Associations between etiologic factors and mortality after endometrial cancer diagnosis: The NRG Oncology/Gynecologic Oncology Group 210 Trial

Abstract

Background

Few studies have analyzed relationships between risk factors for endometrial cancer, especially with regard to aggressive (non-endometrioid) histologic subtypes, and prognosis. We examined these relationships in the prospective NRG Oncology/Gynecologic Oncology Group 210 trial.

Methods

Prior to surgery, participants completed a questionnaire assessing risk factors for gynecologic cancers. Pathology data were derived from clinical reports and central review. We used the Fine and Gray subdistribution hazards model to estimate subhazard ratios (HRs) and 95% confidence intervals (CIs) for associations between etiologic factors and cause-specific subhazards in the presence of competing risks. These models were stratified by tumor subtype and adjusted for stage and socioeconomic status indicators.

Results

Median follow-up was 60 months after enrollment (range: 1 day – 118 months). Among 4,609 participants, a total of 854 deaths occurred, of which, 582 deaths were attributed to endometrial carcinoma. Among low-grade endometrioid cases, endometrial carcinoma-specific subhazards were significantly associated with age at diagnosis (HR=1.04, 95% CI=1.01–1.06 per year, P-trend) and BMI (class II obesity vs. normal BMI: HR=2.29, 95% CI=1.06–4.98, P-trend=0.01). Among high-grade endometrioid cases, endometrial carcinoma-specific subhazards were associated with age at diagnosis (HR=1.05, 95% CI=1.02–1.07 per year, P-trend<0.001). Among non-endometrioid cases, endometrial carcinoma-specific subhazards were associated with parity relative to nulliparity among serous (HR=0.55, 95% CI=0.36–0.82) and carcinosarcoma cases (HR=2.01, 95% CI=1.00–4.05).

Discussion

Several endometrial carcinoma risk factors are associated with prognosis, which occurs in a tumor-subtype specific context. If confirmed, these results would suggest that factors beyond histopathologic features and stage are related to prognosis.

INTRODUCTION

Accepted prognostic factors for endometrial carcinoma include tumor characteristics, such as histology, grade, lymphovascular space invasion, and stage (1). Less is known about the role that etiologic risk factors may have on survival following an endometrial carcinoma diagnosis, which could be of use to clinicians to guide treatment decisions and inform surveillance schedules.

High body mass index (BMI), a strong risk factor for endometrial carcinoma development, has inconsistently been associated with prognosis following an endometrial carcinoma diagnosis (2–14). A systematic review (15) found that eight studies reported no association (2–5, 7–10) while four studies demonstrated increased all-cause mortality associated with high BMI (16–19). Seven studies that were not included in this review found equally mixed results: no association for three (11–13) and increased risk of death with higher BMI in four (14, 20–22). Studies reporting an association between higher BMI and increased mortality more frequently controlled for stage, histologic subtype, and adjuvant therapy compared with studies that did not detect relationships, although exceptions are evident. Most studies did not include sufficient numbers of non-endometrioid carcinomas to fully examine relationships among rare aggressive subtypes and differences in categorization of BMI, small sample sizes, and short follow-up times hamper interpretation of the literature.

Few studies have examined relationships between other endometrial carcinoma risk factors and prognosis. Diabetes, a factor related to increased endometrial carcinoma risk, has been associated with increased mortality in some studies (11, 14, 17, 23, 24). Although smoking is inversely related to endometrial carcinoma risk, smokers may develop more advanced-stage tumors and experience increased endometrial carcinoma mortality than non-smokers (25, 26). Relationships between parity and endometrial carcinoma specific survival have been mixed (11, 17, 27–32), with one study reporting increased survival among women with endometrioid carcinomas, but non-significantly decreased survival among women with non-endometrioid tumors (29). Similarly, relationships between menopausal hormone use and survival among endometrial carcinoma patients remain unclear and potentially vary by preparation type (estrogen-only vs. estrogen plus progestin) (33–36).

