Evaluation of Long-term Clearance Rates of Interventions for Actinic Keratosis

Key Points

Question

What are the long-term clearance rates of treatments used in patients with actinic keratosis?

Findings

This systematic review and network meta-analysis included data from 15 randomized clinical trials reporting sustained clearance rates after at least 12 months of treatment as a proxy for long-term efficacy. Photodynamic therapy with aminolevulinate, imiquimod, 5%, photodynamic therapy with methyl aminolevulinate, and cryosurgery was associated with significant long-term benefits in the network meta-analysis.

Meaning

The findings of this study suggest an evidence-based framework for selecting interventions with long-term efficacy in actinic keratosis.

Abstract

Importance

Multiple interventions are available for the treatment of actinic keratosis (AK). However, most randomized clinical trials and meta-analyses focus on short-term efficacy outcomes.

Objective

To investigate and synthesize the long-term efficacy (≥12 months) of interventions for AK from parallel-arm randomized clinical trials.

Data Sources

Searches in MEDLINE, Embase, and Central were conducted from inception until April 6, 2020. The reference lists of the included studies and pertinent trial registers were hand searched. The study was completed February 27, 2021.

Study Selection

Two reviewers screened the titles and abstracts of 2741 records. Finally, 17 published reports (original studies and follow-up reports) referring to 15 independent randomized clinical trials with an overall sample size of 4252 patients were included.

Data Extraction and Synthesis

Two reviewers independently extracted data on study, patient, and intervention characteristics. Network meta-analysis (NMA) of each outcome was conducted with a frequentist approach. The Grading of Recommendations Assessment, Development, and Evaluation (GRADE) guidance for NMA was used to assess the certainty of evidence. The revised Cochrane risk-of-bias tool for randomized clinical trials was used to evaluate the methodologic quality.

Main Outcomes and Measures

Participant complete clearance, participant partial clearance, and lesion-specific clearance were the outcomes, with each assessed at least 12 months after the end of treatment.

Results

Data from 15 independent randomized clinical trials including 4252 patients were extracted and synthesized. Ten studies were included in an NMA for the outcome of participant complete clearance, with photodynamic therapy with aminolevulinate (ALA-PDT) showing the most favorable risk ratio (RR) compared with placebo (RR, 8.06; 95% CI, 2.07-31.37; GRADE, moderate), followed by imiquimod, 5% (RR, 5.98; 95% CI, 2.26-15.84; GRADE, very low), photodynamic therapy with methyl aminolevulinate (MAL-PDT) (RR, 5.95; 95% CI, 1.21-29.41; GRADE, low), and cryosurgery (RR, 4.67; 95% CI, 1.36-16.66; GRADE, very low). Similarly, ALA-PDT had the highest RR in the NMA for lesion-specific clearance (RR, 5.08; 95% CI, 2.49-10.33; GRADE, moderate). No NMA was possible for participant partial clearance owing to poor reporting of this outcome.

Conclusions and Relevance

This systematic review and network meta-analysis found that therapy including ALA-PDT, imiquimod, 5%, MAL-PDT, and cryosurgery was associated with significant long-term efficacy in the NMA. This study provides data for a possible use in an evidence-based framework for selecting interventions with sustained lesion clearance.

This network meta-analysis evaluates the long-term outcomes of agents used in treatment of patients with actinic keratosis.

Introduction

Actinic keratoses (AK) are common precancerous lesions of the skin as a consequence of lifelong exposure to UV radiation.^1,2 These lesions usually present as erythematous and keratotic or scaling plaques with a rough, sandpaper-like surface on chronically sun-exposed areas, such as the face, ears, arms, and dorsal hands.^2,3 Actinic keratoses rank among the most common skin lesions, with a prevalence of up to 60% in White individuals older than 60 years, and may progress into invasive cutaneous squamous cell carcinoma (cSCC), although the risk is presumably low for single lesions.⁴ However, if multiple AK are present and accompanied by signs of chronic actinic damage or field cancerization, the risk of malignant conversion increases rapidly.^5,6 Visible lesions may be surrounded by tissue that clinically appears unaltered but bears significant UV-induced histologic and genetic abnormalities. This concept has generally been accepted as field cancerization, although an exact clinical definition has not been determined.⁷ Because it is difficult to estimate whether a lesion will become an invasive cSCC, international guidelines recommend early consequent treatment of AK.^8,9,10

Multiple interventions are available for the treatment of AK in clinical practice, including topical drugs and ablative modalities. Numerous studies reported that most interventions are superior to placebo in terms of lesion clearance and improving the cosmetic image.¹¹ However, most randomized clinical trials (RCTs) and meta-analyses focused on short-term outcomes that are evaluated within 3 to 6 months after treatment,^{12,13,14,15,16} although AK is increasingly being considered a chronic condition and reducing the incidence of cSCC should be the ultimate goal of treatment. In addition, most treatment modalities were investigated vs placebo and head-to-head comparisons are widely lacking, limiting the possibility to cross compare distinct active treatments. To this end, no evidence-based recommendation regarding the long-term efficacy of interventions for AK exists.⁸

Methods

This review was conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) reporting guideline and its extension for network meta-analyses (NMA) (PRISMA-NMA).^17,18,19,20 The protocol for this review was defined a priori and registered online in the PROSPERO international prospective register of systematic reviews (CRD42018095903). The methods of this systematic review were prespecified in a published protocol.²¹ The study was completed February 27, 2021.

We included RCTs with an interindividual (parallel-arm) design. Dose-finding studies, pseudorandomized trials, crossover studies, observational studies, retrospective studies, and case series were excluded. Randomized clinical trials with an intraindividual design (ie, treatment applied to opposite sides of the face or scalp) were not considered owing to statistical issues that may arise from pooling interindividual and intraindividual trials in NMA.²² Outcomes were assessed 12 months after the end of treatment or later as a proxy for long-term efficacy. Thus, the observation time of the included studies had to be at least 12 months for inclusion. In the interim (between the end of treatment and outcome assessment) no additional intervention was allowed. For all outcomes, the intention-to-treat population of the primary study was used for the analysis.

