The International Journal of Obesity has just released a short article by Janet Tomiyama (UCLA) and Jeffrey Hunger (UC-Santa Barbara) et al–a team of psychologists. They analyzed NHANES data from 2005-2012 for about 75,000 individuals, and concluded that BMI status doesn’t correlate well with six concrete markers of cardiovascular and metabolic health–blood pressure, blood triglycerides and cholesterol, blood glucose, insulin resistance, and C-reactive protein. Extrapolating a bit from the NHANES study participant numbers, they conclude that millions of Americans–54 million–have been misclassified as unhealthy due solely to their BMI numbers.
According to their analysis, 47% of the overweight people in the study had healthy status (0-1 of 6 markers) other than their weight. About 30% of obese and even 16% of morbidly obese people had healthy status according to their protocol, whereas about 30% of those in the healthy BMI range had more than one actual cardiovascular or metabolic disease marker that would be ignored if only BMI is considered.
Is this really the death-knell for public concern over weight? Should it be?
Here’s how the UCLA press release puts it (with my emphases in italics):
But a new study led by UCLA psychologists has found that using BMI to gauge health incorrectly labels more than 54 million Americans as “unhealthy,” even though they are not. […]
“Many people see obesity as a death sentence,” said A. Janet Tomiyama, an assistant professor of psychology in the UCLA College and the study’s lead author. “But the data show there are tens of millions of people who are overweight and obese and are perfectly healthy.”
Incorrectly labels? Perfectly healthy?
The definition for “healthy” used in this study is 0 to 1 known risk factor for CVD and diabetes. But clinically, one risk factor is often enough to be of acute concern, especially if it’s untreated high blood pressure or blood glucose. Those generally need treatment sooner rather than later.
Furthermore, the study as posted on Hunger’s web page excludes obesity and overweight a priori from that count of risk factors for CVD and diabetes. I don’t know the absolute latest research consensus, other than what was in the Dietary Guidelines for Americans Advisory Committee’s report last February, but my general understanding is that weight does show some statistically independent influence on CVD at least. That picture may be changing as we learn more about its interactions with other risk factors, but if it’s still valid, weight should have been counted as one of the existing “known risk factors” along with the other markers and that would have skewed Tomiyama and Hunger’s analysis considerably.
Even without those considerations, the different weight groups classified by BMI cutpoints do in fact show a significant increase in health risk from one category to the next. Turn Hunger and Tomiyama’s percentages around and you see that 70% of people in the 18.5-24.9 healthy BMI range have 0-1 risk marker other than weight; 53% in the 24.9-29.9 overweight range have more than 1 marker, 70% in the 30-35 range have more than 1 marker, and 84% in the 35 and over BMI range have more than one marker other than weight.
Plot those crude percentages and you’ll see a very sharp rise in risk incidence between the healthy and overweight categories, a reversal of fortune from “most people healthy” to “more than half at risk,” with further solidification of “most people at risk” as you venture further into obese and morbidly obese. There’s really no debating that trend, even given the narrow way this team has defined “healthy.” To say nothing of “perfectly healthy.”
Researchers in biomedicine (i.e., physical as opposed to psychological medicine) have recently reexamined whether the current BMI cutpoints defining healthy, overweight, obese and morbidly obese are in the right places to describe most people’s 10-year risk of overt CVD events (heart attacks and stroke), diabetes, or all-causes mortality, or whether BMI is just a continuous gradient of increased risk without definable cutpoints. At last count, the conclusion was that the current statistical best-fit cutpoints are pretty much correct, even though the data for individuals have a pretty big spread (and that each BMI step or number still has incrementally higher risk than the one below).
The upshot: BMI categories are still a pretty good marker of the overall health status of Americans when you’re talking about trends. Crude, yes. Exceptions for athletes with much more muscle than fat, yes. But the numbers are still strongly in favor of using BMI as a general warning flag to check for more specific cardiovascular and metabolic disease markers in individuals.
It’s very odd to see a paper like this coming from a team of behavioral psychologists, which Tomiyama and Hunger are. They’re at least nominally outside their field here, doing a statistical analysis on physical health data, and the paper’s methodology and definitions (along with some of their position statements in the American Journal of Public Health and elsewhere) show a specific agenda toward deconsecrating BMI and downplaying overweight and obesity in public health policy. Given their application here, I really can’t agree with them that the role of obesity and weight should be played down in medical terms and public health guidance.
The authors’ stated purpose in doing this particular analysis now is laudable–the EEOC has a proposed rule out that employers could penalize workers by increasing their health insurance costs if their BMIs are over 25. The authors argue that this would add to weight stigma and cause prejudicial and undue financial harm to millions of workers without capturing the true picture of their health status.
In that light, it’s quite reasonable for the authors to argue that BMI alone is not a fair standard for excluding or penalizing anyone for insurance rates or costs, and that the concrete markers of incipient disease need to be considered. If they’d stuck to that statement, in the paper and elsewhere, I’d be with them.
My concern here is that (as usual?) news media reporters will run screaming, press release in hand (is it too much to ask that they actually look at the paper or at least the abstract and draw their own conclusions), to claim that weight doesn’t predict CVD and diabetes risk after all, or at all. I can smell the inch-high headlines already–not least because Tomiyama’s UCLA press release for the paper and the news coverage quoting Hunger at UCSB have both headlined the story that way, even though it’s a contention their journal paper doesn’t and frankly can’t support.
Hunger, a doctoral candidate, has mostly stuck to his legitimate conclusions and the specific context in which he undertook this research in the UCSB news release, but overstepped in a recent radio interview, echoing Tomiyama’s “perfectly healthy” quote above and saying people should focus on health-promoting behaviors (presumably other than weight control, which he left out) and “just forget about the numbers on the scale.” Tomiyama, as quoted above, clearly hasn’t limited her statements in the UCLA press coverage. It may not be all that surprising that she is a former student of Traci Mann, a behavioral psychologist whose own public pundit overstatements play pretty fast and loose with the evidence and with common sense when representing research on weight as a physical health risk factor. Both Tomiyama and Hunger imply strongly, if not state openly, that weight is not a health risk factor worth most people’s time. That’s a gross misconstruction and not supported by their analysis, and certainly not by the larger body of evidence in cardiovascular and diabetes nutrition research.
Both researchers had the ability, opportunity, right and, to be blunt about it, the academic responsibility to insist on fact checking and qualifying the scope of their findings for their own university news services and PR coverage. Judging from what came out, I’d say they coasted.
If a biologist, chemist, or physicist made a splashy, overstated and inaccurate announcement to publicize a research paper with methodological and analytical flaws as big as the ones in this paper, that would be considered a serious breach of research ethics. Just announcing it ahead of formal publication is considered bad form. They might lose grant funding, have their labs and papers put under administrative review, possibly lose their jobs. It happens a lot. The media are not kind about it. The universities are not gentle about it. The peer-reviewed journals are not forgiving.
Psychologists? I dunno. New York Times, Wall Street Journal…Oprah, Dr. Oz Show…Instant jukebox heroes.
Yes, there’s a driving need to deal with the potential policy flaw in the proposed EEOC rule, which is due for finalization in April. Yes, their research will make more noise in national public media if sensationalized this way. No, this is not a legitimate way to promote it. That’s letting the agenda drive the bus.