Comparison of spread patterns and survival outcomes in uterine carcinosarcoma versus grade 3 endometrioid endometrial cancer Authors: Burcu Soyak, Nejat Ozgul Published: 2026 ABSTRACT PurposeTo compare the clinicopathologic characteristics, dissemination patterns, and survival outcomes of uterine carcinosarcoma (CS) and grade 3 endometrioid adenocarcinoma (G3EAC), focusing on nodal distribution and the impact of complete surgical staging.MethodsThis retrospective cohort study included 105 patients (43 CS, 62 G3EAC) who underwent primary surgical treatment at a single tertiary center between 2000 and 2016. A focused analysis was performed on a high-quality subgroup of 80 patients (76.2%) who underwent systematic pelvic and para-aortic lymphadenectomy. Progression-free survival (PFS) and overall survival (OS) were estimated using Kaplan-Meier methods, and independent prognostic factors were evaluated via Cox proportional hazards regression models.ResultsCS patients were significantly older than G3EAC patients (p = 0.005). While CS showed lower rates of deep myometrial invasion (44.2% vs. 67.7%; p = 0.016), it was associated with significantly higher positive peritoneal cytology (30.8% vs. 7.0%; p = 0.002). Overall lymph node metastasis rates were comparable between the groups (35.9% vs. 24.6%; p = 0.230). CS was associated with a significantly lower 5-year PFS (35.6% vs. 63.4%; p = 0.042), particularly in Stage IA disease (median PFS 7.7 months), and a trend toward lower 5-year OS (41.1% vs. 64.0%; p = 0.088). Multivariable analysis identified advanced age as the sole independent predictor of poor prognosis (OS HR 1.043, p = 0.021; PFS HR 1.039, p = 0.032). Notably, combined chemoradiotherapy (CRT) effectively eliminated the survival gap between histologies (5-year OS: 72.7% for CS vs. 70.0% for G3EAC; p = 0.590).ConclusionUterine carcinosarcoma exhibits minimal local invasion but high systemic aggressiveness. Traditional markers of local spread, such as myometrial invasion, are less predictive for CS. Systematic surgical staging remains essential, and multimodal adjuvant CRT is crucial to mitigating the inherent survival disadvantage of CS.Supplementary InformationThe online version contains supplementary material available at 10.1007/s00432-026-06523-8. Introduction Uterine carcinosarcoma (CS), malignant mixed Müllerian tumor, was traditionally classified as a subtype of uterine sarcoma. It is a biphasic neoplasm composed of malignant-appearing epithelial and mesenchymal elements (D’Angelo and Prat 2010). However, accumulating evidence suggests that CS is more closely related to carcinoma than to true sarcoma. According to current data, CS represents a metaplastic carcinoma arising from a single malignant clone (McCluggage 2002a). Histologic and molecular studies have consistently demonstrated that carcinosarcomas are monoclonal; the sarcomatous component develops from the carcinomatous element, which is the determining factor in tumor behavior. Metastases of carcinosarcomas mostly contain the carcinomatous component, a finding proven by rigorous histopathological examination (Sreenan and Hart 1995). Consequently, adjuvant treatment strategies for CS should be similar to those used for high-grade endometrial carcinomas, rather than being sarcoma-based. High-grade (grade 3) endometrioid endometrial adenocarcinomas (G3EAC) and CS were compared in the present study with respect to patterns of spread and prognosis. Further biological insights highlight the dominance of the epithelial component. Investigations using vascular endothelial growth factor (VEGF) immunostaining and CD34-based assessment of microvessel density (MVD) demonstrated significantly higher VEGF expression and MVD in epithelial compared with mesenchymal elements (Emoto et al. 1999). Tumors with lymphovascular space invasion (LVSI) exhibited higher VEGF expression, suggesting that angiogenesis is primarily driven by the carcinomatous component. Proliferation studies using MIB-1 (Ki-67) also demonstrated higher mitotic and proliferation indices in the carcinomatous component (Nicòtina et al. 1997). Additionally, matrix metalloprotease-7 expression was significantly stronger in epithelial elements, supporting a key role of the carcinomatous component in invasion and tumor progression (Tanimoto et al. 2000). Molecular analyses of X-chromosome inactivation patterns and loss of heterozygosity have further confirmed the monoclonal histogenesis of CS (Kounelis et al. 1998). Identical p53 and KRAS mutations identified in both tumor components prove their monoclonal nature (Abeln et al. 1997). These biological findings have shifted adjuvant treatment strategies toward regimens used for high-grade endometrial carcinomas, incorporating systemic chemotherapy (CTx) and, where indicated, radiation therapy (RT) to improve local control (Rijswijk et al. 1994; Omura et al. 1983). More recently, integrated molecular analyses from large-scale genomic studies have demonstrated that CS clusters molecularly with high-grade endometrial carcinomas rather than uterine sarcomas (Cherniack et al. 2017; Cancer Genome Atlas Research Network et al. 2013). CS is surgically staged according to the FIGO system, similar to endometrial carcinoma, with disease stage being the most important prognostic factor. Complete surgical staging includes total hysterectomy, bilateral salpingo-oophorectomy (TAH-BSO), and systematic pelvic and