Despite a low level of evidence, our systematic review with meta-analyses found that electrosurgery and CO
2 laser are slightly more efficacious than cryotherapy, but that CO
2 laser causes more erosion. No differences in efficacy and side effects were found between cryotherapy and imiquimod or between cryotherapy and TCA. Podophyllotoxin gel was shown to be slightly more efficacious than podophyllotoxin cream. Imiquimod 5% was found to be more efficacious than a placebo, which is in line with a Cochrane review from 2014 [
10]. The slight quantitative differences with our results may be explained by the fact that the Cochrane review examined all imiquimod concentrations (1%, 5%) and that it likely considered overlapping studies as different studies [
63,
64]. In addition, imiquimod was associated with higher recurrence in our review, likely because the Cochrane review included an intention-to-treat (ITT) analysis that identified patients lost to follow-up as presenting no recurrence. Our findings for patient-administered treatments were similar to those of a recent systematic review by Werner et al. [
88]. However, that review included only 18 RCTs (from Europe and North America) and was restricted to patient-administered treatments. Moreover, it did not evaluate 5-FU and, most importantly, podophyllin, despite the fact that the latter remains the older standard to which most other therapeutic strategies are compared. Lastly, our results are globally consistent with those of Thurgar et al. [
7]; in our review, however, 5-FU was associated with higher clearance and lower recurrence than CO
2 laser, and the two treatments induced the same local mild-grade side effects. Note, however, that we considered only immunocompetent adults, whereas Thurgar et al. indiscriminately included immunocompetent and human immunodeficiency virus-positive patients. Moreover, unlike these authors, we examined polyphenon and KOH, even though the paucity of RCTs and the high risk of bias prevented us from determining their respective efficacies.
Given the low-level evidence of the RCTs examined in our review, we wish to make the following recommendations for future studies of AGW treatments. First, recurrence at 3, 6, and 12 months, patient satisfaction, and QOL should be properly addressed in future RCTs, as they constitute important clinical outcomes. Second, side effects that induce treatment interruption should be better characterized, with data on post-intervention intensity and duration, impact on QOL, and impact on compliance (in the case of patient-administered treatments). Third, efficacy analyses should be conducted not on AGWs but on patients themselves [
89‐
93] for two reasons: on the one hand, the primary goal of therapy is complete healing of the patient; on the other, the observed heterogeneity of outcome measures statistically impedes direct and indirect comparisons and, therefore, the development of general recommendations based on available RCTs. Fourth, split studies should not be used to design new AGW treatments, both for statistical reasons and because of the biases induced by the lack of participant blinding [
94]. Indeed, given the prevalence of performance biases identified in our review, future RCTs should ensure that outcome evaluation is systematically blinded via different approaches [
95,
96] and that outcomes are assessed by an independent committee unaware of treatment group assignment. Fifth, treatments with clearly demonstrated lower efficacy (e.g., podophyllin 20–25%) should be definitively excluded from future RCTs. Lastly, medical-economic evaluation of AGW treatments should be systematically performed.
Limitations
The main limitation of this systematic review is the high risk of bias of the overwhelming majority (66/70, 94%) of included RCTs [
14], which prevented us from developing a clinically meaningful hierarchy of first-line treatments. Note, however, that ITT analysis was performed whenever possible as it comes closest to real-life practices. The lack of information on older therapies or AGW location and characteristics (flat, keratinized, etc.) made it impossible to analyze efficacy based on these criteria. Similarly, sensitivity analyses and assessments of publication bias [
97] were not attempted because of the paucity of RCTs. As was the case in other systematic reviews, authors and pharmaceutical companies could not be contacted to obtain unpublished information [
98,
99]. Another important limitation was restricted access to Chinese databases. While direct comparisons are statistically more robust than pooled analyses, the paucity of RCTs comparing several therapies also prevented the establishment of a hierarchy of treatments. In spite of these limitations, our pooled study found lower recurrence at 12 months for patient-administered treatments, suggesting that these are more relevant than provider-administered treatments as a global therapeutic response [
8].