The clinical behavior of endometrial carcinomas vary substantially by histologic subtype; in particular, non-endometrioid carcinomas portend a worse outcome and certain types such as serous carcinomas spread on peritoneal surfaces, much like their ovarian/tubal primary counterparts (37). Given that the behavior of different endometrial carcinoma subtypes varies, we hypothesize that risk factors that might influence survival for these tumors would also likely vary. Due to the low prevalence of aggressive tumor subtypes in single institution cohorts, we utilized data from the NRG Oncology/Gynecologic Oncology Group (GOG) 210 trial, which included a large number of women with rare but aggressive endometrial tumors, to examine associations between established etiologic risk factors for disease development and prognosis in a large trial of clinically and pathologically well-characterized endometrial carcinoma patients.

METHODS

NRG Oncology/GOG 210 study population

This study population has been previously described (38). The NRG Oncology/GOG 210 molecular staging trial of endometrial carcinoma opened to accrual on September 22, 2003 at 62 U.S. institutions and closed on December 1, 2011. Women with endometrial carcinoma diagnosed by endometrial biopsy or dilation and curettage were approached to participate in this trial. Eligibility required that candidates be suitable for surgery with staging and not have undergone prior retroperitoneal surgery or pelvic/abdominal radiation. Patients with all stages, grades and epithelial histology types (e.g. endometrial carcinoma or carcinosarcoma) were eligible for inclusion in the trial. Prior to hysterectomy and bilateral salpingo-oophorectomy, patients who consented to participate in the trial completed a self-administered questionnaire as previously described (38). On September 23, 2007, eligibility criteria in NRG Oncology/GOG 210 changed from unrestricted enrollment to targeted enrollment of poor prognosis tumor subtypes and cancers occurring among non-obese and non-white patients (38).

Of 6,124 cases enrolled between October 23, 2003 and November 30, 2011, 5,492 (89.7%) completed and returned the questionnaire. We excluded women for the following reasons: incomplete surgical staging (n=20), final diagnosis not endometrial carcinoma (n=53), benign diagnoses (n=6), diagnosis of a second primary (n=2), misclassified pathologic diagnosis based on central pathology review (n=49), inadequate material for pathology review (n=22), protocol deviations (n=17), and improper pre-protocol treatment (n=1). We excluded cases with missing grade (n=23), mixed epithelial cancers (n=556), mucinous cancers (n=18), unusual histologic types (including squamous cell, undifferentiated, and de-differentiated histologies) (n=111), and missing stage (n=5), leaving 4,609 patients for analysis. For this analysis we excluded mixed epithelial cancers as the components of these tumors were not recorded in the 60% of cases with admixtures of endometrioid and adenocarcinoma histology. This study was approved by institutional review boards at the National Cancer Institute and participating study centers.

Risk factor assessment

Prior to surgery, participants self-completed a questionnaire which assessed information on demographic characteristics (age, race, annual income, highest level education attained) and established endometrial carcinoma risk factors including anthropometric measures, reproductive and menstrual characteristics, exogenous hormone use, smoking status, and medical conditions.

Tumor characteristics and outcome assessment

Pathologic information was available from participating NRG Oncology/GOG institutions and through specialized reviews [previously described in (38)] performed by the NRG Oncology/GOG Pathology Committee. Diagnoses of serous, carcinosarcoma, clear cell, grade 3 endometrioid adenocarcinoma, and tumors involving the cervix or with non-nodal metastases were centrally reviewed. Surgical stage and tumor grade were determined post-operatively and coded according to the International Federation for Gynecology and Obstetrics (FIGO) 1988 criteria (39). Final tumor subtype categories included low-grade (grades 1 and 2) endometrioid, high-grade (grade 3) endometrioid, serous, clear cell, and carcinosarcoma. Information on vital status, date, and cause of death was obtained from medical records and supplemented with cancer registry information. Follow-up information was available through December 31, 2013. Time to event (in months) was calculated as the difference between enrollment date and date of death, date last seen, or December 31, 2013, whichever occurred first.

Causes of death were categorized as related to disease-only (n=582), treatment-only (n=11), treatment and disease (n=1), neither treatment nor disease (n=159), cancer treatment from a subsequent trial (n=3), or unknown (n=98). For the purposes of our analysis, we categorized the 582 disease-only deaths as endometrial carcinoma-specific deaths and the remaining 272 deaths as non-endometrial carcinoma deaths.

