First-line Immuno-chemotherapy for extensive-stage small-cell lung cancer: A network meta-analysis and cost-effectiveness analysis

Introduction Many randomized controlled trials have indicated that immuno-chemotherapy could generate clinical benefits, though the cost of immuno-chemotherapy was so prohibitive and the options were varied. This investigation aimed at evaluating effectiveness, safety, and cost-effectiveness for immuno-chemotherapy as a first-line therapeutic option for ES-SCLC patients. Methods Multiple scientific literature repositories were searched for clinical studies where immuno-chemotherapy was regarded as the first-line treatment for ES-SCLC, which were published in English between Jan 1, 2000, and Nov 30, 2021. This study conducted a network meta-analysis (NMA) and cost-effectiveness analysis (CEA) based upon US-resident payer perspectives. Overall survival (OS), progression-free survival (PFS), and adverse events (AEs) were evaluated through NMA. In addition, costings, life-years (LYs), quality-adjusted life-years (QALYs), and incremental cost–benefit ratio (ICER) were estimated by CEA. Results We identified 200 relevant search records, of which four randomized controlled trials (RCTs) (2,793 patients) were included. NMA demonstrated that the effect of atezolizumab plus chemotherapy was ranked at a more elevated position in comparison to other immuno-chemotherapy options and chemotherapy alone, within the general population. The influence of atezolizumab plus chemotherapy and durvalumab plus chemotherapy was ranked higher within populations experiencing non-brain metastases (NBMs) andbrain metastases (BMs), respectively. The CEA revealed that the ICERs of immuno-chemotherapy over chemotherapyalone were higher than the willingness-to-pay (WTP) threshold of $150,000/QALY in any population. However, treatment with atezolizumab plus chemotherapy and durvalumab plus chemotherapy were more favorable health advantages than other immuno-chemotherapy regimens and chemotherapy alone, and the results were 1.02 QALYs and 0.89 QALYs within overall populations and populations with BMs, respectively. Conclusion The NMA and cost-effectiveness investigation demonstrated that atezolizumab plus chemotherapy could be an optimal first-line therapeutic option for ES-SCLC when compared with other immuno-chemotherapy regimens. Durvalumab plus chemotherapy is likely to be the most favorable first-line therapeutic option for ES-SCLC with BMs.


Introduction
Lung cancer has the second-highest morbidity and highest mortalityamong all cancer models globally, with over 2.2 million and 230,000 cases diagnosed, and over 1.79 million and 130,000 deaths occurring globally and within the United States (US) in 2021, respectively (1,2). Small cell lung cancer accounted for more than 10% of lung cancer, and up to 60% were diagnosed as extensive-stage small cell lung cancer (ES-SCLC), with a 5-year survival rate of only 2% (3)(4)(5). The most common distant metastases were brain metastases (BMs), which are prevalent within 10% of such clinical cases at initial diagnosis, accounting for more than 50% incidence within 2 years (6).
During the past 30 years, etoposide plus platinum (EP) was established as a first-line chemotherapeutic option for ES-SCLC, though the survival of patients has not improved significantly, and patients typically endure recurrence within 1-2 years. A phase III clinical data of ES-SCLC demonstrated that the survival time of the chemotherapeutics group increased by only 0.63 days per year (7). Therefore, it is necessary and urgent to develop new drugs to treat ES-SCLC.
However, considering that there is no research to directly compare different immuno-chemotherapy regimens, it is not clear which therapeutic option must be recommended as initial treatment in such clinical cases. Based upon present healthcare scenarios and relevant stakeholders, we need more proof to validate different immunochemotherapy within oncology health care to provide effective medical leverage with decent costings. Consequently, this investigation employed recently reported randomized controlled trials (RCTs) for network meta-analysis (NMA) and cost-effectiveness analysis (CEA) for evaluating effectiveness, safety, and cost-effectiveness for immunochemotherapy and chemotherapy alone as the initial therapeutic option for ES-SCLC clinical cases, from a US payer perspective.

Methods
This work was guided by the PRISMA statement, which included a PRISMA NMA checklist and the consolidated health economic evaluation reporting standards statement (CHEERS) checklist (Supplementary Tables 2, 3 within the Supplementary material).

