Introduction

Small cell lung cancer (SCLC) is a poorly differentiated malignant epithelial tumor and accounts for 13%–15% of lung cancers [1]. It is characterized by rapid multiplication, early widespread metastasis, and abnormal endocrine syndrome. Approximately 30% of patients were diagnosed with limited disease confined to the chest, whereas the rest presented with extensive-stage cancer with metastasis to extrathoracic locations, such as bone, brain, liver, and adrenals [2]. Although SCLC is highly sensitive to chemotherapy and radiotherapy, most patients eventually die of recurrent disease, and long-term survival is rare in extensive-stage SCLC (ES-SCLC) [3, 4]. Etoposide plus platinum (EP) has been the standard treatment for a long time. Despite evaluation of a number of novel drugs (vandetanib, temsirolimus, and bevacizumab) in the treatment of SCLC, little advance has been made because of their failure to show a significant survival benefit [5-7]. The approval of nivolumab as a second-line regimen for SCLC in 2018 marked a quantum leap forward in SCLC, indicating that immunotherapy is changing the overall treatment layout of SCLC [8].

Programmed cell death-ligand 1 (PD-L1) is a protein expressed on the surface of tumor cells that attenuates the activation of T cells and suppresses the antitumor immune response. Durvalumab is a PD-L1 inhibitor and exerts its antitumor efficiency by preventing immune escape mediated by the PD-L1 pathway. Recently, the National Comprehensive Cancer Network (NCCN) Clinical Practice Guidelines has recommended durvalumab in combination with EP as a preferred first-line treatment option for patients with ES-SCLC based on evidence from the CASPIAN phase III randomized clinical trial (ClinicalTrials.gov, NCT03043872), which demonstrated that adding durvalumab to EP significantly improved the median overall survival (mOS; hazard ratio: 0.73, 95% confidence interval: 0.59–0.91) over EP alone [9, 10].

It is reported that the prices of anticancer drugs continue to increase at 10% annually and that the proportion of immunotherapy use is growing [11, 12]. Since the explosion in anticancer medical costs has severely troubled patients, society, and country, more concern should be focused on economic concerns surrounding immunotherapy, including carrying out a comprehensive assessment of its cost-effectiveness. As a newly released immunotherapy drug, durvalumab has been proved to work on urothelial carcinoma and stage III non-small cell lung cancer (NSCLC), but the high cost prohibits its wide usage [13, 14]. Although a pharmacoeconomic study has shown that durvalumab was cost-effective as consolidation therapy for stage III NSCLC from the U.S. health care system perspective, the same conclusion could not be drawn for SCLC because of the difference among stages, process of disease, treatment regimens, etc. [12]. So far, only two immune checkpoint inhibitors (ICIs), durvalumab and atezolizumab, have been approved for use in first-line treatment of SCLC, and durvalumab has showed its superior survival benefit and less incidence of some common severely adverse events (AEs) than atezolizumab [10, 15]. An economic evaluation is, thus, extremely urgent. This study aims to evaluate the cost-effectiveness of durvalumab plus EP (DEP) compared with EP for untreated ES-SCLC from the U.S. payer perspective, expecting to provide information and reference for decision makers.

Materials and Methods Patients and Treatment Basic population information was derived from the CASPIAN trial, mainly including patients of treatment-naïve, histologically, or cytologically documented ES-SCLC, performance status score of 0 or 1, and suitability for first-line platinum-based chemotherapy [9]. Two first-line treatment regimens were evaluated in our model: (a) EP for six 21-day cycles and prophylactic cranial irradiation (PCI) after chemotherapy; (b) DEP for four cycles followed by maintenance with durvalumab. Patients continued to receive treatment until they experienced either disease progression or unacceptable toxicity. The drugs doses were administered as follows: Etoposide 80–100 mg/m2, administered on days 1–3 of each cycle. Carboplatin area under the curve 5–6 mg/mL per minute or cisplatin 75–80 mg/m2, administered on day 1 of each cycle. Durvalumab 1,500 mg every 3 weeks followed by maintenance durvalumab 1,500 mg every 4 weeks.

After progression, second-line alternative regimens consisted mainly of chemotherapy (single-agent or platinum doublet) and immunotherapy (single, immunotherapy plus chemotherapy, or two kinds of immunotherapies). According to the data reported in the CASPIAN trial, 42% and 44% of patients who progressed in DEP and EP arms, respectively, received subsequent therapy. Treatment was discontinued after further progression, at which time all patients received best supportive care (BSC) until death.