para-aortic lymph node dissection. Lymphadenectomy is recommended as it provides accurate staging and may be associated with improved survival (Nemani et al. 2008; Park et al. 2010). Given the historical evolution of treatment paradigms, the objective of the study was to compare CS and G3EAC in a cohort where comprehensive surgical staging was prioritized, evaluating clinical behavior and prognosis within the framework of these biological insights. Beyond comparing clinical characteristics, our study specifically aims to evaluate whether modern adjuvant strategies, particularly combined chemoradiotherapy (CRT), can bridge the survival gap between these two aggressive histologies—a phenomenon we describe as the ‘equalizing effect’. Materials and methods Study design and patients This retrospective cohort study included 105 patients (43 CS and 62 G3EAC) who underwent primary surgical treatment at a single tertiary university hospital between December 2000 and November 2016. All procedures were performed by gynecologic oncologists. The study protocol was approved by the Institutional Ethics Committee (No: GO 17/792 − 37). Patients were staged according to the 2009 FIGO classification system. Follow-up data were collected through electronic medical records and telephone interviews until July 2018 or death. The initial diagnosis was based on the final pathology reports from our institution’s specialized gynecological pathologists. While a pathology re-review was not performed at the time of this analysis, all cases underwent institutional pathology board review during the initial diagnostic process. Progression-free survival (PFS) was defined as the interval from the date of primary surgery to the date of first recurrence or death from any cause. Any of these were considered an event. Overall survival (OS) was defined as the interval from the date of primary surgery to the date of death from any cause. Recurrence was confirmed via radiological imaging (computed tomography [CT], magnetic resonance imaging [MRI], or PET-CT) or pathological validation through biopsy or surgical specimens. Surgical procedures All surgical procedures were performed via open laparotomy. Complete surgical staging—defined as TAH-BSO and systematic pelvic and para-aortic lymph node dissection—was performed in 34 (79%) CS and 46 (74%) G3EAC patients. The remaining patients underwent TAH-BSO with pelvic lymphadenectomy or TAH-BSO alone based on clinical judgement and patient comorbidities. Cytoreductive surgery was required in only two patients (n = 2) due to extensive intraoperative disease. Statistical analysis Continuous variables were analyzed using Student’s t-test or Mann–Whitney U test, based on the assessment of normality. Categorical data were compared using chi-square or Fisher’s exact tests. Survival outcomes, including five-year OS and PFS, were estimated using the Kaplan–Meier method and compared via the log-rank test. Median follow-up time was calculated using the reverse Kaplan–Meier method. Univariable and multivariable Cox proportional hazards regression models were utilized to identify independent prognostic factors, including age, FIGO stage, histology, and peritoneal cytology. From an initial cohort of 105 patients, a complete case analysis was performed for the multivariable models, resulting in a final analyzed cohort of n = 94. Eleven cases were excluded due to missing data in peritoneal cytology (n = 9) or other clinical covariates (n = 2). Variables were selected for the multivariable models based on a twofold approach: univariable significance (p < 0.05 for age and peritoneal cytology) and established clinical relevance in the literature (histology, FIGO stage) regardless of its univariable p-value. To ensure the stability of the regression models and prevent overfitting, the ratio of events per variable (EPV) was maintained above the recommended threshold of 10. The multivariable model for PFS included 43 events, providing an EPV ratio of 10.75 (Omnibus test, p = 0.012). OS model included 41 events, yielding an EPV ratio of 10.25 (Omnibus test, p = 0.018). The proportional hazards assumption was verified by both visual inspection of log-log plots and the use of time-dependent covariates, which confirmed that the assumption was met (p = 0.462 for PFS and p = 0.266 for OS). Given its clinical relevance, the impact of adjuvant RT was evaluated in a separate sensitivity analysis. An interaction term (histology × RT) was tested to assess whether the effect of adjuvant treatment differed between subtypes. Hazard ratios (HRs) and 95% confidence intervals (CIs) were reported. All tests were two-sided, with statistical significance set at p < 0.05. Results Baseline clinicopathologic characteristics A total of 105 patients were included, 43 (41%) had CS and 62 (59%) had G3EAC. CS patients were significantly older (mean age: 67.4 ± 8.8 vs. 62.1 ± 9.8 years; p = 0.005), with a mean age difference of 5.38 years (95% CI 1.67–9.08). While G3EAC was associated with more frequent deep myometrial (≥ 50%) invasion (67.7% vs. 44.2%, p = 0.016), positive peritoneal cytology was substantially more common in the CS group (30.8% vs. 7.0%, p = 0.002). Median tumor size (6.0 cm vs. 5.0 cm; p = 0.051) and LVSI prevalence (60.9% vs. 68.5%; p = 0.516) were comparable (Table 1). The distribution of clinical and pathological features for the entire cohort is presented in Table 1. Discrepancies