Prospective Cohort Study of Pre- and Postdiagnosis Obesity and Endometrial Cancer Survival

Abstract

Background

Disease-free survival (DFS) and overall survival (OS) associations with anthropometric measures of obesity and changes in these exposures remain unknown among endometrial cancer survivors.

Methods

Endometrial cancer survivors diagnosed between 2002 and 2006 completed direct anthropometric measurements and self-reported lifetime weight history during in-person interviews approximately 4 months after diagnosis (peridiagnosis) and approximately 3 years after diagnosis (follow-up). Participants were followed-up until death or March 20, 2019. Cox proportional regression was used to estimate multivariable-adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) for body mass index (BMI), weight, waist circumference, and waist-hip ratio with DFS and OS. Statistical tests were 2-sided.

Results

A total of 540 and 425 cancer survivors were assessed peridiagnosis and follow-up, respectively. During the median 14.2 years of follow-up (range = 0.3-16.5 years), 132 participants had a recurrence and/or died (DFS), with 111 deaths overall (OS). Reduced DFS was noted with greater recalled weight 1 year before diagnosis (HR = 1.88, 95% CI = 1.15 to 3.07), BMI 1 year before diagnosis (HR = 1.88, 95% CI = 1.09 to 3.22), and measured peridiagnosis BMI (HR = 2.04, 95% CI = 1.18 to 3.53). Measured peridiagnosis waist circumference of at least 88 cm was associated with decreased DFS (HR = 1.94, 95% CI = 1.24 to 3.03) and OS (HR = 1.90, 95% CI = 1.16 to 3.13). A twofold decrease in DFS and OS was associated with a BMI of at least 5% or weight change from 1 year before diagnosis to peridiagnosis. No associations were observed for the assessment during follow-up.

Conclusions

One-year before- and peridiagnosis anthropometric measures of obesity were associated with reduced survival among endometrial cancer survivors. Anthropometric changes from 1 year before to peridiagnosis may provide an important indication of future survival in this population.

Endometrial cancer incidence and mortality rates continue to increase and contribute to the estimated 97 000 endometrial cancer–specific deaths that occurred globally in 2020 (1-3). Endometrial cancer survivors live with more comorbidities and in particular have a higher prevalence of obesity compared with women without an endometrial cancer diagnosis (prevalence of body mass index [BMI] ≥30 kg/m² = 60.7% vs 32.4%) (4). Although obesity has primarily been associated with the development and prognosis of lower-stage endometrioid cancers, recent evidence suggests that adiposity may have a prognostic role for nonendometrioid cancer survivors as well (5-7).

To date, most studies involving endometrial cancer survivors have used BMI to approximate total body fat (5,8). However, results from the Nurses’ Health Study suggested that for every approximately 10-cm increase in waist circumference, an indirect indicator of visceral adiposity, healthy women have an 18.0% (95% confidence interval [CI] = 9% to 28%) and 34.0% (95% CI = 23% to 39%) increase in obesity-linked cancer mortality and all-cause mortality, respectively (9). Visceral adiposity has been associated with increased systemic inflammation, insulin metabolism dysregulations, and hyperglycemia, which represent several potential prognostic biological pathways for cancer survivors (5,10). Therefore, it is relevant to consider anthropometrics that assess the distribution of adiposity, and specifically approximate visceral fat, in addition to BMI in this population (5,11,12).

Women living with a BMI of at least 30 kg/m² before or at endometrial cancer diagnosis experience a 1.3- to 1.8-fold decrease in overall survival (OS) compared with those with a BMI in the recommended range (18.5-24.9 kg/m²) (8,13-16). Weight change before any cancer diagnosis has also been associated with worse prognosis among survivors (17). However, the effect of obesity during survivorship or obesity changes pre- to postdiagnosis on prognostic outcomes remains unknown among endometrial cancer survivors (1,8,19). Likewise, although evidence from the general population has reported decreased OS for women with obesity at age 25 years (20), the impact of obesity during young adulthood on later endometrial cancer survival has yet to be determined.

Therefore, the aim of this study was to understand how pre- and postdiagnosis anthropometric measures affect survival after endometrial cancer. The primary objective was to determine the associations of 1 year before, peri-, and 3 years postdiagnosis BMI, weight, waist circumference, and waist-hip ratio with disease-free survival (DFS) and OS among endometrial cancer survivors. Our secondary objective was to assess the associations between lifetime weight change, anthropometric changes 1 year before diagnosis, and anthropometric change after diagnosis with survival outcomes.

Methods

Setting and Participants

The Alberta Endometrial Cancer Cohort Study is a follow-up of women with primary endometrial cancer who participated in a previous case-control study conducted between 2002 and 2006 in Alberta, Canada. Full cohort and eligibility details have been previously described (21,22). Briefly, women who were aged 30-80 years with no prior history of cancer except nonmelanoma skin cancer were eligible to participate and were identified through the Alberta Cancer Registry (ACR) (21). Of the 540 case patients included in the cohort, 425 participants also completed follow-up assessments between 2006 and 2011 and were included in follow-up analyses (Figure  1) (22,23). Ethical approval for the study and follow-up assessments was obtained from the former Alberta Cancer Board, University of Calgary, and University of Alberta. All participants provided informed, written consent.

Figure 1.

Cohort overview of the Alberta Endometrial Cancer Cohort, AB, Canada. BMI = body mass index.

