One of the four major medical journals, JAMA, recently released a study which concluded that antioxidants increase mortality rate. It made big headlines, but unfortunately got blown out of proportion. There were many problems with this study which are described below from an article I copied. Another problem that wasn’t mentioned is the quality/ratio/quantity of the supplements being used. It is assumed that Vitamin E hold a lot of promise, but it is needed in a blanced ratio. Vitamin E has many different forms and taking just one form of the vitamin can throw off the internal milieu balance and uptake of other forms. If you look at your vitamin bottle you’ll probably see Vitamin E in the form of ‘alpha-tocopherol’. If you see a beta, gamma or delta tocopherol then congratulations you may (I say only may) have a better supplement than most. (Many companies will just add a little bit of the other tocopherol forms just to look good, but it really isn’t in a balanced ratio). In nature Vitamin E occurs in these forms as well as tocotrienol forms. There are a plethora of epidemiological studies that show an inverse relationship between Vitmain E intake from foods and disease. I do believe we can find benefit from a pill as it has been proven. However the form must be balanced and not just alpha-tocopheraol (the most predominate and least expensive form). Also many of the studies used in the meta analysis only used one antioxidant and it is well known that antioxidants work synergistically ‘recharging’ each other. (Vit E for Vit C for example).
Any antioxidant also means it is a prooxidant. It can not only decrease but increase free radical damage. So if one give Beta Carotene to smokers without any Vitamin E increase it can cause more damage to lung tissue. It is probably much healthier to get these substances from food as one not only recieves anitoxidant vitamins, but phytochemicals as well.
Dr Steve Austin ND gives a wonderful commentary on the study. (BC is beta-carotene):
At first glance, this analysis appears to prove that antioxidants increase mortality. These findings might be construed to show that supplementing with antioxidants is dangerous and they should arguably not be consumed. On closer analysis, however, this document appears to this writer to have more political than scientific overtones.
The primary effect of the abstract has been to make negative national news. And what is to be said regarding the quality of the science? The inclusion of BC virtually assured an increased mortality in the overall data. It has previously been proven that synthetic BC supplementation increases mortality in smokers.
The authors admit that no treatment duration effect appeared in their analysis. Is it likely that people taking a vitamin that can kill them have the same increased risk when the intervention lasts six months versus ten years? The authors do not address the implausibility of such findings.
In the same way that including BC trials makes the data appear more negative than they otherwise would, the authors purposefully exclude the positive Se trials in parts of their analysis. They claim further to have focused on “low-bias” trials because those trials produce more reliable data. Normally, such is the case, but can trials that randomize with something other than a computer or random-number generator actually run the risk of not being able to determine who is alive or dead? Could such methodological imperfections result in someone’s death? Excluding the so-called “high risk of bias” trials allowed the authors to paint a very different picture than emerges when the other trials are included. Indeed, the fine print shows that the excluded trials, even including those using BC supplementation, produced a statistically significant reduction in mortality compared with placebo (RR 0.91, 95% CI 0.83-1.00, P=0.04 for overall effect). These data were excluded from the abstract. JAMA editors know most of the media are unlikely to read the fine print.
The dosages used in some trials were trivial. For example, as little as 2000 IU of BC, 1,333 IU vitamin A, 60 mg of vitamin C, and 10 IU of vitamin E were used. While there is no plausible way to explain how such trivial amounts could have killed anyone, the mortality data from these trials were nonetheless included in the analysis.
And what of the so-called “low risk of bias” (read higher quality) trials? Some appear irrelevant. As an example, one trial (Age Aging 1992;21:435-9) literally gave subjects only one dose of vitamin A and then followed the subjects for three months. Can such a trial possibly provide meaningful mortality data? The answer appears to be “yes” according to these authors, at least as long as the findings lean toward increased mortality.
That trial was considered to have a “low risk of bias.” Yet in contrast, the respected GISSI trial (Lancet 1999;354:447-455) was classified as low-quality. It is the largest of the trials excluded from the “low-bias” data, and its findings show an overall 8% reduction in mortality with antioxidants compared with placebo, though this difference did not reach statistical significance.
Despite its shortcomings, there is something to glean from the new analysis. First, the data support previous findings that BC supplementation increases mortality. Of four BC trials for which relative risks are individually provided, all at least trend toward increased mortality. In the analysis, smokers are not distinguished from nonsmokers, allowing BC to be painted with a black brush regardless of smoking status.
The overall findings also suggest that the idea perhaps most famously espoused by Linus Pauling, that antioxidants might improve health (and therefore, potentially lower mortality rates), is not supported; when all data were combined (both lower and higher-quality trials) essentially no effect on mortality emerged.
Of four vitamin A trials for which relative risks are individually provided, all at least trend toward increased mortality. Yet, overall, vitamin A trials showed evidence of increased mortality only when a variety of studies were removed from the analysis (those concomitantly using Se and those with a “high risk of bias”). Whether such analysis was made post hoc is not clearly stated.
As with vitamin A, of the four vitamin E trials for which relative risks are individually provided, all at least trend toward increased mortality. These associations were not statistically significant until a number of trials were removed from the analysis (again, those concomitantly using Se or of “high risk of bias”). This finding must be tempered with the recent vitamin E/mortality meta-analysis (previously reviewed in this column) in which the association occurred only at very high dosages while several trials using low doses reflected a reduction in mortality.
Three of four of the vitamin C trials for which relative risks are individually provided showed nonsignificant reductions in risk. All four Se trials for which relative risks are individually provided show at least a trend toward reduced mortality (9-15%). Se is the one antioxidant for which we might have expected a reduction in mortality (particularly in regard to cancer risks in men). These preliminary findings are encouraging.
The authors conclude, “. . . our estimates of increased mortality of about 5% is likely to be conservative.” They suggest this is so because of publication bias, the proven concept that negative or neutral trials are less likely to get published than are trials showing positive findings. Ironically, this line of argument begs the question, can anyone who has been reading JAMA or the New England Journal over the last few years believe that this metaanalysis could have been published in either journal if the overall data had favored the use of antioxidants in lowering mortality?
George Mandler
Licensed Acupuncturist & Nutritionist Dietitian
Acupuncture, Chinese Medicine, & Nutrition
Boston, Jamaica Plain, Brookline, Massachusetts