias
Other bias	High risk	Baseline imbalance in terms of gender and number of participants

Jin 2015.

Study characteristics
Methods	Parallel double‐blinded randomised controlled trial Dates of study: not stated Location: not stated
Participants	Randomised: 43 participants Inclusion criteria: age ≥ 7 years, moderate to severe atopic dermatitis, failed to be treated with topical therapy Exclusion criteria: prior systemic treatments, including systemic steroids, cyclosporine, and phototherapy within 1 month before the study; sensitivity to study drugs; malignancy or lymphoproliferative disease within the past 5 years; history of psychotic disorders; history of substance abuse; pregnancy, intention to become pregnant during the study period, or lactation; diseases or concurrent therapies that could interfere with evaluation of the study Dropouts and withdrawals: 3 patients in group A and 2 in group B were lost to follow‐up without starting treatment
Interventions	Intervention: cyclosporin A (25 mg/capsule) + glucosamine (312.5 mg/capsule): group A (n = 22) Comparator: cyclosporin A (25 mg/capsule) + placebo: group B (n = 21)
Outcomes	SCORAD, AEs
Funding source	National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIP)
Declarations of interest	Not described
Notes	Not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Interactive web response system
Allocation concealment (selection bias)	Low risk	Interactive web response system
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Identical appearance
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Identical appearance
Incomplete outcome data (attrition bias) All outcomes	High risk	Several losses to follow‐up were noted with no imputation
Selective reporting (reporting bias)	Low risk	No potential sources of reporting bias
Other bias	Low risk	Baseline balanced

Khattri 2017.

Study characteristics
Methods	Cross‐over double‐blinded randomised controlled trial Dates of study: not stated Location: not stated
Participants	Randomised: 33 participants Inclusion criteria: 18 to 75 years of age, moderate to severe atopic dermatitis (SCORAD ≥ 15) that could not be managed with conventional therapy Exclusion criteria: previously treated with any agent blocking IL‐12 or IL‐23, current malignancy or autoimmune disease, history of malignancy (except adequately treated or excised basal/squamous cell carcinoma of the skin), evidence of active or latent infection Dropouts and withdrawals: 1 patient from the placebo‐ustekinumab group was retrospectively excluded from analyses due to initial misdiagnosis, 1 patient was excluded from tissue analyses due to quality control issues, 1 patient had dropped out already by week 2 and therefore was not included in tissue analyses
Interventions	Interventions: ustekinumab at weeks 0, 4, and 16 with cross‐over to the other agent at weeks 16, 20, and 32 (n = 16) Comparators: placebo at weeks 0, 4, and 16 with cross‐over to the other agent at weeks 16, 20, and 32 (n = 16)
Outcomes	SCORAD, DLQI, SAEs
Funding source	Research grant from Janssen Pharmaceuticals was allocated to E.G.Y. and J.G.K. P.M.B. and M.S.F. were supported in part by a grant
Declarations of interest	Not described
Notes	Not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated participant randomisation by unblinded pharmacists
Allocation concealment (selection bias)	Low risk	Computer‐generated participant randomisation by unblinded pharmacists
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Double‐blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Double‐blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	< 20% dropout
Selective reporting (reporting bias)	Low risk	No potential sources of reporting bias
Other bias	Low risk	Baseline balanced

Kim 2016.

Study characteristics
Methods	Parallel randomised controlled trial Dates of study: March 2013 to March 2014 Location: Korea
Participants	Randomised: 60 participants Inclusion criteria: moderate to severe atopic dermatitis not controlled by topical treatments Exclusion criteria: receipt of other systemic treatments (systemic immunosuppressive therapy, antibiotics, phototherapy, etc.); uncontrolled chronic disease (diabetes mellitus or hypertension); other clinically relevant systemic disease (abnormal renal or liver function, history of malignancy, etc.); pregnancy or lactation Dropouts and withdrawals: patients who dropped out due to lack of compliance (group A: n = 2, group B: n = 1) and adverse effects (group A: n = 1) were excluded from analysis of treatment success rate
Interventions	Interventions: cyclosporin A + topical agent (medium‐potency topical corticosteroids (methylprednisolone aceponate) on trunk and extremities, or topical calcineurin inhibitors (0.1% tacrolimus ointment) on the face) (n = 30) Comparator: cyclosporin A alone (n = 30)
Outcomes	IGA success rate, EASI final score, total AEs
Funding source	Not described
Declarations of interest	Not described
Notes	Incomplete outcome data
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No details provided
Allocation concealment (selection bias)	Unclear risk	No details provided
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	No details provided
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	No details provided
Incomplete outcome data (attrition bias) All outcomes	High risk	All completed (40/60 (66.7%)) patients enrolled were included in the analysis
Selective reporting (reporting bias)	Low risk	No potential sources of selective outcome reporting
Other bias	Low risk	Baseline balanced

Koppelhus 2014.

Study characteristics
Methods	Cross‐over double‐blinded randomised controlled trial (4 to 8 weeks washout period) Dates of study: not stated Location: not stated
Participants	Randomised: 20 participants Inclusion criteria: severe atopic dermatitis with refractoriness to standard topical treatment (corticosteroid ointments, UVA, UVB, PUVA, tar) Exclusion criteria: pregnancy, uncontrolled hypertension, previous malignancy, infectious disease, liver/kidney disease, active treatment with ECP or immunosuppressants within 4 weeks before start of trial Dropouts and withdrawals: 1 patient dropped out during terminal extracorporeal photopheresis treatment (due to pregnancy); 4 patients chose to leave the study during the terminal course of cyclosporin A
Interventions	Intervention: cyclosporin A 3 mg/kg (n = 10) Comparator: extracorporeal photopheresis 2 J/cm² (n = 10)
Outcomes	SCORAD, Pruritus VAS score
Funding source	Aage Bang’s Foundation
Declarations of interest	Not described
Notes	Not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No details provided
Allocation concealment (selection bias)	Unclear risk	No details provided
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	No details provided
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	No details provided
Incomplete outcome data (attrition bias) All outcomes	High risk	25% dropout
Selective reporting (reporting bias)	Low risk	No potential sources of reporting bias
Other bias	Low risk	Cross‐over

Kwon 2013.

Study characteristics
Methods	Cross‐over double‐blinded pilot study Dates of study: not stated Location: South Korea
Participants	Randomised: 10 participants Inclusion criteria: male and female; more than 12 years old; AD recalcitrant to topical therapy Exclusion criteria: had taken any immunosuppressants including cyclosporin A within 1 month before start of the study. Women who were pregnant or breastfeeding, patients with malignant tumours or psychotic disorders, and individuals with disease or who were receiving concurrent therapies that could interfere with evaluation Dropouts and withdrawals: 2 dropped out due to violation of medication schedule
Interventions	Intervention: cyclosporin A 3 mg/kg with glucosamine 1000 mg/d (n = 5); cross‐over every 2 weeks; no washout period Comparator: cyclosporin A 3 mg/kg alone (n = 5)
Outcomes	% change SCORAD
Funding source	Not described
Declarations of interest	Not described
Notes	Not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No details provided
Allocation concealment (selection bias)	Unclear risk	No details provided
Blinding of participants and personnel (performance bias) All outcomes	Low risk	No details provided
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Two independent dermatologists who were blinded to treatment assignment
Incomplete outcome data (attrition bias) All outcomes	Low risk	Few losses to follow‐up
Selective reporting (reporting bias)	Low risk	No potential sources of selective outcome reporting
Other bias	Low risk	Cross‐over

La Rosa 1995.

Study characteristics
Methods	Cross‐over double‐blinded RCT Dates of study: not stated Location: multi‐centre
Participants	Randomised: 20 participants Inclusion criteria: not stated Exclusion criteria: not stated Dropouts and withdrawals: not stated
Interventions	Intervention: flunisolide 640 mcg (< 3 years old) and 1200 mcg for older patients for 2 weeks (n = 10); 1‐week washout period Comparator: placebo (n = 10)
Outcomes	Total clinical severity score
Funding source	Valeas s.p.a.
Declarations of interest	Not described
Notes	Not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No details provided
Allocation concealment (selection bias)	Unclear risk	No details provided
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	No details provided
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	No details provided
Incomplete outcome data (attrition bias) All outcomes	Low risk	All completed
Selective reporting (reporting bias)	Low risk	No potential sources of reporting bias
Other bias	Low risk	Cross‐over

Meggitt 2006.

Study characteristics
Methods	Parallel double‐blinded RCT Dates of study: February 2001 to September 2002 Location: United Kingdom
Participants	Randomised: 63 participants Inclusion criteria: 16 to 65 years of age with moderate to severe atopic eczema Exclusion criteria: patients who had been admitted for eczema, had used phototherapy or sunbeds, or had been treated with ciclosporin, systemic steroids, Chinese herbal medicine, topical tacrolimus, or evening primrose oil during the preceding 3 months Dropouts and withdrawals: 9 participants were withdrawn: 2 before treatment, 6 during azathioprine treatment, and 1 during placebo treatment, due to azathioprine hypersensitivity, severe nausea, headache, and malaise
Interventions	Intervention: azathioprine 1.0 mg/kg daily for 12 weeks (n = 42) Comparator: placebo (n = 21)
Outcomes	Disease activity (SASSAD), % BSA, itch score (patient assessed), loss of sleep score, quality of life (DLQI), IGA, PsGA
Funding source	Not described
Declarations of interest	Not described
Notes	Not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Minimisation
Allocation concealment (selection bias)	Low risk	Neither investigator nor participants knew which treatment had been allocated until the code was broken after the 3‐month follow‐up assessment
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Neither investigator nor participants knew which treatment had been allocated until the code was broken after the 3‐month follow‐up assessment
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Neither investigator nor participants knew which treatment had been allocated until the code was broken after the 3‐month follow‐up assessment
Incomplete outcome data (attrition bias) All outcomes	Low risk	15% dropout
Selective reporting (reporting bias)	Low risk	No potential sources of reporting bias
Other bias	Low risk	Baseline balanced

Mikhak 2019.

Study characteristics
Methods	Parallel double‐blinded RCT Dates of study: not stated Location: not stated
Participants	Randomised: 82 participants, phase 1a healthy volunteers (n = 50) and phase 1b patients with AD (n = 32) Inclusion criteria: not stated Exclusion criteria: not stated Dropouts and withdrawals: not stated
Interventions	Intervention: KPL‐716 phase 1a (n = 37) and phase 1b patients with AD (n = 20) Comparator: placebo phase 1a (n = 13) and phase 1b patients with AD (n = 12)
Outcomes	Pruritus NRS, SAEs
Funding source	Kiniksa
Declarations of interest	Not described
Notes	Information from abstract
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Insufficient information
Allocation concealment (selection bias)	Unclear risk	Insufficient information
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Insufficient information
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Insufficient information
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Insufficient information
Selective reporting (reporting bias)	Unclear risk	Insufficient information
Other bias	Unclear risk	Insufficient information

Munro 1994.

Study characteristics
Methods	Cross‐over double‐blinded RCT Dates of study: not stated Location: United Kingdom
Participants	Randomised: 24 participants Inclusion criteria: not stated Exclusion criteria: not stated Dropouts and withdrawals: not stated
Interventions	Intervention: cyclosporin A 5 mg/kg for 8 weeks (n = 12); no washout period Comparator: placebo (n = 12)
Outcomes	% BSA, SASSAD, itch and sleep loss scores, AEs
Funding source	Sandoz Pharmaceuticals, Ltd.
Declarations of interest	Not described
Notes	In the second phase, all patients who had completed phase 1 were treated with CyA 5 mg/kg for a further 2 weeks, then were randomised to 1 of 2 protocols for dose reduction: (i) reducing the daily dose of CyA by 1 mg/kg every 2 weeks; and (ii) maintaining the dose at 5 mg/kg but increasing the interval between doses by 1 day every 2 weeks. Assessment and dose reductions were made every 2 weeks
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No details provided
Allocation concealment (selection bias)	Unclear risk	No details provided
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Identical placebo
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Identical placebo
Incomplete outcome data (attrition bias) All outcomes	High risk	50% of patients completed the trial
Selective reporting (reporting bias)	Low risk	No potential sources of reporting bias
Other bias	Low risk	Cross‐over

NCT01552629.

Study characteristics
Methods	Parallel double‐blinded RCT Dates of study: January 2012 to August 2013 Locations: Austria, France, Germany
Participants	Randomised: 22 participants Inclusion criteria: not stated Exclusion criteria: not stated Dropouts and withdrawals: 4 participants were withdrawn ‐ 3 during QGE031 treatment and 1 during cyclosporin A treatment ‐ due to adverse events, consent withdrawn by participant, and loss to follow‐up
Interventions	Interventions: QGE031 SC every 2 weeks (n = 10), cyclosporin A 2.5 to 5.0 mg/kg twice daily (n = 2) Comparator: placebo (n = 10)
Outcomes	EASI, EASI50, IGA, AEs
Funding source	Novartis Pharmaceuticals
Declarations of interest	Not described
Notes	The study consisted of up to a 28‐day screening period (day 28 to day 1), a treatment period of 12 weeks, a follow‐up period of 12 weeks, and an end of study (EoS) evaluation
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No randomisation details provided
Allocation concealment (selection bias)	Unclear risk	No details provided
Blinding of participants and personnel (performance bias) All outcomes	Low risk	A QGE031 matched placebo will be administered as a subcutaneous dose q2 weeks. The study remained blinded to all individuals (including patient, investigator, data analyst, assessor) until the time of pre‐specified unblinding
Blinding of outcome assessment (detection bias) All outcomes	Low risk	No details of blinding provided
Incomplete outcome data (attrition bias) All outcomes	High risk	>20% drop‐out rate, but all participants were included in the analysis
Selective reporting (reporting bias)	Low risk	No potential sources of selective outcome reporting
Other bias	Unclear risk	The trial was stopped early based on efficacy results of the first interim analysis, and it is unclear to what extent results at the point of stoppage were biased in relation to the results, should the trial be allowed to continue until its pre‐specified point of completion

NCT01785602.

