Cost–effectiveness analysis of immune checkpoint inhibitors as first-line therapy in advanced biliary tract cancer

Abstract

Aim: To assess the cost–effectiveness of immune checkpoint inhibitors as first-line treatments for advanced biliary tract cancer (BTC).

Methods: This pharmacoeconomic evaluation employed the fractional polynomial network meta-analysis and partitioned survival model. Costs and utilities were collected from the literature and databases. Sensitivity analyses were used to examine uncertainties.

Results: The incremental cost–effectiveness ratios (ICERs) of first-line treatment strategies were $761,371.37 per quality-adjusted life-year (QALY) or $206,222.53/QALY in the US and $354,678.79 /QALY or $213,874.22/QALY in China, respectively. The sensitivity analysis results were largely consistent with the base case.

Conclusion: From the US and Chinese payer perspectives, adding durvalumab or pembrolizumab to chemotherapy is unlikely to be cost effective in the first-line setting for advanced BTC.

Plain language summary

Summary points.

• The fractional polynomial network meta-analysis and partitioned survival model technique were employed to assess the cost–effectiveness of first-line immune checkpoint inhibitors for treating advanced biliary tract cancer (BTC) from the standpoint of US and Chinese payers.

• In comparison to gemcitabine plus cisplatin (GP), pembrolizumab plus GP (PGP) attained an incremental cost–effectiveness ratio (ICER) of $761,371.37/quality-adjusted life-year (QALY) in the US and $354,678.79/QALY in China.

• The ICER of durvalumab plus GP (DGP) versus PGP was $206,222.53/QALY in the US and $213,874.22/QALY in China.

• One-way sensitivity analysis (OWSA) yielded an ICER for PGP versus GP from $607,790.88/QALY to $1,011,491.52/QALY and $284,362.14/QALY to $471,195.27/QALY in US and China, respectively.

• OWSA generated an ICER for DGP versus PGP from $62,539.45/QALY to $349,905.61/QALY and $122,887.67/QALY to $304,860.78/QALY in US and China, respectively.

• Probabilistic sensitivity analysis (PSA) demonstrated that, when compared with GP in the USA and China, the chances of PGP being cost-effective was 0 and 0% at the threshold of $150,000/QALY and $35,897/QALY, respectively. PGP was preferred to GP in 50% of simulations if the WTP threshold was approximately $754,200/QALY and $350,320/QALY in the USA and China, respectively.

• PSA showed that, when compared with PGP in the USA and China, the probability of DGP being cost-effective was 0.2704% and 0.0002% at the threshold of $150,000/QALY and $35,897/QALY, respectively. DGP was preferred to PGP in 50% of simulations if the threshold was approximately $203,480/QALY and $211,000/QALY in the USA and China, respectively.

• From the US and Chinese payer perspectives, adding durvalumab or pembrolizumab to the GP regimen is unlikely to be cost-effective compared with GP in the first-line setting for advanced BTC at the threshold of $150,000/QALY and $35,897/QALY, respectively.

1. Background

Biliary tract cancer (BTC), accounting for approximately 3% of all gastrointestinal malignancies, is the second most commonly diagnosed primary liver cancer [1]. BTC, mainly including gallbladder cancer and both intrahepatic and extrahepatic cholangiocarcinoma, represent a significant health problem in prevalent regions such as China, Korea, Japan and Thailand with 5.7 to 85 cases per 100,000 individuals [2]. Data from the USA and other countries indicate that the incidence of intrahepatic cholangiocarcinoma has risen for the past 30 years [3]. Majority of BTC individuals receive advanced diagnosis, with limited treatment alternatives and a grim prognosis with a 5-year survival rate of almost 2% [4]. The standard-of-care first-line therapy for advanced BTC (gemcitabine plus cisplatin [GP] chemotherapy) has remained unchanged for the past decade [5]. Therefore, more effective treatment regimens are urgently needed in clinical practice.

