Cost–effectiveness of standard utilization of zoledronic acid for bone metastases from advanced lung cancer in China

The constructed decision analytic model adapted a Markov model approach to simulate the risk of SREs and mortality associated with the two scenarios for the utilization of ZA (i.e., standard utilization and real-world utilization) in Chinese patients with BM developed from advanced lung cancer. According to the clinical practice guidelines, the standard utilization of ZA was defined as monthly intravenous administration of 4 mg of ZA during the model time horizon, which was set to lifetime. The real-world utilization of ZA was based on the observed treatment rate (44.6%) and treatment duration of bisphosphonates (3.0 months) in Chinese patients with BM and advanced lung cancer [8]. The constructed Markov model included three health states – BM only (without SRE), SRE and death – to simulate the risk of developing SRE from BM under the ZA treatment and the risk of mortality associated with BM only and SRE, respectively. The real-world utilization scenario stratified patients by ZA treatment status; then, the risks of having an SRE occur under ZA treatment and no ZA treatment were simulated, respectively. All Markov models were constructed with a 1-month cycle length to account for the rapid disease progression associated with advanced lung cancer. To follow the Chinese health economic evaluation guideline [9], the constructed decision analytic model discounted survival time, quality-adjusted life years (QALY) and medical costs at 3% per annum. The structure of the decision analytic model is illustrated in Figure 1.

The literature search was conducted in both Chinese and English bibliographic databases to identify appropriate evidence to estimate monthly risk of SRE from BM under ZA (or no) treatment in patients with advanced lung cancer and the monthly risk of mortality associated with BM only as well as SRE in Chinese patients with advanced lung cancer. The monthly risks of SRE occurrence and mortality were estimated using a survival function formula with an exponential distribution (survival proportion = e^-λt, where t is the follow-up time and λ is the hazard rate for the risk of SRE occurrence or mortality). Due to the lack of Chinese utility values associated with BM developed from advanced lung cancer, an empirical estimation of the utility was used [10]. The reported reduction of utility associated with the type of SRE relative to BM only and the reported distribution of SRE types in Chinese patients with developed BM from advanced lung cancer were used to estimate the model variables for utility [11,12]. This study identified a clinical cohort of advanced lung cancer patients with BM in a Chinese tertiary care cancer hospital from 1 January 2012 to 31 December 2015 to track their insurance claims records and estimate monthly medical costs associated with BM only and each SRE type. The estimated model variables are summarized in Table 1.

The constructed decision analytic model was used to perform the base-case analysis, one-way sensitivity analyses and a probabilistic sensitivity analysis (PSA) for the comparisons between the two ZA utilization scenarios – standard utilization versus real-world utilization – for the cumulated health benefits and medical costs over a lifetime horizon. The base-case analysis was conducted to estimate the point estimations of cumulated risk of SRE, survival time, QALY and lifetime medical costs associated with the two ZA utilization scenarios. The incremental cost–effectiveness ratio (ICER) per gained QALY was calculated using the ratio between the differences in cumulative medical costs and QALY associated with the two model scenarios. One-way sensitivity analyses were performed to assess the uncertainty associated with each model variable on the cost–effectiveness (indicated by ICER per gained QALY) by varying the model variable from the lower limit to the upper limit of its 95% CI or +/- 25% of its baseline value if their 95% CIs were lacking. A Monte Carlo simulation approach was used to perform the PSA by conducting 10,000 iterations under the gamma or beta distributions of model variables. The distribution of the 10,000 simulations of cost–effectiveness was used to identify the median and 95% credible interval of ICER per gained QALY.

This study used R statistical software to estimate the model variables from the identified literature evidence and estimate medical costs associated with BM only or SRE from the claims data. Microsoft Excel 2013 was used to construct the decision analytic model and conduct the cost–effectiveness analysis.

Based on the simulation of the newly developed SRE in each model cycle, the standard utilization of ZA scenario could reduce the lifetime cumulated risk of SRE by 17.9% (45.7 vs 63.6%) when compared with the real-world utilization scenario. The base-case analysis projected more gained QALY (0.673 vs 0.626 QALY) associated with the standard utilization scenario. Based on the illustrated curves for cumulated medical costs associated with health states over time in the model (Figure 2), the standard utilization scenario began costing less than real-world utilization scenario after 6 months simulation. The simulations of medical costs over lifetime model time horizon projected that the standard utilization scenario increased the cumulative medical costs for BM only (¥112,971 vs ¥79,354; difference ¥33,618, ¥1 = US$0.15 as of 14 November 2018) and drug acquisition costs of ZA (¥19,678 vs ¥4,701, difference ¥14,977) but saved more medical costs associated with SRE (¥210,514 vs ¥292,880, difference -¥82,366) over a lifetime horizon. Thus, the standard utilization scenario saved ¥33,771 in cumulated medical costs when compared with the real-world utilization scenario (¥ 343,163 vs ¥ 376,934). The standard utilization scenario dominated the real-world utilization scenario by gaining more health benefits and saving more cumulated medical costs in the base-case analysis. The point estimations of the cumulated health benefits and medical costs associated with the two model scenarios regarding the utilization of ZA for BM developed from advanced lung cancer in Chinese patients are summarized in Table 2.

One-way sensitivity analyses identified that the utility associated with BM only and the utility of SRE substantially drove up the ICER per gained QALY associated with the standard utilization scenario by ¥958,755 and ¥50,825,320, respectively, when these model values varied from the lower limit to the upper limit of their 95% CIs. Instead, the monthly risk of mortality associated with SRE, the monthly medical costs associated with SRE and the monthly medical costs associated with BM only reduced the ICER per gained QALY by  ¥3,385,073, ¥2,603,536 and ¥1,321,895, respectively. The results of the one-way sensitivity analyses are illustrated in Figure 3. Additionally, increasing the ZA treatment rate from 25% and the ZA treatment duration from 6 months in the standard utilization scenario could allow the standard utilization scenario to cost less than the real-world utilization scenario.

To further demonstrate the potential health and economic benefits associated with improving the utilization of ZA for developed BM from advanced lung cancer in Chinese patients, the one-way sensitivity analysis assessed the impact of increasing the treatment rate and treatment duration of ZA on the cumulated medical costs in the real-world utilization scenario. Increasing the treatment rate by 10% and the treatment duration by 1 month in the real-world utilization scenario could save ¥2219 and ¥828, respectively, in cumulated medical costs on average during the lifetime model simulation.

The generated ICER per gained QALY associated with the standard utilization scenario from the 10,000 simulations in the PSA were plotted to demonstrate the cost–effectiveness distribution (Figure 4). Based on the plotted cost–effectiveness distribution, the cost–effectiveness dominance of the standard utilization scenario was found in 92.8% of the 10,000 simulations. The identified median of the ICER per gained QALY associated with the standard utilization scenario relative to the real-world utilization scenario was -¥598,843 (95% credible interval: -¥2,918,636 to ¥102,492).