Therapeutic efficacy of chemotherapy cycles within different FIGO stage
Exploratory subgroup analyses of PFS and OS were generally consistent with the overall findings that ≥5 cycles of chemotherapy was superior to 1-4 cycles (Figure 3 ). An additional analysis by FIGO stage also suggested that significant improvements in PFS and OS were identified in long-course chemotherapy in FIGO stage IIB-IIIC (n=17 for ≥5 cycles, n=24 for 1-4 cycles; HRPFS 0.41, 95% CI 0.18-0.92; HROS0.41, 95% CI 0.17-0.95), but not entirely significant improvements in FIGO stage I-IIA (n=32 for ≥5 cycles, n=30 for 1-4 cycles; HRPFS 0.67, 95% CI 0.34-1.34; HROS0.88, 95% CI 0.40-1.97); these results were also confirmed in survival curves (Figure 4 ). Additionally, in multivariate analyses, ≥5 cycles of chemotherapy significantly reduced the risk of disease progress and death, and was independently prognostic for PFS and OS in FIGO stage IIB-IIIC after adjusting for age and histology; nevertheless, such significances were not achieved in FIGO stage I-IIA (Table S4 ). These findings indicated that the survival benefit of long-course chemotherapy in SCNEC can be predicted by FIGO stages; ≥5 cycles of chemotherapy was inclined to achieve favourable survival outcomes in FIGO stage IIB-IIIC, but comparable survivals in FIGO stage I-IIA.
Interaction effect of chemotherapy cycles and regimen
Because of the multiple chemotherapy regimens used in the cohort, we then evaluated the optimal regimen for SCNEC. We observed that patients who received EP regimen had obvious prolonged PFS (median PFS: EPvs non-EP, 44.7 months vs 18.0 months; Figure S1A ) and OS (median OS: EP vs non-EP, 63.3 months vs 41.0 months; Figure S1B ) than those treated with non-EP regimen, although the statistic significances were not achieved partly owing to the limited sample size. Next, we investigated the interaction between chemotherapy regimens and courses by incorporating them together, and divided the patients into the following four groups: EP with 1-4 cycles (EP 1-4), EP with ≥5 cycles (EP ≥5), non-EP with 1-4 cycles (non-EP 1-4 cycles), and non-EP with ≥5 cycles (non-EP ≥5). Of the whole cohort, we observed significantly different PFS among these four groups, with EP ≥5 obtaining most satisfied survival (P = 0.011, Figure S2A ); whereas the statistic significance of OS was not achieved (P = 0.160, Figure S2B ). Then we performed subgroup analyses stratified by FIGO stage; interestingly, we identified that these four groups obtained significances of both PFS (P = 0.026, Figure S2E ) and OS (P = 0.042, Figure S2F ) in FIGO stage IIB-IIIC, but neither significant of PFS (P = 0.370, Figure S2C ) nor OS (P = 0.860, Figure S2D ) in FIGO stage I-IIA. Hence, we can conclude that EP regimen with ≥5 cycles should be proposed for patients with FIGO IIB-IIIC SCNEC, nonetheless the optimal regimen and course for FIGO stage I-IIA SCNEC need further investigation.

Discussion

In this retrospective cohort of SCNEC, we observed an inverse correlation between chemotherapy cycles and progression/death; the risks of disease progression and mortality decreased sharply until 5 cycles of chemotherapy. Long-course chemotherapy was associated with significantly superior PFS and OS than short-course chemotherapy. Additionally, FIGO stage was predictive to the therapeutic efficacy of long-course chemotherapy; ≥5 cycles of chemotherapy was superior to <5 cycles of chemotherapy in terms of PFS and OS in FIGO stage IIB-IIIC, whereas such superiority was not observed in FIGO I-IIA. Furthermore, chemotherapy regimen was identified to be relevant to survival outcomes; EP regimen demonstrated obvious prolonged PFS and OS than those treated with non-EP regimen.
SCNEC is one of the most lethal gynecological malignancies that characterized with high mitotic rate, extensive necrosis, frequent lymph-vascular space involvement (LVSI) and strong association with HPV 18 [17-19]. It is highly aggressive with extremely high risk of local and distant failure, even for patients at early stage[14, 19]. In our series, treatment failure was identified in approximately 60% of patients, and the majority presented with hematogenous dissemination to distant organs, including liver, lung, bone marrow and multiple sites; median PFS and OS for the entire cohort were 30.8 (95% CI 24.4-37.2) months and 53.5 (95% CI, 25.2-81.8) months, respectively, which was consistent with previous reports [8, 12, 14]. Therefore, a major challenge in improving the prognosis of SCNEC is to adopt a comprehensive treatment approach to reduce the risk of disease failure. Nonetheless, the optimal treatment strategies, especially the most appropriate chemotherapy regimen and course for patients with SCNEC is under determined as yet [20]. We hence carried out this study to identify the association between chemotherapy intensity and survival outcomes, and aimed to establish a risk-based systemic treatment recommendation for patients with SCNEC.
