| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Letters to the Editor |
Harvard School of Public Health, Nutrition Department, 677 Huntington Avenue, Boston, MA 02115
Dear Sir:
We agree with Truswell that a comprehensive literature search is central to a high-quality meta-analysis. However, we disagree with his statement that our meta-analysis was incomplete. We searched the MEDLINE (National Library of Medicine, Bethesda, MD) database (from January 1966 through June 1996) for literature with the following medical subject headings: "dietary fiber," "cereals (corn, oat, rye, millet, rice, and wheat)," "oat bran", "oatmeal," "gums," "guar," "pectin," "fruit," "vegetables," "psyllium (Plantago ovata husk and seeds)," "blood cholesterol," and "hyperlipidemia." We reviewed 162 clinical studies reporting the effects of soluble fiber on blood cholesterol, including 49 studies of oat products, 26 of psyllium, 20 of pectin, and 57 of guar (1). As stated in our paper, most of the studies excluded lacked either random allocation to low- and high-fiber diets or a randomized crossover between the 2 diet periods. For instance, Truswell cites 2 of his studies that each began with a 14-d control diet followed by a high-fiber diet of either pectin (2) or of oats (3) and that concluded with a 14-d control diet. We did not include either of these studies in the meta-analysis because the order of the control and fiber diets was not randomly assigned.
Truswell is concerned that our inclusion criteria may have been too restrictive. We assessed the influence of certain inclusion criteria on the results by relaxing some of the inclusion criteria in secondary analyses that were reported in our paper. For example, 3 trials [including 1 published by Judd and Truswell in 1982 (4)] that did not use a true low-fiber control but rather compared the effects of high doses with lower doses of the same fiber were included in a secondary analysis. Although the report of the cholesterol-lowering effect of pectin (15 g/d) by Keys et al (5) did not specifically state that the order of treatments was randomized, the investigators administered the pectin and control diets in reverse order to compensate for potential carryover effects. When we included this study in a secondary analysis, the mean total cholesterol–lowering effect of pectin was similar to that of other fibers (-0.04; 95% CI:-0.01, -0.09), making it appear even less likely that type of fiber could have accounted for a significant amount of heterogeneity among the different studies. However, it is possible that small differences in the response of cholesterol (-0.02 to -0.03 mmol/L per gram dietary fiber) to different fibers may not be detectable. In addition to the secondary analyses in the published report, we also performed analyses to include trials in which the background diet was not well characterized or controlled (6–13). The results from these analyses were not materially different from the primary analysis. The mechanism by which fiber lowers blood cholesterol remains unknown. We thank Truswell for his comments on the mechanism of action of viscous fibers.
REFERENCES
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
