Journal of the American College of Cardiology
Pre-Morbid Body Mass Index and Mortality After Incident Heart FailureThe ARIC Study
Author + information
- Received April 14, 2014
- Revision received September 5, 2014
- Accepted September 16, 2014
- Published online December 30, 2014.
Author Information
- Umair Khalid, MD∗,
- Sameer Ather, MD, PhD†,
- Chirag Bavishi, MD, MPH‡,
- Wenyaw Chan, PhD§,
- Laura R. Loehr, MD, PhD‖,
- Lisa M. Wruck, PhD¶,
- Wayne D. Rosamond, PhD‖,
- Patricia P. Chang, MD, MHS#,
- Joe Coresh, MD∗∗,
- Salim S. Virani, MD, PhD∗,††,
- Vijay Nambi, MD††,
- Biykem Bozkurt, MD, PhD∗,††,
- Christie M. Ballantyne, MD∗ and
- Anita Deswal, MD, MPH∗,††∗ (adeswal{at}bcm.edu)
- ∗Department of Medicine, Baylor College of Medicine, Houston, Texas
- †Section of Cardiology, University of Alabama, Birmingham, Alabama
- ‡Department of Medicine, St. Luke's–Roosevelt Hospital Center, New York, New York
- §School of Public Health, University of Texas, Houston, Texas
- ‖Department of Epidemiology, University of North Carolina, Chapel Hill, North Carolina
- ¶Department of Biostatistics, University of North Carolina, Chapel Hill, North Carolina
- #Section of Cardiology, University of North Carolina, Chapel Hill, North Carolina
- ∗∗Johns Hopkins University, Baltimore, Maryland
- ††Section of Cardiology, Michael E. DeBakey Veterans Affairs Medical Center, Houston, Texas
- ↵∗Reprint requests and correspondence:
Dr. Anita Deswal, Michael E. DeBakey VA Medical Center (111B), 2002 Holcombe Boulevard, Houston, Texas 77030.
Abstract
Background Although obesity is an independent risk factor for heart failure (HF), once HF is established, obesity is associated with lower mortality. It is unclear if the weight loss due to advanced HF leads to this paradoxical finding.
Objectives This study sought to evaluate the prognostic impact of pre-morbid obesity in patients with HF.
Methods In the ARIC (Atherosclerosis Risk In Communities) study, we used body mass index (BMI) measured ≥6 months before incident HF (pre-morbid BMI) to evaluate the association of overweight (BMI 25 to <30 kg/m2) and obesity (BMI ≥30 kg/m2) compared with normal BMI (18.5 to <25 kg/m2) with mortality after incident HF.
Results Among 1,487 patients with incident HF, 35% were overweight and 47% were obese by pre-morbid BMI measured 4.3 ± 3.1 years before HF diagnosis. Over 10-year follow-up after incident HF, 43% of patients died. After adjustment for demographics and comorbidities, being pre-morbidly overweight (hazard ratio [HR]: 0.72; 95% confidence interval [CI]: 0.58 to 0.90; p = 0.004) or obese (HR: 0.70; 95% CI: 0.56 to 0.87; p = 0.001) had a protective association with survival compared with normal BMI. The protective effect of overweight and obesity was consistent across subgroups on the basis of a history of cancer, smoking, and diabetes.
Conclusions Our results, for the first time, demonstrate that patients who were overweight or obese before HF development have lower mortality after HF diagnosis compared with normal BMI patients. Thus, weight loss due to advanced HF may not completely explain the protective effect of higher BMI in HF patients.
Obesity has reached epidemic proportions in the United States, with more than two-thirds of adults being either overweight or obese (1). Obesity is linked to the development of cardiovascular diseases including atherosclerosis and hypertension (2). Although independently associated with the development of heart failure (HF) (3,4), obesity also has been shown to be associated with better survival once HF is established (5,6–9); this is often referred to as the “obesity paradox” (10).