Anthropometric and Covariate Data Collection

Trained interviewers used a standardized protocol to assess height, weight, and waist and hip circumference during peridiagnosis and follow-up in-person interviews (24). All measurements were taken in triplicate and the average used as the final value (24). At the peridiagnosis assessment, cognitive interviewing methods were used to assist participants in reporting their weight at each decade starting at age 20 years and 1 year before their cancer diagnosis (21,25). Peridiagnosis and follow-up interviews were conducted a median of 4.4 months (interquartile range = 3.4-5.7 months) and 3.4 years (interquartile range = 3.2-3.8 years) after diagnosis, respectively (23).

Interviewer-administered questionnaires captured demographic information, medical and reproductive histories, hormone use, comorbidities, and family history of cancer. Smoking habits were assessed at peridiagnosis and follow-up interviews. Lifetime alcohol consumption and total caloric intake were also assessed at both time points with self-administered Canadian Diet Health Questionnaire-I (26). Lifetime physical activity was measured with the Lifetime Total Physical Activity Questionnaire at peridiagnosis and a modified Lifetime Total Physical Activity Questionnaire at follow-up assessments (27). At follow-up, comorbidities were summarized using the Charlson Comorbidity Index (28).

Ascertainment of Outcomes

Clinical information including date of diagnosis, cancer histology and stage, primary and adjuvant treatment(s), and cancer recurrence, were obtained from ACR medical chart abstractions. Cancer stage and grade were assessed with the American Joint Committee on Cancer guidelines and from tissue samples according to the International Federation of Gynecology and Obstetrics guidelines, respectively (24,29). Vital status was assessed from record linkage with the ACR that maintains these data from Vital Statistics Alberta and Statistics Canada. Participants were followed up from the date of diagnosis until death or March 20, 2019, whichever occurred first.

Statistical Analysis

BMI and waist-hip ratio were categorized according to the World Health Organization standards (BMI = ≤24.9 kg/m², 25-29.9 kg/m², ≥30 kg/m²; waist-hip ratio: ≤0.85, >0.85) (30,31). Waist circumference was dichotomized based on the Canadian clinical threshold of 88 cm (31,32). Because there are no clinically relevant weight groups, tertiles of peridiagnosis body weight measures were used. Weight and BMI 1 year before diagnosis and weight at age 20 years were also considered. Additionally, the association between waist-hip ratio and survival outcomes was examined by BMI level (<30 vs ≥30 kg/m²) in an exploratory stratified analysis.

Changes in BMI and weight were assessed as percent change from 1 year before diagnosis to peridiagnosis and categorized as increased (≥5% gain), decreased (≥5% loss), and maintained (<5% change) groups. Percent change groups were also created for peridiagnosis to follow-up changes in BMI, weight, waist circumference, and waist-hip ratio. Lifetime weight changes were examined by weight gain since age 20 years, sum of weight change between decades (ie, sum of absolute weight changes between each decade from age 20 years until the peridiagnosis interview), and difference between maximum and minimum weights over adult lifetime (24).

Kaplan-Meier curves and Cox proportional hazards models were used to visualize and estimate multivariable-adjusted hazard ratios and 95% confidence intervals for anthropometric measures with DFS and OS endpoints, respectively. Survival time was measured from cancer diagnosis in 1 year before diagnosis and peridiagnosis analyses. Follow-up analyses time was measured from follow-up interviews. DFS and OS were defined as the time to the first recurrence or death, and death from any cause, respectively.

Age at diagnosis (years), cancer stage (I, II, III/IV), grade (I/II, III, unknown or nonapplicable), primary treatment(s) (hysterectomy, chemotherapy, radiation, any adjuvant, missing), and comorbidities (diagnosis: number of major comorbidities peridiagnosis: 0, 1, ≥2; follow-up: Charlson Comorbidity Index) were selected as covariates a priori based on biological plausibility. Models were additionally adjusted for family history of uterine or colorectal cancer and age based on backwards elimination. There was insufficient evidence that marital status, parity, menopausal status, hormone therapy status, highest education attained, lifetime alcohol consumption, caloric intake, smoking pack years, or total physical activity confounded the associations of interest.

Visual and statistical assessments of Schoenfeld residuals were used to assess the proportional hazards assumption. If an analysis did not satisfy the assumption, models were restricted to the first 10 years of analysis time from diagnosis (before diagnosis and peridiagnosis models) or follow-up interviews (follow-up models). Additionally, results from analysis with less than 10 events per group were not reported. Statistical tests were 2-sided, and a P value less than .05 was considered statistically significant. All analyses were performed with STATA 16 (StataCorp LLC, College Station, TX).

Results

Cohort Characteristics

Among the 540 participants included in this cohort, 98, 142, and 300 participants had a BMI of less than or equal to 24.9 kg/m², 25-29.9 kg/m², and greater than or equal to 30.0 kg/m², respectively (Table  1). Peridiagnosis, participants’ mean (SD) age was 59 (9) years, weight was 85 (22) kg, and waist circumference was 98 (19) cm. Most participants had stage I cancers (79.3%) and had hysterectomies (97.6%). Characteristics of participants who completed follow-up were similar to those who did not. During a median 14.2 years of follow-up (range = 0.3-16.5 years), 132 participants had a recurrence and/or died. Specifically, 73 participants had a recurrence, 50 participants died of endometrial cancer, and there were 111 deaths overall.

Table 1.