Study characteristics
Methods	Parallel double‐blinded RCT Dates of study: June 2013 to November 2014 Location: multi‐centre
Participants	Randomised: 103 participants Inclusion criteria: 18 to 65 years Exclusion criteria: history of hypersensitivity to any of the study drugs; history of serious allergic reactions to any allergen such as anaphylactic shock or life‐threatening asthma; prior intubation; respiratory arrest; hospitalisation due to asthma within the last 3 months or seizures as a result of asthma; history of clinically significant ECG abnormalities or screening/baseline ECG that demonstrated clinical significant abnormalities that could affect patient safety or interpretation of study results; history of long QT syndrome or prolonged QTc interval (Fredericia's) (> 450 msec for males and females) at screening; use of topical prescription treatment (e.g. topical corticosteroids, calcineurin inhibitors, antibiotics) within 2 weeks before initial dosing of study drug. Patient use of emollients was encouraged Dropouts and withdrawals: 22 participants were withdrawn ‐ 13 during QAW039 treatment and 9 during placebo treatment ‐ due to withdrawal by patient, loss to follow‐up, adverse events, and administrative problems
Interventions	Intervention: QAW039 450 mg PO daily (n = 76) Comparator: placebo (n = 27)
Outcomes	Mean change in EASI, AE, SAE, infection
Funding source	Novartis Pharmaceuticals
Declarations of interest	Not described
Notes	Not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Insufficient information to assess the validity of randomisation and allocation concealment
Allocation concealment (selection bias)	Unclear risk	Insufficient information to assess the validity of randomisation and allocation concealment
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Matching placebo
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Double‐blind (patient, investigator)
Incomplete outcome data (attrition bias) All outcomes	Low risk	All completed (all 103 patients initially randomised were included in the analysis)
Selective reporting (reporting bias)	Low risk	No potential sources of selective outcome reporting
Other bias	Low risk	Baseline balanced

NCT01945086.

Study characteristics
Methods	Parallel double‐blinded RCT Dates of study: September 2013 to December 2014 Location: Japan
Participants	Randomised: 79 participants Inclusion criteria: no history of latent or active TB before screening; no signs or symptoms suggestive of active TB; no recent close contact with a person with active TB; negative interferon gamma release assay (IGRA) result within 2 months before first administration of study drug Exclusion criteria: history of or current clinically significant medical illness that the investigator considers should exclude the participant or that could interfere with interpretation of study results; indeterminate initial and repeat IGRA result or newly positive IGRA result and unwilling or unable to undergo TB prophylaxis treatment; has received any of the following medications or therapies within 4 weeks before randomisation including systemic non‐steroid immunosuppressive or immunomodulatory drugs; systemic corticosteroids; high daily dose of inhaled corticosteroids; topical corticosteroids of strongest potency for atopic dermatitis; topical antihistamines (including topical doxepin); topical anesthetics; topical nonsteroidal anti‐inflammatory drugs; topical counter‐irritants (eg, capsaicin, menthol, wintergreen oil); antidepressants or antipsychotics; soporifics; phototherapy including ultraviolet A, ultraviolet B, and psoralen with ultraviolet A (PUVA); hyposensitization (desensitization) therapy Dropouts and withdrawals: 3 participants were withdrawn ‐ 1 during placebo treatment and 2 during ustekinumab 90 mg treatment ‐ due to withdrawal by patient and adverse events
Interventions	Interventions: ustekinumab 45 mg (n = 24) and ustekinumab 90 mg (n = 28) at week 0 and at week 4 Comparator: placebo (n = 27)
Outcomes	Assessment time: at week 2, and every 4 weeks for 24 weeks EASI, IGA, DLQI, Atopic Dermatitis Itch Scale (ADIS)
Funding source	Janssen Pharmaceuticals, K.K.
Declarations of interest	Not described
Notes	For 7 evaluation visits on weeks 2, 4, 8, 12, 16, 20, and 24
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No details provided
Allocation concealment (selection bias)	Unclear risk	No details provided
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quadruple masking
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quadruple masking
Incomplete outcome data (attrition bias) All outcomes	Low risk	20% dropout
Selective reporting (reporting bias)	Low risk	No potential sources of reporting bias
Other bias	Low risk	Baseline balanced

NCT02002208.

Study characteristics
Methods	Parallel double‐blinded RCT Dates of study: October 2013 to February 2016 Location: Sheffield, United Kingdom
Participants	Moderate to severe atopic dermatitis Randomised: 142 participants Inclusion criteria: Atopic dermatitis as defined by a score ≥ 9 on the Nottingham Eczema Severity Score, stratified into moderate (scores 9 to 11 inclusive) and severe (scores 12 to 15 inclusive) disease Fully documented history of the use of topical corticosteroids (TCSs) and/or topical calcineurin inhibitors (TCIs). Patients without a fully documented history will be excluded from the study Male and female patients with moderate to severe atopic dermatitis treated with TCS and/or TCI (with or without emollients) at the time of screening and over the previous month Must have had at least 1 AD flare in the previous 6 months Exclusion criteria: Receipt of any forbidden medication including over‐the‐counter preparations and herbal medicines within 14 days of the first dosing day with the exception of paracetamol up to a maximum of 2 g daily Use of systemic steroids within 4 weeks of the screening visit, light therapy or immunosuppressants within 2 months of the screening visit Use of NSAIDs Patients initially diagnosed with AD and 2 years of age or older will be excluded unless they have either coexisting or a history of asthma and/or allergic rhinoconjunctivitis Patients with contact dermatitis will be excluded Patients with acute skin infection or acute disease flares will be excluded. Those with active flares during screening to randomisation period (visit 1 to 2) may be treated according to normal clinical practice and must be re‐screened once their flares are no longer active Dropouts and withdrawals: 77 participants were withdrawn: 37 during OC000459 treatment and 40 during placebo treatment
Interventions	Intervention: OC000459 (timapiprant) 50 mg daily (n = 69) Comparator: placebo (n = 70)
Outcomes	Mean change in EASI, AEs, SAEs, infection
Funding source	Atopix Therapeutics, Ltd.
Declarations of interest	Not described
Notes	Incomplete outcome data
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Random numbers
Allocation concealment (selection bias)	Low risk	Allocation concealment mentioned
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Matching placebo
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Double‐blind (patient, investigator)
Incomplete outcome data (attrition bias) All outcomes	Low risk	It appeared that all 139 participants initially randomised were included in the analysis for the outcome data available.
Selective reporting (reporting bias)	Low risk	No potential sources of selective outcome reporting
Other bias	Low risk	Baseline balanced

NCT02395133.

Study characteristics
Methods	Parallel double‐blinded RCT Dates of study: March 2015 to October 2016 Location: multi‐centre
Participants	Randomised: 422 participants Inclusion criteria: not stated Exclusion criteria: not stated Dropouts and withdrawals: 47 participants were withdrawn ‐ 14 during placebo treatment, 9 during dupilumab 300 mg q8w treatment, 10 during dupilumab 300 mg q4w treatment, and 14 during dupilumab 300 mg q2w treatment ‐ due to protocol deviation, withdrawal by patient, adverse events, pregnancy, loss to follow‐up, sponsor decision, and reason other than specified
Interventions	Interventions: dupilumab 300 mg q2w (n = 169), dupilumab 300 mg q4w (n = 86), dupilumab 300 mg q8w (n = 84) Comparator: placebo (n = 83)
Outcomes	EASI, IGA, SCORAD, Pruritus BSA, POEM, AEs, SAEs
Funding source	Regeneron Pharmaceuticals
Declarations of interest	Not described
Notes	Not described
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No details provided
Allocation concealment (selection bias)	Unclear risk	No details provided
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Matching placebo
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Double‐blind patient, investigator, assessor
Incomplete outcome data (attrition bias) All outcomes	Low risk	No details (although 155/169 participants (91.7%) completed the trial, all 169 participants were included in the analysis)
Selective reporting (reporting bias)	Low risk	No potential sources of selective outcome reporting
Other bias	Unclear risk	Insufficient information

NCT02594098.

Study characteristics
Methods	Parallel double‐blinded RCT Dates of study: November 2015 to January 2018 Location: not stated
Participants	Randomised: 41 participants Inclusion criteria: Male or female at least 18 years of age If female, not pregnant or nursing Able to provide written informed consent and to comply with requirements of this study protocol Chronic (> 6 months) atopic dermatitis (intrinsic disease with IgE levels < 200, and extrinsic disease with IgE levels > 200) Moderate to severe AD (SCORAD Index ≥ 25, and IGA Index ≥ 3) Women of childbearing potential must have a negative urine pregnancy test at screening and must be practicing an adequate, medically acceptable method of birth control for at least 30 days before day 0 and at least 6 months after last study drug administration. Acceptable methods of birth control include intrauterine device (IUD); oral, transdermal, implanted, or injected hormonal contraceptives (must have been initiated at least 1 month before entering the study); tubal ligation; abstinence; and barrier methods with spermicide. Otherwise, if not of childbearing potential, must have a sterile or vasectomised partner; must have had a hysterectomy or a bilateral oophorectomy, or be clinically diagnosed infertile; or must be in a menopausal state for at least a year Tuberculin purified protein derivative (PPD) or QuantiFERON TB‐Gold test (QFT) negative at the time of screening; if history of positive PPD or QuantiFERON, must have completed appropriate prophylaxis Judged to be in good general health as determined by the principal investigator based upon results of medical history, laboratory profile, and physical examination Stable chronic asthma, treated with inhaled corticosteroids Exclusion criteria: not stated Dropouts and withdrawals: 39 participants were withdrawn ‐ 26 during secukinumab treatment and 13 during placebo treatment ‐ due to lack of efficacy, withdrawal by patient, termination of participation, seeking alternative treatment, or another reason
Interventions	Intervention: secukinumab 300 mg SC at weeks 0, 1, 2, 3, 4 (n = 27) Comparators: placebo at weeks 0, 1, 2, 3, 4 (n = 14); at week 16, all patients received secukinumab 300 mg
Outcomes	Assessment time: at 1, 4, 8, and 13 months IGA, EASI, SCORAD, AEs
Funding source	Icahn School of Medicine at Mount Sinai
Declarations of interest	Not described
Notes	No details of randomisation and blinding
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No randomisation details provided
Allocation concealment (selection bias)	Unclear risk	No details provided
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Both secukinumab and placebo were formulated as subcutaneous injections (pre‐filled syringes), which looked alike
Blinding of outcome assessment (detection bias) All outcomes	Low risk	No details of blinding provided
Incomplete outcome data (attrition bias) All outcomes	Low risk	Per protocol with high degree of withdrawal and differential withdrawal
Selective reporting (reporting bias)	Low risk	No potential sources of selective outcome reporting
Other bias	Low risk	Baseline balanced

NCT02780167.

Study characteristics
Methods	Parallel double‐blinded RCT Dates of study: April 2016 to April 2017 Location: multi‐centre
Participants	Randomised: 267 participants Inclusion criteria: not stated Exclusion criteria: not stated Dropouts and withdrawals: 110 participants were withdrawn: 22 during PF‐04965842 10 mg treatment, 24 during PF‐04965842 30 mg treatment, 19 during PF‐04965842 100 mg treatment, 17 during PF‐04965842 200 mg treatment, and 28 during placebo treatment, due to protocol deviation, lack of efficacy, adverse events, consent withdrawn by participant, and loss to follow‐up
Interventions	Interventions: PF‐04965842 10 mg (n = 49), PF‐04965842 30 mg (n = 51), PF‐04965842 100 mg (n = 51), PF‐04965842 200 mg (n = 51) PO daily Comparator: placebo (n = 56)
Outcomes	IGA, EASI, SCORAD, BSA, Pruritus NRS scores, DLQI, POEM, AEs
Funding source	Pfizer Inc.
Declarations of interest	Not described
Notes	No details of randomisation or blinding
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Patients were randomly assigned (1:1:1:1:1) to receive oral abrocitinib (200, 100, 30, or 10 mg) once daily or placebo once daily for 12 weeks. Randomisation was via an interactive response technology system
Allocation concealment (selection bias)	Low risk	Patients were randomly assigned (1:1:1:1:1) to receive oral abrocitinib (200, 100, 30, or 10 mg) once daily or placebo once daily for 12 weeks. Randomisation was via an interactive response technology system
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Blinded abrocitinib and placebo tablets were delivered to study sites in blister packs. Patients, investigators, and sponsors were blinded to study treatment
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Blinded abrocitinib and placebo tablets were delivered to study sites in blister packs. Patients, investigators, and sponsors were blinded to study treatment
Incomplete outcome data (attrition bias) All outcomes	High risk	Per protocol with high degree of withdrawal and differential withdrawal
Selective reporting (reporting bias)	Low risk	No potential sources of selective outcome reporting
Other bias	Low risk	Baseline balanced

NCT03054428.

Study characteristics
Methods	Parallel double‐blinded RCT Dates of study: March 2017 to June 2018 Locations: United States and Canada
Participants	Randomised: 251 participants Inclusion criteria: not stated Exclusion criteria: not stated Dropouts and withdrawals: 242 participants were withdrawn ‐ 80 during dupilumab 300 mg SC Q4W treatment, 79 during dupilumab 300 mg SC Q2W treatment, and 83 during placebo treatment ‐ due to transition to OLE study, lack of efficacy, physician decision, consent withdrawn by patient, and loss to follow‐up
Interventions	Interventions: dupilumab 300 mg SC Q4W (n = 84), dupilumab 300 mg SC Q2W (n = 82) Comparator: placebo (n = 85)
Outcomes	IGA, EASI, SCORAD, BSA, Pruritus NRS scores, DLQI, POEM, AEs
Funding source	Regeneron Pharmaceuticals, Inc.
Declarations of interest	Not described
Notes	Not described
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No randomisation details provided
Allocation concealment (selection bias)	Unclear risk	No details provided
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Participants received placebo matching dupilumab q2w (including doubling the amount of placebo on day 1 to match the loading dose). Study remained blinded to all individuals (including patient, investigator, carer, assessor) until pre‐specified time of unblinding
Blinding of outcome assessment (detection bias) All outcomes	Low risk	No details of blinding provided
Incomplete outcome data (attrition bias) All outcomes	Low risk	All completed (all 251 participants initially randomised were included in the analysis of endpoint values so far ‐ study is ongoing with final follow‐up assessment pending)
Selective reporting (reporting bias)	Low risk	No potential sources of selective outcome reporting
Other bias	Unclear risk	Insufficient information

NCT03055195.

