Intake of Folate and Related Nutrients in Relation to Risk of Epithelial Ovarian Cancer

Abstract

Assessments of the relation between folate intake and ovarian cancer risk have been limited and inconsistent. Therefore, the authors prospectively examined the association of dietary and supplemental intakes of folate, methionine, and vitamin B₆ with ovarian cancer risk among 80,254 Nurses' Health Study participants. Beginning in 1976, women completed biennial questionnaires assessing ovarian cancer risk factors; starting in 1980, food frequency questionnaires were administered every 2–4 years. During 22 years of follow-up (1980–2002), the authors confirmed 481 incident epithelial ovarian cancers. There were no associations between total folate (top quintile vs. bottom: relative risk (RR) = 1.21, 95% confidence interval (CI): 0.92, 1.60), methionine (RR = 1.00, 95% CI: 0.76, 1.33), dietary vitamin B₆ (RR = 1.09, 95% CI: 0.81, 1.47), or total vitamin B₆ (RR = 1.13, 95% CI: 0.85, 1.51) intake and ovarian cancer risk. Higher dietary folate was associated with a modestly decreased risk after exclusion of cases diagnosed during the 4 follow-up years after dietary assessment (RR = 0.66, 95% CI: 0.43, 1.03) and for the serous subtype (RR = 0.51, 95% CI: 0.31, 0.84). Results did not vary by alcohol intake, multivitamin use, menopausal status, or oral contraceptive use. There was little evidence that folate, methionine, and vitamin B₆ are important in ovarian cancer risk, although dietary folate was inversely associated with risk in some analyses.

Folate, methionine, and vitamin B₆ are nutrients that are important in DNA synthesis and methylation. Folate is required for the synthesis of purine and thymidylate (1), and both low folate intake and low vitamin B₆ intake can result in inadequate levels of 5,10-methylenetetrahydrofolate, leading to abnormal DNA synthesis and repair (2, 3). Methionine is the ultimate methyl donor for DNA methylation, and although humans have de novo methionine synthesis, dietary intake is required for maintenance of adequate levels (4). Low nutrient intakes, especially when combined with a high intake of alcohol (a folate antagonist), may produce a “methyl deficiency” (5, 6). Such a deficiency can lead to global DNA hypomethylation and chromosomal instability, which may be important in carcinogenesis (7).

Low dietary intakes of folate, methionine, and vitamin B₆ have been associated with an increased risk of several cancers, including breast, colon, and prostate cancer (8–13). However, few epidemiologic studies have addressed these associations with regard to ovarian cancer risk. Case-control studies generally have not reported an association (14, 15). However, after conducting a prospective evaluation of dietary folate intake and ovarian cancer in the Swedish Mammography Cohort, Larsson et al. (16) reported a significant inverse association that was strongest among alcohol drinkers. Conversely, in the Iowa Women's Health Study, Kelemen et al. (17) reported a nonsignificant positive association between total folate intake and ovarian cancer, although intake was inversely associated with risk among alcohol drinkers.

Because of the potential biologic importance of the folate pathway in carcinogenesis and a desire to address the somewhat disparate results between the above two prospective studies (16, 17), we examined whether intakes of folate, methionine, and vitamin B₆, from both the diet and supplements, were prospectively associated with ovarian cancer risk in the Nurses' Health Study. We also examined whether this relation varied by alcohol consumption, menopausal status, multivitamin use, oral contraceptive use, or histologic subtype.

MATERIALS AND METHODS

Study population

The Nurses' Health Study cohort was established in 1976 when 121,701 US female registered nurses aged 30–55 years completed and returned a mailed questionnaire. The Nurses' Health Study cohort has been followed by questionnaire every 2 years since, to update exposure variables and ascertain newly diagnosed disease. Data have been collected on numerous ovarian cancer risk factors, including age at menarche, parity, oral contraceptive use, tubal ligation, age at menopause, postmenopausal hormone use, and family history of ovarian cancer. In 1980, we included a 61-item food frequency questionnaire (FFQ) that asked about the use of vitamin and mineral supplements; the questionnaire was expanded to 131 items in 1984. Participants completed the expanded FFQ in 1986, 1990, 1994, and 1998. We used the 1980 FFQ in this analysis because the women were asked about primary sources of folate intake (e.g., breakfast cereal, orange juice, and lettuce) on that questionnaire. Follow-up for women completing the 1980 FFQ was 97.5 percent of person-years through May 31, 2002. The racial/ethnic profile of this population was 97 percent Caucasian, 2 percent African-American, and 1 percent Asian; 1 percent of the women reported being of Hispanic origin.