Descriptive characteristics of the Alberta Endometrial Cancer Cohort participants (N = 540) and follow-up participants (N = 425) by vital status, 2002-2019

Characteristic	All	Alive	Disease-free survival	Overall deaths
Cohort
Total No.	540	429	132	111
Mean age at diagnosis (SD), y	59.1 (9.3)	57.6 (8.8)	63.5 (9.8)	64.9 (9.2)
Highest education, No. (%)
High school diploma	177 (32.8)	134 (31.2)	50 (37.9)	43 (38.7)
Nonuniversity certificate	249 (46.1)	198 (46.2)	61 (46.2)	51 (46.0)
University degree	114 (21.1)	97 (22.6)	21 (15.9)	17 (15.3)
Married or common-law, No. (%)	372 (68.9)	300 (69.9)	84 (63.6)	72 (64.9)
Race or ethnicity, No. (%)
White	507 (93.9)	401 (93.5)	123 (93.2)	106 (95.5)
First Nation and Métis	6 (1.1)	6 (1.4)	1 (0.8)	0 (—)
Asian	18 (3.3)	15 (3.5)	6 (4.6)	3 (2.7)
Hispanic and Latin	3 (0.6)	2 (0.5)	1 (0.8)	1 (0.9)
Not reported	6 (1.1)	5 (1.2)	1 (0.8)	1 (0.9)
Histology
Endometrioid	458 (84.8)	380 (88.6)	97 (73.5)	78 (70.3)
Nonendometrioid	82 (15.2)	49 (11.4)	35 (26.5)	33 (29.7)
Overall AJCC stagea, No. (%)
I	428 (79.3)	362 (84.4)	78 (59.1)	66 (59.5)
II	69 (12.8)	49 (11.4)	25 (18.9)	20 (18.0)
III or IV	43 (8.0)	18 (4.2)	29 (22.0)	25 (22.5)
FIGO grade, No. (%)
<6%	288 (53.3)	250 (58.3)	47 (35.6)	38 (34.2)
6%-50%	125 (23.2)	102 (23.8)	27 (20.5)	23 (20.7)
>50%	73 (13.5)	41 (9.6)	36 (27.3)	32 (28.8)
Other	54 (10.0)	36 (8.4)	22 (16.7)	18 (12.2)
Primary treatmentb, No. (%)
Surgery	527 (97.6)	422 (98.4)	120 (90.9)	105 (94.6)
Chemotherapy	45 (8.3)	26 (6.1)	20 (15.2)	19 (17.1)
Hormone therapy	6 (1.1)	6 (1.4)	2 (1.5)	0 (0)
Radiation therapy	168 (31.1)	124 (28.9)	52 (39.4)	44 (39.6)
Mean weight (SD), kg	84.9 (21.5)	84.7 (20.9)	86.1 (24.3)	85.4 (23.9)
Mean BMI (SD), kg/m2	32.3 (7.9)	33.1 (7.6)	33.2 (9.1)	33.0 (9.05)
Mean waist circumference (SD), cm	98.0 (18.8)	97.4 (18.2)	100.6 (20.1)	100.5 (20.9)
Waist-hip ratio, mean (SD)	0.85 (0.1)	0.84 (0.1)	0.86 (0.1)	0.86 (0.1)
Subcohort
Total No.	425	365	80	60
Mean age at diagnosis (SD), y	58.8 (8.9)	57.9 (8.4)	62.5 (10.4)	64.3 (10.2)
Race or ethnicity, No. (%)
White	407 (95.8)	347 (95.1)	76 (95.0)	60 (100.0)
First Nation and Métis	2 (0.5)	2 (0.6)	1 (1.2)	0 (—)
Asian	12 (2.8)	12 (3.3)	3 (3.8)	0 (—)
Hispanic and Latin	0 (—)	0 (—)	0 (—)	0 (—)
Not reported	4 (0.9)	4 (1.1)	0 (—)	0 (—)
Histology, No. (%)
Endometrioid	368 (86.6)	321 (88.0)	65 (81.3)	47 (78.3)
Nonendometrioid	57 (13.4)	44 (12.0)	15 (18.7)	13 (21.7)
Overall AJCC stage, No. (%)
I	346 (81.4)	305 (83.6)	52 (65.0)	41 (68.3)
II	55 (12.9)	44 (12.1)	16 (20.0)	11 (18.3)
III or IV	24 (5.7)	16 (4.4)	12 (15.0)	8 (13.3)
FIGO grade, No. (%)
<6%	238 (56.0)	206 (56.4)	44 (55.0)	32 (53.3)
6%-50%	102 (24.0)	89 (24.4)	17 (21.3)	13 (21.7)
>50%	50 (11.8)	41 (11.2)	13 (16.3)	9 (15.0)
Other	35 (8.2)	29 (8.0)	6 (7.5)	6 (10.0)
Primary treatment, No. (%)
Surgery	416 (97.8)	359 (98.4)	72 (90.0)	57 (95.0)
Chemotherapy	32 (7.5)	26 (7.1)	8 (10.0)	6 (10.0)
Hormone therapy	5 (1.18)	5 (1.37)	1 (1.25)	0 (—)
Radiation therapy	132 (31.1)	111 (30.4)	29 (36.3)	21 (35.0)
Mean weight (SD), kg	84.9 (21.8)	84.2 (21.0)	88.8 (25.8)	88.8 (25.6)
Mean BMI (SD), kg/m2	32.2 (8.0)	31.9 (7.7)	34.0 (9.6)	34.1 (9.7)
Mean waist circumference (SD), cm	97.5 (18.7)	96.8 (18.4)	101.8 (18.6)	102.3 (19.7)
Waist-hip ratio, mean (SD)	0.84 (0.1)	0.84 (0.1)	0.86 (0.1)	0.86 (0.1)
Postdiagnosis mean weight (SD), kg	85.7 (23.0)	85.2 (22.3)	89.3 (27.4)	88.3 (26.5)
Postdiagnosis mean BMI (SD), kg/m2	32.8 (8.5)	32.5 (8.2)	34.5 (10.4)	34.1 (10.3)
Postdiagnosis mean waist circumference (SD), cm	100.9 (18.3)	100.3 (18.0)	105.1 (19.6)	104.7 (20.2)
Postdiagnosis waist-hip ratio, mean (SD)	0.86 (0.1)	0.86 (0.1)	0.86 (0.1)	0.87 (0.1)

