This study was reported following the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) reporting guideline (Husereau et al., 2022). This economic evaluation was conducted from the perspective of the Chinese healthcare system. The results can be generalized to the situation in China since 75% of the patients in RATIONALE-306 were Asians. A partitioned survival model with three states was constructed for this cost-effectiveness analysis (Williams et al., 2017). The model included three health states: progression-free survival (PFS), progressed disease (PD), and death (D) (Figure 1). Patients in the model move over time from one health state to another, with the probabilities of these transitions being determined by survival functions. These survival functions are partitioned. The time horizon was set as 10 years so that 99% of the patients died in both treatment arms, and the cycle length was set as 21 days (Shao et al., 2022). The primary outcomes were cost, quality-adjusted life-years (QALYs), and the incremental cost-effectiveness ratio (ICER). Both costs and QALYs were discounted at an annual rate of 5% (Guoen, 2020). The willingness-to-pay (WTP) threshold was set as $37,260 (three times the Chinese GDP per capita in 2022) as per the QALY gained (National Bureau of Statistics, 2021). The statistical analysis was conducted through R 4.1.2 (https://www.r-project.org/), and the economic evaluation model was constructed using Microsoft Excel 2019 (Redmond, Washington, United States).

We included patients who were ≥18 years of age with histologically confirmed advanced or metastatic OSCC. Other key characteristics were the same as in the RATIONALE-306 trial (Xu J. et al., 2023). The included patients received tislelizumab or placebo intravenously at a dose of 200 mg every 3 weeks (one cycle). Chemotherapy included cisplatin (75 mg/m² every cycle) or oxaliplatin (130 mg/m² every cycle) plus paclitaxel (175 mg/m² every cycle) or 5-fluorouracil (750–800 mg/m² continuous administration for 5 days in each cycle) or capecitabine (1,000 mg/m² orally twice daily on days 1–14) (Xu J. et al., 2023). Note that oxaliplatin substitution was not permitted in the treatment guidelines in China (Xu J. et al., 2023). Therefore, only cisplatin was considered in this study. According to the chemotherapy exposure in the patients in the RATIONALE-306 trial, we considered a maximum of six cycles of chemotherapy (Xu J. et al., 2023). If the disease progressed or intolerable toxicity occurred, the patients could alter from the first-line treatments to the second-line treatments. In RATIONALE-306, 48% (157/326) patients in the tislelizumab plus chemotherapy group and 55% (177/323) in the placebo plus chemotherapy group received subsequent systemic anticancer therapies; subsequent immunotherapy was provided to 46 (14%) and 72 (22%) patients, respectively. However, detailed information on the subsequent therapies was not reported (Xu J. et al., 2023). Therefore, according to the recent Chinese guideline from the Chinese Society of Clinical Oncology, we assumed the subsequent therapies as the best supportive care (BSC), camrelizumab (immunotherapy), and docetaxel (chemotherapy) (Wang and Han, 2020).

We extracted the PFS and OS data through GetData Graph Digitizer (http://getdata-graph-digitizer.com) from the Kaplan–Meier (KM) curves of the RATIONALE-306 trial because we did not have the individual patient data (IPD). Then, the method of Guyot et al. was used to reconstruct the IPD (Guyot et al., 2012). We validate our reconstructed models by comparing them with the original KM curves, and the results showed that the generated PFS and OS curves closely resemble those provided in the RATIONALE-306 trial. We considered the following parametric survival models: exponential, Weibull, Gompertz, gamma, log-logistic, log-normal, and generalized gamma models (Shao et al., 2023). The minimum of the Akaike information criterion (AIC) goodness-of-fit statistics and visual inspection were used to determine the best fit of the PFS and OS curves (Jrgensen, 2004; Djalalov et al., 2019). The goodness-of-fit results are shown in Supplementary Tables S1 and S2. The validation plot and survival distribution are also shown in Supplementary Figures S1–S4. The detailed clinical information is given in Table 1.

