Prognostic Value of Health Status in Patients With Heart Failure After Acute Myocardial Infarction
Background— Disease-specific health status instruments such as the Kansas City Cardiomyopathy Questionnaire (KCCQ) can quantify symptoms, functional limitations, and quality of life in patients with heart failure. Understanding the relationship between KCCQ scores and prognosis may assist clinicians in both interpreting KCCQ scores and stratifying risk in patients.
Methods and Results— We examined the prognostic value of the KCCQ in a prospective, international cohort of 1516 patients with heart failure after a recent acute myocardial infarction. We focused on the relationship between the KCCQ overall score (KCCQ-os), measured at the first outpatient visit (4 weeks after enrollment), and subsequent 1-year cardiovascular mortality or hospitalization (n=258, 20.3%). KCCQ-os was strongly associated with subsequent cardiovascular events in that those with a score ≥75 had an 84% 1-year event-free survival compared with 59% for those with a score <25 (P<0.001). After demographic and other clinical characteristics were controlled for in multivariable models, KCCQ-os remained strongly associated with outcome (hazard ratio, 2.02; 95% CI, 1.24 to 3.27 for KCCQ-os <25; P<0.001).
Conclusions— In outpatients with heart failure complicating an acute myocardial infarction, KCCQ-os is strongly associated with subsequent 1-year cardiovascular mortality and hospitalization. Use of the KCCQ in outpatient clinical practice can both quantify patients’ health status and provide insight into their prognosis. (Circulation. 2004;110:546-551.)
Received December 1, 2003; de novo received February 16, 2004; revision received April 6, 2004; accepted April 8, 2004.
The American College of Cardiology/American Heart Association (ACC/AHA) guidelines for the management of heart failure (HF) recommend that healthcare providers routinely evaluate patients’ health status during the process of outpatient follow-up and use these assessments to guide care.1,2 Furthermore, recent outpatient performance measures for HF care developed by the ACC/AHA and the American Medical Association’s Physician Consortium for Performance Improvement have established the explicit quantification of patients’ symptoms and function as an indicator of healthcare quality.2 Although patient-reported measures capable of quantifying health status are widely available, they have not yet gained wide acceptance outside clinical trials. Among the possible explanations for this phenomenon is the sense by clinicians that the complexity of collecting and scoring formal health status questionnaires from patients is not offset by the generation of insights that can improve patient care. By describing the association between measures that quantify patients’ health status (their symptoms, function, and quality of life) with prognosis, we can create a better appreciation of how to interpret scores on these measures and support their broader use in clinical care.
The Kansas City Cardiomyopathy Questionnaire (KCCQ) is a disease-specific instrument that quantifies a broad range of health status domains, including patient-reported symptoms, physical and social limitations, and quality of life. It has been shown to be valid, reliable, and highly responsive.3 Furthermore, as a standardized tool that collects data directly from patients, the KCCQ offers a means of patient assessment that is observer independent and highly reproducible in clinically stable patients.3 However, unlike the physician-derived NYHA scale, there is a paucity of data from large prospective studies of HF patients demonstrating the prognostic value of self-reported instruments with respect to traditional clinical end points such as death or hospitalization.
We explored the association between health status, as measured by the KCCQ, and subsequent cardiovascular mortality and hospitalizations in a large international clinical trial enrolling outpatients with congestive HF after an acute myocardial infarction (MI). We focused on the prognostic value of the KCCQ overall score (KCCQ-os), a composite measure of patients’ symptoms, function (both physical and social), and quality of life, because as a single summary score it is readily amenable to rapid clinical review and interpretation.
