RESULTS
A total of 20,228 women diagnosed with endometrial cancers between 2009-2017 were identified (Figure 1). Demographic, clinical and surgical data for patients stratified by a summary marginalization score are presented in Table 1. The highest quintile (Q5) represents patients living in the most marginalized neighborhoods.
Age distribution was slightly skewed, with more patients over seventy in the highest marginalization quintile. Highly marginalized patients were likely to be urban residents. Highly marginalized patients had more comorbidities as reflected by higher Charlson scores (20% versus 27% had a score of 0 in the most and least marginalized quintiles, p<0.001), but severe obesity (BMI>40) was evenly distributed across quintiles.
Of patients with known cancer stage (n=14318), 74% (10,508) were diagnosed with stage I disease. 73% (14,810/20,228) of endometrial cancers were of endometrioid histology. Patients in the most marginalized neighborhoods were more likely to present at more advanced stages: of patients with known stage, 70% (1506/2138) of patients in quintile 5 presented with stage I disease, as compared to 76% (2281/3023) of patients in quintile 1 (p<0.001). Highly marginalized patients were also less likely to be treated surgically for their disease: 11% (332/3111) of patients in the highest marginalization quintile did not receive surgery, as compared to 7% (299/4107) of patients in the lowest quintile (p<0.001).
Figure 2 shows Kaplan-Meier survival curves of patients from diagnosis of endometrial cancer, by summary marginalization quintile. Marginalized patients had significantly worse overall survival (log-rank test, p<0.001). 5-year survival was 77% (95% CI, 75-79) among the most highly marginalized patients (quintile 5) as compared to 83% (95% CI, 81-84) among the least marginalized patients (quintile 1), p<0.001.
Regression analyses of patient- and disease-dependent factors associated with overall survival are shown in Table 2. On univariable analysis, income quintile, as well as two of three marginalization domains assessed (material deprivation and residential instability), were significantly associated with the risk of death (p<0.001 for all), as was the summary marginalization score. On multivariable analysis, after adjustment for year of diagnosis, age, Charlson score, obesity, prior cancer diagnosis, disease histology and stage, marginalization remained a significant independent predictor of survival, with a hazard ratio of 1.05 per quintile for death (95% CI, 1.03-1.08, p<0.001). This translates into a hazard ratio of 1.22 when comparing the highest and lowest marginalization quintiles. Individual marginalization indices, including material deprivation (Q5 vs Q1: HR=1.32, 95% CI, 1.19-1.46) and residential instability (Q5 vs Q1: HR=1.22, 95% CI, 1.10-1.35) were also significantly associated with death after adjustment for year of diagnosis, age, Charlson score, obesity, prior cancer diagnosis, disease histology and stage (Table 3). The ethnic concentration index was not found to be prognostic for the risk of death either on univariable analysis or on multivariable analysis.