Metabolic abnormalities (hypertension, hyperglycemia and overweight), lifestyle (high energy intake and physical inactivity) and endometrial cancer risk in a Norwegian cohort
Article first published online: 30 JAN 2003
Copyright © 2003 Wiley-Liss, Inc.
International Journal of Cancer
Volume 104, Issue 6, pages 669–676, 10 May 2003
How to Cite
Furberg, A.-S. and Thune, I. (2003), Metabolic abnormalities (hypertension, hyperglycemia and overweight), lifestyle (high energy intake and physical inactivity) and endometrial cancer risk in a Norwegian cohort. Int. J. Cancer, 104: 669–676. doi: 10.1002/ijc.10974
- Issue published online: 10 MAR 2003
- Article first published online: 30 JAN 2003
- Manuscript Revised: 27 NOV 2002
- Manuscript Accepted: 27 NOV 2002
- Manuscript Received: 20 SEP 2002
- Norwegian Cancer Society
- Aakre Foundation. Grant Number: TP 49 258
- endometrial cancer;
- metabolic abnormalities;
- Norwegian cohort
Since high energy intake, inactivity, hypertension and diabetes are linked to obesity and an unfavorable hormonal profile, we wanted to test whether energy intake, physical activity, blood pressure and serum glucose are related to the risk of endometrial cancer independent of the body mass index (BMI). A cohort of 24,460 women, aged 20–49 years, attended a Norwegian health screening twice during 1974–1981; they answered questions about diet, physical activity and chronic diseases, and their height, weight, blood pressure and non-fasting serum glucose were measured. By the end of 1996, during 15.7 years of follow-up, 130 cases of endometrial carcinomas were identified. The relative risks (RRs) for endometrial cancer were estimated in proportional hazards models including potentially confounding factors. Obese women (BMI ≥ 30 kg/m2) were at 2.6 times increased risk of endometrial cancer compared to normal weight women (BMI < 25 kg/m2) (RR = 2.57, 95%CI = 1.61–4.10). Among overweight women (BMI ≥ 25 kg/m2), non-fasting serum glucose in the upper quartile vs. in the lower quartile was associated with a 2.4 times increase in risk (RR = 2.41, 95%CI = 1.08–5.37), whereas among obese women, blood pressure above 140/90 mmHg vs. below 140/90 mmHg in both surveys was associated with a 3.5 times increase in risk (RR = 3.47, 95%CI = 1.24–9.70). Especially in women younger than 50 years, high energy intake (5,044–6,401 kJ/day) conferred higher risk compared to low energy intake (< 4266 kJ/day) (RR = 3.40, 95%CI = 1.52–7.60). Increasing recreational activity tended to be protective. Among obese women with non-sedentary jobs at both screenings, RR declined to 0.18 (95%CI = 0.05–0.62) as the level of sustained occupational activity increased (ptrend = 0.03). Our results suggest that inactivity and high energy intake are major risk factors for endometrial cancer independent of BMI, and that hypertension and relative hyperglycemia are significant markers of risk, especially among the heaviest women. © 2003 Wiley-Liss, Inc.
The wide differences in incidence of endometrial cancer (type I-carcinoma) across countries and between urban and rural populations, as well as the changing disease incidence in migrants, indicate that environmental factors have a huge influence on the occurrence of the disease.1 The etiology of endometrial cancer is unknown, yet it has been shown that malignant transformation, cancer cell proliferation, tumor invasion and tumor progression are enhanced by deficiency in progesterone and its action relative to estrogen in the glandular cells of the endometrium.2, 3 A shift in energy balance (e.g., excessive energy intake relative to energy expenditure) might contribute to an unfavorable sex hormone profile in women.2, 4
Epidemiological studies on the association between energy intake and the risk of endometrial cancer are inconclusive.5, 6, 7, 8, 9, 10 A chronically excessive intake of energy relative to requirements may lead to increased body weight. Although adult obesity has consistently been associated with an increased risk of endometrial cancer, the effect may be age and menopausal status dependent.11 The energy expended during physical activity may represent between 15 and 50% of total energy expenditure, depending on the amount of physical activity performed and the body mass.12 Epidemiological studies find that inactivity may be associated with an increased endometrial cancer risk.13, 14, 15, 16, 17, 18 However, some studies indicate that the effect may be dependent on type of activity (i.e., during work and recreation).19, 20 We have previously reported on the effect of activity, energy intake and accurately measured body mass on the incidence of breast cancer21 and similar studies on endometrial cancer are needed.11
Body mass and endogenous levels of estrogens have been associated with risk independently of each other,22, 23 suggesting that the unopposed estrogen/relative progesterone deficiency hypothesis is insufficient to explain endometrial cancer. It has been hypothesized that insulin, a growth factor known for its mitogenic activity,24, 25 plays a major role in endometrial carcinogenesis. Hypertension and diabetes are markers of insulin resistance/hyperinsulinemia.26, 27 Observations linking blood pressure, glucose metabolism and markers of insulin resistance to endometrial cancer come mostly from retrospective studies, which have provided less conclusive results because of self-reported disease history and anthropometry or an absence of adjustment for body mass.1
We analyzed data that included repeated assessments of variables from a cohort of 24,460 Norwegian women to elucidate whether energy intake, recreational and occupational activity, blood pressure and serum glucose concentration are associated with risk of endometrial cancer independent of body mass.
