Long-term dietary heme iron and red meat intake in relation to endometrial cancer risk

ABSTRACT

Background: Heme and total iron, present in meat, have been hypothesized to promote carcinogenesis. Few prospective studies have examined the associations between intakes of heme and total iron, types of meat, and endometrial cancer risk.

Objective: We evaluated the associations between intakes of heme and total iron, types of meat, and risk of endometrial cancer in a large cohort of women.

Design: Among 60,895 women in the Swedish Mammography Cohort, 720 endometrial cancer cases were confirmed during 21 y of follow-up. RRs and 95% CIs were calculated by Cox proportional hazards models.

Results: A comparison of the highest with the lowest quartile showed a 20–30% higher risk of endometrial cancer for higher intakes of heme iron (RR: 1.24; 95% CI: 1.01, 1.53 for ≥1.63 compared with <0.69 mg/d), total iron (RR: 1.31; 95% CI: 1.07, 1.61 for ≥15.09 compared with <12.27 mg/d), and liver (RR: 1.29; 95% CI: 1.06, 1.56 for ≥100 compared with <100 g/wk). No statistically significant associations were observed between intakes of red and processed meats and endometrial cancer risk. RRs did not greatly differ when we stratified by BMI, parity, and intakes of alcohol, vitamin C, or zinc or when we excluded patients with diabetes.

Conclusions: Our study suggests a modest positive association between heme iron, total iron, and liver intakes and endometrial cancer risk; no statistically significant associations were observed for intakes of other red and processed meats and endometrial cancer risk. The Swedish Mammography Cohort was registered at clinicaltrials.gov as NCT01127698.

INTRODUCTION

Dietary heme iron intake has been suggested as a risk factor for cancers of the colon and rectum, breast, and prostate (1–6). Total dietary iron, which includes heme [present in meat (7, 8)] and nonheme iron, can lead to a higher prooxidant load, which is seen when catalytic iron converts hydrogen peroxide to hydroxyl radical. The higher prooxidant load may ultimately lead to more oxidative stress and DNA damage (9, 10). Heme iron has also been shown to be positively associated with the risk of diabetes, obesity, and markers associated with both obesity and diabetes (11, 12)—all risk factors for endometrial cancer (13–18). In addition to heme iron, other components of animal products may promote carcinogenesis. For example, exogenous N-nitroso compounds and heterocyclic amines, which have been shown to be carcinogenic in rodents, are generated as a result of certain curing and cooking techniques, respectively (19, 20).

Recently, one case-control study observed an increased risk of endometrial cancer of ∼2-fold with higher heme iron intakes (21). However, one prospective cohort, the Canadian National Breast Screening Study, examined the association between baseline heme iron intake and endometrial cancer risk and saw no association (22). Because of this inconsistency in the literature, further assessment of the association is warranted. Because meat is one of the main sources of heme iron, prospective evaluation of heme iron and meat intake with endometrial cancer risk will greatly add to the literature to date.

We prospectively investigated the association between intakes of heme iron, total dietary iron, and heme iron–rich animal products (eg, liver, red meat, and processed meat) with risk of endometrial cancer in a population-based cohort study that assessed diet at baseline (1987) and once during follow-up (1997). Given that the effect of iron and meat may vary by risk factors for endometrial cancer, we also considered whether these associations differed by environmental and other nutritional factors.

SUBJECTS AND METHODS

Population

The Swedish Mammography Cohort, a population-based prospective cohort study, was conducted among all women who lived in Uppsala County in central Sweden and were born between 1914 and 1948 and all women who lived in the adjacent Västmanland County and were born between 1917 and 1948. Between 1987 and 1990, 90,303 women received an invitation by mail to participate in a free mammography-screening program and to complete a questionnaire on diet, weight, height, and education; 66,651 of these women returned a completed questionnaire. In this analysis, we excluded women who had an erroneous national registration number, a previous cancer diagnosis other than nonmelanoma skin cancer at baseline, a hysterectomy before baseline, or log_e-transformed energy intakes beyond 3 SDs from the log_e-transformed mean energy intake of the population. This left 60,895 individuals for this analysis.

