2.1 Analytical overview
A partitioned survival model was constructed to simulate the disease process of advanced HCC and to estimate the cost-effectiveness of sorafenib, lenvatinib, atezolizumab plus bevacizumab, and sintilimab plus bevacizumab for patients in China.
Patients with advanced HCC or unresectable tumors without prior systemic therapy were selected for inclusion in the model. Based on disease progression, three transition states of disease were defined: progression-free survival (PFS); progressive disease (PD); and death. The cohort flow was determined by survival curves over time. Each survival curve described the movement of patients out of the health state associated with that curve and into the next state of progression. At a certain time (t), the survival ratio of PD was equal to the OS value minus the PFS value. The proportions of patients with PFS and OS were calculated based on the results of the NCT01761266, NCT03434379, and NCT03794440 trials. The evidence was validated by comparing the PFS and OS results of the model with the observed and extrapolated data [6–8, 10, 11]. The length of each partitioned survival model cycle was 1 month, and survival was adjusted for quality of life based on specific utilities. Direct fees paid by patients at the hospital included medicine drugs, bed fees, testing fees, and adverse reaction processing fees. According to the 2020 edition of the “Guidelines for the Evaluation of Chinese Pharmacoeconomics”, all fees and utilities are discounted by 5% [12]. Models were constructed using the TreeAge Pro 2019 software (2019 TreeAge Software, Inc.).
2.2 Efficacy and safety input
NCT01761266 was a multicenter, randomized, open-label trial. It compared the efficacy and safety of lenvatinib versus sorafenib as first-line systemic therapy for unresectable HCC. A total of 954 patients were randomized in a 1:1 ratio to receive lenvatinib or sorafenib. The study included patients aged ≥ 18 years, diagnosed with advanced HCC, and categorized as Barcelona Clinic Liver Cancer stage B or C and Child–Pugh class A. More information on the inclusion and exclusion criteria is available elsewhere[7].
NCT03434379 was a global, randomized, open-label study. It compared the efficacy and safety of atezolizumab-bevacizumab versus sorafenib as first-line systemic therapy for unresectable HCC. Patients aged ≥ 18 years with locally advanced metastatic or unresectable HCC (or both) and those with advanced HCC who had not received prior systemic therapy were enrolled. More information on the inclusion and exclusion criteria is available elsewhere[8]. The patients were randomized in a 2:1 ratio to receive atezolizumab plus bevacizumab or sorafenib until the occurrence of intolerable toxicity or disease progression. The intent-to-treat population included 336 and 165 patients in the atezolizumab-bevacizumab and sorafenib groups, respectively.
NCT03794440 was randomized, open-label, multicenter study conducted in China. It compared the efficacy and safety of sintilimab-bevacizumab versus sorafenib as first-line systemic therapy for unresectable HCC. Patients aged ≥ 18 years with histologically or cytologically diagnosed or clinically confirmed unresectable or metastatic HCC, no prior systemic therapy, and baseline Eastern Cooperative Oncology Group performance status of 0 or 1 were eligible for selection. A total of 595 patients were randomly assigned in a 2:1 ratio to receive sintilimab plus bevacizumab or sorafenib until disease progression or the occurrence of unacceptable toxicity. More information on the inclusion and exclusion criteria is available elsewhere[9].
In these studies, lenvatinib, atezolizumab plus bevacizumab, and sintilimab plus bevacizumab shared a common comparator (i.e., sorafenib); hence, indirect comparisons were possible. Clinical efficacy inputs for the models were derived from the respective, randomized, controlled trials. Three clinicians or oncologists conducted a blinded review of the three aforementioned studies; based on this review, the studies were comparable (Table 1). The specific review content is shown in the appendix Table S1.
PFS and OS data for lenvatinib and sorafenib were derived from NCT01761266; PFS and OS data for atezolizumab plus bevacizumab were derived from NCT03434379; and PFS and OS data for sintilimab plus bevacizumab were derived from NCT03794440. Models were extrapolated using the method established by Guyot et al. [13]. The GetData Graph Digitizer, version 2.26,9 software (getdata-graph-digitizer, Inc.) was used to obtain data points from the PFS and OS curves. These data points were subsequently used to fit the following parametric survival functions: Weibull; log-normal; log-logistic; exponential; generalized gamma; and Gompertz. The Bayesian Information Criterion is useful for statistical testing methods [14]. The final survival model selection for each study is shown in Table 2. The validation plots of the subgroups are shown in Figures S1–S4.
2.3 Cost and utility inputs
The direct costs that patients need to pay include drugs, testing costs, bed costs, management of adverse reactions, and active treatment after disease progression (Tables 3, 4).The costs are reported in August 2022 US dollars (1 dollar is equal to 6.74 yuan).