Model Structure

A Markov model was constructed to compare the costs and health outcomes of patients treated with DEP and EP strategies from the perspective of the U.S. payer. The model included three mutually exclusive health states: progression-free survival (PFS), progressive disease (PD), and death. A temporary tunnel state was inserted into the model to divide the different treatments received in the PD state. Patients started initial treatment in the PFS state, and the transitions between different states are shown in Figure 1. The cycle length was set to be consistent with treatment cycle length, and this model simulated the lifetime of patients with ES-SCLC.

Markov state transition model.

Abbreviations: BSC, best supportive care; DEP, duralumin plus etoposide and carboplatin or cisplatin; EP, etoposide and carboplatin or cisplatin; M, Markov; PD, progressive disease; PFS, progression-free survival; SCLC, small cell lung cancer.

The primary outcomes of the model were total cost, life-years (LYs), quality-adjusted LYs (QALYs), and incremental cost-effectiveness ratio (ICER). ICER was presented as the incremental cost of gaining an extra unit of QALY. The willingness-to-pay (WTP) threshold was set as $150,000/QALY, and a regimen could be regarded as cost-effective so long as the ICER was less than the WTP threshold [16]. The model was analyzed using TreeAge Pro 2017 (TreeAge Software, Williamstown, MA).

Clinical Parameter Inputs

Transition probabilities were estimated based on Kaplan-Meier survival curves from the CASPIAN trial. An algorithm derived by Hoyle et al. was applied to generate pseudo-individual patient data (IPD), and five survival distributions (Weibull, Log-logistic, Log-normal, Gamma, and Exponential) were used to parameterize the model [17-19]. The final parametric survival distributions were evaluated based on fit statistics (i.e., Akaike information criterion) and visual inspection (supplemental online Table 1; supplemental online Fig. 1). Finally, Weibull distribution was chosen as the optimal fit for the PFS curve of EP as well as overall survival (OS) curves of DEP and EP, whereas the Log-logistic provided the best fit for the PFS curve of DEP. In addition, we used the Exponential model to extrapolate progression rates beyond the duration of the clinical trial.

Costs Inputs

Direct medical costs were composed of drug acquisition and administration, PCI, disease management in PFS and PD, AEs management, BSC, and terminal care. Unit price of first- and second-line treatment drugs were acquired from the Centers for Medicare and Medicaid Services [20]. Drug administration doses were calculated based on the NCCN Guidelines recommendation; body weight was assumed as 70 kg, and body surface area as 1.82 m2 [21, 22]. Cost of PCI was sourced from economic studies associated with PCI in SCLC [23]. Disease management costs referred to hospitalization expenses, laboratory examination fees, and cost of computed tomography and magnetic resonance imaging, which were gained from previous studies [24]. AEs primarily contained events that occurred seriously (grade 3 or 4) and frequency ≥ 5% (Table 1). The costs of AE management were computed through multiplying the incidence by the individual AE unit treatment price, among which the incidence was derived from the CASPIAN trial; the cost was obtained from the Healthcare Cost and Utilization Project based on International Classification of Diseases-10 codes (supplemental online Table 2), as well as the terminal care cost [10, 25]. BSC expenses were obtained from relevant research that assessed the cost-effectiveness of ICIs in lung cancer [26]. All costs were adjusted for inflation to reflect 2019 U.S. dollars using the Consumer Price Index and discounted at a rate of 3% per year [22]. Cost inputs used in the model are listed in Table 2.

