The analysis was conducted from the public health care payer perspective using a monthly cycle over a 20-year time horizon, consistent with Japanese guidelines [15]. A reference value of ¥5,000,000 (¥1=US $0.006 as of July 7, 2026) per quality-adjusted life year (QALY) was used to evaluate cost-effectiveness. The incremental cost-effectiveness ratio (ICER) of CRT-D was assessed relative to optimal medical therapy (OMT).

The modeled cohort reflected the CRT-D group of the RAFT (Resynchronization-Defibrillation for Ambulatory Heart Failure Trial) study [16]. The mean age was 66.3 (SD 9.3) years, and 84.8% of the participants were male. NYHA functional class distribution was 76.7% class II and 23.3% class III. The mean left ventricular ejection fraction was 22.5% (SD 5.4%), and mean QRS duration was 158.1 (SD 23.8) ms. Ischemic etiology accounted for 68.7% of cases, and LBBB morphology was present in 71.7%.

A partitioned survival model was developed to evaluate CRT-D cost-effectiveness in patients with HFrEF, with OMT as the comparator. The model included 3 health states: after treatment (follow-up), hospitalization, and death. Hospitalization events included admissions due to cardiac arrest (CA), worsening HF, and CRT-D–related complications. The CRT-D survival curve was derived by digitizing Kaplan-Meier data from long-term RAFT follow-up [16] and reconstructing individual patient-level survival data using the method proposed by Guyot et al [17]. Parametric survival models, including exponential, Weibull, Gompertz, log-logistic, log-normal, gamma, and generalized gamma distributions, were evaluated. On the basis of the Akaike information criterion (AIC), Bayesian information criterion (BIC), and visual inspection, the gamma distribution provided the best fit and was selected. Comparative AIC and BIC results can be found in Table S1 in Multimedia Appendix 1. Survival for OMT was estimated by applying a hazard ratio (HR) for all-cause mortality from the meta-analysis by Woods et al [3] (0.54, 95% CI 0.43-0.69), assuming proportional hazards.

The primary outcome was QALYs, integrating survival and health-related quality of life. Utility values were assigned according to NYHA functional class and differed by treatment group. Adverse outcomes were modeled as hospitalization events, including HF, CA, device-related complications, and temporary disutility associated with device replacement. The incidence of HF hospitalization was obtained from a prior study [18], which reported cumulative rates of 19.5% in the CRT-D group and 26.1% in the OMT group over a median follow-up of 40 months. These cumulative incidences were converted into monthly transition probabilities for each treatment group, assuming a constant hazard throughout the observation period. The incidence of hospitalization due to CA was estimated from indirect published evidence [19-21]. In the CRT-D group, Tseng et al [19] reported that 6.4% of all-cause deaths following cardiac implantable electronic device therapy were attributable to CA. On the basis of the annual all-cause mortality reported in the MADIT-RIT (Multicenter Automatic Defibrillator Implantation Trial–Reduce Inappropriate Therapy) trial, this proportion corresponded to an estimated CA incidence of 0.14% per year in the CRT-D group. In the OMT group, the annual incidence of CA was derived from a previously reported rate of sudden cardiac death in the conventional therapy group (4.5% per year) [21], used as a surrogate marker for arrhythmic CA requiring hospitalization. This yielded an estimated annual incidence of 6.51%. Device-related complications were modeled exclusively in the CRT-D group during follow-up, with a cumulative incidence of 14% over 48 months [22]. This value was converted into a monthly transition probability assuming a constant hazard. Among previously reported complications [23,24], superficial and systemic infections accounted for 21.7% and 26.6%, respectively, whereas lead dislodgement accounted for 51.7%. Device replacement was assumed in cases of systemic infection or lead dislodgement. Device longevity was assumed to be 111 months based on a recent large-scale observational report from a single manufacturer [25]. In the model, device replacements due to battery depletion were assumed to recur continuously without upper limits as long as the simulated cohort survived.

