Prophylactic pacemaker use to allow h-blocker therapy in patients with chronic heart failure with bradycardia

Transcription

1 Congestive Heart Failure Prophylactic pacemaker use to allow h-blocker therapy in patients with chronic heart failure with bradycardia Eric C. Stecker, MD, MPH, a A. Mark Fendrick, MD, a Bradley P. Knight, MD, c and Keith D. Aaronson, MD, MS b Ann Arbor, MI; and Chicago, IL Background Although the benefits of h-blocker therapy for patients with congestive heart failure (CHF) are independent of pretreatment heart rate, patients with chronic systolic heart failure and low resting heart rates are often excluded from h-blocker therapy. We investigated the effectiveness and cost-effectiveness of prophylactic pacemaker insertion to facilitate h-blocker use in these patients. Methods A Markov model simulated the natural history of a cohort of clinically stable patients with CHF (ejection fraction V35%, mean age 60 years) with resting heart rates of b68 beat/min. Two strategies were evaluated: (1) conventional therapy (conventional) the risks for death and hospitalization were derived from the angiotensin-converting enzyme inhibitor arm of the SOLVD treatment trial; and (2) pacemaker insertion with atrial pacing and carvedilol therapy (pacemaker-carvedilol) risk reductions for death and CHF-related hospitalizations for carvedilol compared with conventional therapy were derived from the US Carvedilol Heart Failure Study. We assumed full carvedilol benefits for 2 years, declining benefits for the next 3 and no additional benefits after 5 years, whereas pacemaker-related adverse events persisted. Results In the base case, the pacemaker-carvedilol strategy increased mean survival by 1.3 years at an incremental cost of $7800, for an incremental cost-effectiveness of $6100 per year of life saved. Results were most sensitive to theoretical pacing-induced harm, changes in hospitalization cost, and reductions in h-blocker benefits. Conclusion Prophylactic pacemaker insertion to facilitate h-blocker treatment in patients with CHF with low resting heart rates has the potential to produce clinical benefits in a highly cost-effective manner. (Am Heart J 2006;151:820-8.) h-blockers are safe and effective drugs to reduce death and hospitalizations in most patients with symptomatic systolic heart failure. Effectiveness and costeffectiveness of carvedilol, metoprolol succinate, and bisoprolol have been demonstrated for patients with mild to moderate chronic systolic heart failure in the US Carvedilol Heart Failure Studies, MERIT-HF and CIBIS-II, respectively. 1-7 However, patients with low resting heart rates were excluded from h-blocker studies. 1-3 This may have reflected concern about reduced or From the Divisions of a General Medicine, and b Cardiovascular Medicine, University of Michigan, Ann Arbor, MI, and c Division of Cardiology, Pritzker School of Medicine, University of Chicago, IL. Supported by a grant from GlaxoSmithKline Pharmaceuticals. Dr Aaronson has received consulting and speaking fees from GlaxoSmithKline Pharmaceuticals. Portions of this paper were presented in abstract form at the 52nd Annual Scientific Session of the American College of Cardiology, March Submitted June 26, 2004; accepted June 5, Reprint requests: Keith Aaronson, MD, MS, University of Michigan Medical Center, 1500 East Medical Center Drive, Room L3623, Ann Arbor, MI keith@umich.edu /$ - see front matter n 2006, Mosby, Inc. All rights reserved. doi: /j.ahj absent effectiveness or the potential for development of symptomatic bradycardia. Although the actual risk for developing clinically significant bradycardia in this setting is unknown, many physicians routinely exclude such patients from h-blocker treatment. Prophylactic atrial pacemaker therapy would allow h-blocker treatment in bradycardic patients with heart failure. Considering the high reliability of contemporary pacing technology, the low complication rates associated with its use, its moderate cost, and the very large benefit of h-blocker therapy, we hypothesized that this approach would reduce death and hospitalizations in a cost-effective manner. Because a clinical trial to address this question has not been performed, we attempted to approximate the results of such a trial through decision analysis/cost-effectiveness modeling. To do this, we used information from existing clinical trials and observational studies. We estimated baseline risk for death and congestive heart failure (CHF) hospitalization in patients receiving an angiotensin-converting enzyme (ACE) inhibitor from the SOLVD treatment trial because this provided much longer follow-up duration than the relevant h-blockers trials, whose follow-up ranged