Recently, immune checkpoint inhibitors such as programmed death 1 (PD-1) and programmed death-ligand 1 (PD-L1) inhibitors have surfaced as promising antitumor agents across various malignancies. Additionally, the integration of immunotherapy with chemotherapy has yielded enhanced outcomes compared with chemotherapy alone in a variety of solid tumor types owing to the immunomodulatory role of chemotherapy [4]. Then the landmark TOPAZ-1 trial explored the efficacy and safety of the anti-PD-L1 antibody durvalumab plus GP (DGP) versus GP as first-line therapy in individuals with advanced BTC [5]. In the TOPAZ-1 trial, significant improvements in progression-free survival (PFS) (hazard ratio [HR]: 0.75; 95% CI: 0.63 to 0.89) and overall survival (OS) (HR: 0.80; 95% CI: 0.66 to 0.97) were achieved in the individuals receiving DGP compared with those receiving GP. Additionally, DGP was related to an acceptable safety profile in the TOPAZ-1 trial. Based on these results, DGP was recommended by guidelines from National Comprehensive Cancer Network (NCCN) and Chinese society of clinical oncology (CSCO) as the first line preferred strategy [6,7]. Following the TOPAZ-1 trial, the KEYNOTE-966 trial assessed the efficacy and safety of the anti-PD-1 antibody inhibitors pembrolizumab plus GP (PGP) versus GP as first-line treatment in individuals with advanced BTC. Along with the absence of any new safety signals, significant improvements in PFS (HR: 0.86; 95% CI: 0.75 to 1.00) and OS (HR: 0.83; 95% CI: 0.72 to 0.95) were observed in the individuals treated with PGP compared with those treated with GP in the KEYNOTE-966 trial [8]. Subsequently, PGP was recommended by the CSCO guideline as the first line preferred treatment regimen [7].

With limited medical resources, it is important to optimize resource allocation and achieve the efficient utilization of these resources. While there is no published pharmacoeconomic assessment of PGP versus GP for the treatment of advanced BTC from the perspective of US and Chinese payers, nor are there studies simultaneously investigating the Cost–effectiveness of DGP, PGP and GP. As such, we conducted the pharmacoeconomic assessment of DGP, PGP and GP as first-line treatments for advanced BTC from the perspectives of US and Chinese payers.

1.1. Population & interventions

Our target population was adults (aged 18 years and older) with histologically confirmed unresectable, locally advanced or metastatic BTC, including extrahepatic cholangiocarcinoma, gallbladder cancer or intrahepatic cholangiocarcinoma. The population either received no previous systemic therapy, or had completed neoadjuvant or adjuvant therapy at least 6 months before the diagnosis of unresectable or metastatic disease. The patient characteristics described above align with those reported in the TOPAZ-1 trial and KEYNOTE-966 trial [5,8]. Eligible individuals received one of three interventions: GP; PGP; and DGP.

1.2. Network meta-analysis

A network meta-analysis was adopted since head-to-head clinical trials of DGP and PGP were unavailable. A systematic review was performed on 30 June 2023, to identify investigations of treatment strategies related to advanced BTC. PubMed, Embase, Web of Science, Cochrane Library and Scopus databases were scanned, and the corresponding search strategies were provided in Supplementary Table S1.

Effectiveness data from PFS and OS Kaplan-Meier curves were collected employing GetData Graph Digitizer (version 2.26). Then individual patient data were reproduced using Guyot's method [9]. Also, the hypothesis of the proportional hazards was tested with log cumulative hazards plots for each trial. To obtain time-varying HRs of treatment regimens, all the first- and second-order fractional polynomial (FP) models listed in the research by Jansen were built [10]. Then FP models were integrated into a fixed-effect Bayesian framework with three parallel Markov chains comprising 100,000 samples following a 10,000 samples burn-in. The best model was identified by fit statistics, namely deviance information criterion (DIC) and visual inspection of fitted curves against observed curves. The R2jags and survival packages in R (version 4.2.3) were required to complete above FP network meta-analysis.