Foremost, this study represents a critical step toward understanding the cycle-dependent effect of chemotherapy on survival outcomes in SCNEC. The necessity of chemotherapy for improving survival is well recognized for SCNEC on account of the very high risk of hematogenous dissemination[21]. Nonetheless, the optimal chemotherapy regimen and course is under determined owing to the rarity of this disease. In SCLC, a regimen of 4-6 cycles of EP is most commonly recommended because of its superiority in both efficacy and toxicity. Similarly, this regimen was also empirically proposed for SCNEC regardless of tumor extent because of its clinical biological similarity with SCLC. Whereas, there were only a small amount of data focus on this topic and the evidence proving the validity and safety of this regimen was limited. This raises questions whether EP regimen chemotherapy still works in SCNEC, and what’s the optimal cycle that possess both satisfied effectiveness and acceptable toxicity. Research into these topics is essential to optimize treatment strategies and individualize treatment plans for SCNEC. Generally, the prognosis of SCNEC is associated with tumor burden, namely FIGO stage. Patients with early staged tumor (FIGO I-IIA) usually demonstrated superior prognosis than those with advanced tumor (FIGO IIB and above). Hence it is reasonable to hypothesize that treatment may be streamlined according to FIGO stage, namely systemic de-intensification for early staged SCNEC, whereas intensification for advanced staged SCNEC to reduce the risk of distant metastasis. In the current study, we observed inverse correlation between chemotherapy cycles with the risks (HR) of disease progression and death; ≥5 cycles was associated with significantly reduced risks of PFS and OS. Nonetheless, not all patients benefited from ≥5 cycles of chemotherapy. Long course chemotherapy with ≥5 cycles mainly showed survival benefit in patients with FIGO stage IIB-IIIC, while a short course of 1-4 cycles was adequate for early staged SCNEC. Hence, we can conclude that chemotherapy was essential for patients with SCNEC, and the intensity of chemotherapy could be modified according to the tumor burden. We proposed a maximum of 4 cycles for those with FIGO stage I-IIA, whereas ≥5 cycles for those with FIGO stage IIB-IIIC was essential.
Additionally, our results support the use of EP chemotherapy regimen as preferred regimen for SCNEC. In our cohort, EP regimen tended to achieve prolonged PFS and OS than non-EP regimen, although the significances were not obtained due to the limited sample size. The combination of cisplatin and etoposide was initially developed in patients with previously treated SCLC[22], and soon became the most commonly used regimen for patients with SCLC due to the strong evidence showing promising results [23-24]. Additionally, EP regimen was whereafter identified as the first-line chemotherapy for small cell neuroendocrine tumors of other sites such as esophagus[25], ileum [26], and bladder [27] due to the akin natural history to SCLC and their propensity for distant spread. Likewise, EP regimen was also recommended for SCNEC [14, 21] considering evidences showing the superiority of EP regimen over other regimens with better outcome and lower toxicity [28-29]. Chang and colleagues used to report significant survival benefit for patients who treated by vincristine, adriamycin, and cyclophosphamide alternating with EP (VAC/PE) regimen compared to those treated with cisplatin, vinblastine, and bleomycin (PVB) combination (5-year survival VAC/PEvs PVB, 68% vs 33%, P =0.0078)[30]. Similarly, Zivanovic et al also reported that the patients who received postoperative EP regimen chemotherapy had a significant higher 3-year recurrence-free survival than those who did not receive adjuvant chemotherapy (83% vs 0%) for in early stage SCNEC [8]. Hence, EP regimen was firstly proposed as preferred regimen for SCNEC in the latest version of National Comprehensive Cancer Network guidelines (version 1 2021)[21].
Notably, we did not focus on the sequence of chemotherapy in terms of efficacy, since this was beyond the scope of the current study. Additionally, selection bias was obvious in this cohort because induction chemotherapy (IC) was mainly performed in patients with higher tumor burden, whereas adjuvant chemotherapy (ACT) was undertaken in almost all patients. Hence, it is infeasible for the efficacy comparison of IC and ACT in our cohort. Generally, IC was considered to have better tolerated and could eradicate micrometastases earlier, compared with adjuvant sequencing. Nonetheless, IC can also put off the time of radical treatment, which may result in tumor progress. Currently, the role of IC in treatment SCNEC is controversial, since the very small sample sizes of relevant studies make it difficult to draw definitive conclusions [15, 31]. Chang et al. observed a high complete response rate in 6 out of 7 patients that treated with neoadjuvant VAC/PE before hysterectomy; however, microscopic residual tumor was identified in all cases, and additional three courses of ACT were given after surgery [31]. Additionally, Lee et al. reported no survival benefit among 6 patients that received IC[15]. In our cohort, IC alone was only performed in 4 patients, but we still observed inferior survival outcomes in these patients compared with IC+ACT and ACT (median PFS, IC vs IC+ACTvs IC+ACT: 5.8 vs 31.7 vs 31.4 months; median OS, 12.4 vs 53.5 vs 44.3 months; Figure S3 ). Hence, we believe that neoadjuvant sequencing was not adequate for patients with SCNEC, and ACT was essential for the treatment of SCNEC. Other caveats of our study included inherent biases considering the retrospective nature of this study and the patient recruitment from a single institution. Nonetheless, because of the rarity of the disease, it is infeasible to conduct a prospective randomized study in SCNEC. Hence, a solution was to include a consecutive, well-characterized cohort that encompassing early to advanced stages of disease to improve the reliability of our conclusion. Additionally, the conclusions of our study need further validation in external cohort.
In conclusion, our results implied that chemotherapy cycles had inverse correlation with risks of disease progression and death; long-course chemotherapy of ≥5 cycles significantly reduced the risks of PFS and OS in patients with FIGO stage IIB-IIIC, while a short course of 1-4 cycles was enough for early staged SCNEC. We proposed a maximum of 4 cycles for those with FIGO stage I-IIA, but ≥5 cycles for those with FIGO stage IIB-IIIC. Although this study was retrospective in design, with a limited number of patients, it is still one of the largest series reported to date. We hope that our experience contributes to the foundation of knowledge regarding this rare but aggressive tumor.