MATERIAL AND METHODS
The women included in our study participated in 2 population-based screening surveys started during 1974–1976 and 1977–1981 in 3 counties of Norway (Finnmark, Oppland and Sogn og Fjordane), as part of the Norwegian National Health Screening Service's program to explore the association of lifestyle with chronic diseases. In 1974–1976 all female residents, aged 35–49 years, and a 10% random sample, aged 20–34 years, received a written invitation to participate. In 4 municipalities of county of Finnmark, all women in the youngest age group were asked to meet. Of 31,509 invited women, 28,562 (90.6%) attended. In 1977–1981, those who registered at the first screening were re-invited, while a 5–11% random sample of women, aged 20–39 years, was invited for the first time. A total of 34,378 women were asked to participate and 31,209 (90.8%) actually did; 26,127 women attended both surveys. Detailed reports on the screenings have been published28, 29 and associations of dietary factors, physical activity and cancer have been found.21, 30, 31, 32, 33
Assessment of lifestyle parameters
The screening procedures were almost identical in the 2 surveys. The written invitation included a questionnaire on ethnicity, chronic diseases, smoking and physical activity. The women marked their usual level of recreational and occupational activity on a scale with the following 4 grades: Recreational activity: Grade 1, Reading, watching television or other sedentary activity; Grade 2, Walking, bicycling or other activity for at least 4 hr per week; Grade 3, Recreational athletics, heavy gardening or similar activities at least 4 hr per week; and Grade 4, Regular (several times a week) training or participation in athletic competitions. Occupational activity: Grade 1, Mostly sedentary work; Grade 2, A lot of walking; Grade 3, A lot of walking and lifting; and Grade 4, Heavy manual work. The same team of trained nurses conducted interviews with the participants at the screening center in both surveys to confirm the information given.
In the second survey, a food frequency questionnaire (FFQ) including questions on 64 explicit food items was also distributed at the screening and returned via mail by 25,892 women (83%) after one reminder. Special emphasis was given to habitual use of fat and fat-rich food items in the diet. The estimated energy intake per time unit was calculated according to the method developed by Gaard and colleagues.21, 34 The partly semi-quantitative FFQ has been described in detail and validated as a useful tool for categorizing individuals according to their intake of energy and fat.35
Assessment of clinical parameters
Body weight was measured to the nearest half kilogram with participants dressed in lightweight clothing. Height was measured in centimeters. As an estimate of relative weight, body mass index (BMI) was obtained by dividing the body weight in kilograms by the height in meters squared (kg/m2). Specially trained nurses measured systolic and diastolic blood pressure by using a mercury sphygmomanometer. After the women had rested for a minimum of 4 min, 2 recordings were made at 1 min intervals with the individual sitting. The lowest blood pressure was used in the analyses. A non-fasting blood sample was drawn and analyzed at the Central Laboratory, Ullevål Hospital, Oslo. Details of the methods have been published.28 Non-fasting serum glucose concentration was estimated according to the method described by Brown36 in every sample in the first survey, while in the second survey this analysis was done on a sub-set only (in Finnmark, part of Oppland).
Follow-up and case identification
We restricted our study to women who were alive with no diagnosis of any malignant disease 1 year after participation in the second survey (1977–1981, n = 25,642). Only women with complete data from both surveys on parity, recreational and occupational activity, and the main clinical parameters [height, weight, blood pressure and non-fasting serum glucose (survey 1)] were included (n = 24,460). As the assessment of energy intake and the double assessments of other analytic variables were completed in the second survey only, we used this survey as the baseline.
Prevalence and incidence of cancer cases were identified through linkage to the Cancer Registry of Norway. Women with an incident, primary, histopathologically confirmed carcinoma of the endometrium were defined as cases. Since 1952, clinicians and pathologists in Norway have been required to report all new cancer cases. An accurate and complete database is achieved through matching with the Register of Deaths at Statistics Norway (SSB). Information about reproductive history, emigration and death was gathered by linkage to SSB. Women were considered to be at risk from the date of the blood sample in the 1977–1981 survey (baseline) through 31 December 1996, or earlier on the occurrence of emigration, death or diagnosis of any cancer. In total, 24,460 women were followed for 384,531 person-years.
To estimate relative risks (RRs) for endometrial cancer associated with high energy intake, overweight, physical inactivity, hypertension and relative hyperglycemia, we used Cox's proportional hazard regression model. The proportional hazard assumption was checked by evaluating the parallelism between graphs of the log-log survivor functions for different categories of the variables. Several models were used to fit, ranging from the simplest model (age at baseline) to the more complex ones, including the analytic variables and co-variates (county of residence, parity, height, alcohol and smoking status), one at a time or simultaneously. Based on substantial modification in RRs derived from inclusion of parity and BMI, and cause-and-effect decisions, we found it appropriate to present the results of equations including only age, as well as those including age, county of residence, parity, height, BMI, physical activity and smoking. Tests for trend were performed by the introduction of “semi-continuous” variables, which were obtained by assigning consecutive integers to levels of categorized variables. Interactions were evaluated by including appropriate product terms in the models.