^aParticipants with incomplete TNM stage (n = 8) were classified based on available lymph node inclusion and metastatic information as stage I. AJCC = American Joint Committee on Cancer; BMI = body mass index; FIGO = International Federation of Gynecology and Obstetrics.

^bThe frequencies for treatment are not mutually exclusive because participants could have multiple treatments.

Anthropometric Measures at 1 Year Before Diagnosis, Peridiagnosis, and Follow-Up

A BMI of at least 30 kg/m² and continuous BMI 1 year before diagnosis were associated with reduced DFS (HR = 1.88, 95% CI = 1.09 to 3.22; Table  2). Reduced DFS was also noted with greater weight tertiles and continuous weight 1 year before diagnosis (T2 = 73.2-90.7 kg: HR_T2 = 1.75, 95% CI = 1.13 to 2.69; T3 > 90.7 kg: HR_T3 = 1.88, 95% CI = 1.15 to 3.07). Only the middle weight tertile was statistically significantly associated with reduced OS (HR_T2 = 1.62, 95% CI = 1.01 to 2.59). Weight at age 20 years was not associated with either survival outcome.

Table 2.

Disease-free and overall survival in relation to BMI, weight, waist circumference, and waist-hip ratio at 1 year before diagnosis, peridiagnosis, and follow-up time points in the Alberta Endometrial Cancer Cohort Study, 2002-2019

Anthropometric characteristic	Disease-free survival		Overall survival
	Events/cases	HR (95% CI)	Events/cases	HR (95% CI)
1 year before diagnosisa
BMIb
≤24.9 kg/m2	19/108	1.00 (Referent)	17/108	1.00 (Referent)
25-29.9 kg/m2	38/144	1.40 (0.79 to 2.47)	31/144	1.13 (0.61 to 2.07)
≥30 kg/m2	74/286	1.88 (1.09 to 3.22)	62/286	1.42 (0.81 to 2.50)
Ptrendc		.004		.16
Per 1 kg/m2	131/537	1.03 (1.01 to 1.06)	110/537	1.02 (0.99 to 1.04)
Weight
T1: <73.2 kg	40/194	1.00 (Referent)	35/194	1.00 (Referent)
T2: 73.2-90.7 kg	52/184	1.75 (1.13 to 2.69)	44/184	1.62 (1.01 to 2.59)
T3: >90.7 kg	39/159	1.88 (1.15 to 3.07)	31/159	1.66 (0.97 to 2.87)
Ptrendc		.005		.15
Per 1 kg	131/537	1.01 (1.00 to 1.02)	110/537	1.01 (1.00 to 1.02)
Weight at age 20 years
T1: <53.5 kg	51/183	1.00 (Referent)	42/184	1.00 (Referent)
T2: 53.5-60.0 kg	32/177	0.82 (0.52 to 1.30)	29/177	0.84 (0.51 to 1.37)
T3: >60.0 kg	49/180	1.50 (0.99 to 2.26)	40/180	1.46 (0.93 to 2.31)
Ptrendc		.28		.06
Per kg	132/540	1.01 (0.99 to 1.02)	111/540	1.02 (1.00 to 1.04)
Peridiagnosisa
BMI
≤24.9 kg/m2	18/98	1.00 (Referent)	17/98	1.00 (Referent)
25-29.9 kg/m2	36/142	1.26 (0.70 to 2.25)	28/142	0.86 (0.46 to 1.60)
≥30 kg/m2	78/300	2.04 (1.18 to 3.53)	66/300	1.51 (0.86 to 2.65)
Ptrendc		.003		.07
Per 1 kg/m2	132/540	1.03 (1.01 to 1.05)	111/540	1.02 (1.00 to 1.05)
Weight
T1: <73.2 kg	43/180	1.00 (Referent)	38/181	1.00 (Referent)
T2: 73.2-90.7 kg	49/180	1.48 (0.96 to 2.28)	40/180	1.32 (0.82 to 2.13)
T3: >90.7 kg	40/180	1.60 (0.99 to 2.57)	33/180	1.49 (0.88 to 2.52)
Ptrendc		.004		.08
Per 1 kg	132/540	1.01 (1.00 to 1.02)	111/540	1.01 (1.00 to 1.02)
Waist circumference
<88 cm	33/173	1.00 (Referent)	29/173	1.00 (Referent)
≥88 cm	96/364	1.94 (1.24 to 3.03)	79/364	1.90 (1.16 to 3.13)
Per 1 cm	129/536	1.02 (1.01 to 1.03)	108/536	1.02 (1.00 to 1.03)
Waist-hip ratio
≤0.85	67/299	1.00 (Referent)	57/299	1.00 (Referent)
>0.85	62/237	1.23 (0.85 to 1.79)	51/237	1.33 (0.89 to 2.00)
Per 0.1 unit change	129/536	1.21 (0.97 to 1.50)	108/536	1.29 (1.01 to 1.65)
At follow-upd
BMI
≤24.9 kg/m2	13/75	1.00 (Referent)	13/75	1.00 (Referent)
25–29.9 kg/m2	15/113	0.83 (0.39 to 1.77)	12/113	0.62 (0.28 to 1.40)
≥30 kg/m2	39/235	1.11 (0.56 to 2.20)	34/235	0.91 (0.45 to 1.84)
Ptrendc		.18		.30
Per 1 kg/m2	67/423	1.02 (0.99 to 1.05)	59/423	1.02 (0.99 to 1.05)
Weight
T1:<73.2 kg	23/138	1.00 (Referent)	21/138	1.00 (Referent)
T2: 73.2-90.7 kg	19/143	0.96 (0.51 to 1.84)	16/143	0.88 (0.44 to 1.76)
T3: >90.7 kg	24/141	1.45 (0.76 to 2.74)	21/141	1.31 (0.66 to 2.60)
Ptrendc		.18		.22
Per 1 kg	67/423	1.01 (1.00 to 1.02)	59/423	1.01 (1.00 to 1.02)
Waist circumference
<88 cm	15/109	1.00 (Referent)	13/109	1.00 (Referent)
≥88 cm	52/314	1.14 (0.63 to 2.06)	46/314	1.18 (0.62 to 2.45)
Per 1 cm	67/423	1.01 (0.99 to 1.02)	59/423	1.01 (0.99 to 1.02)