1.3. Model survival & progression estimates

To begin this procedure, we chose to use the survival outcomes associated with GP (the reference regimen) from the KEYNOTE-966 trial as the baseline effectiveness dataset. This decision was motivated by the fact that the KEYNOTE-966 trail, which had more mature survival data, had a larger sample size [8]. Subsequently, the survival data of GP from the KEYNOTE-966 trial were fitted utilizing the following survival functions: exponential, Weibull, log-normal, log-logistic, gamma, gengamma and Gompertz. The survival function was determined based on the lowest value of the Akaike information criterion (AIC), Bayesian information criterion (BIC), visual inspection and clinical rationality. Corresponding survival curves for DGP, PGP were then calculated by applying the aforementioned time-varying HRs to the baseline survival curve. The survHE package in R (version 4.2.3) was used to build aforementioned parametric survival models.

1.4. Costs & utilities

From the standpoints of payers in both the US and China, direct medical costs associated with treatment were covered, which include drug costs, drug administration costs, treatment-related adverse events (TrAEs) treatment costs, follow-up and examinations costs and terminal care costs. In USA, drug costs were derived from the latest average sale prices of the Centers for Medicare and Medicaid Services (updated July 2023), alongside the Federal Supply Schedule prices published by the National Acquisition Center [11,12]. Meanwhile, in China, drug costs were calculated based on the median winning bid prices found in the YaoZhi database [13]. Drug administration costs in USA were collected from the Medicare Physician Fee Schedule (MPFS) [14]. All other costs were determined based on relevant literature [15–20].

The mean height and weight of adult populations in the US are 1.70 m and 74.7 kg [21], respectively, while in China, they are 1.64 m and 64.8 kg [22]. Considering that most BTC patients are typically middle-aged or elderly and frequently undergo weight loss due to their condition, we assumed that the mean height and weight for US patients were 1.70 m and 70 kg, respectively, and for Chinese patients, they were 1.64 m and 60 kg. For the purpose of simplifying calculations, we only considered grade 3 or higher adverse events with an incidence rate of 5% or higher, which included neutropenia or neutrophil count decreased, anemia, thrombocytopenia or platelet count decreased, and white blood cell count decreased. As clinical trials only offered overall incidence rates for adverse events, the TrAEs treatment costs were treated as one-time costs calculated in the initial cycle of the model.

Based on data from TOPAZ-1 [5] and KEYNOTE-966 [8], the percentages of individuals in the DGP groups, PGP groups and GP groups who subsequently received second-line treatment were 42.5, 47.5 and 49.0%, respectively, while the remaining patients received best supportive care. Following recommendations from the NCCN and CSCO guidelines [6,7], we assumed that, FOLFOX (fluorouracil + leucovorin calcium + oxaliplatin) would serve as the second-line chemotherapy, regorafenib as the second-line targeted therapy, pembrolizumab as the second-line immunotherapy, and XELIRI (irinotecan + capecitabine) as other second-line treatment option. All costs were adjusted to reflect July 2023 values using the Consumer Price Index (CPI) for both China and the US [23,24]. Additionally, we converted Chinese costs into US dollars using the July 2023 exchange rate (USD 1.0 = CNY 7.1619).

As the clinical trials TOPAZ-1 [5] and KEYNOTE-966 [8] did not provide utility values for individuals in PFS and PD states, moreover, considering the relatively low occurrence of biliary tract cancers, there were no studies with sufficient sample sizes to directly assess their utility values. Therefore, we opted to use data from previously published studies, which were 0.76 for PFS states and 0.68 for PD states [25]. Additionally, we adjusted the utility according disutilities due to severe adverse reactions [26,27]. The discount rate for future costs and health benefits is 3% per year in USA [28] and 5% per year in China [29].

