Effect of Caloric Restriction or Aerobic Exercise Training on Peak Oxygen Consumption and Quality of Life in Obese Older Patients With Heart Failure With Preserved Ejection Fraction: A Randomized Clinical Trial

Importance  More than 80% of patients with heart failure with preserved ejection fraction (HFPEF), the most common form of heart failure among older persons, are overweight or obese. Exercise intolerance is the primary symptom of chronic HFPEF and a major determinant of reduced quality of life (QOL).

Objective  To determine whether caloric restriction (diet) or aerobic exercise training (exercise) improves exercise capacity and QOL in obese older patients with HFPEF.

Design, Setting, and Participants  Randomized, attention-controlled, 2 × 2 factorial trial conducted from February 2009 through November 2014 in an urban academic medical center. Of 577 initially screened participants, 100 older obese participants (mean [SD]: age, 67 years [5]; body mass index, 39.3 [5.6]) with chronic, stable HFPEF were enrolled (366 excluded by inclusion and exclusion criteria, 31 for other reasons, and 80 declined participation).

Interventions  Twenty weeks of diet, exercise, or both; attention control consisted of telephone calls every 2 weeks.

Main Outcomes and Measures  Exercise capacity measured as peak oxygen consumption (V̇o₂, mL/kg/min; co–primary outcome) and QOL measured by the Minnesota Living with Heart Failure (MLHF) Questionnaire (score range: 0–105, higher scores indicate worse heart failure–related QOL; co–primary outcome).

Results  Of the 100 enrolled participants, 26 participants were randomized to exercise; 24 to diet; 25 to exercise + diet; 25 to control. Of these, 92 participants completed the trial. Exercise attendance was 84% (SD, 14%) and diet adherence was 99% (SD, 1%). By main effects analysis, peak V̇o₂ was increased significantly by both interventions: exercise, 1.2 mL/kg body mass/min (95% CI, 0.7 to 1.7), P < .001; diet, 1.3 mL/kg body mass/min (95% CI, 0.8 to 1.8), P < .001. The combination of exercise + diet was additive (complementary) for peak V̇o₂ (joint effect, 2.5 mL/kg/min). There was no statistically significant change in MLHF total score with exercise and with diet (main effect: exercise, −1 unit [95% CI, −8 to 5], P = .70; diet, −6 units [95% CI, −12 to 1], P = .08). The change in peak V̇o₂ was positively correlated with the change in percent lean body mass (r = 0.32; P = .003) and the change in thigh muscle:intermuscular fat ratio (r = 0.27; P = .02). There were no study-related serious adverse events. Body weight decreased by 7% (7 kg [SD, 1]) in the diet group, 3% (4 kg [SD, 1]) in the exercise group, 10% (11 kg [SD, 1] in the exercise + diet group, and 1% (1 kg [SD, 1]) in the control group.

Conclusions and Relevance  Among obese older patients with clinically stable HFPEF, caloric restriction or aerobic exercise training increased peak V̇o₂, and the effects may be additive. Neither intervention had a significant effect on quality of life as measured by the MLHF Questionnaire.

Trial Registration  clinicaltrials.gov Identifier: NCT00959660

Heart failure with preserved ejection fraction (HFPEF) is the most rapidly increasing form of heart failure, occurs primarily in older women, and is associated with high rates of morbidity, mortality, and health care expenditures.¹ However, its pathophysiology is poorly understood, and medication trials to date have been neutral.

Quiz Ref IDMost previous HFPEF trials focused on mediating the long-term consequences of hypertension. However, obesity is also an independent risk factor for development of heart failure,^2,3 and more than 80% of patients with HFPEF are overweight or obese.^4,5 Increased adiposity promotes inflammation, hypertension, insulin resistance, and dyslipidemia and impairs cardiac, arterial, skeletal muscle, and physical function;^6-8 all of which are common in HFPEF and contribute to its pathophysiology.⁹Quiz Ref ID It was recently shown that the severity of exercise intolerance, the primary symptom and major contributor to reduced quality of life (QOL) in patients with chronic HFPEF, is significantly correlated with increased body adiposity and skeletal muscle adipose infiltration.^6,10

