The Obesity Paradox: Perception vs Knowledge

To know that we know what we know, and that we do not know what we do not know, that is true knowledge.

Henry David Thoreau (1817-1862)

Data from several epidemiological studies, including the Framingham Heart Study, have found a positive relationship between obesity and mortality, in particular cardiovascular mortality.^1-4 However, several cross-sectional, retrospective database studies, including an interesting study from a clinical US Veterans Administration population in the current issue of Mayo Clinic Proceedings, have found an inverse correlation between body mass index (BMI) and mortality, often termed the obesity paradox.^5-12 The paradox from these cross-sectional databases implies that it could be healthier to be overweight than to be normal or low weight. This is potentially a dangerous message to promulgate from retrospective data in an environment saturated with an obesity epidemic and obesity-related conditions such as type 2 diabetes mellitus and coronary heart disease (CHD). Clinical populations in which an obesity paradox has been described include patients with chronic heart failure,⁵ CHD,^6,7 hypertension,⁸ peripheral arterial disease,⁹ type 2 diabetes,¹⁰ chronic kidney disease,¹¹ and the current exercise testing laboratory database from the Veterans Administration.¹²

The obesity paradox seems relatively easy to refute or explain given that weight loss and physical frailty are often a final common pathway to mortality in patients with chronic heart failure, chronic lung disease, chronic kidney disease, AIDS, and progressive neurologic disease, as well as aging. In contrast, weight loss before death would not seem to explain why a normal-weight person would be at a higher risk of death than an overweight or obese person, unless the normal-weight person had recently been overweight and was on a long-term spiral toward weight loss and death.

Validation of the obesity paradox requires that the inverse correlation between adiposity and mortality persist even when appropriate corrections for confounding variables are added to the analysis. Conversely, if the paradoxical association between body weight and mortality can be dissipated or eliminated by appropriate correction of confounding variables such as the presence of chronic disease, then there is little or no paradox.

The current study by McAuley et al¹² adds an interesting insight into this field of study by analyzing the power of cardiorespiratory fitness as a covariate of its effect on the obesity-mortality association. A compelling rationale as to the importance of fitness vis-à-vis the obesity paradox is that many individuals within a given weight strata with a below average exercise capacity can be considered “unwell” compared with individuals with a higher exercise capacity. If indeed, after adjusting the analysis for fitness (or wellness), the paradoxical association between low body weight and mortality is blunted, this would support the concept that unmeasured confounding variables such as undefined chronic illness mediate the association between lower body weight and poor clinical outcomes. The confounding variable might be chronic lung disease, excessive alcohol intake, illicit drug use, AIDS, or other issues.

See also page 115

Study of the obesity paradox is difficult because of the following factors.

1. Essentially all studies have been retrospective analyses; therefore, each study has been limited by an inability to adjust for all confounding variables.

2. None of the databases used were specifically designed to study the obesity paradox as a primary goal; thus, researchers are limited to analyzing only the available data and covariates. As an example, in the study by McAuley et al, smoking was defined as smoking at baseline, yes or no. If a participant had stopped smoking 1 year previously because of a diagnosis of chronic obstructive lung disease, he or she was listed as a nonsmoker, and the presence of chronic lung disease, an obvious potential confounding variable in the association between weight and mortality, was missed.

3. Few studies have data on recent weight change, in particular, unintentional weight loss in the period preceding the collection of weight data.

4. Few studies have data on excessive alcohol use, illicit drug use, or AIDS as confounding variables, which in the McAuley et al study population may have been a relevant issue.

The study by McAuley et al has several strengths, in particular the collection of cardiorespiratory fitness data that can be used as a proxy for a general state of well-being. Other strengths include its large sample size (N=12,417), an excellent follow-up system to determine mortality, and a computerized medical record to determine presence or absence of cardiac risk factors, such as hypertension, hyperlipidemia, or diabetes mellitus. Additionally, the authors had a computerized list of medications and current smoking habits, and participants who died within the first 2 years of follow-up were censored so that patients who might have had undiagnosed severe chronic disease at the beginning of the study were not included in the analysis.