Assessment of folate, methionine, and vitamin B₆

Details on the semiquantitative FFQ and its reproducibility have been published previously (18–21). Briefly, for each food and beverage item, the questionnaire specifies a common serving size and asks participants to indicate their average intake of that food item during the past year (responses range from “almost never” to “six or more times per day”). Additionally, on the 1980 FFQ, we asked women whether their intake of each food product had greatly increased or decreased over the past 10 years. We calculated intakes of dietary folate, methionine, and vitamin B₆ by multiplying the frequency of food item consumption by the nutrient content of the specified serving size, using food composition values from the US Department of Agriculture (22–24). All nutrient intakes were adjusted for total energy intake using the nutrient residual method (25). If a woman did not provide intake information for a particular food item in a certain year, that food was not included in the calculation of her intake of that nutrient in that year; this was rare, since we only included FFQs that had missing data for 10 or fewer food items. Participants also provided information on the dose and duration of any vitamin supplements used, including brand of multivitamin. We used this information to calculate total folate and vitamin B₆ intakes from both diet and supplements. Other dietary factors, including lactose intake, total calories, and alcohol consumption, were assessed similarly.

In a validation study using two FFQs administered 1 year apart and four 1-week diet records completed 3 months apart, Pearson coefficients for correlation between the FFQs and the diet records were 0.71 for folate, 0.55 for vitamin B₆, and 0.90 for alcohol (18–20). Comparable data were not available for methionine; however, correlations between the FFQ and the diet records were good (0.43–0.56) for the major sources of methionine, such as poultry, fish, red meats, and dairy products. Among 188 Nurses' Health Study women, the correlation between total folate intake calculated from the 1980 FFQ and erythrocyte folate concentration measured in 1987 was 0.55 (26).

Other study data

The women were asked about menopausal status, hysterectomy, oophorectomy, and postmenopausal hormone use during each questionnaire cycle. Women were asked for information on some factors during part of the follow-up period; we updated information on these factors until the questions were no longer asked, at which point the last response given was carried forward until the end of follow-up. For example, the women were asked about oral contraceptive use until 1982, when oral contraceptive use became rare because of the age distribution of the participants. Parity was assessed until 1984, and a question on history of tubal ligation was asked in 1976, 1978, 1980, 1982, and 1994. Data on age at first birth and year of birth were collected on the 1976 baseline questionnaire. Women were not asked about a family history of ovarian cancer until 1992; they were then subsequently queried about this in 1996 and 2000.

Ascertainment of ovarian cancer cases

Incident cases of epithelial ovarian cancer were identified by biennial questionnaire from 1980 to 2002. For women reporting a new case of ovarian cancer, we obtained pathology reports and related medical records to confirm the cancer diagnosis. A gynecologic pathologist (J. H.) who was unaware of the women's exposure status reviewed the records to confirm the diagnosis and identify the histologic type, subtype, morphology, and stage. Deaths occurring in the cohort were identified via family members, the US Postal Service, and the National Death Index, which captures 98 percent of all deaths in this cohort (27, 28); medical records and pathology reports were obtained and reviewed for these cases as well. In a subset of 215 ovarian cancer cases, we compared the histologic type recorded in the pathology report with a standardized review of pathology slides completed by one of the authors (J. H.). Overall, the concordance for invasiveness was 98 percent and the concordance for histologic type was 83 percent; histologic type from the medical record review was used for all cases.

Statistical methods

We excluded women at baseline if they did not complete the 1980 FFQ (n = 23,699); if they had implausibly high or low dietary intakes (<500 kcal/day or >3,500 kcal/day) (n = 5,579); if they reported a diagnosis of cancer, except nonmelanoma skin cancer, before 1980 (n = 3,660); or if they had a history of bilateral oophorectomy or pelvic irradiation (n = 8,509). Person-years of follow-up were accrued from the date of return of the 1980 questionnaire to the diagnosis of ovarian or other cancer (except nonmelanoma skin cancer), report of a bilateral oophorectomy or pelvic irradiation, death, or the end of the follow-up period on June 1, 2002—whichever came first.