We only considered direct medical costs in this study, which included costs of acquiring drugs, costs of routine follow-up, costs for the management of adverse events (AEs), and costs for end-of-life care (EOL). Drug prices in 2022 were obtained from two Chinese public databases health data platforms (MENET, 2022; Yaoch, 2022). Follow-up costs and EOL costs were obtained from one published article (Zhang et al., 2021). The prices in RMB were exchanged for US$ with the exchange rate of 6.90 (6 May 2022). Only severe AEs (≥grade 3) with rates over 5% were considered in this study, which included decreased neutrophil count, decreased white blood cell count, anemia, neutropenia, hypokalemia, and hyponatremia. The utilities of PFS and PD states associated with advanced OSCC were 0.75 and 0.67, respectively, which were derived from a published economic evaluation (Marguet et al., 2021). The disutility values according to AEs were included in this analysis. These disutility values were extracted from other published studies (Shao et al., 2022; Zhou et al., 2022). We assumed that all AEs were incurred during the first cycle (Shao et al., 2022). Then, we subtracted the duration-adjusted disutility from the baseline PFS utility. All cost-related and utility-related parameters are shown in Table 1.

We conducted deterministic sensitivity analysis (DSA) and probabilistic sensitivity analysis (PSA) to test the robustness of the economic evaluation model. In DSA, all parameters were adjusted within the reported 95% confidence intervals (CIs). For parameters not reporting confidence intervals, we assumed a fluctuation of ±20% based on the base-case values. In PSA, a gamma distribution was selected for cost and a beta distribution was selected for probability, proportion, and utility. Scatter plots describing ICER and the cost-effectiveness acceptable curve (CEAC) were obtained by running 10,000 Monte Carlo simulations. In the subgroup analysis, we analyzed the results for patients with the tumor area positivity (TAP) score of ≥10% and <10% for the expression of PD-L1 through the methods of base-case analysis since the RATIONALE-306 trial reported the independent KM curves for OS. Since KM curves for PFS were not reported, we assumed the same PFS as the base-case analysis.

Base-case analysis results including cost, QALYs, and ICERs are shown in Table 2. Tislelizumab plus chemotherapy gained 1.38 QALYs at a cost of $34,965.45, while placebo plus chemotherapy yielded 0.9 QALYs and cost $18,378.25. Compared with the placebo, the tislelizumab combination therapy gained incremental QALYs of 0.48 with an incremental cost of $16,587.2. Thus, the ICER between the placebo plus chemotherapy group and the tislelizumab plus chemotherapy group was $34,699.72 per QALY gained, which was slightly lower than the given WTP of $37,260 per QALY gained.

Subgroup analysis results including cost, QALYs, and ICERs are shown in Table 2. In the subgroup analysis of patients with a TAP score ≥10%, the tislelizumab combination therapy gained an additional 0.49 QALYs at a cost of additional $16907.48 when compared with placebo plus chemotherapy. The ICER was $34,553.83 per QALY gained, which was similar to the base-case analysis results. However, in the subgroup analysis of patients with a TAP score <10%, the tislelizumab combination therapy yielded an incremental QALY of 0.31 with an incremental cost of $12,847.78. The ICER remained $41,848.57 per QALY gained, which meant that the novel therapy was not cost-effective anymore.

The tornado diagram showing the results of DSA is shown in Figure 2. Utilities of PFS and PD were significantly associated with the model outcomes, which even led to the modeled ICERs, exceeding the given WTP threshold. In addition, base-case analysis results were also sensitive to the cost of salvage therapy, laboratory tests and radiological examinations, and tislelizumab. Some other parameters including the subsequent treatment proportions had a slight impact on the model outcomes.

The results of PSA are shown in Figures 3 and 4. According to the cost-effectiveness scatter plot (Figure 3), when WTP was set as $37,260 per QALY, it was obvious that most ICER points were placed under the line of WTP. In addition, the combination therapy had a probability of 77% to be cost-effective. According to the cost-effectiveness acceptability curve (Figure 4), we found that when the WTP threshold was set lower than $34,650 per QALY, tislelizumab plus chemotherapy were unlikely to be cost-effective. When the WTP threshold surpassed $30,300 per QALY, tislelizumab plus chemotherapy had a >50% probability to be cost-effective.