One plausible explanation for this paradox: HF patients who gain or preserve their weight may represent a noncatabolic subgroup of HF patients with different neurohormonal, inflammatory, and metabolic profiles compared with HF patients who lose weight. The known protective effects of the ability to maintain or gain weight in other chronic diseases or catabolic states such as the acquired immunodeficiency syndrome, renal disease, and cancer lend support to this concept (11). Therefore, spontaneous weight loss (cachexia in extreme cases) after the development of HF may characterize a sicker group of patients with HF and, thus, may be associated with greater mortality (12).
However, it is not clear if weight loss after development of HF is the sole contributor to the obesity paradox or whether additional mechanisms, such as pre-existing obesity with possibly greater metabolic reserve prior to HF onset, contribute to the better survival of HF patients. Therefore, we examined the relationship of pre-morbid obesity (i.e., prior to incident HF) with mortality following incident HF. To our knowledge, this is the first study to investigate the prognostic implications of pre-morbid obesity with outcomes after the development of HF.
Methods
Study cohort
The ARIC (Atherosclerosis Risk In Communities) study is an ongoing, community-based cohort study of 15,792 patients, comprised mostly of Caucasian and African-American men and women, age 45 to 64 years at baseline (1987 to 1989), and sampled from 4 U.S. communities: Forsyth County, North Carolina; suburbs of Minneapolis, Minnesota; Washington County, Maryland; and Jackson, Mississippi (13). The institutional review boards from each site approved the ARIC study, and all participants provided written informed consent. Standardized physical examinations and interviewer-administered questionnaires were conducted at baseline (visit 1) and at approximately 3-year follow-up intervals (visit 4: 1996 to 1998). Participant follow-up through annual telephone interviews, hospitalization, and vital status is ongoing. Patients with missing anthropometry (n = 33), with prevalent HF at the first study visit (n = 751), and with missing data to determine prevalent HF at baseline (n = 289) were excluded (14). Participants with race not classified as white or black (n = 48) and blacks not from Jackson or Forsyth County (n = 120) were excluded due to their limited numbers.
Ascertainment of HF cases and follow-up
To determine HF cases, the following methods were used: 1) annual interviews of participants regarding interim hospitalizations (response rate: 93% to 96%); 2) review of discharge lists from local hospitals; and 3) survey of health department death certificate files and the national death index. Incident HF was defined as the first episode of either a hospitalization that included an International Classification of Diseases-9th Revision (ICD-9)-Clinical Modification discharge diagnosis code for HF beginning with “428” (i.e., 428.0 to 428.9) in any position, a death certificate ICD-9 code beginning with “428,” or ICD-10 code “I50” (HF or I50.0 to I50.9) in any position. For this study, incident HF was determined until December 31, 2004; date of last contact; or death (14).
Anthropometry
Participants presented for each study visit after an overnight fast, and measurements were taken in standard scrub attire. Weight was measured using a scale that was zeroed daily and calibrated quarterly. Pre-morbid body mass index (BMI) was defined as a BMI measurement from a study visit that occurred 6 months or more prior to the incident date of HF. Patients with HF were categorized by the pre-morbid BMI into normal (18.5 to <25 kg/m2), overweight (25 to <30 kg/m2), and obese (≥30 kg/m2) groups (15). Patients in the underweight category (BMI <18.5 kg/m2; n = 32) were excluded from this analysis because of small numbers and possible other pre-existing comorbidities that may have led to a cachectic state.
Baseline covariates
Ascertainment of demographics and comorbidities at each study visit has been described in detail previously (13). Age was assessed at the time of incident HF, and sex, race, and education level were obtained from the baseline ARIC visit with interviewer-administered questionnaires. Comorbidities, including hypertension, history of myocardial infarction (MI), coronary heart disease (CHD), diabetes mellitus (DM), and stroke, were assessed as present if these conditions were documented at any of the pre-HF study visits. History of MI was defined as self-report of physician-diagnosed MI or electrocardiographic diagnosis of silent MI. CHD was defined as history of MI, coronary revascularization, or coronary artery bypass surgery. Hypertension was defined by either: 1) a systolic blood pressure (SBP) ≥140 mm Hg or diastolic blood pressure ≥90 mm Hg measured with random-zero mercury manometers; or 2) recent antihypertensive medication use. Presence of DM was defined as either self-reported physician-diagnosed diabetes, recent diabetes medication use, or a blood glucose ≥126 mg/dl fasting or ≥200 mg/dl nonfasting (13).
