Objective: To study the predictive power of serum selenium with regard to cancer mortality in a large sample of the Belgian population given the lack of coherence in the results of observational epidemiological studies in this domain.
Design: A prospective case-control study within a stratified sample of the Belgian male and female population.
Subjects: A total of 201 cases randomly selected from all cancer deaths (N=343) during a 10-y mortality follow-up of a large age- and sex-stratified sample of the total Belgian population aged 25-74 y were matched for age and gender with 603 controls.
Statistics: Conditional logistic regression for both univariate and multivariate analyses using tertile distribution of serum selenium in controls. Odds ratios (ORs) are adjusted for 10 baseline characteristics.
Results: Unadjusted ORs of cancer deaths taking the highest tertile of serum selenium as a reference: in male subjects T1/T3 is 2.2 (CI 1.3-3.7) (P for trend 0.011), whereas in female subjects a nonsignificant OR of 0.8 is observed. In multivariate analyses, no significant modifications of the ORs are observed for the predictive relation of serum selenium with cancer mortality. Besides serum selenium, beta-carotene intake and smoking are independent predictors in male subjects.
Conclusions: In this nested case-control study of a stratified sample from the Belgian population, serum selenium is an independent predictor of cancer mortality in male subjects only, in a country with rather high serum selenium levels with respect to most other European countries.