Systematic Review and Meta-analysis: Loss of Response and Need for Dose Escalation of Infliximab and Adalimumab in Ulcerative Colitis

Abstract

Background

Loss of response to infliximab or adalimumab in ulcerative colitis occurs frequently, and dose escalation may aid in regaining clinical benefit. This study aimed to systematically assess the annual loss of response and dose escalation rates for infliximab and adalimumab in ulcerative colitis.

Methods

A systematic search was conducted from August 1999 to July 2021 for studies reporting loss of response and dose escalation during infliximab and/or adalimumab use in ulcerative colitis patients with primary response. Annual loss of response, dose escalation rates, and clinical benefit after dose escalation were calculated. Subgroup analyses were performed for studies with 1-year follow-up or less.

Results

We included 50 unique studies assessing loss of response (infliximab, n = 24; adalimumab, n = 21) or dose escalation (infliximab, n = 21; adalimumab, n = 16). The pooled annual loss of response for infliximab was 10.1% (95% confidence interval [CI], 7.1-14.3) and 13.6% (95% CI, 9.3-19.9) for studies with 1-year follow-up. The pooled annual loss of response for adalimumab was 13.4% (95% CI, 8.2-21.8) and 23.3% (95% CI, 15.4-35.1) for studies with 1-year follow-up. Annual pooled dose escalation rates were 13.8% (95% CI, 8.7-21.7) for infliximab and 21.3% (95% CI, 14.4-31.3) for adalimumab, regaining clinical benefit in 72.4% and 52.3%, respectively.

Conclusions

Annual loss of response was 10% for infliximab and 13% for adalimumab, with higher rates during the first year. Annual dose escalation rates were 14% (infliximab) and 21% (adalimumab), with clinical benefit in 72% and 52%, respectively. Uniform definitions are needed to facilitate more robust evaluations.

Key Messages.

What is already known?

• A proportion of patients develop loss of response after initial response to anti-TNF for ulcerative colitis, and these patients may regain therapeutic effect following dose escalation.

What is new here?

• In this meta-analysis, we found pooled annual loss of response risks of 10% and 13% and pooled annual dose escalation rates of 14% and 21% for infliximab and adalimumab, respectively—with higher rates in the first year.

How can this study help patient care?

• The results from our meta-analysis support strict disease monitoring during the first year of anti-TNF treatment and the value of dose escalation to regain therapeutic effect.

Introduction

Antitumor necrosis factor alpha (anti-TNF) agents are essential therapeutics in moderate to severe ulcerative colitis (UC). In 2005, infliximab (IFX) was the first anti-TNF agent approved by the Food and Drug administration (FDA) and European Medication Agency (EMA) for the treatment of moderate to severe UC.¹ Infliximab is an IgG1 mouse-human chimeric monoclonal antibody neutralizing the effects of TNF, resulting in potent anti-inflammatory effects. In 2012, adalimumab (ADA), a fully human monoclonal antibody, was approved for the treatment of UC.² According to label, IFX maintenance therapy is dosed at 5 mg/kg intravenously every 8 weeks, with induction infusions at week 0, 2, and 6.¹ Adalimumab maintenance is dosed at a standard dose of 40 mg subcutaneously every 2 weeks, with an induction schedule of 160 mg at week 0, and 80 mg at week 2.²

In general, patients with clinical response or remission after the induction period are considered primary responders to anti-TNF therapy. Disease worsening after primary response is considered loss of response (LOR).³ However, there is a wide variation in definitions in place, and consensus on definitions of primary nonresponse and LOR are lacking.^3,4 To address LOR, dose escalation is possible by either a dose increase (IFX, up to 10 mg/kg; ADA, 80 mg every 2 weeks) or interval reduction (IFX to every 4-6 weeks; ADA to 40 mg weekly).⁵

Loss of response during IFX and ADA use in UC patients has not been systematically assessed. In Crohn’s disease, loss of clinical response at week 52 ranged from 16%-71% for IFX and 16%-47% for ADA. Annual LOR incidences, which take into account different follow-up times among studies, varied between 13%-18% following IFX initiation and 23%-25% following ADA initiation.^6–8 A recent retrospective multicenter study reporting LOR rates in both Crohn’s disease and UC found an LOR percentage of 9% per patient-year in UC.⁹

Dose escalation has shown to be effective in achieving higher rates of mucosal healing in patients with signs of LOR.¹⁰ Annual dose escalation rates were 15% in IFX-treated Crohn’s disease patients and 24%-26% in ADA-treated Crohn’s disease patients, although follow-up varied widely (20-208 weeks); this may have impacted the annual dose escalation rates.^6–8 In UC patients, overall dose escalation rates in primary responders vary between 13%-37% for IFX and 18%-55% for ADA.^11–13 Despite a number of available studies, no comprehensive and systematic evaluation on LOR and dose escalation for anti-TNF in UC is available.

This systematic review and meta-analysis aimed to assess (1) the annual LOR rates to IFX and ADA in adult UC patients with primary response, stratified per LOR definition; (2) the annual dose escalation rates of IFX and ADA in primary responders as a proxy for LOR; and (3) the proportion of patients regaining clinical benefit after dose escalation following loss of response.

