Author + information
- Received February 2, 2010
- Revision received April 20, 2010
- Accepted May 13, 2010
- Published online August 31, 2010.
- Adam S. Fein, MD*,
- Yongfei Wang, MS†,
- Jeptha P. Curtis, MD†,
- Frederick A. Masoudi, MD, MSPH‡∥,
- Paul D. Varosy, MD§,
- Matthew R. Reynolds, MD, MSc*,* (, )
- National Cardiovascular Data Registry
- ↵*Reprint requests and correspondence:
Dr. Matthew R. Reynolds, Beth Israel Deaconess Medical Center, 185 Pilgrim Road, Boston, Massachusetts 02217
Objectives The purpose of the study was to define the extent and nature of cardiac resynchronization therapy (CRT) device usage outside consensus guidelines using national data.
Background Recent literature has shown that the application of CRT in clinical practice frequently does not adhere to evidence-based consensus guidelines. Factors underlying these practices have not been fully explored.
Methods From the National Cardiovascular Data Registry's Implantable Cardiac-Defibrillator Registry, we defined a cohort of 45,392 cardiac resynchronization therapy-defibrillator (CRT-D) implants between January 2006 and June 2008 with a primary prevention indication. We defined “off-label” implants as those in which the ejection fraction was >35%, the New York Heart Association functional class was below III, or the QRS interval duration was <120 ms in the absence of a documented need for ventricular pacing. The relationships between patient, implanting physician, and hospital characteristics with off-label use were explored with multivariable hierarchical logistic regression models.
Results Overall, 23.7% of devices were placed without meeting all 3 implant criteria, most often due to New York Heart Association functional class below III (13.1% of implants) or QRS interval duration <120 ms (12.0%). Atrial fibrillation/flutter, previous percutaneous coronary intervention, and the performance of an electrophysiology study before implant were independently associated with increased odds of off-label use, whereas diabetes mellitus, increasing age, and female sex were associated with decreased odds. Physician training and insurance payer were weakly associated with the likelihood of off-label use.
Conclusions Nearly 1 in 4 patients receiving CRT devices in the study time frame did not meet guideline-based indications. Given the evolving evidence base supporting the use of CRT, these practices require careful scrutiny.
- cardiac resynchronization therapy
- ICD registry
- guideline adherence
- cardiac pacing utilization
Randomized trials have shown that cardiac resynchronization therapy (CRT) in carefully selected patients leads to improved cardiac function, symptoms, quality of life, and survival (1–4). In 2005, the American College of Cardiology/American Heart Association strongly endorsed CRT for patients with a left ventricular ejection fraction (LVEF) ≤35%, QRS interval duration ≥120 ms, and New York Heart Association (NYHA) functional class III or IV heart failure despite optimal drug therapy (Class I recommendation; Level of Evidence: A) (5). These recommendations have subsequently been further strengthened and refined (6,7). Adoption of CRT has been rapid: cardiac resynchronization therapy-defibrillator (CRT-D) systems recently comprised 40% of all implantable cardiac-defibrillator (ICD) implants in the U.S. (8).
While the question of whether patients outside these initially approved treatment criteria could still derive benefit from CRT is a source of continued investigation, these guidelines remain the standard of care (9). However, recent studies suggest that the application of CRT in clinical practice frequently does not adhere to evidence-based consensus guideline recommendations (10,11). To further understand current CRT utilization patterns, we sought to define the extent and nature of CRT implantation outside consensus guidelines using data from the National Cardiovascular Data Registry (NCDR) ICD Registry, and investigate the patient- and provider-related factors associated with “off-label” device usage.