We represented intakes of folate, methionine, and vitamin B₆ in two ways. First, we used the baseline intake reported in 1980, because past intake may be most related to ovarian cancer risk. Second, we used the cumulative average of intakes reported on all previous FFQs; this approach dampens variation due to measurement error and true changes in diet (29). For example, we used averaged folate intake from 1980 and 1984 to predict ovarian cancer risk from 1984 to 1986; similarly, we used averaged folate intake from 1980, 1984, and 1986 to predict risk from 1986 to 1988.

Participants were divided into quintiles of nutrient intake (see Web table 1, which is posted on the Journal's website (www.aje.oxfordjournals.org); e.g., quintiles for baseline dietary folate were ≤178, >178–220, >220–262, >262–323, and >323 μg/day). We also created a methyl group score (6). A high methyl group score was defined as alcohol intake <5 g/day and either folate intake or methionine intake in the top tertile; a low score was defined as alcohol intake ≥10 g/day and either folate intake or methionine intake in the bottom tertile; and all other scores were defined as intermediate. Person-time was allocated to each category of nutrient intake in 2-year increments, allowing participants to change exposure status every 2 years. For years in which the dietary questions were not included in the study questionnaires (1982, 1988, 1992, 1996, and 2000), the participant's response from the previous questionnaire was carried forward.

We calculated incidence rates for each category of nutrient intake by dividing the number of incident ovarian cancer cases by the total person-time in that category. We used Cox regression with time-dependent covariates to estimate relative risks and 95 percent confidence intervals, using women in the lowest intake category as the reference group. In tests for trend, we used nutrient quintile medians. We adjusted for the following potential confounders a priori: lactose intake (in quintiles), duration of oral contraceptive use (never, ≤3 years, >3–5 years, >5–8 years, or >8 years), age at first birth/parity (nulliparous, age at first birth <25 years and 1–3 children, age at first birth 25–29 years and 1–3 children, age at first birth ≥30 years and 1–3 children, age at first birth <25 years and ≥4 children, age at first birth ≥25 years and ≥4 children), postmenopausal hormone use (never use, past use, current use, premenopausal or unknown menopausal status, missing data), history of tubal ligation (yes, no), simple hysterectomy (yes, no), and total caloric intake (in quintiles). We also considered adjustment for other putative ovarian cancer risk factors, including age at menarche, current body mass index (weight (kg)/height (m)²), body mass index at age 18 years, height, and physical activity; however, since adjustment for these variables did not alter the risk estimates, they were not included in the final model. We evaluated potential confounding by the other nutrients; the correlation between nutrient intakes was 0.77 for total folate and total vitamin B₆, 0.73 for dietary folate and dietary vitamin B₆, and 0.16–0.50 between other dietary and total nutrient intake levels (not including correlations between dietary and total intake of the same nutrient). To do this, we included all dietary nutrients (folate, methionine, and vitamin B₆) together in one model; analyses of total folate included additional adjustment for dietary vitamin B₆ and methionine intake, and analyses of total vitamin B₆ included additional adjustment for dietary folate and methionine intake.

In secondary analyses, we excluded cases diagnosed during the 4 years of follow-up after dietary intake assessment. When examining baseline intake, we restricted the analysis to women who indicated on the baseline (1980) FFQ that their intake of the primary food groups comprising the nutrients of interest had not changed substantially over the previous 10 years. We also evaluated the relations between nutrient intakes and specific subtypes of ovarian cancer, including invasive types only and serous/poorly differentiated and endometrioid subtypes; there were too few mucinous and clear-cell cases for separate consideration. To better understand the impact of supplemental folate, we considered the association of ovarian cancer with multivitamin use (the primary source of supplemental folate in our population), examining updated current use, updated duration of use, and the combination of current/past use with duration (never use, past use for <5 years, past use for ≥5 years, current use for <5 years, current use for ≥5 years). Finally, we conducted stratified analyses according to potential effect modifiers, including alcohol intake (<5 g/day vs. ≥5 g/day), menopausal status (premenopausal vs. postmenopausal), multivitamin use (yes vs. no), oral contraceptive use (never vs. ever), and estimated lifetime number of ovulations (<360 vs. ≥360) (30). All p values were two-sided.