in the total number of patients for certain variables, such as peritoneal cytology (n = 9 missing) and LVSI (n = 28 missing), are due to incomplete data in the retrospective records; for peritoneal cytology, occasional procedural omissions or logistical challenges over the 16-year retrospective period. Table 1Baseline clinicopathologic characteristics of the overall cohortCharacteristicCS (n = 43)G3EAC (n = 62)p valueAge, years (mean ± SD)67.4 ± 8.862.1 ± 9.8 0.005 Tumor size, cm (median, IQR)6.0 (4.0–9.0)5.0 (4.0–6.0)0.051Deep myometrial invasion (≥ 50%)19/43 (44.2%)42/62 (67.7%) 0.016 LVSI*14/23 (60.9%)37/54 (68.5%)0.516Cervical stromal invasion*9/43 (20.9%)18/62 (29.0%)0.350Adnexal involvement*8/43 (18.6%)8/62 (12.9%)0.424Omental involvement*6/43 (14.0%)5/61 (8.2%)0.356Positive peritoneal cytology*12/39 (30.8%)4/57 (7.0%) 0.002 FIGO stage 0.335Stage I–II22 (51.2%)39 (62.9%)Stage III–IV21 (48.8%)23 (37.1%)CS, uterine carcinosarcoma; FIGO, International Federation of Gynecology and Obstetrics; G3EAC, grade 3 endometrioid adenocarcinoma; IQR, interquartile range; LVSI, lymphovascular space invasion; SD, standard deviation*Denominators for variables vary due to missing data or limited assessment. Bold values indicate statistical significance (p < 0.05). p-values were calculated using Student’s t-test for age, Mann–Whitney U test for tumor size, and chi-square or Fisher’s exact tests for categorical variables. Categorical variables are presented as ratio and percentage Table 1Baseline clinicopathologic characteristics of the overall cohortCharacteristicCS (n = 43)G3EAC (n = 62)p valueAge, years (mean ± SD)67.4 ± 8.862.1 ± 9.8 0.005 Tumor size, cm (median, IQR)6.0 (4.0–9.0)5.0 (4.0–6.0)0.051Deep myometrial invasion (≥ 50%)19/43 (44.2%)42/62 (67.7%) 0.016 LVSI*14/23 (60.9%)37/54 (68.5%)0.516Cervical stromal invasion*9/43 (20.9%)18/62 (29.0%)0.350Adnexal involvement*8/43 (18.6%)8/62 (12.9%)0.424Omental involvement*6/43 (14.0%)5/61 (8.2%)0.356Positive peritoneal cytology*12/39 (30.8%)4/57 (7.0%) 0.002 FIGO stage 0.335Stage I–II22 (51.2%)39 (62.9%)Stage III–IV21 (48.8%)23 (37.1%)CS, uterine carcinosarcoma; FIGO, International Federation of Gynecology and Obstetrics; G3EAC, grade 3 endometrioid adenocarcinoma; IQR, interquartile range; LVSI, lymphovascular space invasion; SD, standard deviation*Denominators for variables vary due to missing data or limited assessment. Bold values indicate statistical significance (p < 0.05). p-values were calculated using Student’s t-test for age, Mann–Whitney U test for tumor size, and chi-square or Fisher’s exact tests for categorical variables. Categorical variables are presented as ratio and percentage Baseline clinicopathologic characteristics of the overall cohort CS, uterine carcinosarcoma; FIGO, International Federation of Gynecology and Obstetrics; G3EAC, grade 3 endometrioid adenocarcinoma; IQR, interquartile range; LVSI, lymphovascular space invasion; SD, standard deviation *Denominators for variables vary due to missing data or limited assessment. Bold values indicate statistical significance (p < 0.05). p-values were calculated using Student’s t-test for age, Mann–Whitney U test for tumor size, and chi-square or Fisher’s exact tests for categorical variables. Categorical variables are presented as ratio and percentage Nodal and surgical characteristics A comprehensive nodal assessment was performed in 96 patients (39 CS, 57 G3EAC). Median retrieved lymph nodes were 27 (IQR, 17–40), and 33 (IQR, 17–44), respectively (p = 0.573). Overall lymph node metastasis (LNM) rates were 35.9% for CS and 24.6% for G3EAC (p = 0.230). Pelvic (30.8% vs. 19.3%; p = 0.196) and para-aortic lymph node involvement (23.5% vs. 17.4%; p = 0.497) were similar between groups (Table 2). Although the mean number of metastatic nodes was numerically higher in the CS group (2.49 ± 5.15) than in the G3EAC group (1.19 ± 4.55), this difference did not reach statistical significance (p = 0.083). Table 2Nodal characteristics, surgical outcomes, and survival analysis for the overall cohortCharacteristicCS (n = 43)G3EAC (n = 62)p value Surgical and nodal outcomes Retrieved lymph nodes, median (IQR)27 (17–40)33 (17–44)0.573Overall LNM* (pelvic and/or paraaortic)14/39 (35.9%)14/57 (24.6%)0.230Pelvic LNM*12/39 (30.8%)11/57 (19.3%)0.196Para-aortic LNM*8/34 (23.5%)8/46 (17.4%)0.497 Clinical outcomes and survival Recurrence13/43 (30.2%)9/62 (14.5%)0.052Median PFS, months (95% CI)19.4 (7.7–31.1)Not reached 0.042 5-year PFS, %35.663.4Median OS, months (95% CI)32.9 (7.1–58.6)106.1 (92.7–119.6)0.0885-year OS, %41.164.0CI, confidence interval; CS, uterine carcinosarcoma; G3EAC, grade 3 endometrioid adenocarcinoma; IQR, interquartile range; LNM, lymph node metastasis; OS, overall survival; PFS, progression-free survival*Denominators for nodal variables vary due to incomplete or omitted lymph node assessment in some patients. Bold values indicate statistical significance (p < 0.05) Table 2Nodal characteristics, surgical outcomes, and survival analysis for the overall cohortCharacteristicCS (n = 43)G3EAC (n = 62)p value Surgical and nodal outcomes Retrieved lymph nodes, median (IQR)27 (17–40)33 (17–44)0.573Overall LNM* (pelvic and/or paraaortic)14/39 (35.9%)14/57 (24.6%)0.230Pelvic LNM*12/39 (30.8%)11/57 (19.3%)0.196Para-aortic LNM*8/34 (23.5%)8/46 (17.4%)0.497 Clinical outcomes and survival Recurrence13/43 (30.2%)9/62 (14.5%)0.052Median PFS, months (95% CI)19.4 (7.7–31.1)Not