A decision-tree model was employed to assess the budget impact of letermovir for CMV prophylaxis among D+/R− kidney transplant recipients. The model assumed a hypothetical population of healthcare plan members from which the target patient population was derived (Fig. 1). All costs are reported in 2025 US dollars, with costs from earlier years inflated to 2025 values using the Consumer Price Index; no discounting was applied [14]. A targeted literature search was conducted to identify published evidence relevant to the model population, structure, and inputs. Sources were included if they reported data on adult kidney transplant recipients, CMV D+/R− status, letermovir or valganciclovir prophylaxis, leukopenia/neutropenia and associated clinical outcomes, healthcare resource use, costs, or budget impact modeling methods. Peer-reviewed publications, pivotal trial evidence, real-world studies, national data sources, and established modeling guidance were prioritized. When suitable published evidence was unavailable or not directly aligned with the model requirements, internal analyses and expert clinical input were used to inform assumptions and assess clinical plausibility. Where specified, model assumptions were reviewed and validated by expert clinical input obtained through an advisory board meeting or individual consultation.

Budget impact model structure. D+ CMV-positive donor, LN leukopenia/neutropenia, R− CMV-negative recipient

The model was conducted from a US payer perspective and assumed a hypothetical 1,000,000-member plan. The number of D+/R− kidney transplantation patients eligible for CMV prophylaxis was derived using values available in the literature. First, the proportion of people expected to undergo kidney transplantation in the hypothetical plan was calculated using the proportion of adult kidney transplant recipients overall in 2021 in the US, among all adults (i.e., 23,851 kidney transplant recipients [15]/258,213,334 adults [16]). The resulting value showed 0.0092% of adults in the hypothetical population underwent kidney transplantation. The target patient population (i.e., D+/R− kidney transplant recipients) was derived by multiplying the number of adults undergoing kidney transplantation by the prevalence of D+/R− (i.e., 19.50%) as per Jorgenson et al. 2021 [17]. Ultimately, there were 18 target patients per year within the hypothetical plan population (i.e., 1,000,000 plan members × 0.0092% undergoing kidney transplantation × 19.5% with D+/R−).

Utilization rates of each prophylaxis product were applied to the estimated number of eligible patients entering the model each year. In the status quo market scenario (i.e., Scenario 1, without letermovir), all prophylaxis-eligible patients received VGCV. In the alternative scenario (i.e., Scenario 2, with letermovir), some patients received letermovir while the remaining patients received VGCV. The Scenario 2 base case assumed that letermovir utilization was 23% in year 1, 27% in year 2, and 31% in years 3–5. Without available real-world market data, the assumption that utilization would increase in this way was predicated on the idea that letermovir utilization would be lowest in the first year and then increase as it is adopted into institutional standard practices and management guidelines.

Consistent with the MK-8228-002 clinical trial [13], letermovir patients received 480 mg of letermovir daily for 28 weeks as well as acyclovir 400 mg every 12 hours as a preventive measure against HSV and VZV [13]. VGCV was administered orally at a dose of 900 mg. Acyclovir was not required in the VGCV cohort since VGCV acts as an anti-HSV and anti-VZV agent [13]. The average duration of prophylaxis was 176.2 days for letermovir and 160.6 days for VGCV, based on the durations observed in the phase III trial (data on file).

Total prophylaxis drug costs consist of the costs incurred during an average inpatient stay (or all-cause rehospitalization) and outpatient visit during the 28-week prophylaxis period. Specifically, costs attributable to all-cause rehospitalization reflected the prophylactic drug costs during the inpatient stay. It was assumed that hospitals would cover the cost of prophylactic drugs for the duration of rehospitalization.

From the clinical trial analyses, the 28-week all-cause rehospitalization rates were 34.60% for the letermovir group and 40.74% for the VGCV group, with the average duration of inpatient stay being 11.0 days and 11.1 days, respectively (data on file). The costs incurred for prophylaxis drugs during inpatient stays, as well as the costs for outpatient administration during the 28-week period, were calculated using the data provided in Table 1. All other inpatient costs are considered in different sections of the model. Drug cost data used in the model were obtained from the AnalySource database [18], with the lowest wholesale acquisition cost being used. Leukopenia and neutropenia were defined based on patient self-report of symptoms or direct reporting of these events.

