Anemia is common in patients with chronic heart failure (HF), with a prevalence ranging from 10% to 56%, and may be a risk factor for poor outcomes. Anemia in HF remains poorly understood, with significant gaps in its impact on health-related quality of life (HRQoL), with most studies in HF being retrospective or from registries. The purpose of this study was to explore the relation of hemoglobin (Hgb) with HRQoL and training-induced changes in HRQoL in a cohort of patients in Heart Failure: A Controlled Trial Investigating Outcomes of Exercise Training (HF-ACTION). Using data from HF-ACTION, a randomized controlled trial of exercise training in patients with HF and low left ventricular ejection fractions, HRQoL was measured using the Kansas City Cardiomyopathy Questionnaire (KCCQ) at baseline, 3 and 12 months, and annually up to 4 years. Treatment group effects on HRQoL were estimated using linear mixed models according to the intention-to-treat principle. It was hypothesized that baseline Hgb would be correlated with baseline KCCQ scales and that Hgb would moderate the beneficial effect of exercise training on HRQoL. Hgb level was not significantly correlated with baseline HRQoL. Baseline Hgb did not moderate the beneficial effect of exercise training on KCCQ overall or subscales relative to usual care. In conclusion, in the HF-ACTION cohort, there was no correlation with baseline Hgb and baseline HRQoL as measured by the KCCQ. In addition, the beneficial effects of HRQoL from exercise training were not modulated by baseline Hgb.
The purpose of this study was to explore the relationship of hemoglobin (Hgb) with health-related quality of life (HRQoL) and exercise training-induced changes in HRQoL. We examined this in the context of the National Heart, Lung, and Blood Institute–sponsored study Heart Failure: A Controlled Trial Investigating Outcomes of Exercise Training (HF-ACTION), designed to understand the effects of exercise on all-cause mortality and HRQoL in outpatients with heart failure (HF). We hypothesized that patients with HF who had lower Hgb values at baseline would have worse baseline HRQoL and that lower baseline Hgb may moderate the effect of exercise on HRQoL. Previous analysis of the HF-ACTION cohort found significant benefits of the exercise intervention on the Kansas City Cardiomyopathy Questionnaire (KCCQ). In this study, we tested whether patients with higher baseline Hgb levels experience a greater benefit from the exercise intervention than patients with lower baseline Hgb.
HF-ACTION was a multicenter randomized controlled trial of exercise training in outpatients with HF. A complete description of the study design and exercise training protocol has been published previously. Patients were enrolled at 82 centers in the United States, Canada, and France. Inclusion criteria were a left ventricular ejection fraction ≤35%; New York Heart Association (NYHA) class II, III, or IV HF for ≥3 months; optimal HF therapy; and investigator-determined medical stability to begin an exercise program. Patients were excluded if they had major co-morbidities or limitations that could interfere with exercise training, had recent (within 6 weeks) or planned (within 6 months) major cardiovascular events or procedures, were already exercising regularly, or had devices that limited the ability to achieve target heart rates.
Study sites in the HF-ACTION trial were allowed to report Hgb and other baseline laboratory values that had been recorded in patient charts up to 1 year before study enrollment. All sites were experienced HF centers that would have repeated blood testing had there been any suspected changes after their reporting values. Seventy-six percent (n = 1,763) of the total of 2,331 randomized patients had Hgb values recorded at baseline. Thus, for the purpose of this report, the value provided by the sites at the time of enrollment is considered the baseline Hgb.
HRQoL data were collected using the KCCQ, a 23-item self-administered disease-specific health status questionnaire. In addition to an overall summary score, the instrument provides subscales for physical limitations, symptoms, quality of life, and social limitations. The KCCQ is scored from 0 to 100, with higher scores representing better health status. We handled missing values in each KCCQ domain using the standard scoring algorithms to assign the average of the completed items within the domain, assuming that a domain-specific threshold number of items in that domain (usually half) were answered. The KCCQ was self-administered at the baseline clinic visit, at 3-month intervals during clinic visits for the first 12 months, and annually thereafter for up to 4 years.