Study characteristics
Methods	Parallel double‐blinded RCT Dates of study: not stated Location: not stated
Participants	Randomised: 34 participants, mean age 34.6 (13.99) (mepolizumab) (12 male); 30.8 (9.95) (placebo) (9 male) Inclusion criteria: between 18 and 70 years of age inclusive at the time of signing informed consent; AD diagnosed by the Eichenfield revised criteria of Hanifin and Rajka; diagnosis of AD ≥ 2 years before screening visit; IGA score ≥ 3 at screening and baseline visits; AD involvement of ≥ 10% BSA at screening and baseline visits; EASI score ≥ 16 at screening and baseline visits; absolute blood eosinophil count ≥ 350 cells/microlitre at screening visit Exclusion criteria: not stated Dropouts and withdrawals: 26 participants were withdrawn ‐ 14 during mepolizumab 100 mg SC treatment and 12 during placebo treatment ‐ due to lack of efficacy, withdrawal by patient, study terminated by sponsor, protocol violation, and physician decision
Interventions	Intervention: mepolizumab 100 mg SC, every 4 weeks, for 16 weeks (n = 18) Comparator: placebo (n = 16)
Outcomes	Assessment every 4 weeks, for 16 weeks Primary outcomes: number of participants with Investigators' Global Assessment (IGA) score of 0 or 1 and at least a 2‐ grade of improvement at week 16 Secondary outcomes: mean percentage change in Eczema Area and Severity Index (EASI) score from baseline to each study visit; number of participants with IGA score of 0 or 1 and at least a 2‐ grade of improvement at each study visit; numbers of participants with on‐treatment adverse events (AEs), serious adverse events (SAEs), and non‐serious adverse events (nSAEs); numbers of participants with on‐treatment AEs of special interest reported as local site injection reactions and systemic reactions
Funding source	GlaxoSmithKline
Declarations of interest	Not stated
Notes	Research report
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Interactive voice response system
Allocation concealment (selection bias)	Low risk	Interactive voice response system
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Blinded study team, which included a blinded study medical monitor and a blinded operations and science lead
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Blinded study team, which included a blinded study medical monitor and a blinded operations and science lead
Incomplete outcome data (attrition bias) All outcomes	High risk	Trial ended prematurely and technical problems led to unreliable data
Selective reporting (reporting bias)	Low risk	No potential sources of selective outcome reporting
Other bias	Unclear risk	Trial was stopped early based on efficacy results of the first interim analysis, and it is unclear to what extent results at the point of stoppage were biased in relation to results should the trial have been allowed to continue until its pre‐specified point of completion

Nemoto 2016.

Study characteristics
Methods	Parallel double‐blinded RCT Dates of study: not stated Location: Japan
Participants	Randomised: 116 participants Inclusion criteria: participants in parts A (Japanese) and B (white) were healthy volunteers 20 to 49 years of age with body mass index ≥ 185 and < 250 (Japanese) or < 300 (white); part C enrolled Japanese patients 20 to 49 years of age who had been diagnosed with AD of moderate or greater severity, with marked inflammatory cutaneous lesions covering ≥ 5% of their body surface area despite continued treatment with topical corticosteroids for ≥ 12 weeks; a visual analogue scale (VAS) score for pruritus ≥ 50 mm; and receiving topical treatment with hydrocortisone butyrate (Locoid; Torii Pharmaceutical Co., Ltd., Tokyo, Japan) and/or moisturisers during the run‐in period (days 7 to 1) Exclusion criteria: skin disease other than AD, ocular symptoms, eczema herpeticum, molluscum contagiosum, impetigo, allergic disease requiring treatment with corticosteroids or antihistamines, new topical therapy or changes to the dosage regimen of a topical therapy, use of inhaled or intranasal corticosteroids other than topical hydrocortisone butyrate within 2 weeks of the start of or during the run‐in period. Patients were not allowed to use concomitant drugs for AD other than hydrocortisone butyrate and moisturisers during the run‐in period Dropouts and withdrawals: not stated
Interventions	Part A Intervention: PFCIM331 (n = 42) Comparator: placebo (n = 56) Part B Intervention: PFCIM331 (n = 18) Comparator: placebo (n = 6) Part C Intervention: PFCIM331 (n = 27) Comparator: placebo (n = 9)
Outcomes	Pruritus VAS score, AEs
Funding source	Chugai Pharmaceutical
Declarations of interest	Not described
Notes	Not described
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Investigational product randomly assigned to eligible participants. Following eligibility screening, the manager allocated CIM331 and placebo at 3:1
Allocation concealment (selection bias)	Low risk	Randomly assigned eligible participants. Following eligibility screening, the manager allocated CIM331 and placebo at 3:1
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Filled syringes were indistinguishable from each other
Blinding of outcome assessment (detection bias) All outcomes	Low risk	The investigator who administered study drug was different from the person who evaluated safety in the same dose group
Incomplete outcome data (attrition bias) All outcomes	Low risk	Few discontinued
Selective reporting (reporting bias)	Low risk	No potential sources of selective outcome reporting
Other bias	Low risk	Baseline balanced

Neoral group 2008.

Study characteristics
Methods	Parallel open‐label randomised clinical trial Dates of study: not stated Location: multi‐centre
Participants	Randomised: 106 participants Inclusion criteria: patients between the ages of 18 and 64 years who had not responded to conventional therapies Exclusion criteria: not stated Dropouts and withdrawals: 13 participants were withdrawn: 1 before treatment, 6 during cyclosporin A 1 mg/kg/d treatment, 4 during cyclosporin A 3 mg/kg/d treatment, and 2 during cyclosporin A 5 mg/kg/d treatment
Interventions	Interventions: cyclosporin A 1 mg/kg/d (n = 34), cyclosporin A 3 mg/kg/d (n = 37), cyclosporin A 5 mg/kg/d 100 mg (n = 34) PO twice daily for 4 weeks
Outcomes	Disease severity score (8‐area, 4‐sign score)
Funding source	Not described
Declarations of interest	Not described
Notes	‐
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Insufficient information
Allocation concealment (selection bias)	Unclear risk	Insufficient information
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Insufficient information
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Insufficient information
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Insufficient information
Selective reporting (reporting bias)	Unclear risk	Insufficient information
Other bias	Unclear risk	Insufficient information

Oldhoff 2005.

Study characteristics
Methods	Parallel double‐blinded RCT Dates of study: January until July 2002 Location: six centers in Europe
Participants	Randomised: 43 participants with mean age 29 (range 18–57) Inclusion criteria: AD patients diagnosed according to Hanifin and Rajka criteria who had experienced a flare of AD, defined as objective SCORing AD (SCORAD) between 20 and 40 Exclusion criteria: Patients who received local and systemic treatment within 14 days prior to study Dropouts and withdrawals: 4 participants were withdrawn: 3 in the mepolizumab 750 mg IV treatment group (2 exacerbation, 1 lack of effect) and 1 in the placebo group (loss to follow‐up)
Interventions	Intervention: mepolizumab 750 mg IV at day 0 and day 7 (n = 20) Comparator: placebo (n = 23) Fluticasone propionate cream, 0.05% once daily, was given to nonresponders as rescue medication at day 16
Outcomes	Assessment time: Day 0, 2, 14 and 28 for evaluating efficacy outcome and day 30 for evaluating safety outcome Primary outcome: Physicians’ Global Assessment (PGA) at day 14 Secondary outcome: SCORAD, Pruritus NRS score, and safety
Funding source	GlaxoSmithKline Pharmaceuticals
Declarations of interest	Not described
Notes	Not described
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No details provided
Allocation concealment (selection bias)	Unclear risk	No details provided
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	No details provided
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	No details provided
Incomplete outcome data (attrition bias) All outcomes	Low risk	90% completed
Selective reporting (reporting bias)	Low risk	No potential sources of reporting bias
Other bias	Low risk	Baseline balanced

Pacor 2004.

Study characteristics
Methods	Parallel double blinded RCT Dates of study: not stated Location: Italy
Participants	Randomised: 30 participants Inclusion criteria: clinical history of AD with confirmed diagnosis of moderate to severe AD, according to the criteria of Rajka and Langeland Exclusion criteria: treatment with systemic corticosteroids, cytotoxic agents, or phototherapy within 6 weeks before entry; previous treatment with tacrolimus or with ciclosporin; abnormal renal or liver function; hypertension; pregnancy or breastfeeding Dropouts and withdrawals: participants were withdrawn during treatment and during placebo treatment due to protocol deviation, lack of efficacy, adverse events, consent withdrawn by participant, and loss to follow‐up
Interventions	Intervention: tacrolimus 0.1% ointment twice a day with placebo of ciclosporin (n = 15) Comparator: ciclosporin 3 mg/kg PO once a day with placebo of tacrolimus (n = 15)
Outcomes	SCORAD, AEs
Funding source	Ministero Italiano Universita' e Ricerca (MIUR)
Declarations of interest	Not described
Notes	Recorded every 7 days
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No details provided
Allocation concealment (selection bias)	Unclear risk	No details provided
Blinding of participants and personnel (performance bias) All outcomes	Low risk	All treatments were administered by a person unaware of who was participating in the study
Blinding of outcome assessment (detection bias) All outcomes	Low risk	All treatments were administered by a person unaware of who was participating in the study
Incomplete outcome data (attrition bias) All outcomes	Low risk	All completed
Selective reporting (reporting bias)	Low risk	No potential sources of reporting bias
Other bias	Low risk	Baseline balanced

Paul 2002.

Study characteristics
Methods	Parallel single‐blinded RCT Dates of study: not stated Location: France
Participants	Severe atopic dermatitis Randomised: 10 participants Inclusion criteria: adults 18 to 50 years of age with a diagnosis of AD according to the criteria of Hanifin and Rajka; severe AD uncontrolled by conventional treatment with emollients, topical corticosteroids, and phototherapy; minimum disease severity score (severity scoring of AD, SCORAD) of 50. Patients had to stop immunosuppressive therapy with ciclosporin, azathioprine, oral corticosteroids, and phototherapy at least 1 month before the start of the study Exclusion criteria: pregnancy and lactation, history of severe renal or hepatic impairment, history of congenital IgA deficiency Dropouts and withdrawals: 1 participant was withdrawn before treatment
Interventions	Intervention: immediate treatment with IVIG (n = 4) Comparator: delayed treatment after 1 month with IVIG (n = 5)
Outcomes	SCORAD, pruritus and sleep loss scores, AEs
Funding source	Publique Hopitaux de Paris
Declarations of interest	Not described
Notes	Patients were assessed clinically at days 15, 30, 60, and 90. The primary efficacy criterion was measurement of the SCORAD Index at day 30
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated randomisation
Allocation concealment (selection bias)	Low risk	Computer‐generated randomisation
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Evaluator‐blinded trial
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Evaluator‐blinded trial
Incomplete outcome data (attrition bias) All outcomes	Low risk	All completed
Selective reporting (reporting bias)	Low risk	No potential sources of reporting bias
Other bias	High risk	Baseline imbalance in terms of previous treatment with topical steroid, ciclosporin, or topical tacrolimus

Price 2019.

Study characteristics
Methods	Parallel double‐blind placebo RCT Dates of study: not stated Location: not stated
Participants	Moderate to severe atopic dermatitis
Interventions	Interventions: prednisolone 0.75 mg/kg for 5 days, 0.5 mg/kg for 5 days, 0.25 mg/kg for 5 days Comparator: placebo
Outcomes	EASI, SCORAD, IGA, patient‐reported outcomes (POEM and DLQI)
Funding source	Not described
Declarations of interest	Not described
Notes	Data available in abstract form
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Insufficient information
Allocation concealment (selection bias)	Unclear risk	Insufficient information
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Insufficient information
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Insufficient information
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Insufficient information
Selective reporting (reporting bias)	Unclear risk	Insufficient information
Other bias	Unclear risk	Insufficient information

Ruzicka 2017.

Study characteristics
Methods	Parallel double‐blinded randomised controlled trial Dates of study: not stated Location: not stated
Participants	Randomised: 264 patients (1:1:1:1:1 ratio) Inclusion criteria: Patients between the ages of 18 and 65 years who had moderate to severe atopic dermatitis that was inadequately controlled by topical glucocorticoids or topical calcineurin inhibitors Patients with AD who had a score ≥ 10 on the Eczema Area and Severity Index (EASI), a score for pruritus ≥ 50 mm on a visual analogue scale, and a score ≥ 3 on the static Investigators’ Global Assessment (sIGA) Exclusion criteria: not stated Patients were excluded if they had active dermatological diseases concomitant with atopic dermatitis, or if they had received systemic therapy for atopic dermatitis or ultraviolet radiation therapy within 4 weeks before randomisation, potent or very potent topical glucocorticoids or topical calcineurin inhibitors within 2 weeks before randomisation, or mildly or moderately potent topical glucocorticoids or antihistamines (topical or systemic) within 1 week before randomisation Dropouts and withdrawals: 48 (18%)
Interventions	Intervention: nemolizumab 0.1, 0.5, 2.0 mg/kg Q4W or 2.0 mg/kg Q8W subcutaneously Comparator: placebo Q4W subcutaneously
Outcomes	Primary outcomes: percentage of improvement between baseline and week 12 in score on the Pruritus VAS, which was recorded daily by patients Secondary outcomes: efficacy outcomes at week 12 and at each time point (weeks 1, 2, 3, 4, 6, 8, and 10) improvement from baseline: in EASI score in the score on Scoring Atopic Dermatitis (SCORAD) in sIGA score in BSA affected by atopic dermatitis in the pruritus verbal rating scale score and in the sleep disturbance VAS score
Funding source	Chugai Pharmaceutical
Declarations of interest	Not described
Notes	Not described
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Centralised interactive voice or online response system
Allocation concealment (selection bias)	Low risk	Centralised interactive voice or online response system
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Double‐blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Double‐blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	< 20% incomplete
Selective reporting (reporting bias)	Low risk	No source of selective reporting
Other bias	Low risk	Baseline balance

Salek 1993.

Study characteristics
Methods	Cross‐over double ‐blinded placebo RCT Dates of study: not stated Location: multi‐centre
Participants	Randomised: 33 participants (age range 17 to 56 years) Inclusion criteria: severe, long‐standing atopic dermatitis resistant to conventional therapy Exclusion criteria: not stated Dropouts and withdrawals: no withdrawals
Interventions	Intervention: ciclosporin 5 mg/kg for 8 weeks Comparator: placebo for 8 weeks
Outcomes	Final score on EDI
Funding source	Not described
Declarations of interest	Not described
Notes	Not described
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No details provided
Allocation concealment (selection bias)	Unclear risk	No details provided
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	No details provided
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	No details provided
Incomplete outcome data (attrition bias) All outcomes	Low risk	Treatment failure
Selective reporting (reporting bias)	Low risk	No potential sources of selective outcome reporting
Other bias	Low risk	Baseline balanced

Schmitt 2010.

