Non-genetic risk factors for cutaneous melanoma and keratinocyte skin cancers: An umbrella review of meta-analyses.

BACKGROUND: Skin cancers have a complex disease mechanism, involving both genetic and non-genetic risk factors. Numerous meta-analyses have been published claiming statistically significant associations between non-genetic risk factors and skin cancers without applying a thorough methodological assessment.

OBJECTIVE: The present study maps the literature on the non-genetic risk factors of skin cancers, assesses the presence of statistical biases and identifies the associations with robust evidence.

METHODS: We searched PubMed up to January 20, 2016 to identify systematic reviews and meta-analyses of observational studies that examined associations between non-genetic factors and skin cancers. For each meta-analysis, we estimated the summary effect size by random-effects and fixed-effects models, the 95% confidence interval and the 95% prediction interval. We also assessed the between-study heterogeneity (I2 metric), evidence for small-study effects and excess significance bias.

RESULTS: Forty-four eligible papers were identified and included a total of 85 associations. Twenty-one associations were significant at P<10-6 . Fifty-two associations had large or very large heterogeneity. Evidence for small-study effects and excess significance bias was found in fifteen and thirteen associations, respectively. Overall, thirteen associations (actinic keratosis, serum vitamin D, sunburns, and hair color for basal cell carcinoma and density of freckles, eye color, hair color, history of melanoma, skin type, sunburns, premalignant skin lesions, common and atypical nevi for melanoma) presented high level of credibility.

CONCLUSION: The majority of meta-analyses on non-genetic risk factors for skin cancers suffered from large between-study heterogeneity and small-study effects or excess significance bias. The associations with convincing and highly suggestive evidence were mainly focused on skin photosensitivity and phenotypic characteristics.