1.5. Base case analysis

Partitioned survival models (PSM) are frequently employed in pharmacoeconomic evaluations of cancer treatments [30]. The direct correspondence between commonly reported PFS and OS and the survival functions that determine state occupancy estimations in partitioned survival analysis makes the model intuitively appealing, easy to understand and create [31]. The PSM was utilized to represent the survival condition (PFS, progressive disease [PD] and death) of individuals with advanced BTC subjected to different treatments (Figure 1). TreeAge Pro 2021 was employed for PSM building. The 1-month cycle was chosen in accordance with the predefined cycle length in TreeAge Pro 2021. The 10-year simulation horizon was set, given the extremely low 5-year survival rate for advanced BTC, and it is deemed sufficient to accurately capture the progression of the majority of patients to the terminal stage. The primary model output was incremental cost–effectiveness ratios (ICERs). Secondary model outputs were overall costs; life-years (LYs); quality-adjusted life-years (QALYs). According to Neumann et al. [32], the willingness-to-pay (WTP) threshold in USA is assumed to be $150,000 per QALY gained. According to Cameron et al. [33], Tzanetakos et al. [34] and the “China Guidelines for Pharmacoeconomic Evaluations (2020)” [29], the WTP threshold in China is $35,897 per QALY (three-times the Gross Domestic Product per capita in 2022).

Figure 1.

Schematic diagram of three state partitioned survival model.

PD: Progressed disease; PFS: Progression-free survival.

1.6. Sensitivity analysis

The model results were examined for robustness through sensitivity analysis (SA). One-way SA was carried out by adjusting each input over predefined ranges, which were derived from published data or by presuming an adjustment of ±20% from the initial estimates, consistent with previously published literature [16,17]. Due to the scarcity of independent research reports on the health state utility of advanced BTC, only small-sample studies had suggested the utility value of 0.55 for PFS state and 0.541 for the PD state [35]. Therefore, we expanded the range of one-way SA for utility values to ±30%. 10,000 samples of second-order Monte Carlo simulations were employed with each key parameter varied simultaneously under fixed patterns of distributions in the probabilistic SA (PSA). Main parameters for this analysis are detailed in Table 1, while Supplementary Table S2 contains the remaining parameters [11–20,25–29,36,37].

Table 1.

Parameters of partitioned survival model.