Table 1. Probabilities of grade III∼IV AEs AEs EP DEP Anemia 18 9 Thrombocytopenia 9 6 Neutropenia 33 24 Febrile neutropenia 6 5 Data are shown as %. Abbreviations: AEs, adverse events; DEP, durvalumab plus etoposide and carboplatin or cisplatin; EP, etoposide and carboplatin or cisplatin. Table 2. Ranges and distributions of parameters used in sensitivity analyses Parameters Base-case values Minimum Maximum Distributions Reference Cost, $ Durvalumab (50 mg) 361.44 289.15 433.73 Gamma 20 Etoposide (20 mg) 1.33 1.06 1.60 Gamma 20 Carboplatin (10 mg) 1.12 0.90 1.34 Gamma 20 Cisplatin (1 mg) 0.33 0.26 040 Gamma 20 Irinotecan (20 mg) 23.96 19.19 28.75 Gamma 20 Topotecan (4 mg) 196.45 157.16 235.74 Gamma 20 Paclitaxel (6 mg) 1.72 1.38 2.06 Gamma 20 Docetaxel (10 mg) 124.09 99.27 148.91 Gamma 20 Nivolumab (10 mg) 268.03 214.42 321.64 Gamma 20 Pembrolizumab (25 mg) 1,184.20 947.36 1,421.04 Gamma 20 Ipilimumab (5 mg) 741.80 593.44 890.16 Gamma 20 PCI 5,289 4,231 6,347 Gamma 23 BSC 637 478 796 Gamma 26 Terminal care 2,385 1,908 2,862 Gamma 25 Anemia 7,437 5,949 8,924 Gamma 25 Thrombocytopenia 11,461 9,168 13,753 Gamma 25 Neutropenia 12,533 10,026 15,039 Gamma 25 Febrile neutropenia 7,786 6,228 9,343 Gamma 25 Administration 144.72 115.78 173.66 Gamma 21 Disease management in PFS 1,286 1,028.8 1,543.2 Gamma 24 Disease management in PD 1,419 1,135.4 1,702.8 Gamma 24 Utility value PFS 0.673 0.538 0.808 Beta 27 PD 0.473 0.378 0.568 Beta 27 Disutilities Anemia 0.32 0.26 0.38 Beta 36 Thrombocytopenia 0.25 0.2 0.3 Beta 37 Neutropenia 0.46 0.28 0.42 Beta 21 Febrile neutropenia 0.5 0.38 0.56 Beta 21 BSA, m2 1.82 1.6 2.04 Gamma 22 Body weight, kg 70 40 160 Gamma 21 Discount rate, % 3 0 5 Fixed 22 Abbreviations: BSA, body surface area; BSC, best supportive care; DEP, durvalumab plus etoposide and carboplatin or cisplatin; EP, etoposide and carboplatin or cisplatin; PCI, prophylactic cranial irradiation; PD, progressive disease; PFS, progression-free survival. Utility Inputs

The health utility value reflects disease impact on patients’ quality of life (QoL) and usually declines as the disease progresses. A three-level EuroQoL-5D (EQ-5D) questionnaire is routinely used to evaluate QoL in clinical trials. However, the relevant data were not reported together with outcomes of the CASPIAN trial; thus, we extracted the utility values from the relevant pharmacoeconomic studies of NSCLC [27]. To be specific, we assigned a value of 0.673 for first-line therapy, 0.473 for salvage treatment, and 0 for death. Considering the pernicious effects of serious AEs on QoL, utility values were modulated by subtracting the product of multiplying incidences by corresponding disutility values of AEs (Table 2).

Sensitivity Analyses

One-way sensitivity analysis and probabilistic sensitivity analysis were carried out to explore the robustness of the model. Parameters were given a plausible range to test the effect of variation on results in the one-way sensitivity analysis. Minimum and maximum values were obtained from mean ± SD or varied ±20% around base-case values, and the discount rate ranged from 0% to 5% [16, 28]. Probabilistic sensitivity analysis was performed using a Monte Carlo simulation with 1,000 iterations, in which parameters were allotted with a specific distribution, respectively. Specifically, gamma distribution was used for costs, body surface area, and body weight; utility values were assigned a beta distribution; and the discount rate was fixed [16, 28]. Tornado diagram and cost-effectiveness acceptability curve were generated to present the results of one-way sensitivity analysis and probabilistic sensitivity analysis.

Results Base-Case Analysis

The projected life expectancy of patients treated with DEP and EP were 1.73 and 0.87 LYs and 0.93 and 0.49 QALYs, respectively. Adding durvalumab to EP provided an additional survival benefit of 0.86 LYs and 0.44 QALYs and came with an incremental cost of $95,907. The ICER of DEP versus EP was $111,015/LY. Considering the QoL, the ICER was $216,953/QALY, which exceeded the set WTP threshold of $150,000/QALY (Table 3).

Table 3. Summary of cost and effectiveness results Regimen EP DEP Incremental Total cost, $ 38,414 134,322 95,907 Total LYs 0.87 1.73 0.86 Total QALYs 0.49 0.93 0.44 ICER, $ Per LY 111,015 Per QALY 216,953 Abbreviations: DEP, durvalumab plus etoposide and carboplatin or cisplatin; EP, etoposide and carboplatin or cisplatin; ICER, incremental cost-effectiveness ratio; LYs, life years; QALYs, quality-adjusted life years. Sensitivity Analyses

Figure 2 shows the outcome of one-way sensitivity analysis. When varying key parameters individually, the price of durvalumab was found to be the most influential variable for ICER. The model was also sensitive to the alteration of utility values in PD and PFS, discount rate and costs of BSC, etc. However, across the broad variations in the ranges for each parameter, none of them resulted in ICER being below WTP threshold, indicating that parameters have a moderate impact on results.

