The target population for this analysis was patients who had locally advanced or metastatic oesophageal squamous cell carcinoma and did not receive previous systemic treatment, consistent with the patient characteristics in the ORIENT-15 trial (12). A partitioned survival model with 3 health states was developed to conduct the economic analyses from a Chinese healthcare system perspective, and the perspective is consistent with the objective of this study (to estimate the cost-effectiveness of sintilimab plus chemotherapy and help decision-makers in healthcare system to optimally allocate limited health resources) (14). The three mutually exclusive health states were progression-free survival (PFS), progressed disease (PD), and death ( Figure 1 ). In the model, the proportion of patients in each health state at each cycle t was estimated from OS and PFS curves of the ORIENT-15 trial (12). The proportion of patients alive at each cycle (1-week cycle) was estimated by the area under the OS curve, and the proportion alive with PFS was estimated by the area under the PFS curve (15). The proportion of PD was estimated by the difference between the OS and PFS curves (15). The time horizon was 11 years given that more than 99% of the patients died at this time point, and 5 to 20 years were included in the scenario analyses.

This study followed the reporting guideline of Consolidated Health Economic Evaluation Reporting Standards (CHEERS) ( Supplementary Table 1 ) (16). This evaluation was based on a literature review of publicly available data and on modeling techniques, which did not require institutional review board review or exemption by the ethics committee.

To construct the partitioned survival model, graphic data from the ORIENT-15 trial were extracted by GetData Graph Digitizer, version 2.18, and time-to-event data were obtained as described in the study by Guyot et al. (17). The original and reconstructed Kaplan–Meier curves were shown in Supplementary Figure 1 , and summary measures regarding median duration and survival rate in original publication and reconstructed data were shown in Supplementary Table 2 . Because time horizon in this economic study was 11 years (99% of the patients died) beyond the follow-up period in ORIENT-15 trial, the OS and PFS curves of chemotherapy alone were fitted and extrapolated using the following parametric survival functions: Weibull, log-normal, log-logistic, exponential, generalized gamma, Gompertz, and Royston-Parmar spline models and parametric mixture and non-mixture cure models (18–20). The eligible survival function was chosen based on the visual inspection, and lowest value of the Akaike information criterion (AIC) and Bayesian information criterion (BIC) (18). The proportional hazards (PH) modelling method was used to estimate the parametric PFS and OS curves for the sintilimab plus chemotherapy group based on the hazard ratios (HRs) of sintilimab plus chemotherapy vs chemotherapy alone reported in ORIENT-15 trial (12). This could avoid the bias resulted from inconsistency of parametric survival distribution between two groups (18).

The final survival functions of the sintilimab plus chemotherapy vs chemotherapy alone were shown in Table 1 and the goodness-of-fit results were shown in Supplementary Table 3 . The proportions of patients with PFS and OS were calculated by using the selected survival distribution. The validation plot, survival distribution, were shown in Supplementary Figure 2 , 3 . The key clinical inputs were listed in Table 1 .

Only direct medical costs within the healthcare system were included in this study ( Table 1 ), including costs of acquiring and administrating drugs, costs of subsequent treatment after disease progressed, costs for routine follow-up, laboratory tests and scans, costs for end-of-life care and costs for the management of AEs. All costs were reported in 2021 US dollars with the exchange rate in 2021: $1 = ¥6.45, and were adjusted to 2021 value using the consumer price index (26).

According to the ORIENT-15 trial report, sintilimab was given intravenously at a dose of 200 mg every 3 weeks, and it was continued until progressive disease, lost to follow-up, death, or a maximum of 2 years, whichever occurred first (12). And 6.53% patients would receive cisplatin (75 mg/m² every 3 weeks) plus paclitaxel (87.5 mg/m² on the first and second weeks; 175 mg/m² every 3 weeks starting from the fourth week) as chemotherapy regimen, and other patients (93.47%) would receive cisplatin plus 5-fluorouracil (4000 mg/m² every 3 weeks) as chemotherapy regimen (12). Chemotherapy was given within a maximum of 18 weeks (12). To calculate the dosage of paclitaxel, we assumed that the body surface area was 1.72 m² for a typical patient in China (23). The prices of sintilimab, cisplatin, paclitaxel, and 5-fluorouracil were collected from Chinese bid-winning price (21).

