Prediagnostic serum 25-hydroxyvitamin D and melanoma risk

Abstract

Previous studies of serum 25-hydroxyvitamin D (25(OH)D) in relation to melanoma have shown conflicting results. We conducted a nested case-control study of 708 cases and 708 controls, using prediagnostically collected serum, to study 25(OH)D and melanoma risk in the population-based Janus Serum Bank Cohort. Stratified Cox regression was used to estimate hazard ratios (HRs) with 95% confidence intervals (CIs) adjusted for ultraviolet radiation (UVR) indicators and stratified by ambient UVB of residence and body mass index (BMI). Non-linear associations were studied by restricted cubic splines. Missing data were handled with multiple imputation by chained equations. We found an HR of melanoma risk of 1.01 (95% CI: 0.99, 1.04) and an HR_imputed of 1.02 (95% CI: 1.00, 1.04) per 5-nmol/L increase. The spline model showed exposure-risk curves with significantly reduced melanoma risk between 60 and 85 nmol/L 25(OH)D (reference 50 nmol/L). Non-significant J-shaped curves were found in sub-analyses of subjects with high ambient UVB of residence and of subjects with BMI < 25 kg/m². Our data did not yield persuasive evidence for an association between 25(OH)D and melanoma risk overall. Serum levels within the medium range might be associated with reduced risk, an association possibly mediated by BMI.

Figure 1

Overview of study design (selection of cases and controls) and exclusions.

Figure 2

Restricted cubic splines displaying hazard ratios of melanoma with 95% confidence intervals according to prediagnostic serum 25-hydroxyvitamin D in the Janus Serum Bank Cohort, Norway, 1972–2009. Complete case sample: 607 cases and 607 controls. Reference set to 50 nmol/L. Knots located at 36.35, 55.82, 69.83, 85.9, 121.82 nmol/L (first and last at 5 and 95 percentile, the remaining equally spaced), P value for non-linearity 0.15. Adjusted for body mass index, body surface area, lifetime ambient ultraviolet (UV)-B, lifetime sunburns, lifetime sunbathing vacations, occupational UV radiation exposure, education, physical activity, smoking status.

Figure 3

Restricted cubic splines displaying hazard ratios (HRs) of melanoma with 95% confidence intervals according to prediagnostic serum 25-hydroxyvitamin D by ambient UVB of residence in the Janus Serum Bank Cohort, Norway, 1972–2009. (A) Splined HRs for north, mid and southwest with knots located at 41.14, 67.48, 106 nmol/L, P value for non-linearity 0.456. (B) Splined HRs for southeast inland and coast with knots located at 41.43, 70.78, 109.96 nmol/L, P value for non-linearity 0.16. (A,B) Complete case sample: 607 cases and 607 controls. Reference set to 50 nmol/L. Adjusted for body mass index, body surface area, lifetime ambient ultraviolet (UV)-B, lifetime sunburns, lifetime sunbathing vacations, occupational UV radiation exposure, education, physical activity, smoking status.

Figure 4

Restricted cubic splines displaying hazard ratios (HRs) of melanoma with 95% confidence intervals according to prediagnostic serum 25-hydroxyvitamin D by body mass index (BMI) in the Janus Serum Bank Cohort, Norway, 1972–2009. The curves are estimated based on linear combinations of the main effect term and the interaction term. Complete case sample: 607 cases and 607 controls. Reference set to 50 nmol/L. Knots located at 41.22, 69.83, 108.43 nmol/L (10, 50 and 90 percentile, the remaining equally spaced). P_interaction 0.04 and 0.01 for the first (knot 1–2) and second (knot 2–3) curve-segments, respectively. P value for non-linearity 0.011. Adjusted for BMI (continuous), body surface area, lifetime ambient ultraviolet (UV)-B, lifetime sunburns, lifetime sunbathing vacations, occupational UV radiation exposure, education, physical activity, smoking status.