In 1997, a second questionnaire was sent to all 56,030 women in the cohort who were still living in Uppsala and Västmanland counties. The follow-up questionnaire included additional information on physical activity, medical history, age at menarche, history of oral contraceptive use, age at menopause, postmenopausal hormone use, and lifestyle factors such as cigarette smoking history and use of dietary supplements. Approximately 70% (39,227) of the women returned the questionnaire, 3184 of whom had a cancer diagnosis or hysterectomy between baseline and 1997. The study was approved by the Regional Ethical Review Board at the Karolinska Institutet (Stockholm, Sweden) and the Columbia University Institutional Review Board. Completion of the self-administered questionnaire was considered to imply informed consent to participate in this study.

Exposure assessment

Usual frequency of food consumption over the past year was estimated from a 67-item food-frequency questionnaire (FFQ)⁴ of commonly consumed foods in Sweden completed in 1987 and a 96-item FFQ in 1997. Nutrient intakes were estimated by using the food-composition method (23) based on the Swedish National Food Administration Database (24) and were energy-adjusted by using the regression-residual method (23). Intake of iron from diet was estimated from the FFQ, whereas 2 different methods were applied to calculate dietary heme iron. For the main analyses presented, heme-iron content was estimated by multiplying a factor of 0.4 by the total iron content of all meat items—a method proposed by Monsen and used in earlier publications (8, 11, 25). Because this calculation does not take into account variation in heme-iron content relative to the total iron content in animal products, heme-iron content was also calculated by using a method developed by Balder et al (26). Briefly, the average of the measured values of heme-iron content of specific types of meat reported in the published literature (26) was used to generate the heme-iron content–specific factor (instead of the 0.4 used in the other method) for a given food item (ranging from 0.01 for shellfish to 0.70 for blood pudding); for each meat (or fish) item, the type-specific heme-iron factor was multiplied by its total iron content to derive the heme-iron content of that item. Nonheme iron was calculated by subtracting heme iron from total iron.

Women were asked to report how often on average they consumed different foods; they could choose one from 8 prespecified frequencies ranging from “never or seldom” to “4 times per day or more” on the 1987 FFQ and “never” to “3 times a day or more” on the 1997 FFQ. The number of meat questionnaire items on the 1987 questionnaire was 12; the 1997 questionnaire had 14 meat questionnaire items. Similar questions regarding meat, poultry, and fish intake were asked at both time points (1987 and 1997). The summary measure of red meat included minced meat (eg, hamburgers and meatballs), whole beef (eg, steaks), and casserole with beef, pork, or veal. Processed meat consisted of sausage or hot dogs, bacon, ham, salami or lunch meat, and blood pudding/sausage. Blood pudding, sausage, and sausage dishes and cold cuts were summed for the category of total sausage. Total liver included liver, liver pate, and kidney. Chicken and other poultry were summed for the category of poultry. Total meat comprised red meat, processed meat, poultry, and fish intake. The age-specific portion sizes were based on the mean values obtained from 213 randomly chosen women from the study area who weighed and recorded their foods during four 1-wk periods 3–4 mo apart. The baseline FFQ and diet records showed a reasonable correlation ranging from 0.3 to 0.7 for red and processed meat items (A Wolk, unpublished data, 1992).

