Cost-effectiveness analysis of PET-guided BrECADD with chemotherapy in advanced-stage, classical Hodgkin lymphoma

Abstract

Objective

The HD21 trial demonstrated efficacy and safety of brentuximab vedotin, etoposide, cyclophosphamide, doxorubicin, dacarbazine, and dexamethasone (BrECADD) vs. bleomycin, etoposide, doxorubicin, cyclophosphamide, vincristine, procarbazine, and prednisone (eBEACOPP) as frontline therapy for advanced-stage, classical Hodgkin lymphoma. This analysis evaluated the cost-effectiveness of BrECADD from a US healthcare payer perspective.

Methods

Building upon a multicenter, randomized, open-label phase 3 HD21 trial (NCT02661503) that evaluated BrECADD versus eBEACOPP in advanced-stage, classical Hodgkin lymphoma patients, we constructed a Markov model with 3-week cycles over a 50-year horizon. The model’s primary outcomes encompassed total costs, quality-adjusted life years (QALYs), and incremental cost-effectiveness ratios, with all economic parameters discounted at 3.0% annually. Cost-utility analyses employed a willingness-to-pay threshold of $100,000 per QALY, supplemented by comprehensive sensitivity and scenario analyses to verify model robustness.

Results

Economic evaluation demonstrated that compared to eBEACOPP, the BrECADD regimen yielded an additional 2.24 QALYs at an incremental cost of $113,134.17, producing an ICER of $50,411.69/QALY, substantially below the $100,000 WTP threshold. One-way sensitivity analysis identified BrECADD drug acquisition costs and health state utility values as the predominant model drivers; all ICERs remained below the $100,000 WTP threshold across the plausible ranges of all parameters. Probabilistic sensitivity analysis indicated that at a willingness-to-pay threshold of $100,000 per QALY, BrECADD was cost-effective compared to eBEACOPP in 100% of 1000 Monte Carlo iterations. The cost-effectiveness acceptability curve demonstrated that BrECADD achieved a greater than 50% probability of being cost-effective at WTP thresholds above approximately $55,000 per QALY.

Conclusions

BrECADD can be considered a cost-effective treatment versus eBEACOPP in treating advanced-stage, classical Hodgkin lymphoma in America.

Clinical trial number

Not applicable.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12962-026-00757-6.

Introduction

Hodgkin lymphoma (HL) is defined by the presence of malignant Hodgkin and Reed-Sternberg (HRS) cells, encompassing multinucleate giant cells and large mononuclear variants [1]. Hodgkin lymphoma exhibits a bimodal age distribution, with the highest incidence occurring among young adults aged 20–34 years and a secondary peak emerging in adults over 60 years [2]. Hodgkin lymphoma accounts for 10% of newly-diagnosed lymphoma cases in the US (8,720 of 86,500 cases) and 5% of lymphoma deaths, which remains a significant global health burden [3]. It is one of the most common malignancies in young adults, and a large number (40%) of patients present with advanced-stage disease [4].

Treatment of Hodgkin lymphoma is determined by stage of disease. For patients with advanced classical Hodgkin lymphoma, bleomycin, etoposide, doxorubicin, cyclophosphamide, vincristine, procarbazine, and prednisone (eBEACOPP) chemotherapy has been developed as the frontline treatment to eradicate malignant clones upfront with excellent progression-free survival rates [5]. However, beyond the well-documented bleomycin-associated pulmonary toxicity and treatment-related deaths with BEACOPP regimen, the intensive therapy imposes substantial treatment burdens, including persistent multi-organ dysfunction that severely compromises long-term health-related quality of life [6]. Thus, recent large-scale clinical efforts have focused on reducing or eliminating the use of bleomycin altogether from first-line treatment [7].

The CD30 cell surface protein is constitutively expressed at high levels in classical Hodgkin lymphoma and is therefore an attractive therapeutic target for pharmaceuticals, such as the antibody-drug conjugate brentuximab vedotin, which has shown a favourable risk–benefit ratio in relapsed or refractory classical Hodgkin lymphoma as a single agent [8]. Recently, the phase 3 HD21 trial introduced a novel PET-guided approach using BrECADD (BV, etoposide, cyclophosphamide, doxorubicin, dacarbazine, dexamethasone), demonstrating unprecedented efficacy and tolerability. Compared to eBEACOPP, BrECADD achieved superior 4-year PFS (94.3% vs. 90.9%, HR = 0.66, p = 0.035) and significantly reduced treatment-related morbidity (42% vs. 59%, p < 0.0001), while enabling 64% of PET-2-negative patients to receive only four cycles [9]. Crucially, BrECADD mitigated eBEACOPP’s key toxicities: it reduced transfusion requirements by > 50% and restored gonadal function in > 85% of male patients [9]. Despite these benefits, no cost-effectiveness analyses (CEA) have evaluated BrECADD within an individualized, PET-guided framework—a significant gap given its potential to optimize both clinical and economic outcomes. Health care economic concerns are distinctly relevant for Hodgkin lymphoma not only because of its overall incidence, but also because of the well-established association between Hodgkin lymphoma outcomes and patient socioeconomic factors [10]. As such, we sought to evaluate the cost-effectiveness of BrECADD vs. eBEACOPP as frontline therapy for patients with advanced classical Hodgkin lymphoma from a US healthcare payer perspective, leveraging HD21 efficacy data and real-world cost inputs to identify value-based treatment paradigms.

