Cardiac Resynchronization Therapy in Women: US Food and Drug Administration Meta-analysis of Patient-Level Data

Importance  Women were underrepresented in cardiac resynchronization therapy (CRT) trials for heart failure (making up about 20% of enrollees). Combining individual patient data from multiple clinical trials would enable assessment of CRT benefit in women.

Objective  To evaluate whether women with left bundle branch block (LBBB) benefit from CRT-defibrillators (CRT-D) at a shorter QRS duration than men with LBBB do.

Design, Setting, and Participants  Individual patient data were pooled from 3 CRT-D vs implantable cardioverter defibrillator (ICD) trials (4076 patients) enrolling predominantly patients with New York Heart Association (NYHA) class II heart failure and follow-up to 3 years. The effect of CRT-D compared with ICD on outcomes was assessed using random effects Cox proportional hazards.

Main Outcomes and Measures  Time to heart failure event or death (primary) and death alone (secondary).

Results  Women benefited from CRT-D more than men. The main difference occurred in patients with LBBB and a QRS of 130 to 149 milliseconds. In this group, women had a 76% reduction in heart failure or death (absolute CRT-D to ICD difference, 23%; hazard ratio [HR], 0.24, [95% CI, 0.11-0.53]; P < .001) and a 76% reduction in death alone (absolute difference 9%; HR, 0.24, [95% CI, 0.06-0.89]; P = .03), while there was no significant benefit in men for heart failure or death (absolute difference 4%; HR, 0.85 [95% CI, 0.60-1.21]; P = .38) or death alone (absolute difference 2%; HR, 0.86 [95% CI, 0.49-1.52]; P = .60). Neither women nor men with LBBB benefited from CRT-D at QRS shorter than 130 milliseconds, while both sexes with LBBB benefited at QRS of 150 milliseconds or longer.

Conclusions and Relevance  In this population of patients with primarily mild heart failure, women with LBBB benefited from CRT-D at a shorter QRS duration than men with LBBB. This is important because recent guidelines limit the class I indication for CRT-D to patients with LBBB and QRS of 150 milliseconds or longer. While guidelines do give a class IIa indication to patients with LBBB and a QRS of 120 to 149 milliseconds, the present findings are important to communicate because women are less likely to receive CRT-D than men are. This study exemplifies the potential public health and regulatory science value of combining data from multiple clinical trials submitted to the FDA.

As recently outlined by the US Food and Drug Administration (FDA)¹ in the Federal Register, regulatory science research that combines clinical data submitted to FDA has the potential to generate new knowledge and facilitate innovation. One area where this can have value is in determining the safety and efficacy of medical products in patient subgroups that were underrepresented in individual clinical trials, a critical area identified in the 2012 FDA Safety and Innovation Act.²

Cardiac resynchronization therapy (CRT) is a heart failure therapy that improves heart failure symptoms, decreases hospitalizations, and reduces mortality.^3-7 Some studies have shown that women may benefit more than men from CRT^8-11; however, women were underrepresented in CRT trials (making up only about 20% of enrollees), as has been true for other devices,¹² making it difficult to thoroughly assess sex differences. Recent study-level meta-analyses of CRT trials demonstrated that benefit from CRT may be limited to a more restrictive patient population (specifically, patients with a left bundle branch block [LBBB] and QRS ≥150 milliseconds) than that enrolled in the original clinical trials.^13,14 Consequently, 2012 professional society guidelines limited the class I indication for CRT to patients with LBBB and QRS of 150 milliseconds or longer.¹⁵ However, the study-level meta-analyses were unable to assess sex differences because of lack of individual-patient data.

Heart size and QRS duration are generally smaller in women than in men,¹⁶ and recent work has suggested that sex-specific QRS duration criteria for LBBB better predict CRT response.^17,18 We pooled individual-patient data from 3 large CRT-defibrillator (CRT-D) vs implantable cardioverter defibrillator (ICD) trials enrolling predominantly patients with mild heart failure to test the hypothesis that women with LBBB defined by conventional electrocardiographic (ECG) criteria benefit from CRT-D at a shorter QRS duration than men with LBBB do.

This study was approved by the FDA Research in Human Subjects Committee. Informed consent was obtained from patients in the original trials. The inclusion criteria for this meta-analysis required that the included study be a randomized clinical trial comparing CRT-D vs ICD in primarily patients with mild heart failure (New York Heart Association [NYHA] class II), that it report heart failure and mortality outcomes, and that individual-patient data from the study had been submitted to the FDA as a part of a premarket approval application (PMA). Randomized clinical trials performed in patients with moderate to severe heart failure (NYHA class III or IV)^6,7 were not included because the FDA indications for CRT in these patients differ from those in patients with mild heart failure. In addition, the trials that included more severe heart failure were older trials, and the participants in their control groups received only optimal medical therapy, not an ICD as in the more recent trials. Furthermore, the FDA does not have all patient-level data from the older trials in NYHA class III and IV heart failure.