LN event probabilities were defined according to the clinical definition employed in the MK-8228-002 trial [13]. LN events were defined as either events reported by the patient or a white blood cell count of ≤ 3500 cells/µL and absolute neutrophil count ≤ 1000 cells/µL, respectively [13]. The MK-8228-002 trial [13] duration was 52 weeks; however, the 28-week assessment period was applied to derive 52-week LN rates in the model for multiple reasons. First, this time period is in alignment with the pre-specified exploratory safety endpoint of the trial. Second, outcomes observed at 28 weeks were assumed to be representative of 52-week outcomes. This assumption was validated through an advisory board involving experts in HTA, clinical practice settings, and health economics, and through individual consultations with two clinical experts (one clinical expert employed by Merck Sharp & Dohme LLC, a subsidiary of Merck & Co., Inc., Rahway, NJ, USA, and one external physician with transplant/nephrology-relevant clinical expertise). These consultations were used to confirm the clinical relevance and plausibility of the model assumptions.

Second, it is the time period wherein the data are the most robust and reliable, given that there were frequent, pre-specified hematology assessments at the majority of study visits during the treatment period. The trial protocol specified that from the time of treatment allocation through 14 days post-treatment cessation, all AEs had to be reported. After this duration, only significant AEs deemed related to the study treatment were obligated to be reported. Given the structured and systematic reporting of LN during the 28-week prophylaxis period, it is evident that data from this timeframe would be the most comprehensive and unbiased LN rate to be used in the model.

Probabilities of LN were 26% for letermovir and 64% for VGCV as reported in the trial [13].

Assumptions and inputs for deriving LN event costs are summarized in Table 2. The annual LN management costs were considered from both inpatient and outpatient settings whereby the proportion of patients requiring inpatient care (or rehospitalization) was derived from additional ad-hoc analyses of the MK-8228-002 clinical trial (data on file).

LN-associated clinical events refer to the infection-related complications caused by LN. Annual clinical event rates associated with LN (Table 3) were derived from a retrospective database analysis using TriNetX data (data on file) in kidney transplant recipients who initiated VGCV prophylaxis within 30 days after transplantation, reflecting the standard-of-care prophylaxis at the time of letermovir introduction and the availability of mature real-world data for this population. This model also included rates and associated costs of one-time LN-associated clinical events (i.e., acute graft rejection and graft loss). The proportions of patients experiencing these events were derived from a post-hoc analysis of the prophylaxis arm of the MK-8228-002 trial (data on file) stratified by the proportion of patients who had any event of LN (acute graft rejection: 8.25%; graft loss: 1.60%) and those who did not experience LN (acute graft rejection: 6.27%; graft loss: 1.10%). Note that there was an attempt to derive these values from the real-world analysis but challenges in parameterizing acute graft rejection and graft loss with the TriNetX data led to implausible estimates due to factors such as recipient wait time, sensitization, mismatch, or donor quality.

Annual costs of LN-associated clinical events (Table 4) were obtained from a retrospective database analysis using the IBM Truven MarketScan Commercial Claims and Encounters database (2016–2020) among kidney transplant recipients who initiated VGCV prophylaxis within the first 30 days after transplant [21]. While adjusting for baseline characteristics, the regression model generated all-cause costs associated with each clinical event, in addition to the costs in the absence of the respective event for both LN and LN-free (LNF) groups. The precise cost for each clinical event was then determined by subtracting the all-cause costs (without the clinical event) from those directly associated with the event and dividing the result by the total number of annual clinical events. The attribution of LN-associated event costs as inpatient- or outpatient-based was done using outcomes from a claims database analysis by Turzhitsky et al. (2022) [22]. Accordingly, the distribution of costs for LN patients was 65.8% outpatient and 34.2% inpatient; and the distribution of costs for LNF patients was 76.8% outpatient and 23.2% inpatient (detailed calculations in Appendix A, see electronic supplementary material [ESM]).

The primary model outcome was the total annual budget impact of letermovir, which was calculated as the total costs in the scenario with letermovir (i.e., Scenario 2) minus the total costs in the scenario without letermovir (i.e., Scenario 1). Other outcomes included the budget impact on a per member per month (PMPM) basis (i.e., cost difference between the two scenarios divided by the number of plan members divided by 12) as well as the costs by cost component (e.g., CMV prophylaxis, LN event management, and LN-associated events) including further disaggregation by costs incurred in an inpatient versus outpatient setting.

To assess the impact of uncertainty in model inputs and assumptions on the results, one-way sensitivity analyses (OWSA) were conducted. This approach varied each input parameter by ±20% from its base-case value. This range is applied to capture plausible uncertainty around key model inputs in the absence of empirical confidence intervals, while remaining within a realistic and interpretable bound for decision makers. Assessed inputs and parameters included the proportion of the kidney transplantation population eligible for prophylaxis, assumptions regarding the relative utilization of the two prophylaxis products, the time period over which the incidence of LN events was assessed, and the annual direct medical cost of LN events and LN-associated clinical events.