Cost-Effectiveness Analysis of Trastuzumab Emtansine as Second-Line Therapy for HER2 Positive Breast Cancer in China

Objective To evaluate the cost-effectiveness of trastuzumab emtansine (T-DM1) as the second-line treatment for patients with human epidermal growth factor receptor-2 (HER2) positive breast cancer from the Chinese healthcare perspective. Capecitabine (Cap), capecitabine + lapatinib (Cap+Lap), capecitabine + trastuzumab (Cap+Tra), capecitabine + trastuzumab + pertuzumab (Cap+Tra+Pre) were selected as comparators. Methods A three-state Markov simulation model was performed. The state transition probabilities were estimated based on the results of a published network meta-analysis, and utilities were derived from the published literature. The costs populated in the model were acquired from the local charge or previously published studies. Univariate sensitive analysis and probabilistic sensitivity analyses were performed to test the robustness of the results.


Introduction
Breast cancer is the most common malignant tumor and the 5th most common cause of cancer-related death in Chinese women [1]. In 2015, it was estimated 304,000 new cases of breast cancer were diagnosed and nearly 70,000 deaths were due to breast cancer in China [1]. Moreover, the onset age of breast cancer among Chinese women is nearly 10-15 years younger than in western patients [2]. Human epidermal growth factor receptor-2 overexpression has been reported in approximately 20-25% of breast cancers [3,4]. Breast cancers with HER2 overexpression are associated with poor prognosis and shorter patient survival. Nevertheless, with the development of anti-HER2 targeted therapies, there has been a signi cant improvement in the survival of patients with HER2 positive breast cancer [5]. As far as we know, there are ve kinds of anti-HER2 targeted agents available for breast cancer treatment in China,including pyrotinib, lapatinib, pertuzumab, trastuzumab, and Trastuzumab emtansine.
Although therapy with T-DM1 showed certain clinical bene ts, the high cost of T-DM1 is also an important factor affecting treatment decisions. The expenditure for breast cancer treatment in China has been constantly increased over the past few years, which brought a huge economic burden on individuals and families [11]. Therefore, we developed an economic model based on the network meta-analysis to evaluate the cost-effectiveness of T-DM1 compared with Cap, Cap + Lap, Cap + Tra, and Cap + Tra + Pre as the second-line treatment for patients with HER2 positive breast cancer from the Chinese healthcare perspective.

Patient and treatment
The hypothetical cohort matched the inclusion criteria of the network meta-analysis was incorporated into the model [10]. These patients with HER2-positive, unresectable, locally advanced, or metastatic breast cancer had progressed after prior treatment with adjuvant therapy or trastuzumab plus taxane.
The treatment strategies evaluated in our study included: 1)T-DM1 group, a dose of 3.6 mg/kg T-DM1was intravenously infused every 21 days; 2) Cap group, 1,250 mg/m 2 orally twice daily on days 1-14 of each 21-day treatment cycle; 3) Cap + Lap group, lapatinib 1,250 mg orally once daily plus capecitabine 1,000 mg/m2 orally twice daily on days 1-14 of each 21-day treatment cycle; 4) Cap + Tra group, trastuzumab 8 mg/kg loading dose in the rst cycle followed by 6 mg/kg maintenance doses every 3 weeks, plus capecitabine 1,250 mg/m 2 orally twice daily on days 1-14 of each 21-day cycle; 5) Cap + Tra + Pre group, pertuzumab 840 mg initial dose in cycle 1 followed by 420 mg maintenance doses every 3 weeks, trastuzumab 8 mg/kg loading dose in the rst cycle 1 followed by 6 mg/kg maintenance doses every 3 weeks, capecitabine 1,000 mg/m 2 orally twice daily on days 1-14 of each 21-day cycle. Treatment was discontinued when the disease progressed or an intolerable level of toxicity was reached.