Materials and methods

Patients and intervention

The study population was derived from the randomized, multicentre, open-label phase 3 HD21 trial (NCT02661503) conducted by the German Hodgkin Study Group (GHSG), which enrolled 1,500 patients aged ≤ 60 years with newly diagnosed advanced-stage classical Hodgkin lymphoma [9]. Patients were randomized 1:1 to receive either BrECADD or eBEACOPP for four or six cycles. Cycles of BrECADD (full dose level 4) were brentuximab vedotin (1∙8 mg/kg up to a maximum of 180 mg absolute dose intravenously on day 0), etoposide (150 mg/m² intravenously on days 1–3), cyclophosphamide (1250 mg/m² intravenously on day 1), doxorubicin (40 mg/m² intravenously on day 1), dacarbazine (250 mg/m² intra­venously on days 2–3), and dexamethasone (40 mg orally on days 1–4). cycles of BEACOPP (full dose level 4) were escalated doses of etoposide (200 mg/m² intravenously on days 1–3), doxorubicin (35 mg/m² intravenously on day 1), and cyclophosphamide (1250 mg/m² intravenously on day 1), and standard doses of bleomycin (10 mg/m² intravenously on day 8), vincristine (1∙4 mg/m² intravenously on day 8), procarbazine (100 mg/m² orally on days 1–7), and prednisone (40 mg/m² orally on days 1–14). Administration of a G-CSF product was obligatory in both treatment groups. As an alternative to daily G-CSF, the pegylated form could be used [9]. Each treatment cycle had a duration of 21 days. Treatment duration was guided by interim PET-2 response: PET-2-negative patients received 4 cycles, while PET-2-positive patients received 6 cycles. In both treatment arms, approximately 64% of patients achieved PET-2-negative status and were assigned to 4 cycles only, while the remaining 36% with PET-2-positive results received 6 cycles [9]. To maintain consistency with the clinical trial protocol and the chemotherapy administration schedule, the simulation cycle length in our Markov model was set at 3 weeks (21 days).

The safety assessment specifically focused on severe AEs (grade ≥ 3) that occurred with a frequency of 5% or greater in either BrECADD or eBEACOPP groups. As all study data were exclusively derived from the HD21 clinical trial without additional subject enrollment, separate approval from an independent ethics committee was not required. All study procedures were rigorously performed in compliance with applicable ethical guidelines and regulatory requirements. The current analysis utilized data from a randomized, multicentre, open-label phase 3 HD21 trial (NCT02661503). Prior to study enrollment, written informed consent was obtained from all participating individuals or their legally authorized representatives.

Model construction

A Markov model incorporating three distinct health states (PFS, Progressive Disease (PD), and death) was implemented through Excel 2024 and R 4.3.3 to analyze disease trajectories in advanced cancer cases. The fully functional model with source code is available from the corresponding author upon reasonable request. Model initialization placed all subjects in the PFS state, with unidirectional state transitions evaluated at 21-day intervals. This study developed a three-state Markov model (Fig. 1) using Microsoft Excel 2023 software to extrapolate clinical and economic outcomes beyond the trial period, incorporating progression-free survival (PFS), progressive disease (PD), and death states. To assess the appropriateness of candidate parametric distributions for long-term survival extrapolation, we fitted seven standard distributions (Exponential, Gamma, Generalized Gamma, Gompertz, Weibull, Log-logistic, and Log-normal) to the reconstructed PFS and OS data for each treatment arm. The visual comparison of these fitted curves against the reconstructed Kaplan-Meier estimates, together with the corresponding AIC and BIC statistics, is provided in Supplementary Figure S1 and Supplementary Tables S1–S2. Based on these assessments and the clinical considerations described in Section “Model survival and transition probabilities”, a mixture cure model framework was selected for the base-case survival extrapolation (see below). The resulting lifetime extrapolation curves from the mixture cure model are displayed in Supplementary Figure S2. To maintain consistency with the clinical trial protocol, we set the simulation cycle duration at 3 weeks, which corresponds to the chemotherapy administration schedule.

Fig. 1.

Markov model structure. PFS, progression-free survival; PD, progression disease

Given the high cure potential of advanced-stage classical Hodgkin lymphoma, we employed a mixture cure model framework for survival extrapolation beyond the trial observation period [11]. The mixture cure model posits that a proportion of patients (θ, the cure fraction) will never experience the event of interest, while the remaining patients (1 − θ) follow a standard parametric survival distribution. The survival function takes the form: S(t) = θ + (1 − θ) × S_uncured(t), where θ represents the cure fraction and S_uncured(t) denotes the survival function for uncured patients. This framework was selected based on the following rationale: (1) clinical evidence supports substantial cure potential in advanced-stage classical Hodgkin lymphoma, with 4-year PFS exceeding 90% and 4-year OS exceeding 98% in the HD21 trial [9]; (2) the mixture cure model produces clinically plausible long-term survival plateaus, rather than curves that inevitably decline to zero, which is appropriate for a disease with high cure rates in a young patient population (median age 31 years); and (3) this approach ensures structural consistency between PFS and OS curves during extrapolation, preventing implausible crossing of survival curves that can occur with independently fitted standard parametric distributions.