Quiz Ref IDIn obese older individuals without heart failure, weight loss via dietary caloric restriction (diet) improves left ventricular hypertrophy and diastolic function; exercise capacity; glucose, lipid, and blood pressure control; inflammation markers; body composition; and skeletal muscle function.^8,11-13 However, diet is controversial in patients with heart failure; observational studies suggest overweight or mildly to moderately obese patients with heart failure (including HFPEF specifically) survive longer than those who are normal weight or underweight.⁵ There have been no studies of diet in any type of heart failure and current HFPEF management guidelines do not include diet.¹⁴

The objective of this study was to conduct a randomized, single-blind, attention-controlled trial to examine the effects of diet, alone and combined with aerobic exercise training (exercise), on exercise capacity measured as peak exercise oxygen consumption per unit time (V̇o₂, co–primary outcome) and QOL (co–primary outcome), and exploratory outcomes of body composition, leg muscle function, cardiac function, and inflammation in obese older patients with HFPEF.

The trial was conducted at Wake Forest School of Medicine from February 2009 through November 2014, approved by the institutional review board, and registered. Participants provided written informed consent.

Quiz Ref IDParticipants were identified from search lists of medical records.^15,16 Inclusion criteria for participants were 60 years or older; body mass index (BMI; calculated as weight in kilograms divided by height in meters squared) of 30 or higher; symptoms and signs of heart failure defined by the National Health and Nutrition Examination Survey Congestive Heart Failure criteria score of 3 or higher¹⁷ or the criteria of Rich et al,¹⁸ or both; and left ventricular ejection fraction of 50% or more. Major exclusion criteria were left ventricular segmental wall motion abnormalities and significant ischemic or valvular heart disease, pulmonary disease, anemia, or other disorder that could explain the participants’ heart failure symptoms. Participants were clinically stable, had no significant change in cardiac medications for 4 weeks, and were not undergoing regular exercise or diet.

Cardiopulmonary exercise testing was performed on a motorized treadmill using the modified Naughton protocol to the end point of exhaustion.¹⁹ Gas exchange was measured continuously during exercise (Ultima, Medical Graphics). Peak V̇o₂ (mL/kg body mass/min), the co–primary outcome, was the average of measures from the last 30 seconds during peak exercise.¹⁹

The other co–primary outcome was disease-specific QOL assessed as the total score from the Minnesota Living with Heart Failure (MLHF) Questionnaire.^16,17,20 The MLHF score range is 0 through 105, higher scores indicate worse heart failure–related QOL.

Exercise time, 6-minute walk distance, ventilatory anaerobic threshold, and ventilation/carbon dioxide output slope were assessed as previously described.^15,19

Total body lean mass and fat mass were measured by dual energy x-ray absorptiometry (DXA; Hologic) according to standardized protocols.¹⁰ Thigh muscle and fat areas and abdominal, epicardial, and pericardial fat areas were measured using magnetic resonance imaging (MRI), and an image analysis workstation (TomoVision), version 2.1.⁶

Leg press power (measured in W) was assessed using the Nottingham power rig. Muscle quality was calculated as leg power divided by thigh muscle area (W/cm²) from MRI.

Heart failure–specific QOL was assessed with the Kansas City Cardiomyopathy Questionnaire (KCCQ; range 0-100; higher scores indicate better QOL) and general QOL was assessed with the 36-item Short-Form Health Survey (SF-36) physical component score (range 0-100, the average is 50 and higher scores indicate better QOL).^15,16,21,22

Doppler echocardiograms were performed and analyzed per American Society of Echocardiography recommendations.¹⁶ Doppler left ventricular filling patterns and pulse-wave velocity were assessed as described.¹⁶

Left ventricular mass and volumes were assessed by cardiac MRI from multislice, multiphase gradient-echo sequences that were traced manually and calculated by summation.

Blood was collected after overnight fasting and stored at −80°C. B-type natriuretic peptide (BNP) was measured by radioimmunoassay (Phoenix Pharmaceuticals).^15,23 High-sensitivity C-reactive protein and plasma interleukin 6 were measured by enzyme-linked immunosorbent assays (eMethods in the Supplement).

