Menstrual Irregularity and Cardiovascular Mortality

Abstract

Background: Polycystic ovary syndrome, the most common cause of irregular menstrual cycles, is associated with an adverse cardiovascular risk profile. However, there are limited prospective studies confirming the link between polycystic ovary syndrome and cardiovascular mortality.

Methods: We studied 15,005 pregnant women recruited from the Kaiser Foundation Health Plan in California between 1959 and 1966. The menstrual cycle pattern was assessed at baseline according to self-report, physician report, and medical record abstraction. Participants were matched to California Vital Status files annually until 2007 to identify deaths due to overall cardiovascular disease (CVD) and subsets of coronary heart disease (CHD) and cerebrovascular disease based on International Classification of Diseases codes. Cox proportional hazards models were used to estimate the association between irregular cycles and cardiovascular mortality. Missing covariate data were multiply imputated using standard methods.

Results: During 456,298.5 person-years of follow-up, there were 666 CVD deaths, including 301 CHD deaths and 149 cerebrovascular deaths. Compared with women with regular cycles, women with irregular cycles had an increased risk for CHD mortality [age adjusted hazards ratio (HR) 1.42, 95% confidence interval (CI) 1.03–1.94]; however, the association was not statistically significant after adjustment for body mass index (adjusted HR 1.35, 95% CI 0.98–1.85). There was a nonsignificant increase in CVD mortality (age adjusted HR 1.21, 95% CI 0.97–1.52) but not cerebrovascular mortality (age adjusted HR 0.85, 95% CI 0.49–1.47).

Conclusions: In this large prospective cohort of pregnant women, we found an increase in age-adjusted risk for CHD mortality in women with irregular menstrual cycles. This risk was attenuated after adjustment for body mass index.

In a prospective cohort of pregnant women, age-adjusted risk for coronary heart disease mortality was increased in women with irregular menstrual cycles.

Polycystic ovary syndrome (PCOS) is a heterogeneous syndrome among reproductive-aged women characterized by irregular menstrual cycles due to anovulation, hyperandrogenism, and/or polycystic ovaries on ultrasound (1,2). PCOS is associated with an adverse cardiovascular risk profile (3); however, the literature lacks prospective studies based on a well-defined population to confirm the link between PCOS and cardiovascular mortality.

Among healthy reproductive-aged women, PCOS accounts for 87% of women with irregular cycles (4). Using data from a large cohort of pregnant women recruited from the East Bay Area of California (1959–1967) and followed up for approximately 40 yr, we aimed to determine whether menstrual cycle irregularity is associated with increased cardiovascular mortality.

Materials and Methods

Study design and sample

The Child Health and Development Studies (CHDS) was initiated in 1959 to investigate events of pregnancy and childhood development. The recruitment procedures and characteristics of the cohort have been previously described (5). The target population for the CHDS included all members of the Kaiser Foundation Health Plan residing in the East Bay of the San Francisco Bay Area. Women were invited to enroll in the CHDS when they contacted participating Kaiser facilities regarding a confirmed pregnancy. Recruitment efforts resulted in enrollment of more than 98% of eligible women. In all, 15,528 women, aged 14–48 yr, were enrolled between 1959 and 1966. The institutional review board approved the study protocol; informed consent was obtained from all participants. At the time of enrollment, baseline demographic and pregnancy history information was collected by in-person interviews. Clinical measures and information about maternal medical conditions were abstracted from medical records beginning 6 months before pregnancy through labor and delivery. The present investigation included 15,005 women from the cohort with known follow-up status.

Menstrual cycle regularity

Menstrual cycle pattern was assessed at baseline according to the following factors: 1) self-report at the interview and 2) physician report and diagnosis from medical record abstraction. Participants were asked, “Would you say your periods have been regular?” If the participants responded no, they were considered to have irregular cycles. If the participants indicated regular cycles, they were asked, “What is the usual interval between the beginning of one period to the next?” A participant was defined as having irregular cycles if one of the following criteria was met: self-report or physician report of irregular cycles; self-report or physician report of cycle length 36 d or longer; or physician-coded oligomenorrhea, anovulatory cycles, or irregular menses.

Vital Status surveillance and mortality data

After active follow-up of the cohort ended in 1972, Vital Status surveillance was conducted annually. Cohort members were matched to the California Department of Motor Vehicles (DMV) files and California Vital Status records using name and birth date. The DMV files provided both residence and date at last active contact to establish when a participant resided in a given surveillance area. The complete file was then matched to the Vital Status records. Using all names that a participant registered with the DMV to find matches within the Vital Status records substantially reduces the likelihood that cases were missed due to incomplete identifier information. Surveillance efforts identified one or more members of greater than 80% of CHDS families. If a participant did not match to either the DMV or Vital Status files, then the participant was defined as alive at the last active follow-up year.