^aOne year before and peridiagnosis models (n = 540) adjusted for age, grade, stage, treatment, comorbidities, family history of cancer, and age². Number of the major comorbidities includes the following conditions: diabetes, angina pectoris, high cholesterol, pulmonary embolism, myocardial infarction, hypertension, stroke, and thrombosis. BMI = body mass index (kg/m²); CI = confidence interval; HR = hazard ratio.

^bBMI less than or equal to 24.9 kg/m² group includes 3 participants with a BMI less than 18.5 kg/m²; weight tertiles based on peridiagnosis time point.

^cStatistical significance tested with 2-sided dose-response Cox proportional regression.

^dFollow-up models (n = 425) adjusted for age, grade, stage, treatment, postdiagnosis comorbidities, family history of cancer, and age².

Peridiagnosis, BMI of at least 30 kg/m² (HR = 2.04, 95% CI = 1.18 to 3.53) as well as continuous BMI and weight were associated with reduced DFS (Table  2). Waist circumference of at least 88 cm, compared with less than 88 cm, and continuously was associated with decreased DFS (HR = 1.94, 95% CI = 1.24 to 3.03) and OS (HR = 1.90, 95% CI = 1.16 to 3.13). A linear relationship for reduced OS per 0.1 unit change in peridiagnosis waist-hip ratio was observed.

Null associations were observed between follow-up anthropometrics and the survival endpoints (Table  2). Follow-up waist-hip ratio models did not satisfy the proportional hazards assumption in the full or 10-year restricted analyses (results not shown).

Change in Anthropometric Measures Before and After Diagnosis

A gain or loss of at least 5% body weight from 1 year before diagnosis to peridiagnosis was associated with reduced DFS (HR_{≥5% gain} = 1.91, 95% CI = 1.23 to 2.97; HR_{≥5% loss} = 1.83, 95% CI = 1.15 to 2.91) and OS (HR_{≥5% gain} = 2.37, 95% CI = 1.48 to 3.79; HR_{≥5% loss} = 1.76, 95% CI = 1.05 to 2.97) (Table  3). Compared with a stable BMI, a 5% increase and over (HR_{≥5% gain} = 2.12, 95% CI = 1.38 to 3.25) or decrease (HR_{≥5% loss} = 1.85, 95% CI = 1.16 to 2.97) in BMI between these time points was also associated with reduced DFS (Figure  2, A). The 1 year before diagnosis to peridiagnosis percent BMI change and OS model did not satisfy the proportional hazards assumption. Compared with having a stable BMI, a 5% and over increase or decrease in BMI during the year before diagnosis was associated with reduced OS in the 10-year restricted analysis (HR_{≥5% gain} = 2.65, 95% CI = 1.46 to 4.80; HR_{≥5% loss} = 2.38, 95% CI = 1.38 to 4.43).

Figure 2.

Kaplan-Meier survival curves for (A) 1 year before diagnosis and (B) 3 years postdiagnosis percent body mass index change group with disease-free survival in the Alberta Endometrial Cancer Cohort Study, 2002-2019.

Table 3.

Disease-free and overall survival in relation to lifetime, prediagnosis, and postdiagnosis anthropometric change in the Alberta Endometrial Cancer Cohort Study, 2002-2019

