Author + information
- Jean G. Dumesnil, MD∗ ( and )
- Philippe Pibarot, DVM, PhD
- ↵∗Reprint requests and correspondence:
Dr. Jean G. Dumesnil, Québec Heart and Lung Institute, 2725 Chemin Sainte-Foy, Québec, QC G1V-4G5, Canada.
In this issue of the Journal, Rogge et al. (1) publish a substudy of the SEAS (Simvastatin Ezetimibe in Aortic Stenosis) trial that examined whether overweight and obesity might impact outcome in patients with aortic valve stenosis (AS). For this purpose, 1,664 patients enrolled in the trial were grouped according to their baseline body mass index (BMI), and cardiovascular outcomes were recorded during an average follow-up of 4.3 years. As expected and previously reported in other studies (2), overweight and obese patients had higher prevalence of hypertension, more abnormal left ventricular (LV) geometry, and lower stress-corrected midwall shortening compared with normal weight patients. Surprisingly, however, in univariable Cox regression analysis and an initial multivariable Cox regression model adjusting for sex, smoking, hypertension, LV geometry, LV ejection fraction, mean aortic gradient, and randomized study drug treatment, the overweight group of patients (i.e., BMI 25.0 to 29.9 kg/m²) were found to have lower rates of ischemic cardiovascular (CV) events and AS-related events compared with the normal weight group (i.e., BMI 18.5 to 24.9 kg/m²), whereas obesity (i.e., BMI >30.0 kg/m²) did not significantly influence the rates of these events. Moreover, in this first series of multivariable models, the higher BMI classes were not associated with statistically significant increases in CV or total mortality or combined hospital stay for heart failure and death from any cause. In contrast, only when age was added to the covariates in a second series of Cox regression models did the association between overweight and lower rates of AS-related and ischemic CV events become statistically nonsignificant. Furthermore, in this second series of fully adjusted models, overweight and obesity were associated with a 46% to 67% higher risk of death from any cause and a 42% to 69% higher risk of combined hospital stay for heart failure and death from any cause.
Several recent studies in different types of populations (3–5) have indeed reported that, despite a higher prevalence of well-recognized risk factors such as hypertension and dyslipidemia, overweight and/or obese patients might actually have outcomes better than or equal to normal weight patients. Such findings have been termed the “obesity paradox” and have generated much discussion with regard to interpretation and therapeutic implications. The findings of Rogge et al. (1) are very instructive within this context, because they illustrate that, depending on the type of statistical adjustment that is used, completely different interpretations can be derived from the same dataset, and they also provide us with the opportunity to further reflect on this intriguing and unresolved paradox.
BMI: Time to Take Another Look?
In the National Institutes of Health–Centers for Disease Control guidelines (6), the recommendations of weight loss to lower elevated blood pressure or to lower elevated levels of total cholesterol, low-density lipoprotein cholesterol, and triglycerides and to raise low levels of high-density lipoprotein cholesterol in overweight and obese persons are given a category A level of evidence. By contrast, the use of the BMI to assess overweight and obesity is only given a category C level of evidence, and to our knowledge, these recommendations and their levels of evidence have not been revised since 1998. Before that time, the National Institutes of Health consensus since 1985 had been to define overweight-obesity as a BMI of 27.8 or greater for men and 27.3 or greater for women. Very recently, Flegal et al. (7) performed a systematic review and meta-analysis of associations between standard National Heart, Lung, and Blood Institute BMI categories for overweight and obesity and all-cause mortality in a sample of more than 2.88 million individuals from the general population with more than 270,000 deaths. The results show that mortality was significantly lower among those who were overweight (hazard ratio: 0.94, 95% confidence interval: 0.91 to 0.96), compared with normal weight individuals, and they did not observe higher mortality in individuals with grade I obesity. This study thus seems to further confirm the existence of an obesity paradox to the extent that the hitherto arbitrarily defined limits of normalcy might have to be re-examined. For instance, and as appropriately emphasized in the accompanying editorial by Heymsfield and Cefalu (8), the lowest all-cause mortality in many studies is observed between a BMI of 22 and 25 kg/m2, whereas persons with a BMI between 18.5 and 22 kg/m2 have higher mortality. Hence, the higher mortality observed in the “normal” group than in the overweight group could be due to the average resulting from combining persons in the lowest mortality category (i.e., BMI of 22 to 25 kg/m2) with those who have greater mortality (i.e., BMI of 18.5 to 22 kg/m2). Finally, the epidemic increase in type 2 diabetes that has accompanied the overall increase in BMI in the general population is often invoked as an argument in favor of the guidelines, but it can also be mentioned that general life expectancy has also increased in record proportions during the same period and that other factors such as the age at which type 2 diabetes will develop might enter into consideration.
