|
Correction
to
Hu et al.,
|
| |||||||||||||||||||||||||||||||||
The authors also compare the effects of the replacement of 5 percent of calories from saturated fat with the replacement of 2 percent of calories from trans unsaturated fat, concluding that the former would lead to a 42 percent reduction in the risk of coronary heart disease and the latter a 53 percent reduction. It may be that calorie for calorie trans unsaturated fat exerts a greater effect on risk than saturated fat. However, whereas replacement of 5 percent of calories from saturated fat represents a 32 percent reduction in saturated fat intake (the mean being 15.6 percent in their population) and is feasible, replacement of 2 percent of calories from trans unsaturated fat implies a near-total (91 percent) removal of trans unsaturated fats from the diet (mean intake, 2.2 percent), which is not practically achievable. It would have been of interest to compare one-third reductions of both; in such a comparison, the effect of a feasible reduction in saturated fat intake would presumably outweigh the effect of a feasible reduction in trans unsaturated fat intake.
Ira S. Ockene, M.D.
University of Massachusetts Medical Center
Worcester, MA 01655
Robert Nicolosi, Ph.D.
University of Massachusetts at Lowell
Lowell, MA 01854
References
Studies with double-labeled water have demonstrated that estimates of energy consumption, of which fat is a major contributor, are rather inaccurate and biased. In the validation studies,1 correlations for various fats ranged from 0.46 to 0.68 — that is, only 22 to 46 percent of the variance is explained. The inability to detect a clear effect of saturated fat and dietary cholesterol suggests the limited reliability of the data on these components, since metabolic studies have demonstrated that they have a major role in modifying serum lipid levels.2
It is well established that populations consuming low-fat diets have low low-density lipoprotein and high-density lipoprotein (HDL) levels and that mortality from coronary heart disease is also low. Triglyceride levels are not abnormally high. The elevation in triglycerides observed in experimental studies has been shown to be temporary.3 Also, the clinical significance of the lower HDL levels observed with low-fat diets has been questioned.4
The range of total fat intake recorded in these studies is limited and has little relevance to the protective effect of really low fat diets. It will probably be agreed, however, that within these limits, the composition of dietary fat, rather than the level, is of primary importance.
Although high levels of trans fats are clearly undesirable, we should not throw out the baby with the bath water by disparaging all margarines. Tub margarines offer important sources of polyunsaturated fatty acids with minimal amounts of trans acids.
D.M. Hegsted
Harvard Medical School
Southboro, MA 01772
References
Hu et al. cite a book as providing evidence of the validity of the food-frequency questionnaire. The book refers to a 1985 study by Willett and colleagues2 that validated the food-frequency questionnaire by comparing it with four one-week diet records collected over a period of one year in 173 subjects. Thus, validation depended on the assumption that because the two methods agree, each gives valid results. However, diet records have since been shown to underestimate energy intake as compared with measured energy expenditure.3 Furthermore, the reported mean energy intakes in 1985 in repeated food-frequency questionnaires (1418 and 1371 kcal) and diet records (1620 kcal)2 were substantially less than the measured energy expenditure in 123 women of similar ages (2366 kcal).1
The adjustment for total energy intake in the multivariate analysis would only compensate for underreporting if all macronutrients were equally underreported. However, evidence is accumulating of the differential reporting of foods and macronutrients.4 The situation is exacerbated because the probability of underestimation is greater in women than men,5 increases as body-mass index increases,4 and tends to be subject-specific and not eliminated by repeated measures.4 Since obesity is a known risk factor for coronary heart disease, women at higher risk for coronary heart disease (those with a high body-mass index) are more likely to have differentially reported foods and macronutrients. Although as Hu et al. state, "errors in measuring dietary intake could have accounted for a lack of association, but not the presence of an association," bias can both remove and create associations.
Have the authors considered repeating their analysis after excluding women who reported physiologically improbable food intakes? This can be done using cutoff limits for energy intake below which a woman of a given age and body weight could not live a normal lifestyle.1 If the conclusions remain unchanged, they will rest on a sounder foundation.
Rachel K. Johnson, Ph.D., M.P.H., R.D.
University of Vermont
Burlington, VT 05405
Alison E. Black, B.Sc., S.R.D.
Tim J. Cole, Ph.D.
Dunn Nutrition Centre
Cambridge CB2 2DH, United Kingdom
References
To the Editor: Ockene and Nicolosi's inference about the three-dimensional relation among fat intake, physical activity, and body-mass index from the two-dimensional data in Table 2 is inappropriate; the more active women were actually leaner. Ockene and Nicolosi incorrectly characterize our prospective study as "cross-sectional." They also question the generalizability of the findings, but there is no reason to believe that the biologic effect of dietary fat on coronary risk differs between nurses and other women. Finally, a large reduction in trans fat is not as difficult as Ockene and Nicolosi think. Trans fat is the result of food-processing methods that can be changed. Today, most European margarines are free of trans fat, and the consumption of these products can, as Hegsted notes, have important health benefits.
Hegsted repeats the common misconception that the triglyceride-raising and HDL-depressing effects of a low-fat, high-carbohydrate diet are transient. This notion is not supported by a meta-analysis of 27 trials, lasting from 14 to 91 days,1 or a recent large 1-year study.2 Also, there is no evidence that the HDL-lowering effect of a low-fat diet is benign; the population comparisons cited by Hegsted are intractably confounded by major differences in physical activity and body fat.
Hegsted and Johnson et al. were concerned about potential bias in our measurement of diet, including the possibility of an underestimation of total energy intake, citing the validation study of our first questionnaire. The mean energy intake assessed by the revised questionnaire used since 1984 was 1844 kcal,3 a reasonable estimate for women of this age. Even if energy is underestimated, this would not affect our results, because we adjusted for both total energy intake and body-mass index. Also, in several analyses, underreporting of the percentage of calories from fat was not related to obesity.4,5 Furthermore, the relation of total fat intake to the risk of coronary disease did not differ between obese and nonobese women. Our analyses already excluded women with energy intakes outside the range of 500 to 3500 kcal per day. Following the suggestion of Johnson et al., we calculated the ratio of reported caloric intake to predicted caloric intake for each participant using their age and weight. Excluding women with the greatest likelihood for underreporting (the lowest quintile of the ratio) did not change the result (relative risk for total fat [5 percent of energy], 1.02; 95 percent confidence interval, 0.96 to 1.09). The use of four dietary assessments over a period of 14 years was a unique strength of this study, which substantially dampens the likelihood of measurement error and variation due to true changes over time.
In Table 3 of our paper, the unit for cholesterol intake was incorrect. The correct unit is milligrams per 1000 kcal per day.
Frank B. Hu, M.D., Ph.D.
Meir J. Stampfer, M.D., Dr.P.H.
Walter C. Willett, M.D., Dr.P.H.
Harvard School of Public Health
Boston, MA 02115
References
| |||||||||||||||||||||||||||||||||
HOME | SUBSCRIBE | SEARCH | CURRENT ISSUE | PAST ISSUES | COLLECTIONS | PRIVACY | HELP | beta.nejm.org Comments and questions? Please contact us. The New England Journal of Medicine is owned, published, and copyrighted © 2008 Massachusetts Medical Society. All rights reserved. |
Previous
