A systematic review of 19 studies reported a 15% increased risk of melanoma (95% confidence interval (CI) 1.00-1.31) associated with ever use of sunbeds.1 A recent Australian study by Cust et al. demonstrated an increased risk of earlyonset melanoma (<40 years) associated with ever use of sunbeds (adjusted odds ratio (OR) 1.41, 95% CI 1.01-1.96).
2Concurrently with the Australian study and using the same questionnaire, we investigated the relationship between sunbed use and melanoma at any age in the United Kingdom. A similar estimate in the UK, which has higher sunbed usage, would imply that sunbed usage is a major etiological factor for melanoma.Nine hundred and fifty-nine population-ascertained incident melanoma cases diagnosed from September 2000 to December 2005 (age 17-76 years at diagnosis, 22% <40 years at diagnosis), 513 population-ascertained controls and 174 sibling controls were recruited to a case-control study whereby comprehensive sun exposure data, including a life-long residence calendar, were collected as described previously.3 Participants were asked about sunbed or sunlamp use (ever versus never) and about locations they were used. Data were collected on age at first and last use and number of lifetime sessions. Years since first use was calculated and these variables were categorized as presented by Cust et al.: never, <25, !25 years; none, 1-10, >10 sessions; never, 4, >4 and 14, >14 years, respectively. A proxy for sun sensitivity phenotype (categorized as sun-sensitive or not sun-sensitive) was derived, as described previously.
3As far as possible, we repeated the analyses as reported by Cust et al. Spearman correlations, Wilcoxon rank-sum tests and Pearson chi-squared tests were performed for pair-wise associations. ORs and 95% CIs were calculated from unconditional logistic regression models using data from cases and population-ascertained controls to assess the sunbed variables as predictors of melanoma. Population controls were significantly older than cases (median age diagnosis/interview 58 and 53 years, respectively, p < 0.0001) and more educated (v 2 (3) ¼ 6.9, p ¼ 0.03). Cases were significantly more likely to have family history of melanoma in first or second degree relatives compared with controls (v 2 (1) ¼ 8.0, p ¼ 0.01). The primary analyses comparing cases and population controls were therefore adjusted for age (examined as a trend over quartiles), sex, highest educational level (primary/secondary school, sixth form/vocational training, university/post graduate examined as a trend), sun sensitivity phenotype, selfreported family history in 1st or 2nd degree relatives (none, any) and cumulative lifetime total sun exposure (examined as a trend over quartiles). These analyses were repeated in the subset of 157 cases with matched siblings using conditional logistic regression models, adjusted for all of the above-listed factors except family history. We also performed some subgroup analyses stratifying by the factors defined by Cust et al. (sex, age at diagnosis/interview,...