Analyses in this study used data from the NCDR's ICD Registry, a Centers for Medicare and Medicaid Services mandated national, ongoing, prospective, observational database developed in collaboration with American College of Cardiology Foundation and the Heart Rhythm Society. In January 2005, a Centers for Medicare and Medicaid Services national coverage decision stipulated that all Medicare beneficiaries receiving an ICD for primary prevention of sudden cardiac death be enrolled in a national registry. The NCDR's ICD Registry was established later in 2005 and became the sole repository for data collection in April 2006. Although hospitals are not required to submit data for non-Medicare patients, >75% of hospitals participating in the registry have entered data on all ICD implantations regardless of indication or insurance (12). Clinical, demographic, and procedural information for each implant procedure are collected on >130 standardized data elements. Data are submitted by participating hospitals on a quarterly basis using American College of Cardiology-certified software. Data analysis is performed by an independent academic analytic center.
The population of interest was ICD with CRT (CRT-D) device implantations in ICD Registry patients for the primary prevention of sudden cardiac death between January 2006 and June 2008. From the 105,543 CRT-D implanted devices enrolled from 1,300 hospitals in the ICD Registry, we excluded devices placed in patients who had a pacemaker, previous ICD, or any clinical characteristic that would have warranted an ICD for secondary prevention (history of cardiac arrest, sustained ventricular tachycardia, or syncope), leaving 45,776 implants. An additional 384 patients were excluded because of incomplete data on the key variables used to define the use of off-label devices (QRS, ejection fraction [EF], and NYHA functional class). A study cohort of 45,392 implants was identified for analysis, as shown in Figure 1.
The primary outcome measure was the proportion of patients who underwent CRT device placement without meeting the criteria of consensus guidelines that were current during the study period (5,7). We defined such off-label usage as device implantation in patients with any of the following: 1) documented LVEF >35%; 2) NYHA functional class below III; or 3) a QRS interval duration <120 ms without a need for ventricular pacing. Patients with atrial fibrillation (AF) who otherwise met standard CRT implant criteria were considered “on label” (Class IIa recommendation). Further, as contemporaneous consensus guidelines considered CRT reasonable (Class IIa) in patients with a QRS interval duration of <120 ms but expected or actual frequent reliance on ventricular pacing (7), we considered patients with a QRS <120 to still be on label when there was documentation of a significant atrioventricular (AV) conduction abnormality (second- or third-degree AV block or PR interval >300 ms). Although guidelines also advise “optimal medical therapy” as an additional pre-condition for CRT implantation, we did not believe that this criterion was practical to apply in the registry setting. A secondary analysis was performed from within the off-label group to determine the prevalence of more extreme deviations from guideline recommendations, identifying the frequency of CRT implantation in patients with NYHA functional class I, a QRS interval <100 ms, or LVEF >35%.
The relationship of various patient-, physician-, and hospital-related factors with off-label CRT-D implantation was explored using univariate and multivariable analyses. First, the cohort was divided into off- and on-label groups as described above. These groups were compared using chi-square tests for categorical variables and ttests for continuous variables. Variables in this analysis included patient demographics (age, sex, race, insurance type); reason for hospitalization; patient comorbidities and cardiac risk factors; selected baseline electrocardiographic and laboratory findings; physician training; and hospital characteristics (e.g., ownership, size, teaching status, and location).
The independent association of key variables of interest (age, sex, race, insurance status, physician training) as well as other variables associated with the use of off-label devices in univariate comparisons was assessed using hierarchical logistic regression modeling for the use of off-label devices, to account for the clustering of patients among hospitals. A backward stepwise selection algorithm was used in the model, and only variables with a p value <0.05 were kept. Variables used to define off- versus on-label implantation were not considered in the models.
We calculated the proportion of CRT-D implants considered off-label for each of the 10 calendar quarters during the study period and assessed the possible temporal trends using the Cochran-Armitage trend test.
We also hypothesized that patients who deviated with respect to 1 implant criterion would have more abnormal values of the other criteria than on-label patients (e.g., that among CRT recipients with an LVEF >35%, average QRS duration would be greater than among those with an LVEF ≤35%; or that for CRT-D recipients with a QRS duration of <120 ms, mean LVEF would be lower than among those with a QRS duration ≥120 ms). To assess this hypothesis, we stratified the study cohort by each individual implant criterion (NYHA functional class, QRS duration, and EF) and qualitatively compared the other implant criteria according to these groupings.