Statistical Analyses

The competing risks of endometrial carcinoma-specific and non-endometrial carcinoma mortality were modeled using the Fine and Gray model, (40) a semi-proportional model that provides the cumulative incidence of each event while simultaneously considering the competing risk of the other outcome. The Fine and Gray model provides estimates of the cumulative incidence function as subhazard ratios (HRs) and 95% confidence intervals (CIs), analogous to the Cox proportional hazards model. Time since enrollment was used as the underlying time metric and proportional subhazards were assessed by including the risk factors as time varying covariates and evaluating the Wald p-value (41).

We first examined relationships between known prognostic factors, including age, stage, tumor subtype, and adjuvant therapy, with cause-specific subhazards in minimally adjusted models. We then examined relationships between etiologic factors and endometrial carcinoma-specific mortality stratified by tumor subtype. We built a regression model including endometrial carcinoma risk factors that were considered a priori: age at diagnosis, race, parity, BMI, measured in kg/m² (one year before diagnosis), diabetes, smoking status, breast cancer diagnosis and tamoxifen use, oral contraceptive use, and menopausal hormone use. These models were also adjusted for stage, highest education level attained, and annual income. Linear trend tests were performed for BMI, parity, and duration of OC use by including the ordinal form of each categorical variable in the model. As no significant linear trends were observed for parity or duration of OC use, we present dichotomous categories for these exposures. Adjuvant therapy was not included as an adjustment factor in the cause-specific subhazards models as it was not a significant predictor in minimally adjusted models. Missing values for each covariate were modeled as a separate category.

We performed a sensitivity analysis comparing associations between etiologic factors and cause-specific subhazards stratified by enrollment period (i.e. 2003–2007 vs. 2007–2011) for each tumor subtype. We tested for formal interactions by including a multiplicative interaction term between each variable and a binary variable indicating the enrollment period. Statistical analyses were performed using SAS (version 9.3, SAS Institute, Cary, NC, USA) and Stata software (version 11, STATA Corp., Texas, USA). All P values were two-sided; statistical significance was set at P less than 0.05.

RESULTS

Among the 4,609 endometrial carcinoma patients eligible for inclusion in the current analysis, median follow-up was 60 months after diagnosis of endometrial carcinoma (range: 1 day – 118 months). In the multivariable analysis considering established prognostic factors, endometrial carcinoma-specific subhazards were significantly associated with age at diagnosis (HR=1.03, 95% CI=1.02–1.04 per year, P-trend <0.001), stage (HR comparing stage IV to stage I=7.94, 95% CI=5.25–12.00, P-trend <0.001), and tumor subtype (P<0.001) (Table 1). Relative to women diagnosed with low-grade endometrioid tumors, endometrial carcinoma-specific subhazards were increased among women diagnosed with high-grade endometrioid (HR=3.95, 95% CI=2.78–5.62), serous (HR=4.88, 95% CI=3.45–6.90), carcinosarcoma (HR=7.52, 95% CI=5.05–11.20), and clear cell tumors (HR=4.39, 95% CI=2.51–7.68). Adjuvant therapy was not associated with endometrial carcinoma-specific subhazards (P=0.61).

Table 1.

Multivariable subhazard ratios (HRs) and 95% confidence intervals (CIs) for cause-specific mortality by age, tumor characteristics, and adjuvant therapy among 4,609 women with endometrial carcinoma enrolled in the NRG Oncology/GOG 210 trial, 2003–2011