Study characteristics
Methods	Parallel double‐blinded RCT Dates of study: February 2007 to November 2008 Location: multi‐centre
Participants	Randomised: 38 participants (age range 18 to 55 years) Inclusion criteria: severe eczema with SCORAD score ≥ 40, DLQI score ≥ 10 Exclusion criteria: concurrent hypertension, kidney disease, osteoporosis, diabetes, liver disease, or psychiatric comorbidity Dropouts and withdrawals: withdrawals due to AE prednisolone (n = 11), ciclosporin (n = 5)
Interventions	Intervention: prednisolone 0.5 to 0.8 mg/kg, tapered to nil within 2 weeks (n = 21) Comparator: ciclosporin 2.7 to 4.0 mg/kg for 6 weeks (n = 17)
Outcomes	% change in SCORAD, SCORAD50, IGA; withdrawal due to AEs, SAEs, infection; increased LFT
Funding source	Not described
Declarations of interest	Not described
Notes	Incomplete outcome data
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Central randomisation with the use of a permutated‐block randomisation list
Allocation concealment (selection bias)	Low risk	Central randomisation (randomisation and blinding, "The allocation sequence was generated by Stata 8.0 for Windows (StataCorp., College Station, TX, USA) and stored by the clinical trials pharmacist at the Technical University Dresden")
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Participants and assessors were blinded to group assignment during data collection
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Participants and assessors were blinded to group assignment during data collection
Incomplete outcome data (attrition bias) All outcomes	Low risk	Per protocol with a high degree of withdrawal ("although 21/38 participants completed the study as per protocol, all 38 participants were included in the ITT analysis")
Selective reporting (reporting bias)	Low risk	No potential sources of selective outcome reporting
Other bias	Low risk	Baseline balanced

Schram 2011.

Study characteristics
Methods	Parallel single‐blinded RCT Dates of study: July 2009 to December 2010 Location: The Netherlands
Participants	Randomised: 43 participants (age ≥ 18 years) Inclusion criteria: atopic eczema with or without the presence of allergen‐specific IgE Exclusion criteria: pregnancy, breastfeeding, planning pregnancy until 3 months after discontinuation, history of cancer, alcohol abuse, organ transplant. chronic or recurrent infectious disease Dropouts and withdrawals: not stated
Interventions	Intervention: methotrexate initiated 10 mg/week single dose, dose escalation 2.5 to 5 mg per scheduled visit until 22.5 mg/week (n = 20) Intervention: azathioprine initiated 1.5 mg/kg/d single dose, dose escalation 0.5 mg/kg/d each visit until 2.5 mg/kg/d (n = 22)
Outcomes	SCORAD, SCORAD50, IGA, PGA, EASI, POEM, QOL (Skindex‐17), AEs
Funding source	Not described
Declarations of interest	Not described
Notes	Not described
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Concealment of allocation was achieved via a computerised programme
Allocation concealment (selection bias)	Low risk	Concealment of allocation was achieved via a computerised programme
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Clinical outcome measurements were assessed by trained efficacy assessors who were blinded to allocation
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Clinical outcome measurements were assessed by trained efficacy assessors who were blinded to allocation
Incomplete outcome data (attrition bias) All outcomes	High risk	23% dropout
Selective reporting (reporting bias)	Low risk	No potential sources of reporting bias
Other bias	Low risk	Baseline balanced

Simpson 2016.

Study characteristics
Methods	Parallel double‐blinded placebo RCT Dates of study: not stated Locations: North America, Europe, Asia
Participants	Randomised: SOLO1 671 participants, SOLO2 671 participants Inclusion criteria: moderate to severe atopic dermatitis Exclusion criteria: not stated Dropouts and withdrawals: not stated
Interventions	Intervention: dupilumab 300 mg per week for 16 weeks (SOLO1 n = 223, SOLO2 n = 239) or dupilumab 300 mg per every other week for 16 weeks (SOLO1 n = 224, SOLO2 n = 2330) Comparator: placebo (SOLO1 n = 224, SOLO2 n = 239)
Outcomes	EASI, Pruritus NRS score, SCORAD, DLQI, POEM, AEs
Funding source	Regeneron Pharmaceuticals, Inc., and Sanofi
Declarations of interest	Not described
Notes	Not described
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Central interactive voice response system
Allocation concealment (selection bias)	Low risk	Central interactive voice response system
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Matching placebo
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Matching placebo
Incomplete outcome data (attrition bias) All outcomes	Low risk	< 20% dropout
Selective reporting (reporting bias)	Low risk	No potential sources of reporting bias
Other bias	Low risk	Baseline balanced

Simpson 2018.

Study characteristics
Methods	Randomised, placebo‐controlled, double‐blind, phase 2 study Multicenter trials (62 centers) with 2‐week TCS run‐in period before the 12‐week treatment period Dates of study: not stated Locations: 62 centers; United States and Canada, Europe, and other
Participants	Randomised: 209 participants (1:1:1:1), average age: 36.15 Inclusion criteria: participants with moderate‐to‐severe AD with an inadequate response to TCS (⪰ 1‐month history within 3 months before screening) and regular emollient. Eczema Area and Severity Index (EASI) ⪰14 and investigator Global Assessment (IGA) score ⪰ 3 at screening and end of the run‐in period and AD involvement of ⪰10% of body surface area (BSA) and Pruritus Visual Analog Scale (VAS) score ⪰ 3 (measured as part of SCORing Atopic Dermatitis [SCORAD]) at screening. Aged 18 to 75 years Exclusion criteria: participants who used topical calcineurin inhibitors; recent systemic immunosuppressive therapies or phototherapy; and had evidence of other skin conditions, including T‐cell lymphoma or allergic contact dermatitis Dropouts and withdrawals: not stated
Interventions	Intervention: 1) Lebrikizumab 125 mg single dose at baseline (n = 52) 2) Lebrikizumab 250 mg single dose at baseline (n = 53) 3) Lebrikizumab 125 mg once every 4 weeks for 12 weeks (n = 51) Comparator: Placebo every 4 weeks for 12 weeks (n = 53) Topical corticosteroids twice daily was allowed in all participants
Outcomes	Measurement time: 0.25, 1, 1.5, 2, 3, 5 months Primary outcomes: Percentage of participants achieving a 50% reduction in EASI score from baseline (EASI50) at week 12 Secondary outcomes: Percentages of participants achieving EASI75, IGA score of 0 or 1, SCORAD50, Dermatology Life Quality Index and AD Impact Questionnaire (ADIQ) at week 12 Safety outcomes, including treatment‐emergent adverse events (AEs) and serious AEs, were monitored at each visit from baseline to week 20
Funding source	Genentech
Declarations of interest	Not described
Notes	Not described
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	A blocked randomisation scheme stratified by region was mentioned, but there was no sufficient information on how the randomisation and allocation concealment were performed.
Allocation concealment (selection bias)	Unclear risk	A blocked randomisation scheme stratified by region was mentioned, but there was no sufficient information on how the randomisation and allocation concealment were performed.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Placebo was prepared with the same formulation as lebrikizumab without addition of the active agent, and both formulations were identical in appearance.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Placebo was prepared with the same formulation as lebrikizumab without addition of the active agent, and both formulations were identical in appearance.
Incomplete outcome data (attrition bias) All outcomes	Low risk	All completed (it appeared that all 209 participants initially randomised were included in the analysis)
Selective reporting (reporting bias)	Low risk	No potential sources of selective outcome reporting
Other bias	Low risk	Baseline balanced

Simpson 2019a.

Study characteristics
Methods	Parallel double‐blinded placebo RCT Dates of study: not stated Location: not stated
Participants	Randomised: 191 participants. Inclusion criteria: moderate to severe atopic dermatitis Exclusion criteria: not stated Dropouts and withdrawals: not stated
Interventions	Interventions: apremilast 40 mg twice daily (n = 63), apremilast 30 mg twice daily (n = 58) Comparator: placebo (n = 64)
Outcomes	EASI, SAEs, AEs, IGA (sPGA)
Funding source	Celgene
Declarations of interest	Not described
Notes	Not described
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Permuted block method, stratified by geographical region (North America and Japan) and, within region, by baseline EASI score (20 or > 20) with an interactive web response or interactive voice response system
Allocation concealment (selection bias)	Low risk	Permuted block method, stratified by geographical region (North America and Japan) and, within region, by baseline EASI score (20 or > 20) with an interactive web response or interactive voice response system
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Triple (participant, investigator, outcomes assessor)
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Triple (participant, investigator, outcomes assessor)
Incomplete outcome data (attrition bias) All outcomes	Low risk	Lack of efficacy in the placebo and apremilast 30 mg twice‐daily arms and AEs in the apremilast 40 mg twice‐daily arm
Selective reporting (reporting bias)	Low risk	No potential sources of selective outcome reporting
Other bias	Low risk	Baseline balanced

Simpson 2019b.

Study characteristics
Methods	Parallel double‐blinded placebo RCT Dates of study: not stated Locations: Australia, Canada, Germany, Hungary, New Zealand, United States
Participants	Randomised: 113 participants (ages 18 to 75 years) Inclusion criteria: atopic dermatitis with EASI score ≥ 12, IGA score ≥ 3, SCORAD ≥ 40, total AD body surface area ≥10% Exclusion criteria: not stated Dropouts and withdrawals: withdrawal of consent (n = 6)
Interventions	Intervention: tezepelumab + TCS 280 mg (n = 56) Comparator: placebo + TCS (n = 57)
Outcomes	EASI, IGA, SCORAD, Pruritus Scale score
Funding source	MedImmune LLC
Declarations of interest	Not described
Notes	Not described
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Via central interactive voice/web response
Allocation concealment (selection bias)	Low risk	Randomised
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quadruple (participant, care provider, investigator, outcomes assessor)
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quadruple (participant, care provider, investigator, outcomes assessor)
Incomplete outcome data (attrition bias) All outcomes	Low risk	LOCF
Selective reporting (reporting bias)	Low risk	No potential sources of selective outcome reporting
Other bias	Low risk	Baseline balanced

Sowden 1991.

Study characteristics
Methods	Cross‐over double‐blinded placebo RCT Dates of study: not stated Location: not stated
Participants	Randomised: 33 participants (ages 17 to 56 years) Inclusion criteria: severe refractory atopic dermatitis Exclusion criteria: not stated Dropouts and withdrawals: no withdrawals
Interventions	Intervention: ciclosporin Comparator: placebo
Outcomes	Disease activity (SASSAD), extent of disease (% BSA), itch and sleep scores, AEs
Funding source	Not described
Declarations of interest	Not described
Notes	Disease activity, extent of disease, sleep and itch, topical steroid use, and adverse events were assessed every 2 weeks
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No details provided
Allocation concealment (selection bias)	Unclear risk	No details provided
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Double‐blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Double‐blinded
Incomplete outcome data (attrition bias) All outcomes	High risk	30% incomplete data
Selective reporting (reporting bias)	Low risk	No potential sources of reporting bias
Other bias	Low risk	Cross‐over

Thaci 2016.

Study characteristics
Methods	Parallel double‐blinded placebo RCT Dates of study: not stated Location: not stated
Participants	Randomised: 31 participants Inclusion criteria: moderate to severe atopic dermatitis Exclusion criteria: not stated Dropouts and withdrawals: not stated
Interventions	Intervention: dupilumab 300 mg per week (n = 21) Comparator: placebo (n = 10)
Outcomes	EASI, IGA, Pruritus NRS scores, SCORAD, BSA, DLQI
Funding source	Regeneron Pharmaceuticals, Inc., and Sanofi
Declarations of interest	Not described
Notes	Not described
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Central randomisation
Allocation concealment (selection bias)	Low risk	Central randomisation
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Matched placebo
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Matched placebo
Incomplete outcome data (attrition bias) All outcomes	Low risk	92% complete at primary outcome
Selective reporting (reporting bias)	Low risk	No potential sources of reporting bias
Other bias	Low risk	Baseline balanced

Thaci 2018.

Study characteristics
Methods	Paralell double‐blinded placebo RCT Dates of study: not stated Location: not stated
Participants	Randomised: 81 participants Inclusion criteria: moderate to severe atopic dermatitis. Exclusion criteria: not stated Dropouts and withdrawals: not stated
Interventions	Intervention: single ascending dose MOR106 single IV infusion up to 20 mg/kg (n = 56) Comparator: multiple ascending dose MOR106 single IV infusion every week for 4 weeks
Outcomes	EASI, SCORAD, AEs, SAEs
Funding source	Novartis
Declarations of interest	Not described
Notes	Data available in abstract form
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Insufficient information
Allocation concealment (selection bias)	Unclear risk	Insufficient information
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Insufficient information
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Insufficient information
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Insufficient information
Selective reporting (reporting bias)	Unclear risk	Insufficient information
Other bias	Unclear risk	Insufficient information

Tsianakas 2018.

Study characteristics
Methods	Parallel double‐blinded placebo RCT Dates of study: not stated Location: not stated
Participants	Randomised: 64 participants Inclusion criteria: moderate to severe atopic dermatitis with BSA ≥ 10% Exclusion criteria: not stated Dropouts and withdrawals: not stated
Interventions	Intervention: dupilumab 300 mg per week for 12 weeks (n = 32) Comparator: placebo (n = 32)
Outcomes	QoLIAD, EASI, SCORAD, 5‐D Pruritus Scale score, Pruritus NRS score, IGA
Funding source	Regeneron Pharmaceuticals, Inc., and Sanofi
Declarations of interest	Not described
Notes	Assessed every week
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Central randomisation
Allocation concealment (selection bias)	Low risk	Central randomisation
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Drug kits were coded, masking the treatment assigned
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Drug kits were coded, masking the treatment assigned
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	No details provided
Selective reporting (reporting bias)	Low risk	No potential sources of reporting bias
Other bias	High risk	Baseline imbalance in terms of age, duration of disease, SCORAD, BSA, and Pruritus NRS score

Tzaneva 2010.

