Main Menu     Health Library     Clinical and Observation Studies                                                                                                                                            

ANTIOXIDANTS and DEATH from ALL CAUSES -

SUPPLEMENTARY ANALYSIS of FLAWED STUDIES

A meta-analysis of 62 studies of antioxidants published in 2007 claimed little or no benefit for antioxidants in general, and claimed that Vitamin A, Beta Carotene and Vitamin E significantly increased mortality. ((Bjelakovic, G, JAMA 2007, 267:842)  The results of this study are inconsistent with the broader studies of antioxidant benefits developed in the Life Ahead analysis that I reference following..   A review of this study confirms the conclusions that use of very high Vitamin A and Beta Carotene amounts are not beneficial.  But this appears to be true only for the very high amounts of A and Beta used in these studies, and not for lower amounts useful for healthy people.  This study  conclusions on Vitamins E,C and selenium are not  valid. 

The tone of this report was that "Antioxidants  including Vitamin E have little or even negative confirmed value in reducing risk of death."  Other and more pertinent studies for healthy people that I cite show that antioxidants can have significant benefits in reducing risks of cardiovascular diseases, cancer, and all-cause death.   I find that a deficiency in antioxidants may be a major single health deficiency in many US diets, and that their proper correction could prolong millions of lives. 

The study data used in this meta analysis were impossible to view usefully in the paper published because results of each study were presented in three or more tables with no useful organization.  For this reason I have extracted the actual data used in Table 1 following for more convenient view and re-analysis.

First, the conclusions were derived mostly from studies in which multiples of different nutrients were used. Any attempt using mathematical methods to develop such values from such deeply intercorrelated factors is subject to considerable error.  Thus it is essential that such values developed from such multiples be consistent with those developed directly from individual nutrients   I show following that this needed consistency was not obtained.  Many of these multiple antioxidants involved harmful amounts of Beta Carotene and any conclusion on higher than average risk probably was due to this problem. 

Second, the meta analysis included only results on clinical studies, many of which were both very small and conducted for very short times that had little useful quantitative value. Eleven of the studies had only 1 or 2 events in a comparison group.   Antioxidants appear to be related to atherosclerosis and are duration dependent, providing benefits of a reduced risk of about 3% per year of use. Clinical studies are most useful  for medicines and factors that provide near immediate benefits.  But the expected results from usual durations of these studies can be much too small to be detected.  The large majority of about 250 observation studies of antioxidants  included in my present data base show each of the four antioxidants can be  significantly beneficial in reducing risk of cardiovascular diseases, cancer and all-cause death. These studies usually included a likely duration of use that could produce significant benefits.  An exception is that amounts of Vitamin A and Beta Carotene above about 5,000 IU and 6 mg respectively appear to be harmful to health.  Amounts of Vitamin E much above 400 IU also may not be beneficial.

A third problem with the meta analysis is that most clinical study data were secondary on populations that had suffered heart or another diseases.  The body of research now available on each antioxidant shows no or very little effect of antioxidants on secondary populations that probably were taken other medications..  Only two statistically useful primary studies of healthy people individual nutrients were ncluded this meta analysis. There was no consideration of the extensive research done via observation studies and in studies that measure risk base on blood values of nutrients. The inferred presumption that all such research is invalid is baseless.

This was a statistical study of ratios from studies that had no scientific considerations of the mechanisms by which  antioxiands produce benefits. There was no recognition of duration of exposure that appears to produce benefits of about 3% per year.  It is interesting to observe what the actual study data found about the individual antioxidants.

For Vitamin E. There were four primary studies of Vitamin E, as studies E-1 through E-4.  Each of these included only one or two death events in a comparison group. Such studies have no useful accuracy and can destroy usefulness of any statistical comparison that includes them. They are anecdotal and should not be included in any serious analysis of a factor.  I use a minimum of 10 events for recognizing a study for use in my Life Ahead program.  The meta analysis thus included no useful primary clinical study data on Vitamin E for presumably healthy people. 

