Animal food intake and cooking methods in relation to endometrial cancer risk in Shanghai

SUBJECTS AND METHODS

Eligible cases for the study were residents of urban Shanghai, aged 30–69 years, newly diagnosed with endometrial cancer between 1997 and 2003. A total of 1454 endometrial cancer cases were identified through the Shanghai Cancer Registry, and 1204 of them were interviewed, yielding a response rate of 82.8%. All cases were confirmed either by histopathology or by medical history review. Excluded from the analysis were 135 cases who refused to be interviewed (9.3%), 66 who died before interview (4.5%), 23 (1.6%) who could not be located, 12 who were out of town during the period of interview (0.8%), and 14 who could not be interviewed for other miscellaneous reasons (1.0%). The median interval between diagnosis and interview was 5.6 months.

Controls were randomly selected from the Shanghai Resident Registry and frequency matched to cases according to age distribution of endometrial cancer cases in 1996. After exclusion of 63 women who reported a hysterectomy, 1212 out of 1629 eligible controls (74.4%) consented and were interviewed for the study. Reasons for non-participation among controls were refusal (n=340, 20.9%), being out of town during the period of interview (n=61, 3.7%), severe illness (n=13, 0.8%), and other miscellaneous reasons (n=3, 0.2%).

Study participants were interviewed in person by trained nurses and physicians, using a structured questionnaire to obtain information on demographic characteristics, menstrual and reproductive events, exogenous hormone use, diet, cigarette smoking, alcohol consumption, physical activity, family history of cancer, and medical history. Direct body measurements on body height, weight, and waist and hip size were conducted at the time of interview. The quantitative food frequency questionnaire (FFQ) used in the study has been validated by two FFQ surveys and one 24-h diet recall survey (Shu et al, 2004). The FFQ listed 76 food items commonly consumed in Shanghai, including 19 animal food items which were classified into the following 10 food groups: red meat, organ meat, poultry, marine fish, fresh water fish, shrimp and crab, eel, shellfish, eggs, and milk. Red meat included pork, beef, and mutton meat. Each subject was asked to report her usual intake frequency (per day, week, month, or year) and amount in liangs (50 g) for each food item over the past 5 years, ignoring any recent changes. The participant was also asked whether she used deep-frying, stir-frying or roasting/grilling methods to prepare meats and fish, in addition to how frequently she used each cooking method in cooking these foods. The subject was also asked to report whether she usually deep-fried meat and fish to the level that (1) the surface still had a bloody colour; (2) virtually no surface was brown; (3) a small portion of the surface was brown (Goodman et al, 1997); the majority of the surface was brown; or (5) the entire surface was brown with a slightly burnt flavour. Levels 4 and 5 were referred to as well-done.

Consumption of common animal foods was categorised into four groups based on the quartile distribution among controls. Uncommon animal foods, such as organ meat, eel, shellfish, and milk were grouped into two levels, never and ever. The Wilcoxon rank sum test was used for comparisons of the median difference between cases and controls and the χ²-test was used for categorical variables. An unconditional logistic regression model was applied to obtain maximum likelihood estimates of the odds ratios (ORs) and their 95% confidence intervals (CIs). All of the ORs were adjusted for age (as a continuous variable), menopausal status (pre/post menopausal), diagnosis of diabetes (ever/never), alcohol consumption (ever/never), body mass index (BMI) (as a continuous variable), physical activity in metabolic equivalent tasks (as a continuous variable), and total energy intake (as a continuous variable). Other variables, including dietary intake of total vegetables and soy foods, did not appreciably alter the risk estimates and were not adjusted for in our analyses. Tests for trend across quartiles were performed in a logistic regression model by assigning a numerical score (1–4) to each category of categorised variables. Stratified analyses were conducted to evaluate potential modifying effects.

RESULTS

Table 1 shows comparisons of cases and controls for demographic factors and known risk factors of endometrial cancer. No significant differences were observed between cases and controls for age, income, marital status, use of hormone replacement therapy, mean intake of vegetables, or intake of fruits and soy protein. Compared with controls, cases were more likely to have attended higher education, have an earlier age at menarche, a later age at menopause, more years of menstruation, and to be pre-menopausal. Cases were also more likely to be diagnosed with diabetes, have a family history of cancer, have a higher total energy intake and a higher BMI, and were less likely to have had a pregnancy, engaged in regular exercise, used oral contraceptives, or consumed alcohol regularly (Table 1).