2 American Heart Journal Volume 151, Number 4 Stecker et al 821 Figure 1 Decision tree showing the strategies of conventional treatment for CHF in patients with low resting heart rates compared with pacemaker placement with carvedilol initiation. PM, Pacemaker. from only 6.5 months to 1.4 years. 1-4 For the base case, h-blocker benefits were estimated from the US Carvedilol Heart Failure Study 1 because cost and heart rate specific outcome data were available to us for analysis. Sensitivity analyses evaluated the effect of altering basecase estimates of costs, risks, and benefits, including smaller h-blocker benefits similar to those experienced in MERIT-HF, CIBIS-II, and COPERNICUS, 2,3,5 and possible harmful effects of atrial or right ventricular pacing. Methods We used a decision analysis model to estimate the length of life and CHF- and pacemaker-related costs for 2 treatment strategies in patients with systolic heart failure and low resting heart rates. The hypothetical cohort consisted of patients with clinically stable CHF (average age 60 years) with a left ventricular ejection fraction of V35% who were in sinus rhythm with a resting heart rate of V68 beat/min (the lower rate limit for inclusion in the US Carvedilol Trial). Pacemaker therapy included implantation of a dualchamber pacemaker that was programmed to an atrial-based, rate-responsive pacing mode (DDDR mode with long atrioventricular [AV] delay); for the base case, we assumed that all patients would effectively be receiving atrial pacing only. We discounted future costs and benefits at a 3% annual rate. Decision model We developed a Markov model by adapting a previously published model that evaluated the cost-effectiveness of carvedilol treatment in patients with chronic heart failure from a societal perspective. 6 Two strategies were evaluated for the hypothetical patient cohort: (1) conventional, patients treated with ACE inhibitors, diuretics, and digoxin; and (2) pacemaker-carvedilol, patients received conventional therapy and had a permanent pacemaker implanted to maintain adequate heart rate after initiation of h-blocker therapy with carvedilol. We evaluated the model over 20 years with monthly periodicity (ie, 240 cycles of each Markov chain). Because placebo-controlled h-blocker studies in heart failure have been limited in duration, we conservatively modeled the benefits of carvedilol to persist for 2 years and then to taper-off over the subsequent 3 years. After 5 years, the model assumed that carvedilol conferred no additional benefits, whereas the adverse events and costs for this strategy were modeled to persist for the entire 20-year follow-up period. Before entry into each Markov chain, patients assigned to the pacemaker-carvedilol strategy incurred risks and costs of pacemaker therapy before initiation of carvedilol (Figure 1). These risks included pneumothorax requiring active treatment, hematoma requiring reoperation, cardiac perforation, acute infection requiring reoperation, lead displacement, and death. Patients who developed an implantation-related pacemaker infection had the device removed and subsequently did not receive carvedilol. All other patients underwent carvedilol initiation and uptitration over 2 months, during which 5% of patients were assumed to experience an adverse event on the drug, 1 with half of these discontinuing it. Because the benefits of carvedilol in the US Carvedilol Trials were applied to the entire modeled carvedilol cohort on an intention-to-treat basis, the model intrinsically incorporates the rate of carvedilol discontinuation after randomization in the US Carvedilol Heart Failure Trials. The Markov chains contained 5 baliveq states and 1 bdeadq state (Figure 2, A [conventional] and B [pacemaker-carvedilol]). Patients were assigned to an alive state based on their number of previous hospitalizations for CHF (none, 1, 2, 3, or z4), with all patients starting in the no prior hospitalizations state. During each cycle, patients were at risk for death and hospitalization. Those who died were assigned to the dead state and did not enter further cycles. Those who did not die continued on to the next monthly cycle. They continued in