RESULTS

During 22 years of follow-up, we accrued 1,492,550 person-years for 80,254 women. Throughout the follow-up period, we identified 827 potential cases of ovarian cancer, of which 85 percent were self-reported and 15 percent were identified by death certificate. Fifty of these women (6 percent) subsequently denied the diagnosis, and 121 women (15 percent) had a different type of cancer, had metastasis from another primary cancer site, or were found not to have cancer upon medical record review. We were unable to obtain medical records for 65 women (8 percent). Of the 591 confirmed cases, 528 (89 percent) were primary epithelial ovarian cancers; 47 of these women had received a diagnosis of another type of cancer (except nonmelanoma skin cancer) prior to their ovarian cancer diagnosis and were excluded, leaving 481 cases (421 invasive) for analysis.

Table 1 presents the characteristics of the eligible cohort members in 1990, which is the approximate midpoint of the follow-up period, by quintile of dietary folate intake in 1990. Women in the highest quintile were older, more physically active, and more likely to be using postmenopausal hormones and had a slightly lower body mass index than women in the lowest quintile. Other nondietary characteristics did not vary substantially across quintiles. Less than 1 percent of women took folate-only supplements from 1980 to 1994; this figure increased to 5 percent in 1996 and 1998 and 10 percent in 2000. Trends in these characteristics across quintiles of total folate intake (e.g., including supplements) were similar (data not shown).

We did not find a statistically significant association between folate, methionine, or vitamin B₆ intake or methyl group score and risk of epithelial ovarian cancer using baseline (1980) diet or cumulatively updated diet (table 2). However, there was a suggestion of an inverse association with increasing dietary folate intake (for the top quintile vs. the bottom quintile, relative risk (RR) = 0.76, 95 percent confidence interval (CI): 0.52, 1.12; p-trend = 0.30) after results were adjusted for other dietary nutrient intakes. Conversely, there was a modestly higher risk with increasing total folate intake, with a comparable relative risk of 1.21 (95 percent CI: 0.92, 1.60; p-trend = 0.05); this risk was attenuated after adjustment for intake of other nutrients. Associations were similar when baseline analyses were restricted to women who reported having had a stable intake of the primary food groups contributing to nutrient intakes over the previous 10 years (data not shown). We found no association between major food contributors to folate intake (spinach, cereal, and orange juice) and epithelial ovarian cancer risk (data not shown). There was no association between multivitamin use and ovarian cancer risk; for example, the relative risk for current use of >5 years' duration, as compared with never use, was 1.06 (95 percent CI: 0.83, 1.35).

The associations between nutrient intakes and ovarian cancer risk did not vary by alcohol consumption (<5 g/day vs. ≥5 g/day) (table 3) or when we considered cumulatively updated nutrient intake or stratified the data by alcohol consumption of <10 g/day versus ≥10 g/day (data not shown). However, we did observe that among women consuming ≥5 g/day of alcohol, the highest quintile of total folate intake was associated with a modestly increased risk of epithelial ovarian cancer (RR = 1.50, 95 percent CI: 0.94, 2.39; p-trend = 0.02). Associations did not vary by menopausal status, multivitamin use, oral contraceptive use, or estimated lifetime number of ovulations (data not shown).

Increasing dietary folate intake was associated with a modestly decreased risk when we excluded cases (n = 75) diagnosed during the 4 years of follow-up after dietary intake assessment (table 4). For baseline diet, the relative risk comparing the top quintile with the bottom quintile was 0.75 (95 percent CI: 0.49, 1.14; p-trend = 0.37), and for cumulatively updated diet, the corresponding relative risk was 0.66 (95 percent CI: 0.43, 1.03; p-trend = 0.09). We found no other associations in the lag analyses.

