Meta-analysis: Cardiac Resynchronization Therapy for Patients With Less Symptomatic Heart Failure

and infections in 1.4%. Limitation: Subgroup analyses were underpowered and lack data for persons with NYHA class I symptoms, atrial fibrillation, chronic kidney disease, or right bundle branch block. Conclusion: Cardiac resynchronization therapy is beneficial for patients with reduced left ventricular ejection fraction, symptoms of heart failure, and prolonged QRS, regardless of NYHA class.

H eart failure is a common disorder, affecting approximately 2.5% of adults in North America and Europe (1,2). Heart failure substantially reduces quality of life and has high morbidity (with frequent emergency department visits and heart failure hospitalizations) and mortality rates, which create a great economic burden even when patients receive optimal treatment (1,(3)(4)(5)(6)(7). In a previous systematic review of 4420 patients in 14 trials (7), McAlister and colleagues demonstrated a 22% relative risk reduction in all-cause mortality and a 37% relative risk reduction in heart failure hospitalization when cardiac resynchronization therapy (CRT) was added to optimal medical therapy. International guidelines recommend CRT for patients with left ventricular ejection fraction (LVEF) of 0.35 or less, New York Heart Association (NYHA) class III or IV symptoms despite medical treatment, wide QRS duration (Ͼ120 ms), and sinus rhythm (2, 8 -10).
However, important questions remain regarding heart failure and CRT. First, because nearly all participants (91%) in the randomized, controlled trials (RCTs) identified in the previous systematic review had NYHA class III or IV symptoms (7), the effect of CRT in patients with less severe symptoms is unclear. Three RCTs (11)(12)(13) assessing the efficacy of CRT in patients with less severe heart failure symptoms have been published since the previous systematic review (7), and recently the European Society of Cardiology extended its recommendation for CRT to include patients with mildly symptomatic heart failure who have QRS duration of 150 ms or more (14). Second, patients with a narrow QRS duration and severe heart failure symptoms are not considered candidates for CRT, but mechanical and electrical dyssynchrony do not always coexist, raising questions about whether these patients may benefit from CRT (15,16). Finally, pacing with a left ventricular lead (without placement of a concomitant right ventricular lead) may provide the same benefit as a 3-lead CRT device (17).
In this systematic review, we update the previous systematic review (7) and explore the benefits and harms of CRT in patients with less symptomatic heart failure, patients with a narrow QRS duration on electrocardiography, and the use of a left ventricular lead alone versus standard CRT.

Data Sources and Searches
We updated and followed the protocol used for the previous systematic review (7). This included electronic literature searches supplemented by hand-searching reference lists of included studies and review articles, proceedings booklets from meetings, U.S. Food and Drug Administration reports, and contact with primary study authors and device manufacturers (Appendix Table 1, available at www.annals.org) (7). The search was not limited to studies published in English or to publication status. The search was last updated on 20 December 2010. Appendix Table 2 (available at www.annals.org) shows the MEDLINE search strategy.

Study Selection
We included RCTs that 1) enrolled patients with heart failure and LVEF of 0.40 or less, regardless of their baseline NYHA functional class; 2) compared CRT with inactive pacing, right ventricular pacing alone, left ventricular pacing alone, implantable cardioverter-defibrillator (ICD) alone (for trials of CRT plus ICD vs. ICD alone), or usual care; 3) reported all-cause mortality, heart failure hospitalization, change in LVEF, or change in functional outcomes (NYHA class, quality of life, or 6-minute walk test); and 4) included more than 25 participants.
The primary literature search was done by 1 of the authors. Using standardized inclusion or exclusion forms, 2 of the authors then independently reviewed the full texts of all potentially relevant studies. Final decisions about study inclusion or exclusion were reached by consensus.

Data Extraction and Quality Assessment
Data extraction was done by 2 independent reviewers by using standardized data extraction forms. For crossover trials, data from the first period only (before crossover) were used. Quality assessment of all included studies was done by using the 6 domains of the Cochrane tool for assessing risk for bias (18).

Primary and Secondary Outcomes
The primary outcome for this systematic review is allcause mortality. Secondary outcomes include heart failure hospitalizations, quality of life, and functional outcomes (LVEF and 6-minute walk test). Because we expected duration of follow-up to differ among trials, we explored whether the risk ratios (RRs) for the primary outcome varied by duration of follow-up.