Statistical analyses were performed using SAS version 9.1 software (SAS Institute, Cary, North Carolina).
The characteristics of the 45,392 patients who underwent a first-time implantation of a CRT-D who were identified for this study are shown in Table 1.The study population was predominantly (69%) male and had a mean age of 69 years. More than 60% of the cohort had ischemic heart disease, with an average EF of 23%. For the cohort as a whole, 79% were NYHA functional class III and 8% were class IV. Hypertension, diabetes, AF or flutter, and/or prior coronary revascularization were common coexisting conditions. The majority (72%) of the patients were Medicare beneficiaries.
A total of 10,737 (23.7%) of the implants did not meet contemporaneous consensus guideline criteria and were thus considered off label (Fig. 1). This off-label pattern of use was mostly due to device implantation in patients with a NYHA functional class below III (13.1%) and/or a QRS duration <120 ms without a documented need for ventricular pacing (12.0%). Only 1.3% of patients had an LVEF >35%. The CRT-D recipients had a QRS interval ≤100 ms, NYHA functional class I, or LVEF >35% in 8% of cases.
Univariate comparisons of on- versus off-label patient characteristics are also shown in Table 1. Men and nonwhite patients were more likely to have CRT implanted in an off-label fashion than were women and white patients. Off-label CRT recipients were also more likely than on-label patients to have a history of AF/flutter or sinus node dysfunction. Off-label patients were slightly more likely to have had a previous PCI procedure or an electrophysiology (EP) study performed at the time of device implant than on-label patients, and slightly less likely to have diabetes, chronic lung disease, a previous myocardial infarction, or coronary artery bypass graft surgery.
We also examined CRT guideline adherence across hospitals and physicians (Table 2).There were statistically significant but clinically modest differences in measures of implanter's EP training, as well as hospital regional classifications, teaching status, and ownership model. Off-label CRT-D procedures occurred with slightly higher frequency in metropolitan areas and private hospitals, and operators of off-label procedures were slightly less likely to be board certified in EP.
The results of our multivariable analysis are shown in Table 3.Factors associated with increased odds of off-label implantation included hospitalization for noncardiac reasons compared with hospitalization for the procedure, history of previous PCI, the performance of an EP study at the time of device implantation, and higher systolic blood pressure. Of all variables assessed, a history of AF/flutter was most strongly associated with increased odds of off-label CRT-D use (odds ratio: 1.36, 95% confidence interval: 1.29 to 1.43, p < 0.001). Factors significantly associated with a lower adjusted odds of off-label implantation included older age, female sex, nonischemic cardiomyopathy, diabetes, and higher BUN levels. Of note, physician training and insurance payer did show weak statistical associations with guideline adherence. Specifically, surgeons had higher adjusted odds (odds ratio: 1.43, 95% confidence interval: 1.16 to 1.76) of off-label utilization compared with physicians board-certified in EP, while patients with commercial insurance had slightly higher odds of off-label implant than patients with Medicare, and Medicaid patients slightly lower odds. Hospital characteristics such as ownership or teaching status did not correlate with the likelihood of off-label use in multivariate analysis.
In contrast to our hypotheses, we did not find that patients receiving CRT-D devices in off-label fashion for 1 particular reason tended to have greater degrees of abnormality in other implant criteria. For example, patients with a QRS duration of <120 ms or an LVEF >35% were slightly less likely to have NYHA functional class III or IV than were patients with a QRS duration of ≥120 ms (82% vs. 88%) or an LVEF ≤35% (73% vs. 87%), and CRT-D recipients with NYHA functional class I to II had a slightly lower likelihood of baseline left bundle branch block than NYHA functional class III or IV patients (Table 4).