Search strategy and inclusion criteria
A systematic review and NMA were conducted for identifying eligible phase III RCTs to compare regimens containing ICIs plus chemotherapy in first-line treatment. We retrieved the Pubmed, Embase, Cochrane, and Web of Science databases for published articles written in English from Jan 1, 2000, to Nov 30, 2021, with the search terms "PD-1, " "PD-L1, " "immunotherapy, " "chemotherapy, " "extensive-stage small-cell lung cancer, " and "clinical trial" (Supplementary Table 1 in the Supplementary material). In addition, the investigation also focused on abstracts reported by the American Society of Clinical Oncology (ASCO) and the European Society of Medical Oncology (ESMO). Finally, relevant literature was manually screened to avoid missing articles.
Inclusion criteria: (1) patients diagnosed with ES-SCLC; (2) articles in which participants received both types of treatment, one of which was immuno-chemotherapy and the other was chemotherapy; (3) both treatment measures were in the initial treatment environment of ES-SCLC patients; (4) phase III RCTs; (5) the article had the most complete and updated data of the trial; (6) studies published in English. Studies not matching the inclusion criteria were excluded. YWZ and KL carried out literature retrieval and data extraction independently. Whenever duplicate studies were identified, the article having the most comprehensive and recent investigation data were included. Reviews / systematic reviews, meta-analyses, and CEAs were excluded from this investigation.

Data extraction and determination of bias risks
Details were extracted from identified articles, such as author, publication year, trial name or identification, treatment regimens of experimental groups and control groups, number of patients treated, HR of OS and PFS of the overall population, median OS and PFS, together with the incidence of grade 3/4 AEs from each included investigation. Additionally, the odds ratio (OR) of grade 3/4 AEs and the HR of OS and PFS of the population with BMs or non-brain metastases (NBMs) were extracted.
Individual RCT article bias risks were evaluated in line with the Cochrane Collaboration guideline (16), valuating multiple facets for RCT experimental designs, behavior, and detail Frontiers in Public Health 03 frontiersin.org descriptions. Seven tools were used to assess individual RCT results, namely: (1) random sequence generation, (2) allocation concealment, (3) blinding of participants and personnel, (4) blinding of outcome assessment, (5) incomplete outcome data, (6) selective reporting, and (7) other bias.

Statistical analysis
R software (version 4.1.1) 1 with the package "netmeta" was employed for comparative analysis. We combined the HR and 95% CI that was collected. However, since just one RCT informed individual pair-wise comparisons, with paucity in datasets for evaluating heterogeneity across trials, a fixed-effect model was established. Consequently, the frequency method was employed for comparing effectiveness and safety for different schemes. The HR of OS and PFS, corresponding 95% CI, p-value, and OR of AEs were calculated. Subgroup analyses were performed on status with or without BMs. Finally, according to the obtained 95% Cl of HR and p-value, the best treatment schemes were sorted.  Figure 5 in the Supplementary material) was established to assess costings and efficacy for different initial patient treatments for ES-SCLC. The Markov model cycle was determined to be 6 weeks based on the patient's survival and dosing follow-up protocol. Since tremelimumab has not obtained obvious clinical benefits and was not listed, the decision trees included 5 initial therapeutic options: (1) atezolizumab plus chemotherapy, (2) durvalumab plus chemotherapy, (3) pembrolizumab plus chemotherapy, (4) ipilimumab plus chemotherapy, and (5) chemotherapy. Over time, the patient's health status deteriorated and led to mortality, with more than 99% of the registered patients dead over the last 15 years. All patients started PFS status and could receive five kinds of initial treatment strategies randomly. Upon PD or unacceptable toxicity and AEs, some patients received topotecan as subsequent treatment, according to Koichi Goto's recommendations (17); Other patients received supportive treatment (15). To better reflect the current clinical work, the study considered that patients received palliative treatment before the mortality event. All doses and dosing schedules for each treatment regimen were collected from corresponding RCTs (9-12) (Supplementary Table 5 in the Supplementary material).
The study adopted costings and influence from a 3% discounted rate per year (18). The outputs encompassed overall cost, life-years (LYs), quality-adjusted LYs (QALYs), and incremental costeffectiveness ratios (ICERs). The study also focused on population CEA with or without BMs. Depending upon the U.S. consumer-price index, all costings related to healthcare services were inflated to the 1 http://www.r-project.org value of 2021, and willingness-to-pay (WTP) in the United States was $150,000 (19,20

