Author + information
- Received March 14, 2012
- Revision received May 18, 2012
- Accepted May 28, 2012
- Published online October 9, 2012.
- Carl J. Lavie, MD⁎,†,⁎ (, )
- Alban De Schutter, MD, MSc‡,
- Dharmendrakumar A. Patel, MD, MPH⁎,
- Abel Romero-Corral, MD, MSc§,
- Surya M. Artham, MD, MPH⁎ and
- Richard V. Milani, MD⁎
- ↵⁎Reprint requests and correspondence:
Dr. Carl J. Lavie, John Ochsner Heart and Vascular Institute, Ochsner Clinical School, The University of Queensland School of Medicine, 1514 Jefferson Highway, New Orleans, Louisiana 70121-2483
Objectives Our goal was to determine the impact of lean mass index (LMI) and body fat (BF) on survival in patients with coronary heart disease (CHD).
Background An inverse relationship between obesity and prognosis has been demonstrated (the “obesity paradox”) in CHD, which has been explained by limitations in the use of body mass index in defining body composition.
Methods We studied 570 consecutive patients with CHD who were referred to cardiac rehabilitation, stratified as Low (≤25% in men and ≤35% in women) and High (>25% in men and >35% in women) BF and as Low (≤18.9 kg/m2 in men and ≤15.4 kg/m2 in women) and High LMI, and followed for 3 years for survival.
Results Mortality is inversely related to LMI (p < 0.0001). Mortality was highest in the Low BF/Low LMI group (15%), which was significantly higher than in the other 3 groups, and lowest in the High BF/High LMI group (2.2%), which was significantly lower than in the other 3 groups. In Cox regression analysis as categoric variables, low LMI (hazard ratio [HR]: 3.1; 95% confidence interval [CI]: 1.3 to 7.1) and low BF (HR: 2.6; 95% CI: 1.1 to 6.4) predicted higher mortality, and as continuous variables, high BF (HR: 0.91; 95% CI: 0.85 to 0.97) and high LMI (HR: 0.81; 95% CI: 0.65 to 1.00) predicted lower mortality.
Conclusions In patients with stable CHD, both LMI and BF predict mortality, with mortality particularly high in those with Low LMI/Low BF and lowest in those with High LMI/High BF. Determination of optimal body composition in primary and secondary CHD prevention is needed.
Although obesity is considered a risk factor for most cardiovascular (CV) diseases, including coronary heart disease (CHD) (1,2), many studies of cohorts with established CV diseases, including heart failure, hypertension, atrial fibrillation, and CHD, have demonstrated an inverse relationship between obesity, generally determined by body mass index (BMI), and subsequent prognosis, which has been termed the “obesity paradox” (1–4). Although BMI is the most frequently used method to assess overweightness and obesity, this method has been criticized because BMI does not always reflect true body fatness (1–4), which may be better evaluated by assessments of body fat (BF) and fat-free mass (FFM). However, limited data are available on the impact of FFM and BF on the prognosis in patients with established CHD. Therefore, we assessed the impact of both BF and lean mass index (LMI) on 3-year mortality in a cohort with stable CHD.
We reviewed 570 consecutive patients with stable CHD who were referred to cardiac rehabilitation between January 1, 2000, and July 31, 2005, by dividing patients into Low (≤25% in men and ≤35% in women) and High BF (>25% in men and >35% in women) groups, as previously described (2,5–7). Patients also were divided according to LMI into 3 groups on the basis of previously defined 25th to 75th percentiles for analysis of LMI (8): Low (<18.7 kg/m2 in men and <14.9 kg/m2 in women), Medium (18.7 to 21 kg/m2 in men and 14.9 to 17.2 kg/m2 in women), and High (>21 kg/m2 in men and >17.2 kg/m2 in women). Patients were also divided into Low (cutoff ≤18.9 kg/m2 in men and ≤15.4 kg/m2 in women) and High (>18.9 kg/m2 in men and >15.4 kg/m2 in women), as previously defined (9), to be combined with the BF subgroups. Thus, 4 groups were analyzed: Low BF/Low LMI (n = 62), High BF/Low LMI (n = 53), Low BF/High LMI (n = 179), and High BF/High LMI (n = 276).
