Timing of cardiac resynchronization therapy device implantation in heart failure patients and its association with outcomes

Background When used in appropriately selected heart failure (HF) patients, cardiac resynchronization therapy (CRT) reduces mortality and hospitalization. It is not understood whether CRT implantation during hospitalization for HF is associated with similar benefits. Hypothesis Timing of CRT implantation relative to hospitalization for HF is associated with clinical outcomes. Methods This analysis included patients eligible for CRT and discharged alive between January 2005 and December 2012 from 388 hospitals in Get With The Guidelines‐HF. Participants were linked with Centers for Medicare and Medicaid Services data to evaluate outcomes of all‐cause mortality and HF re‐hospitalization based on CRT status (present on admission, placed during hospitalization, and prescribed at discharge; reference = no CRT). Results Of 15 619 CRT‐eligible HF patients, 2408 (15%) had CRT on admission, 1269 (8%) underwent CRT implantation during hospitalization and 643 (4%) had CRT prescribed at discharge. Compared with patients without CRT, mortality was lower in those who received CRT implantation during HF hospitalization (adjusted hazard ratio [HR] 0.63; P < 0.0001) and those prescribed CRT at discharge (adjusted HR 0.78; P = 0.048). A reduction in HF re‐hospitalization was observed in patients with CRT implanted during hospitalization (adjusted HR 0.64; P < 0.0001), but not in those who were prescribed CRT at discharge (adjusted HR 1.02; P = 0.77). Conclusion CRT implantation during HF hospitalization was associated with lower rates of mortality and HF re‐hospitalization. These data suggest that a CRT utilization strategy that does not delay implantation to the post‐discharge period may be appropriate. Randomized data are needed to definitively identify optimal timing of CRT implantation.


| INTRODUCTION
Heart failure (HF) currently affects an estimated 6.5 million adults in the United States, with prevalence continuing to rise. 1,2 When used in appropriately selected HF patients, cardiac resynchronization therapy (CRT) is associated with improved left ventricular Abbreviations: CI, confidence interval; CMS, Centers for Medicare and Medicaid Services; CRT, cardiac resynchronization therapy; GWTG, get with the guidelines; HF, heart failure; HR, hazard ratio; LVEF, left ventricular ejection fraction; ms, milliseconds. function, lower mortality and HF hospitalizations, and improved quality of life. [3][4][5][6] CRT implantation is recommended in current national guidelines (Class I) for patients with symptomatic HF, left ventricular ejection fraction (LVEF) ≤ 35%, sinus rhythm, and left bundle branch block with QRS duration ≥150 ms with Class IIA recommendations for those with left bundle branch block and QRS duration 120 to 149 ms or with non-left bundle branch block and QRS ≥ 150 ms. [7][8][9] Landmark trials showing the benefit of CRT do not specify timing of implantation relative to hospitalization for HF. 4,10 It is therefore not well understood whether CRT implantation during hospitalization for HF is associated with similar improvements in clinical outcomes compared with CRT implantation in an elective setting. CRT implantation during an HR hospitalization could provide benefit by targeting patients at highest risk for adverse outcomes after discharge. 11 Alternatively, patients hospitalized for HF are subject to appreciable clinical instability and medication adjustment, which could attenuate the potential benefit of immediate CRT implantation. This analysis examines whether the timing of CRT implantation relative to admission for HF is associated with differences in all-cause mortality and HF re-hospitalization.

| Data sources
Data for this analysis were obtained from the get with the guidelines-heart failure (GWTG-HF) database as well as the Centers for Medicare and Medicaid Services (CMS) fee-for-service inpatient claims. The GWTG-HF database has previously been described. [12][13][14] Briefly, GWTG is a quality improvement program that collects clinical information from participating hospitals regarding the in-hospital care and outcomes of consecutive patients hospitalized for coronary artery disease, stroke or HF. Data quality is monitored through frequent reports that review the completeness and accuracy of submitted data. Only sites with a high degree of completeness, defined as having medical history data for at least 75% of their patients, were included in this analysis. Individual patients with missing data for a key variable (age, sex, race, ejection fraction, and medical history) were also excluded. Prior reports have demonstrated that patient characteristics and health outcomes of Medicare beneficiaries with HF who are enrolled in the GWTG-HF registry are similar to Medicare beneficiaries in the United States who are hospitalized for HF but are not enrolled. 15,16 All participating institutions are required to comply with local regulatory and privacy guidelines and to submit the GWTG-HF protocol for review and approval by their institutional review board. Data are primarily used at the local site for quality improvement initiatives, and were therefore granted a waiver of informed consent under the common rule.

