Rarely had one seen doctors go at each other such as they did after the publication of a meta-analysis released earlier this year on the relationship between total mortality and obesity. A research group under Katherine M. Flegal of the U.S. Food and Drug Administration Center for Disease Control and Prevention viewed over 7,000 scientific published papers in producing a publication in the Journal of the American Medical Association (2013, 309 (1):71-82). Remaining for a meta-analysis at the end were 97 papers of higher quality, involving a total of at least 2.88 million subjects who had participated, 270,000 of whom had died. It was therefore not a very small data pool. What the CDC experts then produced was a relatively simple analysis. They took the usual BMI categories, that is normal BMI (18.5 up to 25), overweight (25 up to 30), grade I obesity (30 up to 35) and grade II / III obesity (35 and higher). These were correlated with total mortality.
Results: the overweight, with a Hazard Ratio of 0.94, compared to those of normal weight had a significantly lower risk of death. The risk for obese people in all categories was, with a Hazard Ratio of 1.18, significantly elevated. Where only those people with grade I obesity were analysed, the risk of death was the same as for normal weight. So it was primarily the very obese people who made an impact on increased mortality in the general population of overweight individuals. Presumably because they had guessed what they would be unleashing, the scientists provided a second analysis in which they considered only those studies which were adequately adjusted for age, sex and smoking status, but not for chronic diseases. The result went in the same direction.
Not so squeamish: U.S. experts cause a stir
This publication is still making waves. In Germany it has been widely reported, the general tone being that a bit of excess weight is not so bad after all. In the U.S. the main thrust of the media coverage was similar. In particular however it was the medical establishment there that elbowed its way through to also have its say. “A pile of rubbish”, states Walter Willett of the Department of Public Health with regard to the meta-analysis. He even suggested publicly that it should not even be read. He did not say that one should instead rather read his own work, but presumably that is what he meant.
Unfortunately, from a German perspective during the whole hoopla another study published in the European Heart Journal at the same time was somewhat overshadowed (2013 34 (4) :268-277). This study is based on data from the Bavarian stroke network TEMPiS, that is an unselective cohort of stroke patients in clinics “in the hills and dales”. The TEMPiS neurologists looked over how high the risk of death in stroke patients is with reference to the BMI categories. These results are also quite remarkable. Stroke patients who are overweight according to BMI categories have, in comparison with normal-weight people, a Hazard Ratio for all-cause mortality of 0.69. With grade I obesity the Hazard Ratio was 0.50, with grade II / III obesity as low as 0.36, all significantly high.
Boundary shift changes the picture
The question now is: What to do with this information? Not reading is in any case not an option. Accurately reading might help. Whoever interprets from the meta-analysis in JAMA that each additional pound prolongs life – as some media reports have suggested – has simply read it wrong. Across all categories of the overweight the total mortality compared with normal BMI class is higher by 18 percent. And for grade I obesity, it is at any rate not lower, but “merely” equal. Something comes into play obviously in the group who, according to the BMI criteria, are overweight people. These at any rate clearly live longer than the normal-weight people.
That this applies only to the overweight people who are healthy is not true, even though many professional commentators have claimed such. The authors have with their second analysis nonetheless made an effort to come up with, what is in their view, “correctly” adjusted studies, ie studies that were adjusted only for sex, age and smoking status, to take into consideration the factor of obesity-associated illness.
Well trained = Overweight?
The crucial problem of the meta-analysis is another: its stratification via BMI. And here we are at a point where we really learn nothing new. An assessment of excess weight and obesity only using BMI is already problematic because in the lower range of the BMI category “overweight” a lot of well-trained athletes turn up in statistics who – this is the current hypothesis – may be responsible for the surprisingly good performance for the overweight group in the U.S. meta-analysis. More revealing, as every doctor knows, would be waist circumference or the waist-hip ratio. This however in many studies is simply not available.
Aside from this category covering “slightly overweight” people, there may perhaps exist an additional difficulty in a second “problem area” of BMI, namely the transition region between underweight and normal weight. In the meta-analysis by Katherine Flegal, the underweight according to BMI people have a significantly increased risk of death. So another reason for the poor performance of the normal weight group according to BMI could be that the lower limit of normal weight (BMI as 18.5) is far too low. It would be interesting, therefore, to see what would happen if the normal interval of BMI would be shifted to a test range between say 20 and 27.
More reserves in an acute emergency?
An analysis of this unfortunately does not exist in Katherine Flegal’s work, but in another source such data clearly exists. For this reason as well, the virtually global excitement over the Flegal meta-analysis is not really comprehensible. In 2008, the New England Journal of Medicine published a European study in which smaller intervals for the BMI groups were chosen (New England Journal of Medicine 2008, 359:2105). Results: the lowest mortality was found at a BMI of about 25 for men and 24 for women. And overall within a BMI corridor of about 21 up to about 28 there are no major differences. Conclusion: monomaniac fixation on BMI is not a good idea.
Nevertheless: the results of the German stroke study cannot be explained away by shifting the BMI categories. Much more so, in this study the relative benefit rises fairly steadily with BMI. However the situation here is of course completely different: in the meta-analysis from the United States it is about ”chronic” overall survival, with stroke patients it is about acute mortality in a life-threatening situation. Perhaps the rather flat presumption that every pound means more additional reserves in such situations has a justification. This is yet again just a hypothesis.