We included adults (age ≥18 years) with a clinical or histopathologic diagnosis of AK. Studies investigating immunocompromised patients or organ transplant recipients were not included because they are likely to have a treatment response distinct from immunocompetent populations.¹² Furthermore, there is evidence that the natural course of AK is less favorable in immunocompromised individuals with a higher likelihood of conversion to cSCC and lower rates of spontaneous resolution.¹¹

Types of Interventions

The following interventions with practical relevance were eligible: surgical approaches (eg, excisional biopsies, shave excision); cryosurgery; cryopeeling; ablative laser treatment (eg, erbium:YAG or carbon dioxide laser); ingenol mebutate, 0.015% or 0.05%, gel; imiquimod, 2.5%, 3.75%, or 5%, cream; fluorouracil, 5%, cream; fluorouracil, 0.5%, plus salicylic acid, 10%, in solution; diclofenac, 3%, in hyaluronic acid; and photodynamic therapy (PDT) with aminolevulinate (ALA) or its ester (MAL) with illumination from light-emitting diodes or natural daylight. For the NMA, either placebo or the active intervention with the smallest treatment effect served as a comparison.

All main outcomes of interest were dichotomous: (1) participant complete clearance, defined as the number of patients with 100% cleared lesions; (2) participant partial clearance, defined as the number of patients who had at least 75% cleared lesions; and (3) lesion-specific clearance, defined as the number of cleared lesions compared with baseline assessment. The end points had to be reported at least 12 months after completion of treatment or later. If several time points were reported from a primary study, data from the last reported measurement time were extracted and referred to the intention-to-treat population (Table 1).

Table 1. Study Characteristics of the 15 Included Randomized Clinical Trials.

Study ID	Source		Design	Blinding	Sample size, No. a	Localization of lesions	Time of assessment after the end of treatment	Intervention
	Original report	Follow-up report
NCT02799082 (ALA-AK-CT003)	Szeimies et al,23 2010	Dirschka et al,24 2013	Multicenter	Double-blind	122	Face, bald scalp	12 mo	1 (n = 81) or 2 (n = 39) courses of ALA-PDT (37 J/cm2, 50-70 mW/cm2, 630-nm peak wavelength or 170 J/cm2, 196 mW/cm2, wavelength <595 nm); 1 (n = 41) or 2 (n = 34) courses of placebo-PDT treatments (37 J/cm2, 50-70 mW/cm2, 630-nm peak wavelength or 170 J/cm2, 196 mW/cm2, wavelength <595 nm)
NCT02799069 (ALA-AK-CT002)	Dirschka et al,25 2012	Dirschka et al,24 2013	Multicenter	Investigator-blinded	571	Face, bald scalp	12 mo	1 (n = 248) or 2 (n = 123) courses of ALA-PDT (37 J/cm2 and 630 nm peak wavelength or 100 J/cm2 and wavelength 580-1400 nm or 100 J/cm2 and wavelength 600-750 nm); 1 (n = 76) or 2 (n = 68) courses of placebo-PDT (37 J/cm2, 630 nm, and peak wavelength or 100 J/cm2 and wavelength 580-1400 nm or 100 J/cm2 and wavelength 600-750 nm); 1 (n = 247) or 2 (n = 150) MAL-PDT treatments (37 J/cm2 and 630 nm peak wavelength or 100 J/cm2 and wavelength 580-1400 nm or 100 J/cm2 and wavelength 600-750 nm)
Unknown	Foley et al,26 2011	Foley et al,26 2011	Multicenter	Open-label	71	Face, scalp	12 mo	Cryotherapy with liquid nitrogen for 10 s freeze/thaw cycles; persisting lesions were treated at 3, 6, and 9 mo; spray applicator to produce a 1- to 2-mm rim of frozen tissue beyond the outline; application of imiquimod, 5% (20 cm2), 3 times/wk for 3-4 wk; patients with persisting AK underwent another treatment course
NCT00668733	Hanke et al,27 2011	Hanke et al,27 2011	Multicenter	Double-blind	179	Face, balding scalp	14 mo	Application of imiquimod, 3.75%, cream once daily for two 2-wk cycles or 3-wk cycles; application of imiquimod, 2.5%, cream once daily for two 2-wk cycles or 3-wk cycles; application of vehicle cream once daily for two 2-wk cycles or 3-wk cycles
NCT02281682	Jansen et al,28 2019	Jansen et al,28 2019	Multicenter	Single-blind	624	Head, neck	12 mo	Application of fluorouracil, 5%, twice daily for 4 wk; application of imiquimod, 5%, cream 3 times/wk for 4 wk; application of MAL under occlusion for 3 h, followed by illumination (632 nm, 37 J/cm2); application of ingenol mebutate, 0.015%, gel for 3 consecutive days once daily on a maximum area of 100 cm2
Unknown	Jorizzo et al,29 2007	Jorizzo et al,29 2007	Multicenter	Double-blind	246	Head	12-14 mo	Application of imiquimod, 5%, cream once daily 3 times/wk for 4 wk, followed by a 4-wk posttreatment period; patients with persisting AK underwent another treatment course; application of vehicle cream once daily 3 times/wk for 4 wk, followed by a 4-wk posttreatment period; patients with persisting AK underwent another treatment course
Unknown	Krawtchenko et al,30 2007	Krawtchenko et al,30 2007	Single-center	Open-label	75	Head, neck, décolleté	12 mo	Cryosurgery with liquid nitrogen spray for 20-40 s for each lesion; the second session was performed when the treated lesion was insufficiently cleared within 2 wk after the first cryosurgery
Unknown	Ostertag et al,31 2006	Ostertag et al,31 2006	Single-center	Open-label	55	Face, scalp	12 mo	Application of fluorouracil, 5%, cream twice daily for 4 wk; laser ablation in Er:YAG mode in combination with CO2; energy given varied from 7 to 28 J/cm2 in the erbium mode, 10-12 pulses/s with 50% CO2, varying from 2 to 4 W
NCT00847912	Pomerantz et al,32 2015	Pomerantz et al,32 2015	Multicenter	Double-blind	954	Face	12 mo	Application of fluorouracil, 5%, cream twice daily for 4 wk; application of vehicle cream twice daily for 4 wk
Unknown	Stockfleth et al,33 2002	Stockfleth et al,34 2004	Multicenter	Double-blind	36	Face, scalp, dorsal forearms	12 mo (vehicle)-18 mo (imiquimod)	Application of imiquimod, 5%, cream 3 times/wk for 12 wk or until clearance; application of vehicle cream 3 times/wk for 12 wk or until clearance
NCT00987246	Stockfleth et al,35 2012	Stockfleth et al,35 2012	Multicenter	Double-blind	470	Face, balding scalp	12 mo	Application of fluorouracil with salicylic acid once daily until clearance or for a maximum of 12 wk; application of vehicle once daily until clearance or for a maximum of 12 wk; application of diclofenac sodium, 3%, in hyaluronic acid, 2.5% gel twice daily until clearance or for a maximum of 12 wk
NCT00308854 (AK 03)	Hauschild et al,36 2009	Szeimies et al,37 2010	Multicenter	Double-blind	103	Face, scalp	12 mo	Application of ALA patch for 4 h, followed by illumination with red light 37 J/cm2 at 630 ± 3 nm; application of placebo patch for 4 h, followed by illumination with red light 37 J/cm2 at 630 ± 3 nm
NCT00308867 (AK 04)	Hauschild et al,36 2009	Szeimies et al,37 2010	Multicenter	Double-blind for PDT; open-label for cryosurgery	346	Face, scalp	12 mo	Application of ALA-patch for 4 h, followed by illumination with red light 37 J/cm2 at 630 ± 3 nm; application of placebo-patch for 4 h, followed by illumination with red light 37 J/cm2 at 630 ± 3 nm; 1 cycle of cryosurgery using liquid nitrogen open-spraying procedure for a maximum of 10 s
Unknown	Zane et al,38 2014	Zane et al,38 2014	Single-center	Open-label	200	Face, scalp	12 mo	Application of MAL for 3 h under occlusion, followed by illumination for 8 min with 37 J/cm2 red light; application of diclofenac sodium, 3%, in hyaluronic acid, 2.5%, gel twice daily for 90 d
Unknown	Zane et al,39 2014	Zane et al,39 2014	Single-center	Open-label	200	Face, scalp	12 mo	A single session of CO2 laser ablation (application in char-free mode, using 500 l-s pulses at a power of 23 W with a 50-Hz repetition rate); application of liquid nitrogen on the lesion surface for 10-20 s

