risk factor. No evidence is given that the association is causal, and it should not affect the advice on health ..... deliver is important but may often be forgotten.
associations between a putative risk factor and multifactorial diseases, there are at least two points worth comment. The authors discount the possibility that premature death from ischaemic heart disease accounted for the negative association between cholesterol concentrations and cancer, stating that the average age at death from each cause was similar for each concentration of plasma cholesterol. Thus it seems that the ages at death from cancer in their population followed the same pattern as the ages at death from ischaemic heart disease when grouped according to the plasma cholesterol concentration: those with low cholesterol concentrations lived longer before developing cancer, a curious finding if the negative association is true. Also, the direct relation between smoking and cancer should be addressed, particularly in view of the predominance of lung cancer. It is important to know whether those developing cancer were more likely to be smokers and, if so, whether they had high' or low plasma cholesterol concentrations; these concentrations if compared with those from smokers without cancer, matched for other risk factors, age, and sex, might show whether a low plasma cholesterol concentration is an additional risk factor. No evidence is given that the association is causal, and it should not affect the advice on health matters offered by doctors, especially to patients with other risk factors for cardiovascular disease. Public awareness will make the work of those attempting to reduce the incidence of ischaemic heart disease in "at risk" groups even harder.
the British regional heart study, a prospective study of cardiovascular disease in middle aged men drawn from general practices in 24 towns in England, Wales, and Scotland. The data in the table cover a nine year follow up with a total of 645 deaths in the 7690 men for whom a blood cholesterol measurement was available. It is clear from these data that there is no relation between blood cholesterol and cancer mortality in this study of British men. The relation between blood cholesterol and ischaemic heart disease is positive and similar to that seen in virtually all major prospective studies in men, including the Renfrew and Paisley survey. The suggestion of a high death rate from other causes in the lowest fifth of the cholesterol distribution is puzzling as it relates to a varied collection of diseases. The possible cholesterol lowering effect of certain chronic diseases needs to be considered. At present there is no agreement that lower blood cholesterol values are associated with the development of cancer, and there are sound reasons for continuing to advise reduction of intake of saturated fat in the
population. A G SHAPER A N PHILLIPS S J POCOCK
Department of Clinical Epidemiology and General Practice, Royal Free Hospital School of AMedicinie, London NW3 2PF 1 Isles CG, Hall DJ, Gillis CR, Hawthornc VAM, Lever AF. Plasma cholestcrol, coronary hcart disease, anid canccr in the Renfrew and Paislyc survcv. Br Med j 1989;298:920-4. (8 April.)
J S PRICE Clinical Pharmacology Unit, Addenbrooke's Hospital, Cambridge CB2 2QQ 1 Isles CG, Holc DJ, Gillis CR, Hawthorne VM, Lever AF. Plasma cholesterol, coronarv heart disease, and cancer in the Renfrew and Paisley survey. BrMedJ 1989;298:920-4. (8 April.)
SIR,-Trhe Renfrew and Paisley survey by Dr C G Isles and his colleagues' shows an inverse association between blood cholesterol concentration and deaths from cancer in middle aged men. They also show a positive association between blood cholesterol and mortality from coronary heart disease. On the basis oftheir findings they conclude that a reduction in the intake of saturated fat may well decrease coronary heart disease but "other risks might increase." They also noted that the extensive previous research on this topic was inconclusive and contradictory. In 21 previous reports on blood cholesterol and cancer no relation was observed in eight, an inverse association in 12, and a positive relation with colorectal cancer in one. There was no consistent view on whether a preclinical cancer effect could adequately explain the inverse association when it was found. This seems an important time for other major prospective studies to report their findings on this issue to avoid the publication bias incurred when findings not supporting a specific hypothesis remain unpublished. We present the results from
SIR,-Dr C G Isles and his coworkers have provided interesting evidence that in Renfrew and Paisley low plasma cholesterol is associated with an increased risk of mortality from causes other than coronary heart disease. All cause mortality is therefore not associated with plasma cholesterol.' The interpretation of these findings, however, requires careful consideration. Plasma cholesterol was negatively correlated with low social class in their population, especially among men. Social class is an imperfect measure of socioeconomic inequalities, and heterogeneity would be expected within social class categories as well as between them. If the same negative correlation with plasma cholesterol were present within social class categories you would expect to find an increase in diseases related to deprivation among subjects with low cholesterol even after social class had been adjusted for. It would be interesting to know what diseases were predominant among "other cancers" (for men) and "other causes of death" (for women), specifically whether they have strong social class gradients. There is a second problem. Both plasma cholesterol and low social class were positively correlated with age. That they were themselves negatively correlated suggests the presence of a powerful interactive effect, possibly related to social heterogeneity as a result of selective migration patterns in the population studied. This
Annual mortality (per thousand)* among British men in relation to cholesterol concentration. Numbers of deaths are shown in parentheses Serum cholesterol concentration (mmol/I)