Model survival and progression risk estimates
This research implemented GetData Graph Digitizer ® (version 2.26) 3 for gathering data from OS and PFS curve-strategy from RCTs. Consequently, we reconstructed the OS and PFS curves of chemotherapeutics patients depending upon Kaplan-Meier (KM) chemotherapeutic curves of four RCTs and such data were consequently employed for fitting parametric survival models. Peakconsistent Weibull distribution was chosen depending upon Akaike's information criterion (AIC) and Bayesian information criterion (BIC) (Supplementary Table 6 . where Markov cycle = u, arrival at state t after u Markov cycles i = tu, respectively.

Cost and utility estimates
This study considered just immediate medical expenses from a US payer perspective, including drug costs (24), AEs costs (with the assumption that AEs occurred within just 1 cycle during PFS and PD states) (20,23,25,27), administration, tumor imaging, laboratory (23), and death associated costs (25), and best supportive care (28).
Based on four RCTs and clinical practice, carboplatin was selected as the main treatment regimen in the chemotherapeutics group. Once drug cost per cycle was determined, assuming the patient was male-gender, 65 years old, weighing 70 Kg, the height of 170, and body-surface-area 1.84m 2 , area-under-concentration (AUC) curve of 5 mg/ml/min, together with presumed serum creatinine being 1 (29). Medical monitoring costings encompassed financial charges for computed tomography or magnetic resonance imaging (at six-week intervals for the initial 48 weeks and 9-week intervals afterward) (9,11). This study solely added costings for managing grade 3/4 AEs (frequency > 5%) within this model that had distinctly varying probabilities across RCT arms. The entirety of costings linked to healthcare provisions was inflated to correspondent values in 2021, depending upon the US consumerprice index (Table 1).
We used previously published utilities of 0.673 and 0.473 (25) as the mean health utility value for PFS and PD states, accordingly. This investigation also included dis-utility values of grade 3/4 AEs within analysis (23,25,27).
Frontiers in Public Health 04 frontiersin.org

. Sensitivity and scenario analysis
This investigation employed serial sensitivity evaluating predictions for modeling outcome uncertainties. One-way sensitivity evaluation was performed within a variance of 20% baseline values, depending upon varying values for a specific parameter (within the expected range) and pre-determined methodologies for examining individual parameter-driven influences over ICERs (23). This investigation additionally conducted probabilistic sensitivity analyses for evaluating the probability of efficacy by therapeutic regimens through 10,000 Monte Carlo repetitions. A cost-effectiveness adequacy curve for individual therapeutic modalities was assessed to present probabilities of cost-effectiveness.
Subgroup analyses were performed on status with or without BMs of four RCTs. Due to insufficient data for several RCTs, this investigation used identical pooled chemotherapeutics KM to obtain depending upon subgroup-defined HRs, as described by Hoyle (30) for lack of OS and PFS curves regarding BMs status of subgroups.
In addition, we conducted a scenario analysis, where ICIs maintenance phase until death after 4 cycles of first-line treatment, for evaluating if maintenance time for ICIs had a major influence on this investigation's outcomes.

Included studies
We searched 200 records, and 63 eligible articles were searched in full text. After screening, four cluster RCTs, involving 2,793 patients, were included (Supplementary Table 4

Risk-bias proof evaluations
We employed RevMan ® (version 5.4) to summarize risk-bias (Supplementary Figure 2 in the Supplementary material). Two studies were designated as cluster RCTs and employed randomization concealment. Three investigations were described as double-blinded. Three investigations were found to have reduced risk-bias due to blinding of outcome evaluation, while all studies were judged to have a low risk of bias for incomplete outcome data and selective reporting.