There was no evidence of either rising or declining proportion of off-label implants over time, with the proportion of procedures defined as off-label remaining between 22% and 25% for each quarter we assessed (p for trend = 0.30) (Fig. 2).Although the total number of CRT-D implants appeared to increase substantially after the first quarter of 2006, it should be noted that this was the first quarter of data entry into NCDR's ICD Registry; thus, initial procedure volumes were likely underestimated.
Investigating the use of CRT in clinical practice using a large, national patient registry, our study had 4 main findings. First, nearly 1 in 4 patients in contemporary community practice who received new CRT placement from 2006 to 2008 did so outside current clinical guideline recommendations. Second, this pattern of device utilization was not easily explained by geographic, hospital, or physician factors, although weak associations with insurance status and physician training were observed. Third, guideline nonadherence for CRT implantation varied by patient sex, age, history of AF, comorbid conditions, and a prior pattern of invasive cardiac interventions. Last, there was no significant trend toward increasing or decreasing adherence to guideline-based indications over the time period of this study.
A number of randomized clinical trials have established that CRT improves ventricular function and symptom status while reducing hospitalizations and mortality in a majority of carefully selected heart failure patients (1–4). Analysis of the ICD Registry, however, shows that in routine clinical practice, nearly 1 in 4 CRT-D devices were implanted outside current consensus guidelines, which are based mainly on the entry criteria of clinical trials demonstrating the efficacy of CRT devices. Even allowing for evidence from trials published after the timeframe of our analysis, many of these patients would not be expected to derive the same benefits from CRT as patients identified in previous randomized trials. CRT implantation is an invasive procedure associated with important rates of short- and long-term complications, and is more expensive than standard ICD implantation. Until the effectiveness, cost effectiveness, and safety of CRT in larger populations are demonstrated, the wisdom of widespread off-label CRT utilization should be carefully questioned.
While it is relatively straightforward to classify CRT-D procedures in the Registry as on- or off-label, we would caution against equating off-label care with inappropriate care. For example, our study identified AF/flutter as a factor associated with increased likelihood of off-label device usage. This could be because implanting physicians expect that some patients with AF/flutter and a normal QRS duration who are appropriate ICD candidates will pace frequently in the ventricle because of pauses and bradycardia (e.g., with intensification of beta-blocker therapy). It may be reasonable to add a left ventricle pacing lead in these cases, given the known deleterious effects of right ventricle pacing in ICD patients (13). Alternatively, CRT-D devices may be chosen in conjunction with ablation of the AV junction as a strategy to both treat heart failure and symptoms of AF (14–16). Although such procedures are off-label, we would not necessarily consider them inappropriate.
In addition, guidelines can and do evolve as technologies are studied for new indications, and portions of the clinical community may change their practices in advance of clinical trial evidence and guideline revision. For example, both the MADIT-CRT (Multicenter Automatic Defibrillator Implantation Trial With Cardiac Resynchronization Therapy) (9) and longer-term follow-up data from the REVERSE (Resynchronization Reverses Remodeling in Systolic Left Ventricular Dysfunction) study (17) found that CRT therapy was associated with significant reductions in heart failure events (albeit not mortality) and improvements in cardiac structure and function (9,18) when compared with standard ICDs in patients with reduced EF, prolonged QRS duration, and only NYHA functional class I or II heart failure. As our study period preceded the publication of these results, the appropriateness of CRT-D implants in NYHA functional class I and II patients during this time frame is uncertain. While we understand clinicians' desire to provide patients the “benefit of the doubt,” particularly in borderline cases, had MADIT-CRT been a negative trial, we might in hindsight have concluded that these procedures took on unjustified risk and expense.
