|"But Dr. the TV guy said, I would live longer than my normal weight cousin."|
"There is no such thing as an obesity paradox"
In fact the only paradox there is, is the ease with which scientists around the world have been fooled by numerous reports of a lower mortality risk for obese individuals than for normalweight individuals within the past decades (I guess that's wishful thinking controlled by their own beer-bellies ;-). As Greenberg points out, corresponding reports
"[...] contradict the well-accepted, empirically based idea that obesity confers elevated mortality risk. Some of these paradoxical reports resulted from studies that involved cohorts of elderly free-living persons, such as elderly US residents, US veterans, and residents of Jerusalem. Other studies used data from seriously ill patients, including patients on kidney dialysis, post–coronary revascularization patients, livertransplant recipients, and patients with conditions such as wasting disease, AIDS, cancer, chronic obstructive pulmonary disease, heart failure, acute myocardial infarction, and peripheral arterial disease.So exactly those patients, ah... pardon me, subject who are relevant for the few really (not new standards) normal weight individuals out there, right? Accordingly, one of the most frequently heard and in fact reasonable explanations for these counter-intuitive observations was reverse causation:
"Reverse causation is postulated to be caused by factors such as smoking and serious illness that simultaneously induce weight loss and increase mortality risk.These factors are theorized to increase mortality risk at low BMIs, and hence deflate mortality risks for obese individuals relative to normal-weight individuals, thereby yielding an artificially low mortality risk for obese persons. An analytic technique that is commonly used to abate reverse causation involves excluding smokers and participants with serious illness from the analysis. However, this technique cannot be applied in analyses involving samples of seriously ill participant." (Greenberg. 2013)To avoid being accused of falling for the same mistakes, Greenberg decided against the use of the usually subjects and resorted to data of which you'd expect that it had by now already been analyzed: data from the mortality-linked NHANES I, II, and III cohorts. Based on these datasets, which are based on 3 different 15-y follow-up periods: 1973–1988, 1978–1993, and 1991–2006., he set out to evaluate, whether the hypothesis that "mortality risk is lower for obesity than for normal weight only among elderly and/or serious ill Americans" (Greenberg. 2013) - and the results were intriguing:
- compared with normal-weight participants obese participants are older, nonwhite, nonsmokers, and nondrinker
- there was a (shockingly?) steady increase in the prevalence of obesity and a steady decrease in the prevalence of normal weight across the cohorts over time
- significant 2-way interactions were found for cohort and serious illness, cohort and smoking, cohort and sex, or cohort and age for mortality because of all causes - one or more significant 2-way interactions involving age and serious illness, age and sex, illness and sex, smoking and age, smoking and serious illness, or smoking and sex
- a significantly lower mortality risk for obesity than for normal weight was found only among men with serious illness and only in NHANES III
- of all other subgroups none achieved this "effect" - if present (e.g. older (aged.55 y)
men and seriously ill participants) the fat advantage was non-significant, even in NHANES III
- specifically for women, "[t]here was no evidence of lower mortality risk for obese women than for normal-weight women" (Greenberg. 2013)
An often heard, logical, but hard to verify alternative to explain the "seriously ill obesity paradox", is that it could be a result of "clinicians using appropriate diagnostic and therapeutic techniques more intensively and earlier in the disease process for obese male patients than for other patients." (Greenberg. 2013) - If that was the case, even the advantage for seriously ill-patients was nothing but and epidemiological edifice."If reverse causation is the explanation for the RR for all-cause mortality for obesity (with normal weight as the referent) being significantly lower than 1.00 only for men with serious illness, and only in NHANES III, then an increase in reverse causation would have occurred only in seriously ill men between NHANES I and III and would have manifested as a higher mortality rate among normal-weight, seriously ill men in NHANES III compared with NHANES I." (Greenberg. 2013)
To evaluate this, Greenberg thus used NHANES I as a reference and found that the mortality risk normal-weight, seriously ill men in NHANES III, with normal-weight, seriously ill men was 1.11 (0.40, 3.08; compared to NHANES I). A result which would suggest that "an increase in reverse causation between NHANES I and III was unlikely to be the explanation" (Greenberg. 2013) for the "obesity paradox" in the study at hand.
|Usually this would be the place for a "suggested read", but in this case this is more of a "must read" or "obligatory read: "Study Reveals Unsettling Data About How Fat We've Already Gotten Over the Past 40 Years. Plus: Macro Analysis of the Diets of the Leanest & Fattest Yields Surprising Results" (read more)|
- Greenberg JA. The obesity paradox in the US population. Am J Clin Nutr. 2013 Jun;97(6):1195-200.