Flawed Meta-analysis Claims Vitamin E above 150 IU/day Increases Mortality
By Author: Steve Austin, ND
Reference: Miller ER 3rd, Pastor-Barriuso R, Dalal D, et al. Meta-Analysis: High-Dosage Vitamin E Supplementation May Increase All-Cause Mortality. Ann Intern Med 2004; 142:Nov 10 [publication ahead of print, thus no page numbers were available at the time of this writing]
Design: Meta-analysis of 19 randomized clinical trials (RCTs) lasting more than one year and including mortality data showing at least 10 deaths
Participants: A total of 135,967 participants from the 19 RCTs
Study Medication and Dosage: Vitamin E was administered in a variety of forms. Details were not provided in the meta-analysis regarding the form of vitamin E used in each trial. Doses ranged from a low of 16.5 IU/day in one trial to a high of 2,000 IU/day in two trials.
Outcome Measures: All-cause mortality
Key Findings: In trials using doses equal to or less than 330 IU/day, pooled data showed all-cause morality was reduced by a statistically non-significant 2%.
In the three trials using the dose of 400 IU/day exclusively, the largest showed no effect on mortality. One of these trials showed a non-significant increase in mortality. Another showed a statistically significant reduction in mortality. The data from these three trials were not pooled to the exclusion of data from trials using other doses. If they had been, however, the information provided suggests that the 400 IU dosages would have shown essentially no effect on mortality.
Regarding trials using more than 400 IU/day, all used a minimum of 500 IU/day except for one trial in which data for participants given 400 or 800 IU/day could not be separated and were therefore considered to be "600 IU/day" by the meta-analysis authors. Seven of the eight trials using doses of or greater than 500 IU/day showed an increase in mortality. Except for one of these seven, the increased risk became statistically significant only when the data were pooled.
Practice Implications: Many years ago, Mark Twain said there were "lies, damned lies, and statistics." The new meta-analysis appears to use statistics to frighten doctors away from doses of vitamin E that are clearly safe, even according to the data analyzed within the meta-analysis. Rather than stating that there was no evidence suggesting that doses of 400 IU/day increase mortality, the authors attempt to obfuscate this fact by pooling these data with those from trials using higher doses, in which a real problem did appear to emerge. I say "appear" because of the two trials using the highest dose, 2000 IU/day, one showed a non-significant reduction in mortality risk.
Regarding the effects of lower-dose vitamin E on mortality, the authors admit a statistically significant reduction in risk. " . . . [B]ased on four-way data, the pooled risk difference for low-dosage vitamin E [at or below 330 IU/day] trials was -33 per 10 000 persons (p = 0.021)."
The authors do not acknowledge -- though the data demonstrate -- that, as stated above, the three 400 IU/day trials suggest no effect on mortality whatever. In the one trial in which 400 IU/day led to a (statistically non-significant) increased risk, subjects were not given vitamin E alone, so the effect observed might have been due to something other than vitamin E. It is notable that the one 400 IU/day trial reporting no effect on mortality is the only one to use natural vitamin E.
At or above 500 IU/day the data do suggest an increase in mortality -- there was a statistically significant increase in risk for all trials using more than 330 IU/day. Given that the 400 IU/day trials would appear to push the P value for mortality risk up among these "high-dose" trials, it should follow that removing the data from the 400 IU/day trials would probably have improved the statistical significance of the rise in mortality for the higher-dose trials.
By far the largest trial in which a dose above 400 IU/day was associated with a (statistically non-significant) increase in mortality also supplied subjects with synthetic beta-carotene, which may have skewed the data in favor of higher mortality.
How do the authors jump from these findings to a conclusion that any dose above "150 IU/day" is increasing mortality? By using statistics, if not "damned lies." They create a hypothetical calculated dose-dependency curve in which all-cause mortality appears to cross unity at 150 IU/day. However, these artificial numbers do not reflect the findings of individual trials, if we keep in mind that the data from the 400 IU/day trials do not suggest any increase in risk at that dose.
The bias of the authors is demonstrated in other ways, too. They mention that the antioxidant/gastrointestinal cancer meta-analysis discussed in this column last month (Lancet 2004;364:1219-28) found an increase in mortality associated with antioxidant use -- the implication being that vitamin E might have caused the problem. Yet, they avoid mentioning that that increase was essentially due to one trial in which vitamin E-supplementing subjects were also smokers taking synthetic beta-carotene, a combination known to increase mortality independent of vitamin E supplementation.
To further support the idea that vitamin E should be considered toxic, they mention that H Roberts has previously reported "many adverse effects" from taking vitamin E (JAMA 1981;246:129-31) but they neglect to mention that in the 23 years since Dr. Roberts reported these anecdotes, no one has corroborated any of these effects despite the fact that tens of millions of Americans having been taking large doses of vitamin E. They also do not mention that Dr. Roberts has voiced similar concerns about other substances that subsequently have not been found to be toxic by researchers or other clinicians.
If vitamin E is truly killing people -- and the data reported from high-dose trials suggest the possibility of a 3- 7% graduated increase as doses advance from 500 IU up to 2000 IU/day -- then clearly, the cause of death would be a major clue from which to proceed. Let us say, for example, that a significant increase in mortality in those given vitamin E was dependent upon statistically nonsignificant increases in cardiovascular deaths and deaths from homicide. In such a case, we would have to conclude that chance is likely to be the cause of the problem. In contrast, if the increase in mortality had all come from the same cause, and that pattern occurred in multiple trials, we would have to conclude that chance was unlikely to be the culprit. Yet, these authors do not consider the cause of death.
Several reviews of this meta-analysis have focused on the fact that all subjects in most of the trials were chronically ill (for example, with heart disease). This criticism seems unconvincing to me, given that all trials were randomized. Why should people who have a chronic illness be more likely to die from taking vitamin E supplements than from taking placebo? Projecting the findings of this meta-analysis onto healthy populations does require an extrapolation, but I see little reason to assume this to be an unfair extrapolation.
These details aside, the bottom line appears fairly clear: despite claims to the contrary by the authors, the data in this meta-analysis show that vitamin E at or below doses of 400 IU/day do not appear to increase mortality. At and above 500 IU/day, the data suggest that an increase in mortality is possible and should be further explored, though effects on mortality of natural vitamin E (d-alpha) versus synthetic vitamin E (dl-alpha) should be kept separate. Until this issue is better resolved, doctors should feel relatively comfortable in using doses of 400 IU/day. The long-term use of higher doses should be undertaken only with caution and some circumspection.
|
Loading...