The principal investigator and all study investigators, except the biostatistician, were blinded to all study outcomes. Personnel performing the outcome measures were blinded to participant group. For practical considerations, an exception was for cardiopulmonary exercise testing, for which the supervising physician and staff were blinded to the baseline (prerandomization) results. To minimize bias, standardized procedures known to elicit maximal exercise performance were used, including a standardized protocol, guidance by the respiratory exchange ratio (RER, an objective indicator of effort) and Borg scale of perceived exertion, and reading of a standardized participant instruction script prior to each exercise test.

After baseline assessments were completed, participants were randomized using a computer-generated list using SAS software (SAS Institute), version 9.0, maintained by the study statistician and stratified by β-blocker medication and sex to 1 of 4 groups consisting of exercise only, diet only, combined exercise and diet (exercise + diet), or attention control (control). No blocking across time was used.

Participants randomized to either group receiving exercise completed 1-hour supervised exercise sessions 3 times per week for 20 weeks consisting primarily of walking exercise using an individualized exercise prescription based on the exercise test results, and intensity level was progressed as tolerated and based primarily on heart rate reserve.^15,16

Participants randomized to either group receiving diet were prescribed a hypocaloric diet using meals (lunch, dinner, and snacks) prepared by the Wake Forest University General Clinical Research Center metabolic kitchen under direction of a registered dietitian. Participants prepared their own breakfast from a menu. Individual energy needs were calculated from resting metabolic rate (CCM Express, MGC Diagnostics) following an overnight fast and an activity factor based on self-reported daily activity. Prescribed calorie intake deficits were approximately 400 kcal/d for the diet group and approximately 350 kcal/d for the exercise + diet group (the difference between the groups allowed for the energy expenditure of the exercise intervention), but not less than 1000 kcal/d. The diet provided approximately 1.2 g of protein/kg ideal body weight, 25% to 30% fat calories, and the remainder as carbohydrate. Participants were provided daily calcium supplements (600 mg) and kept records of all food consumed, which was monitored weekly.

Participants randomized to control received neither diet nor exercise interventions and voluntarily agreed to not make diet or exercise changes during the 20-week study. They received telephone calls every 2 weeks from staff in an attempt to match that received by participants in the diet and exercise groups.

All analyses were performed with SAS (SAS Institute), version 9.0. The study was a 2 × 2 factorial design to estimate the main effect of the 2 interventions, exercise and diet, and the primary and exploratory outcomes were tested at the 5% 2-sided level of significance. The trial was designed to have 2 co–primary outcomes, the performance measure peak V̇o₂ (mL/kg/min) and the MLHF questionnaire total score. All available outcome data were analyzed in an intention-to-treat analysis. The analysis testing the main effects of diet and exercise and their interaction was performed using analysis of covariance with the baseline measure of the outcome measure, sex, and β-blocker usage as covariates. This method adjusts for differences in the means of the baseline measure of the outcome and other predictor covariates to estimate what the mean in each level of the factor would be had both groups had the same overall mean of the covariates in the model. This method is equivalent to multiple imputation of missing data with the covariates as predictors and infinite iterations. We also performed sensitivity analyses to assess the effect of missing data. The least squares means were presented along with either the standard error or 95% CI. The main effect of each intervention, which is the difference in the LS means between the 2 levels of each of the factors (exercise and diet) is presented along with its 95% CI and a P value. Based on a previous study of participants with heart failure, sample size calculations indicated that 80 evaluable participants would provide more than 80% power to detect a main effect of 6% in peak V̇o₂ and effect size of 20% on MLHF total score. Allowing for up to 20% loss to follow-up, 100 participants were randomized to the 4 groups. Because the test for interaction between the 2 factors, which is a linear contrast between the 4 individual group means, has low power, the 2 interventions were considered additive (complementary) only if the P value for intervention was .10 or lower.

Baseline participant characteristics are presented as mean and standard deviation or frequency and percent. Associations between changes in exercise capacity and other variables were made by Pearson correlations.¹⁹

See the Supplement for additional details on sample size, effect size, testing for interaction, multiple comparisons, multiple stepwise regression, and missing data.