	Endometrial carcinoma-specific mortality		Non-endometrial carcinoma mortality
		deaths=582		deaths=272
	deaths/N	HR (95% CI)1	deaths/N	HR (95% CI)1
Age at diagnosis, per year		1.03 (1.02, 1.04)		1.08 (1.06, 1.09)
p-trend		<0.001		<0.001
Stage
I	184/3344	1.00	170/3344	1.00
II	43/321	1.95 (1.25, 3.06)	23/321	1.47 (0.92, 2.35)
III	231/738	3.70 (2.69, 5.10)	58/738	1.52 (1.03, 2.24)
IV	124/206	7.94 (5.25, 12.00)	21/206	1.82 (1.05, 3.16)
p-trend		<0.001		0.01
Tumor subtype
Low-grade endometrioid	100/2778	1.00	126/2278	1.00
High-grade endometrioid	116/614	3.95 (2.78, 5.62)	34/614	1.16 (0.78, 1.73)
Serous	210/712	4.88 (3.45, 6.90)	71/712	1.70 (1.19, 2.43)
Carcinosarcoma	119/331	7.52 (5.05, 11.20)	32/331	1.62 (1.05, 2.51)
Clear cell	37/174	4.39 (2.51, 7.68)	9/174	0.96 (0.47, 1.96)
chi-square p		<0.001		0.03
Adjuvant therapy
None	155/2619	1.00	154/2619	1.00
Chemotherapy	241/632	1.39 (0.93, 2.07)	39/632	0.56 (0.35, 0.87)
Radiation	85/750	1.12 (0.78, 1.61)	44/750	0.73 (0.52, 1.03)
Chemotherapy and radiation	11/573	1.18 (0.79, 1.75)	33/573	0.63 (0.41, 0.98)
Other	6/35	1.19 (0.35, 4.08)	2/35	0.75 (0.17, 3.23)
chi-square p		0.61		0.06

¹HRs from Fine and Gray semi-proportional competing risk model adjusted for age (continuous), stage (I, II, III, IV), tumor subtype (low-grade endometrioid, high-grade endometrioid, serous, carcinosarcoma, clear cell), adjuvant therapy (none, chemotherapy, radiation, chemotherapy plus radiation, unknown)

Similarly, non-endometrial carcinoma subhazards were significantly associated with age at diagnosis (P= HR=1.08, 95% CI=1.06–1.09 per year, P trend <0.001), stage (HR comparing stage IV to stage I=1.82, 95% CI=1.05–3.16, P-trend =0.01), and tumor subtype (P=0.03), while adjuvant therapy was not significantly related (P=0.06).

Table 2 presents associations between established endometrial carcinoma risk factors and endometrial carcinoma-specific subhazards stratified by tumor subtype with adjustment for all risk factors, education, annual income, and stage. Analyses related to non-endometrial carcinoma mortality are presented in supplemental tables, given the uncertainty related to deaths in this category. Among low-grade endometrioid cases, endometrial carcinoma-specific subhazards were significantly associated with age at diagnosis (HR=1.04, 95% CI=1.01–1.06 per year, P-trend) and BMI (HR comparing class II obesity vs. normal BMI=2.29, 95% CI=1.06–4.98, P-trend=0.01).

Table 2.

Multivariable subhazard ratios (HRs) and 95% confidence intervals (CIs) for associations between etiologic risk factors and endometrial carcinoma-specific mortality stratified by tumor subtype in the NRG Oncology/GOG 210 trial, 2003–2011