For PFS in both treatment arms, the Gompertz distribution was selected for the uncured patient survival function, as it provided the best statistical fit (lowest AIC and BIC) among the eight candidate distributions evaluated (Supplementary Table S1). The PFS cure fractions were estimated directly from the data using maximum likelihood estimation: θ_PFS = 88.6% (95% CI: 81.2%–93.3%) for BrECADD and θ_PFS = 80.1% (95% CI: 61.6%–91.0%) for eBEACOPP. For OS, the limited number of death events (n = 19 per arm) precluded reliable estimation of the cure fraction from the data. Therefore, the OS cure fractions were fixed at 95% for BrECADD and 93% for eBEACOPP, based on observed 4-year OS rates from the HD21 trial (98.6% and 98.2%, respectively [9]), with a conservative downward adjustment to account for potential additional mortality among uncured patients beyond the trial follow-up period. An exponential distribution was used for the uncured patient OS survival function, as it requires estimation of only one parameter, which is more appropriate given the sparse event data. The sensitivity of model conclusions to the fixed OS cure fraction assumptions was assessed through scenario analyses varying these values by ± 5% points.

All mixture cure models were fitted using the flexsurvcure package (version 1.3.0) in R (version 4.3.3) [12]. To ensure clinical plausibility throughout the extrapolation period, the PFS survival curve was constrained to not exceed the OS curve at any time point, such that the proportion of patients in the progressive disease state remained non-negative. Reconstructed Kaplan–Meier curves for PFS and OS from the HD21 trial are presented in Supplementary Figure S1. The lifetime extrapolation curves for both treatment arms are presented in Supplementary Figure S2.

In this study, a 3% annual discount rate was applied for future health utility and cost [13]. A willingness-to-pay (WTP) threshold was set at $100,000 per QALY [14]. A treatment was considered cost-effective if its calculated ICER was lower than this established threshold [15].The overall cost-effectiveness analytical framework, including population, model type, time horizon, and key assumptions, is summarized in Table 1.

Table 1.

Cost-effectiveness analytical framework

Category	Intervention	Comparator
	BrECADD	eBEACOPP
Population	Patients with advanced-stage, classical Hodgkin lymphoma
Decision-analytical model	Markov model
Outcomes	Direct medical costs, quality-adjusted life-year, and incremental cost-effectiveness ratio
Disease model	Progression-free state, progressed-disease state, and death
Cycle length	3 weeks
Time horizon	50 years
Discount rate	3%
Willingness-to-pay	$100,000 per QALY
Measurements	ICER, incremental cost-effectiveness ratio

Model survival and transition probabilities

The OS and PFS data of BrECADD and eBEACOPP were extracted by the GetData Graph Digitizer. The estimates of the OS curve and PFS curve for all groups were reconstructed by R software. The OS and PFS curves for each treatment group were reconstructed using R software (version 4.3.3), and eight parametric distributions (Exponential, Gamma, Generalized Gamma, Gompertz, Weibull, Weibull PH, Log-logistic, and Log-normal) were applied to fit the survival data. Candidate parametric distributions for survival extrapolation were evaluated using a multi-criteria approach, in accordance with recommendations from the NICE Decision Support Unit Technical Support Document 14 [16]. Eight standard distributions (Exponential, Gamma, Generalized Gamma, Gompertz, Weibull, Weibull PH, Log-logistic, and Log-normal) were fitted to the reconstructed PFS and OS data, and their statistical fit was assessed using AIC and BIC (Supplementary Tables S1 and S2). However, for OS, AIC and BIC differences across all candidate distributions were less than 5 points in both treatment arms, indicating that statistical fit alone was insufficient to discriminate among them. Therefore, the final selection of the survival extrapolation approach was guided by the following additional criteria: (1) clinical plausibility of long-term projections, assessed by evaluating whether extrapolated survival curves produced realistic outcomes consistent with published long-term follow-up data from Hodgkin lymphoma cohorts; (2) structural consistency between PFS and OS curves, ensuring that OS remained at or above PFS throughout the entire extrapolation horizon; and (3) compatibility of implied hazard function shapes with the known clinical trajectory of the disease. Based on these considerations, a mixture cure model framework was adopted for the base-case analysis, as described in Section “Model construction“. Detailed parameter estimates for the mixture cure models are reported in Supplementary Table S3.

