The Child Health and Development Studies (CHDS) cohort, initiated in 1959, recruited nearly all pregnant women (over 98%) receiving prenatal care from the Kaiser Foundation Health Plan in the San Francisco East Bay area. In all, 15,528 women were enrolled during the 7-year recruitment period from 1959 until 1966, with deliveries extending into 1967.¹⁷ The CHDS cohort incorporated multiple race/ethnic groups with a broad socio-economic base and uniform access to health care. Pre-existing maternal conditions and clinical measures (e.g., prenatal and obstetric) were abstracted from medical records. Demographic information and health-related behavior were collected from in-person interviews for all pregnancies, generally early in the first trimester. The institutional review board of the Public Health Institute approved the study protocol and informed consent was obtained from all participants.

Prospective surveillance has continued for 5 decades by annual linkage to: i) the California Department of Motor Vehicles, for a history of residence to identify the population at risk for morbidity and mortality, ii) the California Department of Vital Statistics, for identifying deaths and cause of death^18,19 and iii) the California Cancer Registry, for identifying cancer diagnoses.^20–23

CHDS mothers are regularly matched to these sources using an accumulated name and address history for each cohort member. This protects against establishing false matches and failing to identify true matches. Surveillance efforts routinely identify over 90% of CHDS mothers.

Results from linkage to the California Vital Status files were used to append year of death and underlying cause. Since follow-up of the cohort spanned over 50 years, codes for the underlying cause of death from several ICD revisions were used to define ovarian cancer death, including 175 for ICD-7; 183 for ICD-8 and 29 and C56 for ICD-10. Linkages to the National Death Index to evaluate the completeness of mortality surveillance have yielded very few missed cases. California Vital Statistics linkage through 2011 yielded 84 ovarian cancer deaths.

Results from linkage to the California Cancer Registry were used to append ovarian cancer diagnosis, year of diagnosis and tumor histology and grade. The California Cancer Registry has established that its cancer coverage is >99% complete after a lag time of about 2 years.²⁴ Life table analyses estimating expected numbers of breast (unpublished) and testicular cancer cases²⁰ show close comparability to observed numbers. Linkage to the California Cancer Registry through 2011 identified 116 cases of ovarian cancer.

Irregular menstruation was defined as self-report or physician report of irregular menstrual cycles, self or physician report of long cycles (>35 days); or physician coded oligomenorrhea, anovulatory cycles or irregular menses.^18,25 The time frame captured by this measure refers to an assessment of menstrual cycle irregularity that encompasses the time between menarche and study enrollment. Thus, this measures estimates chronic menstrual irregularity in the CHDS over an extended time period, 14 years on average.

Covariates were derived from information provided during pregnancy interview and from medical records. Race was categorized as Caucasian, African-American, Hispanic, Asian or other. Age was based on reported age at entry to the study. Parity reflects number of live-born births prior to the observed pregnancy. Cigarette smoking during pregnancy was based on self-report at entry. BMI was calculated from weight (kg) divided by height (m) squared, measured at first prenatal visit.

Oral contraceptive use, age at menarche and subfertility were determined from interview and from medical record review. Subfertility was defined as physician-coded infertility or participant-reported “trouble getting pregnant”. Since all women in the cohort were pregnant, these characteristics were assessed as subfertility rather than infertility. Age at menarche was evaluated as early (≤11 years) and late (≥15 years) versus the reference Group (12–14 years). Oral contraceptive use, administration of fertility drugs and subfertility were coded as yes versus no and refer to the period just prior to the observed pregnancy.

Kaplan-Meier curves of cumulative ovarian cancer survival and disease-free probability were created to examine the association of irregular menstrual cycles. Age at diagnosis and at death or follow-up was calculated based on year of diagnosis, death or last contact; and, year and age at entry. Hazard ratios (HRs) were calculated using Cox proportional hazards models to estimate ovarian cancer associations for irregular cycles accounting for potential confounding by covariates. The proportional hazards assumption was tested by including a cross-product term between irregular cycles and age at event. Analysis was performed using SAS 9.3.