Outcome assessment

Participants were followed from the date of the baseline questionnaire until the date of diagnosis of endometrial cancer, date of death, date the participant moved out of the study area, date of hysterectomy, or end of follow-up, whichever came first. Invasive endometrial adenocarcinoma was ascertained through computerized record linkage of the study population with the Swedish Cancer Register. Invasive endometrial adenocarcinoma was defined by International Classification of Diseases (ICD)-9 code 182.0 or ICD-10 code C54.9. The Swedish Cancer Register has been estimated to be ∼100% complete (27). Information pertaining to death and hysterectomy was identified through the Swedish Death Register, the National Swedish Population Register, and the nationwide Swedish Patient Register. During the 21 y of follow-up, 720 cases of endometrial cancer were diagnosed: 229 cases from 1987 to 1997, 302 cases from 1997 to 2008 who completed the 1997 follow-up questionnaire, and 189 cases from 1997 to 2008 who did not complete the 1997 follow-up questionnaire.

Statistical analysis

Directly age-standardized demographic and health characteristics were compared across quartiles of the distribution of heme iron intake. Individuals were excluded from the analysis of a particular dietary factor if there were missing data on that specific dietary exposure. Dietary exposures were modeled continuously, categorically according to absolute cutoffs based on grams per week and by quartiles defined within the study.

RRs and 95% CIs were calculated by Cox proportional hazards models. Two different methods were applied to analyze the association between endometrial cancer risk and dietary heme iron and meat intake: using baseline dietary data only and a cumulative average approach (28). Thus, baseline analyses, comparing the association between intakes of heme iron, total iron intake, and specific types of meat and endometrial cancer risk from 1987 to 2008, used only the 1987 food and nutrient intake data. To reduce within-person variation and to best represent long-term diet, the cumulative average approach (28) was used in which dietary data from the baseline questionnaire was used for follow-up from 1987 through 1997 and an average of the dietary intakes from baseline and the 1997 questionnaire was used for follow-up from 1997 to 2008. Two models were used; one model adjusted for age (continuously) and energy (continuously), whereas the multivariate model for the baseline analyses and for the 1987–1997 follow-up of the cumulative average approach additionally adjusted for parity (nulliparous, 1, 2, or ≥3 live births), education (≤9, 10–12, or >12 y of schooling), and BMI (in kg/m²; <21, 21–24.9, 25–29.9, or ≥30). Parity (nulliparous, 1, 2, or ≥3 live births), education (≤9, 10–12, or >12 y of schooling), BMI (<21, 21–24.9, 25–29.9, or ≥30), physical activity (<38.9, 38.9–42.1, 42.2–45.8, or ≥45.9 metabolic equivalent tasks/h per day), smoking history in pack-years (never, past smoker of <15 pack-years, past smoker of ≥15 pack-years, current smoker of <40 pack-years, or current smoker of ≥40 pack-years), menopausal status (premenopausal or postmenopausal), and oral contraceptive use (no or yes) were included in the cumulative average approach model for the 1997–2008 follow-up. In addition, we included different parameterization of age (categories of age) and energy (quartiles of energy) into the model. Because the results were not significantly modified, we present the estimates adjusted for age and energy included as continuous covariates. Indicator terms based on serving-size categories of the distribution of meat intake were entered into the model. To test whether there was a linear trend in risk of disease with increasing intake, ordinal terms for the serving size categories and quartiles of distribution of meat intake and iron and heme iron intake were included in the model, the coefficients for which were evaluated by using the Wald test.

Stratified analyses were conducted by categories of alcohol intake (0 or >0 g/d), BMI (<25 or ≥25), parity (<2, 2, or >2 live births), and categories based on the median dietary intake of vitamin C, vitamin E, and zinc in the 1987 cohort. For each factor of interest, a cross-product term of the ordinal score for the level of each factor and intake of a specific food or nutrient expressed as a continuous variable was included in the model. Participants with missing values for the factor of interest were excluded from these analyses. All statistical analyses were performed by using SAS version 9.1 (SAS Institute Inc).

RESULTS

Baseline cohort characteristics of the Swedish Mammography Cohort, by quartiles of heme iron, are summarized in Table 1. The total study population consisted of 60,895 women, of whom 720 developed invasive endometrial adenocarcinoma during 1,225,047 person-years of follow-up. At the start of follow-up in 1987, participants ranged in age from 40 to 76 y, and ∼40% of the women were classified as overweight (BMI >25) or obese (BMI >30). In 1997, ∼23% of all women were current smokers, 91% were postmenopausal, and more than half (58%) had ever used hormone replacement therapy.