Presented in Table 2 are comparisons of the average intake levels of total and individual animal foods between cases and controls. The median intakes of red meat, poultry, marine fish, fresh water fish, shrimp and crab, eel, and shellfish were significantly higher in cases than in controls. No significant differences were found for intake of organ meat, eggs, and milk.

The associations between animal food intake and endometrial cancer risk are presented in Table 3. The ORs were 1.7 (95% CI: 1.3–2.2) and 2.4 (95% CI: 1.8–3.1), respectively, for those who ate total meat and total fish in the highest intake quartile as compared with those in the lowest quartile. A positive association was observed for consumption of red meat, poultry, marine fish, fresh water fish, shrimp and crab. Consumption of organ meat, eel, shellfish, and milk was not very common in our study population. A moderately increased risk (OR=1.3, 95% CI: 1.1–1.6) was found to be associated with eel or shellfish consumption. Egg and milk consumption were not associated with endometrial cancer risk. No case–control difference was found regarding ever/never eating organ meat. The consumption of organ meat in this study population was too low to evaluate disease risk by intake level.

Neither cooking method nor doneness level of meat and fish were associated with endometrial cancer risk, nor did they modify the association of meat and fish intake with disease risk (Table 4). Additional analyses stratified by consumption of total vegetables and soyfood did not show substantial changes in the above-mentioned associations (data not shown).

DISCUSSION

The positive association of disease with intake of red meat and the null association with eggs and milk found in the current study are consistent with those found in previous studies (La Vecchia et al, 1986; Levi et al, 1993; Shu et al, 1993; Zheng et al, 1995; Goodman et al, 1997; McCann et al, 2000; Jain et al, 2000; Terry and Vainio et al, 2002). However, our finding of a positive association with intake of poultry was inconsistent with previous reports in which null (Okochi et al, 1999; McCann et al, 2000) or inverse (Goodman et al, 1997) associations were reported. Our observation of an elevated risk associated with a high consumption of fish, although inconsistent with some previous studies (La Vecchia et al, 1986; Levi et al, 1993; Goodman et al, 1997; McCann et al, 2000; Terry and Wolk et al, 2002), is in line with that from our earlier report on the same disease (Shu et al, 1993). Endometrial cancer is a hormone-dependent disease. It has been suggested that dietary fat is responsible for the association between animal foods and endometrial cancer risk (Levi et al, 1993; Goodman et al, 1997), probably through influencing oestrogen metabolism and enhancing oestrogen re-absorption in the bowel (Gorbach and Goldin, 1987), although the relation between dietary fat and endogenous oestrogen levels remains controversial (Holmes et al, 2000). In vitro and in vivo studies have found that omega-3 fatty acids, which fish, particularly marine fish, are rich in, inhibit the proliferation and progression of hormone-related cancer (Ip, 1997; Rose and Connolly, 1999). On the other hand, omega-6 fatty acids were found to increase synthesis of cyclooxygenase- and lipoxygenase-catalysed products and thus may increase cancer risk (Nair et al, 1997). It has been shown that farm-raised fish are high in omega-6 but low in omega-3 fatty acids (van Vliet and Katan, 1990). The high consumption of fresh water fish in our study population compared to the high intake of marine fish among Western populations may explain the contradictory findings on fish consumption in our study compared to those conducted in Western countries.

Chemical pollution may be another explanation for our finding on fish. Fish growing in polluted water may have a high level of methylmercury, polychlorinated dibenzo-p-dioxins, dibenzofurans, organochlorine residues, and other chemicals. Some of these chemicals have high toxicity and carcinogenic potency (Rojas et al, 1999). Fresh-water fish have been found to be contaminated with methyl mercury, organochlorines, and other chemicals in China (Hou et al, 1988; Nakata et al, 2002). Evidence that organochlorines are present in seafood is also available (Jiang et al, 2005). Several studies conducted in Shanghai, including this one, have consistently found that fish intake is related to an increased risk of several cancers, including cancers of breast, endometrium, and colorectum (Shu et al, 1993; Guo et al, 1994; Dai et al, 2002; Chiu et al, 2003), suggesting that water pollution may be involved in the aetiology of these cancers. Organochlorines have been shown to have estrogenic activity (Olea et al, 1998; Brouwer et al, 1999) and may increase the risk of hormone-related cancers such as prostate and breast cancer (Terry et al, 2003). However, results from two studies that have directly evaluated the association between organochlorines and endometrial cancer found either no significant association (Weiderpass et al, 2000) or a negative association (Sturgeon et al, 1998). Further studies are needed to search for the factors that are responsible for the positive association between fish consumption and cancer risk in this population.