| Outcomes
Outcomes of interest were all-cause mortality and HF re-hospitalization, defined as a hospital admission with any of the following primary ICD-9-CM diagnosis codes: 428.x, 402.x1, 404.x1, and 404.x3.
Patients who were not reported to have died were censored on December 31, 2012 or the date at which they were no longer enrolled in Part A Fee-for-Service Medicare, whichever occurred first.

| Statistical analysis
For all analyses, patients were grouped by timing of CRT implantation: prior to HF admission and during hospitalization and as CRT prescribed at discharge, or no CRT. Baseline characteristics of the 4 CRT groups are summarized as medians (25th and 75th percentiles) for continuous variables and as percentages (frequencies) for categorical variables. Mortality rates are summarized using Kaplan-Meier rates.
To compare mortality risk between groups, Cox proportional hazard models were used. Cumulative incidence rates and Fine and Gray models, which consider death as a competing risk, were used to compare hospital re-hospitalization rates between groups.
The interaction between time and CRT group was tested in each model, and all models were adjusted; covariates are listed at Table 2.
Covariates were chosen prior to analysis to best account for differences between groups that may have affected the outcomes of inter-  Rates of guideline-recommended medical therapy were high in all groups. Abbreviations: CI, confidence interval; CRT, cardiac resynchronization therapy; HR, hazard ratio. a Event rates are Kaplan-Meier rates. All P-values and hazard ratios are from adjusted Cox models. Follow-up duration shown as median (IQR). Models contain the following covariates: age, gender, race, left ventricular ejection fraction, systolic blood pressure, heart rate, body mass index, medical history (anemia, prior atrial arrhythmia, prior cerebrovascular accident or transient ischemic attack, ischemic heart disease, pulmonary disease, depression, diabetes, hypertension, hyperlipidemia, peripheral artery disease, renal insufficiency, chronic dialysis, smoking, prior revascularization), medications at discharge (angiotensin converting enzyme-inhibitor or angiotensin II receptor blocker, beta blocker, aldosterone antagonist, anticoagulant), and hospital characteristics (geographic region, rural location, teaching hospital, number of beds, and whether the hospital performs heart transplants). Missing values for covariates were imputed using multiple imputations (25 iterations).   Table S1).  Figure 2. A sensitivity analysis confining the no-CRT comparator group to only those with documented prolonged QRS complexes revealed similar hazard ratios (Table S2).