Baseline laboratory parameters, including BF by the sum of the skin-fold method, was performed as described previously (2,5). LMI was determined by (1-BF) × BMI kg/m2, as previously described (8,9). We then divided actual body weight minus ideal body weight proportionally to %BF and used mean LMI for that gender to establish excess lean mass and excess fat mass. The prevalence of hypertension, current smoking, diabetes, and chronic obstructive pulmonary disease (COPD) was also assessed. Patients were followed for an average of 3 years (mean 1,266 ± 527 days, median 1,356 days, range 231 to 2,149 days) to determine all-cause mortality assessed by the National Death Index.
SAS version 9.0 (SAS Institute Inc., Cary, North Carolina) computer software was used for statistical analysis. Mean ± SD or proportions for baseline characteristics were reported, and the 3 LMI groups and the 4 BF/LMI groups were compared with analysis of variance and chi-square analysis. Kaplan-Meier survival curves were constructed to assess survival in the LMI subgroups and by both High/Low BF and High/Low LMI. Cox regression analysis was performed to predict mortality using age, gender, ejection fraction (EF), peak oxygen consumption (VO2), BF, and LMI, with BF and LMI introduced as categoric (low vs. high) and continuous variables. Because of the possibility of overfitting, we also performed the Cox regression without age and gender. This analysis also was performed with 27 patients classified as having COPD, as well as with COPD as a categoric variable in the multivariate analysis. Logistic regression analysis was performed, but the main results were reported using Cox regression because patients had variable follow-up durations. Pearson's correlation was used to correlate BF with LMI and traditional risk factors. Kappa statistic was used to determine the agreement between the LMI and BF groups.
The baseline characteristics of the study population are described in Table 1. During follow-up, 26 patients died; survivors had significantly higher values for BMI, BF, LMI, left ventricular EF, and peak VO2, and borderline higher levels of systolic blood pressure than those who died.
The baseline characteristics of the 3 distinct LMI groups are described in Table 2, with the groups demonstrating significant differences in BMI, % BF, age, high-density lipoprotein (HDL) cholesterol, triglycerides, and gender. The baseline characteristics of the 4 distinct body composition groups based on both BF and LMI are shown in Table 3, with the groups differing in BMI, age, peak VO2, HDL cholesterol, triglycerides, and smoking. All participants in the High BF/High LMI group were considered overweight (39.5%) or obese (60.5%) by BMI criteria, with an average BMI of 31.9 kg/m2 for this group. In the Low BF/High LMI group, the majority of participants were overweight (57.5%), followed by normal (31.3%) and obese (11.1%) by BMI criteria. The High BF/Low LMI group was approximately evenly divided between overweight participants (50.9%) and normal participants (43.4%), with a minority of obese participants (5.7%) by BMI criteria. LMI and BF were independent of each other (r = −0.07; p = 0.07). The BF and LMI groups correlated weakly (r = 0.19; p < 0.0001; kappa 0.10; 95% confidence interval [CI]: 0.03 to 0.17).
Influence of body composition on traditional CHD risk factors
Low HDL cholesterol was associated more with High LMI (odds ratio [OR]: 1.46; 95% CI: 1.09 to 1.97), but not BF (1.15 95% CI: 0.86 to 1.54), after adjusting for gender, age, lipid medication use, and impaired fasting glucose. High low-density lipoprotein cholesterol was associated with high BF (OR: 1.36; 95% CI: 1.02 to 1.82), but not high LMI (OR: 1.02; 95% CI: 0.74 to 1.43), after adjusting for age, gender, and lipid medication use. Both high LMI (OR: 1.39; 95% CI: 1.08 to 1.80) and high BF (OR: 1.46; 95% CI: 1.02 to 2.09) were associated with impaired fasting glucose after adjusting for age and gender. The triglyceride/HDL ratio correlated more, although relatively weakly, with the High LMI group (r = 0.16; p < 0.0001) than with the High BF group (r = 0.09; p = 0.05). Hypertension was associated more with high LMI (OR: 1.31; 95% CI: 1.0 to 1.69) than with high BF (OR: 1.18; 95% CI: 0.82 to 1.69) after adjusting for age and gender. Triglycerides correlated weakly with BF (r = 0.09; p = 0.04) and high LMI (r = 0.16; p < 0.0001). High C-reactive protein was not significantly associated with BF (OR: 1.36; 95% CI: 0.83 to 2.20) or LMI (OR: 1.41; 95% CI: 0.99 to 2.01) after adjusting for age, gender, and lipid medication use.
