Re: Efficacy of vitamin and antioxidant supplements in prevention of cardiovascular disease: systematic review and meta-analysis of randomised controlled trials
The vitamin and antioxidant review by Myung et al.  has severe shortcomings.
1) The study-level meta-analysis suffers from the ecological fallacy
“Ecological fallacy” means that study-level (group-level) analysis can lead to different conclusions than do corresponding individual-level analysis [2,3]. For this reason, examination of individual-level data is recommended, whenever feasible, in order to avoid the potential for the ecological fallacy introduced by study-level analyses [2,3].
For example, in Table 4 the authors divide the 50 included studies to two subgroups by the “duration of treatment” using 5 years as an arbitrary cut-off limit. They test whether a short treatment might differ from a long treatment by looking at the study-level average durations and the study-level average outcomes. However, such an approach cannot give valid information about the time-dependency of a treatment effect. In an individual-level analysis of the ATBC Study, we found that among participants aged 50–62 years at baseline with a dietary vitamin C intake above the median (n = 11,448 with 1,021 deaths), vitamin E started to increase mortality after 3.3 years so that RR = 0.99 (95% CI: 0.82, 1.19) during the first 3.3 years and RR = 1.38 (1.17, 1.63) thereafter . Adding the second vitamin E effect improved the regression model significantly (chi-square (1 df) = 7.1, P = 0.007).
Another time-dependent vitamin E effect in the ATBC Study was decrease in mortality after the participants had reached the age of 70 years. Among 2,284 ATBC Study participants with dietary vitamin C intakes above the median who smoked less than a pack of cigarettes per day, the survival curves of the vitamin E and no-vitamin E participants diverged at about 71 years and vitamin E extended lifespan by 2 years at the upper limit of the follow-up age span .
These kinds of time-dependent effects cannot be examined by comparing study-level averages. The above individual-level analyses refute Myung et al.'s conclusion that the duration of treatment does not influence the effects of any vitamin or antioxidant, which is based on the study-level comparison of 50 studies with diverse vitamins and antioxidants in Table 4.
It is possible that vitamins and antioxidants are beneficial for specific and narrow population groups and such subgroups can be searched for by analyzing individual-level data. For example, it is possible that age, sex, smoking, diet, exercise, etc. might modify the effect of some vitamins and antioxidants, so that some restricted population groups might benefit (or be harmed). Such questions cannot be examined by study-level analysis. In addition to the above mentioned heterogeneity in the vitamin E effect on mortality, its effects on the common cold, pneumonia, and tuberculosis were also heterogeneous over the ATBC study population in the individual-level analyses [6-9]. Therefore, study-level analyses would be misleading also in those cases.
2) The authors are combining apples and oranges
First, let us consider an analogy. If a researcher is interested in the effect of antibiotics on mortality caused by infections, and combines all diverse antibiotics to a single broad category of “antibiotics”, and pools all “antibiotic trials” together, people with basic background in clinical microbiology would consider such a project silly. Different antibiotics kill different bacteria, there is geographic and social group variation in the occurrence of pathogenic bacteria, the usage of antibiotics generates resistant strains (which vary geographically and over time, etc.), etc. The biology of antibiotics is very complex. It is obvious that a single universal estimate for “antibiotic effect on mortality" is meaningless.
“Vitamins and antioxidants” are also a heterogeneous group. Vitamin C is water soluble, vitamin E is fat soluble, beta-carotene is not an essential nutrient, selenium is a mineral, etc. From the point of view of biology, there is no justification to pool 50 trials with diverse substances on the basis that the substances belong to the broad category of “vitamins and antioxidants.” In a proper analysis, each vitamin and antioxidant should be analyzed separately.
For example, in Table 4, the authors examine the effect of “the duration of treatment” in the 50 trials with a cocktail of studies with vitamin A, vitamin B6, vitamin B12, folic acid, vitamin C, vitamin D, vitamin E, beta-carotene, and selenium. Vitamins D and E are fat soluble and therefore there can be lag periods before effects appear, but such lag periods are less likely for the water soluble vitamins. Analyzing the time-dependency of all vitamins and antioxidants together in this way means that the authors assume that all of them have identical time-dependency, which is unreasonable.
Thus, in addition to the problem of comparing study-level averages (point 1), the analysis suffers from a severe apples and oranges problem. The authors do not give any justification for pooling all studies with any vitamin and antioxidant together for a universal “any vitamin and any antioxidant” effect. In their Introduction, Myung et al. criticise other authors for restricting to "individual vitamins or antioxidants". However, specificity is an important goal of modern biomedicine, and of science in general, and specificity conflicts with pooling all vitamins and antioxidants together.
In conclusion, the review by Myung et al. has a number of shortcomings. Furthermore, having read over a dozen previous systematic reviews on vitamins and antioxidants it is hard to see any novelty in this paper. For example, the lack of benefit of vitamin E supplementation for the general population has been shown in numerous previous systematic reviews, only a few of which are cited. Nevertheless, the lack of effect of vitamin E for unselected general population does not refute the possibility that vitamin E might have beneficial effects, for example, on older males [4,5]. Such a hypothesis cannot be examined by using study-level meta-analysis.
1. Myung SK, et al. Efficacy of vitamin and antioxidant supplements in prevention of cardiovascular disease. BMJ 2013;346:f10. http://dx.doi.org/10.1136/bmj.f10
2. Berlin JA, et al. Individual patient- versus group-level data meta-regressions for the investigation of treatment effect modifiers: ecological bias rears its ugly head. Stat Med 2002;21:371-87. http://www.ncbi.nlm.nih.gov/pubmed/11813224
3. Ecological fallacy. http://en.wikipedia.org/wiki/Ecological_fallacy
4. Hemilä H, Kaprio J. Modification of the effect of vitamin E supplementation on the mortality of male smokers by age and dietary vitamin C. Am J Epidemiol 2009;169:946-53. http://dx.doi.org/10.1093/aje/kwn413
5. Hemilä H, Kaprio J. Vitamin E may affect the life expectancy of men, depending on dietary vitamin C intake and smoking. Age Ageing 2011;40:215-20. http://dx.doi.org/10.1093/ageing/afq178
6. Hemilä H, Virtamo J, Albanes D, Kaprio J. The effect of vitamin E on common cold incidence is modified by age, smoking and residential neighborhood. J Am Coll Nutr 2006;25:332-9. http://www.jacn.org/cgi/content/abstract/25/4/332
7. Hemilä H, Kaprio J. Vitamin E supplementation and pneumonia risk in males who initiated smoking at an early age: effect modification by body weight and dietary vitamin C. Nutr J 2008;7:33. http://dx.doi.org/10.1186/1475-2891-7-33
8. Hemilä H, Kaprio J. Subgroup analysis of large trials can guide further research: a case study of vitamin E and pneumonia. Clin Epidemiol 2011;3:51-9. http://dx.doi.org/10.2147/CLEP.S16114
9. Hemilä H, Kaprio J. Vitamin E supplementation may transiently increase tuberculosis risk in males who smoke heavily and have high dietary vitamin C intake. Br J Nutr 2008;100:896-902. http://dx.doi.org/10.1017/S0007114508923709
Competing interests: No competing interests