Three trials met the criteria for inclusion in this meta-analysis: the Multicenter Automatic Defibrillator Implantation Trial with Cardiac Resynchronization Therapy (MADIT-CRT),³ the Resynchronization-Defibrillation for Ambulatory Heart Failure Trial (RAFT),⁴ and the Resynchronization Reverses Remodeling in Systolic left Ventricular Dysfunction Trial (REVERSE)⁵ (Table 1). The data from these trials were presented and discussed at 2 FDA panel meetings.^19,20 For the purpose of the present analysis, it was assumed that data obtained from different manufacturers’ CRT-D systems were poolable. Potential differences between the sponsors’ devices were not considered.

The current analysis included all patients enrolled in MADIT-CRT (n = 1820), all patients in RAFT without a pacemaker at baseline (n = 1663), and all patients from REVERSE with data on QRS morphology (n = 593). We combined the key patient characteristics from all trials (Table 2). The presence of LBBB was defined as reported in each trial using conventional ECG criteria. Time to heart failure event or death (primary) and time to death alone (secondary) were the defined end points. For MADIT-CRT, the definition for heart failure as specified in the MADIT-CRT primary end point was used: heart failure event responsive to oral or intravascular decongestive therapy on an inpatient or outpatient basis.³ For RAFT and REVERSE, heart failure event was defined as heart failure leading to hospitalization.^4,5 Because the 3 trials had different median follow-up times (MADIT-CRT, 2.2 years; RAFT, 4.7 years; and REVERSE, 1.1 years) and the number of patients in subgroups significantly decreased during follow-up, all follow-up was censored at 3 years. Of note, while US patients in REVERSE were only followed up for 1 year, non-US patients were followed up for 2 years,²¹ and the full length of follow-up for REVERSE patients was used.

To address potential differences of CRT-D effect across trials, we used mixed-effects Cox proportional hazards analysis, with and without sex-by-treatment interactions, to calculate the risk for specified end points in the overall population and in subgroups by sex, LBBB, QRS duration, and heart failure etiology (ischemic vs nonischemic). Random effects trial intercept models were used to combine data from all trials in a 1-stage individual-patient data meta-analysis. Cumulative survival curves were created by the Kaplan-Meier method. Patients with LBBB were divided into groups defined by 10-millisecond QRS duration intervals. Analysis was repeated using the QRS groups 120 to 129, 130 to 149, and 150 milliseconds or longer, as QRS durations of 120, 130, and 150 milliseconds have been the thresholds used in FDA indications and/or professional society guidelines for CRT-D. In addition, we performed multivariable analysis, adjusting for ischemic etiology, atrial fibrillation and/or flutter, and cardiac medications; we modeled QRS duration in LBBB as a continuous variable with smoothed splines. For patients without LBBB, analysis by sex was performed in 2 QRS groups: shorter than 150 milliseconds and 150 milliseconds or longer. All statistical analyses were performed using the “coxme package” (version 2.2.3) for R (version 3.0.0) (additional details provided in the Supplement). Ninety-five percent CIs are reported for all hazard ratios (HRs), and 2-sided P<.05 was considered significant for interaction analyses.

Of the 4076 patients included in this pooled analysis, 3198 (78%) were men, and 878 (22%) were women (Table 2). Women were more likely than men to have LBBB (85% vs 68%) and less likely to have ischemic cardiomyopathy (33% vs 67%). The majority of both women and men had NYHA class II heart failure (87% and 82%, respectively).

In women, CRT-D resulted in a 60% relative reduction in heart failure or death (CRT-D to ICD HR, 0.40; absolute difference, 15%) and 55% relative reduction in death alone (absolute difference, 6%) compared with only 26% and 15% relative reductions (absolute differences, 7% and 2%) in the 2 end points, respectively, in men (Figure 1). Subgroup and interaction analysis revealed that there were significant differences in the efficacy of CRT-D by sex in patients with LBBB, QRS shorter than 150 milliseconds, and nonischemic heart failure etiology (Figure 1). Of note, there was no benefit from CRT-D in patients without LBBB, women or men (Figure 1), regardless of QRS duration (eFigure 1 in the Supplement), although the CIs in women were wide.