TABLE 1

Age-standardized baseline cohort characteristics according to quartile of heme iron intake in the Swedish Mammography Cohort¹

	Baseline, 1987: quartile of heme iron2
Characteristic	1	2	3	4
Median (g/d)	0.49	0.90	1.37	2.04
Cases (n)	168	161	193	198
Person-years	301,912	308,216	307,763	307,156
Age (y)3	55.90 ± 9.924	52.83 ± 9.60	53.09 ± 9.59	52.92 ± 9.46
BMI (%)
<21 kg/m2	18	15	14	12
21 to <25 kg/m2	46	46	45	45
25 to <30 kg/m2	28	29	31	32
≥30 kg/m2	8	10	10	11
Parity (%)
0	13	11	10	10
1	18	17	16	16
2	38	39	40	39
≥3	31	33	34	35
Education (%)
≤9 y	87	87	87	88
10–12 y	7	8	8	7
>12 y	6	5	5	5
Energy (kcal)	1639 ± 471	1610 ± 470	1633 ± 416	1470 ± 403

1

Directly age standardized to the age distribution of the analytic cohort.

2

Heme iron content was estimated by multiplying a factor of 0.4 by the total iron content of all meat items.

3

Value is not age standardized.

4

Mean ± SD (all such values).

We examined the association between heme iron and total dietary iron intake and endometrial cancer risk (Table 2). On the basis of baseline data only, a 20–30% higher risk of endometrial cancer was observed for heme iron intake (multivariate RR: 1.24; 95% CI: 1.01, 1.53) and total dietary iron intake (multivariate RR: 1.31; 95% CI: 1.07, 1.61) when the highest and lowest quartiles were compared. Similar but slightly attenuated results (multivariate RR: 1.18; 95% CI: 0.96, 1.47) were observed for heme iron intake when meat-specific factors were applied to derive heme-iron values (26). Similarly, a positive association was observed between heme iron (multivariate RR: 1.24; 95% CI: 1.01, 1.52) and total dietary iron (multivariate RR: 1.25; 95% CI: 1.02, 1.54) and risk of endometrial cancer when the cumulative average approach was applied taking into account changes in diet over time. We further analyzed nonheme sources of iron; a suggestive, but not statistically significant, association was observed for high intakes of nonheme iron (multivariate RR: 1.20; 95% CI: 0.98, 1.48).

TABLE 2

RRs and 95% CIs for endometrial cancer by heme iron and total iron intake in the Swedish Mammography Cohort, 1987–2008¹

		Quartile of intake
Nutrient	Continuous mg/d	1	2	3	4	P-trend2
Heme iron (mg/d)3
Cases	720	168	161	193	198
Median intake	1.13	0.49	0.90	1.37	2.04
Intake range	0–21.13	0–0.69	0.69–1.13	1.13–1.63	1.63–21.13
Age RR4	1.09	1.00	1.05	1.24	1.28	0.007
95% CI	1.00, 1.19	Reference	0.84, 1.30	1.01, 1.53	1.04, 1.58
MV RR5	1.07	1.00	1.03	1.22	1.24	0.02
95% CI	0.98, 1.17	Reference	0.83, 1.28	0.99, 1.50	1.01, 1.53
Updated MV RR6	1.07	1.00	1.05	1.12	1.24	0.03
95% CI	0.98, 1.17	Reference	0.85, 1.30	0.91, 1.39	1.01, 1.52
Iron (mg/d)
Cases	720	156	165	175	224
Median intake	13.62	11.38	12.96	14.31	16.22
Intake range	3.83–63.05	3.83–12.27	12.27–13.62	13.62–15.09	15.09–63.05
Age RR4	1.04	1.00	1.05	1.09	1.36	0.03
95% CI	1.01, 1.07	Reference	0.84, 1.30	0.88, 1.36	1.10, 1.66
MV RR5	1.03	1.00	1.04	1.07	1.31	0.009
95% CI	1.00, 1.06	Reference	0.84, 1.29	0.87, 1.33	1.07, 1.61
Updated MV RR6	1.03	1.00	1.14	1.09	1.25	0.05
95% CI	1.00, 1.07	Reference	0.92, 1.41	0.88, 1.36	1.02, 1.54