reached 0.042 5-year PFS, %35.663.4Median OS, months (95% CI)32.9 (7.1–58.6)106.1 (92.7–119.6)0.0885-year OS, %41.164.0CI, confidence interval; CS, uterine carcinosarcoma; G3EAC, grade 3 endometrioid adenocarcinoma; IQR, interquartile range; LNM, lymph node metastasis; OS, overall survival; PFS, progression-free survival*Denominators for nodal variables vary due to incomplete or omitted lymph node assessment in some patients. Bold values indicate statistical significance (p < 0.05) Nodal characteristics, surgical outcomes, and survival analysis for the overall cohort CI, confidence interval; CS, uterine carcinosarcoma; G3EAC, grade 3 endometrioid adenocarcinoma; IQR, interquartile range; LNM, lymph node metastasis; OS, overall survival; PFS, progression-free survival *Denominators for nodal variables vary due to incomplete or omitted lymph node assessment in some patients. Bold values indicate statistical significance (p < 0.05) Survival outcomes and recurrence patterns Median follow-up for the entire cohort was 61.4 months (95% CI 55.0–67.8). Recurrence was radiologically confirmed in all patients (100%, n = 22), with additional pathological validation performed in 40.9% (n = 9) of these cases. Recurrence occurred in 30.2% (n = 13) of CS and 14.5% (n = 9) of G3EAC patients (p = 0.052, Table 2). Recurrence patterns differed by histology: CS failures were more disseminated, frequently involving distant sites (distant metastasis:7, abdominal:4, pelvic:1, vaginal cuff:1), while G3EAC recurrences were more localized to the vaginal cuff or peritoneum (distant:3, abdominal: 4, pelvic:1, vaginal cuff:1). Five-year PFS was significantly lower in the CS group compared with G3EAC (35.6% vs. 63.4%; p = 0.042; Fig. 1A). Median PFS was 19.4 months (95% CI 7.7–31.1) for CS versus not reached for G3EAC. Similarly, CS showed a clinical survival disadvantage in 5-year OS (41.1% vs. 64.0%; p = 0.088), with a median OS of 32.9 months versus 106.1 months for G3EAC. Although the CS group showed a more aggressive decline in the early follow-up period (Fig. 1B), the difference in 5-year OS did not reach statistical significance. Fig. 1Kaplan–Meier survival curves for the overall cohort by histology. A Progression-free survival (PFS) and B Overall survival (OS) comparing carcinosarcoma and grade 3 EAC. Although the carcinosarcoma group showed a more aggressive decline in the early follow-up period, the difference in 5-year OS did not reach statistical significance. Vertical ticks indicate censored observations. Number at risk is provided below each graph. Abbr: EAC, endometrioid adenocarcinoma. Kaplan–Meier survival curves for the overall cohort by histology. A Progression-free survival (PFS) and B Overall survival (OS) comparing carcinosarcoma and grade 3 EAC. Although the carcinosarcoma group showed a more aggressive decline in the early follow-up period, the difference in 5-year OS did not reach statistical significance. Vertical ticks indicate censored observations. Number at risk is provided below each graph. Abbr: EAC, endometrioid adenocarcinoma. Survival analysis by FIGO stage Early-stage disease (stage IA and IB) In Stage IA disease (n = 21), CS exhibited a markedly aggressive course despite being uterine-confined. Five-year OS and PFS rates for CS were both 28.6%, compared with 60.0% and 63.0% for G3EAC (OS p = 0.095; PFS p = 0.077). Median PFS for Stage IA CS was only 7.7 months (95% CI 5.6–9.8) and median OS was 17.0 months (95% CI 3.9–30.0), whereas medians were not reached for G3EAC. For Stage IB disease (n = 31), 5-year OS was 50.0% for CS versus 72.1% for G3EAC (p = 0.174), and 5-year PFS was 40.0% and 73.2%, respectively (p = 0.122). Median PFS and OS for CS were 17.4 and 32.8 months, while neither was reached for G3EAC. Advanced-stage disease (stage III and IV) In Stage III disease (n = 26), the survival gap was less pronounced; 5-year OS was 54.3% for CS and 77.1% for G3EAC (p = 0.607), while 5-year PFS rates were 50.6% and 68.6%, respectively (p = 0.668). Both groups demonstrated poor outcomes in Stage IV (n = 17). Median OS was 11.1 months (95% CI 7.7–14.4) for CS and 17.9 months (95% CI 8.9–26.9) for G3EAC (p = 0.844). Median PFS was 7.0 months (95% CI 0.0–19.6) for CS and 8.7 months (95% CI 7.5–9.8) for G3EAC (p = 0.614). Impact of adjuvant treatment modalities on survival outcomes Survival outcomes across the cohort were analyzed by adjuvant strategy: CRT, CTx only, RT only, and no adjuvant treatment. Chemoradiotherapy (CRT) group In patients receiving CRT, survival outcomes were comparable between the two histological subtypes, suggesting that intensified multimodal treatment may mitigate the aggressive biological nature of CS. The 5-year OS rate was 72.7% for CS versus 70.0% for G3EAC (p = 0.590), with median OS of 87.6 and 106.1 months (95% CI 13.2–199.0), respectively. The 5-year PFS rates were also similar (63.6% vs. 60.0%; p = 0.502). Notably, median PFS was 70.8 months (95% CI 0.0–169.6) for G3EAC and was not reached for CS, suggesting sustained remission in CS patients treated with CRT. Chemotherapy (CTx) only group Among patients treated with CTx alone, CS showed a trend toward inferior outcomes, though this did not reach statistical significance. The 5-year OS was 22.2% for CS versus 60.0% for G3EAC (p = 0.493). The median OS for CS was 32.9 months (95% CI 0.0–68.4), while the median OS for the G3EAC group was not reached. Similarly, 5-year