Variable	Baseline value	Range		Distribution	Ref.
		Minimum	Maximum
Drugs cost in the USA, $/per mg
Durvalumab	7.976	6.381	9.572	Gamma	[11]
Pembrolizumab	54.811	43.849	65.773	Gamma	[11]
Cisplatin	0.317	0.254	0.381	Gamma	[11]
Gemcitabine	0.019	0.015	0.022	Gamma	[11]
Oxalaplatin	0.136	0.109	0.163	Gamma	[11]
Leucovorin calcium	0.094	0.075	0.113	Gamma	[11]
Fluorouracil	0.004	0.003	0.004	Gamma	[11]
Irinotecan	0.150	0.120	0.180	Gamma	[11]
Capecitabin	0.005	0.004	0.006	Gamma	[11]
Regorafenib	5.543	4.434	6.651	Gamma	[12]
Drugs cost in China, $/per mg
Durvalumab	5.051	4.041	6.061	Gamma	[13]
Pembrolizumab	25.019	20.015	30.022	Gamma	[13]
Cisplatin	0.120	0.096	0.144	Gamma	[13]
Gemcitabine	0.042	0.033	0.050	Gamma	[13]
Oxalaplatin	0.550	0.440	0.660	Gamma	[13]
Leucovorin calcium	0.021	0.017	0.026	Gamma	[13]
Fluorouracil	0.040	0.032	0.048	Gamma	[13]
Irinotecan	1.152	0.922	1.383	Gamma	[13]
Capecitabin	0.001	0.001	0.001	Gamma	[13]
Regorafenib	0.602	0.482	0.722	Gamma	[13]
Drug administration cost in the USA, $
Chemotherapy infusion 1 Hour	132.160	105.728	158.592	Gamma	[14]
Chemotherapy infusion additional hour	28.470	22.776	34.164	Gamma	[14]
Drug administration cost in China, $	19.860	15.888	23.832	Gamma	[16]
Disease management cost in the USA, $/per cycle
Routine follow-up	61.740	49.392	74.088	Gamma	[17]
Laboratory testing	265.460	212.368	318.552	Gamma	[18]
Tumor imaging	194.670	155.736	233.604	Gamma	[18]
Disease management cost in China, $/per cycle
Routine follow-up	20.412	16.330	24.494	Gamma	[16]
Laboratory testing	42.865	34.292	51.438	Gamma	[19]
Tumor imaging	91.839	73.472	110.207	Gamma	[19]
Monthly BSC cost in the USA, $	747.830	598.264	897.396	Gamma	[17]
Monthly BSC cost in China, $	297.478	237.982	356.973	Gamma	[20]
Terminal care cost in the USA, $/per patient	11074.180	8859.344	13289.016	Gamma	[17]
Terminal care cost in China, $/per patient	2003.578	1602.862	2404.293	Gamma	[20]
AEs cost in the USA, $/per unit
Neutrophil count decreased/neutropenia	11935.550	9548.440	14322.660	Gamma	[15]
Anemia	2917.000	2333.600	3500.400	Gamma	[15]
Platelet count decreased/thrombocytopenia	6785.240	5428.192	8142.288	Gamma	[15]
White blood cell count decreased	11935.550	9548.440	14322.660	Gamma	[15]
AEs cost in China, $/per unit
Neutrophil count decreased/neutropenia	454.260	363.408	545.112	Gamma	[16]
Anemia	336.630	269.304	403.956	Gamma	[16]
Platelet count decreased/thrombocytopenia	1523.820	1219.056	1828.584	Gamma	[16]
White blood cell count decreased	454.260	363.408	545.112	Gamma	[16]
Utility
PFS	0.760	0.532	0.988	Beta	[25]
PD	0.680	0.476	0.884	Beta	[25]
Disutility due to AEs
Neutrophil count decreased/neutropenia	-0.090	-0.063	-0.117	Beta	[26]
Anaemia	-0.125	-0.088	-0.163	Beta	[27]
Platelet count decreased/thrombocytopenia	-0.090	-0.063	-0.117	Beta	Assumption
White blood cell count decreased	-0.090	-0.063	-0.117	Beta	Assumption
Discount rate in the USA	0.030	0.000	0.050	Uniform	[28]
Discount rate in China	0.050	0.000	0.080	Uniform	[29]
BSA in the US, m2	1.820	1.456	2.184	Gamma	[36]
BSA in China, m2	1.650	1.320	1.980	Gamma	[37]

AE: Adverse event; BSA: Body surface area; BSC: Best supportive care; GP: Gemcitabine and Cisplatin; PFS: Progression-free survival; PD: Progressed disease.

2. Results

2.1. Network meta-analysis

A total of two relevant Phase III RCTs (TOPAZ-1 and KEYNOTE-966) were included in this network meta-analysis by rigorous screening (Supplementary Figure S1). Log cumulative hazards plots (Supplementary Figure S2) confirmed a violation of the assumption of the proportional hazards. The second-order FP model with a power parameter of (-2, -2) was ultimately chosen for both OS and PFS, the DIC results and the best fitted curves are given in Supplementary Table S3 & Supplementary Figure S3. The first-order FP model (power parameter = 1) for PFS, despite having a smaller DIC, was excluded from consideration due to the clear deviation of clinical reality observed in the fitted survival curve. Finally, the survival functions for chemotherapy were determined to be log-logistic distribution for OS and log-normal distribution for PFS, respectively (Supplementary Table S4).

2.2. Base case analysis

The key economic outputs of the base case simulation are summarized in Table 2. The total LYs in the GP group, the PGP group and the DGP group were 1.46, 1.78 and 2.26 years, respectively. The total QALYs achieved were 1.05, 1.29 and 1.63, respectively. In USA, the overall costs in the GP group, the PGP group, and the DGP group were $44,874.62, $225,550.77 and $296,345.35, respectively. In China, the overall costs for each group were $15,595.48, $99,762.03 and $173,183.37, respectively.

Table 2.

Base case results.