Results of the network meta-analysis
The network plots were built using R software (version 4.1.1), including five immuno-chemotherapy regimens (atezolizumab plus chemotherapy, durvalumab plus chemotherapy, durvalumab with tremelimumab plus chemotherapy, pembrolizumab plus chemotherapy, and ipilimumab plus chemotherapy) and one control regimen (chemotherapy) (Supplementary Figure 3 in  The best treatment results were ranked according to value of p (individual outcomes), where raised values were more successful. Among the overall populations, the regimen having peak value of p for OS was durvalumab plus chemotherapy (p = 0.78), followed by  Table 7 in the Supplementary material). The general safety of immunochemotherapy ranked from high to low for all AEs was as follows: chemotherapy (probability 90%), ipilimumab plus chemotherapy (56%), atezolizumab plus chemotherapy (52%), durvalumab plus chemotherapy (52%), pembrolizumab plus chemotherapy (37%), and durvalumab with tremelimumab plus chemotherapy (13%). The general safety of immuno-chemotherapy ranked from high to low for severe AEs was as follows: chemotherapy (70%), atezolizumab plus chemotherapy (63%), durvalumab plus chemotherapy (57%), pembrolizumab plus chemotherapy (43%), ipilimumab plus chemotherapy (38%), and durvalumab with tremelimumab plus chemotherapy (31%).

Results of the cost-effectiveness analyses
Regarding ES-SCLC cases, this investigation expressed the output effects of five interventions by QALYs (LYs), from more to less was as follows: atezolizumab plus chemotherapy (1.02 QALYs and 1.91 LYs), durvalumab plus chemotherapy (1.01 QALYs and 1.90 LYs), pembrolizumab plus chemotherapy (0.93 QALYs and 1.80 LYs), ipilimumab plus chemotherapy (0.85 QALYs and 1.55 LYs), and chemotherapy (0.77 QALYs and 1.44 LYs). The least total cost of each treatment regimen was ranked from high to low as follows: the total cost of ipilimumab plus chemotherapy was the highest, which was $568,657, followed by pembrolizumab plus chemotherapy ($241,682), durvalumab plus chemotherapy ($229,620), and atezolizumab plus chemotherapy ($213,988). The lowest total cost of chemotherapy was $133,625. Post-further analysis, atezolizumab plus chemotherapy, durvalumab plus chemotherapy, pembrolizumab plus chemotherapy, ipilimumab plus chemotherapy obtained an ICER of $321,452/QALY, $399,978/QALY, $675,358/QALY, and $5,437,894/QALY, respectively, compared with chemotherapy. The baseline results and pairwise comparison of ICER were shown in Table 2; Supplementary Table 8.
The one-way sensitivity analysis showed it was highly sensitive for the utility of PD against chemotherapy. Other considerable influences were the risk of neutropenia in the chemotherapy group or immunochemotherapy group, cost of ICIs, and utility of PD. Alternative factors encompassed within sensitivity analysis, such as the costing and disutilities of AEs, had a minimal impact on ICER (Supplementary Figure 8 in the Supplementary material).
Dataset outcomes for acceptability curves ( Figure 3) and ICER scatterplot (Supplementary Figure 9 in the Supplementary material) demonstrated that the probability of atezolizumab plus chemotherapy, durvalumab plus chemotherapy, pembrolizumab plus chemotherapy, and ipilimumab plus chemotherapy being cost-effective were 32, 29 10, 0% in the overall population, respectively, compared with that of chemotherapy a WTP threshold of $150,000.
Regarding in the Supplementary material) showed that the probability of atezolizumab plus chemotherapy, durvalumab plus chemotherapy, pembrolizumab plus chemotherapy, and ipilimumab plus chemotherapy being costeffective were 12 and 34%, 39 and 20%, 0 and 19%, 0 and 0% in the population with BMs and NBMs, compared with that of chemotherapy a WTP threshold of $150,000, respectively.
Scenario-analysis outcomes suggested that ICIs maintenance therapy resulted in the health costings linked to initial treatment increasing drastically, though this investigation's outcome was not altered. This investigation assumed that clinical cases had ICIs maintenance therapy until death after 4 cycles of initial treatment, whereby health costs of the first-line atezolizumab plus chemotherapy, durvalumab plus chemotherapy, pembrolizumab plus chemotherapy, and ipilimumab plus chemotherapy were $279,513, $326,911, $306,097, and $1,271,747, respectively. An the ICERs were $355,700, $519,417, $731,140, and $5,963,788 per QALY, respectively.

