Last month I wrote about a paper that linked diet soda to an increase in cardiovascular events. The authors (at least first author Dr. Gardener) gave some notes of caution in interpreting the results to some journalists, but overall I strongly disliked how it was covered. It is one study with important limitations that many don’t seem to understand. Aside from Yoni, I didn’t see any critical vetting of the paper.
Luckily we didn’t have to wait much longer for a stronger research design. A new paper by Kiyah Duffey and colleagues sought if food patterns do change the diet beverage-cardiovascular risk link, as there was little statistical adjusting for potential diet-related confounders in the 3 epidemiologic studies that found positive associations between diet soda and metabolic syndrome and diabetes (see previous post). They also point to the paper by de Koning, Malik, Rimm, Willett, and Hu that found no effect in the Health Professionals Follow-Up Study cohort of artificially sweetened beverages on diabetes risk in men after adjusting for a number of confounders- more than the previous studies but still overall diet pattern was not considered.
The new study used data from the 20 year CARDIA prospective study, of 4161 subjects. Diet information was obtained at baseline by a diet history questionnaire followed by a diet history over the previous month. Foods and beverages were classified by an algorithm into 43 food groups for a cluster analysis by diet patterns. The CARDIA study has follow-up exams/diet data 6 times after baseline, but the authors state that cluster analysis with 7 and 20 years after baseline showed stable diet patterns so they could use just baseline data. Diet beverage consumers were classified as “consumers,” and non-drinkers as “nonconsumers.” Diet patterns were divided into “prudent,” or higher in fruit, fish, and whole grains, and “Western“, or higher in fast foods, refined grains, and sugar-sweetened soda. Metabolic syndrome components (i.e. waist circumference, fasting glucose, blood pressure, serum triglycerides, and HDL) were assessed at each of the 6 exams over the course of the study.
The following self-reported data was also collected at baseline: race, sex, age, smoking status, highest education, physical activity score, and family structure. These were included in statistical adjustments. So there are still some inherent limitations in self-reported data, only using baseline data (though they show this is relatively stable over time), sample size, etc but such is the difficulty of nutrition research on tight budgets. This study also did not stratify by dose as Gardener et al. did.
Here is an example of why it is so difficult to disentangle what effect diet beverages really have:
- people who didn’t consume diet beverages consumed more “healthy” nuts, seeds, vegetables, and milk, but also more “less healthy” high-fat refined grains and sugar-sweetened soda
- people who didn’t consume diet beverages who consumed a Western diet pattern got more calories from snacks and fast food than people who did consume diet drinks and ate that pattern
- people who didn’t consume diet beverages who ate a prudent diet pattern got more calories from fruit and low-fat refined grains.
- people who didn’t consume diet beverages in both diet patterns consumed more total calories (interesting find!)
Unsurprisingly, the prudent diet group had a significantly lower risk of metabolic syndrome and several individual components than the Western group (note that it is not all diet responsible for this, as this group also tends to be more educated, active, and have lower BMIs, etc).
Without adjusting for diet pattern, those who didn’t consume diet beverages (nonconsumers) had a lower risk of metabolic syndrome but only 1 of the individual components was significant. Here is the table when diet pattern and beverage consumption was considered together:
Although those consuming the prudent diet pattern that didn’t drink diet beverages had significantly lower metabolic syndrome incidence compared to the Western diet pattern/consumers of diet drinks, it is notable that Western/nonconsumers of diet drinks wasn’t statistically different. In addition, there is little consistency in the results for individual components of metabolic syndrome: sometimes risk is lower if you are not consuming diet drinks, but sometimes it is lower if you are consuming them, and any significance was only reached for the prudent diet pattern. Together, it seems likely that there is no real relationship, though confident conclusions of course cannot be made either way for the various results. These results make little sense in the light of a lack of mechanistic theories, and based on the research we have so far I think those making suggestions to drop diet beverage consumption because of supposed risks is highly inappropriate.
Duffey, K., Steffen, L., Van Horn, L., Jacobs, D., & Popkin, B. (2012). Dietary patterns matter: diet beverages and cardiometabolic risks in the longitudinal Coronary Artery Risk Development in Young Adults (CARDIA) Study American Journal of Clinical Nutrition DOI: 10.3945/ajcn.111.026682