1

MV, multivariate.

2

The P value (test for trend) was evaluated by using the Wald test for the coefficient included in the model that denoted the ordinal terms for the serving-size categories and quartiles of distribution of meat intake and iron and heme iron intakes.

3

Estimated by multiplying a factor of 0.4 by the total iron content of all meat items.

4

Cox proportional hazards models using baseline values of intake adjusted for age (continuously) and energy (continuously).

5

Cox proportional hazards models using baseline values of intake adjusted for age (continuously), energy (continuously), BMI (in kg/m²; <21, 21–24.9, 25–29. 9, or ≥30), parity (nulliparous or 1, 2, or ≥3 live births), and education (≤9, 10–12, or >12 y of schooling).

6

Cox proportional hazards models using cumulative average values of intake (1987 and an average of the 1987 and 1997 values) adjusted for age (continuously), energy (continuously), BMI (in kg/m²; <21, 21–24.9, 25–29.9, or ≥30), education (≤9, 10–12, or >12 y of schooling), and parity (nulliparous or 1, 2, or ≥3 live births) and the following additional covariates for the 1997–2008 follow-up: smoking (never, past smoker of <15 pack-years, past smoker of ≥15 pack-years, current smoker of <40 pack-years, or current smoker of ≥40 pack-years), menopausal status (premenopausal or postmenopausal), physical activity (<38.9, 38.9–42.1, 42.2–45.8, or ≥45.9 metabolic equivalent tasks/h per day), and oral contraceptive use (no or yes).

Total liver intake was associated with an ∼30% higher risk of endometrial cancer when baseline data were examined (Table 3); however, no statistically significant association was observed when analyzed by using the cumulative averaging approach (data not shown). Furthermore, we observed no statistically significant associations between intakes of red meat (multivariate RR: 1.06; 95% CI: 0.68, 1.66 for the comparison of ≥600 with <100 g/wk), processed meat (multivariate RR: 1.12; 95% CI: 0.84, 1.49 for the comparison of ≥300 with <100 g/wk), and sausage (multivariate RR: 1.22; 95% CI: 0.92, 1.62 for the comparison of ≥300 with <100 g/wk) and risk of endometrial cancer when analyzed by using the baseline data only.

TABLE 3

Age- and multivariate-adjusted RRs and 95% CIs for endometrial cancer by grams of red and processed meat intake in the Swedish Mammography Cohort using baseline dietary data, 1987–2008¹

	Continuous				Meat intake				P value2
Red meat
Intake (g/wk)	—	<100	100 to <200	200 to <300	300 to <400	400 to <500	500 to <600	≥600	—
Median (g/100 g wk)	379.93	40.60	162.00	256.07	349.55	446.00	543.03	714.07	—
Cases (n)	718	27	62	136	162	146	94	91	—
Age RR3	1.28	1.00	0.92	1.12	1.17	1.37	1.43	1.21	0.02
95% CI	1.04, 1.63	Reference	0.58, 1.44	0.74, 1.69	0.78, 1.76	0.90, 2.07	0.92, 2.20	0.78, 1.89	—
MV RR4	1.15	1.00	0.89	1.07	1.10	1.26	1.32	1.06	0.11
95% CI	0.90, 1.47	Reference	0.57, 1.40	0.71, 1.62	0.73, 1.65	0.83, 1.91	0.85, 2.04	0.68, 1.66	—
Processed meat5
Intake (g/wk)	—	<100	100 to <200	200 to <300	≥300	—	—	—	—
Median (g/100 g wk)	147.47	53.80	145.73	237.57	367.40	—	—	—	—
Cases (n)	716	194	291	161	70	—	—	—	—
Age RR3	1.28	1.00	1.31	1.38	1.24	—	—	—	0.02
95% CI	0.84, 1.97	Reference	1.09, 1.58	1.11, 1.71	0.93, 1.66	—	—	—	—
MV RR4	1.05	1.00	1.28	1.31	1.12	—	—	—	0.12
95% CI	0.68, 1.63	Reference	1.06, 1.53	1.05, 1.62	0.84, 1.49	—	—	—	—
Sausage
Intake (g/wk)	—	<100	100 to <200	200 to <300	≥300	—	—	—	—
Median (g/100 g wk)	124.00	50.40	139.50	238.00	352.90	—	—	—	—
Cases (n)	716	250	296	117	53	—	—	—	—
Age RR3	1.36	1.00	1.16	1.31	1.15	—	—	—	0.05
95% CI	0.86, 2.15	Reference	0.98, 1.37	1.05, 1.65	0.84, 1.57	—	—	—	—
MV RR4	1.11	1.00	0.99	1.18	1.22	—	—	—	0.24
95% CI	0.69, 1.76	Reference	0.74, 1.31	0.90, 1.53	0.92, 1.62	—	—	—	—
Liver
Intake (g/wk)	—	<100	≥100	—	—	—	—	—	—
Median (g/100 g wk)	23.00	17.00	121.57	—	—	—	—	—	—
Cases (n)	707	577	130	—	—	—	—	—	—
Age RR3	2.90	1.00	1.31	—	—	—	—	—	0.006
95% CI	1.67, 5.02	Reference	1.08, 1.59	—	—	—	—	—	—
MV RR4	2.83	1.00	1.29	—	—	—	—	—	0.01
95% CI	1.58, 5.09	Reference	1.06, 1.56	—	—	—	—	—	—

1

MV, multivariate.

2

The P value (test for trend) was evaluated by using the Wald test for the coefficient included in the model that denoted the ordinal terms for the serving-size categories and quartiles of distribution of meat intake and iron and heme iron intakes.

3

Cox proportional hazards models using baseline values of intake adjusted for age (continuously) and energy (continuously) using baseline values of intake.

4

Cox proportional hazards models using baseline values of intake adjusted for age (continuously), energy (continuously), BMI (in kg/m²; <21, 21–24.9, 25–29.9, or ≥30), parity (nulliparous or 1, 2, or ≥3 live births), and education (≤9, 10–12, or >12 y of schooling).

5

Processed meat consisted of sausage or hot dogs, bacon, ham, salami, lunchmeat, and blood pudding/sausage (baseline intakes).

Following the World Cancer Research Fund (WCRF) and American Institute for Cancer Research report recommendation to not consume >500 g red meat/wk, we also examined the association of ≥500 compared with <500 g red meat/wk (29). We observed no statistically significant association between red meat intakes >500 g/wk and endometrial cancer risk (data not shown). In addition, no statistically significant associations were observed for intakes of total meat or poultry and risk of endometrial cancer by using baseline data only or the cumulative average approach (data not shown).

RRs for heme iron, total dietary iron, and total meat intake and endometrial cancer did not greatly differ when we limited the analyses to participants without a diabetes diagnosis or when we stratified by BMI, smoking (data only available for 1997), parity, and intake of other dietary factors (alcohol, vitamin C, or zinc). However, the positive association between total dietary iron and endometrial cancer risk was stronger among those with a higher BMI (≥25; P-interaction = 0.006), and the positive association between heme iron and endometrial cancer risk was stronger among those who consumed no alcohol (P-interaction = 0.02). In addition, cases that occurred close in time to the completion of the FFQ may represent individuals who altered their diet because of factors such as prediagnostic disease symptoms. RRs were not substantially changed when we excluded cases that were diagnosed within the first year or first 2 y of follow-up, except for poultry, for which a positive association with endometrial cancer risk was observed.

