This analysis modeled a cohort of patients with posterior circulation AIS, reflecting the population enrolled in the EXPECTS trial (Yan et al., 2025). Participants were adults who arrived 4.5–24 h after symptom onset, including those with wake-up or unwitnessed strokes where the estimated onset time met this window. Neurological impairment was confirmed by an NIHSS score of at least 1, and brain imaging demonstrated a PC-ASPECTS of 7 or higher on CT or MRI, indicating limited early ischemic change. Diagnosis was supported by both imaging findings and clinical presentation. All had a pre-stroke mRS below 2 and were not candidates for endovascular intervention. Individuals were excluded for factors such as contraindications to alteplase, substantial hemorrhage risk, recent major surgery or trauma, infarction in the anterior circulation, pregnancy, or any condition likely to compromise treatment safety, protocol adherence, or follow-up.

In the alteplase group, patients received intravenous alteplase at 0.9 mg per kilogram of body weight, not exceeding a total of 90 mg. The initial 10% of the dose was administered as a rapid bolus over 1 min, followed by infusion of the remaining 90% over 60 min. The standard treatment group was managed with antiplatelet therapy—such as aspirin, clopidogrel, or their combination—and other supportive measures, in line with the Chinese Guidelines for Diagnosis and Treatment of Acute Ischemic Stroke 2018 (Neurology and Society, 2018).

We developed a hybrid model combining a decision tree and a Markov model to evaluate the cost-effectiveness of alteplase plus standard medical therapy versus standard therapy alone in patients with posterior circulation ischemic stroke (PCIS) at 4.5–24 h after onset in China. The target population excluded patients with extensive early hypodensity on computed tomography and those scheduled for thrombectomy. The model simulated outcomes over a lifetime horizon of 30 years. This corresponds to a mean starting age of 64 years and an ending age of 94 years, which exceeds China’s average life expectancy of 79 years in 2024. Both the decision tree and the Markov cycles were set at 3-month intervals, resulting in a total of 119 Markov cycles. To minimize potential overestimation of costs and effects, a half-cycle correction was applied. From the perspective of the Chinese healthcare system, only direct medical costs were included; direct non-medical and indirect costs were excluded from the analysis. All future costs and health outcomes were discounted at an annual rate of 5%, in accordance with the China Guidelines for Pharmacoeconomic Evaluations (Guoen et al., 2020). For sensitivity analyses, a range of 0%–8% was applied to reflect plausible variations in long-term economic conditions.

In the decision tree model (Figure 1), patients with posterior circulation stroke were randomly assigned to two groups: one receiving alteplase plus standard therapy, and the other receiving standard therapy alone. At 3 months, clinical outcomes, costs, and adverse events were assessed. The outcome measure of interest was the mRS, which reflects the level of disability, where mRS 0 indicates no disability, mRS 5 indicates the most severe disability, and mRS 6 represents death. Costs collected at 3 months included stroke hospitalization costs, post-discharge care costs, alteplase costs, and costs related to adverse events. The major adverse event considered was symptomatic intracranial hemorrhage (sICH), with its incidence derived directly from the EXPECTS trial (Yan et al., 2025).

After the initial 3-month decision tree phase, patients entered the Markov model, which was designed to capture long-term costs and outcomes after stroke. Since patients in the alteplase plus standard therapy group achieved better mRS scores at 3 months, their initial state distribution in the Markov model was more favorable than that of the standard therapy group. Each Markov cycle represented a 3-month period, with a total of 119 cycles over the lifetime horizon. In each cycle, patients could experience one of three events: no event, recurrent stroke, or death. Deaths were classified as either occurring subsequent to a recurrent stroke or resulting from background mortality. Patients who experienced no event remained in the same mRS state in the next cycle. Those with a recurrent stroke could remain in the same or move to a worse mRS state, while patients who died entered the absorbing state “Dead” and exited the simulation. The overall model structure is shown in Figure 1.

In the decision tree model, the 3-month distribution of mRS outcomes was directly derived from the EXPECTS trial (Yan et al., 2025). In that trial, the proportions of patients in the alteplase group with mRS scores of 0–6 at 3 months were 0.3913, 0.3478, 0.1565, 0.0261, 0.0174, 0.0087, and 0.0522, respectively. The corresponding proportions in the standard therapy group were 0.2564, 0.3504, 0.1197, 0.1197, 0.0342, 0.0342, and 0.0855. The incidence of sICH was 0.0172 in the alteplase group and 0.0087 in the control group.

The initial mRS distribution in the Markov model was derived from the observed outcomes reported above. For subsequent cycle transitions, the EXPECTS trial did not provide annual recurrent stroke probabilities applicable to the Chinese population; therefore, recurrence rates were sourced from the IRIS trial, which demonstrated that higher mRS scores are associated with increased recurrence risk (de Havenon et al., 2024). The case-fatality rate after recurrent stroke was set at 0.21, derived from a separate study of Chinese stroke patients (Pan et al., 2018). Background mortality rates in the Markov model were sourced from the China Health Statistical Yearbook 2023 (Shili et al., 2024). Importantly, these rates were age-specific, ensuring that mortality risk increased appropriately with age throughout the model’s 30-year horizon. Because post-stroke patients experience higher mortality than the general population, the age-specific background mortality rates were further adjusted using a hazard ratio derived from prior studies to reflect excess post-stroke mortality (Chen et al., 2025; Chen et al., 2023).