3 822 Stecker et al American Heart Journal April 2006 Figure 2 Detail in generic format of each Markov chain. Probability of death as an outpatient is conditional on number of previous hospitalizations. There are 5 ambulatory states represented as Ambulatory (i) (i is 0 to 4). A, Markov chains for patients without a pacemaker and not receiving carvedilol. B, Markov chains for patients with a pacemaker and receiving carvedilol. Patients who received a pacemaker and were carvedilol intolerant would go to a Markov chain identical to panel B but experience the higher probabilities of mortality and heart failure hospitalizations experienced by those patients on conventional therapy. their previous alive state if they had not been hospitalized during the prior cycle or in the next higher alive state if they had been hospitalized during the prior cycle. During each cycle, patients in the pacemaker-carvedilol strategy were also exposed to risks and costs for chronic pacemaker complications. Pacemaker complications included pocket erosion, infection, and lead complications, with attendant incremental mortality risks. We assumed that pocket erosion and infections would require device extraction without replacement; these patients received no further benefit from carvedilol. All patients who survived to battery end-of-life underwent pacemaker replacement. Probability estimates Risks for hospitalization and death of patients on conventional therapy were derived from the treatment arm of the SOLVD trial because it provided the longest available follow-up for patients with CHF on ACE inhibitors. 4 The conditional probabilities of death and hospitalization were calculated from data reported in SOLVD. The probability of hospitalization was contingent on the number of previous hospitalizations for that individual, whereas the probability of death was contingent on whether the patient was an inpatient or outpatient. 6 Before applying US Carvedilol Trials risk reductions to our hypothetical sample with resting heart rates of V68 beat/min, we analyzed separately data from the US Carvedilol Heart Failure Trials and from COPERNICUS, 5 and found that both the absolute risk for death and hospitalization and the relative risk reduction from carvedilol were independent of resting heart rate (data on file, GlaxoSmithKline Pharmaceuticals, Research Triangle Park, NC). Similar findings have been reported with metoprolol succinate in the MERIT-HF trial. In that study, placebo group mortality was the same for the lowest 4 quintiles of base heart rate, and the mortality reduction with metoprolol succinate was the same across all 5 quintiles of baseline heart rate. 7 We therefore applied the carvedilol benefits observed in the US Carvedilol Heart

4 American Heart Journal Volume 151, Number 4 Stecker et al 823 Failure Trials, a 65% risk reduction for mortality 1 and a 53% reduction in the mean number of hospitalizations, 8 to the death and hospitalization risks for the conventional arm of our model to obtain base-case death and hospitalization risks for the pacemaker-carvedilol arm of the model. We estimated the risks from acute and chronic complications from permanent pacemaker therapy based on large cohort studies from the literature. The frequency of pacemaker generator replacement for battery depletion was estimated to be 7 years. 9 Variable estimates for event probabilities and costs for the base-case scenario are shown in Tables I and II. Sensitivity analyses were performed on all probability estimates to assess their influence on results. The baseline risk reduction with carvedilol for death of 65% was reduced by half to approximate the 34% mortality reductions observed with bisoprolol and metoprolol succinate in patients with mild to moderate heart failure in CIBIS-II and MERIT-HF, 2,3 and the 35% mortality reduction observed with carvedilol in severe heart failure in COPERNICUS. 5 Because the risk for death during each model cycle is conditional on the number of prior hospitalizations, a 50% reduction in the death benefit can be modeled by reducing only the risk for hospitalization, only the risk for death conditional on hospitalization, or by smaller reductions in both. To achieve a 50% reduction in the mortality benefit with carvedilol, we chose to reduce the difference between the strategies for risk for both hospitalization and death by 50% for each model cycle. This resulted in a concomitant relative reduction in the hospitalization benefit with carvedilol of 43%, in league with the 40% to 50% smaller reductions in heart failure and cardiovascular hospitalizations in MERIT-HF compared with those achieved in the US Carvedilol Trials. 