Pathological evaluation, staging, treatment and follow-up
Detailed information on the diagnosis, treatment, and surveillance for SCNEC were illustrated in Supplementary Materials (online only ). All patients in each cohort were restaged by two radiation oncologists (FPC and XDH) specialized in gynecological cancer according to the 2018 International Federation of Gynecology and Obstetrics (FIGO) staging system [16], with disagreements resolved by consensus.

Statistical analysis

The primary endpoint was progress-free survival (PFS), which was calculated from the initial diagnostic biopsy to date of first relapse regardless of site or a 25% or more increase in size of any pre-existing lesion and/or death from any cause. Secondary endpoint was overall survival (OS), calculated from initial diagnostic biopsy to date of death from any cause, or date of last follow-up visit. Survival analyses were performed using the Kaplan-meier method and compared by the log-rank test. Categorical variables were classified based on clinical findings, and continuous variables were transformed into categorical variables based on median values of our cohort. Because of the complexity of chemotherapy regimens, we classified the multimodal regimens as etoposide plus platinum (EP) and non-EP. Differences of continuous variables were compared using Mann-Whitney U test, and categorical variables were compared using the chi-square or Fisher exact test. The cox proportional hazards regression was performed to calculate hazard ratios (HR) and 95% confidence interval (CI). Multivariable analyses using cox proportional hazards models were performed to evaluate the potential prognostic factors adjusted for age, histology and FIGO stage. The restricted cubic splines (RCS) that estimated from the full cox regression model was used to evaluate the relationships (in HR) between chemotherapy cycles with PFS and OS. All statistical tests were two-sided, and a P -value of <0.05 was considered significant. Statistical analyses were performed in R version 3.4.4 (http://www.r-project.org/), and SPSS 23.0 software (SPSS Inc, IL). All end points were updated in January 2021.

Results

Patient characteristics and treatment outcomes
The characteristics of the patients are listed in Table 1 . Of the 103 patients, FIGO stage I-IIA was diagnosed in 62 (60.2%) patients, and stage IIB-IIIC in 41 (39.8%) patients. All of the enrolled patients have received chemotherapy; EP regimen was carried out in 66 (64.1%) patients, non-EP alone regimens in 37 (35.9%) patient (Table 1 ).
The breakdown of the sites of relapses of these cohorts is detailed inTable S1 (online only) . Median PFS and OS of the entire cohort were 30.8 (95% confidence interval [CI], 24.4-37.2) months and 53.5 (95% CI, 25.2-81.8) months, respectively.
Cycle-dependent effect of chemotherapy cycles on survivals
To quantify the cycle-dependent effect on survival outcomes, we entered chemotherapy cycles into the RCS fitted cox regression to allow for nonlinear relationships between chemotherapy course and survivals. The models identified that the risks (HR) of disease progression and death decreased as the cycles increased; 5 cycles of chemotherapy was identified as threshold that consistent for risk discretization to PFS and OS (Figure 2A and 2B ). Hence, we classified chemotherapy as short-course (1-4 cycles) and long-course (≥5 cycles) based on the threshold (Table S2 ). Through Kaplan-Meier plots, we observed that patients who received ≥5 cycles of chemotherapy demonstrated significantly superior PFS (≥5 vs 1-4: median PFS, 58.6 monthsvs 25.4 months, P =0.027; Figure 2C ) and prolonged OS (≥5 vs 1-4: median OS, 65.1 months vs 37.7 months, P =0.168; Figure 2D ) than those treated with 1-4 cycles. After adjustment for potential prognostic covariates (age, histology and FIGO stage) in multivariable analysis, chemotherapy cycles were independently significant for PFS (HR 0.52, 95% CI 0.30-0.88;P =0.015; Table S3 ), and OS (0.57, 0.31-1.00; P=0.050; Table S3 ). These findings indicate that treatment with ≥5 cycles of chemotherapy contribute to survival improvement in patients with SCNEC. Given that heterogeneity in terms of tumor extension and prognosis was obvious among the patients, we then performed subgroup analyses to determine the therapeutic effect of chemotherapy cycles within different FIGO stages.