Cost and utility estimates

From a US healthcare payer perspective, this study only considered direct medical costs, including drug costs, Routine clinic visit costs, end-of-life palliative care costs, and adverse event management costs (Table 2). Drug acquisition costs were derived from the CMS Medicare Part B ASP Payment Limit files (Q1 2025). For intravenous drugs with multiple manufacturer-specific HCPCS codes, the not-otherwise-specified (NOS) code was used to reflect the weighted average market price. Oral drug costs (dexamethasone, procarbazine, prednisone) were obtained from the CMS National Average Drug Acquisition Cost (NADAC) database. Drug doses were calculated based on a reference patient weighing 70 kg with a body surface area of 1.8 m², with wastage accounted for by rounding up to the next full single-use vial size [17–19]. Detailed per-cycle drug cost calculations are presented in Table S4.

Table 2.

Model parameters, baseline values, ranges, and distributions for sensitivity analyses

Parameters and distributions	Estimated value	DSA range		PSA Distribution	Source
Utility value
PFS	0.75	0.71	0.78	Beta	Christopher et al. [20]
PD	0.66	0.57	0.75	Beta	Christopher et al. [20]
Disutility of Serious neutropenic events	0.2	0.1	0.3	Beta	Tolley et al. [21]
Disutility of Pulmonary toxicity	0.09	0.05	0.12	Beta	Nafees et al. [22]
Disutility of Peripheral neuropathy	0.07	0.04	0.1	Beta	Delea et al. [23]
Disutility of Anemia	0.07	0.04	0.1	Beta	Tolley et al. [21]
Discount rate	0.03	0	0.05	Beta	Cohen et al. [24]
Drug cost, $
BrECADD	31,912.1	25,529.7	38,294.6	Gamma	CMS ASP Q1 2025
eBEACOPP	3,104.1	2,483.3	3,724.9	Gamma	CMS ASP Q1 2025
ASCT	130,698	100,842	179,262	Gamma	Huntington et al. [17]
Allo-SCT	67,582	57,095	83,734	Gamma	Majhail et al. [25]
Consolidative radiation	19,886	15,908.8	23,863.2	Gamma	Avalere Health [26]
Salvage chemotherapy	10,698	8,558	12,837	Gamma	Huntington et al. [17]
Monitoring costs, $
PET	2,500	2,000	3,000	Gamma	Smith et al. [27]
End-of-life care	30,395	24,316	36,474	Gamma	Delea et al. [23]
Routine clinic visit	119	95.2	142.8	Gamma	Huntington et al. [17]
Intravenous chemotherapy administration	145	116	174	Gamma	Huntington et al. [17]
AE cost, $
Serious neutropenic events	13,058	10,446.4	15,669.6	Gamma	Delea et al. [23]
Pulmonary toxicity	19,427	15,541.6	23,312.4	Gamma	Delea et al. [23]
Peripheral neuropathy	8,570	6,856	10,284	Gamma	Delea et al. [23]
Anemia	7,388	5,910.4	8,865.6	Gamma	Delea et al. [23]
Parameters
BrECADD PFS cure fraction	0.886	0.81	0.93	Beta	Estimated from data
eBEACOPP PFS cure fraction	0.801	0.62	0.91	Beta	Estimated from data
BrECADD OS cure fraction	0.95	0.9	1.0	Scenario analysis	Fixed; based on HD21 [9]
eBEACOPP OS cure fraction	0.93	0.88	0.98	Scenario analysis	Fixed; based on HD21 [9]
AE rates
BrECADD neutropenic fever rate	0.28	0.22	0.34	Beta	HD21 [9]
BrECADD pulmonary toxicity rate	0.03	0.02	0.04	Beta	HD21 [9]
BrECADD peripheral neuropathy rate	0.01	0.008	0.012	Beta	HD21 [9]
BrECADD anemia rate	0.30	0.24	0.36	Beta	HD21 [9]
eBEACOPP neutropenic fever rate	0.21	0.17	0.25	Beta	HD21 [9]
eBEACOPP pulmonary toxicity rate	0.05	0.04	0.06	Beta	HD21 [9]
eBEACOPP peripheral neuropathy rate	0.02	0.016	0.024	Beta	HD21 [9]
eBEACOPP anemia rate	0.59	0.47	0.71	Beta	HD21 [9]

In the economic model, drug costs were applied on a cycle-by-cycle basis to reflect the PET-2-guided treatment strategy used in the HD21 trial [9]. During cycles 1–4, the full per-cycle drug cost was applied for all patients in the PFS state. During cycles 5–6, only 36% of the per-cycle drug cost was applied, reflecting the proportion of PET-2-positive patients who continued treatment. From cycle 7 onward, no first-line drug costs were applied. This approach yields a weighted average of 4.72 treatment cycles per patient in both arms.

AEs were considered to include grade ≥ 3 adverse events: serious neutropenic events, pulmonary toxicity, anemia and peripheral neuropathy. Related treatment costs were derived from the published literature. Other cost parameters, such as Routine clinic visit ($119) and Intravenous chemotherapy administration ($145), were derived from published literature [17]. Monitoring frequencies of advanced-stage, classical Hodgkin lymphoma (2024) were driven from drug package inserts and the The LYSA pragmatic guidelines for Hod [20].