Women were assigned to the overweight (BMI = 25–30 kg/m2) or the obese class (BMI ≥ 30 kg/m2) according to the World Health Organization (WHO) classification of adults by BMI. With regard to physical activity, the 2 uppermost levels were merged because only a few participants engaged in the most strenuous activities (grade 4 recreational: n = 52, 0.2% of included and grade 4 occupational: n = 1,331, 5.5% of included). Furthermore, for both recreational and occupational activity, a new variable was created containing information from both surveys. Women who reported sedentary activity (grade 1) in both surveys were characterized as “consistently sedentary.” Those who reported minimum grade 2 level of activity in the first survey and minimum grade 3 in the next were characterized as “consistently active.” The remaining women who were neither consistently sedentary nor consistently active were characterized as “moderately active.”
Concerning systolic and diastolic blood pressure, we tested the WHO criteria for hypertension (140 and 90 mmHg, respectively) as relevant thresholds for endometrial cancer risk. Women with blood pressure below the 140/90 mmHg limit in both surveys were classified as “consistently normotensive” and women with blood pressure above the limit in both surveys were classified as “consistently hypertensive.” Women who were neither consistently normotensive nor consistently hypertensive were classified as “hypertensive in one survey.” With regard to non-fasting serum glucose concentration and energy intake, distribution-based cut-off points (i.e., quartiles) were used. Dietary information was insufficient in observations if there was an estimated daily energy intake below 2,100 kJ and/or less than 20 adequate answers in the FFQ (n = 4,370, 17.9% of included); analyses were conducted on energy intake and fat intake after exclusion of these observations. All significance tests were 2-tailed and the level of significance was set at 5%. The analyses were performed with the SAS statistical package version 8e.
During a mean follow-up period of 15.7 years, there were 130 incident cases of endometrial carcinoma [127 adenocarcinomas (1 serous papillary adenocarcinoma = type II-carcinoma) and 3 unspecified carcinomas] diagnosed among 24,460 women. The mean age at diagnosis (X̄age at diagnosis) was 56.8 years (range 40.3–69.3). Women who developed endometrial cancer were somewhat older and had given birth to fewer children than those who did not (Table I). Baseline mean BMI, mean blood pressure and mean energy intake were higher, whereas mean serum cholesterol was lower in cases vs. non-cases. Those who developed cancer were more likely to be sedentary in their leisure time and at work and to report hypertensive disease.
|Total population (n = 24,460)2||Cases1 (n = 130)2|
|Mean (SD)||Mean (SD)|
|Age (years)||44.7 (6.7)||47.2 (4.5)|
|Parity3||2.6 (1.6)||2.3 (1.7)|
|Age at first birth (years)||23.9 (4.3)||23.6 (4.5)|
|Height (cm)||162.7 (6.0)||162.9 (6.2)|
|BMI (kg/m2)||24.9 (4.1)||26.6 (5.8)|
|Diastolic blood pressure (mmHg)||82.3 (10.7)||83.9 (10.4)|
|Systolic blood pressure (mmHg)||131.4 (18.4)||134.1 (19.3)|
|Serum glucose (mmol/l)4||5.7 (1.0)||5.8 (0.8)|
|Serum cholesterol (mmol/l)||6.2 (1.2)||5.9 (1.3)|
|Serum HDL-cholesterol (mmol/l)||1.4 (0.4)||1.4 (0.3)|
|Serum triglycerides (mmol/l)||1.5 (0.9)||1.5 (0.9)|
|Energy intake (kJ/day)||5,374 (1,562)||5,458 (1,520)|
|Fat intake (g/day)||54.9 (19.4)||54.7 (18.1)|
|Proportions (%)||Proportions (%)|
|Physically active in leisure time||82.9||78.5|
|Physically active at work||86.2||80.8|
Age-adjusted survival analyses of selected background variables confirmed a protective effect of parity and serum cholesterol on endometrial cancer risk. In contrast, no association was observed between smoking and endometrial cancer risk (Table II). Furthermore, there was an overall positive association between BMI and risk of endometrial cancer in age-adjusted analysis [RR = 1.39, 95% confidence interval (95%CI) = 1.21–1.59, per 4.1 kg/m2].
