Wednesday, September 13, 2017

Diet soda and risk of stroke

The Controversy
A few weeks back, just about every major newspaper ran a headline that diet soda consumption causes stroke and dementia. The papers were referring to an investigation that was recently published in the journal Stroke.

The Study
In this study, 2,888 older men and women from the Framingham Heart Study Offspring cohort were followed for 10 years (Reference 1). Beverage intake was measured by means of a food frequency questionnaire that was filled out 3 times during the 10 year follow-up. When compared to subjects that never drank diet soda, subjects that had one or more diet sodas daily had a 296% greater risk of ischemic stroke and a 289% greater risk of Alzheimer’s disease. These are very big increases and the media went crazy with the headlines. But in typical fashion, the newspapers did not tell the whole story here. 

There are some pretty big problems with the methods in this study and a far better designed study performed the same analysis and came up with very different results.

Let’s start with some of the problems with the methods in the Framingham investigation:
1) For a prospective cohort study, this was a pretty small investigation that included only 2,888 subjects and 97 cases in the stroke analysis. The dementia analysis included 1,484 subjects and only 81 cases.

2) There was significant loss to follow-up. Only 72% of subjects completed all 3 food frequency questionnaires.

3) The subjects who drank a lot of diet soda were not nearly as healthy as those that did not. They had a higher waist to hip ratio, body mass index, had more hypertension, lower HDL cholesterol, more diabetes, atrial fibrillation and cardiovascular disease, exercised less and consumed more saturated fat. This would not be a problem if confounders were well measured and adequately controlled for. In my opinion, they were not, which brings us to points 4 and 5.

4) There was an incomplete measurement of confounders, particularly those related to diet. 

5) I believe there were some problems with the modeling in this investigation.

For me, these 5 factors really hurt the generalizability of this study. Harvard University published a study on diet soda consumption and risk of stroke that was much better designed and came up with a very different result (Reference 2). Let’s see some of the ways that this second study was better designed.

1) The Harvard study included 84,085 women from the Nurses Health Study for 28 years of follow-up and 43,371 men from the Health Professional Follow-up Study for 22 years of follow-up. This is compared to only to 2,888 subjects in the Framingham study. The Harvard study included 4,354 strokes while the Framingham study had only 97.

2) Follow-up rates for the Harvard Study were 96% in men and 97% in women. In the Framingham cohort, only 72% of subjects completed all food frequency questionnaires.

3) The Harvard study measured and controlled for many more confounders than the Framingham study. Here is a list of confounders that the Harvard group included that the Framingham group left out of their study: consumption of red meat, poultry, fish, nuts, whole and low fat dairy, fruits and vegetables, cereal fiber, alcohol, parental history of myocardial infarction, trans fat, multivitamin use, aspirin use, vitamin E use, and menopausal status. The Framingham study controlled for diet by including a variable that measured how closely subjects followed the 2005 dietary guidelines. To me, this is inadequate because the guidelines from 2005 don’t really reflect all that can be healthy about one’s diet.

4) The Harvard study included every possible variable in the final model. In other words, they controlled for every potential confounder in the statistical model that they presented in their final results. In the Framingham study, the researchers did not include waist to hip ratio in their final model. This is a major confounder of the relationship between diet soda consumption and risk of stroke because people who are overweight have a much higher risk of stroke and people that are overweight are often more likely to drink diet soda (in an effort to lose weight). I’m not quite sure why this was left out, but the fact that they did not measure for all potential confounders and did not include weight in their final model leaves open the strong possibility of residual confounding in this study.

The Harvard study found a 16% increased risk of stroke for those that consumed 1 or more diet sodas per day when compared to those that did not drink diet soda. This is a far cry from the 296% increased risk found in Framingham. In my opinion, the Harvard results are much more in line with the true association between diet soda consumption and risk of stroke. 

Conclusions And Recommendations
I apologize if this post is a little long or technical. The media has a way of just running with a headline without really understanding the methods or limitations of the study in question. This causes a lot of confusion in my field.

Let me get one thing straight, I am not an advocate for diet soda. It is much better to drink water than any type of soda. I view diet soda the same way as I think of foods like bacon, butter, steak or cheese:  If you eat a ton of these foods, they can cause problems for you. If you have them only occasionally, they are not a major problem. For those that don’t eat sugar at all, an occasional diet soda once or twice a week can be a real treat. 

Incidentally, this level of consumption was not associated with stroke in the better designed Harvard study. In fact, in the discussion section of the Harvard study, the authors mention that the slight increase in risk of stroke with daily diet soda consumption should be interpreted with caution. They mention that there is no known mechanism between diet soda and incident stroke and that in previous analyses in the Harvard cohorts, diet soda has not been associated with weight gain, heart disease, or type 2 diabetes.

1) Pase MP, Himali JJ, Beiser AS, et al. Sugar and artificially sweetened beverages and the risks of incident stroke and dementia. Stroke 2017 DOI: 10.1161/STROKEAHA.116.016027.

2) Bernstein AM, de Koning L, Flint AJ, et al. Soda consumption and the risk of stroke in men and women. American Journal of Clinical Nutrition 2012; 95:1190-9.

No comments: