The cost-effectiveness model was developed using Microsoft Excel^® (Redmond, WA, United States of America) with three health states: progression-free survival (PFS), progressive disease (PD), and death (see Figure 1). A partitioned survival model (PSM) was developed to simulate disease progression in patients with advanced ALK + NSCLC who had not previously received systemic therapy. The distribution of patients across different health states over time was estimated by calculating the area under the survival curve. At each cycle, patients may either remain in the previous state, progress to a disease progression state, or enter the death state, also known as the absorption state.

Patients received iruplinalkib (60 mg once daily for the first 7 days, followed by an increased dosage of 180 mg once daily thereafter) or crizotinib (250 mg twice daily) until disease progression. Subsequent treatment strategies were guided by clinical guidelines and medication data collected in the INSPIRE trial. The model was conducted with 4-week cycles over a 25-year horizon to approximate a lifetime perspective. Patients in the INSPIRE trial entered the model at a mean age of 55 years; thus, the horizon extends to approximately 80 years, exceeding the 2021 Chinese life expectancy (78.2 years). By the end of the model, cumulative mortality of both groups was approximately 90% (iruplinalkib) and 93% (crizotinib), indicating negligible residual survival. Therefore, extending the time horizon further would have minimal impact on model outcomes. In the base-case analysis, both costs and utilities were discounted at 5% annually, as recommended by the China Guidelines for Pharmacoeconomic Evaluation (Liu et al., 2020). Costs from previous years were adjusted using the Consumer Price Index (CPI) and reported in 2024 USD dollars (1 USD = 7.12 CNY). The primary outcomes of this study included costs, quality-adjusted life years (QALYs), life years (LYs), and incremental cost-effectiveness ratios (ICERs). The willingness-to-pay (WTP) threshold was defined as 1-3 times the per capita Gross Domestic Product (GDP) of China in 2024, estimated at $13,452.37 (¥95,749). Additionally, a treatment is considered highly cost-effective if the ICER is less than one times the GDP per capita.

The PFS and overall survival (OS) data for iruplinalkib and crizotinib were obtained from the INSPIRE study (Shi et al., 2024). Parametric survival functions were fitted to the Kaplan-Meier data assessed by an independent review committee (Latimer, 2011). Based on the clinical rationale, visual fit, and statistical goodness-of-fit [Akaike information criterion (AIC) and Bayesian information criterion (BIC)], the log-logistic distribution was more appropriate for the PFS curve to fit for iruplinalkib and crizotinib. Although the AIC and BIC of crizotinib indicated that the log-normal distribution was the most appropriate and the log-logistic distribution was suboptimal, the distribution should remain the same for both groups, as recommended by the National Institute for Health and Care Excellence (NICE) Technical Support Document 14²⁶. The exponential distributions were most appropriate for the OS curve to fit for iruplinalkib and crizotinib. (AIC and BIC for PFS and OS are shown in the Electronic Supplementary Material (ESM) Table 1, the parametric survival curve fits are shown in Figure 2). Given that adverse events (AEs) significantly affect both costs and QALYs, their impact was incorporated into the model. Grade 3-5 adverse events and a frequency ≥5.0% in either treatment arm were included (see Table 1). In the cohort analysis, Chinese natural mortality rates were applied to the fitted survival curves to simulate patient survival more accurately (China Population Census Yearbook, 2020, 2020).

The INSPIRE trial did not include utility measurements; thus, utility values for PFS (0.804) and PD (0.321) were derived from the Chinese subgroup dataset in a study conducted by Nafees et al. (Table 1) (Nafees et al., 2017). Additionally, disutility related to AEs was included in the analysis. Disutility values were obtained from Nafees et al.’s study and a health technology assessment by NICE (see Table 1). In the calculation, the one-time QALY decrements due to AEs were determined by multiplying their disutility values by the corresponding incidence probabilities at the start of the first cycle.