|Variable (per change in variable)2||RR (95%CI)||Cases||All|
|Age (1 year)||1.08 (1.04, 1.11)||130||24,460|
|Parity (1 child)3||0.87 (0.77, 0.97)||130||24,460|
|Age at first birth (1 year)||0.98 (0.93, 1.03)||108||22,275|
|Height (6 cm)||1.06 (0.89, 1.27)||130||24,460|
|BMI (4.1 kg/m2)||1.39 (1.21, 1.59)||130||24,460|
|Serum cholesterol (1.2 mmol/l)||0.77 (0.64, 0.93)||130||24,460|
|Serum HDL cholesterol (0.4 mmol/l)||0.83 (0.68, 1.02)||130||24,023|
|Serum triglycerides (0.9 mmol/l)||1.02 (0.86, 1.21)||130||24,458|
|Fat intake (19.4 g/day)||0.99 (0.82, 1.20)||110||20,090|
|Daily smoking (yes/no)||0.78 (0.53, 1.14)||130||24,460|
We observed an almost doubled risk of endometrial cancer among women in the third quartile of energy intake (5,044–6,401 kJ/day) compared to women in the bottom quartile (< 4,266 kJ/day) (RR = 1.87, 95%CI = 1.09–3.23; Table III). In stratified analyses, the association was confined to women below 50 years at baseline (RR = 2.94, 95%CI = 1.31–6.58 for the second; RR = 3.40, 95%CI = 1.52–7.60 for the third and RR = 1.91, 95%CI = 0.78–4.63 for the upper quartile, ptrend = 0.17; pinteraction = 0.05).
|All women||Age < 50 years||Age ≥ 50 years|
|Cases||RR (95%CI)||Cases||RR (95%CI)||Cases||RR (95%CI)|
|Energy intake1 (kJ/day)|
|4,266–5,043||31||1.54 (0.88, 2.68)||23||2.94 (1.31, 6.58)||8||0.64 (0.26, 1.55)|
|5,044–6,401||35||1.87 (1.09, 3.23)||24||3.40 (1.52, 7.60)||11||0.94 (0.42, 2.11)|
|≥ 6,402||23||1.36 (0.75, 2.48)||13||1.91 (0.78, 4.63)||10||1.01 (0.44, 2.34)|
|p for trend||0.21||0.17||0.83|
|BMI, Quartiles2 (kg/m2)|
|22.1–24.0||27||1.22 (0.69, 2.17)||16||1.17 (0.57, 2.41)||11||1.19 (0.46, 3.07)|
|24.1–26.9||29||1.24 (0.70, 2.19)||18||1.34 (0.66, 2.72)||11||0.95 (0.36, 2.47)|
|≥ 26.9||53||2.26 (1.33, 3.82)||34||2.80 (1.47, 5.34)||19||1.41 (0.58, 3.43)|
|p for trend||0.002||0.0008||0.49|
|BMI, WHO criteria2 (kg/m2)|
|25–30||45||1.43 (0.96, 2.14)||32||1.90 (1.17, 3.09)||13||0.82 (0.41, 1.65)|
|≥ 30||29||2.57 (1.61, 4.10)||15||2.59 (1.39, 4.83)||14||2.28 (1.13, 4.62)|
|p for trend||0.0001||0.0008||0.06|
From the bottom (< 22.1 kg/m2) to the top BMI-quartile (≥ 26.9 kg/m2), the endometrial cancer risk increased by 126% (RR = 2.26, 95%CI 1.33–3.82, ptrend = 0.002, after adjustments; Table III). However, there was only a slight suggested increase in risk associated with BMI between 22.1 and 26.9 kg/m2.To evaluate the possible modification of effect by changes in endocrine exposure around the menopause, we split the population by age; among women younger than 50 the heaviest quartile (≥ 26.9 kg/m2) experienced almost 3 times the risk of the leanest (RR = 2.80, 95%CI = 1.47–5.34, ptrend = 0.0008). In contrast, among the older women the effect disappeared. However, a formal test of interaction was not significant (pinteraction = 0.42). In women younger than 50 with a BMI in the upper quartile, X̄age at diagnosis was 55.3 years. As ethnicity may be a determinant of BMI, we repeated the analyses excluding the Lapp women (n = 1,243) but observed only minimal alterations in the risk estimates (results not shown).
We tested the WHO criteria for classification of adults by BMI. Data were consistent, with an increased risk among obese women (BMI ≥ 30 kg/m2: RR = 2.57, 95%CI = 1.61–4.10), and suggestive of an increased risk among overweight women (BMI = 25–30 kg/m2: RR = 1.43, 95%CI = 0.96–2.14, ptrend = 0.0001) vs. normal weight or lean women (Table III). However, among women younger than 50 at baseline, overweight was evidently associated with an increased risk (RR = 1.90, 95%CI = 1.17–3.09), and obesity conferred a higher risk than in older women (< 50 years: RR = 2.59, 95%CI = 1.39–4.83 vs. ≥ 50 years: RR = 2.28, 95%CI = 1.13–4.62; pinteraction = 0.12). In women younger than 50, X̄age at diagnosis was 54.7 years in overweight and 56.7 years in obese individuals.