One-way sensitivity analysis of DEP versus EP.

Abbreviations: BSC, best supportive care; ICER, incremental cost-effectiveness ratio; PCI, prophylactic cranial irradiation; PD, progressive disease; PFS, progression-free survival; QALY, quality-adjusted life-years.

The results of probabilistic sensitivity analysis suggested that the probability of DEP being cost-effective compared with EP was 9.4% at a WTP threshold of $150,000/QALY (Fig. 3). Price cuts of durvalumab were tested several times to elicit cost-effectiveness of DEP (Table 4). The 1,000 simulation iterations supported the probability that DEP is cost-effective over 50% when discounting the price of durvalumab by 30.7% ($250.58/50 mg).

Cost-effectiveness acceptability curve.

Abbreviations: DEP, durvalumab plus etoposide and carboplatin or cisplatin; EP, etoposide and carboplatin or cisplatin; WTP, willingness-to-pay.

Table 4. Discount price of durvalumab to make DEP cost-effective Price of durvalumab Incremental cost ($) Incremental QALYs ICER ($/QALY) Probability of cost-effective (%) Full cost 95,907 0.44 216,953 9.4 25%-cut cost 71,776 0.44 162,364 37.7 30.7%-cut cost 66,279 0.44 149,931 51.8 50%-cut cost 47,644 0.44 107,775 89.9 Abbreviations: DEP, durvalumab plus etoposide and carboplatin or cisplatin; ICER, incremental cost-effectiveness ratio; QALYs, quality-adjusted life years. Discussion

SCLC is an invasive disease with high malignancy and rapid proliferating speed. Although SCLC carries a lower morbidity than NSCLC, it is accompanied by poorer QoL and prognosis. Whereas new drugs for treatment of NSCLC have become available frequently for decades, it has been a conundrum to prolong the patients’ survival time with SCLC because of the lack of safe and effective therapeutic drug development. Numerous innovative antitumor drugs, especially ICIs, have been approved to treat various pernicious tumors based on significantly superior survival outcomes in the last few years, but many of them, including nivolumab and pembrolizumab, have not shown effectiveness in first-line treatment of neuroendocrine tumors such as SCLC [29-31]. Durvalumab, as the second approved regimen for first-line treatment of ES-SCLC after atezolizumab [10, 15], has shown remarkable survival benefits in ES-SCLC.

Since excessive medical expenses that go along with ICIs hinder their accessibility to all patients, it is important to assess the economic practicality of ICIs. Two studies estimated the cost-effectiveness of atezolizumab for ES-SCLC from China and U.S. perspectives, respectively, and both of them showed that the combination of atezolizumab with carboplatin and etoposide was unlikely to be an economical choice [27, 32]. Similar to atezolizumab, durvalumab is also a PD-L1 inhibitor. Although the unit cycle cost of durvalumab was slightly higher than atezolizumab ($10,843 vs. $9,280), it was superior to atezolizumab in OS benefit (mOS 13.0 vs. 12.3 months). Furthermore, the incidence of some common severe AEs, such as anemia and thrombocytopenia, was lower for durvalumab than for atezolizumab based on the CASPIAN and IMpower133 trials. Because of the improved survival time and QoL of patients with SCLC that durvalumab can promote, it is essential to study the cost-effectiveness of durvalumab for ES-SCLC [10, 15].

This study-evaluated the cost-effectiveness of DEP versus EP for treatment-naïve patients with ES-SCLC. In the base-case analysis from the U.S. payer perspective, DEP regimen costs $216,953 per extra QALY gained compared with EP, which is largely ascribed to the high cost associated with the acquisition of durvalumab. The probabilistic sensitivity analysis demonstrated that EP trumped DEP with a 90.6% possibility to be a cost-effective strategy, which concluded that the cost of durvalumab was not worth the improvement in survival benefits it yielded. In a study by Zhou et al., which appraised the economy of atezolizumab plus chemotherapy for ES-SCLC first-line treatment in the U.S., the ICER was greater than $528,810/QALY, far beyond the WTP threshold of $100,000/QALY [27]. Our ICER is much lower than Zhou's result and closer to WTP threshold, which suggested that adding durvalumab to chemotherapy might be more acceptable than adding atezolizumab. Thus, reflecting on the promising treatment improvement shown by DEP, patients who are able to afford high medical expenditure could be recommended to receive durvalumab treatment more actively.