Alcohol use, SBP, serum creatinine, total serum cholesterol, and insurance status also were collected from the pre-HF ARIC study visit. Patients were defined as smokers if the participants reported a history of current smoking at the pre-HF ARIC study visit. A history of cancer reported at any study visit prior to or including the pre-HF visit was used to define a positive history of cancer. Data for any variables with missing values at the pre-HF visit utilized available data from prior study visits. Renal function was quantified by the estimated glomerular filtration rate using the Modification of Diet in Renal Disease equation (16).
Statistical analysis
Data are shown as mean ± SD and percents. Missing values were found to be <0.8%. Univariate differences among the 3 BMI groups were examined using the chi-square test for categorical variables and the analysis of variance test for continuous variables. To assess the prognostic significance of pre-morbid BMI in HF patients, the endpoint was time to all-cause mortality after incident HF. We performed Kaplan-Meier survival analysis and used the log-rank test to compare time to death after incident HF among the 3 BMI groups. Because the number of patients after 10-year follow-up from the incident HF episode was small, we censored the follow-up after incident HF at 10 years. Cox proportional hazard models were used to examine the relationship between the BMI group and survival. Due to violation of the proportionality assumption, a time-dependent term, the product of BMI group and log time was added to the model to represent the nonhomogeneity of the hazard (17). The adjusted model was fitted by adding age, sex, race, history of MI, hypertension, CHD, DM, stroke, cancer, alcohol use, smoking status, insurance, education level, SBP, and estimated glomerular filtration rate as covariates. Due to concerns related to smoking and history of cancer and their association with BMI status and mortality, as well as the fact that statistical adjustments may not be sufficient to control for duration, intensity, or timing of smoking exposure, we also conducted subgroup analyses by the presence or absence of a history of smoking and cancer. Similarly, due to prior studies demonstrating a possible differential effect of obesity and overweight on mortality in HF patients on the basis of diabetic status, we also performed subgroup analysis by the presence or absence of DM (5,7,18–22). All p values <0.05 were considered to be significant. All analyses were performed using SAS version 9.3 (SAS Institute Inc., Cary, North Carolina).
Results
The study cohort consisted of 1.487 HF patients categorized as normal, overweight, or obese on the basis of BMI obtained at least 6 months before incident HF. Overall, 54% of the HF patients were male and 66% were white; they averaged 67 years of age at the time of diagnosis. Pre-morbidly, the majority of patients were either overweight (35%) or obese (47%). The pre-morbid BMI was measured 4.3 ± 3.1 years before the HF diagnosis at the last ARIC study visit, which occurred at least 6 months before incident HF. Baseline characteristics of patients with incident HF by the 3 pre-morbid BMI groups are shown in Table 1. Compared with HF patients with normal BMI, obese HF patients were younger, were more often African American, were less likely to have health insurance, and attained a lower education level. Overweight and obese HF patients had higher prevalence of comorbidities such as DM and hypertension, as well as higher SBP. Smoking was associated with lower BMI. There was no significant difference in the time period of measurement of BMI before incident HF among the 3 BMI categories.
Baseline Characteristics in Patients With Incident HF
Over a follow-up period of 10 years, a total of 43% of all HF patients died. The Central Illustration shows the Kaplan-Meier survival curves by BMI group over 10-year follow-up after incident HF. The overweight and obese groups had better survival compared with the normal-weight group. Over time, the survival curve for the obese patients appeared to converge toward the other groups, and crossed over the overweight group at ∼7.5 years of follow-up. However, the curves for overweight patients appeared to remain parallel to the normal-weight group. Compared with normal BMI, overweight (hazard ratio [HR]: 0.77; 95% confidence interval [CI]: 0.62 to 0.96) and obesity (HR: 0.75; 95% CI: 0.61 to 0.91) were associated with improved survival. As expected from the Kaplan-Meier curves, the interaction between the BMI group and log time was not significant for the overweight group (p = 0.74), but was significant for the obese group (p = 0.02). The interaction terms were, therefore, included in the models. As shown in Table 2, even after adjusting for covariates, pre-morbid overweight and obesity remained independent predictors of improved survival compared with normal BMI.