For health utility values, the utility for the PFS state was 0.75, and for the PD state, it was 0.66, both derived by mapping EORTC QLQ-C30 data to EQ-5D-3 L using the Young et al. (2015) algorithm in the HD21 trial (NCT02661503) [20]. Adverse-event-specific disutility values were applied rather than a single uniform decrement, to more accurately capture the differential quality-of-life impact of individual grade ≥ 3 adverse events. The disutility values used were: 0.20 for serious neutropenic events [21], 0.09 for pulmonary toxicity [22], 0.07 for peripheral neuropathy [23], and 0.07 for anemia [21]. These values were sourced from published oncology literature, as Hodgkin lymphoma-specific disutility estimates for individual adverse events are not available. Each disutility was applied for the duration of one model cycle (21 days) during the treatment period, consistent with the acute nature of most grade ≥ 3 chemotherapy-related adverse events and with the methodology used in prior cost-effectiveness analyses of frontline Hodgkin lymphoma [17]. Grade ≥ 3 adverse event incidence rates were extracted directly from the HD21 trial [9]. For the BrECADD arm: serious neutropenic events (febrile neutropenia) 28%, pulmonary toxicity 3%, peripheral neuropathy 1%, and anemia 30%. For the eBEACOPP arm: serious neutropenic events 21%, pulmonary toxicity 5%, peripheral neuropathy 2%, and anemia 59%. The QALY decrement attributable to each adverse event was calculated as the product of the disutility value, the arm-specific incidence rate, and the duration of application (21 days).

The detailed parameters and distributions are provided in Table 2.

Sensitive analysis

To assess model stability, we conducted both univariate and probabilistic sensitivity analyses. The univariate approach enabled identification of parameters exerting substantial influence on ICER outcomes, facilitating a thorough evaluation of parameter-related uncertainties across the entire model, result interpretation was enhanced through tornado plot visualizations [18].

We conducted a probabilistic sensitivity analysis using 1,000 Monte Carlo iterations to assess the joint impact of parameter uncertainty on cost-effectiveness conclusions. Gamma distributions were assigned to cost-related variables, and beta distributions were assigned to health utility values, adverse event disutilities, adverse event incidence rates, and PFS cure fractions. The PFS cure fractions estimated by the mixture cure model were sampled from beta distributions parameterized using the estimated mean and standard error (BrECADD: α = 92.82, β = 11.94; eBEACOPP: α = 21.90, β = 5.44). In each PSA iteration, the sampled PFS cure fraction was used to proportionally rescale the PFS survival curve, with a constraint ensuring that PFS survival did not exceed OS survival at any time point. Because the OS cure fractions were fixed rather than estimated from the data, their uncertainty was evaluated separately through scenario analyses varying the values by ± 2 and ± 5% points. Results were presented through cost-effectiveness scatter plots and acceptability curves, demonstrating the probability of cost-effectiveness for BrECADD and eBEACOPP across varying WTP thresholds.

Reults

Base case results

Within a 50-year time frame and after applying discounts, the treatment BrECADD demonstrated the greatest average benefit, achieving 34.32 QALYs and incurring the highest average total costs at $743,281.68. In comparison, the eBEACOPP strategy was estimated to yield an average of 32.08 QALYs with associated costs of $630,147.51 (Table 3). Our analysis revealed that the BrECADD treatment significantly outperformed eBEACOPP. Specifically, when compared to eBEACOPP, the BrECADD regimen was expected to provide an additional 2.24 QALYs on average, with an incremental cost of $113,134.17, leading to an incremental cost-effectiveness ratio (ICER) of $50,411.69 per QALY gained.

Table 3.

CEA results of base-case

Parameters and distributions	Outcomes of Markov model		Incremental changes
	BrECADD	eBEACOPP
Costs	743,281.68	630,147.51	113,134.17
Drug costs	470,497.84	352,160.64	118,337.20
Monitoring costs	229,173.04	224,322.67	4,850.37
AEs costs	43,610.8	53,664.2	-10,053.4
Overall LYs	46.97	44.28	2.69
Total QALYs	34.32	32.08	2.24
Incremental cost per QALY gained			50,411.69

One-way sensitivity analysis

The tornado diagram (Fig. 2) revealed that drug acquisition costs exerted the greatest influence on the ICER among all model parameters. The incremental difference in drug costs between BrECADD and eBEACOPP amounted to $118,336, which slightly exceeded the total incremental cost of $113,134. This reflects a partial offset from reduced adverse event management costs in the BrECADD arm (-$10,053 relative to eBEACOPP), consistent with the lower treatment-related morbidity observed in the HD21 trial [9]. Within each arm, drug acquisition costs represented 63% of total expenditure in the BrECADD arm ($470,498 out of $743,281.68) and 56% in the eBEACOPP arm ($352,162 out of $630,147.51). Varying BrECADD drug costs by ± 20% produced ICERs ranging from approximately $39,000/QALY to $61,000/QALY. Health state utility values for the PFS and PD states were the second and third most influential parameters. Across all parameters evaluated, the ICER consistently remained below the predefined WTP threshold of $100,000/QALY, confirming the robustness of the base-case conclusion favoring BrECADD as a cost-effective treatment strategy.

