Carbohydrate intake, glycemic index and glycemic load in relation to risk of endometrial cancer: A prospective study of Swedish women

Abstract

The associations of carbohydrate intake, glycemic index and glycemic load with endometrial cancer risk were examined among 61,226 participants of the Swedish Mammography Cohort who were cancer-free at enrollment between 1987 and 1990 and completed a food frequency questionnaire. During a mean follow-up of 15.6 years, through June 2005, 608 incident cases of endometrial adenocarcinoma were diagnosed. We observed no overall association between carbohydrate intake, glycemic index or glycemic load and incidence of endometrial cancer; the rate ratios (RRs) for the highest versus the lowest quintile were 1.12 (95% CI, 0.85–1.47) for carbohydrate intake, 1.00 (95% CI, 0.77–1.30) for glycemic index and 1.15 (95% CI, 0.88–1.51) for glycemic load. However, among obese women (body mass index, BMI ≥30 kg/m²), endometrial cancer incidence was nonsignificantly elevated in the top versus bottom quintiles of carbohydrate intake (RR, 1.68; 95% CI, 0.86–3.29) and glycemic load (RR, 1.57; 95% CI, 0.82–2.99). In a subanalysis of women who completed a follow-up questionnaire in 1997, which collected information on physical activity, carbohydrate intake and glycemic load were positively related to endometrial cancer risk among overweight women (BMI ≥25 kg/m²) with low physical activity. In this subgroup, the multivariate RRs comparing extreme quartiles were 1.90 (95% CI, 0.84–4.31) for carbohydrate intake and 2.99 (95% CI, 1.17–7.67) for glycemic load. Results from this cohort study suggest that a high carbohydrate intake and a high glycemic load may increase the risk of endometrial cancer among overweight women with low physical activity. © 2006 Wiley-Liss, Inc.

Evidence is accumulating that insulin resistance and hyperinsulinemia are involved in the etiology of endometrial cancer.1 Obesity, physical inactivity and Type 2 diabetes mellitus are all associated with insulin resistance, hyperinsulinemia and endometrial cancer.2, 3, 4, 5, 6 Moreover, epidemiologic studies have reported increased risk of endometrial cancer in women with high glycosylated hemoglobin (a measure of average blood glucose level over time),7 high prediagnostic serum C-peptide (a marker of insulin production)8 and low circulating adiponectin levels (an indicator of insulin resistance).9, 10, 11

Dietary intake can influence insulin levels, particularly among individuals who are insulin resistant due to other factors, such as obesity and physical inactivity. The concept of glycemic index was developed to quantify the glycemic responses induced by carbohydrates in different foods.12, 13 Dietary glycemic load, defined as the product of the glycemic index of a particular food and its carbohydrate content, was introduced to derive an overall estimate of postprandial glycemia and insulin demand.14 Diets with high glycemic index and load have been related to higher glycosylated hemoglobin,15 higher circulating and urinary C-peptide levels,15, 16 and lower adiponectin levels17, 18 in diabetic and nondiabetic persons. Furthermore, a randomized controlled trial showed that a low glycemic index diet increases serum sex hormone-binding globulin levels and reduces estrogen and testosterone levels.19

To date, only 1 case–control20 and 2 prospective cohort studies21, 22 have provided findings on glycemic index and glycemic load in relation to risk of endometrial cancer. The case–control study20 found a significant 1.9-fold increased risk of endometrial cancer for the highest versus the lowest quintile of glycemic index. In prospective studies, only a moderate increased risk of endometrial cancer associated with a high glycemic load was observed in the Canadian National Breast Screening Study,22 and no overall associations with either glycemic index or glycemic load were reported in the Iowa Women's Health Study.21 However, in both studies, a statistically significant 1.9-fold increased risk of endometrial cancer was found for the highest versus the lowest quintile of glycemic load among obese women.21, 22

Because of scarcity and inconsistent epidemiological data, we examined prospectively the associations of glycemic index and glycemic load as well as carbohydrate intake with the risk of endometrial cancer in a large population-based cohort of Swedish women. We hypothesize that a high carbohydrate intake and diets with high glycemic index and load increase the risk of endometrial cancer primarily in overweight and physically inactive women who are likely to have a greater insulin response to their diet compared with lean and active women.

Material and methods

Study cohort

The Swedish Mammography Cohort was established between 1987 and 1990 when all women who were born between 1914 and 1948 and resident in Uppsala and Västmanland counties in central Sweden received a mailed invitation to be screened by mammography. Enclosed with this invitation was a 6-page questionnaire regarding diet and other exposures; a completed questionnaire was obtained from 66,651 women, representing 74% of the source population. In the autumn of 1997, all 56,030 participants who were still alive and residing in the study area received a second questionnaire; 39,227 women (70%) answered this questionnaire. The study was approved by the ethical committees at the Uppsala University Hospital (Uppsala, Sweden) and the Karolinska Institutet (Stockholm, Sweden).

Assessment of diet

A food-frequency questionnaire with 67 and 96 food items was used to assess diet at baseline and in 1997, respectively. In these questionnaires, women were asked to report how often, on average, they had consumed each food item over the previous year. The questionnaires had 8 mutually exclusive predefined categories for frequency of consumption. Carbohydrate intake was calculated by multiplying the frequency of consumption of each food by its carbohydrate content per age-specific serving, using food composition values obtained from the Swedish National Food Administration Database.23 The age-specific serving sizes were based on mean values obtained from 213 randomly selected women from the study area who weighed and recorded their food intake for a mean of 27.8 days. We computed the average glycemic load over the previous year for each participant by multiplying the carbohydrate content (grams per serving) of each food by its glycemic index; multiplying that product by the frequency of consumption (servings of that food per day), and adding the values from all foods. Glycemic load thus represents the quality and quantity of carbohydrates and the interaction between the two (the product of carbohydrate intake and glycemic index indicates that a higher glycemic index has a greater effect at a higher intake of carbohydrates). Each unit of glycemic load represents the equivalent of 1 g carbohydrate from white bread. Glycemic index values of foods were obtained from published tables.24 The overall glycemic index for each woman was calculated by dividing the glycemic load by total carbohydrates intake; this value represents the overall quality of carbohydrate intake for each woman. Carbohydrate intake, glycemic index and glycemic load were energy-adjusted by using the residual method.25

In a validity study of 129 women randomly chosen from the study cohort, the correlation coefficient between the baseline food-frequency questionnaire and the mean intakes assessed by four 1-week weighted diet records (3–4 months apart) was 0.53 for carbohydrate intake (Wolk, unpublished data). High validity was observed for carbohydrate intake as assessed by the 1997 food frequency questionnaire (correlation coefficient, 0.73).26

Assessment of nondietary exposures

Nondietary exposure data obtained at baseline included education, weight, height, menstrual and reproductive history and use of oral contraceptives and postmenopausal hormones. We calculated body mass index (BMI) as weight in kilograms divided by the square of height in meters. On the second questionnaire, participants also provided information on history of diabetes, smoking and physical activity. For physical activity, women were asked to indicate how many hours per week, on average, during the previous year they engaged in leisure time physical activity. Five possible answers were available, ranging from less than 1 hr per week to more than 5 hr per week. In this analysis, women who reported 2 or more hours of leisure time physical activity per week were defined as physically active, and those who engaged in physical activity 1 or fewer hours per week were defined as physically inactive.

Case ascertainment and follow-up

We ascertained incident cases of endometrial cancer by computerized record linkages of the study population with the national and regional Swedish Cancer registers, both of which have been estimated to be almost 100% complete.27 Only primary invasive adenocarcinomas of the endometrium were included as cases in these analyses. Dates of death and dates of migration were ascertained by linkages to the Swedish Death and Population registers. By linkage to the Swedish Inpatient Register, we obtained information on dates of hysterectomy.

Population for analysis

From the baseline cohort of 66,651 women, we excluded those with an incorrect or a missing national registration number and those lacking date on the questionnaire, date of moving out of the study area or date of death. After further exclusion of women with implausible values for total energy intake (i.e., 3 standard deviations from the mean value for log_e-transformed energy intake) and women with a cancer diagnosis (other than nonmelanoma skin cancer) or who had undergone a hysterectomy before baseline, the study cohort for our main analyses included 61,226 women. For those analyses using data from the second questionnaire, 36,369 women were included after excluding those with implausible energy intake on the second dietary questionnaire and those who had been diagnosed with cancer or had undergone a hysterectomy between baseline and January 1998.

Statistical analysis

The follow-up time for each woman was counted from the date of enrollment (or January 1, 1998 for subanalyses using the second questionnaire) to the date of diagnosis of endometrial cancer, the date of death, the date of hysterectomy, the date of migration or June 30, 2005. Participants were categorized into quintiles of dietary carbohydrate intake, glycemic index and glycemic load. We used Cox proportional hazards models28 with follow-up time as the time scale to estimate rate ratios (RRs) and 95% confidence intervals (CIs) for the association of energy-adjusted carbohydrate intake, glycemic load and glycemic index with endometrial cancer risk. To control as finely as possible for age and calendar time, and possible two-way interactions between these 2 time scales, we stratified the models by age in months at start of follow-up and the year of enrollment. We created multivariate models by stepwise adding covariates to the models. Variables were considered confounders and included in the model if they changed the risk estimate by more than 5%. Potential risk factors for endometrial cancer considered as possible confounders included education, BMI, age at menarche, oral contraceptive use, age at first birth, parity, age at menopause, postmenopausal hormone use and menopausal status. In subanalyses using data from the second questionnaire, we also considered adjustment for physical activity, diabetes, and smoking. Because inclusion of these variables in the models altered the risk estimates by less than 5% they were not included in the main model.

To test for trend, we assigned the median value to each exposure category and treated this value as a continuous variable in the model. Because high BMI and physical inactivity can be strong determinants of insulin resistance, we hypothesized that these factors could modify the relations of carbohydrate intake, glycemic index and glycemic load with endometrial cancer, and we evaluated this hypothesis in stratified analyses. Tests for interaction were conducted using likelihood ratio tests, comparing models with and without product terms representing the variables of interest. All analyses were carried out using SAS statistical software, version 9.1 (SAS Institute Inc., Cary, NC). All p-values are 2-sided.

Results

Baseline characteristics of the study population according to glycemic load are presented in Table I. Compared with women with a low glycemic load, those with higher glycemic load were, on average, older and were less likely to have a postsecondary education and to have ever smoked. Other baseline characteristics did not vary appreciably across glycemic load quintiles.

Table I. Age-Standardized Characteristics of 61,226 Women in the Swedish Mammography Cohort According to Quintiles of Glycemic Load at Baseline (1987–1990)1

Characteristics	Glycemic load quintiles
	<164	164–176	177–186	187–199	≥200
1 All values are means if not otherwise indicated. 2 Information from the second questionnaire. 3 Physically active defined as ≥2 hr/week of leisure time physical activity. 4 Parous women only.
Age at baseline (years)	50.7	52.0	53.6	55.1	57.1
Body mass index (kg/m^2)	24.7	24.7	24.8	24.8	24.8
Postsecondary education (%)	14.1	13.5	12.6	11.7	10.9
History of diabetes (%)2	4.7	4.1	4.5	4.6	4.6
Ever smokers (%)2	53.8	47.5	44.2	42.4	44.5
Physically active (%)2,3	55.3	56.1	57.1	58.2	56.2
Parous (%)	89.1	90.1	89.9	89.0	87.6
Number of children4	2.4	2.4	2.4	2.4	2.4
Age at first birth (years)4	24.0	24.1	24.2	23.9	23.9
Age at menarche (years)	13.2	13.2	13.2	13.2	13.2
Age at menopause (years)	50.7	50.7	50.7	50.7	50.7
Oral contraceptive use (%)	55.2	55.1	53.8	52.2	52.7
Postmenopausal hormone use (%)	45.1	45.4	43.6	44.2	43.1
Total energy intake (kcal/day)	1,547	1,599	1,605	1,608	1,555

During 952,629 person-years of follow-up (mean, 15.6 years), from baseline (1987–1990) through June 2005, 608 incident cases of endometrial adenocarcinoma were diagnosed. In the whole cohort, there were no associations of carbohydrate intake, glycemic index or glycemic load with endometrial cancer risk (Table II). Adjustment for known and potential risk factors for endometrial cancer did not change the results essentially. For example, the RR of endometrial cancer for the highest versus the lowest quintile of glycemic load was 1.13 (95% CI, 0.86–1.48) after controlling for age, education, BMI, age at menarche, oral contraceptive use, parity, age at menopause and postmenopausal hormone use.

Table II. Rate Ratios (RRS) and 95% CIS of Endometrial Cancer According to Quartiles of Carbohydrate Intake, Glycemic Index, and Glycemic Load Among 61,226 Women in the Swedish Mammography Cohort (1987–2005)

	Quintile					ptrend
	1	2	3	4	5
^a Cox proportional hazards models were used to calculate rate ratios adjusted for age in months. Carbohydrate intake, glycemic index and glycemic load were adjusted for total energy intake using the residual method.25
Carbohydrate intake (g/day)
Range (median)	<211 (201)	211–223 (218)	223–233 (239)	234–245 (240)	≥246 (256)
No. of cases	96	124	112	142	134
Person-years	191,736	193,109	190,759	189,306	187,719
Rate ratios (95% CI)	1.00	1.19 (0.90–1.56)	1.03 (0.78–1.36)	1.24 (0.95–1.61)	1.12 (0.85–1.47)	0.42
Glycemic index
Range (median)	<75.7 (73.9)	75.8–78.3 (77.2)	78.4–80.6 (79.6)	80.7–83.3 (81.9)	≥84.4 (85.5)
No. of cases	110	130	126	119	123
Person-years	190,283	196,555	190,717	186,075	188,999
Rate ratios (95% CI)	1.00	1.09 (0.84–1.41)	1.06 (0.81–1.37)	1.01 (0.78–1.32)	1.00 (0.77–1.30)	0.79
Glycemic load
Range (median)	<164 (155)	164–176 (170)	177–186 (181)	187–199 (193)	≥200 (210)
No. of cases	100	123	115	126	144
Person-years	191,609	193,398	190,433	190,368	186,821
Rate ratios (95% CI)	1.00	1.14 (0.87–1.50)	1.03 (0.78–1.36)	1.09 (0.83–1.42)	1.15 (0.88–1.51)	0.41

When we stratified the analyses by BMI, we observed nonsignificant positive associations of carbohydrate intake and glycemic load with endometrial cancer risk among obese women (BMI ≥30 kg/m²) but not among leaner women (Table III). Among obese women, those in the highest quintile of carbohydrate intake or glycemic load had an ∼70 and 60%, respectively, statistically nonsignificant elevated risk of endometrial cancer compared to those in the lowest quintile. Tests for interaction between BMI and carbohydrate intake or glycemic load in relation to risk of endometrial cancer were not statistically significant (p-interaction = 0.42 for carbohydrate intake; p-interaction = 0.11 for glycemic load).

Table III. Rate Ratios and 95% CIS of Endometrial Cancer According to Quartiles of Carbohydrate Intake, Glycemic Index, and Glycemic Load, Stratified by Body Mass Index Among 61,226 Women in the Swedish Mammography Cohort (1987–2005)1

Body mass index	Quintile					ptrend
	1	2	3	4	5
1 Cox proportional hazards models were used to calculate rate ratios adjusted for age in months. Carbohydrate intake, glycemic index and glycemic load were adjusted for total energy intake using the residual method.25 The number of cases does not add up to the total number of cases (n = 608) owing to missing data on body mass index.
<25 kg/m^2 (n = 243 cases)
Carbohydrate intake	1.00	1.06 (0.70–1.61)	0.89 (0.58–1.37)	1.19 (0.79–1.79)	1.01 (0.66–1.54)	0.83
Glycemic index	1.00	1.16 (0.78–1.73)	0.84 (0.55–1.27)	0.89 (0.58–1.37)	1.05 (0.69–1.59)	0.61
Glycemic load	1.00	0.89 (0.59–1.36)	1.13 (0.75–1.70)	1.09 (0.72–1.64)	0.94 (0.61–1.44)	0.96
25–<30 kg/m^2 (n = 192 cases)
Carbohydrate intake	1.00	1.32 (0.79–2.21)	0.91 (0.53–1.57)	1.20 (0.72–2.00)	1.39 (0.84–2.33)	0.27
Glycemic index	1.00	1.35 (0.83–2.19)	1.24 (0.75–2.05)	1.19 (0.73–1.96)	0.86 (0.50–1.46)	0.55
Glycemic load	1.00	1.27 (0.76–2.14)	1.15 (0.69–1.93)	1.07 (0.64–1.79)	1.26 (0.76–2.10)	0.59
≥30 kg/m^2 (n = 147 cases)
Carbohydrate intake	1.00	1.38 (0.70–2.72)	1.19 (0.62–2.28)	0.99 (0.51–1.95)	1.68 (0.86–3.29)	0.35
Glycemic index	1.00	0.82 (0.43–1.56)	1.32 (0.71–2.44)	1.03 (0.55–1.93)	0.93 (0.50–1.74)	0.99
Glycemic load	1.00	1.15 (0.62–2.14)	0.69 (0.34–1.40)	1.05 (0.55–2.01)	1.57 (0.82–2.99)	0.18

A total of 262,993 person-years and 214 incident endometrial cancer cases were available for the analyses using exposure data from the second questionnaire and with follow-up from 1998 through June 2005. As in the main analysis, we observed no overall association between carbohydrate intake, glycemic index or glycemic load and endometrial cancer risk; the age-adjusted RRs for the highest compared with the lowest quartile (we used quartiles rather than quintiles because of a smaller number of cases than in our main analyses) were 1.32 (95% CI, 0.88–1.97) for carbohydrate intake, 1.20 (95% CI, 0.81–1.78) for glycemic index and 1.17 (95% CI, 0.78–1.75) for glycemic load. Removing diabetics from the analyses did not change the results materially (data not shown). In analyses stratified by physical activity, there were nonsignificant positive associations between carbohydrate intake and glycemic load with risk of endometrial cancer among women with low physical activity, but not among physically active women (Table IV). Tests for interaction between physical activity and carbohydrate intake or glycemic load in relation to endometrial cancer were not statistically significant (p-interaction = 0.37 for carbohydrate intake; p-interaction = 0.20 for glycemic load). Among women who were both physically inactive and overweight (BMI ≥ 25 kg/m²), the age-adjusted RRs for the highest versus the lowest quartile were 1.90 (95% CI, 0.84–4.31) for carbohydrate intake and 2.99 (95% CI, 1.17–7.67) for glycemic load (Table IV). Additional adjustment for smoking status (never, past and current) did not substantially change the results for carbohydrate intake (RR, 1.83; 95% CI, 0.81–4.17) or glycemic load (RR, 3.10; 95% CI, 1.18–8.17). After further controlling for education, age at menarche, oral contraceptive use, parity, age at menopause and postmenopausal hormone use, the RRs comparing the highest with the lowest quartile were 1.86 (95% CI, 0.81–4.27) for carbohydrate intake and 2.90 (95% CI, 1.09–7.77) for glycemic load.

Table IV. Rate Ratios and 95% CIS of Endometrial Cancer According to Quartiles of Carbohydrate Intake, Glycemic Index, and Glycemic Load, Stratified Physical Activity and Body Mass Index Among 36,369 Women in the Swedish Mammography Cohort (1998–2005)1

	Quartile				ptrend
	1	2	3	4
1 Cox proportional hazards models were used to calculate rate ratios adjusted for age in months. Carbohydrate intake, glycemic index and glycemic load were adjusted for total energy intake using the residual method.25 The number of cases does not add up to the total number of cases (n = 214) owing to missing data on physical activity and body mass index. 2 Physically active defined as ≥2 hr/week of leisure time physical activity. 3 Physically inactive defined as ≤1 hr/week of leisure time physical activity.
Physically active (n = 110 cases)2
Carbohydrate intake	1.00	1.41 (0.81–2.47)	0.99 (0.55–1.79)	1.18 (0.68–2.07)	0.87
Glycemic index	1.00	1.35 (0.79–2.30)	1.14 (0.66–1.99)	1.19 (0.67–2.09)	0.68
Glycemic load	1.00	0.99 (0.56–1.74)	1.26 (0.74–2.16)	0.95 (0.54–1.68)	0.98
Physically inactive (n = 88 cases)3
Carbohydrate intake	1.00	1.55 (0.82–2.92)	1.49 (0.80–2.81)	1.68 (0.89–3.18)	0.13
Glycemic index	1.00	0.73 (0.36–1.47)	1.24 (0.68–2.29)	1.36 (0.75–2.47)	0.14
Glycemic load	1.00	1.31 (0.67–2.56)	1.83 (0.98–3.42)	1.78 (0.94–3.37)	0.05
Physically inactive and body mass index ≥25 kg/m^2 (n = 53 cases)3
Carbohydrate intake	1.00	1.12 (0.46–2.76)	1.82 (0.81–4.10)	1.90 (0.84–4.31)	0.07
Glycemic index	1.00	1.32 (0.51–3.42)	2.29 (0.98–5.35)	1.73 (0.72–4.17)	0.15
Glycemic load	1.00	2.42 (0.90–6.46)	2.99 (1.17–7.67)	2.99 (1.17–7.67)	0.02

Discussion

In this large prospective study, we found no overall associations between carbohydrate intake, glycemic index or glycemic load and the incidence of endometrial cancer. However, among overweight women with low physical activity, we observed a nonsignificant 1.9-fold increase in risk of endometrial cancer for those who had a high carbohydrate intake and a statistically significant 3-fold increase in risk for those who had a high glycemic load.

Our findings are broadly consistent with those from the Iowa women's Health Study21 and the Canadian National Breast Screening Study.22 The Iowa women's Health Study,21 with 15 years of follow-up and 415 cases, showed no associations of glycemic index or glycemic load with endometrial cancer risk in the whole cohort, but found a statistically significant 1.9-fold increase in endometrial cancer risk for the highest versus the lowest quintile of glycemic load among obese nondiabetic women. Similarly, the Canadian National Breast Screening Study,22 with 16.4 years of follow-up and 426 cases, found a statistically significant 1.9-fold increased risk of endometrial cancer comparing the highest with the lowest quartile of glycemic load among obese women. In agreement with our finding, the Canadian National Breast Screening Study22 also observed an elevated risk of endometrial cancer associated with a high glycemic load among physically inactive women, although this association was not statistically significant (highest versus lowest quartile: RR, 1.50; 95% CI, 0.93–2.42). In a hospital-based case–control study in Italy and Switzerland (with 410 cases),20 there was a suggestion of a stronger positive relation between glycemic index and risk of endometrial cancer in overweight women (highest versus lowest quintile: odds ratio, 2.28; 95% CI, 1.27–4.12) than in lean women (highest versus lowest quintile: odds ratio, 1.62; 95% CI, 1.86–3.05).

Obesity is a major determinant of insulin resistance and hyperinsulinemia.29, 30 Likewise, physical activity strongly influences glucose tolerance and insulin sensitivity.30, 31 Thus, it is likely that individuals who are overweight or inactive have a greater insulin response to their diet as a result of their physiological condition compared to lean and active individuals. In the present study, women who were both overweight and inactive and had a high glycemic load had a statistically significant elevated risk of endometrial cancer. Some previous studies of cancer of the endometrium,21, 22 breast,32, 33 colorectum34, 35, 36 and pancreas37 have also found evidence of effect modification of the association between glycemic index or load and cancer risk by BMI or physical activity.

The major strengths of our study include its prospective and population-based design, a large sample size and detailed information on diet. The prospective design precluded recall bias and the virtually complete follow-up of the study population through linkage to various population-based registers27 largely minimizes the concern that our results have been affected by differential loss to follow-up.

A limitation of our study is that dietary intake was assessed with a self-administered food-frequency questionnaire, which will inevitably lead to some error in the measurement of diet and the calculation of glycemic index and glycemic load. Furthermore, the glycemic index values of some foods are currently based on results reported in only 1 or 2 studies, and those studies often had small sample sizes.24 Thus, measurement error in our study could also be caused by random variation in the estimated glycemic index values. Because of the prospective design of our study, any measurement error and resulting misclassification would be nondifferential. Random nondifferential misclassification of carbohydrate intake, glycemic index and glycemic load may have resulted in underestimation of the relation between these dietary factors and endometrial cancer risk. Another limitation is that information on physical activity was available only in the second (1997) questionnaire. Hence, our analyses stratified by physical activity were based on a relatively small number of incident cases, which limited our statistical power to detect statistically significant interactions. Finally, because our study was observational, we cannot exclude the possibility that unmeasured confounding may have affected our risk estimates.

In summary, in this prospective study, we observed no overall association of carbohydrate intake, glycemic index or glycemic load with endometrial cancer risk. However, our findings suggest that a high carbohydrate intake and a high-glycemic load diet may be associated with an increased risk of endometrial cancer among overweight women with low physical activity.