The study sample was derived from the Eplerenone Post–Acute Myocardial Infarction Heart Failure Efficacy and Survival Study (EPHESUS) Quality of Life Substudy, which has been previously described.4,5 In brief, EPHESUS was a multicenter international clinical trial that examined the efficacy of eplerenone, a selective aldosterone blocker, on reducing morbidity and mortality among patients with HF after an acute MI. Between December 27, 1999, and December 31, 2001, patients meeting the following inclusion criteria were enrolled: evidence of an acute MI as documented by standard criteria, left ventricular (LV) dysfunction defined as an ejection fraction of ≤40%, and HF as documented by physical examination findings or chest radiography showing pulmonary venous congestion. All patients were enrolled between days 3 and 14 after their MI and received cardiac treatment at the discretion of their physician(s), which could have included ACE inhibitors, angiotensin-receptor blockers, β-blockers, digoxin, diuretics, nitrates, antiplatelet agents, and coronary reperfusion. For the Quality of Life Substudy, the KCCQ was administered to all EPHESUS enrollees in the following countries: Argentina, Belgium, Brazil, Canada, France, Germany, the Netherlands, Spain, the United Kingdom, and the United States. The institutional review board or ethics committee at each site approved the protocol, and all patients provided written informed consent before enrollment.
Health Status Data
The KCCQ is a self-administered, disease-specific health status measure for patients with HF consisting of 23 items that quantify the following clinically relevant domains: physical limitations, symptoms (frequency, severity, and recent change over time), self-efficacy, quality of life, and social limitation.3 Each scale is transformed to a score of 0 to 100 in which higher scores indicate better function (eg, less physical limitation, fewer symptoms, and better quality of life). To facilitate interpretability from across the many domains, a summary score was developed. The KCCQ-os is calculated as the mean of scores from the physical limitation, symptom (excluding symptom stability), quality of life, and social limitation domains. For this study, KCCQ data, using linguistically and culturally validated translations, were collected at the time of screening and during follow-up at 4 weeks, 3 months, 6 months, and every 6 months thereafter until the closing date of the study, August 30, 2002.
To focus on the relationship between health status and event-free survival among outpatients with HF, we selected the first follow-up visit (4 weeks after enrollment) as patients’ new, post-MI HF baseline. Clinical parameters collected at this visit included heart rate, systolic and diastolic blood pressures, KCCQ-os, NYHA classification, and medication use. Other data collected at enrollment included demographic characteristics (age, gender, race, country), medical comorbidities [atrial fibrillation, chronic obstructive pulmonary disease (COPD), diabetes, dyslipidemia, hypertension, renal insufficiency, smoking history], cardiovascular disease severity markers [prior congestive HF, prior MI, prior stroke or transient ischemic attack (TIA), unstable angina, post-MI LV ejection fraction], and index MI features (Q-wave versus non–Q-wave MI, attempted reperfusion). Data on the use of ACE inhibitors and β-blockers were collected throughout follow-up.
The 2 primary end points were (1) time to death from cardiovascular causes or first hospitalization for a cardiovascular event, including recurrent MI, HF, stroke, or ventricular arrhythmia, and (2) all-cause mortality. All end points were adjudicated by a blinded critical-events committee. Definitions of all adjudicated end points have been previously published.4
Patients were grouped into the following 4 easy-to-interpret categories on the basis of their KCCQ-os scores: group 1 (0 to <25), high-risk patients; group 2 (25 to <50), moderate- to high-risk patients; group 3 (50 to <75), low- to moderate-risk patients; and group 4 (75 to 100), low-risk patients. One-year survival curves for these KCCQ-os groups were derived by Kaplan-Meier analysis and were compared by use of log-rank tests. Kaplan-Meier analyses were also constructed from the KCCQ data obtained at 3 and 6 months, with subsequent survival reported relative to the time of each assessment (ie, using the 3- or 6-month scores as the patients’ new baselines and observing event rates over the next 12 months).