Strategies	Cost	Incr cost	Cost breakdown		LYs	Incr LYs	ICER/LY	QALYs	Incr QALYs	Outcomes breakdown		ICER/QALY
			PF cost	PD cost						PF QALYs	PD QALYs
US payers
GP	44874.62		7663.10	37211.53	1.46			1.05		0.55	0.50
Pembrolizumab + GP	225550.77	180676.15	193739.63	31811.14	1.78	0.32	566921.10	1.29	0.24	0.74	0.55	761371.37
Durvalumab + GP	296345.35	70794.57	259456.62	36888.93	2.26	0.48	148059.51	1.63	0.34	0.91	0.72	206222.53
Chinese payers
GP	15595.48		3330.84	12264.64	1.46			1.05		0.55	0.50
Pembrolizumab + GP	99762.03	84166.55	88205.70	11556.33	1.78	0.32	264095.68	1.29	0.24	0.74	0.55	354678.79
Durvalumab + GP	173183.37	73421.34	161751.56	11431.81	2.26	0.48	153553.11	1.63	0.34	0.91	0.72	213874.22

GP: Gemcitabine and cisplatin; ICER: Incremental cost–effectiveness ratio; Incr: Incremental; LY: Life-year; PD: Progressed disease; PF: Progression-free; QALY: Quality-adjusted life-year.

Further analysis showed that compared with the GP arm, the PGP arm resulted in an ICER of $566,921.10/LY or $761,371.37/QALY in USA and $264,095.68/LY or $354,678.79/QALY in China. In comparison to the PGP group, the DGP group attended an ICER of $148,059.51/LY or $206,222.53/QALY in USA and $153,553.11/LY or $213,874.22/QALY in China. Both ICERs transcended $150,000/QALY and $35,897/QALY, indicating that in terms of economics, GP is the best option, with PGP second and DGP last.

2.3. Sensitivity analysis

One-way SA yielded an ICER for PGP versus GP from $607,790.88/QALY to $1,011,491.52/QALY and $284,362.14/QALY to $471,195.27/QALY in USA and China, respectively. One-way SA generated an ICER for DGP versus PGP from $62,539.45/QALY to $349,905.61/QALY and $122,887.67/QALY to $304,860.78/QALY in USA and China, respectively. The variables showing the most impact on the ICERs of PGP versus GP were utility of PFS, the cost of pembrolizumab, and utility related to PD. The variables showing the most impact on the ICERs of DGP versus PGP were the cost of durvalumab and pembrolizumab, utility related to PFS. The effect of parameter variation on the ICERs is presented in tornado diagrams (Supplementary Figure S4).

PSA demonstrated that, when compared with GP in the US and China, the chances of PGP being cost effective were 0 and 0% at the pre-set threshold of $150,000/QALY and $35,897/QALY, respectively. PGP was preferred to GP in 50% of simulations if the WTP threshold was approximately $754,200/QALY and $350,320/QALY in the US and China, respectively. PSA revealed that the likelihoods of the DGP group being cost effective were 27.04 and 0.02% at the threshold of $150,000/QALY and $35,897/QALY compared with PGP, respectively. The DGP group was favored to PGP in 50% of simulations if the WTP threshold was approximately $203,480/QALY and $211,000/QALY in USA and China, respectively. Cost–effectiveness acceptability curves (Figure 2) were created to illustrate the possibility of individual strategy being supported at different WTP thresholds.

Figure 2.

Cost–effectiveness acceptability curves. (A) In the USA. (B) In China.

CE: Cost–effectiveness; GP: Gemcitabine and cisplatin.

3. Discussion

Conventional NMA of HRs fails to capture the delayed onset of therapeutic effect or long-term survival discovered in PD-L1/PD-1 inhibitor studies [38,39]. The relative advantage in the DGP arm appeared to increase with extended follow-up in TOPAZ-1, where the OS curves crossed and did not split supporting the DGP arm until around month 6 of treatment [5]. Furthermore, nonproportional hazards were verified in the chosen trials, which has not been discussed by previous studies. The time-varying NMA by fitting FP model relaxed the PH assumption. This analysis is necessary, because if the assumption of the proportional hazards is not established when applying the proportional hazard model, it will lead to serious survival fitting bias.

