Challenges in assessing the sunscreen-melanoma association.

Abstract

Whether sunscreen use affects melanoma risk has been widely studied with contradictory results. To answer this question we performed a systematic review of all published studies, accounting for sources of heterogeneity and bias. We searched for original articles investigating the sunscreen‐melanoma association in humans to February 28, 2018. We then used random‐effects meta‐analysis to combine estimates of the association, stratified by study design. Stratified meta‐analysis and meta‐regression were used to identify sources of heterogeneity. We included 21,069 melanoma cases from 28 studies published 1979–2018: 23 case–control (11 hospital‐based, 12 population‐based), 1 ecological, 3 cohort and 1 randomised controlled trial (RCT). There was marked heterogeneity across study designs and among case–control studies but adjustment for confounding by sun exposure, sunburns and phenotype systematically moved estimates toward decreased melanoma risk among sunscreen users. Ever‐ vs. never‐use of sunscreen was inversely associated with melanoma in hospital‐based case–control studies (adjusted odds ratio (OR) = 0.57, 95%confidence interval (CI) 0.37–0.87, p heterogeneity < 0.001), the ecological study (rate ratio = 0.48, 95%CI 0.35–0.66), and the RCT (hazard ratio (HR) = 0.49, 95%CI 0.24–1.01). It was not associated in population‐based case–control studies (OR = 1.17, 95%CI 0.90–1.51, p heterogeneity < 0.001) and was positively associated in the cohort studies (HR = 1.27, 95%CI 1.07–1.51, p heterogeneity = 0.236). The association differed by latitude (p interaction = 0.042), region (p interaction = 0.008), adjustment for naevi/freckling (p interaction = 0.035), and proportion of never‐sunscreen‐users (p interaction = 0·012). Evidence from observational studies on sunscreen use and melanoma risk was weak and heterogeneous, consistent with the challenges of controlling for innate confounding by indication. The only RCT showed a protective effect of sunscreen.