Limitations of the study include that the population was defined by a need for a clinically indicated stress test and that BMI, rather than a measure of adiposity such as waist size, was used to estimate adiposity. Other issues include the following: mortality data were not delineated into cardiac vs noncardiac causes, and information was lacking on recent weight change, past smoking habits, and the presence of chronic disease (other than cardiac risk factors) and cardiovascular conditions. Risk factors were listed as present or absent rather than as continuous variables, such as lipid levels or blood pressure readings. An additional limitation was nicely described by the authors as the “veteran effect,” ie, at military induction the veterans were required to meet fitness and weight requirements such that, if obesity was present, it most likely developed late in life, mitigating lifelong effects of excessive adiposity on rates of diabetes, hypertension, hyperlipidemia, and the consequent cardiovascular disease. Moreover, in a population characterized by relatively high levels of fitness at baseline, an increased BMI may have been more likely to be related to increased muscularity, rather than to increased adiposity, compared with nonmilitary populations. Furthermore, there may have been a “selective survivor” effect, in which highest-risk obese patients with multiple obesity-related cardiac risk factors may have already died due to cardiovascular disease and thus were not included in the analysis. The obese/overweight patients who remained may have been more resistant to deleterious effects of obesity and thus confounded the results. Finally, an exercise testing—related issue is that cardiorespiratory fitness (quantified in metabolic equivalents and expressed in liters per minute of oxygen consumption) may have been underestimated in more overweight individuals (and overestimated in underweight patients) because metabolic equivalents were estimated from speed and elevation of the treadmill with no attention to the weight of the patient. Heavier patients were actually performing more work than lighter patients at a given workload, yet they would have been listed as having a similar cardiorespiratory fitness.

A recent analysis of data from the Framingham Heart Study provides useful insight into the obesity paradox in clinical populations.⁴ Investigators found that obesity was independently associated with an increased long-term risk of cardiovascular mortality, whereas being underweight was associated with an increased risk of noncardiovascular mortality, with “mixing” in the middle ranges of BMI. Thus, a “U-” shaped curve was traced with increasing mortality on the Y axis and increasing BMI on the X axis. When cardiovascular mortality was the outcome, the curve was J shaped; when noncardiovascular mortality was the outcome, the curve was L (or reverse J) shaped (Figure).

FIGURE.

Conceptual framework of relationship between mortality (cardiovascular or noncardiovascular) and body mass index.

Cross-sectional, noninterventional data relating to the obesity paradox should certainly not be interpreted to state that weight gain should be encouraged in normal-weight or overweight patients with cardiovascular disease or at high risk of cardiovascular disease. This is supported in the analysis by McAuley et al by higher rates of diabetes and hypertension with increasing levels of obesity (albeit with higher rates of smoking in lower BMI strata). Indeed, weight reduction results in the prevention of diabetes in individuals with insulin resistance¹³ and in a decrease of cardiovascular events in individuals with established CHD who are participating in cardiac rehabilitation.¹⁴ Weight reduction in patients with established CHD also is associated with improvements in a host of cardiac risk factors, including insulin sensitivity, blood pressure, hyperlipidemia, high-sensitivity C-reactive protein, plasminogen activator 1, platelet reactivity, and endothelial-related vasodilatory capacity.^15-18

So what do we now know? On the basis of the work by McAuley et al, we know that, when data supporting the obesity paradox are adjusted by cardiorespiratory fitness, a proxy for physical wellness, the paradoxical association between BMI and total mortality is blunted. This suggests that the paradoxical association is mediated, at least in part, by unmeasured confounding variables that link the presence of chronic disease with a poorer outcome. Indeed, if all potential confounding variables (eg, chronic lung disease, AIDS, alcohol use, illicit drug use) had been measured prospectively, the obesity paradox may have further dissipated or disappeared, but that cannot be determined. Nonetheless, residual uncertainty should not cause us to overlook the clear risk-lowering effects of weight reduction in individuals with established CHD or in individuals at high risk of developing CHD.^13,15,16