Given all the commotion over the WACS Antioxidant Study, I contacted Dr. Nancy Cook who served as the lead bio-statistician on the study. She has agreed to let me publish her verbatim replies to a number of questions that are central to the debate. Here they are!
HeartHawk: What was the significance of setting the noncompliance threshold at using 2/3 of supplied study agents?
Dr. Cook: The threshold of 2/3 was pre-specified, but not particularly based on biologic activity. However, I can say that most of the women in the study either took all (or at least 90%) of the study pills or none. We didn't have many at intermediate levels, so the exact choice of cut point would probably not make much difference to the analysis.
HeartHawk: Please comment regarding the significance of the improved risk-ratio for Vitamin-E when noncompliant subjects were censored.
Dr. Cook: It's true that some significant effects emerged in our analyses censoring on noncompliance. This analysis, though, is not intent-to-treat, and can be subject to bias since those who take the study pills are likely to be very different from those who don't. It's also possible that intervening changes in risk factors are related to the lack of compliance, possibly differentially by treatment group (similar to confounding by indication). The outside use of these supplements also needs to be taken into account. The analysis censoring on compliance thus needs to be interpreted with caution. The results among compliers could be due to this self-selection bias, to a true preventive effect in those who keep taking the vitamin E regularly, to a possible early effect that goes away with time, or to a false positive finding due to the multiple comparisons. At this point we can't separate these out. So, while some of the findings for vitamin E are indeed tantalizing, they don't hold up to strict statistical or clinical trial standards. And the body of evidence for anti-oxidant trials in general has largely been disappointing. The most straightforward interpretation is based on our primary result, that there is overall no effect of this particular supplement on the composite endpoint. We cannot trust the secondary analyses to be reliable at this point.
HeartHawk: Any thoughts or comments about what WACS results may intimate about the value of naturally occurring antioxidants in foods as compared to supplements in any dosage or combination.
Dr. Cook: It's true that not everything is known about anti-oxidants, either by type or in combinations. As suggested in the discussion in the paper, some scientists have suggested that other forms of the anti-oxidants (eg. gamma- vs alpha- tocopherol) may be more beneficial. It's also possible that it is a combination of nutrients that has an effect. At this point it seems from a vast array of observational studies that diets high in fruits and vegetables are protective, which has also been supported by intervention studies such as DASH. So perhaps we haven't been studying the right form of anti-oxidant, or maybe a combination is needed, or maybe it is other nutrients in fruits and vegetables that are protective. There is still a lot that is uncertain. The best advice to the public is to eat a diet high in fruits and vegetables as well as to follow other healthy lifestyle habits, such as exercising regularly, rather than relying on supplements.