Primary cause of death was provided by the Vital Status files. Mortality was determined for overall cardiovascular disease [CVD; International Classification of Diseases (ICD), seventh revision code 400–460; ICD, eighth revision (ICD-8) 390–459; ICD, ninth revision (ICD-9) 390–459] and subsets of coronary heart disease (CHD; ICD-8 code 410–414, ICD-9 code 410–414, ICD, 10th revision code I20-I25) and cerebrovascular disease (ICD-8 code 430–438; ICD9-code 430–438; ICD, 10th revision code I60-I-69) (6). Findings presented here are based on events occurring through 2007.

Covariates

Covariates were obtained at the time of enrollment from interviews and medical records. Race was categorized as Caucasian, African-American, Hispanic, Asian, or other. Oral contraceptive use was dichotomized based on participants’ self-report of using oral contraceptive pills since their last pregnancy or indication of oral contraceptive use in the medical records. Parity was defined as the count of previous live-born births. Tobacco use was categorized as never, past, or current. Personal history of diabetes was obtained from physician diagnosis or medical records; this condition was not specified as pregestational or gestational.

Weight and height were measured at first interview. Weight was adjusted to compensate for variation in gestational age by regressing weight on gestational age using the locally weighted scatterplot smoothing technique (7). Adjusted weight was then imputed as the fitted mean weight at 104 d of gestation, the median value at first interview, plus her residual from the locally weighted regression and scatterplot smoothing technique fit. This procedure removes differences in weight due to differences in gestational age at first interview.

Statistical analyses

We compared women with irregular and regular cycles using unadjusted linear regression for continuous variables and logistic regression for categorical variables. The independent associations of irregular cycles with CVD, CHD, and cerebrovascular disease mortality were then assessed using Cox proportional hazards models, adjusting for age, race, body mass index (BMI), oral contraceptive use, parity, and tobacco use.

We dealt with missing covariate values using multiple imputation (8,9), as implemented in the imputation by chained equations package in STATA (StataCorp., College Station, TX) (10). Among 15,005 women, 167 had missing values for age, 282 for race, 2013 for BMI, 122 for parity, and 2912 for tobacco use. The chained equation for each missing covariate was fully specified to capture nonlinearities and interactions in its associations with other covariates. Age was modeled using a cubic spline. There were no missing follow-up times for women who died, and we did not attempt to impute 523 missing censoring times. Thus, a total of 15,005 women were included in the final analyses. Five imputed data sets were created and then analyzed using standard methods for multiply-imputed data, as implemented in the mim package in STATA. These procedures ensure that ses, confidence intervals, and P values properly reflect the extra variability from the imputation of missing variables.

All analyses were conducted using STATA (version 10).

Results

Of the 15,005 women included in the current analyses, 1974 (13.2%) were classified as having irregular menstrual cycles. Compared with women with regular cycles, women with irregular cycles were likely to be younger and have used oral contraceptives (Table 1). The two groups were similar in terms of race, BMI, parity, tobacco use, and prevalence of diabetes.

Table 1.

Baseline characteristics of CHDS cohort as a function of menstrual cycle regularity

	Regular menstrual cycles (n = 13,031)	Irregular menstrual cycles (n = 1974)a	P valueb
Age (yr)c	26.9 ± 6.3	26.5 ± 5.9	0.01
Raced			0.69
White	67.0 (66.2–67.8)	67.8 (65.7–69.8)
Black	23.5 (22.8–24.2)	22.7 (20.8–24.6)
Hispanic	3.1 (2.8–3.4)	3.1 (2.4–3.9)
Asian	3.6 (3.3–4.0)	4.0 (3.1–4.9)
Other	2.8 (2.5–3.1)	2.4 (1.7–3.1)
BMI	22.8 ± 3.7	22.9 ± 4.0	0.30
Oral contraceptive use	3.4 (3.1–3.7)	5.5 (4.5–6.5)	<0.001
Tobacco use (current)	35.9 (34.8–37.0)	37.5 (35.2–39.8)	0.23
Parity	1.4 ± 1.7	1.3 ± 1.6	0.12
Diabetes	0.6 (0.5–0.7)	1.0 (0.5–1.4)	0.08

^aDefined as self-reported irregular cycles, self-reported cycles longer than 36 d, physician-coded anovulatory cycles, physician-coded irregular cycles, or physician-coded oligomenorrhea. 

^bP value calculated by unadjusted linear regression for continuous variables and logistic regression for categorical variables. 

^cContinuous variables represented as mean ± sd. 

^dCategorical variables represented as proportions (95% CI). 

During 456,298.5 person-years of follow-up, there were 2501 deaths observed in the CHDS cohort, of which 666 (26.6%) were due to cardiovascular events. These included 301 CHD deaths and 149 cerebrovascular deaths. There were five deaths due to CVD mortality coded by ICD, seventh revision , which could not be classified as either CHD or cerebrovascular.