PFS was 19.0% for CS versus 62.5% for G3EAC (p = 0.370), with a median PFS of 19.9 months (95% CI 11.8–28.0) for CS, not reached for G3EAC. Radiotherapy (RT) only group RT as a single modality was associated with a survival disadvantage in this sub-group analysis for CS compared with G3EAC, highlighting the systemic nature of CS spread. The 5-year OS was significantly lower for CS at 33.3% versus 75.8% for G3EAC (p = 0.009; median OS: 14.3 months [95% CI 4.8–23.7] vs. not reached). Similarly, 5-year PFS was 25.0% for CS and 77.9% for G3EAC (p = 0.003; median PFS: 7.3 months [95% CI 0.0–19.7] vs. not reached). No adjuvant treatment group Both groups demonstrated poor survival in the absence of adjuvant treatment. The 5-year OS was 30.0% for CS and 31.1% for G3EAC (p = 0.743), with median OS of 9.8 months (95% CI 4.8–14.9) and 32.5 months (95% CI 0.0–88.1), respectively. Five-year PFS rates were 30.0% for CS and 31.1% for G3EAC (p = 0.677). Multivariable and sensitivity survival analyses Cox proportional hazards regression models were used to identify independent prognostic factors for OS and PFS (Table 3). In univariable analyses, increasing age and positive peritoneal cytology were significantly associated with both poorer OS and PFS. Histological subtype was significantly associated with PFS in the univariable model (p = 0.045), with CS demonstrating worse outcomes than G3EAC, although its association with OS did not reach statistical significance (p = 0.091). Notably, advanced FIGO stage (III–IV vs. I–II) was not significantly associated with OS or PFS in the univariable models (p > 0.05). To further investigate this finding, a subgroup analysis of treatment modalities was conducted (Supplementary Table S1). This revealed a distinct treatment distribution: 32.2% (20/62) of Stage I–II patients received no adjuvant therapy, whereas 88.4% (38/43) of Stage III–IV patients underwent intensive multimodal treatment. Table 3Univariable and multivariable Cox proportional hazards regression analysis of prognostic factors for progression-free and overall survivalVariableAnalysisOS HR (95% CI)p valuePFS HR (95% CI)p valueAgeUnivariable1.037 (1.005–1.070) 0.023 1.035 (1.004–1.067) 0.026 Multivariable1.043 (1.006–1.080) 0.021 1.039 (1.003–1.076) 0.032 Stage (III–IV vs. I–II)Univariable1.449 (0.801–2.619)0.2201.432 (0.803–2.556)0.224Multivariable1.314 (0.634–2.722)0.4631.244 (0.604–2.563)0.554Histology (CS vs. G3EAC)Univariable1.660 (0.922–2.989)0.0911.800 (1.013–3.199) 0.045 Multivariable1.272 (0.648–2.496)0.4851.444 (0.746–2.795)0.275Peritoneal cytology (positive vs. negative)Univariable2.231 (1.053–4.728) 0.036 2.277 (1.112–4.664) 0.024 Multivariable1.457 (0.587–3.616)0.4181.461 (0.608–3.511)0.396RadiotherapyUnivariable0.399 (0.220–0.722) 0.002 0.430 (0.241–0.767) 0.004 Sensitivity model†0.476 (0.242–0.935) 0.031 0.527 (0.275–1.010)0.054ChemotherapyUnivariable1.173 (0.652–2.110)0.5931.198 (0.674–2.218)0.538Histology × RTInteraction0.976 (0.247–3.860)0.9720.953 (0.257–3.539)0.943CI, confidence interval; CS, uterine carcinosarcoma; G3EAC, grade 3 endometrioid adenocarcinoma; HR, hazard ratio; OS, overall survival; PFS, progression-free survival; RT, radiotherapyReference categories for the models: Stage I–II, G3EAC histology, and negative peritoneal cytology. The multivariable model simultaneously adjusts for age, stage, histology, and peritoneal cytology. Bold values indicate statistical significance (p < 0.05)†Sensitivity model for adjuvant RT was further adjusted for age, stage, histology, and peritoneal cytology to evaluate its independent impact Table 3Univariable and multivariable Cox proportional hazards regression analysis of prognostic factors for progression-free and overall survivalVariableAnalysisOS HR (95% CI)p valuePFS HR (95% CI)p valueAgeUnivariable1.037 (1.005–1.070) 0.023 1.035 (1.004–1.067) 0.026 Multivariable1.043 (1.006–1.080) 0.021 1.039 (1.003–1.076) 0.032 Stage (III–IV vs. I–II)Univariable1.449 (0.801–2.619)0.2201.432 (0.803–2.556)0.224Multivariable1.314 (0.634–2.722)0.4631.244 (0.604–2.563)0.554Histology (CS vs. G3EAC)Univariable1.660 (0.922–2.989)0.0911.800 (1.013–3.199) 0.045 Multivariable1.272 (0.648–2.496)0.4851.444 (0.746–2.795)0.275Peritoneal cytology (positive vs. negative)Univariable2.231 (1.053–4.728) 0.036 2.277 (1.112–4.664) 0.024 Multivariable1.457 (0.587–3.616)0.4181.461 (0.608–3.511)0.396RadiotherapyUnivariable0.399 (0.220–0.722) 0.002 0.430 (0.241–0.767) 0.004 Sensitivity model†0.476 (0.242–0.935) 0.031 0.527 (0.275–1.010)0.054ChemotherapyUnivariable1.173 (0.652–2.110)0.5931.198 (0.674–2.218)0.538Histology × RTInteraction0.976 (0.247–3.860)0.9720.953 (0.257–3.539)0.943CI, confidence interval; CS, uterine carcinosarcoma; G3EAC, grade 3 endometrioid adenocarcinoma; HR, hazard ratio; OS, overall survival; PFS, progression-free survival; RT, radiotherapyReference categories for the models: Stage I–II, G3EAC histology, and negative peritoneal cytology. The multivariable model simultaneously adjusts for age, stage, histology, and peritoneal cytology. Bold values indicate statistical significance (p < 0.05)†Sensitivity model for adjuvant RT was further adjusted for age, stage, histology, and peritoneal cytology to evaluate its independent impact Univariable and multivariable Cox proportional hazards regression analysis of prognostic factors for progression-free and overall survival CI, confidence interval; CS, uterine carcinosarcoma; G3EAC, grade 