Abbreviations: AK, actinic keratosis; ALA-PDT, photodynamic therapy with aminolevulinate; CO₂, carbon dioxide; MAL-PDT, photodynamic therapy with methyl-aminolevulinate.

We searched MEDLINE, Embase (both via Ovid), and the Cochrane library CENTRAL until April 6, 2020. The search strategies are reported in the eTable in the Supplement. In addition, we searched the following trial registers for the key words actinic keratosis or actinic keratoses: the metaRegister of Controlled Trials, US National Institutes of Health Ongoing Trials Register, Australian New Zealand Clinical Trials Registry, World Health Organization International Clinical Trials Registry Platform, and EU Clinical Trials Register. For ongoing trials and completed trials without data publication, principal investigators or trial sponsors were contacted to obtain preliminary or unpublished data. Reference lists of included records also were screened.

Two of us (T.S., M.V.H.) independently screened titles and abstracts that were identified in the electronic database searches for eligibility. Trial registers were hand searched and assessed for eligibility by 1 of us (T.S.). For records that were considered relevant according to title and abstract screening, full-text articles were obtained and inclusion and exclusion criteria were applied. Whenever discrepancies arose, a resolution was achieved by discussion with 1 of us (C.B.).

Information for each included study regarding design, baseline characteristics, intervention, outcomes, and risk of bias was collected and summarized by 2 of us independently (T.S., M.V.H.) using an internally piloted spreadsheet in Microsoft Excel 2010 (Microsoft Corp). To assess treatment effects and effect sizes, outcomes are expressed as risk ratios (RRs) with 95% CIs. Each patient was the unit of analysis for participant complete and partial clearance, whereas single AK lesions were the unit of analysis for lesion-specific clearance.

Because we expected to include distinct comparisons of interventions, direct and indirect evidence were combined with an NMA assuming consistency of the included studies. We decided to use a frequentist approach using the package netmeta, which is a graph-theoretical model for NMA. This model is appropriate for NMA with few trials per comparison, as in our study. Furthermore, the bias of the model is estimated to be lower than in other approaches as long as the heterogeneity is moderate. Compared with bayesian models, the frequentist approach reveals more accurate and narrower CIs, which may facilitate decision-making for clinicians.^40,41,42 To account for unexplained heterogeneity, we calculated a random-effects model.⁴⁰ Before analysis, data were imported from the collection sheet in an arm-based format, enabling the inclusion of multiarm trials.⁴³ Thereafter, data were transformed from the arm-based to a contrast-based format, which is the necessary input format for the netmeta. For the main analyses, we chose placebo as the reference treatment.

Treatments were ranked with the P score method using the function netrank in which treatments are ranked based on the NMA.^44,45 The P scores were based solely on the point estimates and SEs of the network estimates. The P scores measured the extent of certainty that treatment was better than another intervention, averaged over all competing treatments.