Only direct medical costs were included in this study. In this study, indirect costs were not included. Indirect costs typically refer to productivity losses resulting from morbidity or premature mortality, as well as the economic value of informal caregiving provided by family members. These costs reflect the broader societal burden of stroke but were excluded because the analysis was conducted from the perspective of the Chinese healthcare system, which considers only direct medical expenditures. All costs were converted to 2024 Chinese Yuan (CNY) values. Costs prior to 2024 were inflated to 2024 levels using the healthcare component of the Chinese CPI, while future costs were discounted at an annual rate of 5%.

For both the alteplase group and the standard therapy group, cost components included stroke hospitalization costs, post-stroke care costs, recurrent stroke costs, and adverse event costs. In addition, the alteplase group incurred the drug acquisition cost of alteplase. The hospitalization and post-stroke care costs varied according to mRS levels, with higher disability corresponding to greater resource utilization. These data were obtained from published studies on Chinese stroke patients. The cost of recurrent stroke was sourced from institutional data in China (Chen et al., 2023). The cost of sICH was derived from the Thrombolysis Implementation and Monitor of Acute Ischaemic Stroke in China (TIMS-CHINA) study, a national prospective registry including 1,440 AIS patients treated with intravenous tPA across 67 centers in China (Pan et al., 2018). The cost of alteplase was based on the national procurement price under China’s volume-based purchasing program, which is uniform nationwide: 1,700 CNY for 20 mg and 3,400 CNY for 50 mg. In the base-case analysis, the mean value of these two unit prices were applied, while in the one-way sensitivity analysis, the lower and upper bounds were defined by the respective prices of 50 mg and 20 mg formulations.

Utility reflects the quality of life of patients, and health outcomes were expressed as quality-adjusted life years (QALYs), calculated as utility multiplied by life years. In our cost-effectiveness analysis, the utility values for each mRS state were obtained from a randomized controlled trial conducted in China (Wang et al., 2014). The utility scores for mRS levels 0 to 6 were 0.95, 0.89, 0.67, 0.44, 0.16, 0.10, and 0, respectively. For recurrent stroke, a utility value of 0.42 was applied, while for sICH, a disutility of 0.38 was incorporated into the model (Wang et al., 2024).

The primary outcome of this study was the incremental cost-effectiveness ratio (ICER), defined as the additional cost (in CNY) required to gain one additional quality-adjusted life year (QALY). According to the China Guidelines for Pharmacoeconomic Evaluations (Guoen et al., 2020), alteplase was considered not cost-effective if the ICER exceeded three times the per capita gross domestic product (GDP). When the ICER was between one and three times the per capita GDP, alteplase was deemed cost-effective, and when the ICER was below one time the per capita GDP, alteplase was regarded as highly cost-effective. Secondary outcomes included incremental cost, incremental effectiveness, incremental life years, total cost, total effectiveness, and total life years.

Sensitivity analyses included both one-way sensitivity analysis and probabilistic sensitivity analysis (PSA). In the one-way sensitivity analysis, individual input parameters were varied within their respective confidence intervals to examine the impact of each variable on the ICER. The resulting changes in ICER values were illustrated using a tornado diagram, which highlights the relative influence of each parameter on model outcomes. The PSA accounted for the joint uncertainty across multiple parameters simultaneously. In the PSA, input variables were assigned appropriate probability distributions based on their characteristics: transition probabilities and utility values followed beta distributions, while cost parameters followed gamma distributions. The PSA was conducted to evaluate how combined parameter uncertainty influenced the ICER. Results were presented using scatter plots on the cost-effectiveness plane and cost-effectiveness acceptability curve.

Over a lifetime horizon, patients receiving standard treatment achieved 5.55 QALYs and 8.39 life years, with a lifetime cost of 127,227 CNY. In comparison, those treated with alteplase plus standard treatment achieved 6.49 QALYs and 8.90 life years, incurring a lifetime cost of 144,804 CNY. Thus, alteplase plus standard treatment provided an additional 0.95 QALYs (0.51 life years) at an incremental cost of 17,580 CNY, resulting in an ICER of 18,567 CNY per QALY gained, which is below one times China’s per capita GDP in 2024 (Table 1).

As shown in Figure 2, the input parameters that had the greatest impact on the ICER were the RRs for mRS 4, mRS 5, mRS 2, and mRS 3. Other parameters—such as the proportion of patients with mRS 3 in the control group at 3 months post-stroke, the annual cost for mRS 0–1, the proportion of patients with mRS 5 in the control group at 3 months post-stroke, and the annual cost for mRS 2–5—also influenced the ICER, but to a lesser extent. Notably, variations in all parameters kept the ICER within the range of 11,000–38,000 CNY per QALY.

As shown in Figure 3, the incremental cost-effectiveness scatterplot indicated that all simulated points were located in the first quadrant and lie below the willingness-to-pay threshold line of 95,749 CNY per QALY. Figure 4 showed that when the willingness-to-pay was 24,000 CNY per QALY, alteplase and standard treatment had equal probabilities of being cost-effective. However, when the willingness-to-pay exceeded 24,000 CNY per QALY, the probability that alteplase was cost-effective increased, reaching nearly 100% when the willingness-to-pay reached 35,000 CNY per QALY.