Recreational activity that corresponded to a minimum of 4 hr of walking per week (grades 2–4) at baseline or in both surveys tended to be protective against endometrial cancer even in multivariate analysis including BMI (Table IV). This was further supported by a significant 37% reduction in risk among women in the middle category of recreational activity (grade 2) in the 1974–1976 survey (RR = 0.63, 95%CI = 0.43–0.92, ptrend = 0.12; results not shown). We performed separate analyses for women aged younger and those older than 50 at baseline and data suggested that the oldest women profited most from having an active lifestyle in both surveys (≥ 50 years: ptrend = 0.14; pinteraction = 0.30; results not shown). Among consistently recreationally active women, X̄age at diagnosis was 54.0 years in those younger than 50 and 60.4 years in older women. Stratification by BMI-categories did not reveal any further associations between recreational activity and endometrial cancer risk.
|Cases||RR1 (95%CI)||RR2 (95%CI)|
|Grade 1 (lowest)||28||1.00||1.00|
|Grade 2||85||0.77 (0.50, 1.17)||0.81 (0.53, 1.25)|
|Grade 3 + 4 (highest)||17||0.69 (0.38, 1.27)||0.79 (0.43, 1.45)|
|p for trend||0.20||0.39|
|Moderately active||101||0.67 (0.38, 1.17)||0.72 (0.41, 1.27)|
|Consistently active||15||0.61 (0.30, 1.27)||0.71 (0.34, 1.49)|
|p for trend||0.23||0.40|
|Grade 1 (lowest)||24||1.00||1.00|
|Grade 2||76||0.68 (0.43, 1.07)||0.70 (0.44, 1.11)|
|Grade 3 + 4 (highest)||30||0.59 (0.34, 1.00)||0.61 (0.35, 1.05)|
|p for trend||0.07||0.09|
|Moderately active||85||0.55 (0.33, 0.95)||0.57 (0.33, 0.99)|
|Consistently active||29||0.47 (0.26, 0.87)||0.49 (0.26, 0.91)|
|p for trend||0.04||0.06|
Increasing occupational activity was associated with a reduced risk of endometrial cancer (Table IV). Among women who reported having a job with a lot of walking at baseline, the RR for endometrial cancer was 0.70 (95%CI = 0.44–1.11) and, among those who reported lifting or heavy manual work, RR was 0.61 (95%CI = 0.35–1.05), compared to those who had sedentary work (ptrend = 0.09). By combining the assessments of occupational activity from the 2 surveys, we observed that moderately active and consistently active women had a 43% reduction (RR = 0.57, 95%CI = 0.33–0.99) and 51% reduction (RR = 0.49, 95%CI = 0.26–0.91) in the risk of endometrial cancer, respectively, compared to consistently sedentary women (ptrend = 0.06). Physical activity at work was especially protective in obese women (BMI ≥ 30 kg/m2). In stratified analyses of occupational activity at baseline, and in both surveys combined, the RR decreased to 0.22 (95%CI = 0.08–0.66, ptrend = 0.01; pinteraction = 0.07) and 0.18 (95%CI = 0.05–0.62, ptrend = 0.03; pinteraction = 0.17; results not shown), respectively, as the level of activity increased among obese women. In separate analyses of the effect of occupational activity in women younger and older than 50 at baseline, active work, in both surveys, tended to be most protective among the youngest women (age < 50 years, consistently active vs. consistently sedentary: RR = 0.44, 95%CI = 0.20–0.98, ptrend = 0.07; pinteraction = 0.90; results not shown). Among consistently occupationally active women, X̄age at diagnosis was 54.1 years in those younger than 50 and 61.9 years in older women.
Positive associations between measured blood pressure at baseline in a continuous scale and endometrial cancer risk were suggested in an age-adjusted analysis (RR = 1.14, 95%CI = 0.97–1.34 per 18.4 mmHg for systolic blood pressure and RR = 1.17, 95%CI = 0.98–1.39 per 10.7 mmHg for diastolic blood pressure; results not shown), but in models including BMI the associations disappeared. In analyses using quartiles of systolic and diastolic blood pressure, no associations with endometrial cancer risk were observed (results not shown).
We tested the relevance of the WHO criteria for hypertension (140/90 mmHg) with respect to endometrial cancer risk (Table V). Among obese women (BMI ≥ 30 kg/m2) hypertension in both surveys was associated with a 3.5 times increase in risk of endometrial cancer compared to normotension in both surveys (RR = 3.47, 95%CI = 1.24–9.70, ptrend = 0.02; pinteraction = 0.04). Furthermore, the association was not distorted by residual confounding in analysis adjusted for BMI (continuous term; RR = 3.06, 95%CI = 1.08–8.63; results not shown). Even though we observed 13 incident cases of endometrial cancer among women who reported being treated for high blood pressure at baseline (n = 1,594), hypertensive disease was not associated with risk in survival analysis. Moreover, when excluding women with self-reported hypertensive disease, the estimated RR associated with measured hypertension in both surveys increased to 5.25 (95%CI = 1.46–18.95, ptrend = 0.01) among obese women (results not shown).
|Combined level of arterial blood pressure||All women||BMI < 30 kg/m2||BMI ≥ 30 kg/m2|
|Cases||RR1 (95%CI)||Cases||RR2 (95%CI)||Cases||RR2 (95%CI)|
|Hypertensive in one survey4||25||1.11 (0.70, 1.77)||20||1.42 (0.86, 2.34)||5||0.77 (0.26, 2.27)|
|Consistently hypertensive5||20||1.24 (0.69, 2.25)||6||0.61 (0.24, 1.51)||14||3.47 (1.24, 9.70)|
|p for trend||0.24||0.68||0.02|
Table VI shows that the RR of endometrial cancer increased to 2.12 (95%CI = 1.21–3.71) in the third quartile of non-fasting serum glucose compared to the bottom quartile (ptrend = 0.01). The risk associated with a non-fasting serum glucose ≥ 5.6 mmol/l was even higher in women who were overweight (BMI ≥ 25 kg/m2; RR = 2.45, 95%CI = 1.11–5.42). In contrast, the effect of high serum glucose values was almost absent among women who had a BMI < 25 kg/m2 (pinteraction = 0.92). We did not observe any association between endometrial cancer risk and the continuous non-fasting serum glucose variable (results not shown). None of the women who reported having diabetes at baseline (n = 154) was diagnosed with endometrial cancer during follow-up.
|Serum glucose (mmol/l)||All women||BMI < 25 kg/m2||BMI ≥ 25 kg/m2|
|Cases||RR1 (95%CI)||Cases||RR2 (95%CI)||Cases||RR2 (95%CI)|
|5.2–5.59||27||1.43 (0.78, 2.63)||12||1.36 (0.57, 3.24)||15||1.60 (0.68, 3.78)|
|5.6–6.09||45||2.12 (1.21, 3.71)||19||1.96 (0.89, 4.35)||26||2.45 (1.11, 5.42)|
|≥ 6.1||41||1.88 (1.07, 3.33)||16||1.55 (0.68, 3.51)||25||2.41 (1.08, 5.37)|
|p for trend||0.01||0.22||0.02|
The presence of a cluster of metabolic abnormalities, including self-reported or measured hypertension (diastolic blood pressure ≥ 90 mmHg or systolic blood pressure ≥ 140 mmHg), self-reported diabetes or level of non-fasting serum glucose in the upper quartile (≥ 6.1 mmol/l), level of triglycerides in the upper quartile (≥ 1.76 mmol/l) and level of high-density lipoprotein (HDL) in the lower quartile (< 1.20 mmol/l), was identified in 517 women at baseline. During follow-up, 6 of these women were diagnosed with endometrial cancer and in age-adjusted survival analysis the cluster was associated with a 1.9 times increase in risk (RR = 1.90, 95%CI = 0.84–4.34, results not shown).
Our study of 24,460 women followed for almost 16 years strongly suggests that a cluster of metabolic abnormalities, including hypertension and hyperglycemia, is a significant risk factor for the development of endometrial carcinoma, especially among overweight and obese women. Our results, from models that include energy expenditure (physical activity), body mass and energy intake, suggest that excessive energy intake increases the risk of this malignancy. Furthermore, to our knowledge, our study is the first prospective study to show the protective effect against endometrial cancer of both recreational and occupational activity controlling for measured height and weight (BMI) and other important risk factors.
Our results suggest that relatively high energy intake might be a co-factor in endometrial carcinogenesis, which is in agreement with several case-control studies published in the 1990s.5, 6, 7 In contrast, recent case-control studies did not observe any association with total energy intake.8, 9, 10 In this prospective study, we were able to study simultaneously the effect of energy intake and energy expenditure (physical activity), which is a major advantage considering the laws of thermodynamics. Observations on total energy intake and endometrial cancer from other cohort studies are, however, scarce.
The observed doubling of risk associated with the upper BMI categories was of the same magnitude as in other studies.11 Interestingly, the WHO categories for classification of adults by BMI strongly predicted the risk of developing endometrial cancer. This supports the role of obesity in endometrial carcinogenesis; however a higher number of cases in the reference groups might have facilitated the detection of other significant thresholds (i.e., quartiles). There was a higher proportion of current smokers among non-obese than among obese women (36.4% vs. 25.8%, age-adjusted) and smoking has been inversely related to the risk of endometrial cancer.1, 37 However, the effect of BMI was only minimally altered in the analysis stratified by smoking status (results not shown). The weakening of the associations observed in women older than 50 at baseline vs. younger women may reflect a physiological increase in body weight in post-menopausal women. A tendency among lean women towards more frequent use of hormone replacement therapy,38 which was not allowed for in our analyses, may also have influenced our results.
The suggested slight reduction in the risk of endometrial cancer observed among women with sustained recreational activity is consistent with some observations,13, 14, 20 but in contrast to others,19 yet stronger associations have been reported.15, 16, 18 In our study, there was a clustering of cases in the second level of the 4-point relative measurement scale of recreational physical activity (minimum 4 hr of walking per week). Non-differential misclassification between the second and third levels of activity is likely, and this might have contributed to the weak RR estimates. Furthermore, since ever use of exogenous hormones has been associated with higher social class38 and more time spent in sport activities,39 eventual adjustments for hormonal therapy or social class might have strengthened our results.