All patients with LBBB were divided into groups based on 10-millisecond QRS duration intervals (eTable 2 in the Supplement). There was no difference in outcomes between the CRT-D and ICD groups in either women or men with LBBB at QRS durations of 120 to 129 milliseconds. However, in women with LBBB, there was an 85% relative reduction in heart failure event or death at QRS of 130 to 139 milliseconds and a 69% relative reduction with QRS of 140 to 149 milliseconds. In men, there was no difference between CRT-D and ICD in either of these QRS duration groups. Above 150 milliseconds, CRT-D benefited both women and men with LBBB.

Figure 2 shows Kaplan-Meier graphs for patients with LBBB in the QRS duration range of 130 to 149 milliseconds. Women had a 76% relative reduction in heart failure or death (absolute difference, 23%; HR, 0.24 [95% CI, 0.11-0.53]) and a 76% relative reduction in death alone (absolute difference, 9%; HR, 0.24 [95% CI, 0.06-0.89]) from CRT-D. In contrast, in men with LBBB and QRS duration of 130 to 149 milliseconds, CRT-D did not have a significant effect on heart failure or death (absolute difference, 4%; HR, 0.85 [95% CI, 0.60-1.21]) or death alone (absolute difference, 2%; HR, 0.86 [95% CI, 0.49-1.52]). Interaction analysis in LBBB with QRS of 130 to 149 milliseconds revealed that the difference in treatment efficacy between women and men was significant for the end point of heart failure or death (P = .003) but not for death alone (P = .10).

eFigure 2 and eFigure 3 in the Supplement show Kaplan-Meier curves for women and men with LBBB in the other QRS duration categories (120-129 and ≥150 milliseconds). With LBBB and QRS of 150 milliseconds or longer, both women and men had a significant reduction in heart failure or death (women HR, 0.33 [95% CI, 0.21-0.52], absolute difference, 16%; men HR, 0.47 [95% CI, 0.37-0.59], absolute difference, 14%) and death alone (women HR, 0.36 [95% CI, 0.16-0.82], absolute difference, 5%; men HR, 0.65 [95% CI, 0.47-0.91], absolute difference, 4%). eFigure 4 in the Supplement shows spline curves for QRS duration modeled as a continuous variable in women and men with LBBB.

Adjustment for ischemic etiology, atrial fibrillation and/or flutter, and cardiac medications did not change the results for the QRS groups 130 to 149 milliseconds and 150 milliseconds or longer, while the number of patients in the QRS group 120 to 129 milliseconds was too small for adjustment (Figure 3). eFigure 5 in the Supplement shows the results excluding patients with NYHA class III heart failure (8% of patients overall). The results for the primary end point did not change. For the secondary end point of death alone, the HR point estimates remained almost identical, but CIs widened and crossed HR of 1 in women with LBBB and QRS of 130 to 149 milliseconds and men with LBBB and QRS of 150 milliseconds or longer. In addition, no significant interaction between CRT-D treatment and the individual trials was detected.

The primary finding of this individual-patient data analysis of 3 large randomized CRT-D trials enrolling primarily patients with NYHA class II heart failure is that women with LBBB by conventional ECG criteria derive significant benefit from CRT-D at QRS durations shorter than 150 milliseconds (specifically, ≥130 milliseconds), while men with conventional LBBB derive significant benefit at QRS of 150 milliseconds or longer. These findings are true for both end points of heart failure or death and death alone. This is important because recent professional society guidelines for CRT-D only assign a class I indication to patients with LBBB and QRS of 150 milliseconds or longer.¹⁵ These indications were based on study-level meta-analyses of published trials for which approximately 80% of enrollees were men, and it was not possible to perform robust interaction and subgroup analysis by sex and clinical characteristics due to lack of individual-patient data. Of note, guidelines do give a class IIa indication to patients with LBBB and QRS of 120 to 149 milliseconds,¹⁵ and thus many of these patients would likely be offered CRT-D. However, considering that women receive CRT-D less often than men,²² we believe that the current findings are important to communicate.

The question of the appropriate QRS duration threshold for selecting CRT patients was discussed extensively at the most recent FDA public advisory committee meeting on CRT-D for expanding indications to patients with NYHA class II heart failure.²⁰ The FDA ultimately approved CRT-D for these patients with LBBB, QRS of 130 milliseconds or longer, and ejection fraction of 30% or less. The indication for Boston Scientific also includes patients with ischemic NYHA class I heart failure. The results of the present analysis are not necessarily in conflict with FDA indications because the reason for the lack of CRT-D effect in men with QRS durations of 130 to 149 milliseconds may be that most of these patients diagnosed as having LBBB by conventional ECG criteria do not have a true LBBB.^18,22 Men with a QRS shorter than 150 milliseconds and true LBBB may benefit. The FDA indications for NYHA class III and IV heart failure do not include a requirement for QRS morphology. However, because this study included primarily patients with NYHA class II heart failure (only 8% had class III heart failure), the results cannot be used to make conclusions about patients with NYHA class III or IV heart failure.