We used SAS version 9.1 (SAS Institute Inc., Cary, North Carolina) for all analyses. Baseline patient health status is expressed as mean ± SD. Missing data were comparable across the 2 groups, accounting for death, withdrawn consent, and rolling enrollment into the trial.
For all assessments, we considered a 2-tailed α level of 0.05 to be significant. To examine the relation between baseline Hgb and baseline health status, we used Pearson’s correlation coefficients. We used the bootstrap method to estimate confidence intervals, and we constructed the confidence intervals using the bias-corrected and accelerated method. To examine whether Hgb moderates the effect of exercise training for patients with HF, we used a longitudinal linear mixed-effects model. A mixed model approach models the underlying health trajectory for each patient that gives rise to the observed KCCQ scores at various time points, then compares the average trajectories between treatment groups. This method minimizes the effects of measurement error from any 1 assessment. We used full maximum likelihood estimation to model all available data from each patient without the need to impute missing values or omit patients with missing data from the analysis. This method provides unbiased estimates when the mechanism responsible for missing data can be ignored, that is, when unobserved variables do not explain the probability of missingness over and above what is explained by observed variables. The observed variables included age, gender, NYHA class, cause of HF, systolic and diastolic blood pressures, heart rate at rest, the ejection fraction, diagnosis of diabetes, body mass index, cardiopulmonary exercise duration, smoking status, Canadian angina class, atrial fibrillation or flutter, use of angiotensin-converting enzyme inhibitors, use of angiotensin receptor blockers, use of implantable cardioverter defibrillators, use of β blockers, use of biventricular pacemakers, use of statins, use of nonloop diuretics and loop diuretics, use of aldosterone antagonists (spironolactone or eplerenone), previous myocardial infarction, previous revascularization, history of peripheral vascular disease, history of chronic obstructive pulmonary disease, number of HF hospitalizations in 6 months before baseline, Beck Depression Inventory score, and Perceived Social Support Scale score.
We previously found that the relation between health status and exercise was not strictly linear, and thus we included the effect of time in the model in 2 pieces: a “jump” from baseline to 3 months and a slope after 3 months. The model included fixed and random effects: (1) fixed effects of treatment group (exercise vs usual care), baseline Hgb, and time, 2-way interactions of baseline Hgb by time, treatment by time, baseline Hgb by treatment, and a 3-way interaction of treatment by baseline Hgb by time, (2) fixed effects of the baseline covariates (each of which had <1% missing data) as well as the 2-way interactions of baseline covariates by time (only those for which an omnibus F test of the pair of interactions between each individual covariate and the time-jump and time-slope was significant), and (3) random effects of intercept and time.
We tested whether Hgb had an effect on health status improvement using an omnibus F test on all fixed-effect terms that included baseline Hgb (i.e., baseline Hgb main effect and 2-way and 3-way interactions that included Hgb). We examined interactions separately if the omnibus test was significant.
Table 1 lists the KCCQ scores by 6 Hgb categories. Patients in the higher categories (14 to 15 and >15 g/dl) tended to have better overall scores, which was reflected primarily in the symptom domain. We examined baseline characteristics of HF-ACTION participants by these same 6 categories of Hgb. There were fewer patients in the lowest category (<11 g/dl). In general, the patients with the lower Hgb levels tended to be older, were more likely women, and tended to have shorter exercise times on baseline cardiopulmonary exercise testing. There were more patients with diabetes and higher NYHA classes at the lower levels of Hgb. In contrast, the patients in the highest Hgb category (>15 g/dl) were more likely to be younger, to have lower NYHA classes, to have longer exercise time, and to have no previous HF hospitalizations in the 6 months before baseline entry.