Off-label use of medical technology is perhaps most problematic in cases where the expectation of benefit is low, or when there is actually evidence that benefit is absent. We found that a low proportion of CRT-D devices (1.3%) were implanted in patients with an LVEF >35%, but that a fairly high proportion—12.0%—had a QRS duration <120 ms and no documented need for ventricular pacing (5.6% had a QRS duration <100 ms). Although the use of CRT in patients with echocardiographic evidence of dyssynchrony even with a narrow QRS showed initial promise in nonrandomized studies (19–21), the RETHINQ (Cardiac Resynchronization Therapy in Patients With Heart Failure and Narrow QRS) trial (22) reported a lack of benefit from CRT in patients with a QRS of <120 ms, and the PROSPECT study (23) raised serious doubts about the reproducibility of echocardiographic measures of dyssynchrony as a means of selecting patients for CRT. Despite these negative findings, the implantation of CRT-D devices in patients with narrow QRS complexes appears to have continued.
We find the high frequency of off-label CRT utilization observed in our study potentially concerning for several reasons. First, although major complication rates from CRT procedures are acceptably low in experienced hands, a previous analysis from the ICD Registry has confirmed that the probability of an implant complication is more than twice as high with a CRT-D procedure than with a single-chamber implant (24). Similar results were recently reported from an Ontario registry (25). Implantation of a CRT-D system is also at least $6,000 to $10,000 more expensive than implantation of a single-chamber system. Finally, CRT is well known to have a nonresponse rate of up to 30% even for guideline-recommended indications. For all of these reasons, implantation of CRT-D systems in the context of a low probability of incremental clinical benefit may increase risk for patients, costs for the health care system, and erode the cost-effectiveness of the therapy. In our opinion, the resources expended in at least some of these off-label procedures would be better directed at addressing the documented underuse of CRT in more appropriate candidates (10).
Two previous studies examined the patterns of CRT-D usage and reported on factors associated with off-label use. Using data from the Get With The Guidelines-Heart Failure (GWTG-HF) program, Piccini et al. (10) documented that 9.6% of patients receiving new CRT-Ds had an LVEF >35%, a figure far higher than what we observed in the ICD Registry. These results could have differed for several reasons: the studies covered slightly different timeframes, the GWTG study had far fewer new CRT-D recipients (n = 811) than our study, and the GWTG-HF program only involves patients hospitalized for a heart failure exacerbation at selected centers. Whereas Piccini et al. (10) also demonstrated regional and hospital variation in the use of CRT technology for potentially appropriate candidates, we did not find significant associations with off-label CRT-D use and either geographical region or hospital characteristics.
Contrary to a previous analysis of ICD Registry data (11), our study did not show an association between race and CRT-D therapy use outside published guidelines. An explanation for the discrepancy between these 2 studies is also unclear. Our study had a slightly more recent time frame (2006 to 2008 vs. 2005 to 2007), differed in the definition of off-label device usage (by accepting AV block as on-label), and used different covariates in multivariate modeling. Both the prior ICD Registry study and the GWTG study found that the utilization of CRT for potentially appropriate nonwhite candidates was lower than for white patients.
We found significantly reduced odds of off-label device utilization with increasing age, female sex, diabetes, and elevated blood urea nitrogen levels. A few of these factors have been associated with increased risk of procedural complications (26) and reduced benefit with prophylactic ICD implantation (27). These associations suggest that clinicians may be less likely to extend off-label CRT-D therapy to patients for whom the likelihood of benefit appears reduced and/or the risk of complications appears increased.
Although we did not identify physician or hospital factors predictive of off-label CRT-D usage, we did find increased odds of off-label implants in patients with prior PCI or EP study. Health services literature that has explored the factors associated with medical technology overuse has found a direct relationship with the availability of supply (28). The association between off-label CRT-D implantation and a history of other invasive cardiac procedures factors may, therefore, reflect physician or hospital characteristics not captured in the variables we assessed. It is also possible that information obtained during an EP study (e.g., abnormalities in AV conduction) may influence the choice of device in some cases.
In multivariate analysis, we did find a weak association between guideline adherence and insurer, with Medicaid patients slightly less likely to receive off-label CRT, and commercial insurance slightly more likely, compared with Medicare. All other things being equal, the patient's insurance may therefore be a minor consideration when these decisions are made. We also observed somewhat stronger associations between guideline adherence and physician training, indicating that implanters without EP training of some kind (including surgeons) may be more likely to use a CRT-D in an off-label fashion. A previous registry analysis found a higher rate of early complications when ICDs were implanted by surgeons (24).