After disease progression, 41.0% patients in the sintilimab plus chemotherapy group and 54% in the chemotherapy alone group received subsequent systemic medication therapy, and the proportion and costs for each subsequent therapy were shown in Supplementary Table 4 , which were estimated from ORIENT-15 trial and Chinese bid-winning price (21). The costs of laboratory tests and scans ($134.93 per cycle), costs of subsequent best supportive care ($39.03 per cycle), cost of routine follow-up ($17.16 per cycle) and cost of terminal care in end-of-life ($1,460.30 per patient) were collected from an economic evaluation among patients with advanced OSCC. The analysis included the costs associated with management of grade 3 or higher AEs (probability ≥ 5%), which were extracted from the literature ( Supplementary Table 5 ).

Each health state was assigned a health utility preference on a scale of 0 (death) to 1 (perfect health). The PFS and PD states associated with OSCC were 0.74 and 0.58 respectively, which were derived from a cost-effectiveness analysis considering patients with OSCC (22). The disutility values due to grade 1 or 2 and grade 3 or higher AEs were included in this analysis (24). All AEs were assumed to be incurred during the first cycle.

The incremental cost-effectiveness ratio (ICER) was calculated as the incremental cost per additional quality adjusted life-year (QALY) gained between the sintilimab plus chemotherapy group and the chemotherapy alone group. When the ICER was lower than the prespecified willingness-to-pay threshold (WTP, 3 times GDP per capita per QALY gained, $37,663.26/QALY), cost-effectiveness was assumed according to the recommendation of Guidelines for Evaluation of Chinese Pharmacoeconomics (25). Costs and QALYs were discounted at an annual rate of 5% (25).

The incremental net health benefit (INHB) and incremental net monetary benefits (INMB) were also estimated based on the following formulas: INHB (λ) = (μ _{E 1} − μ _{E 0}) − (μ _{C 1} − μ _{C 0})/λ = ΔE − ΔC/λ; INMB (λ) = (μ _{E 1} − μ _{E 0}) × λ − (μ _{C 1} − μ _{C 0}) = ΔE × λ − ΔC, where μ _Ci and μ _Ei were the costs and effectiveness of sintilimab plus chemotherapy (i = 1) or chemotherapy (i = 0), respectively, and λ was the WTP threshold (27, 28).

To evaluate the robustness of the base-case results, we conducted one-way and probabilistic sensitivity analyses. One-way sensitivity analyses were conducted for all parameters, and the estimated range of each parameter was based on either the reported or estimated 95% CIs in the referenced studies or determined by assuming a 25% change from the base-case value ( Table 1 ). In the probabilistic sensitivity analyses, a Monte Carlo simulation with 1000 iterations was generated by simultaneously sampling the key model parameters from the prespecified distributions. A gamma distribution was selected for the cost parameters, a log-normal distribution for the HRs, and a beta distribution for proportion, and health utility parameters. Based on the data from 1000 iterations, a cost-effectiveness acceptability curve was created to represent the probability that sintilimab plus chemotherapy would be considered cost-effective at various WTP thresholds for health gains (QALYs). To investigate the uncertainty of economic outcomes caused by the subpopulations, exploratory subgroup analyses were performed for the prespecified subgroups that were reported in the ORIENT-15 trial by varying the HRs for OS and PFS. Lastly, additional scenario analyses were conducted: 1) time horizon was 5 or 20 years; 2) four survival curves (OS and PFS curves for sintilimab plus chemotherapy vs chemotherapy alone) were fitted separately to estimate parametric survival curve. Survival curves were fitted with flexsurv packages in R, version 4.1.1, 2021, and the partitioned survival model was developed in Microsoft Excel 2019.

In comparison with chemotherapy alone, sintilimab plus chemotherapy provided an additional 0.37 QALYs and 0.50 overall life-years, with an incremental cost of $8,046.58, which was associated with an ICER of $21,782.24/QALY. The INHB was +0.16 QALYs, and the INMB was +$5,866.61 at a WTP threshold of 3 times GDP per capita per QALY ( Table 2 ).

