According to reporter Nick Tate at Newsmax Health, “If you have bad knees you might want to lay off the soda. New research has found that men with osteoarthritis of the knee who drink sugary soft drinks fare worse than those who avoid the sweet stuff.”
Maybe. Maybe not.
The research reported by Tate was presented at this week’s annual scientific meeting of the American College of Rheumatology. It’s not an actual “college,” but a professional association of physicians who specialize in arthritis and other musculoskeletal diseases. The annual meeting showcased over 600 oral and poster presentations authored by over 3,500 researchers. The abstract and PowerPoint slides for the presentation on arthritis and soft drinks can be obtained from the meeting website. Note that this research study is not yet published, and it has not been peer reviewed.
According to the abstract, researchers followed over four years 3,066 knees of 2,149 participants with osteoarthritis confirmed by x-ray. For each participant, they obtained dietary information from a Block Brief Food Frequency Questionnaire, but apparently only at baseline. The “Block Brief” questionnaire “contains a reduced food list (about 70 food items) and takes 15-20 minutes to complete. It was designed to provide estimates of usual and customary dietary intake.” NutritionQuest, the purveyor of the questionnaire, acknowledges that it has not been validated. The purpose of the study was to determine whether changes in joint space width (JSW) over time were associated with the number of non-diet soft drinks consumed per week.
What confounders did the researchers control for?
According to the abstract, researchers controlled for age, race, education, marital status, household income, employment, follow-up time, depression, knee injury and knee surgery, smoking, milk intake, total energy intake, baseline K-L grade and JSW, weight change, the changes of rim distance and beam angle. No other dietary information obtained from the questionnaire appears to have been used as a control, and the list of non-dietary factors they did not control for is endless.
What did researchers find?
For every tertile of Body Mass Index (BMI), researchers observed a statistically significant difference in the average change in JSW for men who at the beginning of the study consumed five or more soft drinks per week. Average changes were not statistically significant for lesser amounts of consumption. For women in the lowest BMI tertile, the average change was statistically significant for those who consumed ≤ 1 and 2-4 soft drinks, but for women in the middle and higher tertiles the difference in average change in JSW was not statistically significant.
What do these results mean? Probably nothing. Unlike the reporter, the abstract is careful to avoid drawing strong inferences. The researchers say only that “frequent consumption of soft drinks may be associated” with faster arthritis progression. They are not even confident that an association actually exists. Nonetheless, in a conclusion that calls for more research, the researchers imply that soft drink consumption causes arthritis to get worse: “Replication of these novel findings in other prospective studies demonstrating the reduction in soft drink consumption leads to delay in [arthritis] progression are needed to test this hypothesis.” Reporters untrained in science or too quick to discover a story could easily focus incorrectly on just the middle of the sentence: “soft drink consumption leads to delay in [arthritis] progression.”
Is there anything else going on?
A look at the researchers’ PowerPoint slides suggests that participants with different levels of soft drink consumption are different on many other margins. Their Table 1 shows:
- Sex: 29% of participants consuming no soft drinks and 53% of those consuming ≥ 5 soft drinks per week were men.
- Race: 84% of participants consuming no soft drinks and 63% of those consuming ≥ 5 soft drinks per week were non-hispanic whites.
- Education: 40% of participants consuming no soft drinks and 29% of those consuming ≥ 5 soft drinks per week had college educations.
- Income: 12% of participants consuming no soft drinks and 24% of those consuming ≥ 5 soft drinks per week had household income ≤ $25k.
- Employment: 53% of participants consuming no soft drinks and 69% of those consuming ≥ 5 soft drinks per week were employed.
- Depression: 7% of participants who consumed no soft drinks and 15% of those consuming ≥ 5 soft drinks per week were classified as depressed.
- Smoking: 5% of participants consuming no soft drinks were current smokers, but 16% of those consuming ≥ 5 soft drinks per week were current smokers.
It is not obvious what to make of these differences. While the researches say they controlled for them, we cannot tell if their efforts to control made any difference. Controlling for confounders is crucial, but controlling for spurious correlations is not.
A plausible (and perhaps more likely) explanation is that none of these factors — including soft drink consumption — have anything to do with the progression of arthritis. This study may be an example of publication bias, the phenomenon in which positive results are more likely to be reported because positive results are inherently more interesting. That is especially so when they confirm to a popular narrative, such as the notion that the sugar in soft drinks is unhealthy but the sugar in fancy coffee drinks is not.
Why are the results of this study attributed to Harvard?
It is the practice of Neutral Source to attribute work to actual authors, unless the authors cannot be identified, or they work for an institution that has endorsed the work as its own. Public relations outfits, and the newspapers who recycle their press releases, generally do not follow this practice. They prefer to credit the employers of the researchers who actually do the work. This practice is consistent with branding.
Consistent with this practice, Tate and Newsmax attribute these research results, and the medical advice, to Harvard University, the employer of one of four researchers funded by an NIH grant called the Osteoarthritis Initiative (OAI). Harvard is probably cited because it is named on the slides. That begs the question why Harvard is named. It is better known than the employers of the others (Tufts Medical Center and Brown University Alpert Medical School, where the Harvard researcher also holds an adjunct appointment). More generally, universities and their employees both gain when research results are attributed to the university because the university’s brand establishes greater scientific authority. This is especially so when, as in this case, the lead researcher is only an assistant professor, and thus has virtually no brand of his own.
In this case, none of the three universities promoted this research in a press release. A press release was issued, however, by the American College of Rheumatology. Professional societies also are interested in branding, and both the professional society and the researchers whose work they promote gain from it. Precisely because the American College of Rheumatology promoted this study it garnered the attention of Newsmax — and no doubt many other media outlets.
Science-by-press-release is inherently suspect
The advancement of scientific knowledge proceeds in an orderly, if unfortunately slow, way. This process is incompatible with the business of science, which requires more immediate gratification in order to show funders that their past investments were worthwhile and that future investments should be continued. When universities use press releases to promote the work of their researchers, they are trying to get their brands more shelf space in the competition for scarce research funding. A successful press release is one that garners media attention, for media attention is what makes a university’s brand more visible and enjoy greater respect.
For this reason, it is a sensible precaution to be suspicious of any research results promoted by press release.