Fig. 2.

Tornado programs

Probabilistic sensitivity analysis

The cost-effectiveness scatter plot (Fig. 3) shows that all 1,000 Monte Carlo iterations are located in the northeast quadrant of the cost-effectiveness plane, indicating that BrECADD consistently yields greater health benefits but at higher total costs than eBEACOPP. No iterations fall within the southeast quadrant (greater QALYs at lower costs), confirming that BrECADD does not dominate eBEACOPP in the strict health-economic sense. Instead, all iterations fall below the willingness-to-pay threshold line of $100,000 per QALY. At a WTP threshold of $100,000 per QALY, BrECADD was cost-effective compared with eBEACOPP in 100% of iterations. The cost-effectiveness acceptability curve (Fig. 4) further shows that the probability of BrECADD being cost-effective exceeded 50% at a WTP threshold of approximately $50,000 per QALY, reached 91% at $60,000 per QALY, and reached 100% at thresholds of $75,000 per QALY and above.

Fig. 3.

Probabilistic sensitivity analysis

Fig. 4.

Cost-effectiveness acceptability curve

Value of information analysis

At the primary willingness-to-pay threshold of $100,000 per QALY, the per-patient expected value of perfect information (EVPI) was $0, indicating minimal decision uncertainty at this threshold. This is consistent with the probabilistic sensitivity analysis, in which BrECADD was cost-effective in 100% of 1,000 Monte Carlo iterations at the $100,000/QALY threshold. Therefore, under the threshold most relevant to US healthcare decision-making, additional research to reduce parameter uncertainty is unlikely to provide meaningful value.

To provide additional context, Fig. 5 presents the EVPI across a range of WTP thresholds. The EVPI peaked at $6,375.56 per patient at a WTP threshold of $50,000/QALY, which is close to the base-case ICER of $50,411.69/QALY. This is expected, as decision uncertainty is typically greatest when the WTP threshold is close to the base-case ICER. At thresholds above $75,000/QALY or below $30,000/QALY, the EVPI approached zero, indicating that decision uncertainty was concentrated mainly in the intermediate threshold range of $40,000–$65,000/QALY.

Fig. 5.

Excepted value of information

Scenario analysis

To evaluate the impact of structural uncertainty on cost-effectiveness conclusions, we conducted three scenario analyses addressing methodological and structural choices in the model (Tables 4 and 5).

Table 4.

Parameters and distributions of scenario analysis

Scenario	Parameter varied	Base case	Scenario value(s)
1	Time horizon	50 years	5 years
2a	Uncured PFS/OS distribution	Gompertz	Weibull
2b	Uncured PFS/OS distribution	Gompertz	Log-normal
3	OS cure fractions	BrE 0.95, eBE 0.93	± 2% and ± 5% variations

Table 5.

Results of scenario analysis

Scenario	Description	ICER ($/QALY)	<$100K?
Base case	Lifetime horizon, mixture cure model	50,411.69	Yes
Scenario 1	5-year time horizon	94,013.50	Yes
Scenario 2a	Weibull distributions	69,763.86	Yes
Scenario 2b	Log-normal distributions	96,754.36	Yes
Scenario 3a	Equal OS θ = 0.93	61,519	Yes
Scenario 3b Scenario 3c	Equal OS θ = 0.95 BrE OS θ = 0.90	62,821 120,614	Yes No
Scenario 3d	BrE OS θ = 1.00	37,999	Yes
Scenario 3e	eBE OS θ = 0.88	38,653	Yes
Scenario 3f	eBE OS θ = 0.98	121,908	No

Scenario 1: Time horizon. Restricting the time horizon from the base-case lifetime (50 years) to 5 years, which corresponded to the horizon used in our original analysis, yielded an ICER of $94,013.50 per QALY (Table 5). Under this restricted horizon, BrECADD provided 4.789 QALYs at a total cost of $322,220.44, compared with 4.303 QALYs at $276,538.67 for eBEACOPP, representing an incremental gain of 0.486 QALYs at an incremental cost of $45,681.77. While the ICER remained below the $100,000/QALY threshold, it was substantially higher than the base-case ICER of $50,411.69/QALY. This substantial difference demonstrates that restricting the time horizon to 5 years systematically underestimates the long-term value of BrECADD by truncating the survival benefits attributable to its higher cure fraction, while capturing the upfront drug acquisition costs in full. This finding supports the adoption of a lifetime horizon as the base case, consistent with ISPOR Good Practice recommendations for treatments affecting long-term survival in young patient populations.