In Figure 1, mortality was inversely related with LMI, which was highest in the Low LMI group (10.3% [10 of 97]; p < 0.0001 vs. the High LMI; p = 0.003 vs. the Medium LMI) and lowest in the High LMI group (2.7% [7 of 257]). Intermediate mortality was noted in the Medium LMI group (4.2% [9 of 216]; p = 0.2 compared with the High LMI group).
As shown in Figure 2, mortality was highest in the Low BF/Low LMI group (15% [9 of 62]), which was significantly higher than in the other 3 groups (4.5% [8 of 179] for Low BF/High LMI, p = 0.001; 5.7% [3 of 53] for High BF/Low LMI, p = 0.0025; and 2.2% [6 of 270] for High BF/High LMI, p < 0.0001). The High BF/High LMI group had significantly lower mortality than all of the groups (p = 0.003 vs. High BF/Low LMI; p = 0.03 vs. Low BF/High LMI).
Multivariate predictors of mortality
In multivariate analysis using Cox regression (Table 4), low LMI (hazard ratio [HR]: 3.1; 95% CI: 1.3 to 7.1) and low BF (HR: 2.6; 95% CI: 1.1 to 6.4) as categoric variables predicted higher mortality, and as continuous variables, high BF (HR: 0.91; 95% CI: 0.85 to 0.97) and high LMI (HR: 0.81; 95% CI: 0.65 to 1.00) predicted lower mortality. If only peak VO2 and EF were entered into the Cox regression with BF and LMI, both LMI and BF as categoric variables remained significant predictors of mortality, as did BF as a continuous variable, but LMI as a continuous variable was not a significant predictor (HR: 0.86; 95% CI: 0.73 to 1.02). Other independent predictors of 3-year mortality included peak VO2 and EF. Although higher age was associated with a trend for worse survival, this was not statistically significant. Gender also was not significantly associated with survival. By using logistic regression, high BF (OR: 0.90; 95% CI: 0.84 to 0.97) and high LMI (OR: 0.75; 95% CI: 0.58 to 0.97) as continuous variables were independent predictors of lower mortality, and as categoric variables, low BF (OR: 3.3; CI: 1.2 to 9.0) and low LMI (OR: 3.7; 95% CI: 1.4 to 9.8) were independent predictors of higher mortality. In addition, surplus lean mass (OR: 0.90; 95% CI: 0.84 to 0.97) and fat mass (OR: 0.92; 95% CI: 0.86 to 0.99) both independently predicted lower mortality. There was no significant interaction between BF and LMI in any of the subgroups.
Influence of COPD
In our cohort, 567 patients coded yes or no for COPD; mortality was markedly higher (18.5%) in those coded yes for COPD versus only 3.9% in those coded no for COPD (p = 0.004) (Table 5). Compared with those without COPD, those with COPD had lower BMI, LMI, peak VO2, and fasting glucose; a lower prevalence of diabetes; higher age, total cholesterol, HDL-cholesterol, and low-density lipoprotein cholesterol; and a higher prevalence of female subjects and smokers with a trend of lower BF (which was not statistically significant).
However, when COPD was entered into the Cox regression analysis, COPD was not a significant independent predictor of mortality (HR: 0.65; 95% CI: 0.18 to 2.32) and did not have a major effect on the impact of BF (HR: 0.91; 95% CI: 0.85 to 0.97) as an independent predictor of mortality, but it modestly weakened LMI (HR: 0.96; 95% CI: 0.85 to 1.08).
