Dietary Intake and Ovarian Cancer Risk: A Systematic Review

Discussion

This systematic review of 24 large prospective studies, reflecting a much larger number of prospective studies than the earlier review by Schulz and colleagues (25), replicate some, but not all of the associations identified in the earlier review. Here we identified vegetables as protective in one study with point estimates <1.0 in the other 2 studies, consistent with earlier estimates. Specific constituents in vegetables such as isoflavones were also suggested to be associated with lower risk in this review, but were not specifically evaluated in the earlier report. We did not find a significant association with meat intake as reported in the earlier review, but did find evidence of greater risk in relation to nitrites in meat. Although the earlier review suggested risk may be lower in relation to whole grains and low fat milk, this review found no significant reduction in risk with these dietary exposures. Importantly, the large number of studies with nonsignificant and inconsistent findings suggests that either diet is not associated or only modestly associated with ovarian cancer risk or that studies may lack sufficient sample size or accuracy in dietary measurement to identify consistent associations.

Dietary fat and animal fat as well as animal-source nitrite intake are shown to be associated with a higher risk for ovarian cancer in this review, and are supported by mechanistic research. In fact, dietary fat, particularly of animal origin, has been linked to tumor promotion (49), inflammation (50), and elevated estrogen exposure (51), all of which would contribute to ovarian cancer risk. Furthermore, a secondary analysis of the WHI low-fat dietary intervention suggested low-fat diets followed for greater than 4.1 years may lower ovarian cancer risk in postmenopausal women (4). The lack of consistency in the epidemiological studies may reflect variance in the range of fat intake reported. In the AARP study (30), which showed an elevated risk related to total fat intake, high intake was characterized as > 39.9% total energy as fat, an amount slightly greater than the mean intake of 36% of total energy in the California's Teacher Study (20). Furthermore, the specific contribution of saturated versus monounsaturated fats was not evaluated in both cohorts and may vary for women residing in California versus older women residing throughout the United States. Similarly, the increased risk related to dairy shown in the Swedish Mammography cohort may also be a result of higher dairy fat intake reported by the Swedish women (33). Specifically, the upper quintile for the Swedish cohort was ≥4 servings/day, whereas in the European cohort highest intake is estimated at 209 g/day equivalent to approximately a single serving (1 cup) of milk. The AARP is the only cohort study to evaluate nitrates and nitrites in relation to ovarian cancer risk. The higher risk associated with animal-based nitrites could reflect correlated dietary exposures (dietary fat) or may suggest an alternate mechanism of carcinogenesis, such as the formation of nitrites from N-nitrosamine metabolism (52). Efforts to test these associations in other cohorts are needed.

In relation to other sources of dietary fat, 2 meta-analyses and a pooled analysis of prospective cohorts (similar to studies selected in this systematic review) have assessed the relationship between dairy, milk, or related diet exposures and ovarian cancer risk. The results, generally consistent with this systematic review, suggested lactose may be associated with higher risk (53, 54) whereas milk/dairy is not (54, 55). Of note, whole milk and butter, significant sources of dietary fat, were reported to be associated with higher risk for ovarian cancer in the meta-analysis by Qin (55), consistent with the association shown for total dairy intake in the Swedish Mammography cohort described here (33).

Large variance in vegetable exposure estimates (80 g for total vegetables vs. 8 g for selective vegetables such as leafy vegetables) may contribute to the inconsistency in the risk associations related to vegetable intake. Alternatively, select types of vegetables and the related target anticarcinogenic bioactivity of constitutive compounds within vegetables may offer differential risk reduction as is suggested by lower risk in relation to allium vegetable intake in the EPIC cohort (36) in comparison to null associations between isothiocyanate intake and risk in the California Teachers Study (20). Point estimates for risk in relation fruit intake here (35–37) and in the earlier review (27) suggest potential higher risk for ovarian cancer. Although not clearly understood, elevated risk may reflect a propensity for elevated insulin levels with higher glycemic index, particularly if women are older and/or overweight/obese. This hypothesis is supported by results from the NBSS wherein total sugars may be associated with greater ovarian cancer risk (46) and yet, conflict with results from the AARP cohort suggesting that total sugar intake as well as fructose intake (found in fruit and fruit juices) were associated with a lower risk of ovarian cancer (45). Of note, total sugar intake at the upper level in both cohorts was approximately 95 g/day, body mass indexes were similar across studies (approximately 26 kg/m²) and the AARP cohort was only slightly older at 61.5 years versus 58.9 years in the NBSS. Certainly, measurement error variance across instruments is one possible explanation; efforts to assess measurement error using objective urinary measures in sugar intake are underway. Our findings suggesting a lack of consistent associations between total vegetable and fruit intake and risk are supported by a pooled analysis of 12 prospective studies that included 2,130 cases of invasive ovarian cancer (56).

