Intended for healthcare professionals


Polyunsaturated fat and the risk of cancer

BMJ 1995; 311 doi: (Published 11 November 1995) Cite this as: BMJ 1995;311:1239
  1. Walter C Willett
  1. Professor Departments of Epidemiology and Nutrition, Harvard School of Public Health, Boston, MA 02115, USA

    Little evidence that high intake increases the risk

    Increasing the amount of polyunsaturated fat in people's diets has been widely advocated as a means of reducing their risk of heart disease. Compared with carbohydrates, dietary polyunsaturated fat reduces low density lipoprotein cholesterol and raises high density lipoprotein cholesterol concentrations,1 and in several intervention trials people taking diets that were high in polyunsaturated fat and low in saturated fat had significantly lower rates of coronary heart disease.2 3 In the United States the average intake of polyunsaturated fat as a proportion of total energy intake has increased from about 3% in the 1950s to 6% today. This increase, which is due to both increased consumption of vegetable fats and reduced hydrogenation of vegetable oils, may have contributed to the massive decline in the prevalence of coronary heart disease in the United States over the past 30 years. Further increases in consumption, to up to 10% of total energy intake, have been recommended.4 But the question still remains: are such high levels of polyunsaturated fat in the diet safe?

    The main concern is that diets high in polyunsaturated fat might increase the risk of cancer. This concern is partly theoretical: polyunsaturated fat is prone to oxidation, which generates free radicals. More importantly, animal models have shown that high consumption of polyunsaturated fat increases the incidence of tumours.5 6 In some animal studies the incidence of tumours rises with increasing intake of polyunsaturated fat, up to about 5% of total energy. This level is near the middle of the current range of dietary intake in humans.

    In contrast to animal studies, studies in humans have found little evidence of any association between intake of polyunsaturated or vegetable fat and risk of cancer7 8 or between levels of polyunsaturated fat in adipose tissue and risk of breast cancer.9 But despite this generally reassuring evidence there is still room for concern. High intake of polyunsaturated fat is a new phenomenon, and even small increases in the risk of cancer of the breast, colon, or prostate--say 10-20%--would be important if applied to the total population. The available evidence cannot exclude such modest, long term effects.

    The paper by Zureik et al in this issue does little to resolve the issue (p 1251).10 In this 12 year follow up study of 3277 men aged 36-52 only 59 deaths from cancer occurred, and these deaths were from a variety of different types of cancer. In addition, most of the deaths were probably attributable to smoking and excess alcohol consumption. This number of deaths would need to be increased many times for each type of cancer if realistic increases in risk were to be detected, even assuming a large adverse effect of polyunsaturated fat. The fact that the authors found a significant inverse association between intake of polyunsaturated fat and risk of death from cancer is not reassuring. There is no plausible explanation for such a strong protective effect, and the authors correctly recognise that the inverse association is probably not causal. Many alternative explanations are possible, the most likely being chance. Confounding by other dietary factors is also possible--for example, vegetable oils are also the primary source of vitamin E in most diets. Nor can the study exclude an effect on levels of fatty acid in the blood caused by cancers that had not yet been diagnosed at the time the subjects entered into the study. (The authors do not report whether known cancers were excluded at entry.) Most importantly, the finding of a non-causal inverse association does not exclude a modest positive causal association, which would be undetectable because of the small number of deaths.

    In epidemiological studies such as this the central issues are bias, confounding, and how well the factors that are measured, in this case biochemical markers, represent the factors of clinical interest, in this case dietary intake. In these respects a careful dietary assessment may be superior to analysing the proportion of fatty acids in patients' lipids (the authors' “objective” proxy measure of dietary intake). In a prospective study assessments of intake should be free from bias because the measurements are obtained before diagnosis, and a complete dietary assessment allows confounding dietary factors to be controlled for. Using levels of fatty acid in serum cholesterol esters as an indicator of individual intake is likely to have low validity because these levels are under strong homoeostatic control.11 As noted by Zureik et al, monounsaturated fatty acids assessed in this manner reflect diet even less well because they can be synthesised endogenously. Moreover, interpreting levels of fatty acid in serum fractions is complicated because the total adds to 100% and a low level can indicate low consumption of that fatty acid or high intake of other fatty acids. These analyses may be more informative for other fatty acids that are less closely regulated, but in general they are likely to be less informative than carefully collected dietary data.

    Although there is currently little evidence that high consumption of polyunsaturated fat increases the risk of cancer in humans, those responsible for protecting the public's health should not forget the issue. The most reliable information will come from large prospective studies, including the multinational European investigation into cancer and nutrition,12 based on direct dietary assessment as well as on biochemical indicators of dietary intake.


    View Abstract