Serum cholesterol concentration and death from suicide in men: Paris prospective study I

SUBJECTS

Between 1967 and 1972, 7746 native born Frenchmen aged 43-52 and employed by the city of Paris were given initial clinical examinations to study risk factors for coronary disease. For up to four years after the first examination, all subjects were invited to undergo an annual follow up examination at which cholesterol concentration was measured. The number of subjects who had one, two, three, four, or five examinations was 614, 604, 601, 1671, and 4456 respectively.

CHOLESTEROL CONCENTRATIONS

For our study, we included the 6728 men whose serum cholesterol concentration had been measured at least three times. We considered each subject's average serum cholesterol concentration for all the measurements and the change in concentration over the period. We used the within person linear regression method to calculate change in individual serum cholesterol concentration. Serum cholesterol concentrations were regressed onto calendar time to give an average annual change in concentration (regression slope).

FOLLOW UP

The men were followed up for survival until January 1994 by specific inquiries to hospitals, family doctors, and medical services at work. Vital status after an average of 17 years of follow up was available for 6393 of the 6728 subjects. Date and place of death were noted, and cause of death was obtained from death certificates at the National Institute of Health and Medical Research (INSERM). Death from suicide was defined according to ICD-8 and ICD-9 (international classification of diseases, eighth and ninth revisions) as codes E950-959.

STATISTICAL ANALYSIS

We used Student's t test and Cox's proportional hazards model for analysis. All analyses were stratified by the number of measurements of serum cholesterol for each subject (three to five).

CHANGE IN SERUM CHOLESTEROL

The average serum cholesterol concentration for the whole population increased significantly over time (mean change 0.06 (SD 0.25) mmol/l a year, P<0.001). In the men who committed suicide the average cholesterol concentration declined, and the difference with those who did not commit suicide was of borderline significance (-0.02 (SD 0.28) (upsilon) 0.06 (SD 0.25) mmol/l a year, P = 0.07). Men whose cholesterol concentration decreased more than 0.13 mmol/l a year had an age adjusted relative risk of suicide of 2.18 (0.98 to 4.87, P = 0.052) compared with those whose cholesterol concentration changed </=0.13 mmol/l a year. Neither adjustment for other variables nor adding average serum cholesterol concentration to the model (as continuous variable or in three categories) altered the results (table 1).

We divided the men according to the direction in which their cholesterol concentration changed: concentration increased in 4098 men, 14 of whom committed suicide, and declined in 2295, 18 of whom committed suicide. The multivariate adjusted relative risk of suicide for the men with decreasing cholesterol concentrations was 2.13 (1.06 to 4.31, P = 0.03) compared with those with increasing concentrations.

When we stratified the cohort according to the average serum cholesterol concentration we found similar associations between change in concentration and suicide, although this was significant only in men with an average cholesterol concentration </=6.21 mmol/l (table 2). We also found that the association between change in cholesterol concentration and suicide remained for those subjects who committed suicide earlier in the follow up period and for those who committed suicide later on (table 2).

SUICIDE AND LIPID LOWERING DRUGS

Trials of primary prevention of coronary heart disease showed that lowering of serum cholesterol concentration, notably from drugs, was followed by an increased risk of suicide and violent deaths,3 though results were inconsistent.4 However, we do not think that the association we found between declining serum cholesterol concentration and suicide could be explained by the action of hypolipidaemic drugs: few people would have been treated with such drugs in the early 1970s, and, more importantly, the relation between declining serum cholesterol and suicide was similar both for those subjects with elevated serum cholesterol concentrations (>6.21 mmol/l), who would have been candidates for treatment, and for those with lower levels (</=6.21 mmol/l), who would not. Furthermore, two recent large trials of the lipid lowering drugs simvastatin and pravastatin did not show an increased risk of violent death and suicide.5 6

OTHER EXPLANATIONS

Low or decreasing cholesterol concentration might be a consequence of depression7 or of decreased appetite and weight loss in depressed subjects. Effective treatment of depression has been shown to increase serum cholesterol concentration.8 An alternative suggestion is that low or declining serum cholesterol concentrations might alter the metabolism of serotonin,9 10 leading to depression and thus increasing the risk of suicide.

Another alternative is that the associations we found could have been due to confounding factors. Although we took account of smoking habits and mean corpuscular volume (as a proxy of alcohol consumption), residual effects of these factors and the effects of other factors could not be ruled out. Unfortunately, we had no information about individual reasons for declining serum cholesterol concentrations. As in the general French population,11 the number of deaths from suicide in this cohort might have been underestimated. However, our findings should not be notably modified unless underestimation was related to serum cholesterol concentration, which is unlikely.

CONCLUSION

We found that low or declining serum cholesterol concentrations were associated with death from suicide. Mechanisms that might link cholesterol concentrations to suicide should be thoroughly studied.

The Paris prospective study I is organised by the Groupe d'Etude sur l'Epidemiologie de l'Atherosclerose. We thank Annie Bingham for technical assistance with the manuscript and Josiane Difolco, Marie-Laurence Henry, and Francine Renard for collecting the mortality data.