In a recent study low serum total cholesterol was associated with an increased risk of suicide (Reference Partonen, Haukka and VirtamoPartonen et al, 1999). However, the study population was a special subgroup, since the subjects were older male smokers. In addition, the final trial participants were very selected, since the target population included approximately 283 000 subjects, but only 29 133 were recruited (ATBC Cancer Prevention Study Group, 1994).

Continuing our previous research (Reference Vartiainen, Puska and PekkanenVartiainen et al, 1994) and analysing random population samples of Finnish subjects, we prospectively monitored mortality of 18 344 men (aged 25-64 years) through the National Death Register for a mean of 14.6 years. There were 91 suicides among 7649 smokers and 53 suicides among 10 695 non-smokers. In order to replicate the findings of Partonen et al, we classified cholesterol into the same three categories. Using the Cox model the relative risks were adjusted for identical variables except for carbohydrate intake. Among smokers the unadjusted risks (with 95% CIs) of suicide increased from 1.00 to 1.48 (0.63-3.47), and to 1.80 (0.75-4.31) with increasing cholesterol level. The relative hazards changed clearly after adjustment for covariates (1.00, 1.38, 1.62, respectively), but remained non-significant. In the report by Partonen et al, the relative risks did not change at all after adjustment for covariates, which we find surprising. We found no association between cholesterol and suicide in non-smokers.

Inconsistent findings between these two large longitudinal studies may have resulted from several confounding effects. First, 75% of the participants in the ATBC study were treated with alpha-tocopherol alone, beta-carotene alone, or both. It is possible, theoretically, that these antioxidants possess some unknown central nervous system effects. Second, the method of suicide may influence the cholesterol—suicide association. Our own findings implicate that very high serum total cholesterol is associated with the increased risk of violent, but not with non-violent suicide (Reference Tanskanen, Vartiainen and TuomilehtoTanskanen et al, 2000). Third, it has been suggested that cholesterol is only a surrogate marker of changes in dietary polyunsaturated fatty acids, which have been linked to depression (Reference Hibbeln and SalemHibbeln & Salem, 1995) — one of the strongest risk factors for suicide. Probably various other factors also confound this controversial relationship.