From the Director
Balz Frei, Ph.D.
You may have heard about the recent study examining the effect of antioxidant supplements on death rates published in the February 28, 2007, issue of the Journal of the American Medical Association. The study's authors, who are from Denmark, Serbia, and Italy, concluded that antioxidant supplements, in particular beta-carotene, vitamin A, and vitamin E, "may increase all-cause mortality," while "vitamin C and selenium had no significant effect" and their "potential roles on mortality need further study." The study was a meta-analysis, which is a statistical analysis of previously published data pooled together into a single study.
In this study, the authors initially looked at 815 human clinical trials of antioxidant supplements, but only 68 were included in the meta-analysis. A typical clinical trial is a study in which a large group of healthy subjects or patients with a specific disease is randomly divided into two groups that are given either placebo or active treatment (in this case antioxidants) for a long period of time, with neither the participants nor the doctors knowing who is getting which treatment (i.e., the study is double-blinded). The 68 trials deemed acceptable for inclusion in the meta-analysis were divided into either "low bias" or "high bias" trials. While the 47 "low bias" trials showed a 5% increased mortality with antioxidant supplements, the 21 "high bias" trials actually showed a 9% decreased mortality with antioxidant supplements. Furthermore, the increased mortality for beta-carotene, vitamin A, and vitamin E supplements was only seen in "low bias" trials, but not in all 68 trials combined (except for beta-carotene when given singly). In other words, vitamin E supplementation, for example, was only associated with increased mortality after exclusion of data from "high bias" trials.
"High bias" was defined as "one or more unclear or inadequate quality components" of the trial, e.g., "allocation sequence, concealment, blinding, placebo, and follow-up." This is a subjective cut made by two authors of the paper, with "discussion or arbitration by a third author," as needed. Because of this rather subjective cut, two large trials were excluded from overall consideration that found substantial benefit of antioxidant supplements: a trial in Linxian, China, which observed 6% lower mortality associated with supplemental intake of beta-carotene, selenium, and vitamins A, C, and E and was published in the Journal of the National Cancer Institute, and a large trial in Italy called GISSI Prevenzione trial, which found 8% lower mortality with vitamin E supplements and was published in Lancet. The authors of the meta-analysis did not explain why these two large, well-designed trials were excluded. They were good enough to be published in the official journal of the National Cancer Institute and the premier British medical journal.
Hence, the arbitrary separation into "high bias" and "low bias" studies and the exclusion of the "high bias" studies from the overall evaluation seem to reveal a bias by the authors in the analysis and interpretation of the data towards showing harm and confirming their own, previously published data. In other words, we have to trust the judgment of two or three authors on whether these two large studies (Linxian and GISSI Prevenzione) are worthy of inclusion in the overall meta-analysis. If these studies were included, none of the effects on increased mortality reported would be significant, except for beta-carotene.
The data on beta-carotene, however, are driven by two large clinical studies, the so-called Alpha-Tocopherol Beta-Carotene (ATBC) Cancer Prevention Study and the Beta-Carotene And Retinol Efficacy Trial (CARET), which showed increased lung cancer incidence in beta-carotene or vitamin A-supplemented heavy smokers, ex-smokers, and workers occupationally exposed to asbestos. This is nothing new, and these two trials have been extensively reviewed, re-reviewed, and covered by the media. Yes, if you smoke, don't take beta-carotene supplements, but much better, quit smoking!
Also, even among the "low bias" trials, many showed decreased mortality with antioxidant supplements. For example, the Supplementation en Vitamines et en Mineraux Antioxydants (SUVIMAX) Vascular Study in France found that supplementation of over 13,000 middle-aged patients with vitamins C and E and selenium resulted in a 22 percent reduced mortality after seven years.
Furthermore, clinical trials of dietary antioxidants have serious limitations. They may be the "gold standard" for testing safety and efficacy of pharmaceutical drugs, but are hardly suitable to test dietary compounds, which are metabolized in completely different ways from drugs. In addition, dietary compounds are present in the human body before supplementation occurs, which significantly diminishes the power of the trial to detect a statistically significant effect of the supplement.
Here are some additional limitations and problems with most of the 68 trials reviewed in this meta-analysis:
Additionally, mortality data from one 2001 study were misrepresented. The authors of the meta-analysis claimed that there were 13 deaths in the control groups and 17 deaths in the antioxidant groups in that study. Actually, there was only one death in each group.
As Bernadine Healy, former director of the National Institutes of Health and president and CEO of the American Red Cross and current member of the President's Council of Advisors on Science and Technology, put it: "Blenderizing these diverse trials into one giant 232,606-patient-strong study to come up with a seductively simple proclamation is just silly ... Sure, statistics can prove anything. But this study violates a cardinal rule of meta-analysis. Pooled studies must be compatible. That means combining apples and applesor at the least, similar patients and comparable doses and duration of treatments. The first question to ask in evaluating any such study is whether the combination makes sense, both common sense and medical sense, in the first place. On both, the study flunks."
Overall, this is a flawed meta-analysis of flawed data that does not reveal the true health effects of antioxidants, whether beneficial or otherwise, but instead reveals the authors' bias towards showing harm. The analysis is limited to clinical trials of antioxidants, and these trials are well known to have serious limitations. Instead, the totality of evidence from laboratory, animal, and human studies to date shows that antioxidants provide many health benefits, including reduced risk for heart disease and stroke, some cancers, eye diseases, and neurodegenerative diseases, as well as enhanced immunity and resistance to infection. Keep taking your antioxidant supplements, in addition to getting many other micronutrients and phytochemicals from a plentiful intake of fruits and vegetables!
Last updated May 2007