Influence of gender
When stratifying the population by gender, we found similar results. For women, when entered individually, BF and LMI were independently associated with lower mortality (BF HR: 0.84; 95% CI: 0.74 to 0.95 and LMI HR: 0.36; 95% CI: 0.15 to 0.86). However, when combined, BF (HR: 0.87; 95% CI: 0.76 to 0.99), but not LMI (HR: 0.49; 95% CI: 0.21 to 1.14), was associated with lower mortality. For men, BF and LMI were associated with a trend toward lower mortality, both when entered individually (BF HR: 0.94; CI: 0.86 to 1.02 and LMI HR: 0.85; 95% CI: 0.70 to 1.04) and together (BF HR: 0.94; 95% CI: 0.87 to 1.02 and LMI HR: 0.85; 95% CI: 0.68 to 1.06).
We demonstrated that mortality seems to be inversely related with LMI in patients with stable CHD. Mortality is highest in patients with Low BF/Low LMI, lowest in those with High BF/High LMI, and intermediate in those with High BF/Low LMI or Low BF/High LMI. Both Low LMI and Low BF as categoric variables were independent predictors of an approximately 3-fold higher mortality.
Obesity paradox in CHD
Despite the powerful association of overweightness and obesity with CHD risk factors and CHD, numerous studies have reported that in those with established CHD, patients who are overweight and obese have a better clinical prognosis than their lean counterparts with similar CHD, a process that has been termed the “obesity paradox” (1–4), which has been demonstrated in many CHD cohorts (1–5). Although most studies on the obesity paradox used BMI, we have demonstrated that BF also is inversely associated with mortality in patients with CHD (2,5), a fact that is confirmed in the present study, with low BF being an independent predictor of an approximately 3-fold higher mortality in patients with CHD. Of note, these data showed the powerful association of high BF with better survival despite the fact that higher BF was associated with a worse overall CHD risk profile.
One limitation of previous studies assessing the obesity paradox is that FFM or LMI was not accounted for. In our study, we estimated LMI by determination of BF and defined LMI as the BMI that was not accounted for by fat. To our knowledge, this assessment has not been performed in the assessment of patients with CHD and the obesity paradox. We found that unlike BF, which was associated with an obesity paradox in that higher BF was associated with lower mortality, patients with CHD with higher LMI, as would be suggested from the epidemiologic studies, also had a better prognosis. In fact, the best prognosis was noted in those patients with High BF/High LMI, and the highest mortality was noted in those with Low BF/Low LMI. In the multivariate analysis, low LMI as a categoric variable was an independent predictor of a more than 3-fold higher mortality, whereas as a continuous variable, higher LMI was at least strongly associated with better survival. As discussed earlier for BF, a higher LMI also was associated with a worse overall CHD risk profile, so a more favorable impact of LMI with CHD risk factors does not explain the protective effect of higher LMI in a CHD cohort.
As in most other analyses, our study is not able to control for nonpurposeful weight loss before study entry. However, in patients who are referred to cardiac rehabilitation programs, other medical problems are generally stable. Others have suggested that the obesity paradox may be modified by overall physical wellness or by unmeasured confounding factors (10). High levels of fitness significantly alter the association of BMI and other parameters for obesity with subsequent mortality (11–13). Our data were adjusted for fitness, and still BF and LMI were independent predictors of mortality. In addition, only 6 of 570 patients were “underweight” (BMI <18.5 kg/m2), although these patients had 50% mortality; eliminating these patients did not significantly alter any of the major conclusions. Few of our patients were active smokers, and many studies of CHD cohorts have demonstrated that smoking cessation in CHD is associated with a favorable prognosis, nearly equal to never smokers within approximately 6 months of smoking cessation (14). In a large meta-analysis of 900,000 subjects, Whitlock et al. (15) observed an inverse relationship of BMI with mortality at a BMI of 22.5 kg/m2, attributed to respiratory disease, but this was not an independent predictor of mortality in our multivariate analysis. In addition, including COPD in our multivariable model did not change the influence of BF on mortality risk, although it weakened the relationship of LMI. Even in patients with peripheral arterial disease, in whom smoking and COPD are prevalent and strongly associated with this disease, COPD did not completely explain the obesity paradox (14,16). A recent study of more than 50,000 patients with ST-segment elevation myocardial infarction showed that the highest in-hospital mortality was in the “normal” BMI group, followed closely by the class 3 obese or severely obese patients. However, after correcting for baseline factors, mortality was increased only in the severely obese group, suggesting that confounding factors explained the higher mortality in the lower BMI groups (17,18). Nevertheless, this study assessed only in-hospital mortality, and so far confounders have not totally explained the obesity paradox (17–19).