We included costs associated with the initial intervention; device replacement; hospitalizations (CA, HF, and device-related complications); and additional follow-up, including remote monitoring and inappropriate shocks. Costs related to procedure-related complications at initial implantation were incorporated into the initial intervention cost. Cost inputs were derived from analyses of the JROAD-DPC database (approval 20240001) and the Japanese medical fee schedule, which provides standardized reimbursement for each medical service [26]. The database, administered by the Japanese Circulation Society, contains diagnosis procedure combination and per-diem payment system data from approximately 1500 hospitals as of 2024 and covers 2012 to 2024 [27,28]. It represents one of the largest nationwide databases for cardiac and vascular diseases in Japan. For the cost analysis in this study, we extracted data on Japanese patients who underwent initial CRT-D implantation or device replacement between 2012 and 2022. Cost extraction was based on diagnostic and procedure codes. CRT-D implantation and device replacement were identified using procedure codes in the claims database. Hospitalization costs for HF and CA were assumed to be identical between groups as they reflect baseline disease management. Clinical cost differences between groups were captured by adding device-related complications to the CRT-D group. Device-related complications were classified as superficial or systemic infections using International Classification of Diseases, 10th Revision, diagnostic codes. The specific codes used for each definition are provided in Table S2 in Multimedia Appendix 1. Utilities were assigned according to NYHA functional class (I-IV), and cohort-level utility was calculated as a weighted average based on the distribution of NYHA classes [29]. Because treatment was assumed to modify the NYHA class distribution, different stable-state utility values were applied to each treatment group. Specifically, CRT-D was assumed to improve NYHA functional class compared with OMT. On the basis of the COMPANION (Comparison of Medical Therapy, Pacing, and Defibrillation in Heart Failure) trial [30], the proportion of patients with NYHA class improvement at 6 months was 57% in the CRT-D group and 38% in the OMT group. These improvement rates were applied to the baseline NYHA class distribution to estimate follow-up utility. The magnitude of improvement was assumed to be equivalent for patients with baseline NYHA class II and III, and the difference in improvement rates between treatment groups was assumed to persist throughout the time horizon. On the basis of these assumptions, the mean health-related utility during follow-up was 0.805 for CRT-D and 0.781 for OMT. For hospitalization events, mean utility values were 0.649 for HF hospitalization and 0.44 for CA hospitalization. Disutility values of −0.10 for complications and –0.05 for device replacement were applied [31]. All model inputs are summarized in Table 1. Costs and utilities were discounted at an annual rate of 2%, consistent with Japanese cost-effectiveness analysis guidelines [15]. All analyses were conducted using R (version 4.4.3; R Foundation for Statistical Computing).

Formal subgroup analyses were not performed because the individual patient-level clinical characteristics required to define subgroups were unavailable in the data sources. Instead, heterogeneity in treatment effects was evaluated through sensitivity analyses by varying key parameters, including the HR of CRT-D, across plausible ranges reported in the literature.

Scenario analyses examined uncertainty related to treatment effect duration and time horizon. In one scenario, the treatment effect remained constant for 7.5 years and then declined linearly up to 20 years, consistent with prior studies [12,32]. Specifically, this decline was modeled by allowing the HR to approach 1.0 and the difference in utility values between the groups to approach 0. In separate scenarios, the time horizon was set to 10 and 30 years, whereas the treatment effect remained constant and identical to the base case, which used a 20-year horizon.

To evaluate parameter-level and overall model uncertainty, both deterministic sensitivity analysis (DSA) and probabilistic sensitivity analysis (PSA) were performed. Parameter ranges and probability distributions were applied from published sources when available. For parameters lacking reported ranges or distributions, a –20% to +20% variation from base case values was assumed. The discount rate was varied between 0% and 4% in accordance with relevant guidelines [15].