The independent prognostic utility of the KCCQ was assessed through the use of Cox proportional-hazards regression models. Crude hazard ratios were estimated for the KCCQ and the demographic and clinical factors listed in Table 1 (A country effect was included in all models to account for potential confounding). Age, post-MI LV ejection fraction, and baseline (ie, 4-week) vitals were treated as continuous variables; all other variables were categorical. Continuous variables were evaluated for fit linearly and nonlinearly by use of polynomial terms and cut-point categorizations. Next, a base model was developed that incorporated demographic and clinical covariates. Clinical judgment and statistical criteria (stepwise variable selection; P<0.1 for entry and retention) were used in the selection of model terms. Finally, the 4-week KCCQ-os was added to estimate its incremental prognostic contribution after accounting for other model factors; the KCCQ-os was examined as both a categorical and a continuous variable. Model deviance residuals were examined to evaluate the adequacy of parameterization of the individual model terms. Proportional-hazards assumptions were verified with Schoenfeld residuals. Results are reported as hazard ratios and 95% CIs. Analyses were performed with SAS version 8.2 (SAS Institute, Inc) and R version 220.127.116.11
A total of 1988 post-MI HF patients from 10 countries were initially enrolled in the Quality of Life Substudy. Of these, 472 were excluded during the first 4-week period for the following reasons: death (89), study withdrawal (233), missed 4-week follow-up (147), or no KCCQ data available (3). The remaining 1516 patients established the cohort for the present analyses. The average time of the 4-week assessment was 30±4 days (range, 14 to 63 days) after enrollment into EPHESUS; average follow-up time from patients’ 4-week assessment was 16±6 months after study enrollment. Over this period, 179 patients (11.8%) died: 145 (9.6%) died of cardiovascular causes, and 375 (24.7%) had a combined end point of cardiovascular death and cardiovascular hospitalization.
The baseline (4-week) characteristics of the study population are summarized in Table 1. The mean age was 64 years, with 8.6% being ≥80 years of age. Most patients experienced a Q-wave MI (67%), and reperfusion therapy was performed on 60% of patients by the time of study enrollment. The mean post-MI LV ejection fraction was 32±7%. The mean KCCQ-os at 4 weeks was 68±22, with 22.4% of patients attaining a score <50; 15.2% of patients were NYHA class III/IV. Eighty-nine percent of patients were on ACE inhibitors or angiotensin-receptor blockers, and 73% of patients were receiving β-blocker therapy.
Prognostic Value of the KCCQ Overall Score
The Figure shows the Kaplan-Meier event-free survival curves for patients stratified by range of KCCQ-os at 1, 3, and 6 months. Patients with a KCCQ-os ≥75 (ie, the low-risk group) had an 84% 1-year event-free survival compared with 59% for the high-risk group having a score <25 (P<0.001). Similar trends were observed with the 3-month scores (87% 1-year event-free survival for KCCQ-os ≥75 versus 51% for KCCQ-os <25; P<0.001) and 6-month scores (90% 1-year event-free survival for KCCQ-os ≥75 versus 60% for KCCQ-os <25; P<0.001). Looking at all-cause mortality, patients with a KCCQ-os ≥75 had a 94% 1-year survival compared with 81% for those with a score <25 (P<0.001).
The results of bivariate analyses examining the relative hazard of cardiovascular death or hospitalization over the duration of follow-up by demographic characteristics, comorbidities, cardiovascular disease severity markers, index event characteristics, and clinical parameters are summarized in the second column of Table 2. Patients with a KCCQ-os score of 50 to 74 (low- to moderate-risk group) had a 27% higher risk of an event (95% CI, −1% to 63%) than those with a score of ≥75. Patients with scores between 25 and 50 (the moderate- to high-risk group) had a 2-fold-higher risk of an event (hazard ratio, 2.01; 95% CI, 1.53 to 2.64; overall P<0.001), and those with a KCCQ-os of <25 had a >3-fold-higher risk of an event (hazard ratio, 3.26; 95% CI, 2.14 to 4.97; overall P<0.001). No significant differences in crude outcomes were observed by country (overall P=0.23), nor was there any significant difference in the association between KCCQ-os and outcomes across countries (for KCCQ-by country-interaction, P=0.82).