1,3,5 Higher and lower bounds for all other estimates were set at 0.5 and 1.5 times their base-case values unless otherwise stated (Tables I and II). We do not know that atrial pacing is without detrimental effects. There has been no study evaluating the impact of atrial pacing on morbidity and mortality in patients who do not already have symptomatic bradycardia. Therefore, we performed a secondary analysis in which theoretical risks of atrial pacing were introduced into the model. This was done by increasing each of the risks for hospitalization or death conditional on hospitalization. During the course of this study, the results of the DAVID trial were published, raising the possibility that ventricular pacing increases death (61%) and heart failure hospitalizations (54%), although neither individual risk was statistically significant. 17 We therefore performed additional sensitivity analyses by applying these event rates to hypothetical patients in the carvedilol-pacemaker arm who required AV pacing (rather than atrial pacing alone), estimated to occur in up to 1.7% of patients per year. 18 Cost estimates Costs for drugs were estimated based on average wholesale drug prices for furosemide, digoxin, and carvedilol; costs for generic lisinopril were estimated from an individual wholesaler s charges because this drug became available in a generic formulation after publication of average wholesale prices. Costs for carvedilol initiation and outpatient CHF care were estimated from Medicare reimbursement data from Table I. Congestive heart failure base-case values with ranges used in sensitivity analyses Variable Value (range) Reference Clinical variable probabilities Monthly inpatient death 20.3 (NA) [6] rate conventional treatment Monthly inpatient death 8.0 (8-14.2) [6] rate carvedilol treatment Monthly outpatient death 0.7 (NA) [6] conventional treatment Monthly outpatient death 0.3 ( ) [6] carvedilol treatment Monthly hospitalization on [6] conventional treatment No prior hospitalizations 0.8 (NA) 1 prior hospitalization 5.2 (NA) 2 prior hospitalizations 10.6 (NA) 3 prior hospitalizations 12.1 (NA) z4 prior hospitalizations 18.0 (NA) Monthly hospitalization on [6] carvedilol No prior hospitalizations 0.4 ( ) 1 prior hospitalization 2.8 ( ) 2 prior hospitalizations 5.8 ( ) 3 prior hospitalizations 6.6 ( ) z4 prior hospitalizations 9.8 ( ) Monthly initial carvedilol 2.8 ( ) [1] adverse events (each m for first 2 m) Mortality rate 7.1 ( ) [1] Rate of discontinuation of 50 (0-100)... the drug Costs Monthly cost of 28 (14-42) [10] conventional CHF drugs (lisinopril, furosemide, digoxin) Monthly cost of carvedilol 99 (50-150) [10] Initiation of carvedilol 329 ( ) CPT CPT Annual outpatient care for CHF 345 ( ) CPT BMP 3 Inpatient care for CHF on ( ) [8] conventional treatment (per episode) Inpatient care for CHF on carvedilol (per episode) 5632 ( ) [8] Values are percentages or in US dollars. NA, Not applicable; ellipses, data that were assumed because of unavailability of a source; CPT, American Medical Association Current Procedure Terminology code; BMP, basic metabolic panel as listed in Table I. Costs for pacemaker insertion, follow-up, and complications were based on published costs and Medicare reimbursement data from 2002 to 2003 as listed in Table II. Hospitalization costs for CHF were estimated based on a cost analysis study of the US Carvedilol Heart Failure Trials. 8 Variable estimates for costs for the basecase scenario are also shown in Tables I and II. Higher and lower bounds for all costs were set at 0.5 and 1.5 times their base-case values.

5 824 Stecker et al American Heart Journal April 2006 Table II. Pacemaker base-case values with ranges used in sensitivity analyses Variable Value (range) Reference Clinical variable probabilities (% T ) Implantation-related complication rates Pneumothorax 0.9 ( ) [11-13] requiring active treatment Hematoma 0.8 ( ) [11,12] requiring reoperation Cardiac perforation 0.5 ( ) [12,13] Infection requiring 1.0 ( ) [11-13] reoperation Lead displacement 2.1 ( ) [13] Death 0.08 ( ) [13] Annual chronic complication rates System infection 0.3 ( ) [12] Mortality risk 27 ( ) [14] Pocket erosion 0.4 ( ) [12] Lead replacement 0.3 ( ) [15] Mortality risk 0.6 ( ) [16] Battery life (y) 7 (6-12) [9] Require dual-chamber 0 (0-1.7) [17] pacingy Costs ($) Outpatient pacemaker initial implantation 8247 ( ) APC89 + CPT CPT71090 Outpatient pacemaker generator change 6413 ( ) APC90 + CPT33213 Implantation-related complications Pneumothorax ( ) [13] requiring active treatment Hematoma requiring 346 ( ) CPT33222 reoperation Cardiac perforation ( ) [13] Infection requiring