Effect of publication bias on the evidence supporting the use of myopia control interventions.

Abstract

Purpose: Publication bias exists when studies with negative results are less likely to be published, resulting in over-estimation of treatment efficacy. This study aimed to assess the impact of publication bias on current evidence supporting myopia control interventions.

Methods: A systematic literature search was conducted to retrieve systematic-reviews (SRs) and meta-analyses (MAs) assessing myopia control interventions. From eligible SRs & MAs randomised controlled trials (RCTs) evaluating myopia control treatments (spectacle-based, contact lenses (CL), pharmacological) were included. Outcomes were mean changes in axial-length and refractive error. Pooled estimates of efficacy were computed with random effects meta-analysis. Publication bias was evaluated with funnel plots and Egger's test.

Results: This study included 27 SRs and MAs, reporting on 49 eligible RCTs. Considering axial length outcomes, there were 41 RCTs, assessing CL (n = 16, 39 %), spectacles (n = 13, 31.7 %), and pharmacological treatments (n = 12, 29.3 %). Egger's test was significant only for pharmacological treatments (p = 0.045), but trim and fill analysis indicated that that treatment efficacy was not over-estimated. Of 46 RCTs considering refractive error outcomes, 13 (28.3 %), 14 (30.4 %), and 19 (41.3 %) examined CL, pharmacological and spectacle treatments, respectively. Egger's test showed significant asymmetry only for CL (p = 0.006), but there was not over-estimate of treatment efficacy. Among specific treatments, only atropine (refractive outcomes) and multifocal CL (axial length and refractive outcomes) had sufficient RCTs for analysis, and all Egger's test were not significant (all p > 0.05).

Conclusions: This study did not find evidence of publication bias affecting the current literature supporting myopia control treatments, suggesting that the effect of different interventions is unlikely to be overestimated.