In this study we surmised that the non-significant increase in pneumonia risk by vitamin E supplementation in the large study cohort might camouflage a greater effect in participants with low body weight because the dose per weight is greater for them. Consistent with our expectation, we found a significant effect by vitamin E supplementation on participants with the lowest body weight, whereas vitamin E had no effect on participants in the middle range of body weight (Table 2). The harm of vitamin E supplementation was restricted to low-weight participants with high dietary vitamin C intake (Table 3, Fig. 1A).
We also observed increased pneumonia risk by vitamin E supplementation in participants with the highest body weight (Table 2). This finding was unexpected and is not consistent with our rationalization for the subgroup analysis by body weight. It is possible that obesity makes people more sensitive to the potential harms of vitamin E. Although our subgroup selection was based on body weight and not BMI, the heaviest-weight participants had a high median BMI of 33.5 kg/m2 and, furthermore, the harm was more pronounced in participants with BMI over the median. In this subgroup, the harm of vitamin E supplementation was also restricted to participants with dietary vitamin C intake above the median (Table 4, Fig. 1B).
In the heavy weight participants with high dietary vitamin C intake, there was a lag period of 1.5 years before cases of pneumonia started to occur in the vitamin E group (Fig. 1B). Vitamin E is a fat soluble antioxidant and it is accumulated in the adipose tissue. In a one-year pharmacokinetic study, high dose α-tocopherol supplementation increased the ratio of α-tocopherol to γ-tocopherol in the adipose tissue continuously over the study , and in a three-month study, high dose α-alpha-tocopherol increased buccal mucosa cell vitamin E levels over the study , indicating that α-tocopherol accumulation in the adipose tissue is a process of months or years. Accordingly, the lag period in our high weight participants may be explained by slow accumulation of vitamin E in the adipose tissue before the adverse effects start to appear (Fig. 1B).
We assumed that vitamin E supplementation might be more harmful for participants with high dietary vitamin E intake because supplementation could lead to higher systemic levels compared with participants with low dietary vitamin E intakes. In the low body-weight participants we saw a tendency in the expected direction, yet the interaction between dietary and supplemented vitamin E was not statistically significant (Table 3).
Given the proposal that vitamin E supplementation would improve the immune system of elderly people [7, 8], our results are alarming. We do not have a biochemical explanation for our findings indicating that vitamin E may increase pneumonia risk in some population groups. Nevertheless, previous reports have indicated that vitamin E may have harmful effects on the immune system [10–14].
In this study, we carried out several subgroup analyses, which might lead to a few low P-values simply because of the multiple comparison problem. However, at both ends of the body-weight scale, the risk of pneumonia was modified by the same variable, namely dietary vitamin C intake, and in both cases the harm was limited to the subgroup with high dietary vitamin C. Furthermore, previously we found that vitamin E supplementation increased the risk of tuberculosis in participants with high dietary vitamin C intake , and in this respect the current findings are consistent with the earlier. Furthermore, the P-values testing the divergence between the vitamin E and no-vitamin E groups in the identified subgroups are particularly small, P = 0.002 (Fig. 1), and based on reasonable numbers of cases among several hundreds of participants. On these grounds, the harm observed in the identified subgroups probably is real and not explained by multiple testing.
There is evidence indicating that vitamins E and C interact, and based on biochemical studies, it was proposed that high doses of vitamins C and E might be beneficial for smokers because of the interaction . Furthermore, co-supplementation of high-dose vitamin C and vitamin E is common . We found an effect in the opposite direction so that vitamin E supplementation was harmful when combined with high dietary vitamin C intake (Fig. 1). The vitamin E effect modification by dietary vitamin C was not explained by other substances in fruit, vegetables, and berries. We found a similar harmful interaction between high dietary vitamin C and the effect of vitamin E supplementation on tuberculosis risk . Thus, our findings should lead to caution against a combined high-dose vitamin C and vitamin E supplementation, even though the harm may be restricted to selected groups among smokers.
The current US nutritional recommendations  and a recent review  conclude that vitamin E is safe at levels up to 1000 mg/day. However, our findings of the ATBC Study indicate that some population groups may be harmed even by a low dose of 50 mg/day (Fig. 1 and refs. , ). In addition, a Dutch study with elderly people found a significant increase in the severity of respiratory infection by 200 mg/day of vitamin E . Consequently, in some population groups the upper limit of the safe range may be substantially lower than the levels considered conventionally safe.
The two subgroups in which vitamin E increased the risk of pneumonia risk, i.e., the lowest and highest body-weight groups with high vitamin C intake, are small and consist of only 5% of the participant population of the current study (1081/21,657). If these two subgroups are excluded from the study population, the point estimate for the vitamin E supplementation effect on pneumonia risk is reduced from +14% to +4.9%, which demonstrates the restriction of the harm of vitamin E supplementation to these two small subgroups. Nevertheless, the proportion of participants harmed in these two small subgroups, as measured by the increased occurrence of pneumonia, is substantial. In the low body-weight participants with high vitamin C intake, one out of every 13 participants got pneumonia during the trial because of vitamin E supplementation (the number needed to harm, NNH, is 13). In the high body-weight participants with vitamin C intake over the median, the NNH is 28.
There is evidence that vitamin E supplementation may reduce the risk of the common cold and pneumonia in restricted population groups which are so far poorly defined [18, 19, 30]. However, the observed harms of 50 to 200 mg/day vitamin E supplementation in other population groups indicate that vitamin E can also be detrimental to health. Vitamin E self-supplementation should be discouraged until sub-populations that clearly benefit from the supplementation are characterized properly.