3 endometrioid adenocarcinoma; HR, hazard ratio; OS, overall survival; PFS, progression-free survival; RT, radiotherapy Reference categories for the models: Stage I–II, G3EAC histology, and negative peritoneal cytology. The multivariable model simultaneously adjusts for age, stage, histology, and peritoneal cytology. Bold values indicate statistical significance (p < 0.05) †Sensitivity model for adjuvant RT was further adjusted for age, stage, histology, and peritoneal cytology to evaluate its independent impact In the multivariable models, which adjusted for age, FIGO stage, histology, and peritoneal cytology, age remained the only independent prognostic factor for both OS and PFS. In these models, each one-year increase in age was associated with a 4.3% increase in the risk of death (HR 1.043; 95% CI 1.006–1.080; p = 0.021) and a 3.9% increase in the risk of disease progression (HR 1.039; 95% CI 1.003–1.076; p = 0.032). The prognostic impact of histology, stage, and cytology was attenuated after adjustment and did not retain statistical significance. Sensitivity analysis of adjuvant radiotherapy Given the therapeutic impact observed in univariable analysis, a sensitivity analysis was conducted by including adjuvant RT in the multivariable model. In this model, adjuvant RT remained significantly associated with improved OS (HR 0.476; 95% CI 0.242–0.935; p = 0.031), while its association with PFS approached statistical significance (HR 0.527; 95% CI 0.275–1.010; p = 0.054). Importantly, no significant interaction was observed between histologic subtype and RT for either OS or PFS (p > 0.05), indicating that the beneficial effect of RT was consistent across both histologies (Table 3). Outcomes of complete surgical staging (n = 80) To eliminate potential confounding from incomplete lymphadenectomy and stage migration, a focused analysis was conducted on the subgroup that underwent systematic pelvic and para-aortic lymphadenectomy (n = 80; 34 CS, 46 G3EAC) (Table 4). Table 4Nodal characteristics and survival outcomes in the completely staged subgroup (n = 80)VariableCS (n = 34)G3EAC (n = 46)p value Surgical and nodal outcomes Total lymph nodes removed, median (IQR)31.5 (23.0–51.0)37.5 (26.5–54.5)0.174Overall LNM (pelvic and/or paraaortic)13/34 (41.2%)13/46 (28.3%)0.469Pelvic LNM11/34 (32.4%)10/46 (21.7%)0.286Paraaortic LNM8/34 (23.5%)8/46 (17.4%)0.497Total metastatic lymph nodes, median (range)0 (0–22)0 (0–31)0.095Metastatic pelvic lymph nodes, median (range)0 (0–16)0 (0–24)0.146Metastatic paraaortic lymph nodes, median (range)0 (0–16)0 (0–7)0.191 Clinical outcomes and survival Recurrence6/34 (17.6%)6/46 (13.0%)0.569Median PFS, months (95% CI)32.9 (14.3–51.4)Not reached0.0625-year PFS (%)41.772.5Median OS, months (95% CI)43.2 (0.0–111.2)Not reached0.0925-year OS (%)46.473.6CI, confidence interval; CS, uterine carcinosarcoma; G3EAC, grade 3 endometrioid adenocarcinoma; IQR, interquartile range; LNM, lymph node metastasis; OS, overall survival; PFS, progression-free survivalp-values were calculated using Student’s t-test or Mann-Whitney U test for continuous variables and chi-square or Fisher’s exact tests for categorical variables. Categorical variables are presented as ratio and percentage Table 4Nodal characteristics and survival outcomes in the completely staged subgroup (n = 80)VariableCS (n = 34)G3EAC (n = 46)p value Surgical and nodal outcomes Total lymph nodes removed, median (IQR)31.5 (23.0–51.0)37.5 (26.5–54.5)0.174Overall LNM (pelvic and/or paraaortic)13/34 (41.2%)13/46 (28.3%)0.469Pelvic LNM11/34 (32.4%)10/46 (21.7%)0.286Paraaortic LNM8/34 (23.5%)8/46 (17.4%)0.497Total metastatic lymph nodes, median (range)0 (0–22)0 (0–31)0.095Metastatic pelvic lymph nodes, median (range)0 (0–16)0 (0–24)0.146Metastatic paraaortic lymph nodes, median (range)0 (0–16)0 (0–7)0.191 Clinical outcomes and survival Recurrence6/34 (17.6%)6/46 (13.0%)0.569Median PFS, months (95% CI)32.9 (14.3–51.4)Not reached0.0625-year PFS (%)41.772.5Median OS, months (95% CI)43.2 (0.0–111.2)Not reached0.0925-year OS (%)46.473.6CI, confidence interval; CS, uterine carcinosarcoma; G3EAC, grade 3 endometrioid adenocarcinoma; IQR, interquartile range; LNM, lymph node metastasis; OS, overall survival; PFS, progression-free survivalp-values were calculated using Student’s t-test or Mann-Whitney U test for continuous variables and chi-square or Fisher’s exact tests for categorical variables. Categorical variables are presented as ratio and percentage Nodal characteristics and survival outcomes in the completely staged subgroup (n = 80) CI, confidence interval; CS, uterine carcinosarcoma; G3EAC, grade 3 endometrioid adenocarcinoma; IQR, interquartile range; LNM, lymph node metastasis; OS, overall survival; PFS, progression-free survival p-values were calculated using Student’s t-test or Mann-Whitney U test for continuous variables and chi-square or Fisher’s exact tests for categorical variables. Categorical variables are presented as ratio and percentage Clinicopathologic features and peritoneal dissemination In this high-quality cohort, CS patients were significantly older (mean age 67.6 ± 9.1 vs. 61.2 ± 8.6 years, p = 0.002). Regarding uterine-confined features, deep myometrial invasion (≥ 50%) was identified in 55.9% (19/34) of CS and 71.7% (33/46) of G3EAC patients (p = 0.218). Positive peritoneal cytology was significantly more frequent in the CS group (28.1% vs. 8.8%; p = 0.034). Nodal