JAMA in this week published three articles about dietary fat in the etiological development of breast cancer, colon rectal cancer, and cardiovascular disease (CVD). Unexpectedly, all reported negative findings.

The Women’s Health Initiative (WHI) study is a randomized clinical trial in which about 49,000 healthy women with average age 62 were randomized into the intervention group or the comparison group, and were followed up on average 8 years. Women in the intervention group were instructed to adhere on a low fat diet (total fat accounted for 20% of total energy), while women in the comparison group could eat their usual diet (fat accounted for 30-35% of total energy).

Despite its large sample size and long follow up period, the WHI study found no protective effect of dietary fat on the incidence of invasive breast cancer (somewhat weak effect), colon rectal cancer, and heart disease and stroke. Careful analysis suggested that there was also no significant difference in serum low density lipoprotein, blood pressure, and other intermediate factors between the two groups.

What a bummer! The findings from this study are surely a blowout for the long held belief that reducing dietary fat can lower the risk of cancer and in particular, the CVD.

But is that so?

By reading the papers more carefully, I generally agree with the accompanied commentaries that this study is not sufficient to refute the hypothesis that dietary fat is protective against cancer and CVD.

First, this study is not the state of art in the dietary fat research. Recent knowledge suggests that not only the total fat but also the components of fat (such as saturated fat and trans fat) are important. The WHI study failed to address this issue.

Second, the effect of total fat on the etiology of cancer and CVD may be small, and is subjected to the influence of many other factors. In the WHI study, there is no control of other lifestyle factors such as weight maintenance, exercise, and total energy control. But admittedly, it is infeasible to conduct such a multiple factor study.

In fact, in the Multiple Risk Factor Intervention Trial (MRFIT) of 1970s, there was no significant protective effect of lifestyle change on CVD until after a much longer extended follow up period. The lessons from the MRFIT apply to the WHI study too. There was a comparable lifestyle change in both groups, for example, dietary fat also reduced in the comparison group. This kind of contamination will reduce the intervention difference between the two groups, thus reducing the statistical power.

Third, lifestyle changes including dietary fat change are very difficult to adhere. Although the WHI study showed that the adherence to the intervention was not bad, many women didn’t stick to the strict low fat diet. This attrition due to noncompliance may reduce the power too.

There are many other post hoc interpretations to these negative findings which I will not describe here. But the main message is clear: in this largest ever randomized clinical trial of dietary factors, there is no positive finding.

Observational studies have been reporting positive findings for years, and several meta-analyses also yielded promising results. So what on earth could these negative findings happen in the randomized clinical trials?

Now it’s time to rethink about the limit of clinical trials. In these trials, patients are self selected to participate the studies. They tend to be different from the general population. In addition, because of logistic reasons, the interventions employed in the trials are never comprehensive, and most other factors are left to the randomization to balance the differences, which again are never perfect. However, these differences may have significant effect on the intervention (called interaction). In a long follow up, contamination and attrition may negate the purposed effect. In this sense, the power of randomized trials is not much bigger than usual observational studies.

On the other hand, natural experiment such as comparisons between different existing populations and between different time periods may produce insightful findings. For example, the story of lipid and saturated fat is first discovered by ecological studies across countries, and later confirmed by laboratories and clinical practice. The long term trend of population lipid level and fat intake also demonstrates the effect of saturated fat on lipid. A carefully selected natural experiment is not worse if not better than a randomized clinical trial.

One side note, diet in Eskimos and Inuits contains a large amount of animal fat, but their CVD mortality is not significantly higher than that of the general population. This suggests that other lifestyle factors are equally important to influence the risk of CVD.

The controversy surrounding the dietary fat will go on, and as most papers like to conclude: more research is needed.

PS:

One comment in xys states that since results are based on women with age 50 and older, they may not be generalized to other populations.

My response is as follows:

It is true that study conducted on old women may not be generalized to young women and all men, but this study design is valid and so the results for the etiological effect of dietary fat on cancer and CVD is valid. That is, if the biological mechanism of dietary fat on cancer and CVD is independent of age and gender (which may not be totally true), results in this study make sense.

It is also true that the dietary study in the WHI study is an add on study. The WHI study was initially for hormone therapy. However, the dietary intervention was carefully randomized and monitored. It is a standard randomized trial. It is actually a good idea to test different hypotheses on a single cohort, given that the logistics of initiating and recruiting people is tremendous.

On the other hand, if the dietary fat does have effects on cancer and CVD (as the alternative hypothesis states), it should show its effect among old women too. This is a standard refutism framework of Karl Popper. One negative result can refute the hypothesis while tens of positive results cannot prove the hypothesis. Sure this is narrowly defined refutism.

Again, you can say maybe the dietary fat has very small or no effect among old women, given they are quite different from the rest of us. Maybe the dietary fat still has some effects among men or young women. So we are back to the original: more research is needed.