From 1586 records reviewed, 577 participants were further screened by telephone; 167 were scheduled for a screening visit. Ultimately, 100 participants with HFPEF (mean [SD]: age, 67 years [5]; BMI, 39.3 [5.6]) were enrolled and randomized: exercise, 26; diet, 24; exercise + diet, 25; control, 25 (Figure 1). Of these, 92 participants (exercise, 24; diet, 24; exercise + diet, 22; control, 22) completed the intervention and follow-up testing (Figure 1). Participant characteristics were generally in accord with those observed in population studies, with predominantly women (n = 81) and high rates of hypertension, diabetes, left ventricular hypertrophy, and diastolic dysfunction (Table 1).

Both diet and exercise significantly increased exercise capacity as determined by the co–primary outcome, peak V̇o₂ (main effect: diet, 1.3 mL/kg body mass/min [95% CI, 0.8 to 1.8]), P < .001; exercise, 1.2 mL/kg body mass/min [95% CI, 0.7 to 1.7], P < .001). Diet and exercise were additive (complementary) and together produced an increase in peak V̇o₂ of 2.5 mL/kg/min. The co–primary measure of QOL as measured by the MLHF total score was not significantly different with exercise and with diet (main effect: exercise, −1 unit [−8 to 5], P = .70; diet, −6 units [−12 to 1], P = .08) (Figure 2) (Table 2).

By main effects analysis, both diet and exercise significantly increased multiple other measures of exercise capacity as determined by peak V̇o₂ expressed in mL/kg lean body mass/min (exercise, 2.1 [95% CI, 1.0 to 3.1], P = .002; diet, 1.36 [95% CI, 0.2 to 2.3], P = .03) by DXA, in mL/kg lean leg mass/min (exercise, 6.2 [95% CI, 2.7 to 9.7], P = .008; diet, 4.5 [95% CI, 0.9 to 8.0], P = .01), and by MRI in mL/cm² thigh muscle area/min (exercise, 1.1 [95% CI, 0.7 to 1.5], P < .001; diet, 0.6 [95% CI, 0.2 to 1.1], P = .002), as well as V̇o₂ reserve (peak minus rest; exercise, 97 mL/min [95% CI, 44 to 141], P = .005; diet, 59 mL/min [95% CI, 6 to 113], P = .30), exercise time to exhaustion (exercise, 2.0 min [95% CI, 1.4 to 2.6] , P < .001; diet, 1.6 min [95% CI, 1.0 to 2.2], P < .001), peak workload (measured in metabolic equivalents [METs]; exercise, 0.8 MET [95% CI, 0.4 to 1.1], P < .001; diet, 0.7 MET [95% CI, 0.4 to 1.1], P < .001), and 6-minute walk distance (exercise, 32.9 m [95% CI, 18.2 to 46.3], P < .001; diet, 25.9 m [95% CI, 11.9 to 40.2], P = .005) (Table 2). Mean peak RER values were more than 1.10 for all groups at baseline and follow-up, suggesting exhaustive effort. There was an exercise × diet super-additive (synergistic) interaction for 6-minute walk distance (P = .09). There were no other significant exercise × diet interactions (Table 2), suggesting the interventions were additive (complementary) for other variables. With diet, muscle quality significantly improved (main effect, 0.15 W/cm² [95% CI, 0.03 to 0.27]) but leg press power did not significantly change (main effect, 11 W (95% CI, −2 to 23), P = .09) (Table 2).

Diet but not exercise significantly improved the KCCQ score, a heart failure–specific QOL measure, by 7 units (95% CI, 2.6 to 12.3; P = .004), substantially greater than the accepted threshold (5 units) for clinical relevance (Table 2).²¹ Diet also significantly improved the general QOL SF-36 physical component score (diet main effect, 4 units [95% CI, 1 to 7], P = .02) (Table 2). There were no significant exercise × diet interactions.

Body weight was significantly decreased by both diet and exercise (Table 3; eFigure 2 in the Supplement) (main effect: exercise, −3 kg [95% CI, −5 to −1], P < .001; diet, −7 kg [95% CI, −9 to −5], P < .001). With diet main effect analysis, the DXA measures of lean body mass (−2 kg [95% CI, −3 to −1], P < .001), fat mass (−5 kg [95% CI, −6 to −4], P < .001), and percentage of fat mass (−2% [95% CI, −3% to −1%], P < .001) were significantly decreased whereas percentage of lean body mass was significantly increased (2% [95% CI, 1% to 3%], P < .001); in contrast with exercise only, fat mass was decreased (−2 kg [95% CI, −3 to −1], P = .001) (Table 3). With diet main effect analysis, MRI measures of thigh subcutaneous fat (−16 cm² [95% CI, −22 to −10], P < .001), thigh muscle (−6 cm² [95% CI, −9 to −3], P < .001), abdominal subcutaneous fat (−5 cm² [95% CI, −20 to 11], P < .001), and visceral fat were significantly decreased (−31 cm² [95% CI, −43 to −19], P < .001) (Table 3); there were no significant changes with exercise. There was no change in pericardial or epicardial fat. There were no significant exercise × diet interactions (Table 3).

