They did not prove that diet soda causes Alzheimer’s Disease. THEY DID NOT!

I know what the headlines say.

Congratulations, media! You succeeded. You even managed to panic my wife. Here’s the abstract (highlighting mine):

Background and Purpose—Sugar- and artificially-sweetened beverage intake have been linked to cardiometabolic risk factors, which increase the risk of cerebrovascular disease and dementia. We examined whether sugar- or artificially sweetened beverage consumption was associated with the prospective risks of incident stroke or dementia in the community-based Framingham Heart Study Offspring cohort.

Methods—We studied 2888 participants aged >45 years for incident stroke (mean age 62 [SD, 9] years; 45% men) and 1484 participants aged >60 years for incident dementia (mean age 69 [SD, 6] years; 46% men). Beverage intake was quantified using a food-frequency questionnaire at cohort examinations 5 (1991–1995), 6 (1995–1998), and 7 (1998–2001). We quantified recent consumption at examination 7 and cumulative consumption by averaging across examinations. Surveillance for incident events commenced at examination 7 and continued for 10 years. We observed 97 cases of incident stroke (82 ischemic) and 81 cases of incident dementia (63 consistent with Alzheimer’s disease).

This study used the Framingham Heart Study Offspring cohort, which began in 1971. The 10-year incidence of stroke and dementia began in 1998-2001. They excluded people with significant disease before the start of the monitoring period.

Did the participants differ by race or ethnicity? I have no idea. I do know, however, that the authors write about the “absence of ethnic minorities, which limits the generalizability of our findings to populations of non-European decent.” Was that in the coverage you read?

Did they differ by socioeconomic status? No idea. Did they abuse drugs? Work or retire? Live alone or with someone? Have a family history of disease? No idea.

Did they acknowledge that different artificial sweeteners are different molecules with likely different effects or implications? No.

Were there multiple comparisons, meaning some results might be due to chance? Yep. Did they rely on self-report, which might mean recall bias comes into play? Yep.

Was this an observational study? Of course.

Was all of that in the coverage you read?

The study reported the hazard ratios for the Model 2 (which adjusted for demographics, diet, physical activity, and smoking, but still missed a lot, noted above). There was a Model 3, which also accounted for various cardiometabolic factors, but the results weren’t as dramatic. Anyway, in Model 2, compared to drinking no diet soda at all, those who drank at least one a day had HR 2.96 for ischemic stroke and HR 2.89 Alzheimer’s dementia.

Reported in the small text under table 2 is that in Model 2, for this result, there were a total of 58 ischemic strokes in 2137 participants. So… it was rare. For Model 2, there were 47 diagnoses of Alzheimer’s Disease for 1087 participants. Again, pretty rare. (And that is if I’m reading it correctly. It says N/events, and I think they may have meant events/N).

I have pored through the paper and the data supplement, and I can’t find the actual rates of disease reported for each group. I don’t know how they differed.

This type of study, and any discussion of its meaning, would be full of caveats.

It’s an observational study, and it cannot show causation. It’s a well-known and limited dataset, which almost entirely lacks minority participation. They could control for some things, but many other (and important) factors couldn’t be accounted for. There were multiple measurements, and the analysis did not adjust for them. There are other studies which show different findings. The overall rates of dementia and stroke were low in this study, and therefore the scary relative differences aren’t likely as scary in terms of absolute differences. Those absolute rates weren’t clearly reported. There’s also no evidence that changing your behavior with respect to drinking diet soda will change any of these outcomes at all.

Was that in the coverage you read?


P.S. If you do go look at the paper, make sure to note the y-axis labeling in Figure 2. Hint: It doesn’t go from 0-100.

Hidden information below


Email Address*