The results of multivariable analyses examining the relative hazard of cardiovascular death or hospitalization and including all variables except the KCCQ-os are summarized in the third column of Table 2. Predictors retained in the multivariable model include age, post-MI ejection fraction, resting heart rate, β-blocker use, NYHA class, nonwhite race, and the presence of diabetes, prior HF or MI, and COPD. Atrial fibrillation, hypertension, and prior stroke or TIA were not found to be significant independent predictors of outcome. No significant differences in outcomes by country were observed in the multivariable model (P=0.10).
When 4-week KCCQ-os values were added to this multivariable model as a categorical variable (column 4 of Table 2), KCCQ-os remained an independent predictor of cardiovascular events (hazard ratio, 1.37; 95% CI, 1.00 to 1.87 for KCCQ-os between 25 and 50; and hazard ratio, 2.02; 95% CI, 1.24 to 3.27 for KCCQ-os <25; overall P=0.021; c statistic for combined model, 0.70). In this model, NYHA class was no longer independently associated with outcome (overall P=0.2), likely reflecting the high degree of correlation between the KCCQ and NYHA class.3 Of note, the hazard ratios of the other variables did not change significantly, suggesting that the information conferred by the KCCQ about prognosis is distinct and incremental to these other factors. When treated as a continuous variable, KCCQ-os remained highly significant (hazard ratio, 1.097; 95% CI, 1.038 to 1.159) per 10-point decline; P=0.0009), and the hazard ratios of the other variables did not change significantly.
Similar results were obtained for all-cause mortality; the 4-week KCCQ-os remained an independent predictor of cardiovascular events (hazard ratio, 1.29; 95% CI, 0.89 to 1.87 for KCCQ-os between 50 and 75; hazard ratio, 1.65; 95% CI, 1.10 to 2.50 for KCCQ-os between 25 and 50; and hazard ratio, 2.04; 95% CI, 1.12 to 3.72 for KCCQ-os <25; overall P=0.039). Entered as a continuous (linear) variable, hazard for all-cause mortality increased 11.2% (95% CI, 4.1% to 18.9%) per 10-point decline in KCCQ-os (P=0.002).
This study demonstrates that health status, as measured by the KCCQ, is strongly associated with subsequent cardiovascular death or hospitalization in patients with HF after an acute MI. Kaplan-Meier analyses using the KCCQ-os at the 1-, 3- or 6-month assessment (the Figure) consistently demonstrate an increased hazard associated with progressively lower ranges of KCCQ-os. After adjustment for other well-established prognostic variables, including age, LV ejection fraction, revascularization status, presence of renal dysfunction, COPD, and resting heart rate, KCCQ-os remained strongly associated with outcome. In both univariate and multivariable analyses, only prior history of HF or MI was associated with a higher risk of a cardiovascular event. These results concerning its prognostic significance support the utility of the KCCQ in the outpatient setting by providing an additional framework for interpreting results and by assisting in the risk stratification of HF patients.
More than 80 clinical measures have been proposed for risk stratifying HF patients to target the intensity of treatment.7 Strategies such as various echocardiographic, electrophysiological, hemodynamic, biochemical, and exercise/functional determinants have demonstrated various degrees of prognostic utility. However, the clinical value of many of these risk stratification techniques is often limited by the need for specialized and sometimes costly testing, with results usually not immediately available to the treating physicians. In contrast, health status can be a powerful predictor of survival and can be assessed at an outpatient visit.