Model Structure
A Markov model was constructed by Treeage Pro Suit 2011 (Treeage Software, Inc., MA, USA) to compare the cost-effectiveness of T-DM1 with other 4 treatments over a 10-year time horizon. The model included three mutually exclusive health states: progression-free survival, progressive disease (PD) and death ( Fig. 1). The cycle duration was 3 weeks, and the initial health state for all patients was PFS. At the end of each cycle, the patients remained in a PFS state or progressed to the PD state. From the PD state, the patients could either remain at PD or move to death.
Health outcomes were expressed as quality-adjusted life years. The primary result was presented as incremental cost-effectiveness ratio, which was the costs spent to gain one QALY. The willingness-to-pay (WTP) threshold in the analysis was considered as three-times the per capita gross domestic product ($30829.3) of China in 2019, which was suggested by the World Health Organization (WHO). Both cost and health outcomes were discounted at 3% annually after the rst year to allow for current values.

Transition probabilities
The transition probabilities into each Markov state were measured with the clinical e cacy data derived from a published network meta-analysis and the MILIA trial [6,9,10]. OS and PFS values of T-DM1 at multiple time points were read using Engauge Digitizer software version 12.1 (http://digitizer.sourceforge.net) from the Kaplan-Meier (KM) curves of MILIA trial. The parametric model of Weibull was tted to the data extracted from the KM curves by the R statistical software (http://www.rproject.org). The Weibull survival models of other four comparators were derived by applying the HRs for each comparator versus T-DM1, with the following formulas: λ comparator = λ T−DM1 × HR and γ comparator = γ T−DM1 , where λ was the scale parameter of Weibull distribution, γ was the shape parameter of Weibull distribution [12]. The estimated scale (λ) and shape (γ) parameters of T-DM1, and HRs for comparators versus T-DM1 were described in Table 1.
The time-dependency transition probabilities from PFS to PFS, de ned as the ratio of the number of patients at the end of the cycle to the number of patients at the beginning of the cycle, were calculated according to the formulation: tp(t) = s(t)/s(t-1), where t presents the current stage of the Markov model [13,14]. The transition probabilities from PD to death were calculated based on the difference between the estimated OS and PFS Weibull models.

Cost and Utility
From the perspective of Chinese healthcare, only direct medical costs were considered in the model. The direct medical costs consisted of anti-cancer agents, management of adverse events (AEs), hospitalization, and follow-up (Table 1). Average height (155.8 cm) and weight (57.3 kg) of Chinese women were used to calculate the drug doses [15]. There are two package sizes of T-DM1 available in China, 100 mg/vial and 160 mg/vial. According to the dose calculation, each dose required a big vial and a small one. Besides, there is a charity program for T-DM1, that patients would receive T-DM1 for free from cycle 8 to cycle14. Therefore, the costs of T-DM1 were excluded for cycle 8 to cycle14 in the T-DM1 group in the model. The costs of hospitalization included drug administration cost,drug delivery cost, bed fee and nursing fee. Considering that both lapatinib and capecitabine were taken orally, the hospitalization costs were not calculated in the Cap group and Cap + Lap group. Follow-up was associated with radiological examination, laboratory test and echocardiography examination. The radiological examination was performed every 2 cycles, and echocardiography examination was performed at cycle 2, cycle 4, and every 3 cycles thereafter as reported in the EMILIA trial [6].
Grade 3-4 AEs with incidence rates greater than 5% were considered, including diarrhea, Palmar-plantar erythrodysesthesia syndrome (PPES), elevated liver enzymes, neutropenia and thrombocytopenia. The incidence rates of AEs were derived from the relevant RCTs [9,16,17]. AEs management costs were obtained from the published literature and were applied once in the rst cycle after treatment initiation [18][19][20]. It was assumed in the model that the cost of post-progression treatment was the same as the average hospitalization expenditure for breast cancer treatment in China [11]. All the costs were converted into 2019 US dollars (CYN 6.8985 = US $1.00), and adjusted based on the medical care consumer price index (CPI), if necessary.
The health utility values for different health states and disutility values for AEs were obtained from the published literature (Table 1). 0.715 was assigned to the PFS state, and 0.452 to the PD state [21]. Since the disutility of elevated liver enzymes was unavailable, we assumed it was equivalent to the disutility of thrombocytopenia. Like the costs of AEs management, the disutility associated with AEs was only applied to the rst cycle.