With diet main effect analysis, left ventricular mass by MRI (−4 g [95% CI, −7 to 0], P = .03) and left ventricular relative wall thickness by echocardiography (−0.03 [95% CI, −0.05 to −0.01], P = .005) were significantly decreased and mitral E/A velocity ratio (0.10 [95% CI, 0.02 to 0.17], P = .01) was significantly increased (Table 3). No other cardiac MRI or Doppler echocardiography measure was significantly changed (Table 3; eTable 3 in the Supplement). Arterial pulse-wave velocity was unchanged by either diet or exercise (Table 3).

With both diet and exercise, New York Heart Association symptom class significantly improved (main effect: exercise, −0.4 class [95% CI, −0.6 to −0.2], P < .001; diet, −0.4 class [−0.5 to −0.2], P = .001) (Table 2).

With diet but not exercise, high-sensitivity C-reactive protein was significantly reduced (main effect: diet, −2.8 µg/L [95% CI, −4.9 to −0.7], P = .02; exercise, −0.4 [95% CI, −2.5 to 1.6], P = .44); changes in plasma interleukin-6 were nonsignificant (main effect: diet, −0.8 pg/mL [95% CI, −1.5 to −0.1], P = .09; exercise, 0.4 mL [95% CI, −0.3 to 1.1], P = .51); and there was no interaction (eTable 4 in the Supplement). The reduction in high-sensitivity C-reactive protein correlated with the reduction in weight (r = 0.29; P = .005). With diet but not exercise, there were significant reductions in total cholesterol (to convert cholesterol to mmol/L, multiply by 0.0259; main effect: diet, −14 mg/dL [95% CI, −24 to −14], P = .008; exercise, −4 mg/dL [95% CI, −14 to 6], P = .40) and low-density lipoprotein cholesterol (to convert low-density lipoprotein cholesterol to mmol/L, multiply by 0.0259; main effect: diet, −13 mg/dL [95% CI, −21 to −4], P = .008; exercise, −4 mg/dL [95% CI, −12 to 4], P = .35) (eTable 4 in the Supplement); these changes persisted after adjustment for lipid-lowering medications.

In the overall groups combined, change in peak V̇o₂ was inversely related to change in total mass and fat mass, and was positively related to change in percentage of lean body mass and thigh skeletal muscle to intermuscular fat ratio (SM:IMF) (eFigure 3 in the Supplement); there were also correlations with change in left ventricular mass (r = −0.27; P = .02) and high-sensitivity C-reactive protein (r = −0.21; P = .047). Similar results were observed with exercise time as the exercise capacity variable. Multiple stepwise regression showed that sex and change in total mass were the only independent predictors of change in peak V̇o₂ (eMethods in the Supplement).

Participants completing the exercise interventions attended a median of 84% (SD, 14%) of the exercise sessions, and together progressed from an average 19 minutes (SD, 6) at a 2.8 MET (SD, 0.4) at week 1 to an average 49 minutes (SD, 10) at a 3.8 MET (SD, 1.2) at week 20. Further details regarding attendance and progression are in the Supplement.

The average actual caloric intake deficit was −388 kcal/d (SD, 55) for diet-only participants and −355 kcal/d (SD, 23) for exercise + diet participants. Dietary adherence (actual vs prescribed calorie level) from recorded food logs was 99% (SD, 1%) for both diet groups.

Five adverse events judged as possibly related to the intervention occurred among 5 participants: hypoglycemia between meals in 2 participants (diet and exercise + diet groups), ankle pain and swelling later diagnosed as partial tendon tear (exercise + diet group), stress foot fracture (exercise group), and an episode of unusual shortness of breath during exercise (exercise group). Three participants had a total of 6 hospitalizations, all judged unrelated to study participation: 1 participant was hospitalized for pancreatitis (exercise group), 1 participant had 3 hospitalizations for heart failure exacerbation or dyspnea (exercise + diet group), and 1 participant had 2 hospitalizations for leg edema, pain, and erythema (control group). There were no deaths.