Materials and Methods

Study Design

This study followed the methodology for conducting and reporting a systematic review described in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines and guidelines for reporting developed by the Meta-analysis of Observational studies in Epidemiology (MOOSE) group (Supplementary Tables 1, 2, and 3).^14,15 The study protocol was registered in the international database of prospectively registered systematic reviews (PROSPERO; CRD42020160390).

Data Sources and Search Strategy

A systematic search of PubMed, EMBASE, and Cochrane Library was conducted from August 1999 (date of IFX EMA approval for Crohn’s disease) to July 2021, with aid of an experienced librarian. Search algorithms included a combination of the following terms: ulcerative colitis, inflammatory bowel disease, infliximab, adalimumab, and antitumor necrosis factor (for full search strategy, see Supplementary Table 4). There were no language or study design restrictions. Two authors (E.S., P.T.) independently reviewed titles and abstracts of all studies identified in the search for inclusion and exclusion criteria. Subsequently, full text of the selected articles was reviewed for availability of the outcomes of interest. A manual search of references in the included articles was performed to identify additional articles. Discrepancies were resolved by discussion. If no consensus was reached, a third reviewer (F.H.) was involved for final decision-making. All search results were cross-checked for duplicates using Endnote X9.2 (Clarivate Analytics, Philadelphia, PA, USA). Rayyan QRCI (Rayyan Systems Inc.) was used for blinded title/abstract screening.¹⁶

Selection Criteria

Clinical trials and cohort studies assessing IFX and/or ADA use in adult patients with UC were eligible for inclusion if they reported the LOR rates or dose escalation rates due to LOR, with or without the outcome of dose escalation (regaining response or remission). Studies were excluded if they were case reports or case series or if the full text was not available. Additionally, studies not using standard induction or maintenance treatment and studies primarily focusing on proactive drug monitoring were deemed inaccurate for comparison and were excluded. We did not extrapolate LOR from studies only providing clinical remission/response rates since we could not reliably account for other reasons for not reaching remission or response. In case of multiple published studies involving the same cohort, data from the most comprehensive study were included.

Data Extraction

The following information from each study was extracted: first author’s last name, year of publication, study design, anti-TNF use (IFX/ADA), and total study population characteristics (including prior exposure to anti-TNF and baseline concomitant immunomodulator use). Specific outcomes of interest included total follow-up duration following IFX or ADA initiation, number of primary responders, LOR incidence, LOR-related dose escalation incidence, dose escalation strategies, and dose escalation outcomes. Lastly, we extracted the used definitions for primary response, LOR, dose escalation, and clinical benefit of dose escalation. Data were extracted separately by ES and PT, and discrepancies were resolved through discussion.

Outcomes

The primary outcomes included (1) the annual LOR per patient-year to IFX and ADA in UC patients, and (2) the annual dose escalation rates per patient-year for IFX and ADA in UC patients and the subsequent ability to regain therapeutic effect. Loss of response stratified per used definition was added following evaluation of study-specific definitions of LOR. Stratification per follow-up duration (≤65 weeks [52 weeks with a range of one quarter to correct for borderline follow-up time] vs >65 weeks) was added after a recent study implied time dependency of LOR.⁹

Subgroup analyses

Subgroup analyses for annual LOR were carried out, including (1) the extent of disease affecting the colon; (2) the type of anti-TNF agent; (3) concomitant use of mesalamine; (4) concomitant use of immunomodulator; (5) study design; (6) prior anti-TNF agent exposure; (7) adequate trough levels and presence of antidrug antibodies; and (8) quality of studies.

Definitions

Loss of response describes patients who responded to induction therapy but subsequently lost clinical response during maintenance treatment. Primary nonresponse was defined as no response or remission after the induction period.³ Because there is no consensus definition for primary response and LOR, we used the definitions as reported by the authors in each study.³ Consequently, LOR was categorized in 3 groups: (1) treatment discontinuation due to LOR; (2) treatment optimization defined as dose escalation or a combination of dose escalation and treatment switch and/or medical rescue therapy (including steroids and ciclosporin) and/or surgery because of LOR; and (3) increase of disease activity defined as increased clinical disease activity scores (eg, partial Mayo score, physician’s global assessment) and/or increased C-reactive protein (according to provided standards and combined with clinical or endoscopic indication of disease activity) and/or endoscopic disease worsening (according to provided standards or scores). Dose escalation was defined as any dose increase or interval shortening as reported by the authors. Dose escalation rates include all dose escalations performed in primary responders regardless of the reason for dose escalation. Regaining therapeutic effect following dose escalation was defined as recapturing clinical response and/or clinical remission. In case any study used multiple effect outcome measures, short-term clinical response or remission was used.

Risk of Bias Assessment

Information on the methodological quality of each study was extracted, and quality assessment was performed using the ROBINS-I tool for observational cohort studies and risk of bias 2 tool for randomized controlled trials, as recommended by the Cochrane Handbook.^17,18 In the ROBINS-I tool, 7 items can be assessed on a 5-point scale: low, moderate, serious, critical risk of bias, or no information. Details on the subdomains and scoring are provided in Supplementary Figure 1. Two authors (P.T., E.S.) independently assessed risk of bias by rating every item. Discrepancies in the risk of bias judgment were resolved by discussion.