The use of BMI as a predictor of mortality is based on the postulate that, because it predisposes to risk factors such as hypertension, dyslipidemia, and type 2 diabetes known to have a direct relation with mortality, it should thus also be a good predictor of mortality. However, there is no direct causal pathway between an increased BMI and these risk factors, and it is highly probable that the susceptibility of a particular individual to the ill effects of a given level of BMI is also influenced by other factors, such as genetic background, sex, the distribution of fat (visceral vs. sub-cutaneous), lifestyle, and the like. As emphasized by Rogge et al. (1), waist circumference was unfortunately not measured in the SEAS trial such that an unknown proportion of the cohort likely had a more benign form of weight surplus and that the findings with regard to the metabolic syndrome might thus have limited value. The BMI is a measure of total adiposity, but it does not discriminate between visceral fat that likely has a harmful effect versus subcutaneous fat that might be protective. Further studies are thus needed to elucidate the respective impact of visceral versus subcutaneous adiposity on outcomes in patients with AS.
Analyzing the Data: Pathophysiology vs. Statistics
The term paradox implies that the observations are counter to what would be expected on the basis of physiopathology and current concepts. Such an occurrence should, however, prompt further analysis and reflection rather than accepting the findings at face value. Hence, the first statistical model used by Rogge et al. (1) suggested that being overweight had a protective effect with regard to adverse outcomes and that obesity had no effect, even though both overweight and obesity had higher prevalence of hypertension, high fasting glucose, and LV hypertrophy, which are anomalies usually associated with poorer prognosis. However, in view of this result, several observations must be made. First, at the outset, exclusion criteria for the SEAS trial included any indication for statin use, thus excluding patients with dyslipidemia, diabetes, or known ischemic heart disease; and the overweight and/or obese patients in this cohort can thus be considered to be low-risk compared with other cohorts. Second, a very high proportion of the overweight (85%) and obese (95%) had hypertension, which was also included in the multivariable model; the analysis with regard to BMI is thus possibly over-adjusted, because hypertension can be considered to be in the causal pathway between obesity and mortality (7) and because it was present in such a high proportion of the overweight and obese patients. Third, in contrast, age was not included in the initial multivariable analysis, although the overweight patients were significantly younger than the normal weight patients; because age is not considered to be in the causal pathway between overweight/obesity and adverse outcomes, the first model can thus also be seen as having possibly been under-adjusted (7). Finally, the ischemic CV and AS-related events in this study were largely driven by the occurrence of aortic valve replacement and/or coronary artery bypass graft surgery; these interventions are not naturally occurring events but rather decided subjectively by the treating physician on the basis of criteria that might differ from one patient to another. It cannot be assumed, for instance, that physicians are equally aggressive in treating normal weight versus overweight or obese patients.
Fortunately, however, Rogge et al. (1) had the foresight to use a second multiple Cox regression model that included age as a variable. In so-doing, the interpretation of the results and the conclusions that had been drawn from the first model were completely modified to the extent that the so-called obesity paradox was no longer present and that, overall, the findings of the study were more consistent with current pathophysiological concepts. Hence, potential clinical implications of the findings also seemed to be completely different. Because it still included hypertension, the second model might still have been over-adjusted, but the results with regard to BMI would likely not have changed significantly if it had been excluded. As with most randomized clinical trials, the sample of patients recruited in the SEAS trial is a highly selected subset of the general population with AS. In particular, patients with hypercholesteromia, diabetes, and coronary artery disease were under-represented or not represented in this trial, whereas patients with hypertension were likely over-represented. Hence, the results obtained from the SEAS trial might not be directly transposable to the general population, and given that a large proportion of patients with the “at-risk” obesity were excluded from this trial, it is probable that in such a population, the impact of BMI on outcomes is more important than what is reported in the present paper.
Rogge et al. (1) are to be commended for a fine study that demonstrates the extreme caution that should be exercised when facing results that seem to be paradoxical in light of currents concepts and physiopathology. Obviously, this topic needs further investigation to the extent of questioning currently used values for BMI and even the wisdom of using this parameter as an independent predictor of mortality. The statistical analysis of results should also be exhaustive and logical. Finally, it should be emphasized that the benefits of weight loss still have a category A level of evidence. By contrast, the apparent “obesity paradox” reported in many studies requires further investigation and should not be used at this time to justify more lenient lifestyle change recommendations.
↵∗ Editorials published in the Journal of the American College of Cardiology reflect the views of the authors and do not necessarily represent the views of JACC or the American College of Cardiology.
Dr. Pibarot has received a research grant from Edwards Lifesciences for Echo CoreLab Analyses. Dr. Dumesnil has reported that he has no relationships relevant to the contents of this paper to disclose.
- American College of Cardiology Foundation
- Rogge B.P.,
- Cramariuc D.,
- Lønnebakken M.T.,
- et al.
- Lund B.P.,
- Gohlke-Barwolf C.,
- Cramariuc D.,
- Rossebo A.B.,
- Rieck A.E.,
- Gerdts E.
- Hastie C.E.,
- Padmanabhan S.,
- Slack R.,
- et al.