The NYHA classification, which integrates an assessment of symptoms and physical function, is the best-known and most commonly used measure. Despite its prognostic ability, however, this scale has several limitations, including significant interobserver variability, the ability to capture only a limited range of health status information, and application from a physician’s perspective instead of the that of the patient. Data on the prognostic value of other disease-specific health status instruments are limited. The EPICAL Study Investigators demonstrated an association between Minnesota Living with Heart Failure Questionnaire (MLHFQ) scores and hospital-free survival in a small sample of 108 patients with severe HF (NYHA class III/IV)8; a similar association between MLHFQ scores and event-free survival has been reported in another small cohort (n=96) of patients.9 The largest published study to date comes from the SOLVD Study investigators, who demonstrated associations between various domains of the SOLVD Health Quality of Life Questionnaire and subsequent mortality or hospitalization in 5025 patients with predominantly class I/II HF.10 However, results were reported in terms of risk ratios as a function of changes in score by 1 SD across each of the domains, making clinical interpretation and application difficult. In contrast, the results presented here suggest that the KCCQ can be used as a simple, inexpensive, risk-free, self-administered health status measure that can efficiently stratify risk in patients using an easily interpretable scoring system.
A potential limitation of the present study is that it enrolled only patients with ischemic heart disease and LV systolic dysfunction and specifically excluded patients with other possible causes of their HF. It should be noted that as a group with clinical signs and symptoms of HF after a recent MI, the patients enrolled in this study represent a high-risk population. However, it is unlikely that the prognostic significance of the KCCQ is applicable only to HF of ischemic origin because prior work has established the validity of the KCCQ in patients with HF regardless of the underlying HF cause,3,11 and there is no a priori reason to expect that the prognostic value of the KCCQ would vary significantly as a function of etiology. Of note, other parameters in the multivariable model, such as prior MI, might differ in their prognostic importance.
Although there may be little effect of other causes for LV dysfunction on the prognostic value of the KCCQ, there are currently no data on the prognostic significance of the KCCQ in patients with HF and preserved LV function. This relationship needs to be studied in future populations. Finally, the generalizability of these results outside a clinical trial setting may be limited. It is possible that EPHESUS patients received more attentive follow-up and better treatment (eg, 74.4% of patients received β-blockers) than would occur in a typical outpatient clinical practice setting. Although one would not expect these differences to have a significant effect on the relative risk of cardiovascular death or outcome as a function of KCCQ-os, there may be differences in the absolute risks observed herein and those that would be observed in clinical practice. Further studies are needed in a large outpatient cohort outside a clinical trial setting to address this issue.
The number of hospitalizations caused by HF has increased by 165% over the past 2 decades, with a similar increase seen in HF-associated mortality.12 With the incidence of HF now approaching 10 per 1000 among persons >65 years of age, physicians need efficient mechanisms for monitoring the health status of their patients and effective risk stratification tools to target healthcare resources to those at highest risk. The results of this study support the interpretation of KCCQ scores and the use of this instrument as an important prognostic tool in outpatients with HF. Future investigations should explicitly test the clinical utility of using patient-centered measures of health status to improve the care and outcome of HF patients.13
The EPHESUS trial was funded by Pharmacia (now Pfizer). Dr Spertus owns the copyright to the KCCQ. We gratefully acknowledge the steering committees and the clinical sites participating in the EPHESUS Trial that collected the data on which these analyses were conducted.
Physician Consortium for Performance Improvement. Clinical performance measures in heart failure: a consensus document from the ACC, AHA and the Consortium. Available at: http://www.ama-assn.org/go/quality. Accessed February 1, 2004.
R Development Core Team. R: a language and environment for statistical computing. Available at: http://www.R-project.org. Accessed February 1, 2004.
Hülsmann M, Berger R, Sturm B, et al. Prediction of outcome by neurohumoral activation, the six-minute walk test and the Minnesota Living with Heart Failure Questionnaire in an outpatient cohort with congestive heart failure. Eur Heart J. 2002; 23: 886–891.
Spertus J, Conard M, Rinaldi J, et al. The Kansas City Cardiomyopathy Questionnaire is sensitive to clinical change in congestive heart failure. J Am Coll Cardiol. 2002; 39: 460A. Abstract.
Heart Disease and Stroke Statistics—2003 Update. Dallas, Tex: American Heart Association; 2002.