Associations from analyses including only cases of invasive epithelial ovarian cancer (n = 421) were similar to the overall results (data not shown). However, some associations appeared to differ by endometrioid (n = 71) or serous/undifferentiated (n = 299) subtypes (table 5). After adjustment for other dietary factors, the relative risk comparing the top quintile of dietary folate intake with the bottom quintile was 0.51 (95 percent CI: 0.31, 0.84; p-trend = 0.01) for serous tumors and 1.82 (95 percent CI: 0.68, 4.86; p-trend = 0.11) for endometrioid tumors. Conversely, after adjustment for other dietary factors, increasing dietary vitamin B₆ intake was associated with a modestly increased risk of serous tumors (p-trend = 0.08) but a modestly decreased risk of endometrioid tumors (p-trend = 0.13).

DISCUSSION

To our knowledge, this is the largest prospective study to date examining the relation between folate and related nutrients and the risk of primary epithelial ovarian cancer. We found no associations between folate, methionine, and vitamin B₆ and risk of ovarian cancer either overall or when results were stratified by alcohol consumption. However, there was some suggestion that dietary folate intake may be protective, particularly for serous tumors.

Biologic data indicate that folate is important in maintaining the dual processes of DNA nucleotide synthesis and methylation (1). Folate deficiency has been associated with global DNA hypomethylation and chromosomal instability (7). Although we found no association with folate intake and overall ovarian cancer risk, there appeared to be a possible protective effect of dietary folate when we excluded cases diagnosed during the 4 years of follow-up after dietary intake assessment, suggesting that folate intake may not be protective among women with underlying subclinical disease. This is consistent with evidence that ovarian tumors have upregulated expression of folate receptor α (31, 32), which is associated with a higher tumor stage and grade (31, 33, 34). Furthermore, animal data suggest that the effect of folate on carcinogenesis may depend on the dose and timing of the exposure, with folate deficiency enhancing carcinogenesis in normal cells but folate supplementation enhancing progression of existing tumor cells (35).

Although our results were not statistically significant, the relative risks were similar to the overall association for dietary folate observed in the Swedish Mammography Cohort (for the top quartile vs. the bottom quartile, RR = 0.67, 95 percent CI: 0.43, 1.04; p-trend = 0.08) (16). However, in the Iowa Women's Health Study, Kelemen et al. (17) reported a nonsignificantly increased risk with higher dietary folate (for the top quartile vs. the bottom quartile, RR = 1.45). This relative risk was attenuated after exclusion of women diagnosed during the first 2 years of follow-up, suggesting that the increased risk may be due to folate exposure in women with underlying subclinical disease.

Interestingly, in the Iowa Women's Health Study, Kelemen et al. (17) observed a modest, nonsignificantly increased risk (RR = 1.73) of ovarian cancer among women in the highest quartile of total folate intake (e.g., diet plus supplements); the Swedish study did not have detailed information about supplement use (16). We also observed a modestly increased risk among women at the highest levels of total folate intake (RR = 1.21 overall). Although this seems to contradict the dietary folate results, the levels in the fifth quintile of dietary folate are similar to the levels in the fourth quintile of total folate intake (see Web table 1), and the relative risks, at least for serous tumors, are similar in these two analyses (for quintile 5 dietary intake, RR = 0.51, and for quintile 4 total intake, RR = 0.69). It is possible that very high levels of folate may promote tumorigenesis (35), such that women with high levels of supplemental folate may have a slightly increased risk of ovarian cancer, especially since folate from supplements (e.g., folic acid) is absorbed with almost twice the efficacy of dietary folate (36). However, more epidemiologic and animal data are needed to further explore this hypothesis. Alternately, dietary intake may reflect long-term intake better than dietary plus supplemental intake, or another nutrient or set of nutrients found in folate-rich foods may be associated with ovarian cancer risk.

