Wednesday, February 05, 2014

Yet Another Good Reason to Question Observational Diet Studies

Change in self-reported intake of fruits, vegetables, sweets and soft drinks by month
Or at the very least those that utilize food frequency questionnaires (FFQs) to account for dietary intakes (which is almost all of them).

I can't say I'm an ardent fan of research done with FFQs, and nor are many others (remember this letter to the editor begging journals to stop published research that utilizes FFQs?), but supporters will tell you that the errors of FFQs can be accounted for with fudge factor style statistics. While admittedly I'm no stats maven, I did find a study published yesterday in the Journal of Nutrition Education and Behaviour to lend support to my belief these studies aren't as useful as we might hope.

The study highlighted an obvious, and yet important issue that would further undermine FFQ utility - the month the data was collected matters. While the researchers looked at FFQ outcomes collected from adolescent residents of 36 countries, there were 3 - Canada, England and Norway - where data collections took place almost all months of the year. In all 3 researchers found that there were significantly lower reported levels of daily consumption of foods in January and February. In the remaining 36 countries, where FFQs were administered for shorter periods, researchers found lower reported consumption of fruit and sugared soda in the spring vs. the fall and the winter.

Now admittedly, the effect they found here was small, and in part may be explicable on the basis of seasonal food availability, though that wouldn't serve to explain the sugared soda or sweets results. Regardless of affect size or cause however, one thing's definitely clear, FFQs need to go, and I'm praying, in this day and age of smart phone ubiquity, that someone somewhere validates an app that combines daily recall with real-time photography, to help reduce error and markedly amp up the power of observational diet-related study conclusions.

Bookmark and Share


  1. It boggles my mind that researchers aren't already using apps. Two existing apps come to mind in particular -- My Fitness Pal and Lose It. Both are now tracking hundreds of thousands of users, who, if they're actually using the app properly, are consistently tracking everything they eat. Imagine if those apps were to make their data available to researchers (on an anonymized basis, of course)! It's an untapped goldmine!

  2. Eleanor11:17 am

    Larry Istrail, the medical student who developed the Ancestral Weight Loss Registry, has been working on an app to do this, called PhotoCalorie.

    He gave a presentation about both topics recently at Tufts Human Nutrition Research Center...

  3. Anonymous8:00 am

    The limitations of FFQ's are well known. This doesn't mean that the resultant data is useless, just that you have to be careful about the questions you ask. People are terrible at estimating portion size (both relative and absolute). Portion sizes have drifted up and thus the new larger portions become the new normal. People also know what they are supposed to eat, and thus over-report healthy food and under-report unhealthy food. That being said, FFQ can be reliably used to estimate overall dietary patterns. Dietary patterns are also a remarkably good and consistent predictor of multiple health outcomes.