There were 12 secondary studies of Vitamin E.   These are listed as studies E5 through E16 in Table 1.  The average duration was 2.1 years; average of 900 IU/day of Vitamin E, and average risk was a value 0f 0.942.  This happens to confirm the expected time related duration factor of a 3% per year benefit for Vitamin E. of 0.93 for clinical type studies.

Five of these 12 secondary studies included too few events in a comparison group for scientific use.  If I subtract out the studies including only 1 or 2 events, an average result is a risk of  0.93.  If I subtract out all below 10 events the average is 0.95. The only important study is that of GISSI, E8.  Its results were a rik factor of 0.92, again close to my expected value of 0.90 for 3.5 years of nutrient duration.

Conclusion  This actual primary research on Vitamin E per se does not show a negative benefit, much less one that is "statistically significant and negative"   These results confirm a benefit that would  be expected from much other research on Vitamin E.  This is interesting  because most secondary clinical studies of antioxidant use for patients of diseases have not verified a useful effect for any antioxidant.

For Vitamin C:  There is just one study for Vitamin C, # C1 in the table following. It included just 3 events to produce a risk of 0.63 that his little significance.  Conclusion:  There is no information useful for Vitamin C included in this research that could support a claim of  either a benefit or negative.

For Selenium:  Just 3 studies were included. as S1 - S3. Both S2 and S3 have only one event in the comparison group and are worthless. The remaining study is the widely used and excellent one of Clark that showed a risk of 0.84 for  4.5 treatment years and 7.4 treatment plus follow-up years.  The expected value for Clark would be 0.87 for 4.5 years and 0.80 for 7.4 years,. again very close to my forecast for a 3% per year effect of duration..

Two studies included Vitamin E and Selenium:  CES-1 and CES-2. They both showed a benefit with the higher 0.87 risk value for CES-1 as the result most useful.  Two more studies included Vitamin E, Selenium, and 6 mg amounts of Beta Carotene that may be beneficial. .  Their average risk was 0.90, another  benefit. These four studies confirm further a likely benefit for Vitamin E and Selenium combinations from E5-E16 and SE- in amounts that would be expected from my Global Analysis of antioxidants.

For Vitamin A and Beta Carotene:  Three studies of Vitamin A included  2 useful ones, A2 and A3.  Both have extraordinary amount of Vitamin A, and or Beta Carotene far above the level of 5,000 IU of Vitamin A I found as maximum useful for health..  They showed as expected a negative benefit of 1.15.  The three studies of Beta Carotene did average a 0.92 risk value, but other data have confirmed Beta Carotene in such high mounts to be harmful.   All of these values of course are subject to substantial statistical variability. Other studies not listed in Table 1 included multiple nutrients with large amounts of Vitamin A or Beta Carotene that could seriously confound the valuation of another included  nutrient . I do not feel that valid values of individual nutrients can be derived from results of multiple nutrients that include large amounts of Vitamin A or Beta Carotene by the method used in this study.

In Conclusion::  This Meta Analysis has no merit and has provided seriously misleading conclusions that probably will harm the health of many people.  It deserves a balanced rebuttal.

References:  For far more extensive studies of antioxidants, please go to www.lifeahead.net, and access the 'Health Research Library'

Papers in the library include:

Vitamin E and Cardiovascular Disease.  This more extensive paper should be reviewed first

Vitamin E and Cancer    This paper provides a more extensive treatment of the problem of btaining valid research on cancer

Vitamin A, Beta Carotene and Cardiovascular Disease.

Vitamin A and Cancer.

Vitamin C and Cardiovascular disease

Vitamin C and Cancer

Selenium and Heart Disease

Selenium and Cancer

Antioxidants and Death from all Causes

Antioxidants, All Causes, a Global Analysis

Antioxidants, Death from All Causes, a Supplementary Analysis

                                                                                                       Table 1

                       Antioxidant Supplements from Clinical Studies and Risk of Death from All Causes