The one-way sensitivity analysis revealed that the price of durvalumab has greatest influence on ICER, hinting that the most realistic action to make DEP the preferred strategy would be a reduction in the price of durvalumab. In light of probabilistic sensitivity analysis results and price cut tests, the reduction should be at least 30.7% at the given WTP threshold of $150,000/QALY in the U.S. This may provide reference for government and pharmaceutical companies and help them to take measures to promote a wider availability of durvalumab, like adding it to the Medicare pharmacopeia or implementing a strategy for a patient assistance program. Noticeably, there was no standard for WTP threshold established for cost-effectiveness analysis in the U.S., and a WTP threshold of $100,000/QALY to $150,000/QALY was commonly used [33]. Regardless of the level of WTP threshold, DEP cannot be considered as a cost-effective regimen. The cost-effectiveness of DEP would be overestimated, even if the price were discounted by 30.7%. In addition, according to the drug label and NCCN Guidelines, durvalumab is approved for the treatment of urothelial carcinoma and stage III NSCLC with a dose of 10 mg/kg every 2 weeks, but the recommended dose by NCCN is 1,500 mg every 3 or 4 weeks for SCLC [9, 34], which is higher than other treatments. That is to say, if the dose can be reduced with guaranteed efficacy, the cost-effectiveness of DEP may be improved to some extent.

Our study has some strengths. Generally, a considerable impact on analysis results will be exerted by a subtle change in curve fits. In this study, we generated the pseudo-IPD based on the time-to-event information from Kaplan-Meier curves. Compared with the traditional fitting model, ours avoided the unfavorable influence of curve tails’ uncertainty resulting from a small number of patients at risk so as to enhance the accuracy of analysis results [17]. In addition, multiple methods were tested to extrapolate OS past the end of the clinical trial periods to promote the accuracy of the simulation as much as possible. The mOS values were 13.3 and 10.4 months for DEP and EP, respectively, in our model cohort analysis and were commensurate with those reported in the CASPIAN trial, 13.0 and 10.3 months, respectively (supplemental online Fig. 1). Based on our analysis, durvalumab at its current price is economically less advantageous. The provision of cost-effective care requires reasonable pricing and a supportive payment system. The process for approving new drugs and incorporating them into treatment guidelines should balance costs and clinical benefits. The price cutting tests in our research can offer decision-making information for this purpose.

Some limitations in analysis should be noted, mainly governed by model assumptions and data availability. First, we hypothesized all patients receive BSC if their disease progressed after second-line treatment. Actually, 12% and 14% of patients in our trial accepted third-line therapy, so the model may not accurately reflect the disease course of ES-SCLC and, thus, undervalue the effectiveness of regimens. However, one-way sensitivity analysis showed that the cost of second-line therapy and the probability of patients receiving subsequent therapy exert only a minor impact on the result. Second, the utility values of this model referred to that of NSCLC [27, 32] because there is no study directed at QoL in patients with SCLC, and, further, the utility values used in pharmacoeconomic studies of SCLC mainly referenced those of NSCLC [35]. Despite the slight discrepancy of utility values between patients with SCLC and NSCLC, the variation in any given range did not qualitatively change the outcome. The EQ-5D questionnaire can be used to investigate patients’ QoL in routine clinical practice of any specific protocol to make the results more accurate in future research. Third, grade I–II AEs and those with an incidence of <5% were excluded from our model because of their low cost and low impact on outcomes, and this may result in overvaluation of utility and undervaluation of AE management costs. Fortunately, the bias this factor introduces is expected to be small because there was no significant difference in incidence of excluded AEs between the EP and DEP groups. Meanwhile, we found that the incidence of most common grade III–IV AEs, such as hematological toxicities, was numerically higher in the EP group than in the DEP group, possibly owing to the two more cycles of chemotherapy received by patients in the EP group, suggesting that moderate AEs may be caused by durvalumab. A real-world analysis should be conducted to compare the incidence of severe AEs between EP and DEP groups in the same situation. Finally, our study directly compared durvalumab plus chemotherapy with chemotherapy alone based on the CASPIAN trial. Although there are other potential first-line treatments for ES-SCLC, our study did not indirectly compare them because of the lack of robust head-to-head trial data. Further analyses could be conducted to evaluate economy among a couple of protocols or ICIs.

Conclusion

According to this analysis, the combination of