Subgroups and Sensitivity Analysis
A priori, we assessed the efficacy of CRT among studies that included patients with NYHA class I or II symptoms compared with NYHA class III or IV symptoms as a separate subgroup analysis; trials were classified as having patients who were predominantly (Ͼ50% but Ͻ100%) or exclusively (100%) in one NYHA subgroup or the other. Other prespecified subgroups were sex, age, ischemic etiology, QRS duration, year of enrollment, and whether patients received an ICD. Left ventricular lead-only pacing

Context
Guidelines recommend cardiac resynchronization therapy (CRT) for patients with reduced left ventricular ejection fraction and advanced symptoms of heart failure.

Contribution
This meta-analysis of 25 trials includes new evidence that CRT reduces mortality and heart failure hospitalizations in patients with left ventricular systolic dysfunction, prolonged QRS duration, and milder symptoms. The relative magnitude of the benefits in patients with milder symptoms seemed to be similar to those in patients with New York Heart Association class III or IV symptoms.

Caution
Few trial participants had atrial fibrillation or asymptomatic (New York Heart Association class I) heart failure.

Implication
Some patients with reduced left ventricular ejection fraction and mild symptoms may benefit from CRT.

-The Editors
Review CRT in Patients With Less Symptomatic Heart Failure trials versus biventricular lead trials were evaluated separately.

Statistical Analysis
For dichotomous outcomes (mortality and heart failure hospitalization), RRs and 95% CIs were calculated. For continuous outcomes (such as the 6-minute walk test and quality-of-life scores), weighted mean differences (WMDs) and 95% CIs were calculated. Intention-to-treat analyses were performed by using the same end point definitions as in the primary studies. We included results from primary study reports and not from their extended follow-up analyses, although these were reviewed for consistency of results. When reported, the components of a primary outcome were analyzed separately.
Because we expected studies to differ in length of follow-up and study participants, we decided a priori to use a DerSimonian-Laird random-effects model for all outcomes (18). The I 2 statistic was used to quantify heterogeneity; a value greater than 50% was considered to indicate substantial heterogeneity (19).
Meta-regressions were run to explore potential sources of heterogeneity among studies. The studies were weighted by size and variance and regressed against year of publication, age, sex, percentage of patients with key baseline characteristics of interest (ischemia, atrial fibrillation, and left bundle branch block), percentage in each NYHA class, mean QRS duration, and background ICD use. We examined the effect of duration of follow-up on the RR for all-cause mortality by using an additional meta-regression model.
Review Manager, version 4.2 (Cochrane Collaboration, Copenhagen, Denmark), was used to generate the forest plots and unadjusted RRs; meta-regression and other analyses were done by using R, version 2.12 (R Foundation for Statistical Computing, Vienna, Austria), using the metafor command (20).

Role of the Funding Source
The study was not supported by external funding.

Study Selection and Evaluation
The primary literature search yielded 3964 studies (Figure 1). Of these, 11 RCTs (11-13, 17, 21-27) met the inclusion criteria and were added to the 14 trials (28 -41) from the previous systematic review (7). All of the newly included trials were published, except for Greater-EARTH (27) (for expansions of all study names, see the Glossary). Greater-EARTH was presented at the 2010 Heart Rhythm Society meeting and was included because the principal investigator provided us with the unpublished data for this review. Additional data and clarifications were provided by the principal investigators of another 5 trials.

All-Cause Mortality
Pooled data from all 25 trials show that CRT reduced all-cause mortality by 19% (RR, 0.81 [95% CI, 0.72 to 0.90]); there was no appreciable statistical heterogeneity among trials (I 2 ϭ 0%). Excluding trials without events in 1 or both groups did not affect mortality estimates (RR, 0.80 [CI, 0.72 to 0.89]). In the 6 trials that predominantly included patients with NYHA class I or II symptoms, CRT reduced the risk for all-cause mortality (RR, 0.83 [CI, 0.72 to 0.96]; I 2 ϭ 0%) (Figure 2 Four studies compared CRT with left ventricular pacing: Two included patients with NYHA class III or IV symptoms (17,25); 1 included patients with NYHA class II, III, or IV symptoms (22); and 1 included patients with NYHA class I, II, or III symptoms (27). Left ventricular pacing alone did not affect all-cause mortality compared with CRT (RR, 0.83 [CI, 0.32 to 2.13]; I 2 ϭ 27%), although the number of events was small (28 deaths in 677 patients).
Because the trials had different durations of follow-up (ranging from 1 month to approximately 40 months), we examined the effect of follow-up duration on the RR of all-cause mortality. The RR (approximately 0.80) was constant over time (Appendix Figure, available at www.annals .org).

