Dietary fat affects obesity rate.

Recently, we published a paper in the Journal titled “Dietary Fat Intake Does Affect Obesity!” (1). This was in response to an earlier paper published by Willett titled “Is Dietary Fat a Major Determinant of Body Fat?” (2). The editorial written about our paper was also prepared by Willett (3) and deserves rebuttal because it was essentially a critique of parts of our article. Our paper attempted to use epidemiologic data to counter similar data that Willett used in his argument published earlier in the Journal (2). We stated, “Ecologic studies...[are] useful for raising hypotheses” but they have far too many weaknesses to be used for more than that. We went on to refute each of Willett’s examples with examples that we felt were more appropriate. Our rebuttal of Willett’s use of median body mass index and weight and his selection of correlational analyses from China are the 2 areas that he focused on in his editorial. Our analysis presented regression results relating overweight prevalence data from a wide range of countries to the proportion of energy from fat to show that there appears to be a progression from lower to higher fat intake that is correlated with the percentage of individuals who are overweight. We never stated that this showed causation; rather, as the above quote and others would show, we were cautious in using such data. In the editorial by Willett (3), he presented correlation coefficients between dietary fat and obesity from 65 rural Chinese counties, a study that related household dietary intake data with adult body mass indexes. We rebutted this by presenting data from a nationwide longitudinal study of the effects of diet and activity on Chinese adults. Willett claimed that we did not control for a range of confounders and he noted quite correctly that “in China...dietary fat has increased concurrently with increases in wealth and food availability, reductions in infectious disease, and declines in physical activity... it is implausible that many Chinese would voluntarily resume the diets they consumed during times of poverty.” We agree. In fact, in many papers we have modeled the structure of these relations from income changes to diet and activity to changes in prevalence of overweight. We were only contesting the point in his piece that stated, “in China no correlation was found between dietary fat intakes...and body weight” (2). Note that we examined the effect of changes in diet while controlling for physical activity, smoking, sex, age, and residence. This powerful fixed-effects approach addresses his concerns. Elsewhere, we addressed his concerns in more detail (4, 5). He noted that the size of the coefficient of energy from fat (using adjustment and partitioning methods) was small. He is correct. We would expect this because we were looking at the effect of fat while controlling for energy intake, as we showed in the paper. One of the more important contributions to our conclusion was the meta-analysis of the 28 clinical trials. Willett (2) selected 6 of these studies to make his point—we felt this was an inappropriate way to analyze the data. We took all extant published work, omitted 1 trial that was clearly an outlier (although it benefited our argument), and did a weighted regression of the others to show a small effect of the reduction in the proportion of energy from fat on weight loss. Willett pointed out that we overlooked the study by Knopp et al (6) in which low-fat diets (22–27% of energy) were given to hyperlipidemic men. A weight loss of 6 kg was seen in the heavier group of men with combined hyperlipidemia eating 25% fat compared with a loss of 2–3 kg in the other groups. We did not find this study when we prepared the database for our work. Since that time, a 6-mo multicenter trial in 400 overweight men and women reported a significantly greater weight loss of 0.94–1.81 kg associated with a decrease in fat intake of 7.9–10% of energy. The control groups in this study gained 0.82–0.18 kg during the same interval. Whether the low-fat diet had complex or simple carbohydrates made no significant difference. These data were presented at the Eighth International Congress of Obesity (WHM Saris, A Astrup, AM Prentice, FJ Zunft, X Formiguera, unpublished observations, 1998). Willett did not address the range of experimental animal and human studies we reviewed, in particular, he did not comment on those that showed the ways that reducing fat with fat modifiers or other means does not lead to full energy compensation. This led to our major point. We did not expect the thermic effect of fat reduction to be important. Rather, we felt that changing the fat content of food had a major effect on energy density, and in turn, this significantly affected total energy intake. In other words, our main argument is that the effect of energy density on food intake affects total energy intake. We used a set of animal and human studies to address this topic. Willett’s editorial did not address this half of our paper. In summary, we never expected a strong effect of reduction in fat intake while keeping energy constant and we reviewed a range of experimental data to back up this point. Thus, we would agree with Willett on that point. However, as we noted, controlling fat intake is an important element of our effort to reduce total energy intake. To prevent the increase in the prevalence of obesity, controlling the fat in the diet is an important component. Most importantly for those in the United States and other highincome countries, we concluded that “to reduce the prevalence of obesity, there must be an increase in energy expenditure, a reduction in total energy intake, or both. This goal can be facilitated by reducing the amount of fat in the diet.” Letters to the Editor