The cost comparison between NAC and AC for lung cancer patients has not been extensively studied. We began our study searching the PubMed database to identify studies published before January 2020 that analyzed the cost-effectiveness of NAC and AC in NSCLC using the following search terms: adjuvant chemotherapy and neoadjuvant chemotherapy, cost-effectiveness; cost-effectiveness, adjuvant chemotherapy, neoadjuvant chemotherapy and lung cancer; cost-effectiveness, preoperative, postoperative, chemotherapy and lung cancer. Seven studies evaluated the cost-effectiveness of NAC and AC; however, none assessed the cost-effectiveness specific to lung cancer. Of the seven cases mentioned, six were studies related to ovarian cancer and one related to head and neck cancer [14,15,16,17,18,19,20].
Among these studies, head and neck cancer revealed that NAC is more cost-effective than AC . Four of the studies [14,15,16,17] related to ovarian cancer showed similar results and other two studies [18, 19] indicated AC as the dominant strategy. Findings from previous research studies stated that the therapeutic regimen is more cost-effective but these findings are not consistent.
The findings in our study showed that NAC is more cost-effective than AC, with a cost saving of ¥3064.90 and a QALY increment of 0.10 years per patient. In contrast to previous studies, the input parameters in our model included the cost of chemotherapy AEs; only one study by Tran et al.  explicitly incorporated the chemotherapy AE into their model. One possible explanation for is that there were no significant differences in the chemotherapy-related toxicities for NAC and AC in ovarian cancer and head and neck cancer [15, 29]. For NSCLC patients however, the tolerability of chemotherapy and the ratio of AE are significantly different in NAC and AC as supported by the NATCH three-phase trial  and the study by Brant et al. . Nonetheless, the difference in tolerability of chemotherapy and the ratio of AE does not contribute to OS. In addition, the treatment expense of grade 3 and 4 AEs are even higher than the surgery procedure cost . Thus, although the result in our model was not sensitive to the ratio and cost of AEs, we believe the cost comparison between NAC and AC needs to consider the impact of AEs.
The sample population in our study is cT2-4N0-1 NSCLC patients excluding stage IV patients (for whom NCCN guidelines recommend two treatment strategies). The choice of NAC and AC is a tough one in the initial treatment phase. The patients who are less clinically at-risk benefit more from AC, while the stage IV patients are recommended systemic therapy by NCCN and there is robust evidence in support of same . Thus, our study focused on the sample population of patients whose treatment strategies were controversial.
However, most studies compared NAC or AC with the treatment of surgery alone, and estimated the survival benefit; very few studies directly compared the two chemotherapy approaches [7, 8]. The head-to-head comparison of the studies of NATCH and Brandt et al. in light of NAC and AC, showed that there were no statistically significant differences in the OS and DFS. However, the NATCH trial was criticized for being overly optimistic and over representing the study design [7, 8]. The percentage of stage I diseased patients in the cohort who did not benefit from chemotherapy is 75%. In comparison with the meta-analysis , the stage I diseased patients in the NAC cohort accounted for nearly 50% of the group. This is the reason base-case probabilities are centered on the study of Brandt et al. in our model.
Furthermore, our study used real-world data. The study generated two groups (92 in NAC and 92 in AC) with comparable characteristics through strict exclusion criteria and propensity score matching analyses to prevent selection bias related to a nonrandomized cohort. The ratio of males and females more closely reflects the real-word population of NSCLC patients who need to receive either NAC or AC.
What is more, the study sample population excluded the patients with microscopic and macroscopic residual disease (R1/R2 resection), which avoids the influence of surgery discrepancy (since the surgery which results in resection to minimal or no gross residual disease may be associated with a long-term survival advantage). The single-center data source reduced the effectiveness of surgery.
There are some limitations to our model. As with all cost-effectiveness analyses, assumptions in clinical base-cases, cost and QOL are important to the projected outcomes determined by the model. Consequently, one-way and probability sensitivity analyses were performed to test our assumptions. The sensitivity analyses showed that our model was robust enough to handle to the variation of cost, QOL, ratio of complication and AEs. However, the variation of OS would change the conclusion of the cost-effectiveness analysis in our model.
The median OS is the most sensitive parameter in our cost-effectiveness analysis model. The studies of Brandt et al., the NATCH trial and Tim et al. all showed that the median OS of NAC and AC have no significant differences [3, 10, 12, 13]. In fact, the difference (< 0.19 years) of the median of NAC and AC (9.22 vs. 8.98 year in Brandt et al.) is enough to change the conclusion of our model. If 9.22 and 8.98 years as the OS of NAC and AC in our model are used, then NAC is more cost-effective with the ICER of ¥22,560/QALY. Given the concern of survival in lung cancer treatment for NSCLC patients, it is important to evaluate sensitivity of OS in cost-effectiveness analysis.
Simultaneously, there are several assumptions in the cost. To make the model clear and accurate, our cost measures were intentionally confined to the associated costs of the initial treatment phase. This was also based on the assumption that there were no significant differences between treatment and ongoing care in the NAC and AC groups beyond the initial recovery period. However, if long-term surgery complication or chemotherapy AEs affected one group and increased the follow-up medical treatment, the difference of NAC and AC cost may be improperly over or underestimated. Furthermore, one patient may not have once AE in the chemotherapy treatment.
In addition, probabilities used in estimating surgery complication and postoperative death may be overrated in NAC because patients with comorbidities or more complex diseases may be more likely to receive NAC. Hence, the reason the ratio of related complication in NAC is higher than AC in our model. In the NATCH trial however, the postoperative death of AC is higher than NAC (5% vs. 7.5%) and the ratio of complication in the multicenter randomized controlled trial (RCT) is influenced by the level of the surgery team. For the case of chemotherapy tolerance, our model did not consider the probability of completed chemotherapy (full dose and full cycles). The chemotherapy AEs of NAC and AC had no significate difference (25.4% vs. 27.3%) in the NATCH trial. Thus, the base-case probability may change in the future with more comparative research in the area of NAC and AC.
Currently, no comparative studies have examined QOL in NAC and AC for NSCLC patients . Hence, we assumed that the health utility weight of NAC and AC is the same at the various treatment stages. We also used health utility weights from previously published literature with NSCLC treatment phase whenever possible. There is also a difference in psychological effects after NAC and AC.