Commentary: Having too much evidence (depression, suicide, and low serum cholesterol)

BMJ 1996; 313 doi: (Published 14 September 1996) Cite this as: BMJ 1996;313:651
  1. Malcolm Law, readera
  1. a Wolfson Institute of Preventive Medicine, Department of Environmental and Preventive Medicine, St Bartholomew's and Royal London School of Medicine, London EC1M 6BQ

    Another observational study on low serum cholesterol concentration and suicide appears in this week's BMJ,1 complementing other BMJ publications in the past two years.2 3 4 These papers all take as their premise that the controversy on low serum cholesterol, depression, and accidents and suicide remains unresolved. They conclude that it remains unresolved.

    The past two years have also seen the publication of two large randomised controlled trials of cholesterol reduction using hydroxymethyl glutaryl coenzyme A reductase inhibitors (“statins”).5 6 These drugs produce large reductions in serum cholesterol concentration (20-30%) and hence are particularly informative on safety. If smaller trials are taken into account as well, published randomised trials of statins have now accumulated 38 700 person years of observation on active treatment and 33 800 person years on placebo. With a total of 11 deaths from accidents and suicide in treated patients and 14 in controls in these trials, there is no indication of hazard. Most important, however, since one can never prove a negative, is the upper limit of the 95% confidence interval of the combined risk estimate from the numbers in these trials. The greatest likely hazard, if there were any, would be small—one death in about 5000 person years of treatment. Few such potent treatments have so much evidence of safety from randomised trials.

    The older trials of cholesterol lowering (average reduction about 10%) are also reassuring. Concern relating to two trials that recorded more deaths from accidents and suicide in treated patients than controls (though neither difference was statistically significant) was resolved by the demonstration that the extra deaths occurred among men who had not taken their allocated tablets (active or placebo).7 There had been a chance allocation of more patients with psychiatric illness on entry to active treatment than to placebo.7 8 The rest of the older trials show no excess8; table 1 summarises the results.

    Table 1

    Deaths from accidents and suicide in randomised controlled trials of reduction of serum cholesterol concentration

    View this table:

    The issue of low serum cholesterol and depression was directly examined in three randomised, placebo controlled trials of statins in which indices of depression were measured in all the participants—a total of 7400 people taking active treatment and 2400 taking placebo.9 10 11 Depression was no more common among those taking active treatment.

    With these trial results, why is there continued publication of observational studies that foster the view that the uncertainty persists? The observational and trial data have tended to be seen in isolation from each other. Some cross sectional studies show an association between low serum cholesterol concentration and depression,12 and, given this, there is also an association between low serum cholesterol and suicide.8 But the associations may arise because low serum cholesterol causes depression or because it is a consequence of depression (simply because depressed people eat less). Cross sectional studies cannot determine which is the cause and which the consequence. Other observational evidence indicates that the low cholesterol concentration is a consequence of the depression since (a) there was no excess mortality with low serum cholesterol in prospective studies of working men (less likely to have serious psychiatric illness on entry to the studies by virtue of being in work),8 (b) the excess mortality in the prospective studies was short term with no significant excess after six years,8 and (c) treating depression has been shown to increase serum cholesterol concentration.13 The recent observational studies have merely introduced variations on the same theme: low serotonin concentrations (which accompany and may cause depression) are, not surprisingly, also associated with low cholesterol,2 14 people who attempt suicide have low serum cholesterol concentrations,3 and, in this week's issue, men with declining serum cholesterol concentrations are particularly likely to commit suicide.1 If these were the only studies one could not distinguish cause from consequence. The randomised trials resolve the matter, providing compelling evidence that low cholesterol concentration does not cause depression, accidents, or suicide.

    As Sherlock Holmes observed, one can have too much evidence: “What was vital was overlaid and hidden by what was irrelevant.”15 The randomised trial data are vital, but, given the trial evidence, observational data that are unable to distinguish cause from consequence have become irrelevant. It is all too familiar to find one vital piece of evidence that resolves an issue being drowned by much other data that serve only to obfuscate, leaving an overall impression of uncertainty. We should, like Holmes, “from all the facts presented to us, pick just those which we deem to be essential.”


    1. 1.
    2. 2.
    3. 3.
    4. 4.
    5. 5.
    6. 6.
    7. 7.
    8. 8.
    9. 9.
    10. 10.
    11. 11.
    12. 12.
    13. 13.
    14. 14.
    15. 15.
    View Abstract