It is interesting to speculate that although high BF may be a risk factor for CV diseases and CHD, by some mechanism this may be protective in cohorts with known disease. For example, before a CV event, positive caloric balance leading to adiposity may result in pathogenic adipose tissue responses that cause metabolic diseases, increasing CV risk. Paradoxically, during a time of negative caloric balance, as may occur during a CV event or major interventional procedure, adipose tissue may respond with enhanced function, which may improve CV and other clinical outcomes (5,20). Another interesting possibility is that higher BF and especially higher LMI may be associated with muscular strength, which is associated with better prognosis (21,22) and survival in several populations, even independently of aerobic fitness.
This is a relatively small, retrospective study of a select cohort, and the follow-up was relatively short. We assessed BF by the sum of the skin-fold method, which has potential limitations. We calculated LMI on the basis of the assessment of BF; the cut-points used for LMI have been validated (8,9), but not using the skin-fold assessment; also, because we estimated subcutaneous BF, visceral BF may be included in our LMI. Although we measured body composition, we did not measure BF and lean mass distribution; 2 recent studies (both including Mayo Clinic data) demonstrated that central obesity was associated with mortality in CHD (12,23), whereas a study in heart failure indicated that central obesity was associated with better survival (24,25). Although we corrected for COPD, our study did not use assessment of pulmonary function. In addition, we assessed total mortality (which may be the most important and reliable end point), but we did not assess other CHD morbidity. Finally, because our study was small and several variables were used in the Cox regression analysis, a potential limitation is that overfitted models may capitalize on the idiosyncrasies of the sample at hand. However, even when limiting the variables studied in the Cox regression analysis, both LMI and BF as categoric variables and BF as a continuous variable remained significant predictors of mortality, whereas LMI as a continuous variable became a borderline significant predictor.
Despite the recognized study limitations, our findings indicate that both high LMI and high BF are independent predictors of better survival in those with stable CHD, with mortality especially high in the Low BF/Low LMI group and lowest in the High BF/High LMI group. Prospective studies are needed to validate our findings and to determine the optimal body composition in both primary and secondary CHD prevention.
All authors have reported they have no relationships relevant to the contents of this paper to disclose.
Presented in part as 2 abstract presentations to the American College of Cardiology Annual Scientific Sessions, April 2 to 5, 2011, New Orleans, Louisiana.
- Abbreviations and Acronyms
- body fat
- body mass index
- coronary heart disease
- confidence interval
- chronic obstructive pulmonary disease
- ejection fraction
- fat-free mass
- high-density lipoprotein
- hazard ratio
- lean mass index
- odds ratio
- oxygen consumption
- Received March 14, 2012.
- Revision received May 18, 2012.
- Accepted May 28, 2012.
- American College of Cardiology Foundation
- Lavie C.J.,
- Milani R.V.,
- Ventura H.O.
- AACE/ACE Obesity Task Force
- Das S.R.,
- Alexander K.P.,
- Chen A.Y.,
- et al.
- Lavie C.J.,
- Milani R.V.,
- Ventura H.O.
- Bays H.E.
- Artero E.G.,
- Lee D.C.,
- Ruiz J.R.,
- et al.
- Artero E.G.,
- Lee D.C.,
- Lavie C.J.,
- et al.
- Coutinho T.,
- Goel K.,
- Corrêa de Sá D.,
- et al.