While the registry includes a substantial number of data elements, it may still lack sufficient clinical detail to judge the appropriateness of device utilization in many cases. Another limitation common to all ICD registry analyses is that the registry has no systemic approach to auditing the data. Although many consistency checks were instituted in the data collection process, the overall accuracy of the data is not known. Additionally, the registry does not easily facilitate interpretation of a patient's drug regimen, as it lacks complete information on drug doses or intolerance, did not specifically include aldosterone antagonists, and is limited to discharge medications after device implantation. With recently published data indicating substantial underutilization of beta-blockers before ICD implantation (29), we would hypothesize that incorporation of an “optimal medical therapy” criterion into our definition of off-label treatment would only increase the prevalence of guideline nonadherence. Finally, the large sample sizes involved in analyses like this one increase the potential for both type I (false positive) statistical error and the identification of true associations that are clinically unimportant.
From the ICD Registry experience, the use of cardiac resynchronization defibrillators in clinical practice is frequently noncompliant with evidence-based consensus guidelines. Nearly 1 in 4 devices placed from 2006 to 2008 did not conform to contemporaneous guidelines, and many of these might presently be considered inappropriate even when allowing for potential future guideline changes. This pattern of practice is not easily explained by geographic, hospital, or physician factors. Implantation of a CRT-D (vs. standard ICD) in patients with a low probability of incremental benefit should be discouraged.
Dr. Curtis has received research support from the National Cardiovascular Data Registryand has equity interest in Medtronic. Dr. Masoudi has contracts with the American College of Cardiology and Oklahoma Foundation for Medical Quality, and research funding from the NHLBI, AHRQ, ACCF, and Heart Rhythm Society. Dr. Varosy has received consultant income from Boston Scientific. All other authors report that they have no relationships to disclose.
- Abbreviations and Acronyms
- atrial fibrillation
- cardiac resynchronization therapy
- ejection fraction
- implantable cardiac-defibrillator
- left ventricular ejection fraction
- New York Heart Association
- percutaneous coronary intervention
- Received February 2, 2010.
- Revision received April 20, 2010.
- Accepted May 13, 2010.
- American College of Cardiology Foundation
- Hunt S.A.,
- Abraham W.T.,
- Chin M.H.,
- et al.
- Jessup M.,
- Abraham W.T.,
- Casey D.E.,
- et al.
- Epstein A.E.,
- DiMarco J.P.,
- Ellenbogen K.A.,
- et al.
- Piccini J.P.,
- Hernandez A.F.,
- Dai D.,
- et al.
- Gasparini M.,
- Auricchio A.,
- Metra M.,
- et al.
- Gasparini M.,
- Auricchio A.,
- Regoli F.,
- et al.
- Daubert C.,
- Gold M.R.,
- Abraham W.T.,
- et al.
- St. John Sutton M.,
- Ghio S.,
- Plappert T.,
- et al.
- Achilli A.,
- Sassara M.,
- Ficili S.,
- et al.
- Yu C.M.,
- Chan Y.S.,
- Zhang Q.,
- et al.
- Bleeker G.B.,
- Holman E.R.,
- Steendijk P.,
- et al.
- Chung E.S.,
- Leon A.R.,
- Tavazzi L.,
- et al.
- Lee D.S.,
- Krahn A.D.,
- Healey J.S.,
- et al.
- Al-Khatib S.M.,
- Greiner M.A.,
- Peterson E.D.,
- Hernandez A.F.,
- Schulman K.A.,
- Curtis L.H.
- Goldenberg I.,
- Vyas A.K.,
- Hall W.J.,
- et al.
- Lucas F.L.,
- DeLorenzo M.A.,
- Siewers A.E.,
- Wennberg D.E.
- Hauptman P.J.,
- Swindle J.P.,
- Masoudi F.A.,
- Burroughs T.E.