Based on the incessant ovulation hypothesis, which postulates that less frequent ovulatory cycles reduce ovarian cancer risk, we expected that irregular cycles would be protective. Instead we found that irregular cycles were a marker for higher risk of ovarian cancer. Further, the increased risk of ovarian cancer observed for women with irregular cycles was significantly and positively age-dependent, demonstrating substantial risk beginning at age 70, suggesting that the effects of irregular cycles may require the susceptibility of advanced age to enhance risk.

A number of studies have reported ovarian cancer associations with menstrual cycle variability,^{14,15,26–32} and four studies have reported associations with polycystic ovary syndrome (PCOS)³³ (and reviewed in Refs. 16,34). Findings for these studies have been inconsistent showing negative, positive and null associations. Three of these studies observed statistically significant results—two showing decreased risk^14,26 and one showing increased risk.¹³ All three were based on case-control study designs and may have been subject to errors in recall and recall bias, differences in participant age and parity and differences in how irregular cycles was defined, strictly as a specific length or as variable lengths, and whether anovulatory cycles were included.

This is the first prospective study showing a significant and positive ovarian cancer association with irregular cycles. Information about menstrual cycle variability was captured at a young age during the peak of reproduction and long before diagnosis of ovarian cancer, minimizing the potential for recall bias. The prevalence of irregular cycles (13%) in the CHDS is comparable to that observed in other studies.^35,36 The long follow-up period allowed for a sufficient accumulation of ovarian cancer diagnoses and deaths to examine risk by age. Since all the women in our study had achieved a pregnancy, we were able to rule out lifetime infertility as an explanation of the observed association. Infertility occurs with relatively low frequency (ranging from 11% in 1965% to 6% in 2010)³⁷ and most women are able to achieve a pregnancy, including those with irregular cycles or PCOS, a condition associated with irregular cycles.³⁸ Thus, our results apply to the many women with regular and irregular cycles who do conceive.

We were able to adjust for a number of important confounders measured prospectively from medical records (oral contraceptive use and history of infertility) and/or from self-report (age at menarche, difficulty becoming pregnant) assessed at entry during early pregnancy. Prior studies have demonstrated strong protection against ovarian cancer for oral contraceptive use and higher risk for infertility (reviewed in Refs. 9, 10, 14, 39). Oral contraceptive use (4% in our cohort) and infertility or difficulty becoming pregnant prior to entry (12% in our cohort) were infrequent in the CHDS as would be expected for a cohort of pregnant women, enrolled in the early 1960s just as the Food and Drug Administration approved use of oral contraception for birth control.⁴⁰ Although CHDS women with irregular cycles were more likely to be subfertile (18% versus 11% in women with regular cycles), there was no evidence of interaction between irregular cycles and either oral contraception or subfertility in predicting ovarian cancer. Neither subfertility nor oral contraceptive use was a predictor; and, adjustment for these variables did not change the association with irregular cycles, likely due to their low prevalence in this cohort. Young age at menarche has also been observed to enhance ovarian cancer risk (reviewed in Ref. 9). However, young age at menarche (≤11 years) was not associated with ovarian cancer risk in the CHDS and did not confound the association with irregular cycles. None of the examined covariates—subfertility, timing of menarche, weight, age, race and parity—explained the observed increased risk for women with irregular cycles.

It is possible that unmeasured factors account for the association we observed between menstrual irregularity and ovarian cancer. We do not have information about changes in menstrual cycle regularity or events (e.g., obstetric and gyneco-logic) and behavior after the study period, thus we were unable to examine the influence of these factors known to both increase (hormone replacement therapy [HRT]) and decrease (oral contraceptive pills [OCP] and tubal ligation) risk of ovarian cancer. Some of these unmeasured changes should have enhanced the association we observed rather than explain it, such as oral contraceptives and bilateral oophorectomy. Oral contraceptive use may have increased in this population over time regardless of menstrual cycle length as this form of birth control became increasingly more popular. And, it is likely that oral contraceptive use and possibly bilateral salpingo-oophorectomy (with hysterectomy) would have been more frequent in women with irregular cycles to regulate unpredictable bleeding. If so, the association we observed between irregular cycles and ovarian cancer may be larger than we were able to estimate.