Study characteristics
Methods	Randomised, observer‐blinded, cross‐over trial Dates of study: not stated Location: not stated
Participants	Randomised: 40 participants Inclusion criteria: AD participants aged ≥ 18 years with severe disease (SCORAD Score > 45) Exclusion criteria: Pregnant or lactating women, participants who have any severe systemic diseases, history of abnormal UVA sensitivity, intake of photosensitising drugs, local therapy within 2 weeks and photo(chemo)therapy or other systemic treatment within 4 weeks before study entry Dropouts and withdrawals: not stated
Interventions	Intervention: Ultraviolet (UV) A1 Control: Psoralen plus UVA (PUVA)
Outcomes	Assessment time: at baseline and after 10 and 15 treatments (at 1, 3, 6, and 12 months after cessation of treatment) Final SCORAD score, AEs
Funding source	Medical University of Vienna
Declarations of interest	Not described
Notes	‐
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Patients were allocated by means of coin tossing
Allocation concealment (selection bias)	Low risk	Patients were allocated by means of coin tossing
Blinding of participants and personnel (performance bias) All outcomes	High risk	Open‐label, observer‐blinded
Blinding of outcome assessment (detection bias) All outcomes	High risk	Open‐label, observer‐blinded
Incomplete outcome data (attrition bias) All outcomes	High risk	Per protocol with high degree of withdrawal and differential withdrawal, large number of losses to follow‐up (23/40; 57.5%); participants included in the study were analysed in the results for short‐term outcomes, and 13/40 (32.5%) participants were analysed for long‐term outcomes
Selective reporting (reporting bias)	Low risk	No potential sources of selective outcome reporting
Other bias	Low risk	Cross‐over

van Joost 1994.

Study characteristics
Methods	Parallel double‐blinded, placebo RCT Dates of study: not stated Location: multi‐centre
Participants	Severe atopic dermatitis Randomised: 46 participants Inclusion criteria: severe generalised atopic dermatitis Exclusion criteria: pregnant or lactating women, history of abnormal UVA sensitivity Dropouts and withdrawals: 4 from ciclosporin, 14 from placebo withdrew because of lack of response
Interventions	Intervention: ciclosporin 5 mg/kg/d for 6 weeks (n = 23) Comparator: placebo (n = 23)
Outcomes	Severity of disease (SASSAD), extent of disease (BSA), itch and loss of sleep scores
Funding source	Not described
Declarations of interest	Not described
Notes	Not described
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No information on randomisation
Allocation concealment (selection bias)	Unclear risk	No information on randomisation
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Use of placebo suggested that at least patients were blinded
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	No information on whether the other personnel involved in the study were blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	41/46 (89.1%) participants were included in the analysis.
Selective reporting (reporting bias)	Low risk	No potential sources of selective outcome reporting
Other bias	Low risk	Baseline balanced

Wahlgren 1990.

Study characteristics
Methods	Cross‐over double‐blinded RCT Dates of study: not stated Location: not stated
Participants	Moderate to severe atopic dermatitis Randomised: 10 participants Inclusion criteria: moderate to severe atopic dermatitis Exclusion criteria: pregnant or lactating women, fertile women, UV therapy during previous month Dropouts and withdrawals: not stated
Interventions	Intervention: ciclosporin solution 100 mL dose 5 mg/kg/d Comparator: placebo
Outcomes	Number of patients free of itch, final itch score, AEs
Funding source	Karolinska Institutet, Swedish Medical Research Council
Declarations of interest	Not described
Notes	Days 2 to 5, days 9 to 10
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No randomisation details provided
Allocation concealment (selection bias)	Unclear risk	No details provided
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	No details provided
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	No details provided
Incomplete outcome data (attrition bias) All outcomes	Low risk	All completed (all 10 patients completed the study)
Selective reporting (reporting bias)	Low risk	No potential sources of selective outcome reporting
Other bias	Low risk	Cross‐over

Wollenberg 2019.

Study characteristics
Methods	Parallel double‐blinded placebo RCT Dates of study: not stated Locations: Australia, Canada, Germany, Japan, Poland, United States
Participants	Randomised: 204 participants (ages 18 to 75 years) Inclusion criteria: atopic dermatitis with BSA ≥ 10%, EASI ≥ 12, SCORAD ≥ 25 Exclusion criteria: not stated Dropouts and withdrawals: not stated
Interventions	Interventions: tralokinumab 45 mg (n = 50), 150 mg (n = 51), 300 mg (n = 52) Comparator: placebo (n = 51)
Outcomes	EASI, IGA, SCORAD, Pruritus NRS score, Dermatology Life Quality Index, EASI50, EASI75, SCORAD50, AEs
Funding source	MedImmune LLC
Declarations of interest	Not described
Notes	Not described
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Interactive web and voice response system
Allocation concealment (selection bias)	Low risk	Interactive web and voice response system
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Double‐blind study in which patients, investigators, and sponsor staff were blinded
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Double‐blind study in which patients, investigators, and sponsor staff were blinded
Incomplete outcome data (attrition bias) All outcomes	High risk	31% dropout (even ITT)
Selective reporting (reporting bias)	Low risk	No potential sources of reporting bias
Other bias	Low risk	Baseline balanced

Zonneveld 1996.

Study characteristics
Methods	Parallel open‐label RCT Dates of study: not stated Location: not stated
Participants	Randomised: 78 participants (ages 18 to 70 years) Inclusion criteria: AD patients aged 18 to 70 years with severe disease, long‐standing AD resistant to conventional therapy and/or causing the patient significant suffering and disability Exclusion criteria: not stated Dropouts and withdrawals: not stated
Interventions	Intervention: ciclosporin with starting dose at 5 mg/kg/d for 14 days; if clinical response was satisfied, the dose was lowered to 4 mg/kg/d for the next 14 days; After that if clinical response was favourable, dose was lowered to 3 mg/kg/d. Patients who showed no satisfactory response to 5 mg/kg per day at Day 14 were kept on that dose for a further 14 days. Treatment was stopped in patients who did not show a favourable response to 5 mg/kg per day after 28 days. Comparator: ciclosporin with starting dose at 3 mg/kg/d for 14 days; if clinical response was favourable, the patient continued on that dose. If no response, the dose was increased to 4 mg/kg per day (dose of ciclosporin was increased every 2 weeks if no response) Concomitant medication: Treatment with systemic corticosteroids, cytotoxic agents or photo (chemo)therapy (PUVA/UVB) was stopped at least 2 weeks before entry into the study and was not allowed during the study. Systemic antihistamines, topical emollient therapy and antibiotics were permitted. Use of topical corticosteroids was permitted if necessary.
Outcomes	Assessment time: at baseline, every 2 weeks for 2 months, then every month for 10 months % BSA, SASSAD, itch and sleep loss scores, AEs
Funding source	Sandoz Pharmaceuticals
Declarations of interest	Not described
Notes	2‐month dose‐finding period for optimal dosage of CsA for each patient. After this, patients from both groups were treated with their optimal dosage for a further 10 months. Patients were followed up for a minimum of 2 months after ciclosporin therapy was discontinued.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No information on randomisation
Allocation concealment (selection bias)	Unclear risk	No information on randomisation
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Open‐label study
Blinding of outcome assessment (detection bias) All outcomes	High risk	Open‐label study with patient outcome assessment
Incomplete outcome data (attrition bias) All outcomes	High risk	30/78 patients (38.5%) completed the study and were included in the analysis
Selective reporting (reporting bias)	Low risk	No potential sources of selective outcome reporting
Other bias	Low risk	Baseline balanced

Zurbriggen 1999.

Study characteristics
Methods	Cross‐over double‐blinded pilot RCT Dates of study: not stated Location: not stated
Participants	Randomised: 14 participants Inclusion criteria: Patients with severe atopic dermatitis, atopic skin diathesis Exclusion criteria: body weight < 40 kg, hypertension (> 160/95 mmHg), history or presence of malignancy Dropouts and withdrawals: not stated
Interventions	Each formulation was administered for 8 weeks, followed by switching to the other treatment group for another 8 weeks. Intervention: New formulation of ciclosporin (microemulsion of CsA with improved pharmacokinetic properties) Comparator: Standard formulation of ciclosporin (old formulation)
Outcomes	Disease activity (SASSAD), extent of disease (% BSA)
Funding source	Novartis Pharma Schweiz AG
Declarations of interest	Not described
Notes	‐
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation was performed via a computer‐generated random number sequence (Patients and methods, paragraph 2)
Allocation concealment (selection bias)	Unclear risk	No information in the paper to enable a meaningful assessment of independence between sequence generation and allocation
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Blinded to patients and investigators
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Insufficient information
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Insufficient information
Selective reporting (reporting bias)	Low risk	No evidence of selective outcome reporting
Other bias	Unclear risk	Insufficient information

AD: atopic dermatitis; ADIS: Atopic Dermatitis Itch Scale; ADR: adverse drug reaction; AE: adverse event; BMI: body mass index; BSA: body surface area; CAP: community‐acquired pneumonia; CsA: cyclosporin A; DLQI: Dermatology Life Quality Index; EASI: Eczema Area and Severity Index; EASI50: 50% improvement in Eczema Area and Severity Index; EASI75: 75% improvement in Eczema Area and Severity Index; ECP: eosinophil cationic protein; EDI: Eczema Disability Index; EoS: end of study; FcERI: high‐affinity receptor for IgE; GBR: abbreviated name of new medicine (GBR830) for dermatitis which is an anti‐OX40; GISS: Global Individual Sign Score; HADS: Hospital Anxiety and Depression Scale; HBcAb: hepatitis B core antibody; HBsAg: hepatitis B surface antigen; IFN: interferon; IGA: Investigators' Global Assessment; IgE: immunoglobulin E; IgG: immunoglobulin G; IGRA: interferon gamma release assay; IL: interleukin; IPSA: Investigators' Pruritus Severity Assessment; ITT: intention‐to‐treat; IVIG: intravenous immunoglobulin; LFT: liver function test; LOCF: last observation carried forward; MPA: methylprednisolone aceponate; NRS: Numerical Rating Scale; nSAE: non‐serious adverse event; NSAID: non‐steroidal anti‐inflammatory drug; OLE: open‐label extension; PGA: Physicians’ Global Assessment; POEM: Patient‐Oriented Eczema Measure; PPD: purified protein derivative; PUVA: ultraviolet light therapy for skin disease; QFT: QuantiFERON TB‐Gold test; QOL: quality of life; QoLIAD: Quality of Life Index for Atopic Dermatitis; RCT: randomised controlled trial; ROB: risk of bias; SAE: serious adverse event; SASSAD: Six Area, Six Sign Atopic Dermatitis; SBA: serum bactericidal assay; SC: subcutaneous; SCORAD: severity SCORing of Atopic Dermatitis; SCORAD50: 50% improvement in SCORAD; sIGA: static Investigators’ Global Assessment; sPGA: static Physicians’ Global Assessment; TB: tuberculosis; TBSA: total body surface area; TCI: topical calcineurin inhibitor; TCS: topical corticosteroid; Tdap: tetanus component of tetanus, diphtheria, and pertussis vaccine; TEAE: treatment‐emergent adverse event; UVA: ultraviolet A; UVAB: combined ultraviolet A and ultraviolet B; UVB: ultraviolet B; VAS: visual analogue scale.

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Arm 2014	Not related to moderate to severe AD (Mixed type of eczema)
Atakan 1998	Non‐RCT
Belloni 2008	Review
Buckley 1998	Review
Bunikowski 2001	Non‐RCT
Dickey 1976	Not related to moderate to severe AD (mixed type of eczema)
Hamilton 2014b	Non‐RCT
Jolles 2003	Non‐RCT
Kondo 1994	Non‐RCT
Litzlbauer 2014	Animal study
NCT00914186	Topical product
NCT01949311	Non‐RCT
NCT02324972	Not related to moderate to severe AD (included mild to moderate AD)
NCT02595073	Topical product
Patel 2012	Review
Purohit 2019	Topical product
Samrao 2012	Non‐RCT
Schneider 1998	Non‐RCT
Wolff 2005	Not intervention of interest
Zheng 2019	Review

AD: atopic dermatitis; RCT: randomised controlled trial.

Characteristics of ongoing studies [ordered by study ID]

Irvine 2018.

Study name	A Randomised Controlled Trial Protocol Assessing the Effectiveness, Safety, and Cost‐effectiveness of Methotrexate versus Ciclosporin in the Treatment of Severe Atopic Eczema in Children
Methods	Multi‐centre, parallel‐group, assessor‐blind, pragmatic RCT of 36 weeks' duration with a 24‐week follow‐up period
Participants	102 children ages 2 to 16 years with moderate to severe atopic eczema, unresponsive to topical treatment, will be randomised (1:1)
Interventions	Methotrexate (MTX; 0.4 mg/kg/week) or ciclosporin (CyA; 4 mg/kg/d)
Outcomes	Primary outcomes: Change from baseline to 12 weeks in Objective Severity Scoring of Atopic Dermatitis (oSCORAD) and time to first significant flare following treatment cessation
Starting date	29/06/2017
Contact information	Guy’s and St Thomas’ NHS Foundation Trust and King’s College London Telephone: 020 7188 7188, extension 51161 Email: carsten.flohr@kcl.ac.uk
Notes	‐

NCT00232076.

Study name	Verification Study of Ciclosporin for Atopic Dermatitis
Methods	Allocation: randomised Intervention model: single group assignment Masking: double Primary purpose: treatment
Participants	Severe atopic dermatitis
Interventions	Drug: ciclosporin
Outcomes	Not stated
Starting date	May 2004
Contact information	Novartis Pharmaceuticals, Novartis
Notes	Completed with No Results Available

NCT00376129.

Study name	Open‐Label, Single‐Center Study of Alefacept in Patients With Atopic Dermatitis
Methods	Allocation: randomised Intervention model: single group assignment Masking: none (open‐label) Primary purpose: diagnostic
Participants	Atopic dermatitis
Interventions	Drug: alefacept
Outcomes	Primary endpoints: Change in EASI at Visit 13 compared to baseline via paired t‐test Additional endpoints: Percentage of patients reaching PGA of "clear" or "almost clear" and/or a reduction in EASI of ≥ 50 or ≥ 75% compared to baseline at any visit after baseline Percentage of patients reaching a pruritus score of none or mild Several immunological endpoints
Starting date	January 2006
Contact information	University Hospital Inselspital, Berne|Biogen‐Dompic AG
Notes	Completed with No Results Available

NCT02211417.