DISCUSSION

In this prospective population-based cohort study, we observed a slightly higher risk of endometrial cancer with high dietary iron and heme-iron consumption. A higher intake of liver—a very rich source of heme iron—was associated with a higher risk of endometrial cancer when the 1987 baseline dietary data were examined but was attenuated when the cumulative average of both time points was examined. No statistically significant associations were observed for intakes of other meat products (eg, processed meat or sausage) and endometrial cancer risk.

Our results for heme iron and total dietary iron intake were not in line with those of the Canadian National Breast Screening Study, which observed no association between heme iron and endometrial cancer risk (22). However, our results were similar, but less pronounced, to those of a recent case-control study in which a 2-fold risk was observed for iron from animal sources (multivariate RR: 1.86; 95% CI: 1.22, 2.85 for the comparison of >4.14 with <1.92 mg/d) (21). This large population-based, case-control study examined animal-derived iron intake (iron from nondairy animal foods) and not heme iron intake; thus, these measures represent slightly different exposures.

A few different mechanisms underlying the association of endometrial cancer risk with dietary iron and heme iron intake have been suggested. First, total dietary iron and heme iron may lead to a higher prooxidant load, which may ultimately lead to more oxidative stress and DNA damage (9, 10). Second, heme iron has also been shown to be positively associated with the risk of diabetes, obesity, and markers associated with both obesity and diabetes (11, 12)—all of which are suspected or established risk factors for endometrial cancer (13–18). Because obesity and personal history of diabetes may be in the causal pathway between heme iron and endometrial cancer risk, we also conducted analyses with and without obesity as a covariate and limited the analysis to those who were not diabetic at baseline; the results were similar. Finally, other components found in animal products or specifically processed meat, which may be highly correlated with heme and total iron, such as N-nitroso compounds and heterocyclic amines, are thought to be carcinogenic (19, 20). Liver is one such meat, which is known to contain high concentrations of purines—heterocyclic aromatic organic compounds. Because we did not collect cooking and processing information on meat intake, we were unable to examine heterocyclic amines and endometrial cancer risk.

Many studies have examined the association between meat intake, a major contributor of heme iron, and endometrial cancer risk. In contrast with our nonsignificant results, a recent meta-analysis (30) observed random-effects dose-response summary estimates of 1.51 (95% CI: 1.19, 1.93) per 100 g red meat/d based on 7 case-control studies. Furthermore, the meta-analysis observed a random-effects dose-response summary estimate of 1.26 (95% CI: 1.03, 1.54) per 100 g total meat/d based on 8 case-control studies. Similarly, in the Canadian Study of Diet, Lifestyle and Health, positive, but not statistically significant, findings were observed between red meat, processed meat, and total meat intakes and risk of endometrial cancer (31). The Iowa Women’s Health Study, which was not included in the meta-analysis because of a lack of cutoff information for intake of types of meat, examined meat intake and endometrial cancer risk; no association was observed for intake of total meat and red meat, whereas only a modest positive association was observed for intake of processed meat and poultry (32). None of these estimates were adjusted for BMI, weight, or smoking—potentially important covariates. In addition, our study observed a positive association between liver intake and endometrial cancer risk. However, results from 3 previous case-control studies, conducted in Europe (33, 34) and China (35), were heterogeneous, and a suggested elevated (33), null (35), and inverse (34) association was reported. The WCRF/American Institute for Cancer Research report (29) recommended limited consumption of red meat to <500 g/wk (18 oz) and avoidance of processed meats to prevent cancers of the bowel. Specific to endometrial cancer, the WCRF panel determined that there is limited evidence suggesting that red meat is a risk factor (29). Of the endometrial cancer studies included in the meta-analysis and WCRF report, most have been based on case-control data, which may be subject to selection and recall biases. In addition, many studies did not adjust for BMI or energy intake—potential confounding factors. However, we observed no association between intake of red meat >500 g/wk and risk of endometrial cancer. One reason why we may observe a statistically significant association between endometrial cancer and heme and total iron intake, but not with all the individual major sources of iron as red and processed meat (eg, sausage and processed meat), may be the result of a smaller contrast and precision in the measurement of the etiologic agent when assessing dietary iron and heme iron intake per se compared with the use of total meat or specific type of meat consumption as a surrogate of dietary iron and heme iron intake. The incorporation of all sources into the dietary iron and heme-iron variable also allows for examination of a wider contrast in the exposure of interest compared with when just select food groups of meat are examined; therefore, we may have more power to detect the association between the nutrient and the risk of endometrial cancer. For example, the analysis of red meat does not include all sources of heme iron such as fish or poultry or dietary iron such as nonmeat sources of select vegetables; therefore, the analysis of red meat may not adequately capture or actually dilute the association between the true etiologic or causal agent (eg, heme iron and dietary iron) and endometrial cancer risk.