This study was reviewed and approved by the institutional review board of the Japanese Heart Rhythm Society (approved 20240001) This study used anonymized data from the JROAD-DPC registry. The requirement for individual informed consent was waived by the Institutional Review Board, and participants were provided with the opportunity to opt out in accordance with the applicable ethical guidelines.

In the base case analysis, per capita costs were ¥12,258,410 for CRT-D and ¥640,056.90 for OMT, yielding an incremental cost of ¥11,618,353. CRT-D produced 7.07 QALYs compared with 4.75 QALYs for OMT, corresponding to an incremental gain of 2.32 QALYs. The resulting ICER was ¥5,009,880 per QALY (Table 2).

In the scenario analysis, the ICER increased to ¥5,423,235 per QALY when the treatment effect was assumed to decline after 7.5 years (Table 2). When the time horizon was shortened to 10 years, the ICER increased to ¥8,523,072 per QALY. Conversely, extending the time horizon to 30 years reduced the ICER to ¥4,386,803 per QALY.

DSA identified the primary drivers of the ICER as CRT-D efficacy (HR), discount rate, and initial treatment cost (“costINDEX”; Figure 1). Variation in CRT-D efficacy produced ICERs ranging from ¥3,892,821 to ¥7,625,335 per QALY. Variation in the discount rate resulted in ICERs ranging from ¥4,465,415 to ¥5,624,129 per QALY, whereas variation in initial treatment cost yielded ICERs ranging from ¥4,519,272 to ¥5,500,488 per QALY.

PSA results are shown in Figure 2. The median ICER was ¥5,022,618 per QALY, with a range of ¥3,804,418 to ¥7,178,795 per QALY across the 2.5th and 97.5th percentiles. At a willingness-to-pay threshold of ¥5,000,000 per QALY, the probability that CRT-D was cost-effective was 49.2%.

Using the Japanese JROAD-DPC database, we assessed the cost-effectiveness of CRT-D compared with OMT from a public health insurance perspective. Over a 20-year horizon, CRT-D generated an incremental cost of ¥11,618,353 and a QALY gain of 2.32, yielding an ICER of ¥5,009,880 per QALY. This estimate approximated the Japanese reference value. When a waning treatment effect was incorporated, the ICER increased to ¥5,349,656 per QALY, remaining comparable to the reference threshold. In UK settings, Mealing et al [11] reported that CRT-D was cost-effective in only 10 of 24 patient subgroups at a £30,000 per QALY threshold (£1=US $1.34 as of July 7, 2026), highlighting substantial heterogeneity. In 4 subgroups similar to our population (QRS duration of ≥150 ms and NYHA class II and III with or without LBBB), the ICER for CRT-D vs OMT ranged from £15,000 to £20,000, below the accepted UK threshold of £30,000. In a US study [12] using a reference value of US $100,000 per QALY, CRT-D was consistently the most cost-effective strategy among patients with QRS duration above 150 ms regardless of LBBB status or NYHA class. Although reference thresholds differ across the United Kingdom, United States, and Japan, our findings appear conservative in this international context. An ICER of approximately ¥5 million per QALY suggests that further gains in cost-effectiveness will likely depend on refined patient selection.