Direct and indirect evidence were investigated by node splitting with the function netsplit, and inconsistency parameters were calculated to explore inconsistency.⁴⁶ Furthermore, net-heat plots were created by the function netheat as described previously and analyzed visually to locate spots accounting for inconsistency.⁴⁷ Forest plots and 2-dimensional graphic presentations of the network structure were created with the same package in R, version 3.4.0 (R Foundation).⁴⁸ Heterogeneity among trials was quantified using the I² statistic.⁴⁹

The risk of bias was assessed for each study independently by 2 of us (T.S., M.V.H.) by judgment according to the revised Cochrane risk-of-bias tool for randomized trials.⁵⁰ If at least 10 RCTs reported a specific comparison, publication bias was assessed with comparison-adjusted funnel plots whose asymmetry was tested for significance with the Egger test.⁵¹ The certainty of the evidence for the outcomes was evaluated using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) for NMA.^52,53,54 A detailed summary of the GRADE process is available in the protocol.²¹

Results

Our literature search identified 2741 references. After title and abstract screening and removal of duplicates, 37 records underwent full-text review. Of these articles, most were dismissed because they did not meet the eligibility criteria regarding study design (n = 14) or outcome (n = 1), or data were incompletely reported regarding long-term follow-up (n = 5). In addition, 4 more duplicates were identified. Cross-referencing of the full texts revealed 4 more records. Seventeen published records (original studies and follow-up reports) referring to data from 15 independent RCTs with an overall sample size of 4252 patients met the eligibility criteria and were included (eFigure 1 in the Supplement; Table 1).^{24,26,27,28,29,30,31,32,34,35,37,38,39} All studies included AK located in the head region, and 1 study included AK located on the dorsal forearms (Table 1).³⁴ All studies were similar in clinical and methodologic features. Thus, we assumed that the transitivity assumption was given for both the NMA and GRADE assessments.

Risk of Bias Assessment

Most studies were graded at low risk for the domains missing outcome data, measurement of the outcome data, randomization process, and deviations from the intended intervention (Figure; Table 2 and Table 3). Six studies were assessed at high risk for the domain selection of the reported results. Thus, the estimations of the overall bias of the included studies were heterogeneous. Because 10 studies were included for the NMA of the outcome participant complete clearance, a comparison-adjusted funnel plot was created showing no risk for publication bias (eFigure 2 in the Supplement).

Figure. Risk of Bias Evaluation of Included Studies.

Summary of the evaluation of the included studies for each risk of bias item. Our judgments about each risk of bias item are presented as percentages across all included studies. For the assessment of the risk of bias, the initial randomized clinical trials were used.

Table 2. Quality of Evidence for the Distinct Outcomes and Comparisons for Interventions of AK.