The findings confirm prior analyses by Arshad et al¹⁰ and Zareba et al²³ from MADIT-CRT demonstrating that women, but not men, benefited from CRT-D at QRS shorter than 150 milliseconds. However, the prior MADIT-CRT analyses did not divide patients by the presence of LBBB and QRS duration, and thus the greater benefit in women with QRS shorter than 150 milliseconds could have been explained by a higher incidence of LBBB in women. The larger sample size of patients with QRS shorter than 150 milliseconds (1209 men and 297 women) in the present analysis allowed us to investigate the combination of LBBB presence and QRS duration in 10-millisecond intervals. Interestingly, a different individual-patient data meta-analysis that included significantly more patients with NYHA class III heart failure found that benefit from CRT became significant at a QRS of 140 milliseconds, while findings of interaction analysis by sex were not significant.²⁴

The hypothesis of this study was based on the premise that patients with true, complete LBBB benefit from CRT-D. With complete LBBB, activation of the interventricular septum and left ventricular lateral wall is uncoupled, resulting in about a 100-millisecond delay between initial activation of the 2 walls. In contrast, in normal conduction or incomplete LBBB conduction delay, most of the left ventricle is activated synchronously via the rapidly conducting His-Purkinje system. The beneficial effect of CRT-D in LBBB likely derives from attenuating the dyssynchronous contraction caused by the large activation delay, while in other cases, CRT-D can overdrive nature’s rapid activation of the left ventricle. The lack of CRT benefit in patients without LBBB has also been observed in recent studies.^{13,22,23,25,26}

In the present analysis, LBBB was defined by the conventional criteria used in the trials. While for MADIT-CRT, the World Health Organization criteria for LBBB were used,²³ in RAFT and REVERSE, the exact definition of LBBB was not specified. In the current study, women had a higher percentage of LBBB than men (85% vs 69%). However, the difference is likely even greater because men have longer QRS durations than women¹⁶ and are more likely to have a false-positive LBBB diagnosis.¹⁸ New sex-specific strict LBBB criteria were proposed that require a QRS of 130 milliseconds or longer in women and 140 milliseconds or longer in men, along with mid-QRS notching and/or slurring.¹⁸ Recent single-center studies demonstrated that patients not meeting strict LBBB criteria had a 4-fold higher rate of heart failure hospitalization or death and did not respond to CRT-D compared with patients who met the strict LBBB criteria.^17,27 The current study seems to support the use of sex-specific criteria for LBBB. Based on the strict LBBB criteria, one might expect men with QRS durations of 140 to 149 milliseconds to derive benefit; however, the present study did not involve analysis of mid-QRS notching and/or slurring. In addition, while there was no benefit for women or men with LBBB and a QRS of 120 to 129 milliseconds, the number of patients in this group was too small to make definitive statements about CRT-D efficacy.

The fact that men were more likely than women to have ischemic cardiomyopathy and atrial fibrillation, which are both associated with a worse prognosis,²⁸ might have contributed to a greater benefit from CRT-D in women. While controlling for these variables did not affect the results, it is difficult to differentiate the effect of LBBB vs nonischemic cardiomyopathy because, as demonstrated by a recent cardiac magnetic resonance study,²⁹ complete LBBB in patients referred for ICD and CRT-D is most commonly caused by nonischemic cardiomyopathies.

The effectiveness and safety of medical products such as drugs, devices, and biologics can differ between women and men due to differences in prevalence of disease, physiology, body size, and a plethora of other intrinsic and extrinsic factors.³⁰ The FDA released a draft guidance on evaluating sex differences in device trials to improve the quality and consistency of available data regarding the performance of medical devices in both sexes. The draft guidance document discusses the importance of ensuring that representation by sex is consistent with disease prevalence and that data from studies are appropriately analyzed for sex differences.³¹ Pooling individual-patient data from multiple clinical trials in a specific product area provides an additional powerful tool to analyze sex differences.