Unlike the previous prospective studies (16, 17), we did not observe that the association between dietary or total folate intake and ovarian cancer varied by alcohol consumption. Both previous prospective studies reported that increased dietary (16) or total (17) folate intake was protective among women consuming alcohol. The cutpoints defining alcohol drinkers were similar in all three studies (this study: 5 g/day; Larsson et al. (16): 3 g/day; Kelemen et al. (17): 4 g/day). Different populations and sample sizes may explain the discrepancies between studies. The Swedish population had a much lower intake of folate than women in our study. It is possible that we did not observe the folate-alcohol interaction because such an interaction may only be apparent for women with the lowest levels of folate intake. Our first quintile of intake (≤178 μg/day) incorporated the first two quartiles of intake in the Swedish study, and our second quintile (>178–220 μg/day) incorporated the top two quartiles in that study. However, we conducted an analysis comparing women in the first (≤151 μg/day) and second (>151–178 μg/day) deciles of intake to the remaining women and found no associations (data not shown). Previous studies also had smaller sample sizes, possibly leading to different associations. For example, in the Iowa study, there were only seven ovarian cancer cases in the high alcohol/high folate group (17). Given that the folate-alcohol interaction has been observed for other cancers, including colon cancer and breast cancer (8–13), it is important to further investigate this association in ovarian cancer.

We found few associations between methionine or vitamin B₆ intake and risk of epithelial ovarian cancer. One prospective study also reported no significant associations with methionine intake, even after stratification by folate intake (17). A case-control study examined vitamin B₆ and found no association with ovarian cancer (14). We also did not observe any association between multivitamin intake and risk of ovarian cancer. This is consistent with previous Nurses' Health Study analysis with follow-up from 1980–1996 (37) and a Canadian case-control study (38).

We found some evidence that the associations for folate and other nutrients differed by histologic subtype. For example, high dietary folate intake was associated with a decreased risk of serous tumors but possibly an increased risk of endometrioid tumors; in the Swedish study, however, Larsson et al. (16) reported that the decreased risk for dietary folate intake appeared to be strongest for endometrioid tumors. Interestingly, there appeared to be a slightly increased risk of both tumor subtypes at the highest level of total folate intake. Although these results highlight the importance of examining risk factors separately for various histologic subtypes, the potential underlying biologic mechanisms are unclear. Given the somewhat small numbers of cases in these analyses, the results should be interpreted with caution.

This study had several strengths and weaknesses. We had a large number of cases for the primary analyses, which increased our power to detect statistically significant associations, but our power to evaluate specific histologic subtypes was somewhat limited. We also had repeated dietary and supplement intake assessments, which allowed us to minimize random within-person variation in measurement of food and nutrient intake (39). Folate as measured by this FFQ had a reasonable correlation with diet records (correlation = 0.71), although our results may have been slightly attenuated. However, folate has previously been associated with coronary heart disease (40), hypertension (41), breast cancer (8, 42), colon cancer (43), and colon adenomas (26, 44) in the same cohort, suggesting that the lack of an observed association with ovarian cancer is not likely to be a result of exposure misclassification. One limitation of this study is that most women had an adequate intake of the dietary nutrients studied. It is possible that risk of ovarian cancer is seen only at very low intakes of folate. We constructed a methyl group score (6), which integrated folate, methionine, and alcohol intakes, to better identify people with the most harmful intake profile; however, we found no associations between this score and ovarian cancer risk.

In conclusion, our results suggest that nutrients in the one-carbon metabolism pathway may not play an important role in the etiology of ovarian cancer. However, in several subset analyses with a 4-year lag and histologic subtypes, there was a suggestion of an inverse association with dietary folate intake. We did not find the interaction between folate intake and alcohol consumption that was reported in two previous prospective studies (16, 17). Because of the potential importance of this interaction, the possible reasons for the discrepancies between studies should be examined further. Although the data are limited and not entirely consistent, evidence from our study and others, in conjunction with biologic data, is compatible with the hypothesis that very low folate intake (e.g., deficiency) may increase ovarian cancer risk, possibly early in the carcinogenic pathway, but very high folate intake late in carcinogenesis may be unlikely to lower risk. Since few lifestyle risk factors have been identified for ovarian cancer, further assessment of these associations is warranted.

This project was supported by National Institutes of Health grants P01 CA87969 and P50 CA089393. Dr. Shelley Tworoger was partially supported by a training grant in cancer epidemiology (T32 CA090001) from the National Cancer Institute.

The authors thank Drs. Graham Colditz and Walter Willett for their helpful comments on the manuscript.

Conflict of interest: none declared.

References