Study name	Oral DS107G in Moderate to Severe Atopic Dermatitis
Methods	Allocation: randomised Intervention model: parallel assignment Masking: quadruple (participant, care provider, investigator, outcomes assessor) Primary purpose: treatment
Participants	Atopic dermatitis
Interventions	Drug: DS107G Comparator: placebo
Outcomes	IGA (Investigators' Global Assessment) EASI (Eczema Area and Severity Index) POEM (Patient‐Oriented Eczema Measure) DLQI (Dermatology Life Quality Index) SCORAD (Score Atopic Dermatitis) VAS (visual analogue scale) pruritis score BSA (body surface area) affected TEAE (treatment‐emergent adverse event)
Starting date	January 2015
Contact information	Dignity Sciences, Ltd.
Notes	Completed with No Results Available

NCT02226068.

Study name	Photophoresis versus Ciclosporine in Severe Atopic Dermatitis
Methods	Allocation: randomised Intervention model: cross‐over assignment Primary purpose: treatment
Participants	Atopic dermatitis
Interventions	Drug: cyclosporin A (CsA) Other: extracorporeal photopheresis (ECP)
Outcomes	Change in SCORAD (SCORing Atopic Dermatitis)
Starting date	June 2002
Contact information	University of Aarhus
Notes	Completed with No Results Available

NCT02864498.

Study name	Efficacy and Safety Study of Orally Administered DS107 in Moderate to Severe Atopic Dermatitis Patients
Methods	Allocation: randomised Intervention model: parallel assignment Masking: double (participant, investigator) Primary purpose: treatment
Participants	Atopic dermatitis
Interventions	Drug: DS107 Comparator: placebo
Outcomes	IGA of 0 (clear) or 1 (almost clear) and decrease in IGA ≥ 2 points Proportion of patients achieving an IGA score of 0 (clear) or 1 (almost clear) and a decrease in IGA ≥ 2 points Change from baseline in Eczema Area and Severity Index (EASI) Change from baseline in Numerical Rating Scale (NRS) for Pruritus Proportion of patients achieving a decrease in IGA ≥ 2 points
Starting date	January 2017
Contact information	DS Biopharma
Notes	Completed with No Results Available

NCT02908698.

Study name	Effect of Oral Steroids on Skin Outcomes in Atopic Dermatitis
Methods	Allocation: randomised Intervention model: parallel assignment Masking: quadruple (participant, care provider, investigator, outcomes assessor) Primary purpose: basic science
Participants	Atopic dermatitis
Interventions	Drug: prednisone Comparator: placebo control
Outcomes	Late cutaneous response (LCR) Comparison of eosinophils and basophils in LCR between drug and placebo using histopathology Comparison of eosinophils and basophils in LCR between drug and placebo using flow cytometry
Starting date	2017 January 24
Contact information	McMaster University
Notes	Terminated with No Results Available

NCT03050151.

Study name	Study of Dupilumab Auto‐injector Device When Used by Patients With Atopic Dermatitis
Methods	Allocation: randomised Intervention model: parallel assignment Masking: none (open‐label) Primary purpose: treatment
Participants	Atopic disorders Eczema, atopic
Interventions	Drug: dupilumab Device: auto‐injector device Device: pre‐filled syringe
Outcomes	Number of validated AI device‐associated product technical failures (PTFs) during the treatment period divided by total number of actual injections Type of validated AI device‐associated PTFs during the treatment period divided by total number of actual injections Number of patients with an AI device‐associated PTF Percentage of patients with an AI device‐associated PTF Number of AI device‐associated product technical complaints (PTCs) divided by total number of actual injections Type of AI device‐associated PTCs divided by total number of actual injections|Number of patients with an AI device‐associated PTC Percentage of patients with an AI device‐associated PTC Number of AI device‐associated failed drug deliveries (defined as patient failure to administer the full dose at a given attempt, excluding PTF) divided by total number of actual injections Type of AI device‐associated failed drug deliveries (defined as patient failure to administer the full dose at a given attempt, excluding PTF) divided by total number of actual injections Number of patients with AI device‐associated failure to deliver dose Percentage of patients with AI device‐associated failure to deliver dose Number of patients with response to patient satisfaction questions with the AI device Percentage of patients with response to patient satisfaction questions with the AI device
Starting date	2017 February 28
Contact information	Regeneron Pharmaceuticals, Inc. Sanofi
Notes	Completed with No Results Available

NCT03100344.

Study name	Randomized, Double‐blind, Multi‐center, Parallel‐group, Placebo‐controlled Dose‐ranging Study to Assess the Efficacy and Safety of Nemolizumab in Moderate‐to‐severe Atopic Dermatitis Subjects With Severe Pruritus Receiving Topical Corticosteroids (TCS)
Methods	Randomised, double‐blind, multi‐centre, parallel‐group, placebo‐controlled trial Primary purpose: treatment
Participants	Moderate to severe AD with severe pruritus receiving TCS, whose condition was not adequately controlled with topical treatments
Interventions	Drug: nemolizumab Drug: placebo
Outcomes	Primary outcome: percent change from baseline in Eczema Area and Severity Index (EASI) at week 24 (Time frame: from baseline to week 24) Secondary outcomes: Number of participants achieving Pruritus Categorical Scale (PCS) success (defined as a weekly prorated rounded average PCS ≤ 1 (none to mild)) at week 24 [Time frame: week 24] Number of participants with an improvement of weekly average peak Pruritus Numerical Rating Scale (NRS) ≥ 4 at each time point up to week 24 [Time frame: from week 1 to week 24] Percent change from baseline in SCORing Atopic Dermatitis (SCORAD) at week 24 [Time frame: baseline, week 24] Absolute change from baseline in SCORing Atopic Dermatitis (SCORAD) at week 24 [Time frame: baseline, week 24] Percent change from baseline in weekly average sleep disturbance Numerical Rating Scale (NRS) at week 24 [Time frame: baseline, week 24] Absolute change from baseline in weekly average sleep disturbance Numerical Rating Scale (NRS) at week 24 [Time frame: baseline, week 24] Number of participants achieving Investigators' Global Assessment (IGA) success (defined as IGA 0 (clear) or 1 (almost clear)) at each time point up to week 24 [Time frame: from week 1 to week 24] Number of participants with Eczema Area and Severity Index (EASI)50 (defined as achieving 50% reduction from baseline in EASI score) at each visit up to week 24 [Time frame: from week 1 to week 24] Number of participants with Eczema Area and Severity Index (EASI)75 (defined as achieving 75% reduction from baseline in EASI score) at each visit up to week 24 [Time frame: from week 1 to week 24] Number of participants with Eczema Area and Severity Index (EASI)90 (defined as achieving 90% reduction from baseline in EASI score) at each visit up to week 24 [Time frame: from week 1 to week 24] Number of participants achieving Investigators' Global Assessment (IGA) success (defined as IGA 0 (clear) or 1 (almost clear)) and a reduction ≥ 2 points at each visit up to week 24 [Time frame: week 1 to week 24] Percentage change from baseline in Eczema Area and Severity Index (EASI) at each visit up to week 24 [Time frame: from baseline to week 24] Percentage change from baseline in weekly average of the peak Pruritus Numerical Rating Scale (NRS) at each visit up to week 24 [Time frame: at baseline and week 24] Number of participants with adverse events [Time frame: from screening to follow‐up visit (week 32)] Absolute change from baseline in weekly average of the peak pruritus Numerical Rating Scale (NRS) at each visit up to week 24 [Time frame: baseline to week 24]. Pruritus NRS is a scale to be used by participants to report the intensity of their pruritus (itch) during the last 24 hours. For maximum itch intensity, scores were provided on a scale of 0 to 10, with 0 being 'no itch' and 10 being 'worst itch imaginable'. Higher scores indicate worse outcome Absolute change from baseline in weekly average of the average Pruritus Numerical Rating Scale (NRS) at each visit up to week 24 [Time frame: baseline to week 24] Percentage change from baseline in weekly average of the average Pruritus Numerical Rating Scale (NRS) at each visit up to week 24 [Time frame: baseline to week 24]
Starting date	2017 June 14
Contact information	Galderma R&D
Notes	Published after the last search of this review: Silverberg JI, Pinter A, Pulka G, Poulin Y, Bouaziz JD, Wollenberg A, Murrell DF, Alexis A, Lindsey L, Ahmad F, Piketty C, Clucas A. Phase 2B randomized study of nemolizumab in adults with moderate‐to‐severe atopic dermatitis and severe pruritus. J Allergy Clin Immunol 2020;145(1):173‐182. doi: 10.1016/j.jaci.2019.08.013. Epub 2019 Aug 23

NCT03131648.

Study name	Tralokinumab Monotherapy for Moderate to Severe Atopic Dermatitis ‐ ECZTRA 1 (ECZema TRAlokinumab Trial no. 1)
Methods	Allocation: randomised Intervention model: parallel assignment Masking: double (participant, investigator) Primary purpose: treatment
Participants	Atopic dermatitis
Interventions	Intervention: tralokinumab Comparator: placebo
Outcomes	Subjects with Investigators' Global Assessment (IGA) score of 0 (clear) or 1 (almost clear) at week 16 Subjects achieving ≥ 75% reduction in Eczema Area and Severity Index (EASI) Change in Scoring Atopic Dermatitis (SCORAD) from baseline to week 16 Reduction in worst daily Pruritus Numerical Rating Scale (weekly average) ≥ 4 from baseline to week 16 Change in Dermatology Life Quality Index (DLQI) score from baseline to week 16
Starting date	2017 May 30
Contact information	Not stated
Notes	RCT protocol ‐ no results

NCT03160885.

Study name	Tralokinumab Monotherapy for Moderate to Severe Atopic Dermatitis ‐ ECZTRA 2 (ECZema TRAlokinumab Trial no. 2)
Methods	Allocation: randomised Intervention model: parallel assignment Masking: double (participant, investigator) Primary purpose: treatment
Participants	Atopic dermatitis
Interventions	Drug: tralokinumab Comparator: placebo
Outcomes	• Subjects with Investigators' Global Assessment (IGA) score of 0 (clear) or 1 (almost clear) at week 16 • Subjects achieving ≥ 75% reduction in Eczema Area and Severity Index (EASI) • Change in Scoring Atopic Dermatitis (SCORAD) from baseline to week 16 • Reduction in worst daily Pruritus Numerical Rating Scale (weekly average) ≥ 4 from baseline to week 16 • Change in Dermatology Life Quality Index (DLQI) score from baseline to week 16
Starting date	2017 June 12
Contact information	LEO Pharma
Notes	Completed with No Results Available

NCT03334435.

Study name	A Phase 3 Multicenter, Double‐Blind Study to Evaluate the Long‐Term Safety and Efficacy of Baricitinib in Adult Patients With Atopic Dermatitis(I4V‐MC‐JAHN)
Methods	Allocation: randomised Intervention model: parallel assignment Masking: double (participant, investigator) Primary purpose: treatment
Participants	Atopic dermatitis
Interventions	Drug: baricitinib Comparator: placebo
Outcomes	Proportion of participants with a response on Investigators' Global Assessment (IGA) of 0 or 1 at 16 weeks Proportion of participants with a response on IGA of 0 or 1 at 36 weeks Proportion of participants with a response on IGA of 0 or 1 at 52 weeks Proportion of participants achieving IGA 0, 1, or 2 Proportion of participants achieving IGA 0 or 1 (non‐responders) Proportion of participants achieving response on Eczema Area and Severity Index (EASI)75 from baseline of originating study Proportion of participants with a 4‐point improvement from baseline of originating study in Itch NRS
Starting date	2018 March 28
Contact information	Eli Lilly and Company clinicaltrials.gov@lilly.com
Notes	‐

NCT03428100.

Study name	A Phase 3, Multicenter, Double‐Blind, Randomized, Placebo‐Controlled Study Evaluating the Safety and Efficacy of Baricitinib in Combination With Topical Corticosteroids in Adult Patients With Moderate‐to‐Severe Atopic Dermatitis Who Have Experienced Failure to Cyclosporine or Are Intolerant to, or Have Contraindication to, Cyclosporine (I4V‐MC‐JAIN)
Methods	Allocation: randomised Intervention model: parallel assignment Masking: double (participant, investigator) Primary purpose: treatment
Participants	Atopic dermatitis
Interventions	Drug: baricitinib Drug: placebo Drug: topical corticosteroid
Outcomes	Proportion of participants achieving Investigators' Global Assessment (IGA) of 0 or 1 with 2‐point Improvement (high or mid dose) Proportion of participants achieving IGA of 0 or 1 with 2‐point improvement (low dose) Proportion of participants achieving Eczema Area and Severity Index 75 (EASI75) Proportion of participants achieving EASI90 Percent change from baseline in EASI score Proportion of participants achieving 75% improvement in SCORing Atopic Dermatitis (SCORAD75) Proportion of participants achieving 4‐point improvement on Itch Numerical Rating Scale (NRS) Change from baseline in the score of item 2 of the Atopic Dermatitis Sleep Scale (ADSS) Change from baseline in skin pain NRS Proportion of participants achieving EASI50 Proportion of participants achieving IGA of 0 Change from baseline in SCORAD|proportion of participants achieving SCORAD90 Change from baseline in body surface area (BSA) affected Proportion of participants developing skin infections requiring antibiotic treatment Mean number of days without topical corticosteroid (TCS) use Mean gram quantity of TCS use (tube weights)|percent change from baseline on Itch NRS Change from baseline in total score of the Patient‐Oriented Eczema Measure (POEM) Change from baseline in Patient Global Assessment of Severity ‐ Atopic Dermatitis (PGI‐S‐AD) score Change from baseline on the Hospital Anxiety Depression Scale (HADS) Change from baseline in the Dermatology Life Quality Index (DLQI) Change from baseline on the Work Productivity and Activity Impairment ‐ Atopic Dermatitis (WPAI‐AD) questionnaire Change from baseline in the European Quality of Life 5 Dimensions 5 Levels (EQ‐5D‐5L) score
Starting date	2018 May 15
Contact information	Eli Lilly and Company
Notes	‐

NCT03443024.

