Patient demographics. The study inclusion criteria were met by 2,644 eligible patients (1,055 non-chemotherapy and 1,589 chemotherapy). In this population, only 16.4% were aged <45 years, 56.1% were aged 45-65 years, and 27.5% were aged ≥65 years. A large majority of patients were white (n=2,243; 84.8%). Stage IA accounted for 44.2%, stage IB for 5.7%, stage IC for 38.4%, and stage IIA for 11.7%. Of the 2,644 patients, 59.1% were diagnosed with grade 1, and 40.9% with grade 2. Tumor laterality consisted of right (n=1,112; 42.1%), left (n=1,121; 42.4%) and other/unknown (n=411; 15.5%). There were 44.3% of patients who had tumors <10 cm in size, and 42.3% with tumors >10 cm. In the present study, endometrioid carcinoma accounted for 68.9%, and serous cystadenocarcinoma for 31.1% of cases. Patients characteristics are presented in Table 1.
Comparison of univariate covariates. Before the propensity score matching, patients in the chemotherapy groups tended to be younger than those in the non-chemotherapy groups (≤45 years: 17.2 vs. 15.3%; 45-65 years: 59.5 vs. 51.0%; P<0.001). They were less likely to be in stage IA (34.2 vs. 59.3%; P<0.001), and more likely to be in stage IB (5.8 vs. 5.7%; P<0.001) and stage IC (46.0 vs. 26.8%; P<0.001). Compared with non-chemotherapy groups, grade 1 cases were less among the chemotherapy groups (52.9 vs. 68.3%; P<0.001) and there were more grade 2 cases (47.1 vs. 31.7%; P<0.001). There was a larger number of patients that had serous cystadenocarcinoma in the chemotherapy groups (33.5 vs. 27.4%; P=0.001). To eliminate the heterogeneity and imbalance between groups, the present study performed propensity score matching and a logistic regression analysis. The results demonstrated that the two groups both had an equal number of patients, and all clinical factors were well balanced without significant differences, indicating the potential covariates between groups were greatly decreased (Table 2).
Association between chemotherapy and survival. The present study analyzed the association between chemotherapy and 5-year OS for stages IA-IIA. There were no statistically significant differences between the chemotherapy and non-chemotherapy groups (stage IA: 46.5 vs. 53.7%; P=0.110; stage IB: 49.0 vs. 49.1%; P=0.059; stage IC: 46.1 vs. 48.1%; P=0.750; stage IIA: 39.0 vs. 37.2%; P=0.249). Patients with early-stage disease could not benefit from chemotherapy to prolong their 5-year OS (Table 3 and Fig. 2).
Univariate analysis of clinical factors with survival. The present study performed a univariate analysis of the matched population in order to investigate the prognostic effects of the clinical factors. No significant differences were observed between the chemotherapy and non-chemotherapy groups for 5-year OS (P=0.245). Older age was a risk factor for 5-year OS (P<0.001). Tumors >10 cm in size had lower 5-year OS (P=0.014). Furthermore, 5-year OS of endometrioid carcinoma was higher than serous cystadenocarcinoma (P<0.001; Table 4 and Fig. 3).
Cox proportional hazards model. The cox proportional hazards model is often used in medical research to investigate the association between survival time of patients and one or more predictive variables. The Kaplan-Meier method and log-rank tests describe survival according to one factor under investigation, but to not include the impact of any others. In addition, they are available only when the predictive variables are categorical, and do not easily work for quantitative predictors, such as age. Given these issues, an alternative method is the cox proportional hazards regression analysis, which works for both quantitative predictive variables and categorical variables. Furthermore, the cox regression model extends survival analysis methods to simultaneously assess the effects of several risk factors on survival time.
In order to investigate how clinical factors jointly impact on survival, the present study took all the factors associated with survival into a multivariate cox regression analysis. The results are presented in Table 5. The analysis revealed that elderly patients (age, ≥65 years) had a higher risk and worse prognosis (HR, 1.486; CI, 1.208-1.827; P<0.001). Endometrioid carcinoma was associated with improved 5-year OS (HR, 0.697; CI, 0.584-0.833; P<0.001). Chemotherapy still had no statistically significant effect on the 5-year OS after excluding the influence of all the confounding factors (HR, 1.092; CI, 0.954-1.249; P=0.201).