Cause-Specific Mortality
The mortality benefit of CRT was largely driven by a reduction in heart failure-related mortality in the 12 trials that reported this outcome (218 events

Heart Failure Hospitalization
Overall, CRT was associated with a reduction in the risk for hospitalization with heart failure (RR, 0.69 [CI, 0.58 to 0.82]; I 2 ϭ 50%) (Figure 3); no appreciable difference was found between trials enrolling predominantly patients with NYHA class III or IV symptoms (RR, 0.65 [CI, 0.50 to 0.86]; I 2 ϭ 57%) and those enrolling predominantly patients with NYHA class I or II symptoms (RR, 0.71 [CI, 0.57 to 0.87]; I 2 ϭ 37%), although the absolute rate of heart failure hospitalization was higher in the former trials (22% vs. 17% in the NYHA class I or II trials). Cardiac resynchronization therapy was associated with a reduction in heart failure hospitalization in the 2 studies exclusively of patients with NYHA class I or II symptoms ( Given the degree of statistical heterogeneity in the analyses of heart failure hospitalization, which was not explained by NYHA class at baseline, bivariate metaregression models were used to explore the reasons for statistical heterogeneity. These models demonstrated that the percentage of patients with ischemic heart failure enrolled in the trials explained most of the heterogeneity, because these patients seemed to derive less benefit from heart failure hospitalization than nonischemic patients. Each 5% increase in the percentage of patients with ischemic heart failure in an RCT was associated with an 8% relative reduction (CI, 3.9% to 12.8%) in the benefits of CRT on heart failure hospitalization.

Quality of Life
Quality of life was reported in 15 of the 25 trials. Overall, CRT was associated with an improvement in scores on the Minnesota Living with Heart Failure Questionnaire (MLHFQ) compared with control participants (14 trials, 4283 participants; WMD, 6.56 points [CI, 4.08 to 9.04 points]), but substantial statistical heterogeneity was found (I 2 ϭ 72%) that was largely attributable to symptom status at baseline. Two of the 3 trials (787 participants) including patients with NYHA class I or II symptoms had better MLHFQ scores at baseline (mean scores, 40 [35] and 28 [11]) and did not show any appreciable improvement with CRT (WMD, 1.82 points [CI, Ϫ0.77 to 4.41 points]; I 2 ϭ 0%). The remaining trial in patients with NYHA class I or II symptoms (12, 42) re-

Assessment for Publication Bias
We tested for publication bias by using a funnel plot for all-cause mortality. Although the funnel plot was asymmetrical, the area missing consisted of small positive studies; if anything, this indicates that our estimates of all-cause mortality may be conservative. A funnel plot for heart failure hospitalization was asymmetrical, indicating potential publication bias; the plot was missing small neutral or negative trials.