From the perspective of the Chinese healthcare system, direct costs, including first-line and subsequent-line drugs, disease management across different disease stages, and management of AEs, were incorporated into the model. Treatment costs for each cycle were calculated based on the dosing schedule and unit costs from the MENET database and prices of drugs at local providers (see Table 1; Excellence, 2021; Yunnan Healthcare Security Administration, 2025). The cost of initial drug treatment in the PFS health state was calculated under the assumption that the patients were treated until disease progression or death. Since iruplinalkib and crizotinib are both oral drugs, administration costs were excluded. Costs per cycle in the PD health state were calculated as a weighted aggregate based on subsequent treatment regimens and their respective utilization proportions from the INSPIRE trial (see Table 1). AEs’ costs were calculated once in the first cycle. Most costs were derived from the published literature, and on this basis, opinions of clinical experts in lung cancer were considered (including 38 consultant oncologists specializing in lung cancer treatment from more than 30 grade A tertiary hospitals in 3 Chinese provinces) (see Table 1).

Specifically, the subsequent treatment regimens for the intention-to-treat (ITT) population in clinical trials have been prospectively documented for patients who progressed following first-line treatment. This study hypothesized that the second-line therapeutic agents administered to patients do not overlap with those used in first-line treatment. For example, patients receiving iruplinalkib as a first-line therapy would not receive the same drug in subsequent treatment lines. When processing subsequent-line treatment records, cases were excluded if the same drug was used in both first-line and second-line treatment, if the drugs used in subsequent treatment accounted for less than 1%, or if the subsequent-line drugs had not yet been approved for clinical use. In instances where treatment records did not specify the exact drugs but only described the treatment method (e.g., a combination of targeted therapy and chemotherapy), the cost per cycle of the treatment was calculated by summing the average price of the second-line targeted therapies and the per cycle chemotherapy cost obtained through consultation with clinical experts.

The proportion of patients receiving radiotherapy has been documented in the post-line treatment records of clinical trials. Only a subset of patients receiving chemotherapy after PD also receive concurrent radiotherapy, which overlaps with chemotherapy patients. Therefore, the proportion of patients receiving radiotherapy was calculated separately and not included in the calculation for patients receiving drug treatments. The per capita radiotherapy cost was derived from expert consultation. The total cost of radiotherapy in the model was estimated as a one-time value, calculated by multiplying the number of newly progressed patients by the radiotherapy proportion and the average per-patient cost. (see in the ESM Table 2).

Supportive disease management costs were also included in the PFS and PD health states, including outpatient and examination (see in the ESM Supplementary Table S3). The items and costs associated with disease management per cycle were sourced from the INSPIRE study and consultations with clinical experts. In the PD state, increased attention is required to monitor patients for potential brain metastases. Consequently, the per-cycle disease management cost of $418.07(¥2976.231) in the PD state is higher than that of $37.67 (¥268.123) in the PFS state.

This study also considers the cost of end-of-life care, assuming that all patients transitioning to the death state incur a one-time end-of-life care cost of ¥34,421.49 ($4,844.15). This cost parameter is derived from a study in China and has been adjusted to 2024 values using the CPI (see Table 1) (Li et al., 2018).

To address the uncertainty in the model, deterministic sensitivity analysis (DSA) and probabilistic sensitivity analysis (PSA) were conducted. Ranges were based on 95% CIs or varying the default input by ±10%. The PSA was performed by using a Monte Carlo simulation with 1000 simulations. Distributional assumptions were based on recommended guidelines (Andrew et al., 2006). Beta distributions were assigned for utilities of health states, and gamma distributions were assumed for costs. Full details of values, ranges, distribution types, and sources are reported (see Table 1) Two scenario analyses were also carried out. First, time horizons of 20 years and 30 years were set, respectively. The survival of patients has been extended due to advancements in ALK inhibitors (Christopoulos et al., 2021). By shortening or lengthening the study period, the cost-effectiveness and sustainability of iruplinalkib can be further assessed in both the short and long term. Second, the log-normal distribution was used to fit both the PFS curve of the two groups, because the AIC and BIC supported a log-normal distribution for the PFS curve of crizotinib.

