点击显示 收起
1 Hormonal and Reproductive Epidemiology Branchm, Division of Cancer Epidemiology and Genetics, National Cancer Institute, Rockville, MD 20852 E-mail: althuism{at}mail.nih.gov
2 Statistics Collaborative, Inc, Washington, DC
Dear Sir:
We thank the authors of the letters for their comments on our article. Our review and meta-analysis summarized randomized clinical trials (RCTs) designed to assess glucose and insulin responses to dietary chromium supplements (1). We limited our review to RCTs to avoid the potential for bias inherent in nonrandomized studies. We attempted to include every RCT in the literature.
The letters by Kalman and by Juturu and Komorowski cite several studies they say our review omitted (2–11). Our review in fact included 4 of these studies (2–5); the other 6 studies they mentioned (6–11) were not RCTs and therefore were not eligible for inclusion. Specifically, 1 of the 4 RCTs cited as being omitted was both discussed in the review and combined analytically in the meta-analysis (3). In addition, we discussed in detail the findings from the other 3 RCTs although they were not included in the meta-analysis (2, 4, 5). One of these RCTs (4) was excluded from the meta-analysis because the study population—women with gestational diabetes—was not a focus of our review; one of the others was excluded simply because data presented in the original report were insufficient for abstraction, and updated data were not available from the investigators (5).
The authors of all 3 letters express concern that we did not analytically combine the study by Anderson et al (2) in the meta-analysis. First, the Chinese population described by Anderson et al was very different from the populations of the other trials, such that its inclusion would lead to violation of the statistical assumption of heterogeneity in models pooling all 4 studies. Second, because the odds ratio estimated by meta-analytic techniques is weighted more heavily for large studies, pooling the data from Anderson et al (n = 155) with the data for the 38 subjects from the other 3 studies would overwhelm the results, making the effects of the smaller studies—ie, studies from populations more similar to that of the United States—difficult if not impossible to assess. Thus, we believe that separating the presentation of the results of the Western studies from that of the results from the one non-Western study better facilitates critical review.
The remaining 6 studies cited as being omitted were not RCTs, but rather uncontrolled investigations (6–11). In addition to an uncontrolled study, one report described a small controlled clinical trial that assessed 10 subjects who were not randomly assigned to receive treatment or placebo (11). Although the data were not presented in the report, those authors reported no difference between the placebo and chromium groups (11).
McCarty is correct that the studies we reviewed did not address the use of high doses of chromium. He points out that data from uncontrolled and animal studies suggest that chromium may be valuable as a dietary supplement or in pharmaceutical doses. Nonetheless, before making recommendations for use by the general public, we urge that investigators test dietary chromium supplements, particularly those with high doses, in a well-designed RCT. The limited data from RCTs on dietary chromium supplementation have yet to prove that it is either efficacious or safe for healthy persons or for those with type 2 diabetes.
REFERENCES