The MADIT-CRT (Multicenter Automatic Defibrillator Implantation Trial With Cardiac Resynchronization Therapy) showed the combined cardiac resynchronization therapy (CRT) and implantable cardioverter-defibrillator (ICD), compared with ICD only, had a 34% reduction in risk of death from any cause or a nonfatal heart failure event among patients with mild heart failure (New York Heart Association [NYHA] functional classes I and II). This reduction was primarily driven by a 41% reduction in the risk of heart failure events (1). Moreover, Zareba et al. (2) and Goldenberg et al. (3) reported that CRT-ICD therapy was primarily beneficial among those with a left bundle branch block (LBBB) conduction disturbance.

In this paper, we address whether adding CRT to ICD comes at the expense of a decrease in quality of life, or whether it provides an improvement among patients with mild heart failure (NYHA functional classes I and II). We assess changes in quality of life among those with and without LBBB conduction disturbance.

Detailed information about the MADIT-CRT study design, randomization, recruitment, and outcome has been published ((1),4). The MADIT-CRT trial enrolled 1,820 patients at 110 centers in 14 countries (1,271 patients in the United States) from December 22, 2004, through April 23, 2008; the trial was stopped on June 22, 2009. Patients enrolled in the study had ischemic cardiomyopathy (NYHA functional class I or II) or nonischemic cardiomyopathy (NYHA functional class II only), sinus rhythm, an ejection fraction of 30% or less, and prolonged intraventricular conduction with a QRS duration of 130 ms or more.

The analyses presented here were based on 1,699 patients, which comprise the subset of the original 1,820 patients who had baseline observations and at least 1 additional observation on our outcome variables up to the close of the study.

To assess quality of life, we used 6 basic scales and 3 summary scales of the Kansas City Cardiomyopathy Questionnaire (KCCQ) instrument that comprises heart failure–specific quality-of-life–related measures ((5),(6),(7),(8),(9),(10),11). The 6 basic scales included the Symptom Stability, Symptom Frequency, Symptom Burden, Physical Limitation, Quality of Life, and Social Limitation scales. The 3 summary scales included the Total Symptom, Clinical Summary, and Overall Summary scores. See the Online Appendix for a discussion of these scales.

We used feasible generalized least squares (12) to estimate models of the KCCQ measures in which all variables are centered on hospital means, thereby eliminating hospital fixed effects, and multiplied by heteroskedasticity weights to account for the different numbers of subjects at each hospital.

The data were slightly different in baseline blood pressure across the arms; therefore, to reduce error variance, we adjusted analyses for baseline systolic and diastolic blood pressure levels.

For each KCCQ scale, we estimated 2 models: Model 1 tested the difference in the change from baseline in the CRT-ICD group compared with the change from baseline in the ICD-only group on the KCCQ scales across the 4.5 years of the study period. Model 2 tested whether the treatment effect varied across time period quintiles within the 4.5 years of the study period: joint tests were used of the interaction terms between indicators of the time quintiles and the indicator of CRT-ICD group.

We repeated the analyses of Models 1 and 2 on the LBBB and non-LBBB subgroups. We investigated whether the differences in effects were statistically different between the LBBB and non-LBBB subgroups in Model 1 by testing the interaction of the LBBB indicator and the CRT-ICD group indicator using data that included all patients.

See the Online Appendix for our investigation of whether date of enrollment, dying or being lost to follow-up, or the patient switching device during the study impacted the KCCQ scores differentially across study arms.

(Table 1) shows baseline characteristics and baseline KCCQ scores of the study cohort by treatment arm and LBBB status.

(Figure 54_gr1) shows the estimated difference in effects between the CRT-ICD and ICD-only groups, and the corresponding 95% confidence intervals, from Model 1 for each KCCQ measure, adjusted for baseline systolic and diastolic blood pressure levels. A positive value means the CRT-ICD group had a larger effect (i.e., change from baseline) in KCCQ scores than the ICD-only group. The first panel depicts estimates for the entire study population. The second panel depicts the estimates for the LBBB subgroup. The third panel depicts estimates for the non-LBBB subgroup, which were not significant for each measure. The interactions between the LBBB indicator and ICD-CRT group indicator, estimated using all patients, were significantly different between the LBBB and non-LBBB subgroups for all KCCQ measures (p < 0.05) except for the Symptom Severity and Symptom Burden scales (p = 0.91 and p = 0.06, respectively).