The validity of the model was assessed using the Assessment of the Validation Status of Health-Economic Decision Models Checklist (AdViSHE). The checklist comprises five main components, including the conceptual model, the input data, the implemented software program, and the model outcomes. The conceptual model used in this study adhered to the methodologies recommended by NICE. The input data were reviewed by the Independent Review Committee (IRC), pharmacoeconomic experts, and clinical specialists. Survival functions were derived from individual patient data from clinical trials, and the best-fitting function was chosen based on the AIC and BIC. Additionally, Sensitivity analyses were conducted to include as many parameters as possible to minimize model uncertainty. The model was developed in Microsoft Excel (Redmond, WA, United States) and reviewed by all authors.

In the base-case analysis, over a 25-year time horizon, the results demonstrated that Iruplinalkib provided greater health benefits compared to crizotinib. In first-line treatment, the projected median PFS for iruplinalkib was 36.92 months compared with 12.92 months for crizotinib. Iruplinalkib was associated with a 0.50 increment in LYs compared with crizotinib. After accounting for health-related quality of life, first-line Iruplinalkib was estimated to result in 3.44 QALYs, which represented an increase of 2.11 QALYs compared with crizotinib.

The results of the base–case analysis are shown in Table 2 The longer PFS time with first-line iruplinalkib resulted in higher costs in the PFS state compared to crizotinib. The difference in first-line drug costs was $61,311.21(¥436,535.85), while the difference in disease management costs was $1,301.05 (¥9,263.49). Patients in the crizotinib group progressed to the PD state more rapidly, leading to higher drug costs during the PD state compared to iruplinalkib, with a cost difference of $46,480.22 (¥330,939.16). Additionally, the disease management costs in the PD state for patients receiving crizotinib were slightly higher than those of iruplinalkib, with a cost difference of $11,470.85 (¥81,672.45). The remaining cost components showed minimal differences between the two intervention groups (see Table 2). In total, treatment with iruplinalkib resulted in an increase of $4,325.55 (¥30,797.94) and an additional 1.48 QALYs, translating into an ICER of $2921.17(¥20,798.73) per QALY gained. The base-case ICER was below the WTP threshold of $13,447.89 (¥95,749) (1.0 times GDP per capita in China for 2024) per QALY gained (see Table 2).

In the DSA, the main model drivers were the cost of iruplinalkib, cost of alectinib, cost of ensartinib and discount rate estimates (see Figure 3). At the highest end of the cost of iruplinalkib (adjusted to $ 20.16(¥143.55) per 60 mg), in DSA, the ICER remained under a $3,060.41/QALY threshold (1 times GDP per capita in China for 2024), indicating the stability of the study results.

PTE_CT Positron emission tomography-computed tomography, PFS progression-free survival; A PSA was performed for the base-case analysis, and the cost-effectiveness acceptability curve (CEAC) is shown in Figure 4. Cost-effectiveness acceptability curve at different thresholds for willingness to pay. The PSA demonstrated that iruplinalkib had a 100% probability of being cost-effective at a WTP threshold of $13,447.89 (1.0 times GDP per capita in China for 2024)/QALY. The cost-effectiveness plane is shown in Figure 5 where all scatter points are located entirely below the WTP threshold, further supporting the robustness of the model and the cost-effectiveness of iruplinalkib.

In the first scenario analysis, which varied the time horizon to 20 and 30 years, the ICERs were $ 2403.32 (¥17,111.65) and $17,111.654 (¥¥22,728.35) per QALY, respectively. Both ICERs were below the WTP threshold of $13,447.89 (¥95,749) (1 times GDP per capita in China for 2024), indicating that iruplinalkib was cost-effective. In the second scenario analysis, where a log-normal function was used to fit the PFS curves of the two groups, the resulting ICERs were $9817.186 (¥¥69,898.36). The ICERs were still below WTP, indicating that iruplinalkib was cost-effective.