Scenario 2: Scenario 2 evaluated alternative parametric distributions for uncured patient survival. The Weibull distribution yielded PFS cure fraction estimates similar to the Gompertz base case (BrECADD: 88.7% vs. 88.6%; eBEACOPP: 83.2% vs. 80.1%) and an estimated ICER of $69,763.86/QALY, remaining below the $100,000/QALY threshold. The Log-normal distribution produced markedly different cure fraction estimates with substantially wider confidence intervals (BrECADD: 80.8%, 95% CI: 57.6%–92.8%; eBEACOPP: 78.8%, 95% CI: 70.9%–85.1%), suggesting less stable model convergence with this distribution. Under the Log-normal specification, the narrower gap in cure fractions between arms resulted in an estimated ICER of $96,754.36/QALY.

Scenario 3: OS cure fraction assumptions. Because the overall survival cure fractions could not be reliably estimated from the HD21 trial data owing to the limited number of observed death events (n = 19 per arm) and were therefore fixed in the base-case analysis, a dedicated scenario analysis was undertaken to quantify the influence of this structural choice on the cost-effectiveness conclusion. Two categories of variation were examined. First, within a narrow clinically plausible margin of ± 2% points, both treatment arms were assigned identical OS cure fractions of 0.93 (Scenario 3a) and 0.95 (Scenario 3b). Under these equal-cure specifications, the incremental QALY gain attributable to BrECADD is driven solely by the difference in PFS cure fractions between the two arms (88.6% vs. 80.1%), and the resulting ICERs were $61,519/QALY and $62,821/QALY, respectively—both comfortably below the $100,000/QALY WTP threshold. Second, a wider ± 5% point perturbation was applied to each arm individually while the other arm remained at its base-case value. Raising the BrECADD OS cure fraction to 1.00 (Scenario 3d) or lowering the eBEACOPP OS cure fraction to 0.88 (Scenario 3e) widened the long-term survival gap between arms, yielding ICERs of $37,999/QALY and $38,653/QALY, respectively. Conversely, lowering the BrECADD OS cure fraction to 0.90 (Scenario 3c) or raising the eBEACOPP OS cure fraction to 0.98 (Scenario 3f) reversed the direction of the OS benefit, producing ICERs of $120,614/QALY and $121,908/QALY, which exceeded the $100,000/QALY threshold. It should be emphasized, however, that both of these threshold-exceeding scenarios rest on the counterfactual assumption that BrECADD delivers a lower long-term OS cure fraction than eBEACOPP—a premise that is inconsistent with the HD21 trial, in which BrECADD demonstrated numerically superior 4-year OS (98.6% vs. 98.2%) together with a substantially lower incidence of treatment-related morbidity (42% vs. 59%, p < 0.0001). Across all clinically plausible configurations of the OS cure fractions, the ICER remained below the $100,000/QALY threshold, reinforcing the robustness of the base-case finding that BrECADD is cost-effective relative to eBEACOPP from a US healthcare payer perspective.

All scenario analyses supported the robustness of the base-case conclusion favoring BrECADD as a cost-effective first-line treatment for advanced-stage classical Hodgkin lymphoma.

Discussions

HD21, a large randomised, multicentre, parallel, open-label trial, was recently reported comparing standard BrECADD to eBEACOPP in the first-line setting for advanced-stage, classical HL [9]. At a median follow-up of 48 months, investigators found a significant progression-free survival benefit of BrECADD versus eBEACOPP [28]. However, to our knowledge, no manuscript published to date has evaluated the cost-effectiveness of BrECADD in the first-line, definitive treatment of advanced-stage, classical HL.

In this cost-effectiveness analysis of the BrECADD regimen versus eBEACOPP in advanced-stage, classical HL, we found that BrECADD would be considered cost-effective at current US payer-perspective prices over a lifetime horizon. Our model was sensitive to several inputs, most notably drug acquisition costs and health state utility values. Results demonstrated that BrECADD yielded an incremental gain of 2.24 QALYs at an incremental cost of $113,134.17, producing an ICER of $50,411.69/QALY, substantially below the $100,000/QALY WTP threshold. The finding suggests that BrECADD offers cost-effectiveness advantages over eBEACOPP despite its considerably higher per-cycle drug acquisition cost. Although BrECADD demonstrated superior clinical efficacy (4-year progression-free survival 94.3% vs. 90.9%; HR = 0.66, p < 0.001) and substantially reduced treatment-related morbidity (42% vs. 59%, p < 0.0001) at current pricing, future indication expansions and therapeutic advancements may enable price optimization to further improve cost-effectiveness.

Drug costs were calculated from a US healthcare payer perspective using CMS Medicare Part B ASP Payment Limit files (Q1 2025) for intravenously administered agents and the CMS NADAC database for oral agents. Per-cycle doses were computed for a reference patient weighing 70 kg with a body surface area of 1.8 m², and drug wastage was explicitly accounted for by rounding each calculated dose up to the nearest full single-use vial size [19, 29]. In addition, model inputs were conservatively specified to bias against the intervention. Although HD21 reported a higher incidence of serious neutropenic events in the BrECADD arm (28% vs. 21% for eBEACOPP), primary prophylactic granulocyte colony-stimulating factor (G-CSF) was assumed to reduce serious neutropenic events with BrECADD to rates observed with eBEACOPP [30]. Critically, model conclusions remained stable even under scenarios where primary G-CSF prophylaxis reduced serious neutropenic events below eBEACOPP levels.