The analysis of Model 2, using the entire study population, resulted in rejecting the hypothesis that the treatment effect does not vary by time for each KCCQ measure (significant at the 0.05 level), except for the Physical Limitation, Symptom Frequency, and Symptom Stability scales, which were not significant. The significant findings on these joint tests suggested that a difference in effect exists for at least 1 of the time quintiles. However, we cannot rule out that at least 1 time quintile had no difference (the joint test of the hypothesis that at least 1 parameter was 0 was insignificant for each measure—all tests had p values >0.49). The LBBB subgroup analysis of Model 2 resulted in in the same inferences, but results of Model 2 applied to the non-LBBB subgroup were not significant for each measure.

(Figure 54_gr2) shows the mean Overall Summary score of the KCCQ among the LBBB subgroup for each 6-month period between enrollment and Year 3. (Table 2) shows the sample sizes for each of these time periods by treatment group. Similar to the preceding analyses, these differences based on the non-LBBB subgroup were not statistically significant, whereas those based on the full sample fell in between.

The results of this study provide evidence that CRT-ICD is associated with better improvement in heart failure–specific quality of life than ICD-only among minimally symptomatic patients. Specifically, better effects were found among the CRT-ICD group with respect to all KCCQ measures we evaluated. These findings were statistically evident among patients with LBBB but not discernible among those without this conduction disturbance. Evidence suggests that the difference between groups varies across time since enrollment, particularly during the first 2 years.

The KCCQ scores range from 0 through 100 points. The magnitudes of treatment-assignment effects on changes in the KCCQ measures reported here are on the order of approximately 1 to 3 points. Spertus et al. (13) found that differences of 5 points on KCCQ Overall Summary score correspond to changes in patient heart failure status that were judged by cardiologists to be small in terms of clinical changes. Changes in patient clinical heart failure status judged to be large by cardiologists corresponded to changes in KCCQ Overall Summary scores of between 20 and 25. Flynn et al. (14) found a 5-point change in the KCCQ overall score was associated with a 2.50-ml/kg change in peak Vo₂ and a 112-m change in 6-min walking distance; they judged these changes to be clinically meaningful. What is not clear, however, is how patients, rather than cardiologists and researchers, would deem the difference in life experiences, rather than clinical indicators, that correspond to the differences in effects that we observed. This is an important factor for providers assisting patients to attain the patients' goals, particularly for patients who value the quality of their life experiences (15).

There is insufficient previous research on minimally symptomatic patients (i.e., research restricted to NYHA functional class I and II) to provide a direct comparison for our results. The aforementioned studies by Spertus et al. (13) and Flynn et al. (14) evaluated sicker patient groups that include NYHA functional classifications greater than II. This is true of other studies that evaluate KCCQ measures as well. For example, Myers et al. (16), Sullivan et al. (16), Ekman et al. (17), Eurich et al. (6), Heidenreich et al. (18), and Soto et al. (19), each found associations between KCCQ measures and relevant outcomes, but each were assessing a sicker patient population (including NHYA functional classes III and/or IV). Because the MADIT-CRT patient population is minimally symptomatic, with baseline KCCQ scores starting in the upper half of the scale range, changes in KCCQ measures are expected to be lower than among patients with worse initial scores (20). Consequently, any quality-of-life improvement is notable, but additional research is necessary to identify the extent that the effect sizes found in our study impact the lives of patients in this population and thereby impact treatment decisions and satisfaction.

Notwithstanding the need for future research to identify the importance to patients of the effects identified in this study, the improvement found across all KCCQ measures suggests that the clinical benefits of resynchronization therapy (1), as an adjunct to the ICD, do not come at the expense of quality of life—a finding that may be reassuring to patients. Instead, the CRT-ICD is associated with a greater increase in quality of life than ICD-only among the minimally symptomatic patients, who might otherwise have been presumed to not garner a discernible improvement.