Quiz Ref IDThe major novel findings of this randomized clinical trial are that among older obese participants with chronic, stable HFPEF, intentional weight loss via Caloric restriction diet was feasible, appeared safe, and significantly improved the co–primary outcome of exercise capacity. The combination of diet with exercise, the only intervention previously shown to improve exercise capacity in HFPEF,^15,16,25 produced a robust increase in exercise capacity (+2.5 mL/kg/min V̇o₂, substantially greater than the accepted clinically meaningful increase of 1.0 mL/kg/min). The co–primary outcome of QOL, as measured by the MLHF total score, did not show a significant change with either exercise or diet compared with control. There was also a significant change in 2 other standardized measures of QOL, the KCCQ score (a heart failure–specific QOL instrument) and the SF-36 physical score (a general QOL instrument). These were exploratory QOL measures that raised the possibility of an effect on QOL.

These results are credible because studies of diet alone or in combination with exercise in non–heart failure clinical populations have shown similar overall findings.^11,12 In a randomized clinical trial with similar design and sample size of obese, frail, older adults, diet, exercise, and their combination significantly improved peak V̇o₂ and other measures of physical function and the effects of diet and exercise were additive (complementary).¹¹ Other studies also have shown significantly greater improvements in body composition with diet than exercise.^11,26 Our finding of more improvement in exploratory QOL measures with diet than with exercise is also credible based on prior exercise and diet studies in obese older persons¹¹ and because exercise has not consistently improved QOL in HFPEF.^15,16,25 Furthermore, preliminary reports have indicated that weight reduction via bariatric surgery can prevent the onset of heart failure, and can improve exercise capacity in patients with heart failure and reduced ejection fraction (HFREF).^27,28

As others have indicated, peak V̇o₂ relative to body weight (mL/kg/min), the preplanned co–primary outcome, is the most relevant measure of exercise capacity during weight-bearing (treadmill) exercise.^11,29 A true increase in exercise tolerance with diet is further supported by (1) significant increases in 4 other measures that are relatively independent of body mass (V̇o₂ reserve, exercise time to exhaustion, workload, 6-minute walk distance); (2) preservation of absolute peak V̇o₂ (mL/min); and (3) improvement in leg power that occurred despite significant loss of muscle mass. The largest increase in exercise capacity was with exercise and diet combined.

What are the potential mechanisms underlying improved exercise capacity? Increased adipose tissue mass promotes inflammation, hypertension, insulin resistance, and dyslipidemia resulting in impaired cardiac, arterial, and skeletal muscle function, all of which contribute to reduced exercise capacity in patients with HFPEF^9,30-34 and can be reversed with diet.^6,8,9,13,35 Using DXA, we recently reported that percentage of body fat and percentage of leg fat were significantly increased whereas percentage of body lean and leg lean mass were reduced in older participants with HFPEF vs controls and were related to reduced exercise capacity.¹⁰ Using MRI, we found that older participants with HFPEF have increased thigh IMF, despite a normal amount of subcutaneous fat.⁶ Furthermore, the SM:IMF ratio was increased, and both IMF area and SM:IMF ratio were independent predictors of peak V̇o₂.⁶ IMF may compete with muscle tissue for critical blood flow during exercise reducing perfusive oxygen.³⁶ IMF may also reduce diffusive oxygen transport by increasing the distance oxygen must traverse from the capillary to the muscle mitochondria. Furthermore, increased IMF may reduce skeletal muscle capillary density and mitochondrial biogenesis and oxidative metabolism, all of which are abnormal in HFPEF.^32,34,37 In our study, increased peak V̇o₂ was associated with reduced fat mass, increased percentage of lean mass and thigh SM:IMF ratio, and reduced inflammation biomarkers. Thus, improvement in peak V̇o₂ from diet and exercise may be due to reduced inflammation and enhanced mitochondrial function, attenuated reactive oxygen species generation, increased vascular oxidative stress resistance, increased nitric oxide bioavailability, and improved microvascular function. Together, these may increase diffusive oxygen transport or oxygen utilization by the active muscles.³⁸