Discussion
Recently, the promotion of ICIs has vastly shifted therapeutic options for ES-SCLC patients. Some encouraging results of phase III clinical studies demonstrated that introducing atezolizumab, durvalumab, durvalumab plus tremelimumab, pembrolizumab, and ipilimumab to chemotherapy shows clinical activity. Considering that these expensive drugs have brought a heavy burden on social health resources and patients, it is unclear which treatment regimen has the best efficacy and safety in the first-line treatment of ES-SCLC. Consequently, this investigation pioneered a comprehensive comparative clinical trial of immuno-chemotherapy and proved that one of the ICIs has better efficacy, safety, and overall economic outcomes. The results of NMAs indicated that atezolizumab plus chemotherapy and durvalumab plus chemotherapy regimens produced more survival benefits in patients with NBMs and BMs than other immuno-chemotherapy regimens and chemotherapy, respectively. Furthermore, the survival advantages of atezolizumab plus chemotherapy and durvalumab plus chemotherapy translated into the highest QALYs in patients with NBMs and BMs, respectively. All five immuno-chemotherapy regimens were associated with all levels of AEs risk, and ipilimumab plus chemotherapy strategy was linked to lowered risk for all-grade AEs (all levels) in comparison to chemotherapy. Unexpectedly, the safety of immuno-chemotherapy regimens is lower than that of chemotherapy strategy, which could be due to the combined regimens summarize AEs of ICIs and chemotherapy. Consequently, this assessment reflects the universal profiles of the current research results.
Frontiers in Public Health 08 frontiersin.org Hazard ratios (gray and brown cell) and p-values (blue cell) of the network meta-analysis of the overall survival in the overall population (A), population with brain metastases (B), and population with non-brain metastases (C). AC, atezolizumab plus chemotherapy; DC, durvaluma plus chemotherapy; DTC, durvalumab with tremelimumab plus chemotherapy; PC, pembrolizumab plus chemotherapy; IC, ipilimumab plus chemotherapy; C, chemotherapy.
Frontiers in Public Health 09 frontiersin.org Hazard ratios (gray and brown cell) and p-values (blue cell) of the network meta-analysis of the progression-free survival in the overall population (A), population with brain metastases (B), and population with non-brain metastases (C). AC, atezolizumab plus chemotherapy; DC, durvaluma plus Frontiers in Public Health 10 frontiersin.org  The baseline results of the CEA indicated that atezolizumab plus chemotherapy and durvalumab plus chemotherapy were the most effective strategies and provided the best treatment outcome in the NBMs and BMs populations, respectively. When it talks about costeffectiveness according to relevant studies, immuno-chemotherapy regimens would be favored by clinical cases having reduced HRs for OS, while in patients with higher HRs it can become worse than chemotherapy (25,31). Although atezolizumab plus chemotherapy and durvalumab plus chemotherapy provided 1.02 and 0.89 QALYs in patients with NBMs and BMs, respectively, whose QALYs were much higher than the other four treatment measures, they increased the survival benefit by 0.25 and 0.12 QALYs and the additional cost of $80,363 and $74,562, resulting in an ICER = 321,452 and 621,350/ QALY, that is higher than WTP in the US, making it not cost-effective, in comparison to chemotherapy, respectively. Finally, modeling outcomes demonstrated that neither treatment plans were costeffective in comparison to chemotherapy, in line with outcomes of several past investigations. However, chemotherapy alone was not enough to greatly improve the survival and prognosis of patients with ES-SCLC. Therefore, in addition to chemotherapy in first-line treatments, the most effective treatment strategy was to use atezolizumab plus chemotherapy for NBM cases and durvalumab plus chemotherapy for BM cases. Sensitivity analysis shows that the utility of PD was the most important factor influencing ICER value, followed by the incidence of AEs, and the price of ICIs are also factors that cannot be ignored. Since the price of ICIs is much higher than chemotherapy in the US, subsequent probabilistic sensitivity analysis results confirmed that atezolizumab plus chemotherapy and durvalumab plus chemotherapy were cost-effective in 32, 29, and 12%, 39% of the overall population and population with BMs, respectively. The results of the acceptable curve revealed that the US-based ICER value was affected by the shift in WTP value, while the US-based WTP value was affected by the per capita GDP. The average per capita US-based GDP value was adopted in our investigation (32). However, the per capita GDP of different regions in the US varies, so for several economically underdeveloped regions, the optimal strategy could be chemotherapy among the overall population. Regarding economically developed regions, atezolizumab plus chemotherapy and durvalumab plus chemotherapy were the preferred treatment options for the overall population and brain metastases, respectively.