Because diet was measured before the diagnosis of endometrial cancer, it is unlikely that the reporting of meat intake would be systematically biased by disease status in this prospective study. The possibility of misclassification of meat intake exists; however, this would likely be nondifferential misclassification, and such misclassification would have attenuated the relative risk estimates for the relation between intakes of meats and nutrients and risk of endometrial cancer. Measurement of dietary intake was updated during the follow-up to reduce within-person variation and to best represent long-term diet, so measurement error was potentially reduced but could not be ruled out. In addition, we were not able to measure the potentially carcinogenic compounds that are found in meats, including N-nitroso compounds, heterocyclic amines, or polycyclic aromatic hydrocarbons. Because all covariates were not measured in 1987 and 1997, we were not able to adjust for all covariates at both time points. At both time points, we adjusted for the important endometrial cancer risk factors (eg, BMI and parity) and, in general, results from age-adjusted and multivariate models and models in which the dietary data were cumulatively averaged were similar except for liver intake. Nonetheless, we could not rule out uncontrolled confounding by an unknown or unmeasured factor or residual confounding from measurement error in the included covariates; however, our results suggest that residual or unmeasured confounding would be small.

The strengths of this study included the prospective assessment of diet, repeated measurement of diet 10 y after baseline, a long follow-up time (21 y), a population-based cohort design, and a large number of endometrial cancer cases. Recall or selection biases were minimized in this study because exposure data were collected before disease and because of the cohort’s high follow-up rate and completeness of the cancer registry.

In summary, a modest statistically significant elevation in the endometrial cancer risk was seen for dietary iron, heme iron, and liver intake. No statistically significant associations were observed for other specific types of red and processed meat intakes and endometrial cancer risk. These dietary factors may work through oxidation and inflammatory mechanisms. Further examination of these factors is warranted.

The authors’ responsibilities were as follows—JMG: study concept and design, data analysis, interpretation of the results, manuscript writing, and primary responsibility for the final content of the manuscript; EF and RAG: interpretation of the results, critical revision of the manuscript, and provision of significant advice or consultation; and AW: research design, data collection, interpretation of the results, critical revision of the manuscript, and provision of significant advice or consultation. All authors read and approved the final manuscript. None of the authors had a conflict of interest.

FOOTNOTES

2

Supported by grants from the World Cancer Research Fund International, The Swedish Cancer Foundation, The Swedish Research Council for infrastructure, and the Swedish Foundation for International Cooperation in Research and Higher Education, and by the Karolinska Institutet’s Distinguished Professor Award (to AW).

4 ABBREVIATIONS

 
• FFQ

  food-frequency questionnaire

 
• ICD

  International Classification of Diseases

 
• WCRF

  World Cancer Research Fund