Vitamin C intake and ovarian cancer risk suggested no significant association in 2 studies (19, 39), but one study showed a significantly higher risk (20); all 3 studies showed point estimates greater than 1.0 and the Chang study, while showing significant relative risk, also reported a nonsignificant P_trend. Differences in mean vitamin C intakes ranging within the upper quintiles from 130 mg/day (19) to >665 mg/day (20) were reported and may suggest that supplemental vitamin C versus dietary may be contributing to risk in the Chang study as intake above 200 mg/day would be difficult to achieve with dietary sources alone. The relationship between nutrient supplementation and a possible increase in cancer risk related to pro-oxidant properties has been demonstrated for other nutrients such as vitamin E (57) and β-carotene (58), and has also been suggested in relation to supplemental vitamin C (59).

Finally, our results suggest that tea intake may be associated with lower risk of ovarian cancer (42), although not consistently (48), whereas coffee and caffeine showed no association (47). A meta-analysis that included 1,244 cases of epithelial ovarian cancer did not support a lower risk with either coffee or tea intake (48).

Our systematic review has limitations. First, the lack of consistency in risk associations makes it difficult to define conclusively the role of specific dietary choices in modifying ovarian cancer risk. Furthermore, all the studies evaluated used food frequency questionnaires (FFQs) to evaluate self-reported diet. Measurement error is a well-known problem for FFQs and may have biased toward the null, a common finding here (60). None of the FFQs corrected for measurement error using objective measures. All studies used baseline diet to assess ovarian cancer risk over time; repeated measures or assessment of change in diet was not included in any of the studies. Use of repeat diet measures can be problematic in that cases occurring before (or even shortly after) diet assessments are collected would need to be excluded thus reducing sample size and statistical power to test these associations. However, repeatedly measuring diet over time may provide more accurate assessment of dietary exposures. In addition, for some dietary exposures there may be a threshold that must be reached to demonstrate a lower or a higher risk of disease. For select populations, these lower or upper thresholds for disease modulation may not be achieved. Furthermore, looking across populations internationally, the quantity as well as the nutrient or bioactive density of the food items may vary widely, further contributing to the inconsistent associations demonstrated. Finally, residual confounding cannot be ruled out. For example, higher intake of vegetables and fruit has been associated with higher self-reported physical activity and also lower rates of overweight/obesity. The selected studies controlled for confounding and mediating variables, but not consistently.

Few of the studies had adequate sample size to evaluate relationships for specific foods or specific types of fruits, vegetables, etc. In addition, data evolving for breast cancer systematic reviews and meta-analysis suggest that tumor subtype may be a pertinent factor to assess in relation to diet–disease associations. Here, 13 studies provided some data regarding specific associations for tumor subtypes; however, overall the confidence intervals were wide because of small subgroup sample size. The NHS evaluated risk by tumor subtype, in this case for lactose intake, and found a 2-fold higher risk in women in the highest quintile of lactose intake for serous tumors only (34) and kaempferol intake (42)—a bioactive primarily derived from intake of broccoli and tea was associated with a 43% reduction in risk for serous tumors. The EPIC cohort suggested that both total meat and poultry intake were associated with higher risk for serous tumor subtypes (12) whereas the Swedish Mammography cohort similarly found higher risk for serous subtype (33), but in this case related to total dairy and milk intake, associations not seen in the Netherlands cohort (32). Evidence from the AARP study cohort suggested that total fat was associated with a 14% higher risk for serous tumors; a nonsignificant elevated risk for mucinous tumors was also demonstrated (RR = 1.42; 95% CI, 0.93–2.17; ref. 30).

Strengths of our systematic review include stringent inclusion criteria to evaluate only prospective studies and those with a larger number of ovarian cancer cases. The significant associations identified were relatively weak, but relevant in terms of modifying the risk of a cancer that carries with it a high mortality rate. We included studies that evaluated more common diet exposures as well as several less common exposures such as lactose, bioactive compounds, and beverage consumption.

In conclusion, there is a need for additional studies to evaluate the role of diet in ovarian cancer risk, including pooled studies of sufficient sample size to assess associations by tumor subtype as well as randomized, controlled trials of dietary exposures. The findings from this review suggest the potential for a higher risk of ovarian cancer in women consuming higher amounts of dietary fat, animal products (including dairy and nitrites), and possibly vitamin C or fruit; alternately in women who consume higher amounts of isoflavones, tea, and possibly vegetables, risk may be reduced, although the dietary exposures evaluated seem to be weak-moderate and not consistently demonstrated.