Comparison	Direct evidence			Indirect evidence		Network meta-analysis
	No. of comparisons	RR (95% CI)	Certainty of evidence	RR (95% CI)	Certainty of evidence	RR (95% CI)	Certainty of evidence
Participant complete clearance
Fluorouracil vs ALA-PDT	0	NA	NA	0.35 (0.07-1.62)	Very lowa,b	0.35 (0.07-1.62)	Very low
Fluorouracil vs CO2 laser	0	NA	NA	2.14 (0.31-14.65)	Very lowa,b	2.14 (0.31-14.65)	Very low
Fluorouracil vs cryosurgery	1	8.33 (0.78-88.86)	Very lowa,b	0.16 (0.03-0.84)	Very lowb,c	0.59 (0.15-2.28)	Very low
Fluorouracil vs diclofenac	0	NA	NA	2.48 (0.24-25.23)	Very lowb,c	2.48 (0.24-25.23)	Very low
Fluorouracil vs imiquimod, 2.5%	0	NA	NA	3.88 (0.63-23.94)	Very lowb,c	3.88 (0.63-23.94)	Very low
Fluorouracil vs imiquimod, 3.75%	0	NA	NA	3.50 (0.58-21.31)	Very lowb,c	3.50 (0.58-21.31)	Very low
Fluorouracil vs imiquimod, 5%	1	0.46 (0.11-1.86)	Lowa,d	0.48 (0.09-2.57)	Very lowb,c	0.47 (0.16-1.37)	Very low
Fluorouracil vs MAL-PDT	0	NA	NA	0.47 (0.08-2.81)	Very lowb,c	0.47 (0.08-2.81)	Very low
Fluorouracil vs placebo	1	1.95 (0.54-7.03)	Very lowb,c	5.46 (0.95-31.28)	Very lowe,f	2.80 (0.99-7.86)	Very low
ALA-PDT vs CO2 laser	0	NA	NA	6.16 (1.04-36.45)	Lowb	6.16 (1.04-36.45)	Low
ALA-PDT vs cryosurgery	1	1.35 (0.37-4.94)	Lowb	3.65 (0.33-40.16)	Very lowd,g,h	1.69 (0.54-5.29)	Low
ALA-PDT vs diclofenac	0	NA	NA	7.15 (1.03-49.54)	Lowb	7.15 (1.03-49.54)	Low
ALA-PDT vs imiquimod, 2.5%	0	NA	NA	11.18 (1.48-84.46)	Very lowb,i,j	11.18 (1.48-84.46)	Very low
ALA-PDT vs imiquimod, 3.75%	0	NA	NA	.09 (1.35-75.29)	Very lowd,k	10.09 (1.35-75.29)	Very low
ALA-PDT vs imiquimod, 5%	0	NA	NA	1.35 (0.35-5.21)	Very lowe,f	1.35 (0.35-5.21)	Very low
ALA-PDT vs MAL-PDT	1	1.51 (0.42-5.50)	Lowb	0.20 (0.00-43.27)	Lowk	1.35 (0.39-4.75)	Low
ALA-PDT vs placebo	1	12.84 (1.99-83.02)	Moderatee	4.76 (0.65-34.59)	Lowk	8.06 (2.07-31.37)	Moderate
CO2 laser vs cryosurgery	1	0.27 (0.07-1.08)	Moderatel	NE	NE	0.27 (0.07-1.08)	Moderate
CO2 laser vs diclofenac	0	NA	NA	1.16 (0.09-14.96)	Moderatel	1.16 (0.09-14.96)	Moderate
CO2 laser vs imiquimod, 2.5%	0	NA	NA	1.81 (0.17-19.66)	Very lowb,i,j	1.81 (0.17-19.66)	Very low
CO2 laser vs imiquimod, 3.75%	0	NA	NA	1.64 (0.15-17.56)	Very lowd,k	1.64 (0.15-17.56)	Very low
CO2 laser vs imiquimod, 5%	0	NA	NA	0.22 (0.04-1.22)	Very lowe,f	0.22 (0.04-1.22)	Very low
CO2 laser vs MAL-PDT	0	NA	NA	0.22 (0.03-1.77)	Lowk	0.22 (0.03-1.77)	Low
CO2 laser vs placebo	0	NA	NA	1.31 (0.20-8.35)	Very lowb,c	1.31 (0.20-8.35)	Very low
Cryosurgery vs diclofenac	0	NA	NA	4.22 (0.49-36.65)	Very lowa,b	4.22 (0.49-36.65)	Very low
Cryosurgery vs imiquimod, 2.5%	0	NA	NA	6.60 (0.94-46.52)	Very lowb,i,j	6.60 (0.94-46.52)	Very low
Cryosurgery vs imiquimod, 3.75%	0	NA	NA	5.95 (0.86-41.46)	Very lowa,b	5.95 (0.86-41.46)	Very low
Cryosurgery vs imiquimod, 5%	2	0.70 (0.22-2.19)	Very lowd,g,h	1.54 (0.12-20.37)	Very lowa,b	0.79 (0.28-2.26)	Very low
Cryosurgery vs MAL-PDT	0	NA	NA	0.80 (0.16-3.88)	Very lowa,b	0.80 (0.16-3.88)	Very low
Cryosurgery vs placebo	0	NA	NA	4.76 (1.36-16.66)	Very lowb,c	4.76 (1.36-16.66)	Very low
Diclofenac vs imiquimod, 2.5%	0	NA	NA	1.56 (0.11-21.88)	Very lowb,i,j	1.56 (0.11-21.88)	Very low
Diclofenac vs imiquimod, 3.75%	0	NA	NA	1.41 (0.10-19.56)	Very lowd,k	1.41 (0.10-19.56)	Very low
Diclofenac vs imiquimod, 5%	0	NA	NA	0.19 (0.02-1.74)	Very lowe,f	0.19 (0.02-1.74)	Very low
Diclofenac vs MAL-PDT	1	0.19 (0.04-0.83)	Moderatel	NE	NE	0.19 (0.04-0.83)	Moderate
Diclofenac vs placebo	0	NA	NA	1.13 (0.13-9.89)	Lowk	1.13 (0.13-9.89)	Low
Imiquimod, 2.5%, vs imiquimod, 3.75%	1	0.90 (0.23-3.47)	Very lowd,i,j	NE	NE	0.90 (0.23-3.47)	Very low
Imiquimod, 2.5%, vs imiquimod, 5%	0	NA	NA	0.12 (0.02-0.72)	Very lowe,f	0.12 (0.02-0.72)	Very low
Imiquimod, 2.5%, vs MAL-PDT	0	NA	NA	0.12 (0.01-1.08)	Very lowb,i,j	0.12 (0.01-1.08)	Very low
Imiquimod, 2.5%, vs placebo	1	0.72 (0.16-3.22)	Very lowb,i,j	NE	NE	0.72 (0.16-3.22)	Very low
Imiquimod, 3.75%, vs imiquimod, 5%	0	NA	NA	0.13 (0.02-0.78)	Very lowd,k	0.13 (0.02-0.78)	Very low
Imiquimod, 3.75%, vs MAL-PDT	0	NA	NA	0.13 (0.02-1.18)	Very lowd,k	0.13 (0.02-1.18)	Very low
Imiquimod, 3.75%, vs placebo	1	0.80 (0.18-3.51)	Very lowd,k	NE	NE	0.80 (0.18-3.51)	Very low
Imiquimod, 5%, vs MAL-PDT	0	NA	NA	1.01 (0.19-5.30)	Very lowe,f	1.01 (0.19-5.30)	Very low
Imiquimod, 5%, vs placebo	2	6.85 (1.95-24.05)	Very lowe,f	4.88 (1.05-22.77)	Very lowb,c	5.98 (2.26-15.84)	Very low
MAL-PDT vs placebo	1	8.50 (1.31-55.28)	Lowk	2.30 (0.11-49.10)	Lowb	5.95 (1.21-29.41)	Low