Data Synthesis and Analysis

Due to large variation in follow-up among studies, we used annual rates per patient-year per study to describe the LOR and dose escalation incidence rates. The annual LOR per patient-year was calculated as follows: LOR events/ (primary responders x years of follow-up). The same method was used to calculate the annual dose escalation per patient-year. Regaining therapeutic effect following dose escalation was reported as a percentage of the total number of dose escalations. We used a random-effects model to calculate the pooled annual LOR incidence rates, annual dose escalations incidence rates, and proportion of patients regaining therapeutic effect after dose escalation, using the DerSimonian and Laird method.¹⁹ Proportion transformation was performed using the generalized linear mixed model.²⁰ Additionally, subgroup analyses on predefined LOR definitions (described previously) were conducted for the annual LOR rates. Heterogeneity across the pooled studies was evaluated using I² statistics, with low, moderate, substantial, and considerable heterogeneity defined as <30%, 30%-59%, 60%-75%, and >75%, respectively.²¹ To assess the robustness of the pooled effects, influence analyses using a manual leave-one-out sensitivity analysis was performed.²² Publication bias was evaluated through visual inspection of funnel plots. All analyses were conducted in R Version 3.6.2 (Free Software Foundation, Inc.), using the following packages: meta and metafor for meta-analyses, and robvis for risk of bias visualization.

Results

Search Results

After removal of duplicates, the database search yielded 26 508 potentially relevant articles that were assessed for title and abstract (Supplementary Figure 2). We assessed the full text for 166 studies. Finally, we included 50 unique observational cohort studies and no randomized controlled trials on adult UC patients treated with either IFX, and/or ADA. These studies provided data on LOR incidence (n = 38),^11,23–59 and/or dose escalation incidence (n = 31),^{11,24,26–28,31,33,37–41,47–50,53–57,59–68} and/or dose escalation outcomes (n = 15).^{11,12,27,31,39,50,53–56,59,62,66,69,70} Outcomes for IFX were described in 34 studies. ^{12,23–39,49–52,56,57,59–66,69,70} Outcomes for ADA were described in 23 studies.^{11,24,25,28,31,33,37,40–48,53–55,57,58,67,68} Forty-one were retrospective cohort studies, and 9 were prospective cohort studies. Characteristics of all included studies are provided in Tables 1 and 2. Definitions of primary response, LOR, and outcomes of dose escalation per study are provided in Supplementary Table 5.

Table 1.

Characteristics of studies included in the meta-analysis of loss of response to infliximab or adalimumab.^*

Study	Year	Design	Anti-TNF	Prior Anti-TNF Exposure, in %	Concomitant IM use, in %	Follow-up, in weeks	Primary responders, n	LOR n (%)	Annual LOR in %
Loss of response defined as discontinuation
Angelison23	2016	Retro	IFX	3	53	52	190	22 (12)	11.6
Barberio24	2020	Retro	IFX	10	15	52	60	6 (10)	10.0
Bertani25	2020	Pros	IFX	NA	0	48	27	5 (19)	20.1
Fernandez26	2015	Retro	IFX	0	77	160	170	32 (19)	6.1
García-Bosch27	2016	Retro	IFX	0	85	302	40	16 (40)	6.9
Gies28	2010	Pros	IFX	0	37	60	18	4 (22)	19.3
Hayes29	2014	Retro	IFX	4	54	52	71	8 (11)	11.3
Kolar30	2017	Retro	IFX	0	NA	46	26	3 (12)	13.0
Kronsten31	2021	Retro	IFX	0	65	52	74	5 (7)	6.8
Petitdidier32	2020	Retro	IFX	30	69	54	118	37 (31)	30.2
Pouillon33	2019	Retro	IFX	13	43	282	105	26 (25)	4.6
Russo34	2008	Retro	IFX	0	50	65	31	3 (10)	7.7
Teisner35	2010	Retro	IFX	0	67	96	12	5 (52)	22.6
Tursi36	2010	Pros	IFX	0	100	83	23	2 (9)	5.4
Tursi37	2021	Retro	IFX	NA	33	209	172	9 (5)	1.3
Viola38	2019	Retro	IFX	100	12	52	59	1 (2)	1.7
Yamada39	2014	Retro	IFX	0	49	156	24	1 (4)	1.4
Baert40	2014	Retro	ADA	100	21	52	55	16 (29)	29.1
Balint41	2015	Pros	ADA	67	52	52	55	4 (7)	7.3
Barberio24	2020	Retro	ADA	58	23	52	30	8 (27)	26.7
Bertani25	2020	Pros	ADA	NA	0	48	20	8 (40)	43.3
Christensen42	2015	Retro	ADA	100	27	52	21	2 (10)	9.5
Gies28	2010	Pros	ADA	0	37	60	20	6 (30)	26.0
Kamat43	2019	Retro	ADA	10	71	52	13	4 (31)	30.8
Kronsten31	2021	Retro	ADA	0	54	52	48	6 (13)	12.5
Kumei44	2021	Pros	ADA	14	37	52	23	3 (13)	13.0
Macaluso45	2021	Pros	ADA	34	NA	38	32	3 (9)	12.8
McDermott46	2012	Retro	ADA	87	NA	96	17	8 (47)	25.5
Oh47	2020	Retro	ADA	44	68	74	92	30 (33)	22.9
Pouillon33	2019	Retro	ADA	13	43	282	25	10 (40)	7.4
Tursi48	2018	Retro	ADA	41	11	78	102	8 (8)	5.2
Tursi37	2021	Retro	ADA	NA	17	209	111	2 (2)	0.4
Van de Vondel11	2018	Retro	ADA	63	32	177	101	5 (5)	0.6
Loss of response defined as treatment optimization
Alzafiri49	2011	Retro	IFX	4	81	62	19	10 (53)	44.1
Arias50	2014	Retro	IFX	0	48	245	184	113 (61)	13.0
Endo51	2016	Retro	IFX	0	53	107	27	8 (30)	14.4
Nishida52	2017	Retro	IFX	0	46	83	37	14 (38)	23.7
Otsuka59	2018	Retro	IFX	0	44	52	17	4 (24)	23.5
Renna53	2018	Retro	ADA	50	11	40	93	50 (54)	69.9
Taxonera54	2017	Retro	ADA	58	55	100	184	91 (50)	25.7
Zacharias55	2017	Retro	ADA	39	67	55	36	18 (50)	47.3
Loss of response defined as increase of disease activity
Juliao56	2013	Retro	IFX	0	71	52	21	4 (19)	19.0
Ma57	2015	Retro	IFX	0	65	159	66	39 (59)	19.3
Ma57	2015	Retro	ADA	0	65	159	36	21 (58)	19.1
Sugimoto58	2018	Retro	ADA	19	67	350	14	6 (43)	6.4