Meat and fish cooked at high temperatures can produce HCAs, potent experimental mutagens, or carcinogens (Dolara et al, 1979; Sinha and Knize et al, 1998; Sinha and Rothman et al, 1998; Wong et al, 2005). Heterocyclic amines begin to form at temperatures of 150°C or higher, and their production can be increased up to three-fold when the cooking temperature is increased from 200 to 250°C (Dolara et al, 1979). Frying, broiling, grilling, and baking are associated with formation of large amounts of HCAs (Sinha and Knize et al, 1998; Sinha and Rothman et al, 1998). In our population, deep-frying is the most common high-temperature method for cooking red meat and fish, whereas roasting or grilling is less common, although it may be used to cook poultry. Oil temperature is normally around 240–270°C during deep-frying. It has been reported that deep-frying also generates fumes containing mutagenic compounds (de Meester and Gerber, 1995; Shields et al, 1995). These mutagens or carcinogens have been linked to an increased risk of cancers among Western populations (Zimmerli et al, 2001; Sinha et al, 2005). However, in our study, we found no association between cooking patterns or doneness levels and the risk of endometrial cancer, nor did they modify the risk associated with meat intake with the exception of the positive association with deep-fried cooking observed among women whose consumption of fish was in the lowest quartile. The latter is likely to be attributed to a chance finding owing to the lack of an underlying biological mechanism and the multiple comparisons conducted in this study. Our findings may reflect the fact that deep-frying, roasting, and grilling are much less common among the Chinese compared to people living in Western countries (Koh et al, 2005). In Shanghai, stir-frying is the most common cooking practice. Thus, the exposure to HCAs in our study population may not have been high enough to increase cancer risk.

Compared with Western populations, our study population has a much lower level of meat consumption. On average, the total meat intake is only about half to one-third of that consumed in Western countries (Sinha et al, 2005; Gonzalez et al, 2006). Our findings provide potent evidence of the unfavourable role of meat intake in the development of endometrial cancer, even at a low average level of consumption, and thus strongly support public health messages suggesting that women should be encouraged to lower their meat intake.

The strengths of our study include a population-based design, a large sample size, high response rates, and the availability of detailed dietary information collected using a validated FFQ. Given the nature of the case–control study design, however, recall bias is a major concern in our study. In this study, the response rate among controls (74.4%) is lower than that in cases (82.8%), raising a concern about selective participation bias. Although neither the potential study participants nor the interviewers knew our study hypotheses, we cannot exclude the possibility that study participation would be influenced by general health knowledge including healthy dietary habits. However, we found that the controls of the current study had patterns of meat and fish intake similar to the controls of the Shanghai Breast Cancer Study which was conducted in the same population 2–3 years before the current study which had a much higher response rate (91.1% for cases and 90.3% for controls) (Dai et al, 2002). Therefore, selection bias is unlikely to be the main explanation for the study findings. The dietary intake of cases may have changed as a result of cancer diagnosis and treatment, and current diet may influence the recall of usual diet (Willett, 1998). In China, cancer patients are commonly advised to avoid or lower meat intake for improving prognosis. Thus, dietary change after cancer diagnosis would probably lead to an underestimation of the association of meat intake with endometrial cancer risk. In our study, the median interval between diagnosis and interview for cases was only 5.6 months. Study participants were asked to report their dietary habits before cancer diagnosis, ignoring any recent dietary changes. This may have minimised the misclassification of the dietary assessment. However, we cannot exclude the possibility that some patients might link meat intake to the cause of their disease and thus the result may have been over-reporting of meat consumption among cases.

In summary, this large population-based case–control study of endometrial cancer, conducted in a population that has a much lower total meat consumption compared to people in Western countries, found that diets high in animal foods increase the risk of endometrial cancer.