Our analysis also incorporates PET-2-guided adjustments to first-line therapy. As reported by Borchmann et al., PFS for PET-2-positive patients exceeded 90% at three years with the use of escalated BEACOPP in both treatment groups [31–33]. Patients with an incomplete PET response to first-line therapy received salvage chemotherapy, ASCT, or allo-SCT, with the remaining patients successfully treated with consolidative radiation alone. HD21 demonstrated the highest numbers of progression-free survival events in patients with PET-2-positive disease, confirming the appropriateness of an individualised PET-guided strategy and the rationale for the corresponding changes to first-line therapy in our model.

This study employed a Markov model with a lifetime (50-year) horizon to assess the cost-effectiveness of first-line BrECADD combination therapy for advanced Hodgkin lymphoma. While BrECADD significantly reduced transfusion requirements (> 50%) and restored gonadal function in > 85% of male patients, distributional effects across socioeconomic groups were not quantified. Given known disparities in Hodgkin lymphoma outcomes [10], future distributional cost-effectiveness analyses are warranted. Several limitations should be noted: (1) long-term PFS and OS were extrapolated using a mixture cure model framework. Although this approach is well-suited to diseases with substantial cure potential and produced clinically plausible lifetime projections, survival extrapolation inevitably introduces uncertainty, particularly regarding the fixed OS cure fractions. We addressed this uncertainty through probabilistic sampling of the PFS cure fractions and dedicated scenario analyses for the OS cure fractions and alternative survival distributions. (2) Our economic evaluation exclusively considered direct medical costs from a US healthcare payer perspective, thereby omitting potentially substantial indirect costs such as productivity loss, caregiver burden, and late effects including secondary malignancies. This limitation may affect the comprehensiveness of our cost-effectiveness assessment. (3) These data were extracted from aggregated published trial results rather than individual patient-level data, which may not fully capture detailed patient characteristics. In real-world clinical practice, eligibility for subsequent treatment is influenced by multiple factors; patients with poor performance status in particular are more likely to opt for best supportive care rather than active anti-cancer therapy. (4) Uncertainties existed regarding both the model structure and parameter sources; we addressed these through one-way sensitivity analysis, probabilistic sensitivity analysis, and three structural scenario analyses (time horizon, alternative survival distributions, and OS cure fraction assumptions), all of which supported the robustness of the base-case conclusion. (5) Adverse-event-specific disutility values were derived from oncology populations other than Hodgkin lymphoma (chronic lymphocytic leukaemia, non-small cell lung cancer, and relapsed/refractory Hodgkin lymphoma), as disease-specific disutility estimates for individual adverse events in advanced-stage classical Hodgkin lymphoma are currently unavailable. However, the grade ≥ 3 adverse events evaluated (febrile neutropenia, pulmonary toxicity, peripheral neuropathy, and anemia) represent common chemotherapy-related toxicities whose quality-of-life impact is expected to be broadly comparable across cancer types. Subgroup analyses (e.g., by age or PET-2 status) were not performed due to trial data limitations. These limitations could influence the precision of the economic evaluations for the treatment strategies.

Conclusions

From a US healthcare payer perspective, the ICER for BrECADD vs. eBEACOP based on the HD21 study falls within the range of thresholds for cost-effectiveness for oncology therapies in the US. Thus, BrECADD is likely to be a cost-effective treatment for patients with advanced-stage, classical Hodgkin lymphoma receiving frontline therapy.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

BrECADD

Brentuximab vedotin, etoposide, cyclophosphamide, doxorubicin, dacarbazine, and dexamethasone

eBEACOPP

Bleomycin, etoposide, doxorubicin, cyclophosphamide, vincristine, procarbazine, and prednisone

WTP

Willingness-to-pay

ICER

Incremental cost-effectiveness ratio

QALY

Quality-adjusted life year

ASCT

Autologous stem-cell transplantation

allo-SCT

Allogeneic stem-cell transplantation

PET

Positron emission tomography

EVPI

Expected value of perfect information

AIC

Akaike information criterion

BIC

Bayesian information criterion

PFS

Progression-free survival

PD

Progressive disease

OS

Overall survival

AEs

Adverse events

Author contributions

HZ, GZ, YL, HL, and BC contributed to the study design, data collection, and analysis. GZ drafted the manuscript. All authors read and approved the final manuscript.

Funding

This study was funded by the General Project of Humanities and Social Science Research for the Ministry of Education in 2022 (No.22YJAZH147); Research team of Pharmacoeconomics and Health Technology Evaluation, Guangdong Pharmaceutical University (2024ZZ13); General Project for Planning Philosophy and Social Sciences in Guangdong Province in 2024 (No.GD24CGL29); Guangdong University Innovation Team Project (No.2022WCXTD011).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethical approval and consent to participate

Not applicable. This study was based on secondary data analysis and modeling, with no direct involvement of human participants.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.