^*Treatment optimization is defined as dose escalation or a combination of dose escalation and treatment switch and/or medical rescue therapy and/or surgery because of LOR. Increase of disease activity is defined as increase of disease scores (eg, partial Mayo score), physician’s global assessment and/or increased C-reactive protein (according to provided standards) and/or endoscopic disease worsening (according to given provided standards or scores).

Abbreviations: ADA, adalimumab; IFX, infliximab; IM, immunomodulator including thiopurines and methotrexate; LOR, loss of response; NA, not available; TNF, tumour necrosis factor.

Table 2.

Characteristics of studies included in the meta-analysis of dose escalation during infliximab or adalimumab treatment

Study	Year	Design	Anti-TNF	Prior Anti-TNF Use, in %	Conco-mitant IM Use in %	Follow-up, in Weeks	Pre-DE Group, n	DE N (%)	Annual DE rate, %	Responsive to DE, n (%)	DE Strategy	Median Time to Escalation, in Weeks
Alzafiri49	2011	Retro	IFX	4	81	62	19	8 (42)	35.3	NA	DI or IS	NA
Arias50	2014	Retro	IFX	0	48	245	184	91 (50)	10.5	Overcoming clinical relapse: 46 (51)	DI to 10mg/kg: 14 IS 6 weeks: 28; IS 4 weeks: 11 Both IS and DI: 38	28 (IQR 15- 48)
Armuzzi60	2014	Retro	IFX	0	56	180	96	15 (16)	4.5	NA	NA	NA
Balint61	2018	Pros	IFX	7	43	54	73†	20 (27)	26.4	NA	DI to 10 mg/kg	12 (IQR 4-27)
Barberio24	2020	Retro	IFX	10	15	52	60	17 (28)	28.3	NA	IS to 6 or 4 weeks	NA
Barreiro-de Acosta62	2009	Pros	IFX	0	65	104	13	6 (46)	23.1	Deep remission: 3 (50)	IS to 6 weeks	NA
Cesarini69	2014	Retro	IFX	2	46	NA	NA	41	NA	C-Response at next infusion: 37 (93%); C-Remission week 52: 28 (68)	DI to 10mg/kg: 15 IS: 26 (4-6 weeks)	126 (range 17-413)
Dumitrescu70	2015	Retro	IFX	NA	40	NA	NA	113	NA	Week8: C-Response: 61 (54) C-Remission: 10 (11) Week 24: C-Response: 45 (40) C-Remission: 16 (14)	DI 10 mg/kg	31 (IQR 16-68)
Fernandez26	2015	Retro	IFX	0	77	160	187†	78 (42)	13.6	NA	DI: 47 IS: 9 Both: 13	NA
García-Bosch27	2016	Retro	IFX	0	85	302	40	10 (25)	4.3	C-Response: 9 (90)	DI 10 mg/kg IS to 6 or 4 weeks	29 (IQR 24-60)
Gies28	2010	Pros	IFX	0	37	60	18	7 (39)	33.7	NA	DI or IS	NA
Juliao56	2013	Retro	IFX	0	71	119	28†	4 (14)	6.2	C-Response: 4 (100)	IS to 6 weeks	NA
Kronsten31	2021	Retro	IFX	0	65	52	78†	10 (13)	12.8	No discontinuation: 6 (60)	DI or IS	35 (IQR 22-48)
Llao63	2016	Retro	IFX	0	80	136	15	8 (53)	20.4	NA	NA	NA
Ma57	2015	Retro	IFX	0	65	159	66	36 (55)	17.8	NA	DI or IS	NA
Otsuka59	2018	Retro	IFX	0	44	52	17	4 (24)	23.5	C-Response: 3 (75)	IS 6 weeks	NA
Oussalah64	2010	Retro	IFX	0	67	78	80	36 (45)	30.0	NA	NA	NA
Pouillon33	2019	Retro	IFX	6	43	282	140‡	77 (55)	10.1	NA	DI or IS	NA
Rostholder65	2011	Retro	IFX	0	38	52	50	27 (54)	54.0	5	DI or IS	NA
Taxonera12	2015	Retro	IFX	0	77	NA	NA	79	NA	C-Response week 12: 54 (68) C-Remission week 12: 41 (52) Sustained Clinical benefit: 46 (58)	DI: 30 IS to 4 weeks: 21 IS to 6 weeks: 28	NA
Tursi66	2014	Retro	IFX	0	NA	182	110	3 (3)	0.8	Remission: 3 (100)	DI 10mg/kg: 1 IS to 4 weeks: 1 Both: 1	NA
Tursi37	2021	Retro	IFX	NA	33	208	184†	9 (5)	1.2	NA	DI or IS	NA
Viola38	2019	Retro	IFX	100	12	52	53	26 (49)	49.1	NA	IS: 11 DI: 5 Both: 10	NA
Yamada39	2014	Retro	IFX	0	49	156	24	17 (71)	23.6	C-Remission: 16 (94)	DI: 2 IS: 8 Both: 7	NA
Baert40	2014	Retro	ADA	100	21	52	61	22 (36)	36.1	NA	NA	12
Balint41	2015	Pros	ADA	67	52	52	73	13 (18)	17.8	NA	IS	NA
Barberio24	2020	Retro	ADA	58	23	52	30	11 (37)	36.7	NA	DI or IS	NA
Gies28	2010	Pros	ADA	0	37	60	20	8 (40)	34.7	NA	IS	NA
Iborra67	2016	Retro	ADA	67	46	52	258†	93 (36)	36.0	NA	NA	21
Kronsten31	2021	Retro	ADA	0	54	52	63	16 (25)	25.4	No discontinuation: 7 (44)	DI or IS	24 (IQR 17-42)
Ma57	2015	Retro	ADA	0	65	139	36	18 (50)	18.7	NA	DI of IS	NA
Oh47	2020	Retro	ADA	44	68	74	92	37 (40)	28.3	NA	IS	28 (IQR 14-59)
Pouillon33	2019	Retro	ADA	35	43	282	63‡	38 (60)	11.1	NA	DI or IS	NA
Renna53	2018	Retro	ADA	50	11	40	93	50 (54)	69.9	“Clinical benefit”: 23 (46)	IS	NA
Taxonera54	2017	Retro	ADA	58	55	100	184	76 (41)	21.5	C-Response: 36 (47) C-Remission: 15 (20)	IS: 72 DI: 4	17 (IQR 9-39)
Taxonera68	2010	Retro	ADA	100	50	48	23	4 (17)	18.8	NA	IS	NA
Tursi48	2018	Retro	ADA	41	11	78	56	9 (16)	10.7	NA	NA	NA
Tursi37	2021	Retro	ADA	NA	17	208	118†	6 (5)	1.4	NA	DI or IS	NA
Van de Vondel11	2018	Retro	ADA	63	32	177	196§	129 (66)	19.3	Two PGA ≥13 weeks apart: 77 (60)	IS	12 (IQR 7-22)
Zacharias55	2017	Retro	ADA	39	67	55	36	8 (22)	21.0	Continuation of therapy: 5 (63)	IS	NA