Participant partial clearance
Cryosurgery vs fluorouracil	0	NA	NA	0.30 (0.08-1.07)	Low	0.30 (0.08-1.07)	Low
Cryosurgery vs imiquimod, 5%	1	0.42 (0.12-1.48)	Lowd,m	NE	NE	0.42 (0.12-1.48)	Low
Cryosurgery vs ingenol mebutate	0	NA	NA	0.77 (0.21-2.84)	Lowd,m	0.77 (0.21-2.84)	Low
Cryosurgery vs MAL-PDT	0	NA	NA	0.56 (0.15-2.06)	Lowd,m	0.56 (0.15-2.06)	Low
Fluorouracil vs imiquimod, 5%	1	1.41 (1.17-1.71)	High	NE	NE	1.41 (1.17-1.71)	High
Fluorouracil vs ingenol mebutate	1	2.60 (1.97-3.44)	High	NE	NE	2.60 (1.97-3.44)	High
Fluorouracil vs MAL-PDT	1	1.91 (1.51-2.40)	High	NE	NE	1.91 (1.51-2.40)	High
Imiquimod, 5%, vs ingenol mebutate	1	1.85 (1.36-2.50)	High	NE	NE	1.85 (1.36-2.50)	High
Imiquimod, 5%, vs MAL-PDT	1	1.35 (1.04-1.75)	High	NE	NE	1.35 (1.04-1.75)	High
MAL-PDT vs ingenol mebutate	1	0.73 (0.53-1.02)	High	NE	NE	0.73 (0.53-1.02)	High
Lesion-specific clearance
Fluorouracil vs fluorouracil with salicylic acid	0	NA	NA	1.72 (0.65-4.59)	Very lowd,c	1.72 (0.65-4.59)	Very low
Fluorouracil vs ALA-PDT	0	NA	NA	0.31 (0.13-0.76)	Very lowd,c	0.31 (0.13-0.76)	Very low
Fluorouracil vs cryosurgery	0	NA	NA	0.44 (0.20-0.97)	Very lowd,c	0.44 (0.20-0.97)	Very low
Fluorouracil vs diclofenac	0	NA	NA	2.12 (0.85-5.30)	Very lowd,c	2.12 (0.85-5.30)	Very low
Fluorouracil vs imiquimod, 5%	0	NA	NA	0.56 (0.19-1.68)	Very lowd,n	0.56 (0.19-1.68)	Very low
Fluorouracil vs ablative laser	1	0.84 (0.39-1.79)	Very lowd,n	1.12 (0.28-4.48)	Very lowd,c	0.90 (0.46-1.74)	Very low
Fluorouracil vs MAL-PDT	0	NA	NA	0.38 (0.14-1.00)	Very lowd,c	0.38 (0.14-1.00)	Very low
Fluorouracil vs placebo	1	1.70 (0.80; 3.62)	Very lowd,c	1.28 (0.32-5.11)	Very lowd,n	1.59 (0.82-3.09)	Very low
Fluorouracil with salicylic acid vs ALA-PDT	0	NA	NA	0.18 (0.07-0.47)	Very lowd,o	0.18 (0.07-0.47)	Very low
Fluorouracil with salicylic acid vs cryosurgery	0	NA	NA	0.26 (0.10-0.69)	Very lowd,o	0.26 (0.10-0.69)	Very low
Fluorouracil with salicylic acid vs diclofenac	1	1.06 (0.50-2.26)	Very lowd,o	14.91 (0.68-327.06)	Very lowd,o	1.23 (0.59-2.57)	Very low
Fluorouracil with salicylic acid vs imiquimod, 5%	0	NA	NA	0.33 (0.09-1.13)	Very lowd,o	0.33 (0.09-1.13)	Very low
Fluorouracil with salicylic acid vs ablative laser	0	NA	NA	0.52 (0.18-1.50)	Very lowd,o	0.52 (0.18-1.50)	Very low
Fluorouracil with salicylic acid vs MAL-PDT	0	NA	NA	0.22 (0.09-0.56	Very lowd,o	0.22 (0.09-0.56	Very low
Fluorouracil with salicylic acid vs placebo	1	1.08 (0.50-2.30)	Very lowd,o	0.08 (0.00-1.67)	Very lowd,o	0.92 (0.44- 1.93)	Very low
ALA-PDT vs cryosurgery	1	1.25 (0.59-2.68)	Moderated	2.93 (0.44-19.74)	Moderated	1.41 (0.70-2.86)	Moderate
ALA-PDT vs diclofenac	0	NA	NA	6.77 (3.05-15.00)	Very lowd,o	6.77 (3.05-15.00)	Very low
ALA-PDT vs imiquimod, 5%	0	NA	NA	1.79 (0.63-5.06)	Lowd,m	1.79 (0.63-5.06)	Low
ALA-PDT vs ablative laser	0	NA	NA	2.86 (1.17-7.00)	Very lowb,l	2.86 (1.17-7.00)	Very low
ALA-PDT vs MAL-PDT	1	1.62 (0.76-3.48)	Moderated	0.43 (0.12-1.74)	Moderatel	1.20 (0.61-2.34)	Moderate
ALA-PDT vs placebo	1	3.16 (1.30-7.71)	Moderated	11.58 (3.57-37.53)	Moderatee	5.08 (2.49-10.33)	Moderate
Cryosurgery vs diclofenac	0	NA	NA	4.80 (1.97-11.69)	Very lowd,o	4.80 (1.97-11.69)	Very low
Cryosurgery vs imiquimod, 5%	1	1.27 (0.59-2.72)	Lowd,m	NE	NE	1.27 (0.59-2.72)	Low
Cryosurgery vs ablative laser	1	2.19 (0.98-4.90)	Very lowb,l	1.64 (0.42-6.40)	Very lowd,n	2.03 (1.02-4.06)	Very low
Cryosurgery vs MAL-PDT	0	NA	NA	0.85 (0.35-2.07)	Moderated	0.85 (0.35-2.07)	Moderate
Cryosurgery vs placebo	1	2.52 (1.03-6.16)	Moderatee	6.64 (2.06-21.39)	Moderated	3.60 (1.77-7.33)	Moderate
Diclofenac vs imiquimod, 5%	0	NA	NA	0.26 (0.08-0.85)	Very lowd,o	0.26 (0.08-0.85)	Very low
Diclofenac vs ablative laser	0	NA	NA	0.43 (0.16-1.13)	Very lowd,o	0.43 (0.16-1.13)	Very low
Diclofenac vs MAL-PDT	1	0.12 (0.05-0.28)	Moderatel	0.47 (0.12-1.82)	Very lowd,o	0.18 (0.09-0.36)	Lowp
Diclofenac vs placebo	1	1.01 (0.47-2.17)	Very lowd,o	0.27 (0.07-1.09)	Very lowd,o	0.75 (0.38-1.46)	Very low
Imiquimod vs ablative laser	0	NA	NA	1.60 (0.57-4.47)	Very lowb,l	1.60 (0.57-4.47)	Very low
Imiquimod 5% vs MAL-PDT	0	NA	NA	0.67 (0.20-2.15)	Lowd,m	0.67 (0.20-2.15)	Low
Imiquimod, 5%, vs placebo	0	NA	NA	2.83 (1.00-8.02)	Lowd,m	2.83 (1.00-8.02)	Low
Ablative laser vs MAL-PDT	0	NA	NA	0.42 (0.15-1.15)	Very lowb,l	0.42 (0.15-1.15)	Very low
Ablative laser vs placebo	0	NA	NA	1.77 (0.81-3.90)	Very lowd,n	1.77 (0.81-3.90)	Very low
MAL-PDT vs placebo	0	NA	NA	4.24 (1.91-9.41)	Moderated	4.24 (1.91-9.41)	Moderate