With diet, left ventricular mass and relative wall thickness decreased and left ventricular E/A ratio increased, but we observed no other improvements in resting cardiac function. We also observed no significant changes in epicardial or pericardial fat, in contrast with reduced adipose tissue elsewhere. Although we did not measure cardiac function during exercise, these data suggest that the improvements we observed with diet and its combination with exercise may be due primarily to favorable “noncardiac” peripheral adaptations, in accord with reports of exercise in HFPEF.^16,33

Because of the reported “heart failure obesity paradox” (lower mortality observed in overweight or obese individuals),⁵ before diet can be recommended for obese patients with HFPEF, further studies likely are needed to determine whether these favorable changes are associated with reduced clinical events. However, a recent meta-analysis of randomized trials among older patients without heart failure indicates that intentional weight loss from diet is associated with a 15% reduction in total mortality.³⁹

As observed in other diet studies and despite adequate protein intake, there was a significant decrease in muscle mass with diet that was not prevented by exercise. Although the long-term consequences of this are unclear, the muscle loss did not prevent increases in exercise capacity or leg power. Inclusion of strength training may have reduced loss of muscle mass during diet.

This was a randomized clinical trial with frequent monitoring, professionally administered diet, and medically supervised exercise; safety and efficacy could differ under other conditions. The minimum BMI was 30, which includes most patients with HFPEF.^4,5 However, our data do not address safety and efficacy of diet in patients with lower BMI.

Our participants had typical clinical features of HFPEF (including severe exercise intolerance, left ventricular hypertrophy, and diastolic dysfunction, 76% on maintenance diuretics) and met predetermined criteria for HFPEF utilized in prior publications and recommended by the American Heart Association and the American College of Cardiology¹⁴ and European Society of Cardiology. The relatively modest BNP levels likely result from (1) nonhospitalized, stable participants who were clinically well-compensated as required for exhaustive exercise testing and the 20-week intervention; (2) a strong, inverse relation that exists between BNP and BMI,⁴⁰ such that, when matched for other disease variables, BNP is much lower in obese than nonobese patients with HFPEF, frequently less than 100 pg/mL^40,41; (3) BNP levels are significantly lower in HFPEF than HFREF due to lower left ventricular wall stress.²³

Among obese older patients with clinically stable HFPEF, caloric restriction or aerobic exercise training increased peak V̇o₂, and the effects may be additive. Neither intervention had a significant effect on quality of life as measured by the MLHF Questionnaire.

Corresponding Author: Dalane W. Kitzman, MD, Wake Forest School of Medicine, Medical Center Blvd, Winston Salem, NC 27157-1045 (dkitzman@wfubmc.edu).

Author Contributions: Drs Kitzman and Morgan had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Kitzman, Morgan, Hundley, Nicklas.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Kitzman, Brubaker, Morgan, Haykowsky, Eggebeen.

Critical revision of the manuscript for important intellectual content: Kitzman, Brubaker, Haykowsky, Hundley, Kraus, Eggebeen, Nicklas.

Statistical analysis: Kitzman, Morgan, Eggebeen.

Obtained funding: Kitzman, Nicklas.

Administrative, technical, or material support: Kitzman, Brubaker, Hundley, Eggebeen, Nicklas.

Study supervision: Kitzman, Brubaker, Eggebeen, Nicklas.

Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Kitzman reports receiving personal fees from Icon, Relypsa, Abbvie, Regeneron, GlaxoSmithKline, Merck, Forest Labs, and Corvia Medical; receiving grant support from Novartis; and owning stock in Gilead Sciences and stock options in Relypsa. No other disclosures were reported.

Funding/Sponsor: This study was supported by research grants R01AG18915, P30AG021332, R01HL093713, and R01AG020583 from the National Institutes of Health. Also supported in part by the Kermit Glenn Phillips II Chair in Cardiovascular Medicine at Wake Forest School of Medicine (Dr Kitzman) and the Mortiz Chair in Geriatric Nursing Research in the College of Nursing and Health Innovation at The University of Texas at Arlington (Dr Haykowsky).

Role of the Funder/Sponsor: The sponsors had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.