replacement 8247 ( ) APC89 + CPT CPT71090 Lead displacement 2246 ( ) APC106 + CPT33220 Death ( ) [13] Chronic complications System infection 7392 ( ) DRG118 + CPT CPT CPT CPT CPT CPT Pocket erosion 7392 ( ) DRG118 + CPT CPT CPT CPT CPT CPT Lead complications 7498 ( ) DRG117 + CPT CPT CPT71090 Outpatient follow-up of pacemaker 213 ( ) CPT Results Base case For the base-case estimates, the pacemaker-carvedilol strategy resulted in 99 months of life versus 80 months with the conventional strategy, an undiscounted gain of 19 months (1.6 years) with the pacemaker-carvedilol strategy. The discounted gain in life expectancy with the pacemaker-carvedilol strategy was 1.3 years. The discounted cost estimates were $ for the conventional strategy versus $ for the pacemakercarvedilol strategy, a difference of $7800. Therefore, the pacemaker-carvedilol strategy resulted in a cost of $6100 per year of life saved when both costs and benefits were discounted Table III. Sensitivity analyses Sensitivity analyses confirmed the cost-effectiveness of the pacemaker-carvedilol strategy under a range of reasonable assumptions. The most economically significant variables in univariable analyses are listed in Table IV. The results were moderately sensitive to changes in the cost of hospitalizations for patients on conventional therapy. Increasing the cost by 50% resulted in the conventional strategy being dominated by the pacemaker-carvedilol strategy (ie, higher costs and a lower survival rate for the conventional strategy), whereas decreasing the cost by 50% resulted in an incremental cost-effectiveness ratio of $ per year of life saved ($20000 to save 15 months of life). Increasing the cost of hospitalizations for patients on carvedilol, increasing the cost of carvedilol itself, or decreasing the hospitalization and mortality benefits of carvedilol by 43% and 50%, respectively (resulting in a mortality benefit similar to that observed in MERIT-HF, CIBIS-II, and COPERNI- CUS) had minimal effect on the cost-effectiveness of the pacemaker-carvedilol strategy. Individually increasing each acute pacemaker implantation risk or the cost of each complication had virtually no impact on the costeffectiveness of the pacemaker-carvedilol intervention. The outcome was also insensitive to increasing the cost of pacemaker implantation. Increasing the risks for AV block and theoretical harm from atrial pacing reduced the cost-effectiveness of the pacemaker-carvedilol strategy, but cost-effectiveness ratios remained in a favorable range. A 1.7% annual rate of AV block, assuming concomitant increases in hospitalizations and mortality from the DAVID trial, 17 reduced the cost-effectiveness of the pacemaker-carvedilol strategy to $19300 per year of life saved ($2900 to save Notes to Table 2: APC, Center for Medicare and Medicaid Services Ambulatory Payment Classification; DRG, Center for Medicare and Medicaid Services Diagnosis Related Group code. TUnless otherwise indicated. ydevelopment of AV block requiring use of ventricular lead for pacing.

6 American Heart Journal Volume 151, Number 4 Stecker et al 825 Table III. Baseline results by year of follow-up Year 1 Year 5 Year 10 Year 20 PM-carv Conv PM-carv Conv PM-carv Conv PM-carv Conv Survival rate (%) Hospitalizations (per 100 patients) Life years (per patient) Total costs (per patient) ($) Incremental CER ($) PM-carv, Pacemaker-carvedilol strategy; Conv, conventional strategy; CER, cost-effectiveness ratio. Table IV. Selected variables from a 1-way sensitivity analysis on incremental cost-effectiveness Table V. Selected variables from a secondary analysis on incremental cost-effectiveness Variable Low variable estimate High variable estimate Variable Low variable estimate High variable estimate CHF-related variables Cost of hospitalization on pacemakercarvedilol therapy Cost of hospitalization on conventional therapy Reduction in carvedilol benefit (50% for mortality and 43% for hospitalization) Pacemaker-carvedilol strategy dominant 9600 Base case Cost of carvedilol Pacemaker-related variables Rate of AV blockt Base case Hospitalization due to Base case atrial pacingy Cost of pacemaker insertion Variables with highest cost-effectiveness ratios were selected for illustration. All costs and benefits are discounted at 3% annual rate. Ranges tested for each variable are listed in Tables I and II. Values are in US dollars. TAssuming increased risks for mortality and hospitalization from the DAVID trial apply to this cohort; see text. ytheoretical risk only; see text. CHF-related variables Cost