	Low-grade endometrioid n=2,278 deaths=100		High-grade endometrioid n=614 deaths=116		Serous n=712 deaths=210		Carcinosarcoma n=331 deaths=119		Clear cell n=174 deaths=37
Characteristic	HR (95% CI)1	p2	HR (95% CI)1	p2	HR (95% CI)2	p2	HR (95% CI)2	p2	HR (95% CI)2	P2
Age at diagnosis, per year	1.04 (1.01, 1.06)	0.002	1.05 (1.02, 1.07)	<0.001	1.01 (0.99, 1.02)	0.48	1.00 (0.97, 1.03)	0.77	1.05 (0.98, 1.13)	0.16
Ethnicity		0.34		0.13		0.82		0.90		0.23
White	1.00		1.00		1.00		1.00		1.00
Black	1.55 (0.74, 3.26)		1.44 (0.80, 2.61)		1.13 (0.77, 1.64)		1.12 (0.67, 1.87)		0.24 (0.05, 1.25)
Other	1.71 (0.58, 5.06)		2.58 (0.87, 7.70)		1.00 (0.51 (1.96)		1.04 (0.31, 3.47)		2.73 (0.04, 189.77)
Parity		0.88		0.78		0.003		0.05		0.21
Nulliparous	1.00		1.00		1.00		1.00		1.00
Parous	0.94 (0.54, 1.62)		1.09 (0.61, 1.94)		0.55 (0.36, 0.82)		2.01 (1.00, 4.05)		0.25 (0.05, 1.25)
BMI (kg/m2)		0.04		0.92		0.89		0.21		0.001
Normal (18.5–24.99)	1.00		1.00		1.00		1.00		1.00
Underweight (<18.5)	3.61 (0.53, 24.61)		2.94 (0.19, 45.13)		1.47 (0.25, 8.68)		1.04 (0.06, 17.23)		NE
Overweight (25–29.99)	1.04 (0.47, 2.30)		1.24 (0.69, 2.22)		0.98 (0.63, 1.54)		0.78 (0.42, 1.44)		4.87 (1.64, 14.48)
Class I Obese (30–34.99)	0.86 (0.38, 1.91)		1.01 (0.52, 1.95)		1.01 (0.63, 1.60)		0.62 (0.30, 1.29)		2.21 (0.29, 16.93)
Class II Obese (35–39.99)	2.29 (1.06, 4.98)		1.02 (0.47, 2.23)		1.14 (0.70, 1.84)		1.05 (0.51, 2.15)		0.55 (0.04, 7.63)
Class III Obese (≥40)	1.92 (0.86, 4.31)		1.30 (0.66, 2.54)		0.79 (0.43, 1.45)		1.55 (0.74, 3.24)		11.96 (3.51, 40.75)
p-trend3	0.01		0.63		0.86		0.38		0.17
Diabetes		0.28		0.75		0.95		0.14		0.49
Never	1.00		1.00		1.00		1.00		1.00
Ever	0.73 (0.38, 1.92)		0.92 (0.55, 1.54)		0.99 (0.69, 1.41)		0.66 (0.38, 1.15)		0.56 (0.10, 2.96)
Smoking status		0.36		0.49		0.96		0.51		0.34
Non-smoker	1.00		1.00		1.00		1.00		1.00
Current smoker	1.65 (0.73, 3.74)		1.59 (0.74, 3.41)		1.11 (0.53, 2.31)		1.47 (0.63, 3.44)		0.41 (0.11, 1.45)
Former smoker	0.88 (0.53, 1.44)		1.00 (0.64, 1.57)		1.00 (0.72, 1.37)		1.23 (0.78, 1.95)		0.37 (0.05, 2.75)
Breast cancer and tamoxifen use		0.15		0.83		0.04		0.35		0.007
No breast cancer/no tamoxifen	1.00		1.00		1.00		1.00		1.00
Breast cancer/no tamoxifen	2.45 (1.00, 5.99)		0.76 (0.27, 2.26)		0.09 (0.01, 0.71)		1.21 (0.39, 3.74)		NE
No breast cancer/tamoxifen use	NE		NE		1.00 (0.72, 1.37)		NE		NE
Breast cancer/tamoxifen use	0.98 (0.24, 3.94)		0.77 (0.22, 2.72)		1.15 (0.70, 1.87)		1.55 (0.85, 2.83)		0.07 (0.01, 0.48)
Oral contraceptive use		0.33		0.56		0.43		0.49		0.61
Never	1.00		1.00		1.00		1.00		1.00
Ever	0.79 (0.50, 1.27)		0.88 (0.56, 1.37)		1.15 (0.81, 1.61)		0.83 (0.49, 1.41)		1.38 (0.40, 4.78)
Menopausal hormone use		0.92		0.83		0.34		0.56		0.06
None	1.00		1.00		1.00		1.00		1.00
Estrogen only	0.87 (0.42, 1.68)		1.19 (0.57, 2.45)		0.72 (0.37, 1.40)		1.01 (0.48, 2.13)		0.20 (0.03, 1.25)
Progestin only	1.33 (0.46, 3.86)		0.58 (0.06, 5.93)		0.20 (0.02, 1.83)		1.92 (0.50, 7.30)		NE
Estrogen plus progestin	1.07 (0.57, 1.98)		0.87 (0.53, 1.43)		0.84 (0.54, 1.31)		0.75 (0.41, 1.36)		0.09 (0.01, 0.85)