| DISCUSSION
This analysis utilizing the GWTG-HF registry has several notable findings: (a) CRT implantation during HF hospitalization is associated with improvements in risk-adjusted mortality and HF re-hospitalization compared with no CRT, (b) CRT prescribed at discharge is associated with improved risk-adjusted mortality compared with no CRT, but this effect is not as large as that seen in CRT implanted during HF hospitalization, (c) CRT prescribed at discharge is not associated with improved HF re-hospitalization rates compared with no CRT, and (iv) pre-existing CRT at the time of HF hospitalization is associated with worse clinical outcomes compared with no CRT.
Current guidelines do not address timing of CRT implantation relative to admission for HF. Our analysis shows that in eligible patients admitted with a primary diagnosis of HF, CRT implantation prior to discharge is associated with improved mortality and HF hospitalization rates compared to those who do not receive CRT. In current practice, it is uncommon for CRT to be implanted during HF hospitalization, a fact that is supported in our patient cohort, as only 8% of patients without CRT already in place received the therapy prior to discharge.
These results suggest that it is appropriate to consider CRT implantation in patients admitted for HF prior to discharge.
The present study also shows that patients who are prescribed CRT at discharge with a plan for subsequent placement is associated with only modest clinical benefit compared to those who receive CRT during HF hospitalization. Although it is unknown whether patients in this group received post-discharge CRT as prescribed, this represents real-world practice making the results clinically applicable. Given the known clinical benefit of CRT, long-term follow-up to further investigate rates of CRT implantation when prescribed at discharge could highlight important areas for improvement in discharge transitions and planning. Additionally, prospective, randomized trials aimed at comparison of clinical outcomes between patients who receive CRT during admission vs after discharge could further our understanding of appropriate timing of device implantation.
In our analysis, patients who require hospitalization for HF with pre-existing CRT have higher mortality and HF re-hospitalization rates than other groups studied. Based on differences in baseline characteristics highlighted in Table 1, this group represents a patient cohort with more cardiovascular and non-cardiovascular comorbidities.
Despite adjusting for the majority of these differences, worsened clinical outcomes persisted. This suggests the presence of unmeasured clinical attributes, likely including markers of HF severity or duration of disease, which may account for the recurrent hospitalizations and worsened mortality. Prior studies and subsequent guidelines have focused on indications for CRT implantation, but have not identified FIGURE 1 Relationships between time of CRT implantation and mortality-the risk of death is lower for CRT implantation during hospitalization for HF vs CRT in place at admission or CRT prescription at discharge. These relationships remain consistent over time.CRT, cardiac resynchronizatio therapy; HF, heart failure; HR, hazard ratio; CI, confidence interval markers of HF disease severity that may be associated with lack of clinical benefit or worsened outcomes with CRT use. Our results suggest that a better understanding of specific factors that may affect certain patients' ability to receive the intended clinical benefits of CRT is needed.

| Strengths and limitations
The major strength of our analysis is that it utilized data collected as part of a large, national registry. Because of the large number of subjects included, the results are statistically robust with enhanced data consistency. In addition, registry analysis provides a diverse patient population not typically included in clinical trials.
A major limitation of our study is that patients were not ran-  Abbreviations: CI, confidence interval; CRT, cardiac resynchronization therapy; HR, hazard ratio. a Event rates are cumulative incidence rates. All P-values and hazard ratios are from adjusted Fine and Gray models. Models contain the following covariates: age, gender, race, left ventricular ejection fraction, systolic blood pressure, heart rate, body mass index, medical history (anemia, prior atrial arrhythmia, prior cerebrovascular accident or transient ischemic attack, ischemic heart disease, pulmonary disease, depression, diabetes, hypertension, hyperlipidemia, peripheral artery disease, renal insufficiency, chronic dialysis, smoking, prior revascularization), medications at discharge (angiotensin converting enzyme-inhibitor or angiotensin II receptor blocker, beta blocker, aldosterone antagonist, anticoagulant), and hospital characteristics (geographic region, rural location, teaching hospital, number of beds, and whether the hospital performs heart transplants). Missing values for covariates were imputed using multiple imputations (25 iterations).
that the patients included represent an unbiased sample from the larger population.

| CONCLUSION
Landmark trials and current clinical practice guidelines do not address optimal timing of CRT implantation relative to hospitalization for HF. The primary finding of this analysis is that CRT implantation during HF hospitalization is associated with lower all-cause mortality and reduced HF re-hospitalization rates compared with patients who do not receive CRT. Clinical benefit is greater when CRT is implanted during HF hospitalization compared with when it is prescribed at discharge. These data suggest that CRT utilization strategy that does not delay implantation to the post-discharge period following HF hospitalization may be appropriate. Patients who require hospitalization for HF despite having CRT already in place had worse outcomes compared with patients with no CRT. This observation is likely explained by these patients representing a particularly high-risk subgroup. Randomized trials should be aimed at identifying the appropriate timing of CRT implantation relative to hospitalization for HF, as well as identification of specific characteristics that may affect patients' ability to derive the intended benefits of CRT.

SUPPORTING INFORMATION
Additional supporting information may be found online in the Supporting Information section at the end of the article.