Data submitted to the FDA in PMAs has confidentiality protections, including that FDA cannot disclose receipt of PMAs until a decision is made or reveal trade secrets and commercial or financial information.³² However, after the FDA issues an order approving or denying approval of any PMA, summary safety and effectiveness data are available for public disclosure.³³ An individual-patient data meta-analysis such as this one is a summary of safety and effectiveness data and is a logical mechanism for reporting safety and effectiveness in patient subgroups that are underrepresented in individual trials. Other examples of FDA regulatory science with drug trials include individual-patient data meta-analyses in patients with hepatitis C³⁴ and human immunodeficiency virus.³⁵

While the present analysis was performed by the FDA, it is not possible for the FDA to conduct all research of this nature. Currently, nonsummary safety and efficacy data from marketing applications are not available to researchers outside the FDA. However, the FDA recently requested public comment on the “Availability of Masked and De-identified Non-Summary Safety and Efficacy Data.”¹ The posting indicates that making data sets available to non-FDA experts for regulatory science research could further facilitate innovation in the development and evaluation of medical products and maximize benefit to society that patients provide by participating in clinical trials. Separately, industry,³⁶ academic consortiums,³⁷ and other medical product regulatory agencies³⁸ have proposed other “open data” initiatives.

A limitation of this study is that it is a post hoc analysis of the included clinical trials, and multiple comparisons were performed. In addition, the results could have been influenced by different follow-up time between trials and larger size of MADIT-CRT³ and RAFT⁴ compared with REVERSE.⁵ To partially address this, follow-up was censored at 3 years and time-to-event analysis was performed incorporating random effects by trial. It should be noted that differences between trials are a limitation of prior study-level meta-analyses. The number of patients in the LBBB and 120 to 129 milliseconds QRS duration group was small, and this particular analysis might therefore be underpowered to detect significant results. In addition, findings related to mortality should be interpreted with caution because of the low mortality rate in these patients with mild heart failure symptoms. Two large older trials including patients with more severe heart failure were not included in this analysis. These trials differed significantly from the included trials in that they did not include an ICD control arm, enrolled predominantly patients with NYHA class III heart failure,^6,7 and in 1 trial required the presence of mechanical dyssynchrony in patients with QRS of 120 to 149 milliseconds.⁷ The findings from the current study cannot be extended to more severe heart failure.

In summary, in an individual-patient data meta-analysis of 3 major clinical CRT trials primarily limited to patients with mild heart failure symptoms (NYHA class II), women were found to benefit from CRT-D at a shorter QRS duration than men. While current guidelines only give a class I indication for CRT-D to patients with LBBB and QRS of 150 milliseconds or longer, this analysis found that women with LBBB and QRS of 130 to 149 milliseconds have a 76% reduction in heart failure events and mortality from CRT-D. While professional society guidelines do give a class IIa indication for these patients, and thus most women in this group are likely be offered CRT-D, these findings are important to communicate because women are less likely to receive CRT-D than men. The fact that women normally have smaller ventricles and shorter QRS duration than men provides an anatomical and/or physiological explanation for the findings, but the higher rate of nonischemic cardiomyopathy in women compared with men may have also contributed. Overall, this study highlights the importance of sex-specific analysis in medical device clinical studies and the public health value of combining individual-patient data from clinical trials submitted to the FDA.

Corresponding Author: David G. Strauss, MD, PhD, Center for Devices and Radiological Health, US Food and Drug Administration, 10903 New Hampshire Ave, 62-1126, Silver Spring, MD 20993 (david.strauss@fda.hhs.gov).

Accepted for Publication: January 26, 2014.

Published Online: June 23, 2014. doi:10.1001/jamainternmed.2014.2717.

Author Contributions: Drs Zusterzeel and Strauss 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: Zusterzeel, Selzman, Caños, O’Callaghan, Piña, Strauss.

Acquisition, analysis, or interpretation of data: Zusterzeel, Selzman, Sanders, Caños, O’Callaghan, Carpenter, Piña, Strauss.

Drafting of the manuscript: Zusterzeel, Strauss.

Critical revision of the manuscript for important intellectual content: Zusterzeel, Selzman, Sanders, Caños, O’Callaghan, Carpenter, Piña, Strauss.

Statistical analysis: Zusterzeel, Caños, Carpenter, Strauss.

Obtained funding: O’Callaghan, Strauss.

Administrative, technical, or material support: Zusterzeel, Caños.

Study supervision: Strauss.

Conflict of Interest Disclosures: None reported.

Funding/Support: This project was supported in part by the FDA Office of Women’s Health and by a research fellowship from the Oak Ridge Institute for Science and Education through an interagency agreement between the US Department of Energy and the FDA.

Role of the Sponsors: The funders 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.

Disclaimer: The mention of commercial products, their sources, or their use in connection with material reported herein is not to be construed as either an actual or implied endorsement of such products by the Department of Health and Human Services.