Study name	A Study of Lebrikizumab in Patients With Moderate‐to‐Severe Atopic Dermatitis
Methods	Allocation: randomised Intervention model: parallel assignment Masking: triple (participant, investigator, outcomes assessor) Primary purpose: treatment
Participants	Atopic dermatitis
Interventions	Biological: lebrikizumab Drug: placebo
Outcomes	Percent change in Eczema Area and Severity Index (EASI) from baseline to week 16 Proportion of patients with 75% improvement from baseline in EASI (EASI75) at week 16 Proportion of patients with IGA score of 0 (clear) or 1 (almost clear) and a reduction of 2 points from baseline to week 16 (5‐point scale) Proportion of patients with EASI < 7 at week 16 Proportion of patients achieving EASI50 and EASI90 at week 16 Percent change in the sleep loss Numerical Rating Scale (NRS) score from baseline to week 16 Percent change in Pruritus NRS score from baseline to week 16 Proportion of patients with Pruritus NRS change of 3 from baseline to week 16 Proportion of patients with Pruritus NRS change of 4 from baseline to week 16 Change in body surface area (BSA) involved with AD from baseline to week 16 Change from baseline in Atopic Dermatitis Impact Questionnaire (ADIQ) score
Starting date	2018 January 30
Contact information	Dermira, Inc.
Notes	Completed with No Results Available

NCT03526861.

Study name	Tralokinumab Monotherapy for Adolescent Subjects With Moderate to Severe Atopic Dermatitis ‐ ECZTRA 6 (ECZema TRAlokinumab Trial no. 6)
Methods	Allocation: randomised Intervention model: parallel assignment Masking: double (participant, investigator) Primary purpose: treatment
Participants	Atopic dermatitis
Interventions	Drug: tralokinumab Drug: placebo
Outcomes	Investigators' Global Assessment (IGA) score of 0 (clear) or 1 (almost clear) at week 16 ≥ 75% reduction in Eczema Area and Severity Index (EASI75) at week 16 Change in Scoring Atopic Dermatitis (SCORAD) from baseline to week 16 Reduction of adolescent Pruritus Numerical Rating Scale (NRS) (weekly average) ≥ 4 from baseline to week 16 Change in Children's Dermatology Life Quality Index (CDLQI) score from baseline to week 16 Number of adverse events Occurrence of anti‐drug antibodies (yes/no)
Starting date	2018 June 19
Contact information	LEO Pharma
Notes	‐

NCT03533751.

Study name	Efficacy, Safety, and Pharmacokinetic Profile of ANB020 in Adults With Moderate to Severe Atopic Dermatitis
Methods	Allocation: randomised Intervention model: parallel assignment Masking: quadruple (participant, care provider, investigator, outcomes assessor) Primary purpose: treatment
Participants	Atopic dermatitis
Interventions	Drug: etokimab (ANB020) Comparator: placebo
Outcomes	Percent change in Eczema Area and Severity Index (EASI) Safety of ANB020 in adult patients with moderate to severe atopic dermatitis Proportion of subjects who achieve Investigators' Global Assessment (IGA) response of 0 (clear) or 1 (almost clear) Proportion of subjects who achieve IGA score reduction of 2 Percent change in peak weekly averaged Numerical Rating Scale (NRS) for pruritus score from baseline Proportion of subjects who achieve weekly averaged NRS score reduction from baseline of 3 Proportion of subjects with EASI50 (50% improvement from baseline) Proportion of subjects with EASI75 (75% improvement from baseline) Percent change in EASI score from baseline to clinical assessment time points prior to week 16 Absolute change in EASI score from baseline Percent change in percent body surface area (BSA) from baseline Absolute change in percent BSA from baseline Percent change in Scoring Atopic Dermatitis (SCORAD) score from baseline Absolute change in SCORAD scores from baseline
Starting date	2018 May 23
Contact information	AnaptysBio, Inc.
Notes	‐

NCT03559270.

Study name	Study of Baricitinib in Adult Patients With Atopic Dermatitis
Methods	Intervention model: single group assignment Masking: none (open‐label) Primary purpose: treatment
Participants	Atopic dermatitis
Interventions	Drug: Baricitinib Comparator: Not stated
Outcomes	Proportion of participants achieving Investigators' Global Assessment (IGA) of 0 or 1 Proportion of participants achieving Eczema Area and Severity Index (EASI75) Proportion of participants achieving a body surface area of 3% Proportion of participants achieving 4‐point improvement in Itch Numerical Rating Scale (NRS)
Starting date	2018 June 27
Contact information	Eli Lilly and Company Incyte Corporation
Notes	‐

NCT03568071.

Study name	A Study to Assess Efficacy, Safety, Tolerability and Pharmacokinetics (PK)/Pharmacodynamics (PD) of MOR106 in Subjects With Moderate to Severe Atopic Dermatitis
Methods	Allocation: randomised Intervention model: sequential assignment Masking: quadruple (participant, care provider, investigator, outcomes assessor) Primary purpose: treatment
Participants	Atopic dermatitis
Interventions	Drug: MOR 106 Comparator: placebo
Outcomes	Percent change in Eczema Area and Severity Index (EASI) score Proportion of subjects who achieve 50% overall improvement in Eczema Area and Severity Index (EASI) score Proportion of subjects who achieve an Investigators' Global Assessment (IGA) score of 0 or 1 Proportion of subjects who achieve Investigators' Global Assessment (IGA) score reduction of 2 Percent change in Scoring Atopic Dermatitis (SCORAD) score Number of incidents of treatment‐emergent adverse events (TEAEs), adverse events of special interest (AESIs), serious adverse events (SAEs), and discontinuations due to adverse events (AEs) Characterisation of the MOR106 immunogenetic profile MOR106 (AUC0‐inf)
Starting date	2018 April 26
Contact information	Galapagos NV
Notes	‐

NCT03568136.

Study name	Investigation of Efficacy of Secukinumab in Patients With Moderate to Serve Atopic Dermatitis
Methods	Allocation: randomised Intervention model: parallel assignment Masking: double (participant, investigator) Primary purpose: treatment
Participants	Dermatitis, atopic Eczema, atopic Neurodermatitis, atopic
Interventions	Drug: secukinumab 300 mg Drug: placebo
Outcomes	Reduction in EASI Reduction in EASI Reduction in SCORAD (Scoring Atopic Dermatitis) Change in pruritus score (visual analogue scale) Change in IGA score (5‐point Investigators' Global Assessment) Serum biomarkers CCL17 and CCL22 Increase in DLQI (Dermatology Life Quality Index) Consumption of topical methylprednisolone aceponate Serious and non‐serious adverse drug reactions Gender distribution in patients with atopic dermatitis
Starting date	2018 September 18
Contact information	GWT‐TUD GmbH Novartis Pharmaceuticals
Notes	‐

NCT03703102.

Study name	Study of an Anti‐OX40 Monoclonal Antibody (KHK4083) in Subjects With Moderate to Severe Atopic Dermatitis
Methods	Allocation: randomised Intervention model: parallel assignment Masking: quadruple (participant, care provider, investigator, outcomes assessor) Primary purpose: treatment
Participants	Atopic dermatitis
Interventions	Drug: KHK4083 Drug: placebo
Outcomes	Percent change from baseline to week 16 in Eczema Area and Severity Index (EASI) score
Starting date	2018 October 22
Contact information	Kyowa Kirin Pharmaceutical Development, Inc. Kyowa Kirin Co., Ltd.
Notes	‐

NCT03720470.

Study name	Study Evaluating Efficacy and Safety of PF‐04965842 and Dupilumab in Adult Subjects With Moderate to Severe Atopic Dermatitis on Background Topical Therapy
Methods	Allocation: randomised Intervention model: parallel assignment Masking: quadruple (participant, care provider, investigator, outcomes assessor) Primary purpose: treatment
Participants	Dermatitis Dermatitis, atopic Eczema Skin diseases Skin diseases, genetic Genetic diseases, Inborn Skin diseases, eczematous Hypersensitivity Hypersensitivity, immediate Immune system diseases
Interventions	Drug: PF‐04965842 100 mg Drug: PF‐04965842 200 mg Drug: dupilumab Drug: oral placebo Drug: injectable placebo
Outcomes	Investigators' Global Assessment (IGA) Eczema Area and Severity Index (EASI) Pruritus Numerical Rating Scale (NRS) Body surface area (BSA) Patient Global Assessment (PtGA) Dermatology Life Quality Index (DLQI) EuroQol Quality of Life 5 Dimensions 5 Levels Scale (EQ‐5D‐5L) Hospital Anxiety and Depression Scale (HADS) Patient‐Oriented Eczema Measure (POEM) Pruritus and Symptoms Assessment for Atopic Dermatitis (PSAAD) Scoring Atopic Dermatitis (SCORAD) Steroid‐free days
Starting date	2018 October 29
Contact information	Pfizer, Inc.
Notes	‐

NCT03736967.

Study name	2a Study to Assess the Efficacy and Safety of REGN3500 Monotherapy and Combination of REGN3500 Plus Dupilumab in Adult Patients With Moderate‐to‐Severe Atopic Dermatitis
Methods	Allocation: randomised Intervention model: parallel assignment Masking: quadruple (participant, care provider, investigator, outcomes assessor) Primary purpose: treatment
Participants	Atopic dermatitis
Interventions	Drug: REGN3500 Drug: dupilumab Drug: REGN3500 + dupilumab combo Drug: placebo
Outcomes	Percent change in Eczema Area and Severity Index (EASI) score Proportion of patients achieving EASI50 Proportion of patients achieving EASI75 Proportion of patients achieving EASI90 Absolute change in EASI score Proportion of patients with an Investigators' Global Assessment (IGA) score of 0 or 1 (on a 5‐point scale) Proportion of patients with an IGA score reduction of 2 points Absolute change in weekly average of daily peak Pruritus Numerical Rating Scale (NRS) Percent change in weekly average of daily peak Pruritus Numerical Rating Scale (NRS) Proportion of patients with improvement (reduction) in weekly average of daily peak Pruritus NRS 4 Time to onset of effect on pruritus Percent change in SCORing Atopic Dermatitis (SCORAD) Change in percent body surface area (BSA) of AD involvement Incidence of treatment‐emergent adverse events (TEAEs) from baseline through end of week 16 Incidence of treatment‐emergent SAEs from baseline through end of treatment Incidence of treatment‐emergent adverse events of special interest (AESIs) from baseline through end of treatment Incidence of TEAEs from baseline through end of study Incidence of treatment‐emergent SAEs from baseline through end of study Incidence of treatment‐emergent AESIs from baseline through end of study Serum REGN 3500 concentration Serum dupilumab concentration Anti‐REGN3500 antibody Anti‐dupilumab antibody
Starting date	2018 November 12
Contact information	Regeneron Pharmaceuticals, Inc. Sanofi
Notes	‐

NCT03738397.

Study name	A Study to Compare Safety and Efficacy of Upadacitinib to Dupilumab in Adult Participants With Moderate to Severe Atopic Dermatitis
Methods	Allocation: randomised Intervention model: parallel assignment Masking: quadruple (participant, care provider, investigator, outcomes assessor) Primary purpose: treatment
Participants	Atopic dermatitis
Interventions	Drug: upadacitinib (ABT‐494) Drug: dupilumab Drug: placebo
Outcomes	Percentage of participants achieving a 75% reduction in Eczema Area and Severity Index (EASI) (EASI 75) from baseline Percent change in worst Pruritus Numerical Rating Scale (NRS) Percentage of participants achieving a 100% reduction in EASI (EASI100) Percentage of participants achieving a 90% reduction in EASI (EASI90)
Starting date	2019 February 21
Contact information	AbbVie
Notes	‐

NCT03754309.

Study name	A Study of KY1005 in Patients With Moderate to Severe Atopic Dermatitis
Methods	Allocation: randomised Intervention model: parallel assignment Masking: triple (participant, investigator, outcomes assessor) Primary purpose: treatment
Participants	Dermatitis, atopic
Interventions	Drug: KY1005 Drug: placebo
Outcomes	Percentage change in Eczema Area and Severity Index (EASI) Incidence of treatment‐emergent adverse events (TEAEs) Percentage and absolute change from baseline in EASI over time Change in epidermal thickness Change in keratin 16 staining of skin biopsies Percentage of patients with ≥ 50% reduction in EASI (EASI50) Percentage of patients with ≥ 75% reduction in EASI (EASI75) Percentage of patients with ≥ 90% reduction in EASI (EASI90) Change in Validated Investigators' Global Assessment (vIGA) Percentage of patients with a response on vIGA of 0 or 1 Change in SCORing of Atopic Dermatis (SCORAD) Index Change in affected body surface area (BSA) Change in Patient‐Oriented Eczema Measure (POEM) Change in Patient‐Oriented SCORing of Atopic Dermatitis (PO‐SCORAD) Index Change in Dermatology Quality of Life Index (DLQI) Change in Numerical Rating Scale (NRS) for pruritus
Starting date	2018 December 13
Contact information	Kymab Limited
Notes	‐

NCT03761537.

Study name	Tralokinumab in Combination With Topical Corticosteroids in Subjects With Severe Atopic Dermatitis Who Are Not Adequately Controlled With or Have Contraindications to Oral Cyclosporine A
Methods	Allocation: randomised Intervention model: parallel assignment Masking: double (participant, investigator) Primary purpose: treatment
Participants	Atopic dermatitis
Interventions	Drug: tralokinumab Other: placebo
Outcomes	≥ 75% reduction in EASI score (EASI75) IGA score of 0 (clear) or 1 (almost clear) Change in SCORAD EASI75 response Reduction in worst daily Pruritus NRS (weekly average) ≥ 4 Change in DLQI score Adverse events Presence of anti‐drug antibodies
Starting date	2018 December 28
Contact information	LEO Pharma
Notes	‐

NCT03817190.

Study name	Efficacy and Safety of Orally Administered DS107 in Adult Patients With Moderate to Severe Atopic Dermatitis
Methods	Allocation: randomised Intervention model: parallel assignment Masking: double (participant, investigator) Primary purpose: treatment
Participants	Atopic dermatitis
Interventions	Drug: DS107 Drug: placebo
Outcomes	Validated Investigators' Global Assessment (vIGA) Eczema Area Severity Index (EASI) Numerical Rating Scale (NRS) Body surface area
Starting date	DS Biopharma
Contact information	September 2019
Notes	‐

NCT03912259.