Abbreviations: AK, actinic keratosis; ALA, aminolevulinate; MAL, methyl aminolevulinate; NA, not available; NE, not estimable because the intervention was not connected in a loop in the evidence network; PDT, photodynamic therapy; RR, risk ratio.

^aThe study by Krawtchenko et al³⁰ showed some concerns for bias; mainly in the domain deviations from the intended interventions (−1).

^bLarge CIs crossing the line of no effect (−2).

^cThe study by Pomerantz et al³² was at high risk of bias owing to the rating of the domains deviations from the intended interventions and selection of the reported results (−2).

^dConfidence interval is crossing the line of no effect (−1).

^eLarge CI (−1).

^fThe studies by Jorizzo et al²⁹ and Stockfleth et al³⁴ were estimated to be at high risk for overall bias.

^gThe studies by Foley et al²⁶ and Krawtchenko et al³⁰ were at unclear risk of bias (−1).

^hI²>70% (−2).

ⁱImiquimod, 2.5%, is licensed in only a few countries, and thus is not generalizable (−1).

^jVery wide CI (−2).

^kThe study by Hanke et al,²⁷ was at high risk for bias (−2).

^lThe study by Zane et al³⁹ had an unclear risk of bias (−1).

^mThe study by Foley et al²⁶ was at unclear risk for bias (−1).

ⁿThe study by Ostertag et al³¹ was rated at high risk of bias in the domains measurement of outcome and selection of the reported result (−2).

^oThe study by Stockfleth et al³⁵ was rated at a high risk of bias owing to the domain selection of the reported result (−2).

^pIncoherence.

Table 3. Risk of Bias Evaluation for Each Included Study.

Source	Randomization process	Deviations from intended intervention	Missing outcome data	Measurement of the outcome	Selection of the reported result	Overall bias
NCT027990823,24	Low risk	Low risk	Low risk	Low risk	Some risk	Low risk
NCT0279906924,25	Low risk	Low risk	Low risk	Low risk	Low risk	Low risk
Foley et al,26 2011	Low risk	Low risk	Low risk	Some risk	Low risk	Some risk
NCT0066873327	Some risk	Low risk	Low risk	Low risk	High risk	High risk
NCT0228168228	Low risk	Low risk	Low risk	Low risk	Low risk	Low risk
Jorizzo et al,29 2007	Some risk	Low risk	High risk	Low risk	High risk	High risk
Krawtchenko et al,30 2007	Low risk	Some risk	Low risk	Low risk	Low risk	Some risk
Ostertag et al,31 2006	Low risk	Some risk	Low risk	High risk	High risk	High risk
NCT0084791232	Low risk	High risk	Low risk	Low risk	High risk	High risk
Stockfleth et al,33 2002	Low risk	High risk	Low risk	Low risk	High risk	High risk
NCT0098724635	Some risk	Low risk	Low risk	Low risk	High risk	High risk
NCT0030885436,37	Some risk	Low risk	Low risk	High risk	Low risk	Some risk
NCT0030886736,37	Some risk	Some risk	Low risk	Some risk	Low risk	Some risk
Zane et al,38 2014	Some risk	Some risk	Low risk	Low risk	Low risk	Some risk
Zane et al,39 2014	Some risk	Low risk	Low risk	Low risk	Low risk	Some risk

Outcomes of Interest

Ten studies with 16 direct comparisons investigating 10 distinct interventions were included for the participant complete clearance outcome. Pairwise comparisons were derived mainly from 1 trial each. Data from 2 trials were available for the comparisons of cryosurgery vs imiquimod, 5%, and imiquimod, 5%, vs placebo (eFigure 3A in the Supplement).

The studies were included in an NMA with placebo as reference intervention (I² = 67.7%). Heterogeneity was mainly due to differences in study designs derived from 2 trials.^26,30 ALA-PDT (RR, 8.06; 95% CI, 2.07-31.37; GRADE, moderate); imiquimod, 5% (RR, 5.98; 95% CI, 2.26-15.84; GRADE, very low); MAL-PDT (RR, 5.95; 95% CI, 1.21-29.41; GRADE, low); and cryosurgery (RR, 4.76; 95% CI, 1.36-16.66; GRADE, very low) were significantly superior to placebo in the NMA based on a random-effects model (eFigure 3B and C in the Supplement). ALA-PDT showed the highest probability of being the best treatment among all interventions in this NMA, followed by imiquimod, 5%, and MAL-PDT (eFigure 3D in the Supplement). Direct and indirect evidence were consistent for all comparisons except for fluorouracil vs cryosurgery. For this comparison, disagreement of direct and indirect evidence was statistically significant (P = .005) and visually detected in the net-heat plot (eFigure 3E and F in the Supplement).

Participant partial clearance was reported only in 2 RCTs.^26,28 Owing to the low number of studies reporting this outcome, an NMA was not performed. In a 4-armed, head-to-head trial, fluorouracil (69.7%) was significantly superior to imiquimod, 5% (49.4%), MAL-PDT (36.5%), and ingenol mebutate (26.8%) (Table 2).²⁸ In a 2-armed trial, the participant partial clearance was 8.3% for cryosurgery and 20.0% for imiquimod, 5%, albeit without statistical significance.²⁶ Another trial suggested partial clearance for patients treated with fluorouracil, although the reported data did not match our definition of this outcome.³²

Eight studies with 12 direct comparisons investigating 9 distinct interventions were included for examination of lesion-specific clearance. ALA-PDT (RR, 3.16; 95% CI, 1.30-7.71; GRADE, moderate) and cryosurgery (RR, 2.52; 95% CI, 1.03-6.16; GRADE, moderate) were significantly superior to placebo, and diclofenac was significantly inferior to MAL-PDT (RR, 0.12; 95% CI, 0.05-0.28; GRADE, moderate). All other interventions did not show a significant difference in the treatment effect regarding this outcome derived from the pairwise comparisons (eFigure 4A in the Supplement).

The studies were included in an NMA with placebo as reference intervention (I² = 92.1%). Heterogeneity was mainly due to differences in study designs derived from 3 trials.^24,35,38 ALA-PDT (RR, 5.08; 95% CI, 2.49-10.33; GRADE, moderate), MAL-PDT (RR, 4.24; 95% CI, 1.91-9.41; GRADE, moderate), cryosurgery (RR, 3.60; 95% CI, 1.77-7.33; GRADE, moderate), and imiquimod, 5% (RR, 2.83; 95% CI, 1.00-8.02; GRADE, low), were significantly superior to placebo in the NMA based on a random-effects model (eFigure 4B and C in the Supplement). ALA-PDT showed the highest probability of being the best treatment among all interventions in this NMA, followed by MAL-PDT and cryosurgery (eFigure 4D in the Supplement). Direct and indirect evidence were consistent for all comparisons of the network, and no major inconsistency was detected in the net-heat plot or with node splitting (eFigure 4E and F in the Supplement).