distribution and metastatic burden As detailed in Table 4, median retrieved lymph nodes were 31.5 for CS and 37.5 for G3EAC (p = 0.174). Despite lower rates of deep invasion, LNM was numerically more frequent in CS (41.2%) than G3EAC (28.3%; p = 0.469). Pelvic LNM occurred in 32.4% of CS and 21.7% of G3EAC (p = 0.286), while para-aortic involvement was detected in 23.5% and 17.4%, respectively (p = 0.497). Survival analysis of the staged cohort Survival outcomes within the completely staged subgroup demonstrated a persistent clinical disadvantage for CS (Table 4). Median PFS for CS was 32.9 months (95% CI 14.3–51.4), while it was not reached for G3EAC (p = 0.062; 5-year PFS rates: 41.7% vs. 72.5%). Median OS for CS was 43.2 months (95% CI 0–111.2), vs. not reached for G3EAC (p = 0.092; 5-year OS rates: 46.4% vs. 73.6%). Although these differences did not reach statistical significance, they reflect the inherent biological aggressiveness of CS even with accurate nodal mapping. Recurrence patterns showed that upper abdominal/peritoneal relapses occurred in 17.6% of CS and 13.0% of G3EAC patients. Discussion The present study provides a comparative analysis of CS and G3EAC, focusing on their distinct pathophysiological behaviors and prognostic trajectories. Our results demonstrate that while these two high-grade malignancies are managed under similar protocols, they exhibit fundamentally different dissemination modes. CS displays a more systemic and aggressive phenotype from the clinical outset, supporting its understanding as a metaplastic carcinoma (McCluggage 2002a, 2002b; Cherniack et al. 2017). Our observations regarding the distinct clinical behavior of CS and G3EAC are further supported by recent 2025 literature. Dagher et al. (2025) similarly emphasized the heterogeneity among high-grade endometrial carcinomas, demonstrating that CS exhibits both higher peritoneal cytology positivity and the poorest survival outcomes when compared to G3EAC. Furthermore, in advanced-stage disease, our results align with Conforti et al. (2025), who demonstrated that histological subtype is a major determinant of survival outcomes even beyond disease stage. Age and the biological paradox of dissemination A primary distinction in our cohort was the significantly higher age of patients in the CS group, consistent with the established literature characterizing CS as a disease of the elderly, typically presenting in the sixth and seventh decades of life (Cherniack et al. 2017). Beyond demographics, the most striking finding was the “surgical paradox”: while G3EAC exhibited a higher incidence of deep myometrial invasion (67.7% vs. 44.2%; p = 0.016), CS demonstrated a more aggressive systemic profile. Despite less frequent deep local penetration, the CS group showed significantly higher positive peritoneal cytology (30.8% vs. 7.0%; p = 0.002) and distant organ metastasis. This suggests that CS possesses an inherent systemic aggressiveness, likely driven by EMT (Chaffer et al. 2016). This cell plasticity allows for early peritoneal seeding and distant failure independent of traditional markers of uterine invasion, such as myometrial depth (Tanimoto et al. 2000). Lymphophilic nature independent of myometrial depth In endometrial cancer, deep myometrial invasion is traditionally established as the primary predictor of LNM. Across our entire cohort, however, the incidence of overall LNM was not significantly different between the two subtypes (35.9% for CS vs. 24.6% for G3EAC, p = 0.230), despite the significantly higher deep myometrial invasion in G3EAC. This indicates that while deep myometrial invasion remains a reliable predictor for G3EAC, it is less predictive for CS. In the CS group, nodal risk remained high even with superficial myometrial invasion. This underscores that CS histology itself drives lymphatic dissemination independent of anatomical depth (Park et al. 2010). Survival disadvantage and the frailty of stage IA disease When evaluating the entire cohort, the clinical impact of this biological aggressiveness is most evident in survival outcomes. Our findings suggest a trend towards a survival disadvantage for CS compared to G3EAC, most notably in PFS (5-year PFS: 34.9% vs. 61.3%; p = 0.016). This gap was particularly notable in Stage IA disease (n = 21), where CS exhibited a markedly aggressive course despite being uterine-confined. For Stage IA CS, the 5-year OS and PFS rates were both 28.6%, compared with 60.0% and 63.0% for G3EAC (OS p = 0.095; PFS p = 0.077). The median PFS for Stage IA CS was only 7.7 months (95% CI 5.6–9.8) and the median OS was 17.0 months (95% CI 3.9–30.0), whereas medians were not reached for G3EAC. These clinical observations in Stage IA CS should be interpreted with caution given the small subgroup size; however, they may be partially explained by the adjuvant treatment distribution in our cohort. Specifically, 66.7% (6/9) of Stage IA CS and 41.7% (5/12) G3EAC patients received no adjuvant therapy during the study period. This lack of adjuvant treatment, combined with the inherent systemic aggressiveness of CS, likely contributed to the rapid disease progression observed even in early-stage disease. These findings suggest that even at the earliest FIGO stage, CS carries a substantially worse prognosis than G3EAC (D’Angelo and Prat 2010; Cherniack et al. 2017), necessitating its management as a systemic disease from diagnosis. However, as