of hospitalization on pacemakercarvedilol therapy Cost of hospitalization on conventional therapy Pacemaker-carvedilol strategy dominant Cost of carvedilol Pacemaker-related variables Rate of AV blockt Base case Conventional strategy dominant Hospitalization due to atrial pacingy Base case Conventional strategy dominant Mortality due to atrial pacingy Base case Conventional strategy dominant Cost of pacemaker insertion All variables were tested against a 50% decrease in mortality benefits and a 43% decrease in hospitalization benefits conferred by carvedilol. Those with highest costeffectiveness ratios were selected for illustration. All costs and benefits are discounted at 3% annual rate. Ranges tested for each variable are listed in Tables I and II. TAssuming increased risks for mortality and hospitalization from the DAVID trial apply to this cohort; see text. ytheoretical risk only; see text. 1.8 months of life). Because we cannot be certain that atrial pacing alone was not responsible for a portion of the adverse impact observed in DAVID, we also introduced variables to simulate a theoretical increase in risk for hospitalization or death from atrial pacing. When the atrial pacing variables were individually varied, only the impact of hospitalization risk was significant. Increasing this risk by 50% resulted in similar incremental cost-effectiveness of $ per year of life saved. However, it resulted from a greater increase in cost ($9300 to save 5.4 months of life). Additional sensitivity analyses of decreased carvedilol benefits (a 43% decrease in the hospitalization and 50% mortality benefits conferred by carvedilol) in combination with other factors showed the pacemaker strategy to be dominated by the conventional strategy (ie, lower costs and a better survival rate for the conventional strategy) or economically unattractive only when there was atrial- or AV-pacing induced harm (Table V). In these analyses, the widely accepted $ discounted cost per year of life saved threshold was exceeded only when rates of AV block exceeded 0.55%, the relative risk for atrial pacing induced hospitalization increased by N25%, or the relative risk for atrial pacing induced mortality increased by N22%. Figure 3 displays the results of a sensitivity analysis in which carvedilol mortality and morbidity benefits were reduced as previously, and both atrial pacing-induced mortality and hospitalization risks were simultaneously

7 826 Stecker et al American Heart Journal April 2006 Figure 3 as previously, the rates of AV block requiring ventricular pacing were increased, and the morbidity and mortality rate increases of ventricular pacing observed in the DAVID trial were applied. The incremental costeffectiveness of the carvedilol-pacemaker strategy reached the $40000 per year of life saved threshold of economic attractiveness at an annual rate of developing AV block of 0.57%. Results of a sensitivity analysis of theoretical increases in risk for death and hospitalization as a result of atrial pacing. The relative increases were applied equally and simultaneously to both risks. Figure 4 Results of a sensitivity analysis in which carvedilol associated mortality and hospitalization benefits were reduced by 50% and 43%, respectively, and rates of AV block requiring ventricular pacing were increased. Based on the DAVID trial results, we assumed that ventricular pacing induced a 61% increase in mortality and a 54% increase in heart failure hospitalization rates. increased up to 35% (ie, a 35% relative increase from the risks for death and hospitalization used in the base-case model). This shows that the pacemaker strategy is economically attractive until the atrial pacing-induced rates of both hospitalization and death are increased by 31% relative to the baseline. When both rates are increased by 33% the conventional strategy is dominant (less costly with better outcomes). Figure 4 examines the potential adverse impact of ventricular pacing in the context of reduced carvedilol benefits. In this analysis, carvedilol associated mortality and hospitalization benefits were reduced Discussion Using a Markov model, we demonstrated that pacemaker use to facilitate h-blocker therapy in patients who are currently denied their benefits may be a costeffective method of reducing death and hospitalizations from heart failure. These results were insensitive to reasonable reductions in benefits and increases in costs. Previous trials such as the US Carvedilol Heart Failure Trials, COPERNICUS, and MERIT-HF used a lower heart rate limit of 