More frequent use of HRT in later life by women with irregular cycles could have contributed to the associations we report, although it is unclear why HRT use would be higher among these women. The increased risk of ovarian cancer associated with HRT use is significant but moderate (HR 5 1.53; CI 1.4–1.66).⁴¹ Therefore, the frequency of HRT use would have to differ markedly for women with irregular cycles in order to account for the sizeable association we found.

Our definition of irregular cycles may capture heterogeneous biological and hormonal traits. Further, menstrual cycle status refers to the time between menarche and the observed pregnancy but we were unable to capture change in cycle variation after the study. Misclassifying women who became regular after the observed pregnancy would bias results toward the null. However, despite its lack of specificity, “irregular cycles”, as defined, is a strong predictor of ovarian cancer risk in the CHDS suggesting that cycle regularity in early reproductive life is the relevant risk factor.

It is possible that women with irregular cycles or PCOS have a hormonal profile that favors the neoplastic process through chronic unopposed estrogen, higher androgens or some other mechanism. The protection against ovarian cancer risk observed for oral contraception formulations containing only progestin and progestin-plus-estrogen supports the possibility that unopposed estrogen, rather than low progesterone, could be associated with increased risk.^9,12 Alternatively, oral contraceptives may protect by reducing circulating androgen.¹² Molecular evidence also suggests that androgen synthesis and action are involved in ovarian cancer etiology.⁴² Therefore, the increased risk of ovarian cancer among PCOS women who express symptoms of hyperandrogenism⁴² is consistent with an androgen-related mechanism.

This study provides new prospective evidence that woman with irregular menstrual cycles, who are more likely to have PCOS, are at an increased risk of ovarian cancer. If confirmed this will add to the burden of liability for women with PCOS who have higher risk for endometrial cancer,^43,44 metabolic syndrome⁴⁵ and cardiovascular disease (CVD), including prospective evidence of higher CVD risk in this cohort.¹⁸ The cooccurrence of increased risk for multiple diseases in women with PCOS complicates the implications for prevention. Oral contraception could possibly provide greater protection against ovarian cancer for women with PCOS compared to the general population adding to the known benefit of oral contraceptive use for reducing risk of endome-trial cancer.⁴⁶ However, the benefit of oral contraceptives for women with PCOS, must be considered against the cost of potential adverse effects on insulin resistance, glucose intolerance and vascular risk known to be more frequent in these women.⁴⁵ The relatively higher frequency of CVD in women (lifetime risk for women with optimal risk factor levels is4.1%)⁴⁷ compared to ovarian cancer (lifetime risk is 1.3%)² and endometrial cancer (lifetime risk is 2.8%)² suggests that it is essential to give significant consideration to the individual risk for cardiovascular consequences of OCP use. The risk reduction in ovarian cancer associated with metformin to control diabetes observed in a large case-control study³³ may provide a promising alternative for women with PCOS who are at a higher risk of diabetes. However, more research from ongoing prospective clinical trials is needed to confirm whether metformin protects against ovarian cancer among women who do not have diabetes⁴⁸ and whether there are other unforeseen, harmful consequences of metformin treatment.

Other factors to consider for women with PCOS are intrauterine devices (IUDs) and HRT use. Given the potential increased ovarian cancer risk associated with IUD use either directly or indirectly via increased inflammation,⁹ and with HRT,⁴¹ further research is needed to determine whether women with irregular cycles require special guidelines governing use of these.

Current screening procedures using trans-vaginal ultrasonography and serial assay for CA-125 have been relatively unsuccessful in reducing ovarian cancer mortality^11,49 and have shown limited effectiveness for early detection.^11,50,51 Developing a risk algorithm that integrates known risk markers such as germline mutations (e.g., BRCA), with phenotypes (e.g., irregular cycles) and behavioral characteristics (e.g., oral contraceptive and HRT use) may offer a way to improve screening targets until better biomarkers are identified. The identification of higher risk phenotypes such as irregular menstruation may provide opportunities to narrow the search for early biomarkers. However, additional work to understand the mechanisms underlying the association between irregular menstruation and ovarian cancer will be necessary in order to more specifically define the clinical phenotype driving this observed association.