Kaplan-Meier Survival Curves in HF Patients by Pre-HF BMI Categories
Survival after the development of heart failure (HF) differed significantly among body mass index (BMI) groups defined by pre-HF BMI. The overweight and obese patients had better survival compared with the normal-weight group.
Hazard Ratios for All-Cause Mortality After Incident HF
A sensitivity analysis was performed after excluding patients with a presentation of fatal incident HF. Again, the overweight (adjusted HR: 0.72; 95% CI: 0.57 to 0.90; interaction with log time p = 0.37) and obese (adjusted HR: 0.69; 95% CI: 0.56 to 0.86; interaction with log time p < 0.001) HF patients had significantly better survival compared with HF patients with normal BMI.
Subgroup analyses were conducted in HF patients on the basis of smoking, history of cancer, and the presence or absence of diabetes (Figure 1). The beneficial trends associated with pre-morbid overweight and obesity compared with normal weight, similar to those observed in the overall cohort, were noted in all subgroups, although not all differences reached statistical significance.
Adjusted Risk of Mortality Associated With BMI Category
The adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) for mortality risk for the obese and overweight groups compared with the normal-weight group (reference group; HR: 1.00) are shown on a logarithmic scale for the overall cohort and for the subgroups stratified by smoking, cancer, and diabetes. The results are consistent across all subgroups.
Discussion
This study demonstrates that the majority of patients who develop HF in the community are pre-morbidly overweight or obese. Additionally, for the first time, we have shown that patients who are overweight or obese before incident HF have better survival after they develop HF compared with patients with normal BMI. This association is independent of the patients’ demographic profile and comorbidities. Furthermore, this trend occurred irrespective of smoking status, history of cancer, or diabetes.
Our findings of an association between a higher pre-morbid BMI and improved survival following incident HF could suggest that obese patients have a higher metabolic reserve compared with normal-weight patients, providing them with a survival advantage when cardiac cachexia ensues after HF development. Several studies have demonstrated an obesity paradox in patients with chronic and acute decompensated HF (i.e., a survival advantage of higher BMI measured in patients with established HF) (5,7–9,18–22). It must be noted that none of the previous studies demonstrating the obesity paradox in HF used the BMI (or BMI-equivalent variable) prior to development of HF. Moreover, very few studies had a follow-up >5 years (23,24). Because prior studies used the BMI of patients with established HF, they were unable to determine the effect of weight loss between the time of development of HF and the BMI measurement as a marker of more advanced HF versus the possible survival advantage of pre-existing obesity or overweight. Our study goes a major step further than prior studies by demonstrating that higher pre-morbid BMI is independently associated with a long-term survival advantage over a long (10-year) follow-up period.
Several proposed mechanisms could contribute to this apparent obesity paradox, including the fact that HF is a catabolic state leading to cachexia, and obese and overweight patients may have better outcomes as they have higher metabolic reserves (12,25). Another hypothesis is that obesity alters the natural history of HF through neurohumoral pathways. Higher levels of serum lipoproteins may neutralize bacterial lipopolysaccharides and thus attenuate the detrimental cytokine response in HF (26–29). Adipose tissue may produce higher levels of soluble tumor necrosis factor (TNF) receptors that serve as a reservoir for harmful circulating TNF (30). Levels of circulating stem cells are also higher in obese patients (31). Furthermore, obese patients have decreased adiponectin levels and an attenuated renin-angiotensin system and catecholamine response, both of which are associated with improved HF survival (26,32).