|≤11 (n = 129)||>11–12 (n = 215)||>12–13 (n = 347)||>13–14 (n = 414)||>14–15 (n = 348)||>15 (n = 310)|
|KCCQ overall summary scale||68.2 ± 20.5||70.4 ± 20.7||70.1 ± 19.3||70.8 ± 19.1||73.1 ± 18.7||72.1 ± 18.9|
|Physical limitations||66.9 ± 21.2||67.6 ± 23.3||66.9 ± 22.5||68.6 ± 22.0||71.4 ± 21.3||70.2 ± 20.8|
|Symptoms||69.3 ± 23.4||72.9 ± 22.0||73.4 ± 20.2||72.9 ± 19.9||74.8 ± 19.5||74.0 ± 20.2|
|Quality of life||57.6 ± 25.5||59.8 ± 24.8||59.3 ± 23.8||59.4 ± 23.6||61.8 ± 24.2||56.8 ± 25.1|
|Social limitations||60.1 ± 27.2||62.3 ± 27.9||62.2 ± 27.5||62.2 ± 27.9||62.7 ± 26.9||60.5 ± 27.5|
Table 2 lists the Pearson’s correlation coefficients for Hgb and KCCQ overall and subscale scores at baseline. There were negligible correlations between Hgb and health status. The correlations between Hgb and the KCCQ physical limitations subscale and the KCCQ symptoms subscale were statistically significantly different from zero; however, they were still negligible in magnitude (0.068 and 0.055 respectively) and not clinically significant.
|KCCQ||r (95% Confidence Interval) ∗|
|Overall summary scale||0.031 (−0.017 to 0.081)|
|Physical limitations subscale||0.068 (0.022 to 0.112)|
|Symptoms subscale||0.055 (0.008 to 0.105)|
|Quality-of-life subscale||−0.003 (−0.051 to 0.044)|
|Social limitations subscale||0.007 (−0.040 to 0.048)|
Figures 1 and 2 show the distributions of Hgb and KCCQ overall scores at baseline for men and women (in red), on the y and x axes, respectively. They also display the LOESS curves and underlying scatterplots (gray dots) of the relations between Hgb and KCCQ for men and women. These relations appear consistent with the magnitudes of the reported correlation coefficients, that is, no relation: the scatterplots and the LOESS curves are flat.
We reported previously that the improvement in the KCCQ overall summary score in the exercise training group was higher than the improvement reported by the usual care group (p <0.001) and that there was no attenuation of this early benefit over time, as neither group experienced significant changes in KCCQ scores after 3 months. In the present analysis, which includes all participants who had at least a baseline KCCQ score and recorded Hgb value, we found that the beneficial effect of exercise on health status was not moderated by Hgb. In the analysis adjusted for baseline covariates, parameter estimates for the 3-way interaction terms of time (either jump or slope) by Hgb by exercise training were not significant for the overall summary scale (p = 0.65 for the jump of baseline to 3 months, p = 0.56 for the slope of 3 months to the end of the study; Table 3 ). Results for the KCCQ subscales were similar to the results for the overall summary scale; none of the 3-way interaction terms were statistically significant.