our observations regarding Stage IA CS are based on a very small subgroup, these findings should be interpreted with caution and regarded as exploratory only. The equalizing effect of multimodal adjuvant therapy A major finding is that while CS patients receiving only RT had poorer outcomes than their G3EAC counterparts (p = 0.009), this survival gap was effectively eliminated in the combined CRT group (5-year OS: 72.7% for CS vs. 70.0% for G3EAC; p = 0.590). Our exploratory observations on the ‘equalizing effect’ of CRT suggest a potential systemic nature for CS. While single-modality RT provides local control, it fails to address the high risk of distant and peritoneal failure inherent to CS (Galaal et al. 2013; Powell et al. 2010). Therefore, a multimodal approach incorporating both the local benefits of RT and the systemic coverage of CTx is essential to achieve comparable outcomes. Independent prognostic factors: multivariable analysis Our multivariable analysis, identified advanced age as the sole independent predictor of poor prognosis for both OS (HR: 1.043; p = 0.021) and PFS (HR: 1.039; p = 0.032). Interestingly, FIGO stage did not emerge as a significant predictor of survival in our cohort. This finding likely reflects a combination of the intensive multimodal management of advanced-stage patients and the significant proportion of early-stage patients (32.2%) who did not receive adjuvant therapy during the study period. The high recurrence potential of CS and G3EAC even in early stages, when left untreated, appears to have narrowed the prognostic difference between early and advanced-stage disease in our series. Furthermore, in sensitivity analysis adjuvant RT was significantly associated with improved OS (HR: 0.476; p = 0.031), providing a 52.4% reduction in the risk of mortality. This emphasizes that beyond the histology, patient-specific factors—primarily age—and the optimal treatment combinations are the primary determinants of outcome. Insights from the completely staged subgroup (n = 80) To eliminate the confounding effect of stage migration, we performed a focused analysis on the completely staged subgroup (n = 80). In this high-quality surgical cohort, the survival difference between histologies did not reach statistical significance (PFS p = 0.062; OS p = 0.092). However, the 23.5% para-aortic involvement in CS underscores the necessity of systematic staging (Park et al. 2010). Omitting this step could lead to significant understaging, as superficial uterine findings in CS do not preclude a substantial upper lymphatic burden (Nemani et al. 2008). Strengths and limitations We acknowledge that our study is primarily limited by its retrospective design and small subgroup sample sizes, rendering secondary analyses exploratory and hypothesis-generating rather than confirmatory. However, despite these limitations, a major strength is the high quality and uniformity of the surgical management. Specifically, 76.2% of our patients underwent systematic pelvic and para-aortic lymphadenectomy and all patients operated via laparatomy, providing a “clean” cohort for the accurate analysis of nodal distribution and significantly reducing the risk of stage migration. Furthermore, the single-center nature of our study at a tertiary university hospital provided a significant advantage in terms of consistency across surgical techniques, standardized pathological interpretations, and uniform adjuvant treatment strategies, thus minimizing the inter-institutional variability often encountered in multi-center datasets. Additionally, the long median follow-up of 61.4 months (95% CI 55.0–67.8) ensures a robust and reliable assessment of 5-year survival outcomes and long-term recurrence patterns in these aggressive malignancies. Conclusion In conclusion, our study suggests that CS and G3EAC exhibit markedly different local invasion patterns despite a similar risk for lymphatic dissemination. CS follows a more aggressive systemic phenotype characterized by early peritoneal seeding and a high lymphatic burden, even in cases with superficial uterine involvement. While our results are limited by the cohort size and should be considered hypothesis-generating, this “surgical paradox“ highlights that traditional markers of local invasion, such as deep myometrial invasion, are significantly less predictive of spread in CS compared with G3EAC. Importantly, our multivariable analysis identified advanced age as the sole independent predictor of poor prognosis for both OS and PFS in these high-grade malignancies. Furthermore, while CS carries an inherent survival disadvantage, our findings suggest that this gap can be effectively mitigated through a multimodal adjuvant approach. The ‘equalizing effect’ observed in our cohort provides a clinical signal that the survival gap might be mitigated through a multimodal approach. However, these findings require validation in larger, prospective trials. While CS exhibits a general survival disadvantage across the entire cohort (PFS p = 0.016), this disparity narrows in completely staged patients and is effectively eliminated by combined CRT. Therefore, comprehensive surgical staging remains essential, and treatment strategies should be optimized with a focus on multimodal therapies to improve clinical outcomes in this high-risk patient population. Electronic Supplementary Material Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1