One-way sensitivity analyses showed that the model was particularly sensitive to utility of PFS and the cost of sintilimb ( Figure 2 ). When the lower boundary of the utility (0.56) for PFS was adopted, the ICER of sintilimab plus chemotherapy vs chemotherapy alone was $28,892.78/QALY, and when the upper boundary (0.93) was adopted, the ICER was $17,480.31/QALY. When the cost of sintilimab was discounted by 25%, the ICER was $18,402.56/QALY. The remaining parameters, such as the HR (for PFS and OS), cost and utility related to AEs, had only moderate or low associations with the ICER. All parameters were not associated with ICERs exceeding the threshold.

Compared with chemotherapy alone, the probabilistic sensitivity analysis showed that sintilimab plus chemotherapy added a mean of 0.40 QALYs with an additional mean cost of $8.560.06, which resulted in a mean ICER of $21,621.89/QALY ( Figure 3A ). The cost-effectiveness acceptability curve showed that the probability of sintilimab plus chemotherapy being cost-effective was 0.01%, 76.80% and 98.60% at 1, 2- or 3-times GDP per capita per QALY threshold, respectively ( Figure 3B ).

The subgroup analyses, which were conducted by varying the HRs for OS and PFS, revealed that sintilimab plus chemotherapy was associated with primarily positive INHBs in the most subgroups at the threshold of 3-times GDP per capita per QALY ( Figure 4 ). Only among patients with CPS<1, the point estimate of INHBs was negative, and in fewer subgroups (women, local advanced, cisplatin plus 5-fluorouracil, CPS<1), the lower bound of INHBs by varying the HRs for OS were negative.

When the time horizon was 5 or 20 years, the overall cost of sintilimab plus chemotherapy was $23,671.40 or $24,909.54, resulting in an ICER of sintilimab plus chemotherapy vs chemotherapy alone of $ 22,919.51 or $21,680.02 per QALY. When four survival curves (OS and PFS curves for sintilimab plus chemotherapy vs chemotherapy alone) were fitted separately, the overall cost of sintilimab plus chemotherapy was $26,784.74, resulting in an ICER of $18,329.16/QALY ( Table 3 ).

The strengths of this study are worth highlighting. To the best of our knowledge, this study is the first to evaluate the economic outcomes of sintilimab plus chemotherapy in the first-line treatment for advanced or metastatic OSCC by integrating the latest evidence through a pharmacoeconomic modeling technology. Besides, the partitioned survival model is employed in this economic analysis for the advantages that survival curve can be directly used to obtain the proportion of patients with different health states, and complex risk functions can be directly reconstructed by extrapolation method, so the calculation is relatively simple. Furthermore, there is no need to make additional assumptions on the model to calculate the probability of transition between states, so it is more suitable for the actual survival of patients.

There are also several limitations in our study. First, we only evaluated the cost-effectiveness of sintilimab plus chemotherapy vs chemotherapy alone for patients with advanced OSCC. In fact, the cost-effectiveness of sintilimab vs similar drugs (other ICIs) should also be assessed. However, due to the lack of robust head-to-head trial data, we could not compare the economics of other immunotherapies with that of sintilimab directly, such as pembrolizumab, camrelizumab, and nivolumab, which had also shown favorable health benefits for patients with advanced OSCC. Therefore, future research will perform network meta-analyses to obtain HRs for OS and PFS between different ICIs regimens, subsequently achieving cost-effectiveness comparisons. Second, though ORIENT-15 is a large and well-designed trial, any biases such as rigorous eligible patients, high medication adherence and racial difference, can affect its validity and generalizability. Thus, there’s a difference in generalizing our cost-effectiveness analysis results outside the Chinese setting. Third, clinical benefits beyond the observation time of the ORIENT-15 trial were assumed through fitting of parametric distributions to the reported Kaplan-Meier PFS and OS data, which may lead to uncertainty in the model outcomes, although flexible parametric models were applied and the modeled and observed data were validated. However, we have conducted a series of sensitivity, subgroup and scenario analyses to evaluate the uncertainty, and they showed the model results were robust. Fourth, the costs of grade 1 and 2 AEs were excluded, which might result in the overestimation of the economic results associated with sintilimab plus chemotherapy. Lastly, the disutility value for AEs extracted from a study outside Chinese setting, may also cause some bias. These limitations may not be major factors, as showed in the one-way sensitivity analysis indicating that the influence of the costs/disutility values of AEs were minor.