In a recent analysis of the VITAL (VITamins And Lifestyle) study cohort,1 Asgari et al2 failed to identify an association between antioxidant supplementation and the risk of incident melanomas. On this basis, they question the association previously observed in the SU.VI.MAX study (Supplementation en Vitamines et Mineraux Antioxydants).3,4 However, the 2 studies do not address the same question. The SU.VI.MAX study was a prospective randomized clinical trial testing the hypothesis that supplementation would influence cancer risk.4 The VITAL study was an observational study testing the hypothesis that individuals who use antioxidant supplementation would differ in their cancer risk from those who do not. The reasons why this does not directly address whether supplementation influences cancer risk are clearly explained in the discussion of the original article describing the VITAL study.1 The individuals who used supplementation may have had a healthier lifestyle or cultivated more health-oriented behavior than those who did not. For example, those using supplementation also exercised more and had a lower body mass index and a healthier diet. Although these confounders can to some extent be accounted for in the risk determination, and certain confounders were indeed included in the reported analysis, the 2 populations are different. It is thus impossible to attribute any difference observed to supplementation or to inherent between-group differences. The only way to study supplementation reliably is through a randomized prospective study where between-group differences are minimized through randomization. This was the strategy pursued in the SU.VI.MAX study. In addition, failure to reject the null hypothesis (no difference between groups) in the VITAL study analysis does not mean that no such difference exists but only that one was not observed (absence of proof is not the same as proof of absence).
Ezzedine K, Hercberg S, Malvy D. Antioxidant Supplementation and Risk of Skin Cancers. Arch Dermatol. 2010;146(5):567-568. doi:10.1001/archdermatol.2010.83