Anthropometric characteristic	Disease-free survival		Overall survival
	Events/cases	HR (95% CI)	Events/cases	HR (95% CI)
Lifetime weight changea
Weight gain since age 20 yb
T1: <18.7 kg	42/174	1.00 (Referent)	35/174	1.00 (Referent)
T2: 18.7-32.0 kg	45/173	1.66 (1.06 to 2.62)	37/173	1.36 (0.84 to 2.21)
T3: >32.0 kg	40/173	1.50 (0.93 to 2.39)	34/173	1.47 (0.87 to 2.48)
Ptrendc		.02		.34
Per 1 kg	127/520	1.01 (1.00 to 1.02)	106/520	1.01 (0.99 to 1.02)
Sum of weight change between decades
T1: <21.4 kg	35/179	1.00 (Referent)	28/179	1.00 (Referent)
T2: 21.4-36.3 kg	57/185	1.78 (1.16 to 2.75)	49/185	1.76 (1.10 to 2.82)
T3: >36.3 kg	39/173	1.49 (0.92 to 2.43)	33/173	1.55 (0.91 to 2.64)
Ptrendc		.12		.38
Per 1 kg	131/537	1.01 (1.00 to 1.01)	110/537	1.00 (1.00 to 1.01)
Difference between maximum and minimum weights over adult lifetime
T1: <19.1 kg	36/182	1.00 (Referent)	33/182	1.00 (Referent)
T2:19.1-32.2 kg	54/178	1.65 (1.07 to 2.54)	41/178	1.23(0.77 to 1.98)
T3: >32.2 kg	42/180	1.56 (0.97 to 2.52)	37/180	1.50 (0.90 to 2.50)
Ptrendc		.02		.32
Per 1 kg	132/540	1.01 (1.00 to 1.02)	111/540	1.01 (0.99 to 1.02)
Change 1 y before to peridiagnosisa
BMId
≥5% decrease	29/75	1.85 (1.16 to 2.97)	24/75	—
Maintain	68/345	1.00 (Referent)	56/345	1.00 (Referent)
≥5% increase	34/118	2.12 (1.38 to 3.25)	30/118	—
Per 1%	131/537	1.02 (0.99 to 1.05)	110/537	1.03 (1.00 to 1.06)
Weight
≥5% decrease	30/79	1.83 (1.15 to 2.91)	25/79	1.76 (1.05 to 2.97)
Maintain	71/351	1.00 (Referent)	58/351	1.00 (Referent)
≥5% increase	30/108	1.91 (1.23 to 2.97)	27/108	2.37 (1.48 to 3.79)
Per 1%	131/537	1.02 (1.00 to 1.05)	110/537	1.03 (1.00 to 1.06)
Change peridiagnosis to follow-upe
BMI
≥5% decrease	13/59	1.16 (0.60 to 2.24)	13/59	1.41 (0.71 to 2.83)
Maintain	39/231	1.00 (Referent)	27/231	1.00 (Referent)
≥5% increase	27/133	1.21 (0.73 to 2.00)	19/133	1.09 (0.60 to 1.97)
Per 1%	79/423	1.02 (0.99 to 1.06)	59/423	1.02 (0.99 to 1.06)
Weight
≥5% decrease	14/63	1.25 (0.67 to 2.35)	13/63	1.40 (0.71 to 2.76)
Maintain	38/240	1.00 (Referent)	28/240	1.00 (Referent)
≥5% increase	27/120	1.52 (0.91 to 2.52)	18/120	1.22 (0.67 to 2.24)
Per 1%	79/423	1.03 (0.99 to 1.06)	59/423	1.03 (1.00 to 1.07)
Waist circumferencef
≥5% decrease	9/48	—	9/48	—
Maintain	38/191	1.00 (Referent)	26/191	1.00 (Referent)
≥5% increase	30/181	0.78 (0.46 to 1.30)	22/181	0.87 (0.47 to 1.58)
Per 1%	79/423	1.01 (0.98 to 1.05)	59/423	1.01 (0.97 to 1.06)

^aLifetime, 1 year before, and peridiagnosis models (n = 540) adjusted for age, grade, stage, treatment, comorbidities, family history of cancer, age². BMI = body mass index; CI = confidence interval; HR = hazard ratio.

^bWeight gain since age 20 years: peridiagnosis interview weight minus recalled weight at age 20 years (only includes positive values, n = 520). Sum of weight change between decades: sum of absolute weight changes between each decade from weight at age 20 years until the peridiagnosis interview weight. Difference between maximum and minimum weights over adult lifetime: greatest weight minus lowest recalled weights.

^cStatistical significance tested with 2-sided dose-response Cox proportional regression.

^dEstimates not reported due to proportional hazard violation.

^eFollow-up models (n = 425) adjusted for age, grade, stage, treatment, postdiagnosis comorbidities, family history of cancer, and age².

^fEstimates not reported due to insufficient event numbers (n < 10).

Reduced DFS was observed for the middle (T2 = 18.7-32.0 kg) compared with the lowest tertile (T1 < 18.7 kg) of weight gain since age 20 years (HR_T2 = 1.66, 95% CI = 1.06 to 2.62) and as an inverse linear trend (Table  3). Compared with the lowest tertile of total weight change between each decade of life before cancer diagnosis (T1 < 21.4 kg), the middle tertile (T2 = 21.4-36.3 kg) was associated with reduced DFS (HR_T2 = 1.78, 95% CI = 1.16 to 2.75) and OS (HR_T2 = 1.76, 95% CI = 1.10 to 2.82). Greater differences between maximum and minimum weights during adult lifetime were associated with reduced DFS (HR_T2 = 1.65, 95% CI = 1.07 to 2.54). Lastly, null associations were observed between changes in anthropometrics from peridiagnosis to follow-up with survival outcomes (Figure  2, B).

Stratified Analyses by BMI

Among survivors with a peridiagnosis BMI less than 30 kg/m², an inverse association for every 0.1 increase in waist-hip ratio with DFS was observed, but not for those greater than or equal to 30 kg/m² (Table  4). Null associations were observed for categorical waist-hip ratio with survival endpoints when stratified by a BMI less than 30 kg/m² or greater than or equal to 30 kg/m² at peridiagnosis and follow-up.