†Including both primary responders and primary nonresponders.

‡Patients were included in the study if they have received at least 6 months of treatment with infliximab or adalimumab.

§Patients with early discontinuation were excluded, yet dose escalation occurred within time needed to assess primary response and may therefore also include partial responders or primary nonresponders.

Abbreviations: ADA, adalimumab; C-Response, clinical response; C-Remission, clinical remission; DE, dose escalation; DI, dose intensification; IFX, infliximab; IM, immunomodulator including thiopurine and methotrexate; IQR, interquartile range; IS, interval shortening; NA, not available; PGA, physician global assessment; TNF, tumour necrosis factor.

Annual Loss of Response to Infliximab

Twenty-four cohort studies were included in the meta-analysis to assess annual LOR to infliximab maintenance treatment.^{23–39,49–52,56,57,59} These studies collectively included 1591 primary responders and provided a total of 4075 patient-years of follow-up. Follow-up ranged from 46 weeks to 302 weeks. Only 1 study included >50% of patients with prior anti-TNF exposure. A wide variety of definitions (n = 10) was used to define LOR (Supplementary Table 5). Overall, estimates of the annual LOR rates to IFX ranged from 1.3% to 43.5%, and the pooled annual LOR rate to IFX was 10.1% (95% CI, 7.1-14.3; I² = 86.6%; Figures 1 and 2). Subgroup analysis showed a random effects summary of 13.6% annual LOR (95% CI, 9.3-19.9; I² = 69.1%) for studies with a follow-up of ≤65 weeks (n = 13) and 7.6% (95% CI, 4.4-13.1; I²= 90.7%) for studies with a follow-up >65 weeks (n = 11; Figure 2 and Supplementary Figure 3). The leave-one-out sensitivity analysis resulted in incidence rates ranging from 9.5%⁴⁹ to 11.3%,³⁷ indicating low outlier influences (Supplementary Table 6).

Figure 1.

Forest plot of annual incidence rates of loss of response to infliximab maintenance therapy. Annual incidence rates are plotted per LOR definition group and combined. Abbreviation: PY, patient-years.

Figure 2.

Summary of all meta-analyses’ random effects models.