¹HRs from Fine and Gray semi-proportional competing risk model adjusted for all variables shown in the table, stage (I, II, III, IV), education (High school/GED, some college/technical school, college graduate/beyond), income ($20,000–$39,999, $40,000–$69,999, ≥$70,000)

²Wald chi-square p-value excluding the unknown category

³p-value (two-sided) for trend was calculated using the Wald test for the ordinal variable based on the categories (excluding underweight) and referent group shown

NE=not estimable

Only age at diagnosis was related to endometrial carcinoma-specific subhazards (HR=1.05, 95% CI=1.02–1.07 per year, P-trend<0.001) among high-grade endometrioid cases. Among serous cases, endometrial carcinoma-specific subhazards were decreased with parity relative to nulliparity (HR=0.55, 95% CI=0.36–0.82) and a history of breast cancer that did not include tamoxifen treatment compared with lack of both exposures (HR=0.09, 95% CI=0.01–0.71), although the confidence interval was wide. We observed increased endometrial carcinoma-specific subhazards associated with parity among carcinosarcoma cases (HR=2.01, 95% CI=1.00–4.05).

Finally, among clear cell cases, increased endometrial carcinoma-specific subhazards were observed for underweight (HR=4.87, 95% CI=1.64–14.48) and class III obesity (HR=11.96, 95% CI=3.51–40.75) compared with normal-weight; however the confidence intervals were wide due to the small number of cases and events for this tumor subtype. We also observed decreased endometrial carcinoma-specific subhazards related to a history of breast cancer with tamoxifen treatment compared with lack of both exposures (HR=0.07, 95% CI=0.01–0.48); however this finding was based on 1 death among 9 exposed clear cell cases.

Relationships between etiologic factors and non-endometrial carcinoma subhazards are shown in Supplemental Table 1. Non-endometrial carcinoma subhazards were associated with age at diagnosis among women with all tumor subtypes, except clear cell. Further, non-endometrial carcinoma subhazards were associated with certain factors among the tumor subtypes as follows: parity, BMI, diabetes and smoking among low-grade endometrioid cases; smoking among high-grade endometrioid cases; and BMI among serous cases.

Supplemental Tables 2–6 show associations between risk factors and endometrial carcinoma-specific subhazards by enrollment period. Only among serous cases did the relationship between endometrial carcinoma-specific subhazards and risk factors (MHT use) significantly differ in the pre-amendment vs. post-amendment period (Supplemental Table 4).

DISCUSSION

Our study confirms the established findings that non-endometrioid histologic subtype, advanced stage and older age are associated with worse prognosis. In the U.S. population, low-grade endometrioid carcinomas comprise approximately 72% of malignant endometrial tumors, and despite an overall favorable prognosis, these tumors account for 26% of cancer-specific deaths (42). Accordingly, identifying non-traditional factors associated with worse prognosis in this group is particularly important, as these women may benefit from more intensive medical management.

In agreement with some (14, 16–22), but not all prior analyses (2–5, 7–13), our findings demonstrate that obesity is associated with an increased risk of death among women with low-grade endometrioid tumors. Furthermore, older age was related to increased risk of death, a finding well supported by the literature (43–51). The possible reasons for the worse prognosis for older or obese low-grade endometrioid endometrial carcinoma patients are speculative. Given that BMI and circulating estrogen levels are positively correlated among postmenopausal women, we hypothesize that tumors among heavier women have increased capacity to grow in metastatic environments. In addition, BMI is posited to represent a pro-inflammatory state and insulin resistance, which may drive carcinogenesis through mechanisms independent of hormones. The associations of worse prognosis with age could represent residual confounding related to more advanced disease within stage.