|Effect||Estimate (95% Confidence Interval)||p Value †|
|KCCQ overall summary scale|
|Exercise training vs usual care||9.19 (−2.56 to 20.94)||0.13|
|Hgb level (g/dl)||0.42 (−0.24 to 1.08)||0.21|
|Hgb level × exercise training||−0.68 (−1.55 to 0.19)||0.12|
|Baseline to 3-month visit (jump)||−15.54 (−27.62 to −3.45)||0.01|
|Jump × exercise training||−1.04 (−12.56 to 10.49)||0.86|
|Jump × Hgb level||0.32 (−0.33 to 0.97)||0.33|
|Jump × Hgb level × exercise training||0.20 (−0.65 to 1.05)||0.65|
|3-month visit to end of study (slope)||0.80 (0.24 to 1.36)||0.01|
|Slope × exercise training||−0.13 (−0.66 to 0.40)||0.62|
|Slope × Hgb level||−0.03 (−0.06 to 0.00)||0.07|
|Slope × exercise training × Hgb level||0.01 (−0.03 to 0.05)||0.56|
|KCCQ physical limitations subscale|
|Exercise training vs usual care||6.42 (−8.02 to 20.87)||0.38|
|Hgb level (g/dl)||0.29 (−0.52 to 1.10)||0.48|
|Hgb level × exercise training||−0.47 (−1.53 to 0.60)||0.39|
|Baseline to 3-month visit (jump)||−19.83 (−34.45 to −5.22)||0.01|
|Jump × exercise training||1.43 (−12.63 to 15.49)||0.84|
|Jump × Hgb level||0.49 (−0.30 to 1.28)||0.22|
|Jump × Hgb level × exercise training||0.07 (−0.96 to 1.11)||0.89|
|3-month visit to end of study (slope)||0.77 (0.13 to 1.40)||0.02|
|Slope × exercise training||−0.31 (−0.92 to 0.29)||0.31|
|Slope × Hgb level||−0.04 (−0.07 to 0.00)||0.04|
|Slope × exercise training × Hgb level||0.02 (−0.02 to 0.07)||0.27|
|KCCQ symptoms subscale|
|Exercise training vs usual care||15.40 (2.52 to 28.28)||0.02|
|Hgb level (g/dl)||0.82 (0.10 to 1.54)||0.03|
|Hgb level × exercise training||−1.12 (−2.07 to −0.17)||0.02|
|Baseline to 3-month visit (jump)||−7.78 (−19.16 to 3.61)||0.18|
|Jump × exercise training||−1.95 (−14.42 to 10.52)||0.76|
|Jump × Hgb level||0.16 (−0.53 to 0.85)||0.65|
|Jump × Hgb level × exercise training||0.25 (−0.67 to 1.17)||0.59|
|3-month visit to end of study (slope)||0.82 (0.33 to 1.31)||0.001|
|Slope × exercise training||−0.21 (−0.75 to 0.33)||0.44|
|Slope × Hgb level||−0.03 (−0.06 to −0.01)||0.02|
|Slope × exercise training × Hgb level||0.02 (−0.02 to 0.06)||0.42|
|KCCQ quality-of-life subscale|
|Exercise training vs usual care||10.08 (−4.31 to 24.46)||0.17|
|Hgb level (g/dl)||0.42 (−0.39 to 1.22)||0.31|
|Hgb level × exercise training||−0.77 (−1.83 to 0.29)||0.16|
|Baseline to 3-month visit (jump)||−17.21 (−32.49 to −1.93)||0.03|
|Jump × exercise training||−1.37 (−16.15 to 13.42)||0.86|
|Jump × Hgb level||0.30 (−0.52 to 1.13)||0.47|
|Jump × Hgb level × exercise training||0.19 (−0.90 to 1.28)||0.74|
|3-month visit to end of study (slope)||0.90 (0.28 to 1.53)||0.005|
|Slope × exercise training||−0.40 (−1.02 to 0.22)||0.21|
|Slope × Hgb level||−0.04 (−0.08 to −0.01)||0.02|
|Slope × exercise training × Hgb level||0.03 (−0.01 to 0.08)||0.17|
|KCCQ social limitations subscale|
|Exercise training vs usual care||5.49 (−12.41 to 23.40)||0.55|
|Hgb level (g/dl)||0.26 (−0.74 to 1.25)||0.61|
|Hgb level × exercise training||−0.42 (−1.74 to 0.90)||0.54|
|Baseline to 3-month visit (jump)||−13.93 (−32.78 to 4.91)||0.15|
|Jump × exercise training||−3.20 (−20.51 to 14.12)||0.72|
|Jump × Hgb level||0.24 (−0.73 to 1.20)||0.63|
|Jump × Hgb level × exercise training||0.34 (−0.93 to 1.62)||0.60|
|3-month visit to end of study (slope)||0.57 (−0.21 to 1.35)||0.16|
|Slope × exercise training||0.44 (−0.27 to 1.15)||0.23|
|Slope × Hgb level||−0.01 (−0.05 to 0.03)||0.72|
|Slope × exercise training × Hgb level||−0.03 (−0.08 to 0.03)||0.32|