Table 4.

Disease-free and overall survival in relation to waist-hip ratio by BMI status in the Alberta Endometrial Cancer Cohort Study, 2002-2019

Anthropometric characteristic	Disease-free survival				Overall survival
	BMI <30 kg/m2		BMI ≥30 kg/m2		BMI <30 kg/m2		BMI ≥30 kg/m2
	Events/cases	HR (95% CI)	Events/cases	HR (95% CI)	Events/cases	HR (95% CI)	Events/cases	HR (95% CI)
Peridiagnosisa
Waist-hip ratio ≤0.85	33/172	1.00 (Referent)	34/127	1.00 (Referent)	29/172	1.00 (Referent)	28/127	1.00 (Referent)
Waist-hip ratio >0.85	21/68	1.74 (0.97 to 3.14)	41/169	0.71 (0.42 to 1.18)	16/68	1.51 (0.78 to 2.91)	35/169	1.02 (0.59 to 1.75)
Waist-hip ratio per 0.1 unit change	54/240	1.52 (1.09 to 2.11)	75/296	0.86 (0.62 to 1.18)	45/240	1.38 (0.95 to 1.99)	63/296	1.08 (0.77 to 1.52)
Follow-upb,c
Waist-hip ratio ≤0.85	18/115	1.00 (Referent)	12/83	1.00 (Referent)	17/115	1.00 (Referent)	10/83	1.00 (Referent)
Waist-hip ratio >0.85	10/73	0.63 (0.28 to 1.45)	26/151	0.94 (0.46 to 1.95)	8/73	—	23/151	0.89 (0.41 to 1.94)
Waist-hip ratio per 0.1 unit change	28/188	1.35 (0.80 to 2.28)	38/234	0.86 (0.54 to 1.37)	25/188	1.25 (0.72 to 2.17)	33/234	0.84 (0.51 to 1.40)

^aPeridiagnosis models (n = 540) adjusted for; age, grade, stage, treatment, comorbidities, family history of cancer, and age². BMI = body mass index; CI = confidence interval; HR = hazard ratio.

^bFollow-up models (n = 425) adjusted for age, grade, stage, treatment, postdiagnosis comorbidities, family history of cancer, and age².

^cEstimates not reported due to insufficient event numbers (n < 10).

Discussion

The timing and degree of change in anthropometric measures of obesity was an important indicator of future prognosis among endometrial cancer survivors in this cohort. Endometrial cancer survivors with elevated 1 year before and peridiagnosis weight, BMI, or waist circumference experienced a nearly twofold decrease in their survival. Greater peridiagnosis waist-hip ratios were also associated with worse OS in this study, although these results may be more relevant to survivors with a BMI less than 30 kg/m². Reduced survival after cancer diagnosis was noted for participants with greater lifetime weight change and anthropometric changes 1 year before diagnosis. Conversely, no anthropometric measurements from follow-up or changes in anthropometric measurements after diagnosis were associated with survival in this cohort.

Similar to a previous study that reported a null association between weight at age 25 years and later endometrial cancer-specific mortality (33), we did not observe any associations between weight at age 20 years and survival outcomes. However, the presence of obesity before endometrial cancer diagnosis has been associated with increased mortality during cancer survivorship (13–16). Specifically, endometrial cancer survivors with a BMI of at least 30 kg/m² at 5 years or at least 35 kg/m² at 1 year before their cancer diagnosis are estimated to have approximately one-half the OS of survivors with a BMI in the recommended range (13–16). We also observed that greater weight and BMI 1 year before diagnosis were associated with reduced DFS, thus corroborating these previous findings.

A recent systematic review on excess peridiagnosis body fat and survival reported a 1.3-fold decrease in OS for endometrial cancer survivors with greater (generally ≥30 kg/m²) compared with lower (18.5-24.9 kg/m²) BMI and a positive relationship between BMI and endometrial cancer recurrence (summary HR = 1.28, 95% CI = 1.06 to 1.56) (8). Therefore, the reduced DFS observed with elevated peridiagnosis BMI, weight, and waist circumference noted in this study are in agreement with previous literature. Lastly, no associations between follow-up anthropometrics and either outcome were observed in this cohort, which aligns with the null postdiagnosis BMI and OS associations previously reported by the Women’s Health Initiative (18) and the NIH-AARP Diet and Health Study (19).

Generally, weight changes in early adulthood have been associated with increased all-cause, cardiovascular disease, and, to a lesser extent, cancer-specific mortality (33–35). Greater lifetime weight changes and fluctuations were associated with reduced survival among participants of this study. Increased accumulation of visceral fat is often associated with weight cycling (36), which may partially explain the worse prognosis observed among survivors with a history of weight fluctuations. Our results also suggest that participants with a weight or BMI gain or loss of at least 5% from 1 year before diagnosis to peridiagnosis experienced an approximately 50.0% reduction in DFS and OS compared with survivors who maintained their weight. This observation aligns with previously reported associations between 10.0% and greater weight change 6 months before cancer diagnosis and reduced survival (17). Different underlying biological mechanisms may contribute to the weight gain and weight loss survival associations. Weight loss is the primary characteristic of cancer cachexia that has been associated with more aggressive histology, advanced tumor stage, and increased treatment toxicities (17). In contrast, excess adiposity via weight gain has been associated with progression of endometrial cancer and other comorbid conditions, which may have a longer latency period but still contribute to reduced survival (5,12). Cancer survivors with any weight loss at diagnosis are more likely to experience additional weight loss after diagnosis, which has been associated with worse prognosis 6 months postdiagnosis (17,37,38). Furthermore, the association between postdiagnosis weight change and prognosis may be partially dependent on at-diagnosis BMI such that worse DFS has been noted among survivors with obesity who experience postdiagnosis weight gain compared with their normal-weight counterparts (37). In our study, anthropometric changes between peridiagnosis and follow-up were not associated with survival. However, these analyses were limited to follow-up participants who had a longer interval of postdiagnosis change (approximately 3 years) compared with participants in prior studies (generally 6-24 months). Hence, survivor bias may have contributed to these null findings, because our follow-up participants may represent healthier, more resilient participants who were alive 3 years after their diagnosis.

