br Results br On bivariate analysis
On bivariate analysis, significant differences were noted across BMI categories in age at diagnosis, histology, stage, race, grade, hyperten-sion, hyperlipidemia, diabetes, and treatment modalities (Table 1). Overall, there was a significant inverse linear association of age at diag-nosis with BMI (p b 0.001, R2 = 0.04) (Fig. 1). As expected, women with endometrioid histology had a younger age at diagnosis than women with non-endometrioid histologies. The inverse linear association was observed in both subgroups (p b 0.001, Fig. 1). Significant differences (p b 0.05) associated with higher BMI in the endometrioid subgroup included uterine confined disease at diagnosis, grade 1 tumors, non-white race, diagnosis of hypertension, and no further adjuvant treat-ment (Supplementary Table 1). Stage was not associated with BMI in women with non-endometrioid histology.
In the cohort of evaluable patients for this AZD-2281 analysis, 543 women had died overall, 393 (22.7%) from disease specific causes. Median follow up was 3.4 years [Range: 1 day–17.8 years] for the entire cohort, 4.3 years [Range: 1 day–17.8 years] for endometrioid cancers, and 2.1 years [Range: 10 days–17.4 years] for non-endometrioid cancers. There was a significant difference in disease specific and overall survival curves and trend across BMI categories (log rank test p b 0.001, test of trend p b 0.001) (Fig. 2) for the entire cohort. In univariable analysis, younger age at diagnosis, higher BMI, endometrioid histological subtype, lower stage, white race, lower grade, hyperlipidemia, statin use, metformin use, beta blocker use, and adjuvant therapy were associated with decreased hazard of disease specific death (p b 0.01) (Table 2). In
Demographic, histopathologic and treatment variables by BMI category of women treated for endometrial cancer.
ANOVA, Pearson's Chi-squared, or Fisher Exact as appropriate.
multivariable analysis, younger age, lower stage, lower grade, and use of statins and metformin were independently associated with decreased hazard of disease-specific death (p b 0.05). Hyperlipidemia and statin use were deemed to be collinear (ρ = 0.91, variance inflation factor (VIF) N 5) and hyperlipidemia was subsequently excluded from the cho-sen model in order to retain possible associations of medication use. After controlling for other factors, BMI overall was not associated with disease-specific death (Wald p = 0.10). However, in multivariable anal-ysis comparing women with Class 2 obesity to normal weight women there was a decreased hazard of death for women with Class 2 obesity (HR 0.65; 95% CI 0.2–0.97, p = 0.04).
In the subgroup of women with endometrioid histology there remained a difference in overall and disease specific survival curves on Kaplan Meier analysis (log rank p = 0.03, Supplementary Fig. 1). In univariable analysis of this subgroup, age, BMI, stage, grade, hyperlipid-emia, and use of beta blockers, were associated with improved disease-specific survival (p b 0.05). Diabetes and need of any adjuvant therapy (chemotherapy, radiation therapy or both) were associated with in-creased hazard of death for patients with endometrioid histology by univariable analysis. In multivariable analysis, independent predictors of improved disease-specific survival for patients with endometrioid cancers included younger age at diagnosis, earlier stage disease, low
Fig. 1. Linear association of younger age at diagnosis with increased BMI. (A. entire cohort; B. endometrioid histology; C. non-endometrioid histology). Bars reflect 95% confidence intervals.
Fig. 2. Kaplan–Meier overall survival estimates by body mass index for women with endometrial cancer.
Multivariable Cox Regression Modeling of hazard for disease specific death for women with endometrial cancer.
Variable Univariable hazard ratio Multivariable hazard ratio3
grade, and hyperlipidemia (p b 0.05) (Table 3). Any additional treat-ment (chemotherapy, radiation therapy, or both) was associated with increased hazard of death in multivariable analysis of women with endometrioid cancers. BMI was not significantly associated with DSS after controlling for the aforementioned variables (Wald p = 0.15).
In the subgroup of women with non-endometrioid histology, sur-vival curves were significantly different between BMI categories as women with higher BMI having improved survival (log rank p b 0.003, Supplementary Fig. 2). In this cohort, there were significant differences in age, race, hypertension, and diabetes associated with obesity catego-ries (p b 0.05). Unlike the endometrioid patients, stage, grade, metfor-min use and use of adjuvant therapy were not associated with obesity. In univariable Cox regression analysis, age at diagnosis, BMI, stage, hy-pertension, diabetes, hyperlipidemia, adjuvant therapy, and metformin and statin use were associated with disease-specific survival (p b 0.05). In multivariable Cox regression analysis of women with non-endometrioid cancers, younger age at diagnosis, lower stage, treatment utilizing both chemotherapy and radiation, and statin use remained in-dependently associated with decreased hazard of death (p b 0.05). BMI was not independently associated with survival after controlling for these factors (Wald p = 0.11). However, in comparison of patients with non-endometrioid cancers and Class 2 obesity to normal weighted patients with Class 2 obesity there was an improved survival for the obese group (HR 0.59; 95% CI 0.37–0.94, p = 0.03).