68 beat/min as a cutoff for trial entry. CIBIS- II used a lower limit of 60 beat/min. 1-3,5 The number of patients with heart failure who might not be eligible for h-blocker based on these concerns about bradycardia is not trivial. Based on an analysis of the University of Michigan heart failure database, we estimate that in the absence of a h-blocker, 15% of patients with heart failure have heart rates of V68 beat/min. Extending the heart rate lower limit to 60 beat/min would still leave 10% of patients with heart failure ineligible for h-blocker use. Sensitivity analyses addressed the impact of uncertainty inherent in economic modeling, particularly assumptions in our model that could provoke debate and affect the results. The first is that atrial pacing is not harmful with respect to mortality or CHF hospitalizations. Although there is no evidence that atrial pacing increases mortality, we cannot absolutely exclude the possibility that the DAVID findings resulted from atrial pacing rather than ventricular pacing as has been widely assumed. However, even assuming increased individual risks for mortality and hospitalization from atrial pacing, the model s results were still well within the widely accepted cost-effectiveness threshold of $ per year of life gained. When both mortality and hospitalization risks from atrial pacing were increased equally and simultaneously, cost-effectiveness remained lower than the threshold for up to a 31% increase (Figure 3). More than this limit, the ratio rapidly increased, with the pacemaker-carvedilol strategy being dominated (ie, both higher mortality and higher costs) for a z40% increase in both risks. The second assumption is that the estimates of the mortality and hospitalization benefits from the US Carvedilol Heart Failure Trials accurately represent the benefits of carvedilol in CHF. The benefits from this trial were roughly double those observed in similar patients treated with metoprolol succinate (MERIT-HF)

8 American Heart Journal Volume 151, Number 4 Stecker et al 827 or bisoprolol (CIBIS II), or in more advanced heart failure treated with carvedilol (COPERNICUS). The only head-to-head trial between h-blockers in heart failure (COMET) showed lower mortality with carvedilol than with metoprolol tartrate, although hospitalization was the same. 19 Nonetheless, reducing the benefits of carvedilol by half in our model to approximate the results of MERIT-HF and CIBIS-II in similar patients still resulted in a very economically attractive improvement in survival, with an incremental cost-effectiveness ratio of $9600 per year of life saved. Based on our model, the pacemaker-carvedilol strategy would be economically unattractive or clinically harmful only if the benefits of carvedilol were half those found in the US Carvedilol Heart Failure Study and atrial or ventricular pacing proved sufficiently harmful. There were several conservative assumptions inherent in our model that reduced the cost-effectiveness of the pacemaker-carvedilol intervention. We included full pacemaker-related risks and costs regardless of whether patients tolerated carvedilol. Thus, a carvedilol intolerant patient alive 7 years after initial pacemaker implantation (our base-case estimate for the length of battery life) would still have the pacemaker generator replaced, although it would not have been used. In a similar fashion, patients accrued the risks and costs associated with pacemaker and carvedilol therapy for the entire 20 year follow-up period, although full benefits from carvedilol were assumed for only 2 years and were tapered completely off by 5 years. Given the 17% mortality reduction over a mean follow-up of nearly 5 years seen with carvedilol versus metoprolol tartrate in COMET, our model assumption appears to be overly conservative. We applied the full increase in risk for death and heart failure hospitalizations found in the DAVID trial, although taken individually, neither was statistically significant. Furthermore, it is not clear that these risks are applicable to the patients defined by the cohort in our model. Lastly, to simplify the model, we assumed all patients required a pacemaker, when in practice, some could be started and titrated successfully on carvedilol and a pacemaker implanted only if clinically significant bradycardia developed. For these reasons, we believe that our model estimates provide very conservative lower bounds on the benefits that would be seen with a pacemaker and carvedilol strategy. To reflect contemporary pacemaker implantation practices