Compared with women with regular cycles, women with irregular cycles had an increased risk for CHD mortality [age adjusted hazards ratio (HR) 1.42, 95% confidence interval (CI) 1.03–1.94]; however, the association was not statistically significant after adjustment for BMI (adjusted HR 1.35, 95% CI 0.98–1.85) (Table 2). Further adjustment with potential confounders did not alter the association (adjusted HR 1.33, 95% CI 0.97–1.83). There was a nonsignificant increase in overall CVD mortality (age adjusted HR 1.21, 95% CI 0.97–1.52) but not cerebrovascular mortality (age adjusted HR 0.85, 95% CI 0.49–1.47). The results did not differ in a complete-case analysis of 11,679 participants with complete covariate data.

Table 2.

HRs (95% CI) for cardiovascular mortality as a function of menstrual cycle regularity in the CHDS cohort

	Regular cycles (n = 13,031)	Irregular cycles (n = 1974)a
CVD mortality
Cases, n	577	89
Age-adjusted model		1.21 (0.97–1.52)
Age, BMI-adjusted model		1.16 (0.93–1.45)
Multivariate modelb		1.14 (0.91–1.43)
CHD mortality
Cases, n	255	46
Age-adjusted model		1.42 (1.03–1.94)
Age, BMI-adjusted model		1.35 (0.98–1.85)
Multivariate model		1.33 (0.97–1.83)
Cerebrovascular mortality
Cases, n	135	14
Age-adjusted model		0.85 (0.49–1.47)
Age, BMI-adjusted model		0.83 (0.48–1.44)
Multivariate modelb		0.82 (0.47–1.43)

^aDefined as self-reported irregular cycles, self-reported cycles longer tohan 36 d, physician-coded anovulatory cycles, physician-coded irregular cycles, or physician-coded oligomenorrhea. 

^bAdjusted for age, race, BMI, parity, current tobacco use, and oral contraceptive use. 

Baseline diabetes was a strong predictor of CHD mortality (HR 5.10, 95% CI 2.41–10.81) in an unadjusted analysis. However, addition of diabetes to the multivariable model for CHD death did not change the results for irregular cycles (adjusted HR 1.34, 95% CI 0.98–1.84). Similarly, when women with baseline diabetes were excluded from the analysis, the association remained the same (adjusted HR 1.33, 95% CI 0.97–1.83).

Discussion

In a large prospective cohort of women with more than 40 yr of follow-up, we found an increase in age-adjusted risk for CHD mortality among women with irregular menstrual cycles. There was suggestive but inconclusive evidence for an association independent of BMI. We noted a nonsignificant increase in an age-adjusted risk for overall CVD mortality but not cerebrovascular mortality.

One previous study provided evidence for increased CHD risk among women with a history of irregular cycles. Over 14 yr of follow-up among 82,439 postmenopausal women in the Nurses’ Health Study, women who recalled a history of irregular cycles at 20–35 yr of age had higher CHD mortality (age adjusted relative risk 1.67, 95% CI 1.35–2.06) as well as nonfatal CHD (age adjusted relative risk 1.25, 95% CI 1.07–1.47) (11). Our findings for fatal CHD are consistent with the Nurses’ Health Study results but attenuated, likely because of two factors. First, the assessment of menstrual cycle regularity in our study was not subject to recall bias, avoiding potential bias away from the null. Second, given that our cohort recruited only pregnant women and irregular cycles are known to be associated with anovulatory infertility (12), our participants may be a healthier group of women with irregular cycles. However, a recent study has shown that although oligomenorrhea is associated with a lower fecundability, women with oligomenorrhea had at least one birth as often as women without oligomenorrhea (13). In addition, we found a 13.2% prevalence of irregular cycles in our cohort, which is consistent with other studies (14,15). Despite the possibility of a selection bias, the study strengths include the large cohort of ethnically diverse reproductive-aged women and prospective ascertainment of irregular cycles with assessment of cardiovascular mortality.

Natural history studies following up cohorts of PCOS women, identified by ovarian wedge resection, have found inconclusive evidence for cardiovascular morbidity (16,17). More recently a study sponsored by the National Heart, Lung, and Blood Institute showed that among postmenopausal women suspected of myocardial ischemia, a history of irregular cycles and elevated postmenopausal androgen levels were associated with worse 5-yr cardiovascular event-free survival (18). However, because circulating androgen levels in women with PCOS commonly normalize before menopause (19), this study may identify a subgroup of postmenopausal women who are at a higher risk for coronary artery disease.

Certain limitations should be noted in interpreting our results. First, although PCOS accounts for the majority of women with irregular cycles (4), our exposed group likely incorporates other etiologies of irregular cycles including hypothalamic abnormalities. In addition, oral contraceptive use is an important influence on the risk of CHD mortality, and our study does not take into account oral contraceptive use during the follow-up period. Also, our outcome assessment was limited to the California Vital Status files instead of the National Death Index. However, we do not expect women with irregular cycles to preferentially move out of California compared with women with regular menstrual cycles; thus, limiting our outcome assessment to California should not be a source of bias.

In this large prospective cohort of pregnant women, irregular menstrual cycles were associated with an increased age-adjusted risk for CHD mortality. This risk was attenuated after accounting for BMI. Future prospective studies based on a well-defined PCOS population are needed to assess the relationship between PCOS, BMI, and cardiovascular morbidity and mortality.