The four treatment regimens in the model followed the dosing intervals and doses provided in their respective labels. In the NCT01761266 trial, the recommended dosage for patients with a body weight < 60 kg and ≥ 60 kg was 8 mg/day and 12 mg/day lenvatinib, respectively; the recommendation for patients who received sorafenib was 400 mg twice daily [7]. According to the data, there are more male than female patients with HCC in China. This model used the average Chinese male weight of 69.6 kg estimated in 2021 to calculate the drug dose[15]. Therefore, in the model, the dosage in the lenvatinib group was 12 mg/day. In the NCT03434379 trial, the recommended dosages were 1,200 mg intravenous atezolizumab and 15 mg/kg bevacizumab once every 3 weeks [8]. In the NCT03794440 trial, patients received 200 mg intravenous sintilimab plus 15 mg/kg bevacizumab, once every 3 weeks [9]. Hence, one treatment cycle required 1,044 mg bevacizumab. The prices of the drugs were obtained from YaoZhi Internet (https://www.yaozh.com/), in which the latest negotiated prices for medicines are occasionally reported.
Except for the drug costs of the four treatment regimens, other costs were calculated based on consultations with physicians from three hospitals; the data provided by the physicians were averaged. Patients on all treatment regimens were followed up once every 2 months. The examinations included abdominal ultrasound, magnetic resonance imaging, blood testing, and liver function testing. Patients receiving intravenous atezolizumab plus bevacizumab or sintilimab plus bevacizumab were required to pay the inpatient bed cost. For all four treatment regimens, costs were calculated for grade 3 adverse reactions with an incidence > 10%, including hand-foot skin reaction and hypertension. Since adverse reactions occurred throughout the course of treatment, the incidence of adverse reactions was allocated evenly to each month.
To ensure model consistency, it was assumed that the four treatment regimens were discontinued at the time of disease progression. Active treatment after progression followed, assuming sequential treatment with regorafenib. For patients experiencing treatment intolerance and progression, the cost of next treatment was US$1,774/month.
The measure utility-specific quality-adjusted life year (QALY) based on the health status was used to determine treatment outcomes. The utility values for PFS, PD, and death were 0.76, 0.68, and 0, respectively (values were derived from Thompson et al.) [16].
2.4 Comparative cost-effectiveness
The cost and utility of the four treatment options were compared using the incremental cost-effectiveness ratio (ICER). According to the recommendations of the World Health Organization and the Guidelines for the Evaluation of Chinese Pharmacoeconomics (2020), this study used three times the per capita GDP of China (US$36,600) reported in 2021 as the willingness-to-pay (WTP) threshold [17–19]. In the comparison with sorafenib, a scheme with an ICER less than WTP is considered to have economic value.
2.5 Perspectives
In China, there is a social insurance and house fund system, which includes medical insurance. Individuals need to pay their contributions on a monthly basis[20]. Through the medical insurance fund, costs related to outpatient visits and hospitalizations can be reimbursed. However, the reimbursement rates differ depending on the diseases and drugs. This complicated process involves medical care diagnosis-related groups/diagnosis-intervention packet payments [21, 22]. In the present study, we analyzed medical care costs from the patient perspective. We considered only the proportion of the patient-copayment for the cost of examination fees, bed fees, adverse reaction processing fees, first-line drug fees, and second-line treatment fees. According to the medical insurance reimbursement policy of Yunnan Province, for ordinary urban residents, the outpatient self-pay rate for sorafenib and lenvatinib was 35% and 40%, respectively. The inpatient self-pay rate for sintilimab plus bevacizumab was approximately 45%. Atezolizumab plus bevacizumab was not covered by the medical insurance reimbursement scheme; hence, the self-pay rate for this regimen was 100% [23].
2.6 Sensitivity analyses
In this study, one-way and probability sensitivity analyses were used to explore the influence of different factors on the results of the model. Upper and lower inputs for one-way sensitivity analyses are shown in Table 5. In one-way sensitivity analyses, the incremental net monetary benefit (INMB) was calculated based on the following formula:
INMB(λ) = (µE1 − µE0)*λ − (µC1 − µC0) =ΔE*λ − ΔC,
where µEi and µCi are the effectiveness and cost of alternative therapy (i = 1) or sorafenib (i = 0), respectively [24], and λ is three times the GDP per capita of China reported in 2021 (WTP).
The one-way sensitivity analysis was represented using a tornado diagram. The probabilistic sensitivity analysis was performed using Monte Carlo simulation sampling, and the final results were presented using a cost-effectiveness acceptability curve.