DSA identified populations in which higher economic value may be achieved. The principal ICER driver was the HR, indicating that greater treatment effects improve cost-effectiveness. A meta-analysis by Woods et al [3] showed that, compared with OMT, the HR for CRT-D was lower in female than in male patients irrespective of QRS duration or age. The same analysis demonstrated that, when CRT-D was compared with OMT, the HR approached 1.0 among patients without LBBB or with QRS duration below 150 ms, indicating attenuation of survival benefit in these subgroups. Because survival was modeled using all-cause mortality, CRT-D is likely to be more cost-effective in patients with a lower burden of high-mortality comorbidities. This interpretation aligns with a recent patient-level meta-analysis of 8 randomized trials [33], which reported substantial survival benefit in patients with no comorbidities (adjusted HR 0.54) but diminished benefit in those with 3 or more comorbidities (adjusted HR 0.83). Accordingly, patients with a lower comorbidity burden may tend to exhibit a more favorable ICER, although this trend requires careful interpretation regarding cumulative longitudinal costs. However, a detailed analysis of specific comorbidities was not feasible in the present study, and further research is warranted to evaluate cost-effectiveness stratified by comorbidity profiles. Additionally, the crucial trials underpinning our efficacy estimates were conducted over a decade ago. Since then, contemporary OMT has advanced significantly with modern foundational therapies (ie, angiotensin receptor–neprilysin inhibitors and sodium-glucose cotransporter 2 [SGLT2] inhibitors) [34,35]. In particular, SGLT2 inhibitors are now widely used in Japan as a key component of guideline-directed medical therapy [36]. If these modern drugs substantially improve baseline outcomes, the incremental benefit of CRT-D over OMT might be attenuated in current practice, potentially increasing the ICER. Future modeling studies incorporating clinical data from the modern therapy era are warranted. Recent incremental innovations in CRT-D, including rATP and optimization algorithms, have improved efficacy but increased device costs. Several studies have shown these enhancements to be cost-effective in Japanese and other national settings [13,14,37]. Overall, the additional costs associated with these functional upgrades appear acceptable from a cost-effectiveness perspective. Another influential determinant was the initial treatment cost, driven largely by device price. The current reimbursement for CRT-D devices in Japan is between ¥3,090,000 and ¥4,750,000 depending on the model. Together with lead costs, this accounts for more than half of the initial treatment cost. Device price also affects downstream costs related to battery depletion and infection requiring replacement; therefore, reductions in device price would likely generate substantial savings. In PSA, the 95% credible interval for the ICER was ¥3,804,418 to ¥7,178,795, with a 49.2% probability of remaining below ¥5 million per QALY. These findings indicate moderate parameter uncertainty, with ICER estimates distributed within approximately –¥2.5 million to +¥2.5 million of the reference value.

Several limitations of this study should be acknowledged. First, although the JROAD-DPC database captures a substantial proportion of acute care hospitals in Japan, it does not include all medical institutions. Furthermore, the JROAD-DPC database was not used to derive clinical event probabilities because this database possesses inherent limitations in tracking long-term, longitudinal patient outcomes across different medical institutions once patients are discharged to outpatient care. Consequently, cost inputs may be subject to selection bias, limiting generalizability to all patients with HF in Japan. Second, the 20-year time horizon required extrapolation of survival and costs beyond observed follow-up. To mitigate this limitation, survival curves were generated using parametric modeling of long-term data from the RAFT study. Although using overseas trial data (RAFT) requires caution regarding generalizability, the trial’s mean age (66.3 years) is comparable to that of major Japanese CRT-D registries (66.5 or 68.4 years) [38]. Nevertheless, this population-level cost-effectiveness model does not incorporate individual risk factors. Future Japanese studies incorporating patient-level subgroup analyses are needed to refine estimates and support value-based allocation within the health care system. Finally, detailed subgroup analyses by clinical characteristics, such as QRS duration, LBBB status, or comorbidity burden, were not feasible because of data constraints. Instead, the potential influence of these factors was examined through DSA by varying HRs rather than directly estimating subgroup-specific effects.

Despite these limitations, to our knowledge, this study represents one of the first comprehensive cost-effectiveness evaluations of CRT-D in the Japanese health care setting using real-world cost data.

CRT-D for patients with HFrEF in Japan demonstrated cost-effectiveness within an acceptable range relative to the reference value. Sensitivity analyses identified the HR and initial treatment cost as key determinants of value. These findings emphasize the importance of identifying patient subgroups with a high expected treatment response. Specifically, future cost-effectiveness analyses stratified by comorbidity profiles will provide further valuable insights. From a policy perspective, reducing device costs may further enhance the societal value of this intervention.