Study name	Evaluation of Dupilumab in Chinese Adult Patients With Moderate to Severe Atopic Dermatitis
Methods	Allocation: randomised Intervention model: parallel assignment Masking: quadruple (participant, care provider, investigator, outcomes assessor) Primary purpose: treatment
Participants	Atopic dermatitis
Interventions	Drug: dupilumab SAR231893 Drug: placebo Drug: emollient (moisturiser)
Outcomes	Investigators' Global Assessment (IGA) Eczema Area and Severity Index (EASI)75 response (score reduction 75%) Peak score on Pruritus Numerical Rating Scale (NRS) (score reduction) Peak score on Pruritus NRS (score reduction) Change in peak score on pruritus NRS Percent change in peak score on pruritus NRS Change in EASI score Change in BSA affected by atopic dermatitis (AD) Change in Dermatology Life Quality Index (DLQI) score Change in Patient‐Oriented Eczema Measure (POEM) score Change in Pruritus NRS score Change in EuroQol 5 Dimensions questionnaire (EQ‐5D) Percent change in EuroQol 5 dimensions questionnaire (EQ‐5D) IGA score reduction Patients with EASI50 Patients with EASI90 Patients achieving IGA 0 to 1 and a reduction of points EASI score changes EASI score changes, percent Change in weekly average of peak daily Pruritus NRS score Percent change in weekly average of peak daily Pruritus NRS score Patients with "absence of pruritus" or "mild pruritus" on the pruritus categorical scale Sick leave/missed school days Sick leave/missed school days, patient proportion Adverse events Immunogenicity: dupilumab
Starting date	2019 January 28
Contact information	Sanofi Regeneron Pharmaceuticals, Inc.
Notes	‐

NCT03948334.

Study name	A Study to Assess the Safety and Efficacy of ZPL389 With TCS/TCI in Atopic Dermatitis Patients
Methods	Allocation: randomised Intervention model: parallel assignment Masking: double (participant, investigator) Primary purpose: treatment
Participants	Atopic dermatitis
Interventions	Drug: ZPL389 Drug: TCS or TCI, or both
Outcomes	Frequency of AEs Frequency of AEs Investigators' Global Assessment response Eczema Area and Severity Index (EASI) score
Starting date	2019 April 4
Contact information	Novartis Pharmaceuticals Novartis
Notes	‐

ADIQ: Atopic Dermatitis Impact Questionnaire.

ADSS: Atopic Dermatitis Sleep Scale.

AE: adverse event.

AESI: adverse event of special interest.

AI: auto‐injector.

BSA: body surface area.

CDLQ: Children's Dermatology Life Quality Index.

CsA: cyclosporin A.

DLQI: Dermatology Life Quality Index.

EASI: Eczema Area and Severity Index.

EASI75: 75% improvement in Eczema Area and Severity Index.

EASI90: 90% improvement in Eczema Area and Severity Index.

ECP: Eosinophil cationic protein.

EQ‐5D‐5L: European Quality of Life 5 Dimensions 5 Levels.

HADS: Hospital Anxiety and Depression Scale.

IGA: Investigators' Global Assessment.

LCR: late cutaneous response.

MTX: methotrexate.

NRS: Numerical Rating Scale.

oSCORAD: Objective Severity Scoring of Atopic Dermatitis Scale.

PGA: Physicians' Global Assessment.

PGI‐S‐AD: Patient Global Assessment of Severity ‐ Atopic Dermatitis Scale.

POEM: Patient‐Oriented Eczema Measure.

PO‐SCORAD: Patient‐Oriented SCORing of Atopic Dermatitis Index.

PSAAD: Pruritus and Symptoms Assessment for Atopic Dermatitis.

PTC: product technical complaint.

PTF: product technical failure.

RCT: randomised controlled trial.

SAE: serious adverse event.

SCORAD: severity SCORing of Atopic Dermatitis.

SCORAD75: 75% improvement in severity SCORing of Atopic Dermatitis.

TCS: topical corticosteroid.

TEAE: treatment‐emergent adverse event.

vIGA: validated Investigators' Global Assessment.

WPAI‐AD: Work Productivity and Activity Impairment ‐ Atopic Dermatitis questionnaire.

Differences between protocol and review

Types of studies

In this review, we described types of studies by adding the following sentence: “We included both parallel and cross‐over design trials, but we excluded cluster RCTs".

Types of participants

We added hand eczema as one of our exclusion criteria for this review. In the protocol, we planned to include eczema of any severity, but we limited the focus of the review to moderate to severe eczema because systemic interventions generally are not considered appropriate for mild eczema because mild eczema is easily controlled with topical treatments, and systemic eczema treatments are considerably more expensive and potentially more toxic than topical eczema treatments.

Types of interventions

Our previous protocol did not clearly define the node of analysis for each intervention in network meta‐analysis. Therefore, in this review, we clearly defined nodes in network meta‐analysis based on interventions regardless of dose, while splitting interventions with different doses in subgroup analysis.

We excluded infliximab, rituximab, tocilizumab, tacrolimus, pimecrolimus, and alitretinoin from this review because these treatments generally are not considered as systemic treatments for moderate to severe eczema.

Types of outcome measures

Outcomes of interest were not clearly specified in our previous protocol. The protocol defined many measures for clinical severity of eczema, including EASI, SCORAD, SASSAD, IGA, and others. However, we did not specify whether these measurements should be continuous or dichotomous outcomes. In this review, we decided to use dichotomous outcomes as our outcomes of interest and continuous outcomes when dichotomous outcomes were not available. We made this decision because dichotomous outcomes could reflect the clinical difference in clinical severity better than continuous outcomes. Based on that decision, we selected EASI75 as one of our primary outcomes, along with POEM50. However, only a few studies reported POEM50; therefore, we selected POEM reported as a continuous score instead of POEM50. In addition, the proportion of participants who achieved IGA 0/1 was selected as a secondary outcome because this outcome provided data obtained by comparing biological treatment and conventional treatment. Therefore, revised primary and secondary outcomes were as follows.

Primary outcomes

• Proportions of participants who achieved EASI75 (achieving 75% improvement in EASI score) at short‐term (≤ 16 weeks) and long‐term (> 16 weeks) durations were the primary outcomes for assessing clinical severity of eczema. We selected this outcome because of consensus agreement by the HOME initiative that this should be a core instrument for clinician‐reported signs (Schmitt 2014), and because dichotomous outcomes could reflect clinical differences in severity better than the continuous outcomes

• Proportions of participants who achieved POEM50 (achieving 50% improvement in POEM score) at short‐term and long‐term durations were planned to be the primary outcomes for participant‐reported symptoms because of consensus agreement by the HOME initiative that this should be a core instrument for patient‐reported symptoms (Spuls 2017). However, only a few studies reported POEM50; therefore, we used POEM instead of POEM50 in continuous scores at short‐ and long‐term durations

• We considered the proportion of participants with serious adverse effects (SAEs) and the proportion of participants with infection at short‐term and long‐term durations used to evaluate the safety of each systemic treatment

Secondary outcomes

• Proportions of participants who achieved Investigators' Global Assessment or Physicians' Global Assessment value of 0 or 1 (clear or almost clear) (IGA 0/1) at short‐term and long‐term durations

Measures of treatment effects

We did not report in this review the number needed to treat for an additional beneficial outcome (NNTB) or the number needed to treat for an additional harmful outcome (NNTH) for each specific comparison, so we deleted from the Methods section the sentences related to NNTB and NNTH in measures of treatment effect.

We added details for measures of treatment effect (why we report safety outcomes in terms of RR, how to interpret) for greater clarity, and we added the following sentences.

• "We used risk ratios to compare adverse events as all included studies were randomised controlled trials. RR is an appropriate outcome measure for comparison of adverse events in this situation".

• "The SMD expresses the intervention effect in standard units rather than in the original units of measurement. When SMDs were estimated, the absolute values of the intervention and comparison groups were not useful because studies have used different measurement instruments with different units. Therefore, “Cohen’s effect sizes” were used for interpreting SMDs as follows: 0.2 represents a small effect, 0.5 a moderate effect, and 0.8 a large effect (Cohen 1988). Variations were represented as small, moderate, and large if values were < 0.40, 0.40 to 0.70, and > 0.70, respectively (Cohen 1988)".

• “For categorical outcomes such as POEM and IGA, the data were dichotomised to the proportion of participants achieving clear or almost clear eczema (POEM score 0 to 2, IGA score 0/1) or mild disease (SCORAD 0 to 24). Then RR was used for analysis in the same way as dichotomous outcome data. In addition, the proportion of participants achieving 50% improvement in SCORAD or POEM was recorded and analysed as RR”.

Ranking probabilities

We added information about ranking probabilities for specific primary and secondary outcomes for greater clarity. "The SUCRA was used to estimate relative rankings in all four primary outcomes (EASI75, POEM, SAE, and infection) and in one secondary outcome (IGA 0/1) at week 16 of follow‐up time or nearest (short‐term duration) and at time points over 16 weeks of follow‐up (long‐term duration). We also developed and used rankograms to reflect uncertainty in ranking probabilities visually and graphically (Chaimani 2017)".

Unit of analysis issues

We added the reasons to describe why we use only data from the first part of a cross‐over study. "To avoid carry‐over effects in cross‐over studies, we only included data from the first part of any included study (i.e. before cross‐over), assuming data for the first phase are available. The data from the first part of a cross‐over study were included because we were not confident that the washout period was sufficient to prevent a carryover effect. According to the Cochrane Handbook for Systematic Reviews of Interventions, data from the first period can be used in situations where a carryover effect might be suspected”.

Assessment of heterogeneity

We added details for assessment of heterogeneity and changed criteria to consider heterogeneity from a cut point of I² > 50% to a range of I² as detailed below.

• "We used the Q‐test and the I² statistic to measure heterogeneity in the study results. We interpreted the I² statistic according to the following thresholds (Section 9.5.2 of the Cochrane Handbook for Systematic Reviews of Interventions; Deeks 2017): 0% to 40% might not be important; 30% to 60% may represent moderate heterogeneity; 50% to 90% may represent substantial heterogeneity; 75% to 100% represents considerable heterogeneity. We considered the range of I² of 30% to 60% and a P value < 0.05 as indicating substantial statistical heterogeneity. If statistical heterogeneity was significant, we explored the potential for heterogeneity. When we considered that the studies are suitable to combine in a meta‐analysis based on clinical and methodological characteristics, we pooled the effect estimates using a random‐effects model".

Data synthesis

We added information from data synthesis for specific primary and secondary outcomes for greater clarity.

• "We performed NMA on all outcomes including both efficacy and safety outcomes at week 16 of F/U time or nearest (short‐term duration) and at time points > 16 weeks (long‐term duration). Our primary analysis included RCTs that included participants with moderate to severe eczema who used systemic immunosuppressive treatments administered by any route. In the network meta‐analysis, our main comparisons included dupilumab, mepolizumab, tralokinumab, omalizumab, or other biological agents, and systemic conventional treatments such as cyclosporin A (ciclosporin), methotrexate, azathioprine, mycophenolate, and systemic corticosteroid compared with placebo (plus standard topical therapies) or other active agents".

Subgroup analyses

Given that outcome data in the included studies were not separately available for ages ≤ 12 and > 12 years nor for moderate and severe eczema, we were not able to perform subgroup analyses by age of participant and severity of eczema at baseline. Therefore, we performed subgroup analysis based on dosage schedule and duration of follow‐up for both pair‐wise analysis and network meta‐analyses of EASI75, POEM score, SAEs, and infection outcomes.

Sensitivity analysis

We planned to perform a series of sensitivity analyses using Stata as follows (Stata 2017).

• Exclusion of studies with high risk of bias.

• Exclusion of studies with small sample size (i.e. < 25th percentile of total number of participants).

• Exclusion of unblinded studies.

• Analysis without industry‐sponsored trials.

• Exclusion of studies with high risk of attrition bias and incomplete outcome data.

However, available data are limited, so we performed sensitivity analyses only for the following: (1) inclusion of studies with low risk of bias, and (2) exclusion of studies with small sample sizes (i.e. < 25th percentile of total number of participants) for EASI75 outcomes during short‐term and long‐term durations.

'Summary of findings' tables and GRADE assessments

Our previous protocol did not comprehensively define the interventions and outcomes to be presented in the 'Summary of findings' tables. Therefore, we revised the detail and pattern to report 'Summary of findings' tables following Yepes‐Nuñez 2019, and we added more information for grade assessment to improve the clarity of the review. We added a rationale for our selection of comparisons for the ‘Summary of findings’ tables, to tell the reader that the selection wasn’t led by the review’s findings.

Contributions of authors

NC was the contact person with the editorial base.
RS, PD, ALL, NML, RD, and NC co‐ordinated contributions from the co‐authors and wrote the final draft of the review.
RS and PD screened papers against eligibility criteria.
RS and PD obtained data on ongoing and unpublished studies.
RS, PD, NML, and NC appraised the quality of papers.
RS, PD, NML, and NC extracted data for the review and sought additional information about papers.
RS, PD, and NML entered data into RevMan.
RS, PD, and NC analysed and interpreted data.
RS, PD, ALL, NML, and NC worked on the Methods sections.
RS, PD, ALL, NML, RD, and NC drafted clinical sections of the background and responded to the clinical comments of referees.
RS, PD, NML, and NC responded to the methodological and statistical comments of referees.

Disclaimer

This project was supported by the National Institute for Health Research via Cochrane Infrastructure funding to the Cochrane Skin Group, and was also supported by the Complex Reviews Support Unit, also funded by the National Institute for Health Research (project number 14/178/29). The views and opinions expressed therein are those of the authors and do not necessarily reflect those of the Systematic Reviews Programme, NIHR, NHS, or the Department of Health.

Sources of support

Internal sources

• Mahasarakham University, Thailand

  This research project was financially supported by Mahasarakham University.

External sources

• The National Institute for Health Research (NIHR), UK

  The NIHR, UK, is the largest single funder of the Cochrane Skin Group.

Declarations of interest

Ratree Sawangjit: none declared.
Piyameth Dilokthornsakul: none declared.
Antonia Lloyd‐Lavery: none declared.
Nai Ming Lai: none declared.
Robert Dellavalle: serves as Cochrane Skin Joint Co‐ordinating Editor (starting 15 January 2018). He receives an editorial stipend for serving as Social Media Editor and reimbursement for annual editorial board meeting expenses from the Journal of the American Academy of Dermatology, an editorial stipend from the Journal of Investigative Dermatology for serving as Podcast Editor, and reimbursement for travel to the Dermatology Foundation annual grant review meeting, where he serves as a grant reviewer. He receives royalties from UpToDate, for which he serves as a dermatology section editor. He has served as the principal investigator on independent grants from Pfizer Pharmaceuticals to the University of Colorado, which develops patient decision aids for dermatological diseases.
Nathorn Chaiyakunapruk: none declared.

New