Loss of response based on treatment discontinuation

Seventeen studies were included for the meta-analysis (Figures 1 and 2).^{23–35,37–39} The random-effects summary showed an annual LOR of 7.5% (95% CI, 4.9-11.4; I² = 87.1%). Subgroup analysis showed a random-effects summary of 11.4% annual LOR (95% CI, 7.5-17.2; I² = 70.1%) for studies with a follow-up ≤65 weeks (n = 10), and 4.5% annual LOR (95% CI, 2.4-8.4; I²  = 81.5%) for studies with a follow-up >65 weeks (n = 7).

Loss of response based on treatment optimization

Five studies were included for the meta-analysis (Figures 1 and 2).^49–52,59 The random-effects summary showed an annual LOR of 19.4% (95% CI, 12.7-29.4; I² = 76.9%). Subgroup analysis showed a random-effects summary of 35.0% annual LOR (95% CI, 20.7-59.1; I²  = 7.2%) for studies with follow-up ≤65 weeks (n = 2), and 13.7% annual LOR (95% CI, 11.6-16.3; I² = 55.3%) for studies with follow-up >65 weeks follow-up (n = 3).

Loss of response based on increase of disease activity

Two studies were included for the meta-analysis (Figures 1 and 2).^56,57 The random-effects summary showed an annual LOR of 19.3% (95% CI, 14.3-26.0; I² = 0%).

Annual Loss of Response to Adalimumab

A total of 21 studies included 1128 UC patients with primary response to ADA that were treated for 2219 patient-years of follow-up.^{11,24,25,28,31,33,37,40–48,53–55,57,58} Follow-up ranged from 40 weeks to 350 weeks. Seven studies included >50% of patients with prior anti-TNF exposure. A wide variety of definitions (N = 8) was used to define LOR (Supplementary Table 5). Overall, estimates of the annual LOR rates to ADA ranged from 0.4% to 69.9%, and the pooled annual LOR rate to ADA was 13.4% (95% CI, 8.2-21.8; I² = 89.8%; Figures 2 and 3). Subgroup analysis showed a random effects summary of 23.3% annual LOR (95% CI, 15.4-35.1; I² = 78.6%) for studies with a follow-up of ≤65 weeks (n = 12), and 7.2% (95% CI, 3.1-16.8; I²  = 92.0%) for studies with a follow-up >65 weeks (n = 9; Figure 2 and Supplementary Figure 4). The leave-one-out sensitivity analysis resulted in incidence rates ranging from 12.2⁵³ to 15.9,³⁷ indicating low outlier influences (Supplementary Table 7).

Figure 3.

Forest plot of annual incidence rates of loss of response to adalimumab maintenance therapy. Annual incidence rates are plotted per LOR definition group and combined. Abbreviation: PY, patient-years.

Loss of response based on treatment discontinuation

Sixteen studies were included in the meta-analysis (Figures 2 and 3).^{11,24,25,28,31,33,37,40–48} The random-effects summary showed an annual LOR of 10.7% (95% CI, 6.1-18.9; I² = 84.8%). Subgroup analysis showed a random-effect summary of 18.8% annual LOR (95% CI, 13.2-26.9; I² = 43.6%) for studies with a follow-up ≤65 weeks (n = 10), and 5.0% annual LOR (95% CI, 1.6-15.5; I² = 92.7%) for studies with a follow-up >65 weeks (n = 6).

Loss of response based on treatment optimization

Three studies were included for the meta-analysis (Figures 2 and 3).^53–55 The random-effects summary showed an annual LOR of 42.8% (95% CI, 26.1-70.0; I² = 93.9%).

Loss of response based on increase of disease activity

Two studies were included for the meta-analysis (Figures 2 and 3).^57,58 The random-effects summary showed an annual LOR of 11.7% (95% CI, 5.5-25.1; I² = 82.1%).

Infliximab Dose Escalation Incidence Rates and Regaining Therapeutic Effect

Twenty-one studies were included for the meta-analysis of annual dose escalation rates during IFX use (Figure 2 and Figure 4A; Table 2).^{24,26–28,31,33,37–39,49,50,56,57,59–66} Dose escalation strategies were reported in 19 studies. Four studies solely performed either interval shortening or increase of dosage, whereas in 15 studies, both strategies were used. Follow-up ranged from 52 weeks to 302 weeks. Only 1 study included >50% of patients with prior anti-TNF exposure. Random-effects summary showed an annual dose escalation rate of 13.8% (95% CI, 8.7-21.7; I² = 92.6%). Reported time to dose escalation medians ranged from 12 to 126 weeks (n = 6). Subgroup analysis showed a random-effects summary annual dose escalation rate of 31.8% (95% CI, 23.5-43.2; I² = 62.4%) for studies with ≤65 weeks follow-up (n = 8), and 8.5% (95% CI, 4.8-15.2; I² = 91.8%) for studies with >65 weeks follow-up (n = 13). The leave-one-out sensitivity analysis resulted in incidence rates ranging from 12.8%⁶⁵ to 15.7%,⁶⁶ indicating low outlier influences (Supplementary Table 8). Pooled effectiveness of dose escalation as reported by the authors was 72.4% (95% CI, 53.7-85.6; I² = 73.8%; n = 12; Figure 4C).^{12,27,31,39,50,56,59,62,65,66,69,70}

Figure 4.

Forest plot of annual incidence rates of dose escalation and subsequential clinical benefit. A, Dose escalations during infliximab maintenance therapy. B, Dose escalations during adalimumab maintenance therapy. C, Proportion of patients regaining clinical remission, clinical response, or continued use following dose escalation for loss of response during infliximab maintenance treatment. D, Proportion of patients regaining clinical remission, clinical response, or continued use following dose escalation for loss of response during adalimumab maintenance treatment. Abbreviation: PY, patient-years.