and the uncertain effects of h-blockade on the AV node of patients with relatively slow sinus rates (but not slow enough to have a recognized indication for a pacemaker), we modeled implantation of a dualchamber (AV) pacemaker. However, many electrophysiologists advocate single-chamber (atrial lead only) pacemakers for patients with sinus node dysfunction without AV block. Simplification of the model presented here to simulate insertion of an atrial pacemaker would result in only modest enhancement in cost-effectiveness for the pacemaker-carvedilol strategy because the difference in implantation-related costsissmall. Under a variety of reasonable assumptions, a strategy of pacemaker therapy to allow h-blocker use in patients who have been previously excluded from treatment because of low resting heart rates may be highly costeffective. This conclusion is robust with respect to reasonable reductions in h-blocker benefits, increases in drug, pacemaker and hospitalization costs, and to potential increased risks for hospitalization or death from atrial or AV pacing. We conclude that pacemakers can be used to extend the benefit of h-blockers to a large and previously excluded group of patients with systolic heart failure. We thank Mary Ann Lukas, MD, for her critical review of the manuscript. References 1. Packer M, Bristow MR, Cohn JN, et al. The effect of carvedilol on morbidity and mortality in patients with chronic heart failure. US Carvedilol Heart Failure Study Group. N Engl J Med 1996;334: CIBIS-II Investigators. The Cardiac Insufficiency Bisoprolol Study II (CIBIS-II): a randomised trial. Lancet 1999;353: MERIT-HF Investigators. Effect of metoprolol CR/XL in chronic heart failure: Metoprolol CR/XL Randomised Intervention Trial in Congestive Heart Failure (MERIT-HF). Lancet 1999;353: SOLVD Investigators. Effect of enalapril on survival in patients with reduced left ventricular ejection fractions and congestive heart failure. N Engl J Med 1991;325: Packer M, Coats AJ, Fowler MB, et al. Effect of carvedilol on survival in severe chronic heart failure. N Engl J Med 2001;344: Delea TE, Vera-Llonch M, Richner RE, et al. Cost effectiveness of carvedilol for heart failure. Am J Cardiol 1999;83: Deedwania PC, Giles TD, Klibaner M, et al. Efficacy, safety and tolerability of metoprolol CR/XL in patients with diabetes and chronic heart failure: experiences from MERIT-HF. Am Heart J 2005;149: Fowler MB, Vera-Llonch M, Oster G, et al. Influence of carvedilol on hospitalizations in heart failure: incidence, resource utilization and costs. US Carvedilol Heart Failure Study Group. J Am Coll Cardiol 2001;37: Scherer M, Ezziddin K, Klesius A, et al. Extension of generator longevity by use of high impedance ventricular leads. Pacing Clin Electrophysiol 2001;24: Anonymous drug topics red book. Montvale, (NJ):Thomson Medical Economics; Aggarwal RK, Connelly DT, Ray SG, et al. Early complications of permanent pacemaker implantation: no difference between dual and single chamber systems. Br Heart J 1995;73: Kiviniemi MS, Pirnes MA, Eranen HJ, et al. Complications related to permanent pacemaker therapy. Pacing Clin Electrophysiol 1999;22:

9 828 Stecker et al American Heart Journal April Tobin K, Stewart J, Westveer D, et al. Acute complications of permanent pacemaker implantation: their financial implication and relation to volume and operator experience. Am J Cardiol 2000; 85:774-6 [A9]. 14. Klug D, Lacroix D, Savoye C, et al. Systemic infection related to endocarditis on pacemaker leads: clinical presentation and management. Circulation 1997;95: Helguera ME, Maloney JD, Pinski SL, et al. Long-term performance of endocardial pacing leads. Pacing Clin Electrophysiol 1994;17: Smith HJ, Fearnot NE, Byrd CL, et al. Five-years experience with intravascular lead extraction. US Lead Extraction Database. Pacing Clin Electrophysiol 1994;17: Wilkoff BL, Cook JR, Epstein AE, et al. Dual chamber pacing or ventricular backup pacing in patients with an implantable defibrillator: the Dual Chamber and VVI Implantable Defibrillator (DAVID) Trial. JAMA 2002;288: Kristensen L, Nielsen JC, Pedersen AK, et al. AV block and changes in pacing mode during long-term follow-up of 399 consecutive patients with sick sinus syndrome treated with an AAI/AAIR pacemaker. Pacing Clin Electrophysiol 2001;24: Poole-Wilson PA, Swedberg K, Cleland JG, et al. Comparison of carvedilol and metoprolol on clinical outcomes in patients with chronic heart failure in the Carvedilol Or Metoprolol European Trial (COMET): randomised controlled trial. Lancet 2003;362:7-13.