Although the unopposed estrogen hypothesis is proposed as a main biologic mechanism between obesity and endometrial cancer risk, the mechanisms for mortality remain poorly understood (5,11). Given that endometrial cancer survivors are estimated to be 3 times more likely to die of cardiovascular disease than their endometrial cancer (39), mechanisms linked to visceral adiposity, and thus cardiovascular disease, including increased low-grade systemic inflammation, altered insulin metabolism, and hyperglycemia may be more relevant for mortality pathways in this population (5,10). Such inflammatory and insulin pathways driven by visceral adiposity may partially explain why a relationship between elevated waist-hip ratio and reduced DFS was observed among survivors without obesity. Additionally, the loss of skeletal muscle during cancer cachexia may contribute to decreased total body weight without reduced adipose tissue and has been associated with decreased survival (17).

Several gaps remain in this research area, including a limited understanding of the impact early-life obesity, lifetime weight trajectories, and weight fluctuations have on later endometrial cancer survivorship. Given the biologic plausibility of these associations, future research should focus on how the timing, type, and distribution of adiposity affects survival after an endometrial cancer diagnosis. Additionally, the effectiveness of intentional diet and physical activity weight loss interventions on survival and the prognostic role of cancer cachexia remains unknown in this population (17,40).

Strengths of this study include direct, standardized anthropometric measurements from multiple time points and lifetime weight histories obtained by interviewers trained in cognitive interviewing methods. Study participants were identified through a population-based cancer registry that has at least 95% case ascertainment rates, and all cancers were histologically confirmed, incident cases. High-quality data were available for many potential confounding factors, which permitted a thorough examination of these associations. Outcome, treatment, and follow-up care data were obtained through a combination of data linkage with Alberta’s electronic medical record system and medical chart abstractions conducted by qualified health record technicians. Despite the nearly 17-year follow-up, the relatively small sample size and generally high endometrial cancer survival rates limited our power, and the postdiagnosis estimates should be interpreted with caution given the small number of outcome events. Additionally, estimates from analysis that did not satisfy the proportional hazards assumption could not be presented.

Although obesity is an established risk factor for endometrial cancer, its impact on endometrial cancer recurrence and survival is less understood (7). This study observed that endometrial cancer survivors with elevated BMI, weight, and waist circumference 1 year before or peridiagnosis had an approximately twofold decrease in their DFS and OS. Our results also indicate that both weight gain and loss of at least 5% in the year before diagnosis were important indicators of future survival among women with endometrial cancer and highlight weight fluctuations as an important prognostic factor in this population. Monitoring of weight fluctuations at the time of endometrial cancer diagnosis coupled with interventions to assist these at-risk individuals may help improve survival in this population. In addition, future research that includes larger case numbers, more sophisticated measurements of adiposity, and body composition are needed to further our understanding of the impact that postdiagnosis weight has on survivorship.

Funding

This work was supported by 3 separate grants from the National Cancer Institute of Canada through the Canadian Cancer Society (NCIC No. 12018, NCIC No 13010, NCIC Grants No 17323) and from 1 grant awarded by the former Alberta Cancer Board (ACB Grant 22190). CM Friedenreich received career awards from the Canadian Institutes of Health Research and the Alberta Heritage Foundation for Medical Research/Alberta Innovates (AHFMR/Alberta Innovates). LS Cook and KS Courneya held Canada Research Chairs and LS Cook also received career award funding from AHFMR. LS Cook receives support from the US National Cancer Institute (NCI P30CA118100).

Notes

Role of the funders: The funders had no role in study design and conduct of the study, data collection and analysis, data interpretation, or manuscript preparation and decision to submit the manuscript for publication.

Disclosures: The authors have no conflicts of interest to disclose.

Author contributions: Conceptualization, RLKP, JM, ARM, KSC, LSC, CMF; Data curation, CMF; Formal analysis, RLKP; Funding acquisition, KSC, LSC, CMF; Investigation, KSC, LSC, CMF; Methodology, KSC, LSC, CMF; Project administration, KSC, LSC; Resources, CMF; Supervision, CMF: Visualization, RLKP; Writing—original draft, RLKP; Writing—review & editing, RLKP, JM, ARM, KSC, LSC, CMF.

Acknowledgements: We would like to thank the participants and staff of the Endometrial Disease and Physical Activity Study and Alberta Endometrial Cancer Cohort Study for their contributions to the original case-control and follow-up cohort study.

Prior presentations: Aspects of this work has been previously presented at the Canadian Society for Epidemiology and Biostatistics 2021 Virtual Conference, June 1-3, 2021, Canada.

Data Availability

The data underlying this article will be shared on reasonable request to the corresponding author.