Relationships of risk factors with endometrial carcinoma-related deaths among non-endometrioid cases were less clear in our analysis. Parity was related to a lower risk of death among women with serous carcinomas, but increased risk of death among women with carcinosarcomas. These effects on behavior could reflect differences in the biology between the pure serous carcinoma and the carcinosarcomas, which demonstrate mixed epithelial and stromal differentiation. It is of further interest that carcinosarcomas are viewed by some researchers as a form of “metaplastic carcinoma” and descriptive studies suggest that the epithelial component may appear endometrioid in some cases and non-endometrioid in others, suggesting these tumors may arise via separable etiological process, akin to the Type I versus Type II distinction among endometrial carcinomas (52). Moreover, the molecular profile of carcinosarcomas suggests two distinctive profiles – one that is endometrioid-like, with mutations in PTEN and ARID1A and one that is serous-like, with mutations in TP53 and PPP2R1A(53).

Similar to the etiology of carcinosarcomas, some pathology investigators have posited that clear cell carcinoma is not readily classifiable as a Type I or Type II tumor (54). Data suggest that these tumors have some specific genetic abnormalities that distinguish them from endometrioid carcinomas, and these tumors portend a worse prognosis (42). Perhaps the better prognosis of clear cell carcinomas associated with use of estrogen plus progestin menopausal hormones is an indicator of a more indolent Type I phenotype.

The underlying causes of non-endometrial carcinoma death, while unknown in our study population, are likely to be driven by cardiovascular events, the leading cause of death among women with endometrial carcinoma (55). In line with this notion, we observed that older age, elevated BMI, smoking, and diabetes – exposures that affect longevity more generally – were associated with non-endometrial carcinoma mortality among endometrial carcinoma patients.

Our results demonstrate that etiologic risk factors known to influence endometrial carcinogenesis have prognostic impact following a diagnosis. Overall, women with low-grade endometrioid tumors have excellent prognosis; however, incorporating information on additional prognostic factors may identify women who can safely avoid adjuvant therapy. Several endometrial carcinoma prediction models have been developed for overall survival (56, 57), lymph node metastasis (58–60), loco-regional or distant recurrence (61), and receipt of radiotherapy (62). On the whole, these models only incorporate information on pathologic characteristics obtained from surgery with the exception of a recent risk-scoring model that included patient-level characteristics. (57). Our results, along with those from AlHilli et al. (57) suggest that patient-level factors are relevant for endometrial carcinoma prognosis, and that histology-specific risk-scoring models integrating these factors may be more accurate for individualized treatment planning and follow-up. Conversely, women with high-grade endometrioid and non-endometrioid tumors have a poor prognosis regardless of stage or age. An identification of non-traditional factors related to prognosis may enhance our understanding of the underlying biology of these tumors as well as suggest tailored treatment strategies.

Our study had several strengths including a large overall sample of patients, standardized review of histologic diagnoses by a panel of experts, and information on numerous epidemiologic risk factors. Furthermore, our ability to examine associations separately by tumor subtype is a major advantage over previous studies which either considered overall estimates or stratification by broad categories (Type I and Type II). Finally, we demonstrated that these risk factors are independently related to outcomes after adjustment for clinical factors such as stage and measures of socioeconomic status, which are known to influence endometrial carcinoma survival (1, 63). Despite the large overall sample size, some exposure categories for non-endometrioid subtypes had small numbers, resulting in imprecise estimates. Further, changes in enrollment criteria affected the distribution of tumor subtypes we observed in our study as compared with national estimates; however, our sensitivity analyses did not demonstrate major differences in relationships between risk factors and prognosis before and after these changes occurred. Finally, we lacked information on causes of death other than endometrial carcinoma which limits our ability to interpret relationships observed for this outcome.

In conclusion, we observed that factors associated with the incidence of endometrial carcinoma are related to endometrial carcinoma-specific mortality following a diagnosis. Additional studies, particularly those performed within pooling projects, are needed to confirm our findings. A better understanding of factors related to endometrial carcinoma progression and survival is needed to improve approaches for treatment and provide insights that may improve long-term prognosis.

Research highlights.

• Non-endometrioid histologic subtype, advanced stage and older age are associated with worse endometrial carcinoma prognosis.

• Obesity is related to higher mortality among low-grade endometrioid cases.

• Parity is associated with lower mortality in serous cases but higher mortality in carcinosarcoma cases.