Another possible explanation is that obese or overweight patients may present with and be diagnosed with HF at an earlier stage due to symptoms exacerbated by excess body weight, such as dyspnea and edema (i.e., the obesity paradox may represent a lead-time bias). Furthermore, obese patients have a higher prevalence of comorbidities such as hypertension and DM, as supported by our study, and may represent a higher-risk population for HF. Another possibility is that the higher prevalence of hypertension, as well as higher blood pressures, in the overweight and obese patients may allow for greater up-titration of disease-modifying HF therapies. It is interesting to note that the protective effect of obesity was greatest during the initial years (Central Illustration), with significant interaction between BMI group and time, suggesting perhaps that during the later years, the complications of obesity-associated comorbidities catch up and lead to a greater decline in survival. This finding would also be expected if a lead-time bias is contributing with an earlier presentation with HF due to obesity-enhanced symptoms. Unlike our analysis, most studies that have examined the obesity paradox in patients with established HF have had shorter follow-up periods, usually <5 years. In contrast, the protective effect of being overweight did not appear to decrease over time. Although a history of cancer and smoking are associated with lower BMI and higher mortality, perhaps confounding the analyses of BMI and survival, our subgroup analyses did suggest that the observed results were independent of smoking or cancer status.
Previous studies that have evaluated the association between obesity and cardiovascular outcomes have used various indexes of obesity, including BMI, waist circumference and waist-hip ratio, and percent body fat (22,33). Whereas waist-hip ratio and waist circumference are better predictors of central obesity, BMI reflects generalized obesity (34,35). On the basis of previous analyses conducted on the ARIC cohort (14), which had confirmed that obesity and overweight are independent risk factors for developing HF, the degree and pattern of relationships for the development of HF were comparable for all 3 indexes of obesity. Also, because most previous studies that have evaluated the association between obesity and cardiovascular outcomes have used BMI, we used BMI as the index of obesity in our study.
The recent HF guidelines from the American College of Cardiology and American Heart Association do not specifically recommend weight reduction in obese patients with HF on the basis of the lack of data demonstrating a beneficial effect in this population (36). Although our study suggests that patients who are overweight/obese before the development of HF have better survival compared with patients of normal weight, it does not answer whether targeted weight reduction in obese patients with HF is beneficial or not. Only a randomized controlled trial of targeted weight reduction in obese patients with HF could help resolve that question.
Study limitations
Our study has inherent limitations associated with an observational cohort study, including those of possible residual confounding from unmeasured covariates. In addition, identification of the cases relied on ICD-9 codes; only hospitalized HF and incident fatal HF were included because we lacked consistent data on outpatient HF. However, validation of HF hospitalizations in an ARIC community surveillance study in 2005 has shown that the sensitivity and positive predictive value of ICD-9 code 428.x in any position for HF classified by subsequent medical record review by ARIC criteria were 0.95 and 0.77, respectively, for combined acute decompensated HF and chronic HF (in comparison to 0.83 and 0.78, respectively, by Framingham criteria) (37). Also, the fact that there was a long time period between measurement of BMI and incident HF (mean 4.2 years) makes it unlikely that our cohort included HF cases in whom weight loss as a result of HF would have occurred.
Furthermore, community surveillance reports have indicated that 74% of outpatient HF cases are hospitalized within 1.7 years (38). Because a diagnosis of HF in obese patients may be less specific than that in normal-weight patients, there is a possibility of differential misclassification bias. Also, we were unable to adjust for level of fitness, which has been shown to modify the association of BMI with prognosis in HF (39,40). Fatal initial HF episodes were determined from death certificates, which may overestimate or underestimate the true number of cases. We did not have a record of medical therapies instituted following incident HF and were unable to adjust for potential differences in therapy by BMI status. Additionally, the type of HF (HF with preserved or reduced ejection fraction) was not known.
Conclusions
The majority of patients with incident HF in the community have pre-existing overweight or obesity. Once the overweight/obese patients develop HF, they have lower mortality compared with HF patients with prior normal BMI. These results suggest that a significant component of the obesity paradox is driven by pre-morbid obesity, and it is, therefore, unlikely that cardiac cachexia due to advanced HF is the only mechanism contributing to the observed obesity paradox in established HF. Future studies are needed to confirm our observations in other cohorts.
