expert opinion

Micronutrient studies: a statistical analysis – Part 2

November 1, 2012


Prof. Dr. Manfred Wilhelm, biostatistician, Ulm University of Applied Sciences, Germany

Study 3: Mursu et al. (2011): Dietary Supplements and Mortality Rate in Older Women – The Iowa Women’s Health Study.

The increased mortality risk reported by the Iowa study for older women who take dietary supplements is not generalizable, even though the study examined a total of 38,772 older women. The authors themselves simply talk about a possible association between the two in their conclusion: ‘In older women, several commonly used dietary vitamin and mineral supplements may be associated with increased total mortality risk.’

A detailed look at the paper reveals serious weaknesses in planning, execution and evaluation of the study, which means no concrete statement can be made about the harmfulness of dietary supplements. Originally, the study wasn’t even designed to test for differences in mortality risk between older consumers and those not taking dietary supplements. The original aim of the study was to investigate the relationship between demographic factors (such as nutrition and lifestyle habits) and incidences of cancer among post-menopausal women. The now published results of the Iowa study therefore address a question that was posed only after data collection. Precise details about the dose and length of time for which participants took supplements are not given. Participants were only asked three times over the course of the study period from 1986 to 2008 about their supplements intake. Even then, this was done using a tool which was not valid for testing dietary supplement intake. Possible switches from one group to another – i.e. from taking supplements to not or vice versa – were not recorded. As a result, as the statistical method that was chosen requires that the two groups be kept independent of one another, the statistical results based on these groups are worthless.

One of the most important quality criteria when it comes to planning clinical and epidemiological studies is that the starting conditions of both groups to be compared are similar enough to be comparable, i.e. struc-tural sameness. At the beginning of the Iowa study, there were already hugely significant differences between the older women taking supplements and the older women not taking supplements. Age aside, the two groups differed in every other aspect that was reported, particularly concerning their nutrition habits. The fact that the two groups were non-comparable in structure could therefore mean that the statistically significant difference in mortality observed between these two groups was more to do with personal differences than taking supplements. As the groups were not comparable, this therefore means the study is systematically distorted (biased). The quality and reliability of the Iowa study are therefore extremely limited. Even though these differences were later taken into account by the complex statistical models, this can only partly make up for the lacking quality and nonetheless leads to difficulties when interpreting the statistical results.

In studies where mortality is the dependent variable, life-threatening diseases should normally be taken into account. This was not the case for the Iowa study, which isn’t really surprising as mortality was not the originally-planned dependent variable. Good risk models are those that explain mortality risk based on a limited number of factors that may influence it. In the Iowa study, however, adjusted mortality risk was determined according to up to 16 different factors. It therefore remains unclear whether these influencing factors were significant enough, or even valid, to be included in the model. The different risk models that adjust for factors in this way deliver very different results: 47 analyses showed that supplements were associated with a significantly reduced adjusted mortality risk, 33 analyses showed that supplements were linked to an increased adjusted mortality risk. Such a robust and generalizable conclusion cannot therefore be made based on the results of an analysis conducted in this way."

Study 4: Klein et al. (2011): Vitamin E and the Risk of Prostate Cancer – The Selenium and Vitamin E Cancer Prevention Trial (SELECT).

The general claim that taking vitamin E for a number of years will increase the risk of contracting prostate cancer among healthy men cannot be substantiated by the SELECT study (2), which examined 35,333 parti-cipants from August 2001 to July 2011. A pre-planned interim analysis conducted in October 2008 was unsuc-cessful in showing a significant association between consumption of vitamin E, selenium or a combination of the two and the risk of getting prostate cancer. As a result of this interim analysis, the results of which were published a number of years ago, the study, which until then had been conducted as a blind study, was aban-doned early and participants were asked to stop taking the dietary supplements.

A second statistical analysis was conducted in July 2011 on all participants who had been tracked since the interim stage. This analysis found that the previous group of vitamin E consumers was at greater risk of developing prostate cancer. As the study participants had taken no vitamin E supplements between October 2008 and July 2011, an increased risk of prostate cancer cannot be attributed to vitamin E. Other factors not considered in this study must therefore have been responsible for this risk. Solely the study results from the time of the interim analysis in October 2008 are valid."

"Study 5: Omenn et al. (1996): Risk Factors for Lung Cancer and for Intervention Effects in CARET, the Beta-Carotene and Retinol Efficacy Trial.

As the participants of the CARET study (n = 18,314) were all heavy smokers or workers who had been ex-posed to asbestos, the conclusions of the study relating to the risk of catching lung cancer are not applicable to the general population. The associations between beta-carotene and vitamin A intake and increased lung cancer risk or mortality among those already at high risk of lung cancer are also not as clear-cut as they are made out to be in the study summary. There it states that, on the basis of all participants’ results, there is a significant link between increased lung cancer risk and high doses of beta-carotene + vitamin A. But if one considers the originally-planned study population, consisting of two separate groups – a pilot cohort (n = 1,845, 15 mg or 30 mg of beta-carotene) and an efficacy cohort (n = 16,469, only 30 mg of beta-carotene) – neither of these cohorts had a significantly increased lung cancer risk due to beta-carotene + vitamin A.

The results of two different statistical tests on various subgroups (men, women, those exposed to asbestos, former smokers, etc.) showed that in 34 out of 49 analyses, there was no significant increase in lung cancer risk due to intake of beta-carotene + vitamin A. The link between significantly increased mortality and beta-carotene + vitamin A that was reported in the study summary stands in contrast to the findings for life expectancy between those who took beta-carotene + vitamin A and those who did not: there were no significant findings between the two groups. Separate data relating to mortality for the pilot and efficacy cohorts are not given. For the subgroup of former smokers not exposed to asbestos, there were even potentially positive effects noted of beta-carotene + vitamin A: ‘…that the response to the study vitamins in former smokers with no asbestos exposure may not be adverse and might even be favorable.’ Such contradictory results mean that no universal conclusions can be made, neither for the heavy smokers and workers exposed to asbestos that were examined in the CARET study, nor for the general population.”

Based on: Prof. Dr. Manfred Wilhelm and Martin Braun. People unsettled by lack of communication. Vitamin Report 2012 – Articles on provision of micronutrients. TRIAS-Verlag, 2012.


  1. Mursu, J. et al. Dietary Supplements and Mortality Rate in Older Women – The Iowa Women’s Health Study. Arch Intern Med. 2011; 171(18):1625–1633.
  2. Klein, E. A. et al. Vitamin E and the Risk of Prostate Cancer – The Selenium and Vitamin E Cancer Prevention Trial (SELECT). JAMA. 2011; 306(14):1549–1556.
  3. Omenn, G. S. et al. Risk Factors for Lung Cancer and for Intervention Effects in CARET, the Beta-Carotene and Retinol Efficacy Trial. J Natl Cancer Inst. 1996; 88(21):1550–1559.