by Bjelakovic et al, The Lancet 364
(Oktober 2004)
The aim of the meta-analysis by Bjelakovic and coworkers (1) was to assess whether antioxidant supplements reduce the incidence of gastrointestinal cancer and mortality. The meta-analysis included 14 randomized trials (n=170 525) with b-carotene, vitamin A, vitamin C, vitamin E, selenium, and other vitamins and minerals. The trials investigated primary as well as secondary prevention within different population. Types, doses, dose regimens, and duration of supplementation vary considerably between the trials.
Only in 3 studies healthy individuals were assessed, whereas in the remainder subjects with established high risk (i.e. long-term heavy smokers, asbestos workers, patients with oesophageal dysplasia, previous skin cancer, hepatitis B surface antigen carriers, relatives of patients with previous liver cancer) were included. In many of the trials gastrointestinal cancers were secondary endpoints only. Please note, antioxidants are thought to prevent from chronic diseases, thus being important at the onset of the disease (i.e. mutations, LDL oxidation) rather than treating patients or people at high risk. Antioxidants cannot be expected to reverse the negative effects created by smoking or poor dietary habits.
The authors of the meta-analysis summarized in their findings section that “neither the fixed-effect nor the random-effects meta-analyses showed significant effects of supplementation versus placebo on cancer incidences”. We would like to add that the reported relative risks of 0.96 and 0.90, respectively, at least indicate favorable trends for the antioxidant supplements. For a subgroup of ‘seven high quality trials’ the authors reported an increased mortality with a relative risk of 1.06. This effect was statistically significant based on the fixed effects-model but non-significant by the random-effects meta-analysis. The authors “resolved this difference by exclusion of the smallest trial”, achieving statistical significance for the random-effects model.
From a statistical point of view, this manipulation of excluding a high quality trial is totally inappropriate. Using a fixed-effects model to combine treatment estimates assumes that no heterogeneity exists between the study results. In the present case, the Cochrane test for heterogeneity resulted in a P value of 0.02, which is a clear indication for relying on the random effects model. Thus, to generate homogeneity by the removal of high quality studies from a meta-analysis is arbitrary and inadequate. A statistical multiple testing issue may have arisen from the emphasis of the authors on the subgroup of high quality studies. As the study protocol was not accessible, we could not verify if this subgroup had been predefined in the protocol. As we were interested in the overall relative risk across all 9 studies providing mortality data (Figure 4), we estimated the relative risk of antioxidant supplementation versus placebo: Relative risk = 1.035, 95%-confidence limits (0.963 to 1.1122) with a P value of 0.346 for the random-effects model, by far away from the usual limits of statistical significance. While performing this calculation we discovered a major typing error in Figure 4 on mortality data. The number of deaths in the placebo group of the Li et al study should read 167 instead of 67, but the authors in their risk calculations obviously used the correct number. The results of the meta-analysis are dominated (biased) by the CARET2 and ATBC3 studies in heavy smokers and asbestos workers using high doses of b-carotene and vitamin A. The authors themselves correctly stated “heterogeneity remained significant because of the strong effect of the Omen 1996 trial [CARET]”, a very peculiar population receiving a daily supplement of 25,000 IU vitamin A and 30 mg b-carotene. CARET and ATBC account for more than 28% of the population of the meta-analysis. The inclusion of other recent findings including the Supplémentation en Vitamines et Minéraux Antioxydants (SUVIMAX)4 trial would have lead to very different results. Therefore, it is correctly stated in the accompanying comment by Forman & Altman5 that Bjelakovic and colleagues1 analysis of mortality is “only to be considered exploratory” and does not offer convincing proof of hazard because it did not include other trials than with gastrointestinal cancer endpoints. Epidemiological and mechanistic studies point to the fact that antioxidants play a pivotal role in primary rather than secondary prevention. Two major studies are still ongoing which include antioxidant supplements: Physicians’ Health Study (PHS II)6 and Women’s Health Study (WHS)7. Both of these studies target healthy populations. The outcome of these studies will be important for any conclusion that may be drawn on efficacy and safety of antioxidant supplements.
Bjelakovic G, Nikolova D, Simonetti RG, Gluud C. Antioxidant supplements for prevention of gastrointestinal cancers: a systematic review and meta-analysis.
Lancet 2004; 364:1219-8.Omenn GS, Goodman GE, Thornquist MD, et al. Effects of a combination of beta carotene and vitamin A on lung cancer and cardiovascular disease.
N Engl J Med 1996; 334:1150-5.Alpha-Tocopherol, Beta-Carotene Cancer Prevention Study (ATBC). The effect of vitamin E and beta carotene on the incidence of lung cancer and other cancers in male smokers.
N Engl J Med 1994; 330:1029-35.SUVIMAX. istna.cnam.fr/sites/suvimax.
Forman D, Altman D. Vitamins to prevent cancer: supplementary problems.
Lancet 2004; 364:1193-4.Christen WG, Gaziano JM, Hennekens CH. Design of Physicians' Health Study II--a randomized trial of beta-carotene, vitamins E and C, and multivitamins, in prevention of cancer, cardiovascular disease, and eye disease, and review of results of completed trials.
Ann Epidemiol 2000; 10:125-34.Lee IM, Cook NR, Manson JE, et al. Beta-carotene supplementation and incidence of cancer and cardiovascular disease: the Women's Health Study.
J Natl Cancer Inst 1999; 91:2102-6.