DISCUSSION
In this systematic review, we confirm that CRT improves LVEF and reduces all-cause mortality and heart failure hospitalization in patients with milder symptoms of heart failure (NYHA class I or II), left ventricular systolic dysfunction, and prolonged QRS duration. The relative magnitude of these benefits (risk reductions of 17% for mortality and 29% for heart failure hospitalization) are similar to that seen in patients with NYHA class III or IV symptoms, left ventricular systolic dysfunction, and prolonged QRS duration. Our findings contrast with those of a recent meta-analysis (43) of 2 trials in patients with NYHA class I or II symptoms (compared with the 6 trials in our analysis) that report no survival benefit with CRT, but a significant reduction in a composite outcome of "any heart failure events." Of note, 98% of the control patients in our analyses of trials including NYHA class I or II symptoms had an ICD; thus, the benefits of CRT that we found represent incremental benefits additional to the expected benefits from the ICD implanted in both groups in each study. However, CRT did not improve quality of life or functional outcomes, such as results of the 6-minute walk test, in patients with mildly symptomatic heart failure-in contrast to their marked beneficial effects on these outcomes (similar in magnitude to those of angiotensin-converting enzyme inhibitors [44]) for patients with NYHA class III or IV symptoms at baseline. This is not surprising, given that patients with NYHA class I or II heart failure have less symptom burden and impairment of quality of life at baseline. The improvements in LVEF that we documented for trial participants regardless of NYHA class are consistent with results from other studies (7,36,45,46). Although data from the REVERSE trial and MADIT-CRT suggested that the benefits of CRT on left ventricular remodeling were greatest in patients with longer QRS durations and nonischemic heart failure (47,48) and a substudy from MIRACLE also suggested greater left ventricular remodeling with CRT in patients with nonischemic disease (46), without access to individual-patient data, we could not explore whether this finding persisted in other trial data sets. Certainly, the benefits of CRT on the composite clinical outcome was greatest in patients in MADIT-CRT and RAFT who had a QRS duration greater than 150 ms. Of note, CRT is the only positive inotropic therapy that has been shown to improve both cardiac systolic function and patient survival.
An important question about CRT, as with any intervention that has been tested in only a selected range of patients and depends on specialized technical expertise to implant, is how generalizable the benefits demonstrated in RCTs will be when the device is used in clinical practice by less experienced clinicians working in smaller-volume centers (49 -51). This is particularly relevant for CRT, because approximately 38% of the patients (18 of the RCTs) in our efficacy analysis were randomly assigned only after successful device implantation. As a result, these RCTs may overestimate the potential benefit from CRT and underestimate the risk, because patients who could not tolerate the procedure or in whom implantation was unsuccessful were not included in the trial data. We anticipate that data from the National Cardiovascular Data Registry and ongoing cohort studies will be vital in establishing the clinical effectiveness and safety of CRT and tracking changes over time as device implanters, the tools for implantation, and the sophistication of the devices change-complication rates for left ventricular lead placement may be higher in the community. Such data will also be important to inform future cost-effectiveness analyses of CRT; current estimates (52, 53) based on analyses using trial data and restricting use of CRT in their models to patients with NYHA class III or IV symptoms will not be applicable as indications for CRT expand.
Although we followed current recommendations for performing a systematic review and obtained unpublished data from several of the primary studies included in our meta-analysis, our study has limitations. Substantial statistical heterogeneity was present in some analyses and could not be explained by the variables considered in the metaregressions; however, subgroup analyses and metaregressions are post hoc analyses and generally underpow-ered. In addition, the conclusions about the implications for clinical practice are limited for some subgroups of patients who were excluded from or underrepresented in the trials: those with bradyarrhythmias, atrial fibrillation, chronic kidney disease, or right bundle branch block. Finally, most of the trial participants were younger and relatively healthier than patients with heart failure encountered in clinical practice.
What are the implications of our findings? Our data support the expansion of indications for CRT to less symptomatic patients with heart failure who have LVEF less than 0.35 and QRS duration greater than 120 ms and are in sinus rhythm (Table). However, 85% of less symptomatic patients in these trials had NYHA II symptoms, and high-quality evidence to support this therapy in patients with asymptomatic left ventricular dysfunction or NYHA class I symptoms is inconclusive.
Our data also illuminate other issues about CRT for which randomized trial evidence is sparse and thereby highlight research priorities. For example, whether CRT is as efficacious in patients with atrial fibrillation (54) as in those with sinus rhythm is unclear (55). This is an important research question for future randomized trials because less than 1% of participants in CRT trials had atrial fibrillation, but almost 30% of all CRT devices are implanted in patients with atrial fibrillation (56,57). Moreover, although preliminary observations (58) suggest that CRT reduces symptom burden in patients with LVEF greater than 0.35, prolonged QRS, and NYHA class III or IV symptoms that are refractory to optimal medical therapy, an RCT is needed before practice recommendations can be made (59). Nonetheless, 10% to 15% of patients who received CRT devices in the United States and Europe have LVEF greater than 0.35 (56,57,60). Finally, the most pressing research priority for CRT should be to establish a uniform definition of "CRT response." A recent review pointed out the poor correlations among the 17 most frequently used definitions for CRT response and the fact that although 99% of the PROSPECT (Predictors of Response to Cardiac Resynchronization Therapy) participants would have been defined as CRT responders by at least 1 of these commonly used criteria, 94% would also have been defined as CRT nonresponders by at least 1 of the criteria (61).
Of note, our meta-regression analysis showed that inclusion of a higher proportion of patients with ischemic heart failure in the RCTs was associated with less benefit from CRT in reducing heart failure hospitalization, but no differential effect on mortality was observed. Studies in patients with NYHA class I or II symptoms (62) and class III or IV symptoms (46) have shown that an ischemic cause of heart failure is associated with less benefit from CRT. Thus, understanding which patients with ischemic heart disease should receive a CRT device, and the roles of scar tissue, wall thinning, limited myocyte viability, and subendocardial ischemia in making this decision, also warrant future research.
It had been estimated that CRT was indicated in fewer than 10% of symptomatic patients with heart failure who have left ventricular systolic dysfunction (63, 64). However, as our systematic review reveals, the evidence base has evolved substantially since these earlier estimates, and CRT may now be indicated for most of the 40% of patients with systolic heart failure who have a QRS duration greater than 120 ms (65). However, more than one third of current CRT recipients do not have functional or echocardiographic improvement after activation of their CRT (7), indicating that relying on RCT eligibility criteria to define which patients should undergo device implantation is im-perfect. As such, we believe establishing criteria for case selection so that CRT devices are preferentially implanted in the patients who are most likely to benefit is of vital importance for researchers, clinicians, and policymakers.