COMPETENCY IN MEDICAL KNOWLEDGE: Although higher BMI is associated with an increased risk of developing clinical HF, patients with HF who are overweight or obese have better survival rates than those with normal weight. This obesity paradox also applies to patients who were overweight or obese before incident HF hospitalization, suggesting that the paradox is not entirely accounted for by weight loss or cardiac cachexia due to HF.
TRANSLATIONAL OUTLOOK: Further research is needed to understand the mechanisms responsible for the protective effect of pre-morbid overweight and obesity and the effect of intentional weight loss on clinical outcomes in patients with HF.
Acknowledgments
The authors thank the staff and participants of the ARIC study for their important contributions.
Footnotes
The ARIC study is carried out as a collaborative study supported by National Heart, Lung, and Blood Institute contracts (HHSN268201100005C, HHSN268201100006C, HHSN268201100007C, HHSN268201100008C, HHSN268201100009C, HHSN268201100010C, HHSN268201100011C, and HHSN268201100012C). Dr. Nambi has served on the regional advisory board of Sanofi; and was a coinvestigator on a patent filed along with Roche and Baylor College of Medicine on use of biomarkers in prediction of heart failure. Dr. Ballantyne has received grant/research support (all paid to institution, not individual) from Abbott Diagnostic, Amarin, Amgen, Eli Lilly, Esperion, GlaxoSmithKline, Genentech, Merck & Co., Novartis, Pfizer, Inc., Regeneron, Roche, Roche Diagnostic, and Sanofi-Synthelabo; and has served as a consultant to Abbott Diagnostics, Aegerion, Amarin, Amgen, Arena, Cerenis, Esperion, Genentech, Genzyme, Kowa, Merck, Novartis, Pfizer, Inc., Resverlogix, Regeneron, Roche, and Sanofi-Synthelabo. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
Listen to this manuscript's audio summary by JACC Editor-in-Chief Dr. Valentin Fuster.
You can also listen to this issue's audio summary by JACC Editor-in-Chief Dr. Valentin Fuster.
- Abbreviations and Acronyms
- BMI
- body mass index
- CHD
- coronary heart disease
- DM
- diabetes mellitus
- HF
- heart failure
- MI
- myocardial infarction
- Received April 14, 2014.
- Revision received September 5, 2014.
- Accepted September 16, 2014.
- American College of Cardiology Foundation
References
- ↵
- ↵
- ↵
- Kenchaiah S.,
- Sesso H.D.,
- Gaziano J.M.
- ↵
- Horwich T.B.,
- Fonarow G.C.,
- Hamilton M.A.,
- et al.
- Kenchaiah S.,
- Pocock S.J.,
- Wang D.,
- et al.
- Lavie C.J.,
- Alpert M.A.,
- Arena R.,
- et al.
- Shah R.,
- Gayat E.,
- Januzzi J.L. Jr..,
- et al.
- ↵
- ↵
- ↵
- ↵
- The ARIC Investigators
- ↵
- Loehr L.R.,
- Rosamond W.D.,
- Poole C.,
- et al.
- ↵
- ↵
- ↵
- Ata N.,
- Sozer M.T.
- ↵
- ↵
- ↵
- Kistorp C.,
- Faber J.,
- Galatius S.,
- et al.
- Rauchhaus M.,
- Clark A.L.,
- Doehner W.,
- et al.
- ↵
- ↵
- Weber M.A.,
- Neutel J.M.,
- Smith D.H.
- ↵
- ↵
- Yancy C.W.,
- Jessup M.,
- Bozkurt B.,
- et al.
- ↵
- Rosamond W.D.,
- Chang P.P.,
- Baggett C.,
- et al.
- ↵
- ↵
- Lavie C.J.,
- McAuley P.A.,
- Church T.S.,
- et al.
Podcast