Sensitivity Analysis
A series of one-way sensitivity analyses were performed to identify the in uence of key model parameters on the model. As shown in Table 1, the parameters of costs and utilities were varied at a range of ± 20% of their baseline values, HRs were varied over their 95% CIs, and the range of discount rate was from 0-8%. The results of the one-way sensitivity analyses were presented as the tornado diagrams. Probabilistic sensitivity analysis (PSA) was further performed to assess the robustness of the estimated costeffectiveness ratio using Monte Carlo simulations of 1,000 patients, where samples were taken randomly from the distribution of the included parameters. Gamma, normal, beta distributions were adopted for costs, HRs, and utilities respectively. WTP acceptability curves were generated to illustrate the result of PSA.

Sensitivity analysis
The results of univariable sensitivity analyses were shown in Fig. 2. The parameters with the greatest in uence on the ICERs were HRs of OS, HRs of PFS, cost of T-DM1, and the utility of PFS state in all comparisons. The impacts of AEs managing costs, hospitalization costs, follow-up costs, and capecitabine costs were almost negligible. Other parameters had a minor in uence on the robustness of the model. The ICERs were lower than the WTP threshold of $30,829.3/QALY in three scenarios, one of which was Cap + Tra dominated, and two of which were Cap + Tra + Pre dominated. Other scenarios did not alter the cost-effectiveness conclusion that ICERs exceeded the WTP threshold.

Discussion
HER2-positive breast cancers tend to develop more rapidly and spread more aggressively than HER2negative cancers. As the rst HER2-targeted antibody-drug conjugate, trastuzumab Emtansine has been proven to be very effective and well tolerated in the second-line or later treatment for HER2-positive breast cancer [23].  [24]. However, the guideline provided by the National Institute of Health and Healthcare (NICE) did not recommend trastuzumab emtansine, because the probability of trastuzumab emtansine being the most cost-effective was 0% [25]. With the current increase in healthcare expenditure in China, the medical costs of breast cancer remain a signi cant economic burden particularly on advanced patients [26]. The cost-effectiveness factor needs to be considered before T-DM1 is widely used in clinical treatment.
We conducted an economic analysis to compare the T-DM1 with four alternative regimens as second-line The present study has some limitations that deserve to be mentioned. First, due to the lack of randomized controlled trials, HR from indirect comparison was used to calculate the transition probabilities. The univariate sensitive analysis indicated that HR was the most important in uential factor on ICER, especially for T-DM1 versus Cap + Tra, and T-DM1 versus Cap + Tra + Pre. Second, grade 1/2 AEs and dose reduction due to toxic reactions were not considered in the model. We assumed that the majority of mild AEs were self-limiting, and the costs of AEs management had minimal impact on the ICER base on the univariate sensitive analysis. Dose adjustment (for example, the rst dose reduction for T-DM1was to 3.0 mg/kg, and the second to 2.4 mg/kg) was not applied, which may result in a higher total cost than clinical practice. Third, the utilities of PFS and PD were obtained from published literature based on United Kingdom populations, which may differ in Chinese. Besides, the utilities were assumed to be the same in each therapy, which may be different from the real-world data.

Conclusion
Although T-DM1 as second-line therapy in the treatment of HER2-positive advanced breast cancer showed excellent clinical e cacy, the results of our study suggested that T-DM1,compared with capecitabine monotherapy, capecitabine plus lapatinib, capecitabine plus trastuzumab, capecitabine plus trastuzumab and pertuzumab, was unlikely to be cost-effective from the perspective of the Chinese healthcare system. A signi cant reduction in the price of T-DM1 may be a potential measure to improve its cost-effectiveness. We expect that the results of this study will be useful for decision-making by patients, doctors and governments. When high-quality head-to-head clinical trials are available, our research will be updated. Declarations