Our finding of a 40–50% protective effect from increasing levels of occupational activity is in agreement with that observed in a Swedish cohort.40 However, we found that occupational activity tended to be most protective among younger women (mean age at baseline = 42.0 years, Xage at diagnosis = 54.6 years), whereas Moradi et al.40 found that the effect of this activity was limited to women aged 50–69 years in analyses adjusted for attained age, calendar year of follow-up and place of residence. As we observed that women who reported sedentary jobs at baseline had fewer children, a lower BMI and were more sedentary in their leisure time, relative to those who were physically active at work (p < 0.05, age adjusted; results not shown), confounding may partly explain the discrepancy between studies. A number of retrospective studies13, 14, 16, 19, 41 support an inverse relationship between occupational activity and risk of endometrial cancer, although in the case-control study by Moradi et al.,16 the effect was confined to women who were not obese and smokers. In our study, the effect modification by weight was in favor of the obese women (BMI ≥ 30 kg/m2) and the protection experienced by sustained occupational activity tended to be confined to non-smokers (results not shown). Only in 1 case-control study, which was hampered by a low participation rate and limited sample size, was occupational activity not related to the risk of endometrial cancer.20
An innovation in the study of hypertensive disease and endometrial cancer is our equation, which includes information on 2 repeated measurements of blood pressure, measured height and weight (BMI), and other co-variates (county of residence, parity, smoking and physical activity) used to describe variation in endometrial cancer incidence during follow-up. The increase in risk by more than 3 times, observed among obese women with blood pressure ≥ 140/90 mmHg in 2 assessments at an interval of 3–5 years, is convincing and, similarly, Weiderpass et al.42 found that self-reported hypertension was a risk factor only among obese women.
To our knowledge, ours is the first prospective study of the relationship between a marker of glucose metabolism and risk of endometrial cancer adjusted for BMI, physical activity and other potential co-variates. We found an overall positive association between non-fasting serum glucose concentration and risk of endometrial cancer, which was confined to overweight women (BMI ≥ 25 kg/m2). In a case-control study, mean fasting levels of glucose and insulin were higher in endometrial cancer patients than in healthy controls.43 In another small study, the cancer patients (n = 10) had significantly higher fasting glucose and insulin levels and higher insulin responses after glucose administration than the normal women (n = 10), even though their glucose responses were similar.44 However, Troisi et al.45 found that adjustments for BMI and other risk factors eliminated an apparent positive association between fasting C-peptide levels and risk of endometrial cancer in a retrospective study. In our study, only non-fasting serum glucose levels were available. Nevertheless, the fact that adjustment for time since last meal did not affect the risk estimates strengthens our data (results not shown). The lack of any effect of serum glucose on the endometrial cancer incidence among normal weight women in our cohort is in agreement with a prospective study in which an increased risk of endometrial cancer associated with self-reported diabetes was confined to women with BMI ≥ 27.4 kg/m2.46 Likewise, most retrospective studies of this relationship have reported an increased risk among overweight (BMI ≥ 25 kg/m2)47 or obese (BMI ≥ 30 kg/m2, BMI ≥ 31.9 kg/m2)17, 48 women, while Weiderpass et al.42 saw no difference in the effect of diabetes when comparing obese women with non-obese. In our study, it is likely that the majority of women who reported to have diabetes at baseline were type 1 diabetics and thus the absence of any association with endometrial cancer risk is in agreement with a suggested increase in risk confined to type 2 diabetes.47
In light of the observed effects of overweight/obesity, hypertension and relative hyperglycemia, we suggest that a metabolic syndrome with insulin resistance as a major feature is a risk factor for endometrial cancer as already hypothesized by Weiderpass et al.42 In order to address this question in further detail, we analyzed the effect of a cluster of metabolic abnormalities (e.g., high blood pressure, low levels of HDL and high levels of triglycerides and blood glucose) linked to insulin resistance26 and newly introduced as a predictor of colorectal cancer mortality.27 We found a suggested positive association in age-adjusted analysis between this cluster of metabolic abnormalities and the risk of endometrial cancer.