The current assessment has several implications. On the one hand, patient survival has improved significantly with the introduction of ICIs. However, data was scarce for its efficacy within BM cases, and few clinical trials have been conducted for BMs alone. Patients with BMs were either excluded or only included in subgroups within key trials. The brain micro-environment itself has immunosuppressive effects, so it can promote the development of various tumor tissues and block anti-tumor immune responses (33-35). It is currently well established that chemotherapy can increase the efficacy of ICIs (36 (38). Regarding the balance of the health care system, ensuring that patients with specific characteristics have access to safe, effective, and innovative treatments is as important as minimizing economic toxicity. On the other hand, immunotherapy was improving the therapeutic efficacy of SCLC. Physicians and administrators need to select proper patients who can benefit from this type of therapy to maintain our healthcare system and establishing prognostic and response predictive markers was critical. PD-L1 expression, tumor mutational burden (TMB), and tumor-infiltrating lymphocytes (TILs) can be reliable prognostic biomarkers in small-cell lung cancer (SCLC) (39)(40)(41). However, our study did not perform an analysis of biomarkers, so further studies are needed in future work to explore biomarkers to determine which patients with heterogeneous diseases are likely to benefit more from treatment so that treatment can be tailored to the individual.
Although this study has important strengths, some limitations should be considered. Firstly, when using the NMA method to indirectly compare immuno-chemotherapy regimens, we assumed that the included studies did not differ in patient characteristics and summarized the chemotherapy groups. Secondly, the inference of long-term survival benefit is depending upon short-term survival data of each experiment, which will alter upon change of long-term follow-up. This is an inevitable limitation in our model. Consequently, it is necessary to evaluate the concordance of such modeled health outcomes with real-world data. Thirdly, for enhanced analysis, this investigation assumed that all chemotherapy regimens used carboplatin, which was safer in the clinic. The cost of carboplatin was higher than that of cisplatin, so the cost of chemotherapy can be overestimated. However, sensitivity analysis demonstrated that the cost of carboplatin has little impact on the model results. Fourthly, several trials lacked survival data from subgroups, and the original group balance was produced by Hoyle's methods. Consequently, the results of the subgroups analysis should be interpreted carefully. Fifth, this investigation analyzed the cost-effectiveness of patients with or without BMs. However we did not investigate the economic results of other subgroups, such as age, gender, smoking status, and liver metastasis. Sixth, due to the lack of complete QoL data to calculate the utility values, we referred the mean health utility value of NSCLC in PD state, and corrected the utility values by considering the disutility values of AEs and only 3/4 AEs were included, which might lead to overestimates or underestimates of the utility values. Finally, this investigation did not include social costs, including those related to the informal and non-health sectors.
In conclusion, immuno-chemotherapy regimens appear to be superior to standard chemotherapy. Among the five immunochemotherapy strategies, atezolizumab plus chemotherapy regimen seem to have the best effect on ES-SCLC patients other than BMs; durvalumab plus chemotherapy option can be a favorable condition for the population with BMs. Whereby, from the perspective of the US payer, the first-line use of four clinically effective immunochemotherapy regimens to treat ES-SCLC patients is not cost-effective in comparison to chemotherapy, though atezolizumab plus The cost-effectiveness acceptability curves for the atezolizumab plus chemotherapy, durvaluma plus chemotherapy, pembrolizumab plus chemotherapy, ipilimumab plus chemotherapy strategies compared to the chemotherapy strategy in the overall population.
Frontiers in Public Health 12 frontiersin.org chemotherapy regimen can provide a more effective balance across ICER and QALYs in the overall population. Within BM clinical cases, durvalumab plus chemotherapy program obtain more health benefits. This finding can help physicians make decisions in clinical work and aid policy formulation in medical reimbursement.

Data availability statement
The original contributions presented in the study are included in the article/Supplementary material, further inquiries can be directed to the corresponding author.

Ethics statement
This article is based on previously conducted studies and does not contain any new studies with human participants or animals performed by any of the authors, it does not require the approval of the independent ethics committee.

Author contributions
YZ, KL, QY, MZ, and LP: designed experiment, analyzed the data, wrote the manuscript, and complete the revision. YZ and KL: performed the experiments. LP and MZ: contributed analysis tools and funding. All authors have read and approved the manuscript.