Adalimumab Dose Escalation Incidence Rates and Regaining Therapeutic Effect

Sixteen studies were included for the meta-analysis of annual dose escalation rates during ADA use (Figure 2 and Figure 4B; Table 2).^{11,24,28,31,33,37,40,41,47,48,53–55,57,67,68} Random-effects summary showed an annual dose escalation rate of 21.3% (95% CI, 14.4-31.3; I² = 89.6%). Reported time to dose escalation medians ranged from 12 to 28 weeks (n = 6). Subgroup analysis showed a random-effects summary annual dose escalation rate of 32.7% (95% CI, 24.5-43.5]. I² = 74.2%) for studies with ≤65 weeks of follow-up (n = 9), and 13.4% (95% CI, 6.8-26.3; I² = 88.4%) for studies with >65 weeks of follow-up (n = 7). The leave-one-out sensitivity analysis resulted in incidence rates ranging from 19.6%⁵³ to 25.8%,³⁷ indicating low outlier influences (Supplementary Table 9). Pooled effectiveness of dose escalation as reported by the authors was achieved in 52.3% (95% CI, 45.0-59.6; I² = 21.2%; n = 5; Figure 4D).^{11,31,53–55}

Subgroup Analyses

Subgroup analyses, including exploration of LOR predictors, could not reliably be performed for the following parameters: use of therapeutic drug monitoring, extent of disease, concomitant mesalamine or immunomodulator use, prior anti-TNF agent exposure, trough levels, and antidrug antibody presence. These features were only described for the total study population including primary nonresponders and not specifically for patients experiencing LOR. There were insufficient prospective studies for solid stratification for study design. We decided not to perform meta regression analyses given the lack of reliable and complete data per included study.

Risk of Bias

The ROBINS-I tool was used for all included studies. The overall risk of bias was low, moderate, serious, and critical in 4%, 24%, 62%, and 10%, respectively (Supplementary Figure 1). Visual inspection of funnel plots suggested a higher annual incidence of LOR and (to some extent) dose escalations in larger studies (ie, studies with more patient-years of follow-up; Supplementary Figures 5-6).

Discussion

In this systematic review and meta-analysis, we addressed the incidence rates of annual LOR and dose escalation during IFX and ADA treatment for UC. We found an overall pooled annual LOR of 10% for IFX and 13% for ADA. The annual LOR defined as treatment discontinuation was lower compared with treatment optimization (8% vs 19% for IFX and 11% vs 43% for ADA, respectively). In subgroup analyses, LOR rates were higher in studies with shorter follow-up (≤65 weeks) compared with studies with longer follow-up. Annual dose escalation rates were 14% for IFX and 21% for ADA; therapeutic effect was regained in the majority of cases. Definitions of primary response, LOR, successful dose escalation, and follow-up varied markedly among included studies.

As follow-up time varies largely between studies, we used the incidence rate per patient-year as an approximation of LOR. Recently, a large retrospective cohort study reported an LOR to anti-TNF for UC of 9% per patient-year, based on discontinuation due to disease activity, after a median of 2.4 years.⁹ The authors reported a gradual decrease in annual LOR over time, which is in line with our findings and could explain the higher overall LOR incidence in our study due to inclusion of studies with shorter follow-up. In the first year of treatment, the authors described an annual LOR of 30% for UC. This is significantly higher than the rates we found for LOR due to discontinuation for either IFX (14%) or ADA (23%) during the first 65 weeks of treatment. However, the authors only included a time frame of 8 months following 4 months of induction in the first year. Therefore, it is difficult to compare these LOR rates. It is likely that most causes for LOR, such as immunogenicity or loss of sensitivity to the mechanism of anti-TNF, occur relatively early given the predisposed susceptibility of this group of patients. The annual LOR for IFX and ADA may be different between Crohn’s disease and UC. As a comparison, in Crohn’s disease overall annual LOR rates to IFX and ADA in patients are 13%-18% and 23%-25%, respectively.^6–8 In these studies, LOR was most often based on increase of clinical symptoms, which is expected to lead to higher LOR rates compared with discontinuation or treatment optimization used in our study.^71,72 Moreover, the authors did not adjust for the follow-up time per individual study but reported the total LOR events in the total follow-up time of all studies. Two recent large retrospective studies suggested that the annual LOR risk to anti-TNF is higher in UC patients compared with Crohn’s disease patients.^9,73

Dose escalation may be used as a proxy of LOR and can be used to recapture response following LOR. Yet, dose escalation may also be performed to achieve desired trough levels in absence of disease activity. Although current guidelines only recommend reactive therapeutic drug monitoring and we have excluded studies primarily focusing on proactive drug monitoring, dose escalation rates may overestimate the true LOR rate to a small extent, as proactive dose escalations cannot be ruled out completely.⁷⁴ Similar to the LOR rates, the annual dose escalation rate was higher in studies reporting shorter follow-up time compared with studies with a longer follow-up. The need for dose escalation seems to be higher during the early phases of IFX and ADA treatment. In a previous systematic review, cumulative dose escalation rates after primary response were 15%-37% for IFX and 18-55% for ADA use.¹³ However, these rates were not adjusted for follow-up time and are therefore difficult to interpret. In Crohn’s disease, annual dose escalation rates (15% for IFX, 26% for ADA) were similar to our rates.⁸