The mechanism of increased endometrial cancer risk linked to obesity and inactivity is unclear, but it is biologically plausible that changes in metabolism and hormonal activity may be implicated.49 The normal menstrual cycle reflects the refined balance between the proliferative actions of estrogen and the antiestrogenic and secretory transforming actions of progesterone on the endometrium.49 In our study, obesity conferred a higher risk of endometrial cancer in women younger than 50 years than in older women. This points to the importance of the progesterone insufficiency experienced by obese premenopausal women as a result of an increased number of anovulatory cycles23, 49 in the etiology of endometrial cancer. The smaller increase in risk observed among obese women aged 50 years or more might be the result of increased circulating levels of free estradiol associated with increased conversion of androgens to estrogens in the adipose tissue and lower levels of sex hormone-binding globulin (SHBG) in obese postmenopausal women.23, 49
A proposed mechanism for the protective effect on risk of endometrial cancer by exercise is lowered body mass.19 However, we found a significant effect of physical activity even after adjustment for BMI, and inactive women have been shown to have higher serum estrogen levels than active women independent of body weight.50, 51 Experimental and epidemiological evidence suggest that insulin-like growth factor (IGF)-I, a hormone with strong mitogenic and antiapoptotic actions, is important in endometrial carcinogenesis. While BMI and physical activity also appear to affect blood level of IGF-I, dietary energy intake seems to be the critical regulator even in obese subjects.52 Thus, the observed increased risk associated with relatively high energy intake even after adjustments (BMI and physical activity) in our study may reflect an increased proliferative action of IGF-I in response to excessive energy intake. In the case of negative energy balance (low energy intake relative to energy expenditure), which may be reflected in a low BMI, the body will experience energy restriction, which has been associated with decreased level of IGF-I.52 Besides, energy restriction is known to enhance DNA repair, modulate oxidative damage to DNA and reduce oncogene expression.53 Moreover, energy restriction also affects insulin metabolism,53 while inactivity and obesity are associated with insulin resistance and the resultant hyperinsulinemia54 and there is an established link between diabetes, a disease characterized by hyperinsulinemia when diagnosed as type 2 diabetes, and endometrial cancer.1 There is increasing evidence that insulin is a growth factor for tumor formation.24, 25 The mechanisms underlying insulin-mediated neoplasia may include enhanced DNA synthesis with resultant tumor cell growth, inhibition of apoptosis and an altered sex hormone milieu.54 In our study, high blood glucose levels and hypertension, which are associated with insulin resistance/hyperinsulinemia, were strong markers of endometrial cancer risk, in particular, in combination with high body weight. This suggests that in overweight and obese women, coexisting metabolic disturbances may act synergistically and facilitate malignant transformation of the endometrial glandular cells. In light of this, the observed protective effect from relative hypercholesterolemia, which is also generally associated with obesity, is interestingly odd and, to our knowledge, there is at present no well-founded hypothesis about the molecular events, which may explain this suggested antagonistic effect on endometrial carcinogenesis.
The high attendance rate in the population-based screenings, the almost complete follow-up for almost 16 years, the measurements at 2 different time points of height, weight and blood pressure by a constant team of trained nurses, the repeated assessments of self-reported recreational and occupational activity, the dietary data and the definition of cases based on a diagnosis of endometrial carcinoma by a pathologist strengthen our results and are clear advantages over previous studies published in this field. Moreover, the accuracy of the levels of recreational activity reported on the questionnaire has been validated and the level of recreational activity correlates positively with physical fitness.55, 56, 57 The findings in our study that women who were sedentary in their leisure time had a higher BMI and a significant dyslipidemia relative to those who were active (p < 0.05 for higher BMI, serum LDL/HDL ratio and serum triglycerides in inactive vs. active, age adjusted; results not shown) strengthen the validity of our assessment.
The wide confidence intervals of our risk estimates are the result of the limited number of events available for analysis. Additionally, information about actual hysterectomy rates in the population was not accessible. However, the prevalence of extirpated corpora uteri among women aged 15 years or more in Norway in the 3-year period 1988–1990 was estimated to be 198 per 100,000 women58 and hysterectomy is unlikely to be a real confounding factor in our study. Norwegian women were restrictive in their use of estrogens for peri- and post-menopausal symptoms (HRT) until the 1990s, when the rate of prescription started to increase, with combined estrogen-progestin hormone replacement being most frequently prescribed.38 Likewise, use of oral contraceptives has been limited in Norway compared to neighboring countries, and the low-dose estrogen pills have been the predominant preparation since 1975,59 making the missing information on exogenous hormone use of minor importance in this context.
In conclusion, our study advocates public health efforts aimed at reducing physical inactivity and excessive energy intake in order primarily to reduce the prevalence of obesity and metabolic syndrome and to have an additional effect on endometrial cancer morbidity. Furthermore, we find it reasonable that women with high body weight might benefit from regular measurements of serum glucose and blood pressure, and eventual treatment of established diabetes and hypertension, with respect to endometrial cancer morbidity through calorie restriction and exercise, among other standard measures.
- 11Cancer-preventive effects. In: IARC Handbooks of cancer prevention, vol. 6: weight control and physical activity. Lyon: IARC Press, 2002. p 83–206.
- 12Assessment of human energy expenditure. In: BouchardC. Physical activity and obesity. Champaign: Human Kinetics, 2000. p 103–31., , .
- 29National Health Screening Service. The cardiovascular disease study in Norwegian counties: results from second screening. Oslo: National Health Screening Service, 1988. 264p.
- 34Calculation of nutrient intake based on dietary data from 2nd cardiovascular disease survey in Finnmark, Sogn og Fjordane, and Oppland 1977–83. Technical report 1. Oslo: The Cancer Registry of Norway, 1994., , .
- 35The reproducibility of a self-administered diet questionnaire. Vår Føda 1987; Suppl 1: 33–7., .
- 59Comments on the use of contraceptives. In: Legemiddelforbruket i Norge 1981–1985 [Drug consumption in Norway 1981–1985]. Oslo: Norsk Medisinaldepot [Norwegian Medicinal Depot]. 1986: 121–2..