No Association of Consumption of Animal Foods with Risk of Ovarian Cancer

A potential role of dietary factors on the risk of ovarian cancer (OVC) has been suggested by ecologic studies due to observed differences in international incidence rates (1). The contribution of dietary factors to the etiology of OVC has been suggested through the modulation of the endogenous hormonal milieu (2, 3) or through antioxidant and anticarcinogenic mechanisms (4). Some case-control studies have suggested that OVC risk is increased with high intakes of fat or dairy products, but the data are inconsistent (5-15). This relates particularly to foods of animal origin and specifically to consumption of fish, dairy products, and meats (16).

Given the paucity of prospective data with a sufficiently large number of cancer cases, we examined animal food consumption as predictors of OVC risk in the large-scale multicenter European Prospective Investigation into Cancer and Nutrition (EPIC) Study.

Details of the EPIC Study have been described in detail elsewhere (17). Briefly, study participants from 10 European countries (Denmark, France, Germany, Greece, Italy, the Netherlands, Norway, Spain, Sweden, United Kingdom), mostly from the general population, were recruited into the study between 1992 and 2000 (366,521 women; 153,521 men). For the present study, females free of any cancer at baseline, with at least one intact ovary, and with non-missing dietary and follow-up information have been included (n = 325,731). All participants signed an informed consent agreement at enrollment. A detailed description of this study population can be found in ref. 18.

At baseline recruitment, habitual diet of the past 12 months was assessed by means of country-specific food frequency questionnaires or diet histories. Foods of animal origin examined in the present study were total meat, fish, eggs and total dairy products and selected subgroups of meat (red meat, poultry, processed meat) and dairy products (milk, yogurt, cheese).

In EPIC, case ascertainment was based upon linkage to cancer registries or active follow-up. To classify ovarian tumors, International Statistical Classification of Diseases and Related Health Problems 10th Revision (ICD-10) (code C56) and International Classification of Diseases O-2 were used. As of April 2004, 620 OVC cases have been reported to the common database at IARC, Lyon. Of those, 581 were primary malignant cancers used for the analysis. Histologic subtype was specified for 61%.

To make dietary exposures comparable across participating countries, dietary intakes were calibrated using a fixed-effects linear model in which center and gender-specific 24-h recall data from an 8% random sample of the total cohort (19) were regressed on questionnaire intakes controlling for covariates (20). Cox's Proportional Hazards models were used to evaluate the association between animal food consumption and OVC occurrence. The models were stratified by study center to control for (unmeasured) center effects. Age was used as the primary time variable with the subjects' age at recruitment as entry time and the subjects' age at diagnosis or censoring (death, emigration, or last complete follow-up) as exit time. Models were controlled for body mass index; total energy intake (continuous); parity (parous, nulliparous); ever use of oral contraceptives; hormone replacement therapy (yes, no, unknown); menopausal status (pre-, postmenopausal, not defined); education (three categories); smoking (never, ever, unknown); and unilateral ovariectomy (yes, no). All statistical tests were two-sided, and a P value <0.05 was considered statistically significant. We calculated a power of 95% to detect a significant hazard ratio (HR) of ≥1.5 for the highest versus the lowest quintile (α = 0.05; ref. 21).

Baseline characteristics of the study population can be found in Table 1. We observed no significant association between the major animal food groups (total meat, eggs, fish, total dairy products) and risk of OVC, neither with the quintile analysis nor with the linear analysis (Table 2). In addition, meat subgroups (red meat, poultry, processed meat) and dairy products (milk, yogurt, cheese) did not show any relationships with incident OVC (Table 2). Further adjustment for fruit and vegetables or other animal products made little difference to these estimates (data not shown). We found no evidence for effect modification by menopausal status, ever oral contraceptives use, and baseline hormone replacement therapy use for any of the animal foods. Nulliparous women seemed to benefit from a high consumption of total dairy products [HR, 0.37; 95% confidence interval (95% CI), 0.14-0.97, per increment of 39.4 g/day (1 SD)] compared with parous women (HR, 1.01; 95% CI 0.69-1.47; P_interaction = 0.0025); however, none of the dairy subgroups nor other animal foods showed significant associations. Histology-specific [serous (n = 228), mucinous (n = 51), endometrioid tumors (n = 56)] models yielded mostly nonsignificant risk estimates except for associations between serous tumors, total meat and poultry (HR, 1.27; 95% CI, 1.02-1.60; and HR, 1.31; 95% CI, 1.07-1.61 per increment of 1 SD in intake, respectively; data not shown).

Exclusion of women who were diagnosed within 1 year of recruitment (n = 81) did not materially change these associations.

The present study on >325,000 European women does not provide evidence for an association between consumption of animal foods (meat, fish, eggs, dairy products) and risk of OVC. Although cohort evidence is still limited, a direct association between meat consumption and OVC risk has been suggested in several case-control studies (10, 12, 15, 22-24). Egg consumption has been related to OVC in most of the cohort studies (25-27) but not all (28). To date, there is only one prospective analysis of fish consumption and OVC risk reporting a null finding (28), whereas an inverse association was indicated by several case-control studies (8, 22, 29). With respect to dairy foods, study results are mixed (9, 12, 14, 15, 22, 23, 26, 30-33).

To our knowledge, this is the largest prospective study to report on a variety of animal foods in relation to OVC risk. Apart from its large sample size, its specific strength is the wide variation in food consumption due to the multicenter design of EPIC. Limitations of the study include the potential of misreported food consumption, which could have obscured weak associations.

In conclusion, in the present study, we found no evidence of a significant association between animal food consumption and OVC risk. Our findings from subgroup analyses (parous versus nulliparous women; histology-specific analysis) need confirmation in future studies because the number of cases per subgroup was relatively small, and we cannot rule out that these findings might have occurred by chance.