Clinical benefit after dose escalation of IFX and ADA was observed in the majority of patients. However, similar to studies in Crohn’s disease, there is a large variation in regained clinical response and remission rates between studies.⁷⁵ Dose escalation outcomes are subject to used definitions for success. Most studies evaluated success based on short-term clinical parameters without using objective, long-term biochemical and endoscopic measurements following dose escalation. Moreover, dose escalations were not performed in a standardized manner and therefore lead to potential selection bias. It should also be noted that a proportion of dose escalations are performed for dose optimization in (partial) responders. Another proportion may have lost sensitivity to this mechanism and may eventually require a biological agent with a different mechanism of action.

Because the total LOR rate was higher in the first year of treatment (≤65 weeks), physicians should be more aware of signs of LOR during the first year of anti-TNF treatment and consider concomitant immunomodulator therapy, especially for IFX.⁷⁶ Patients should be informed on the higher risk of LOR during the early stages of therapy, and timely assessment of disease activity may be crucial to prevent a severe disease relapse. It is important to consider predictors that may identify patients at higher risk for LOR for improvement of therapy determination. For example, the HLA-DQA1*05 allele is associated with the development of antidrug antibody formation and consequently LOR.⁷⁷

It is important to report a uniform definition of primary response and LOR and work towards implementing these uniform definitions in clinical practice. This will enable physicians and researchers to gain more insight in the magnitude of LOR to medical therapies for inflammatory bowel diseases (IBD), establish predictors related to LOR, and evaluate the effectiveness of interventions following LOR. First, there is large variation in definitions used for LOR, which substantially impacts the number of patients classified as LOR. Loss of response may be defined as increase of clinical symptoms based on a variety of clinical indices that cannot be compared reliably. Moreover, increase of clinical disease scores may not always indicate clinically relevant LOR and may also not be related to the underlying IBD.⁷¹ The use of more reliable outcome measures such as (combinations of) biomarkers (eg, fecal calprotectin), endoscopic assessment, treatment optimization (either dose escalation or initiation of steroids), or treatment discontinuation may be more appropriate and applicable.^4,7,78 Second, as the LOR risk does not seem to be linear over time, varying follow-up time among studies reporting LOR hamper the ability to compare LOR rates.⁹ Loss of response rates are best reported at predefined time points using a survival curve that adjusts for drop-outs.

Strengths of our study include the extensive literature search and robust methodology. We stratified LOR according to different definitions and showed the impact of these definitions on the LOR rates. Limitations include the substantial heterogeneity among the included studies. Our findings on annual LOR in UC patients on IFX and ADA are based on different used definitions of LOR and follow-up in literature and should therefore be interpreted carefully. In order to adjust for varying follow-up times, we calculated the annual LOR and dose escalation rates per study. We recognize that this approach also has downsides including the expected nonlinear LOR and dose escalation risk over time. In addition, studies with a longer follow-up were consequently weighted more heavily compared with studies with a short follow-up. Annual LOR and dose escalation are therefore only an approximation of the “true” rates. In order to account for the expected nonlinear time relation, we performed a subgroup analysis for studies with a comparable follow-up time (≤65 weeks). We would have preferred to use individual data such as concomitant immunomodulator use for separation and prediction of LOR, but unfortunately these data were only available at an aggregate level in the majority of studies. Similarly, separate use of the definitions for response or remission was not feasible. No endoscopic data were available for further analysis. We did not compare LOR, dose escalation and dose escalation success rates between IFX and ADA as this was beyond the scope of this study. Lastly, our risk of bias assessment reflected the absence of well-defined outcomes of interest, which further underlines the need for definitions that are widely supported and can be implemented in daily practice.

In conclusion, annual LOR incidence in UC patients was 10% for IFX and 13% for ADA treatment. The annual LOR incidences were higher during the first 65 weeks of treatment for both IFX (14%) and ADA (23%). Dose escalation was performed in 14% (IFX) and 21% (ADA) of UC patients and resulted in clinical benefit in 72% and 52%, respectively. A uniform definition for LOR is warranted to gain more insight on the magnitude of this problem and for future evaluation of clinical interventions.

Glossary

Abbreviations

ADA

adalimumab

Anti-TNF

antitumor necrosis factor alpha

IBD

inflammatory bowel diseases

IFX

infliximab

LOR

loss of response

UC

ulcerative colitis

Author Contributions

P.T. and E.S. contributed to the conceptualization, methodology, data curation, formal analysis, visualization, and writing of the original draft. L.D., N.dB., T.R., and F.H. contributed to the conceptualization, methodology, supervision, review, and editing of the manuscript. All authors approved the final version of this manuscript and agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. No additional writing assistance was used for this manuscript.

Funding

No funding has been received for this study.

Conflicts of Interest

P.T., E.S., and N.dB. report no conflicts of interest. L.D. has served as speaker or on advisory boards of Sandoz, Janssen, and Dr Falk. T.R. has served as speaker or consultant for Ferring, Janssen, and Takeda. F.H. has served on as speaker or consultant for Abbvie, Takeda, and has received unrestricted grants from Janssen-Cilag, Abbvie.

Data Availability

Data, analytic methods, and study materials can be requested through contacting the corresponding author.