Discussion

This NMA addressed the long-term efficacy after at least 12 months for interventions for AK. Overall, we analyzed data from 15 RCTs with an interindividual design. We specifically investigated 3 efficacy outcomes that were to be reported at least 12 months after the end of treatment of the trial interventions as a proxy for the long-term efficacy. No other treatment was allowed to account for the efficacy of the initial interventions applied to the intention-to-treat populations. Other NMA for AK were conducted with a focus on short-term efficacy outcomes^15,55,56,57 and special localization subgroups, such as the nonscalp and nonface areas.⁴²

Finding the most appropriate end point to determine the long-term efficacy of interventions for AK is a challenge. There is increasing evidence that AK is a chronic condition showing a variable disease course where even resolution without any active treatment is possible.⁵ Most clinical trials habitually report the short-term clearance of treated lesions or areas at 3 to 6 months as a primary efficacy end point. To assess the long-term efficacy, we investigated clearance rates at a time of at least 12 months in a sense of sustained clearance. Methodically, we referred these rates to the baseline intention-to-treat population to account for the initial trial randomization, although the clearance rates were commonly reported in separate follow-up trials. This approach can result in overly conservative estimates because some patients achieving sustained clearance may have been lost to follow-up, which then would decrease the clearance rates. Another way to examine the long-term efficacy is to look at the recurrence rates of lesions after a period of complete clearance. However, to reliably determine the recurrence rates requires a continuous and spatially mapped observation of treated AK to distinguish whether lesions are either relapsing at the original site of origin or if they arise de novo in adjacent sites. Furthermore, it is debatable whether the 12-month time represents a valid proxy for the long-term and not short-term efficacy. This point is likely too early to capture the risk of progression toward cSCC, which may take several years. In contrast, assessing at later times can introduce bias by a high risk of attrition and intercurrent treatments.

The most clinically relevant readout for the long-term efficacy of interventions might be the prevention of cSCC formation because AK are regarded as precursor lesions for cSCC. The rates and latency of malignant progression have been, however, subject to intense and controversial debate for years. We limited our inclusion criteria to RCTs owing to methodologic considerations and to incorporate the highest possible evidence. Data from nonrandomized observational or postmarketing surveillance studies are another important source for the long-term results of interventions for AK, in particular because they have large sample sizes and long follow-up periods.⁵⁸ However, observational and postmarketing studies are not necessarily randomized, which is a prerequisite for an NMA, and were therefore excluded from our analysis.

The recently published postmarketing surveillance studies LEIDA 1 and 2 compared diclofenac, 3%, gel with imiquimod, 5%, cream regarding long-term clinical outcomes for 36 months.⁵⁹ The primary end points were treatment-induced inhibition of histologic change to grade III AK and the occurrence of invasive cSCC in the treated areas. These end points require a histopathologic assessment of lesions, which may not be performed in other trials, and patient care outside of trials as a diagnosis of AK is usually made on clinical grounds. Short-term clearance and the long-term outcomes were discordant in the LEIDA trials, underlining the importance of a longer follow-up even after initial clearance.⁵⁹ However, because no data on sustained clearance rates were reported, the trials were not included in this analysis.

Recently, a core outcome set was developed proposing a collection of 6 core outcomes that should be reported in future RCTs on AK.⁶⁰ The efficacy outcomes that were developed are complete clearance of AK and the percentage of AK cleared, yet the ideal time to assess these outcomes is to be determined. When long-term follow-up is possible, treatment location-specific incidence and progression of cSCC should be reported, although this is not required in all studies. The long-term efficacy was not regarded as essential for trial reporting and not placed in the innermost circle of an onion model of this core outcome set.⁶⁰ Thus, it remains unclear when and how the long-term efficacy of interventions for AK should be assessed.

We performed 2 separate NMAs for the outcomes participant complete and lesion-specific clearance. The interventions ALA-PDT, MAL-PDT, imiquimod, 5%, cream; and cryosurgery were significantly superior to placebo in both models. Although ALA-PDT showed the most favorable RR and was ranked best among all interventions, the relative efficacy values and treatment rankings must be interpreted with caution. In particular, it remains elusive how to translate the distinct RR values into clinical relevance. We are hesitant to derive hierarchical or algorithmic treatment recommendations from our results. Fluorouracil did not show significant long-term efficacy over placebo for participant complete or lesion-specific clearance. These findings are in contrast to the 4-armed, head-to-head trial comparing fluorouracil with imiquimod, MAL-PDT, and ingenol mebutate, in which fluorouracil showed the highest participant partial clearance.²⁸ This outcome was reported in only 1 additional trial, making an NMA not reasonably feasible. Furthermore, previous NMAs on the short-term efficacy of interventions for AK also identified fluorouracil formulations as the most efficacious.^15,55,56 We conclude that the long-term efficacy of fluorouracil remains unclear despite the high-grade evidence from the trial conducted by Jansen et al.²⁸ Nevertheless, we believe that field-directed treatments offer advantages in sustained clearance and prevention of malignant progression of AK because subclinical changes are also addressed, which may give rise to cSCC.⁶¹

Limitations

This study has limitations. Although sustained clearance rates have not been specifically investigated and synthesized, to our knowledge, both NMAs were poorly connected and the comparisons were mainly derived from few trials. There was unexplained heterogeneity and some inconsistency, in particular, regarding the comparison of fluorouracil vs cryosurgery for the outcome participant complete clearance. Furthermore, field-directed treatments, such as imiquimod, PDT, and fluorouracil, were compared with lesion-directed approaches, such as cryosurgery, which may limit the generalizability of our results.

Conclusions

The performance of NMA appeared to enable the comparison of multiple interventions that were not investigated in head-to-head trials. This study provides data that might contribute to an evidence-based framework to guide the selection of interventions for AK with proven long-term efficacy and sustained AK clearance.

Supplement.

eTable. Search Queries in MEDLINE, Embase (via Ovid), and the Cochrane Library CENTRAL

eFigure 1. PRISMA Flowchart of the Study

eFigure 2. Comparison-adjusted Funnel Plot for the Outcome “participant complete clearance”

eFigure 3. Analysis of the Outcome Participant Complete Clearance

eFigure 4. Analysis of the Outcome Lesion-Specific Clearance