In the formulation of treatment strategies, clinicians require the best available evidence. This research revealed that the effectiveness and cost were ordered as below: DGP > PGP > GP. Cost–effectiveness assessments help resolve the dilemma of balancing between the effectiveness and cost of pharmaceutical technology. The sequence of cost–effectiveness stood as follows: DGP < PGP < GP. GP is currently the most cost-effective treatment for those with advanced BTC, despite the fact that combination therapies (DGP, PGP) have significantly enhanced patient survival benefits. Extensive SA confirmed the base case results.

To the best of our knowledge, this research marks the first examination of the cost–effectiveness of pembrolizumab combined with GP versus GP for the treatment of advanced BTC from USA and Chinese payer perspective and no pharmacoeconomic evaluation simultaneously comparing DGP, PGP and GP. Several studies have reported the cost effectiveness of DGP, PGP and GP among individuals with advanced BTC but were limited to the comparison of two drugs [37,40–42]. According to the research of Zhuomiao et al. [37], durvalumab combined with GP is not a cost-effective alternative for the primary treatment of BTC compared with GP alone from the standpoint of payers, both in China and USA. The study of Zhiwei et al. [40] found that, when assessed from the perspective of the Chinese healthcare system, combining pembrolizumab with GP as a first-line treatment option for advanced BTC does not seem to be a cost-effective approach compared with GP. Our analysis was consistent with analyses of Zhuomiao et al. [37] and Zhiwei et al. [40]. However, the evaluation of Youwen et al. [42] showed that from the standpoint of healthcare systems, first-line DGP versus PGP was a cost-effective option for patients with advanced BTC in China, but not in USA. As for the difference between the study of Youwen et al. and this study the following two reasons may explain. First, the effectiveness of DGP versus PGP were based on traditional network meta-analysis. Second, the perspectives of the two studies were different.

This research had several potential limitations. First, although TOPAZ-1 trial and KEYNOTE-966 trial are currently the large-scale and high internal validity Phase III trial for the first-line treatment of advanced BTC, as with other trial-based economic evaluations, any bias in the trials will be reflected in this study. In addition, clinical data from the TOPAZ-1 trial and KEYNOTE-966 trial should be interpreted in the context of strictly screened patients and relatively short follow-up periods and may not mirror the actual long-term outcomes of such regimens. Second, health utility values were retrieved from the relevant literature. Therefore, estimations of health utility might not be perfect. While SA of utilities have been performed, these findings may be insensitive to the utilities. The findings in this report should verified when real-world long-term data and ideal utility data become available. Third, costs for grade 1 or 2 TrAEs were not included in this report. However, according to the one-way sensitivity simulations, costs related to TrAEs had a negligible influence on the ICERs. Thus, these findings may be insensitive to the disutilities and costs for grade 1 or 2 TrAEs. Fourth, considering the difference in cost, efficacy and perspective, findings in the evaluation are unlikely to be generalizable.

4. Conclusion

From the perspectives of US and Chinese payers, adding durvalumab or pembrolizumab to the GP regimen is unlikely to be cost effective compared with GP alone in the first-line setting for advanced BTC, given the willingness-to-pay thresholds of $150,000 per QALY and $35,897 per QALY, respectively.

Supplemental material

Supplemental data for this article can be accessed at https://doi.org/10.1080/1750743X.2024.2347822

Author contributions

R Liu and Y Zhao developed the economic model and performed the analyses. R Liu and Y Zhao interpreted the results and wrote the draft of the manuscript. R Liu, Y Zhao, F Shi, J Zhu and M Huang reviewed, analyzed and interpreted the data. K Qiu contributed to the design of the primary model and the interpretation of the results. All authors reviewed and approved the final manuscript.

Financial disclosure

The authors have no financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

Competing interests disclosure

The authors have no competing interests or relevant affiliations with any organization or entity with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, stock ownership or options and expert testimony.

Writing disclosure

No writing assistance was utilized in the production of this manuscript.