Evidence for increased cardiovascular events in the fathers but not mothers of women with polycystic ovary syndrome

Abstract

BACKGROUND

Polycystic ovary syndrome (PCOS) is a familial syndrome, associated with multiple cardiovascular disease (CVD) risk factors. Thus, parents of affected women may have a higher prevalence of CVD events than the general population.

METHODS

PCOS probands (n = 410) and their participating parents (n = 180 fathers and 211 mothers) were queried for CVD events in themselves and non-participating family members. In order to include the family CVD history of all parents, agreement between the proband and parental reports of CVD events was assessed. Estimated 10-year coronary heart disease (CHD) risk was calculated using the Framingham risk calculator. The National Health and Nutrition Examination Survey (NHANES) 2001–2002 database was used to generate gender, age and body mass index-relevant population parameters of CVD prevalence in the USA population.

RESULTS

Ninety-eight percent of the parents' self-reporting of CVD events agreed with the proband's report of parental heart attack history [Kappa = 0.82; 95% CI: (0.69, 0.94)] and 99% with parental stroke history [Kappa = 0.79; 95% CI: (0.62, 0.97)]. Fathers of women with PCOS had a higher prevalence of heart attack and stroke compared with the reference NHANES population (heart attack: 11.1 versus 5.3%, P < 0.0001; stroke: 3.0 versus 1.0%, P = 0.002). Fathers of women with PCOS had an elevated 10-year risk for CHD (11.5 versus 9.9% in NHANES, P = 0.03). No statistically significant increased prevalence of CVD events or 10-year risk was noted in probands or mothers.

CONCLUSIONS

Fathers, and not mothers, may be disproportionately burdened with CVD in PCOS families. The strengths of this study include the size of our cohort, the consistent phenotyping and the validation of proband's reporting of parental CVD events.

Introduction

Cardiovascular disease (CVD) is the leading cause of mortality in women in the USA. Polycystic ovary syndrome (PCOS) is a common endocrine disorder, affecting 4–8% of reproductive-age women (Azziz et al., 2004). In addition to ovulatory dysfunction and hyperandrogenism, PCOS is associated with a constellation of metabolic and CVD risk factors otherwise known as metabolic syndrome (Ehrmann et al., 2006). In addition to dyslipidema (Legro et al., 2001), women with PCOS have been noted to have high rates of glucose intolerance and type 2 diabetes (Ehrmann et al., 1999; Legro et al., 1999), that are present even in adolescents (Palmert et al., 2002).

Women with PCOS have an increased risk of subclinical atherosclerosis as indicated by endothelial dysfunction (Paradisi et al., 2001), increased carotid artery intima media thickness (Talbott et al., 2000), and increased coronary artery calcification compared with reproductively normal control women of similar age and weight (Christian et al., 2003). It has been suggested that, due to its prevalence and its metabolic risk factors, PCOS may account for a significant proportion of atherosclerotic heart disease observed in women (Dahlgren et al., 1992). Post-menopausal women with current or past stigmata of PCOS have been shown to have a more severe CVD event history and risk profile (Krentz et al., 2007; Shaw et al., 2008). However, no prospective studies have established an association between women with well-characterized PCOS and actual cardiovascular events, or their premature onset (Legro, 2003).

Since age is a significant risk factor for CVD events, it may be that studies have not followed cohorts of women with PCOS to an age where CVD events become more common in women, the seventh and eighth decades of life. The Nurse's Health Study has shown an increased CVD event rate in post-menopausal women with a history of oligomenorrhea during their reproductive years (Solomon et al., 2002). Similarly, prospective studies of post-menopausal women with androgen excess have shown them to have more severe coronary artery disease and higher CVD event rates (Shaw et al., 2008).

PCOS is likely a complex genetic disease (Ewens et al., 2010) that is clearly modified by the environment (i.e. more severe phenotype with obesity). We and others have shown that elevated CVD risk clusters in first-degree relatives, including dyslipidemia, insulin resistance and diabetes (Legro et al., 2002b; Yildiz et al., 2003; Sam et al., 2006). Accordingly, parents of affected women provide an opportunity to investigate PCOS-related CVD risk in a population at an age where these events become common, and also represent a group with environment and genes shared with their daughters. We hypothesized that there would be an increased CVD risk in parents of women with PCOS and an increased prevalence of coronary heart disease (CHD) risk based on the Framingham Risk Assessment Tool for estimating the 10-year risk of developing CHD (myocardial infarction and coronary death).

Materials and Methods

Study subjects

Women with PCOS and their parents were recruited into an ongoing phenotype/genotype study of PCOS families between the years of 2001 and 2005. The study was approved by the Institutional Review Boards of Northwestern University Feinberg School of Medicine and the Penn State College of Medicine. Written informed consent was obtained from all participants.

PCOS probands were defined by the 1990 NIH-NICHD criteria of hyperandrogenic chronic anovulation with exclusion of other etiologies (Zawadski and Dunaif, 1992). To qualify as a PCOS proband, a woman had to have six or fewer menses per year in addition to hyperandrogenemia defined by either total testosterone (T) more than 58 ng/dl and/or non-sex hormone-binding globulin-bound T more than 15 ng/dl, levels >2 SD values above the mean value established in reproductively normal women aged 18–40-year-old in the early follicular phase of the menstrual cycle (Legro et al., 1998). Probands were not taking any confounding medications that alter sex steroid metabolism or insulin sensitivity and were age 15–44 years. Non-classical 21-hydroxylase deficiency, hyperprolactinemia, androgen-secreting tumors, thyroid disease and pregnancy were excluded by appropriate tests before the diagnosis of PCOS was made (Legro et al., 1998).

Study design

A total of 410 PCOS probands from the two participating sites were studied, along with participating parents. The majority of the participants resided in Pennsylvania; however, 35 states were represented for probands, 35 states for mothers and 34 states for fathers. We obtained medical history, biometric information and fasting blood according to our standard phenotyping protocol (Legro et al., 1998). Standardized forms were used to obtain personal medical history, including history of cardiovascular events such as myocardial infarction and stroke, tobacco use and current prescription medications. The participant was asked to recall any ‘heart attack’ or ‘stroke’ within their immediate family, requiring that they specify during which of three age ranges the event occurred for that family member: <55 years, 55–65 years, >65 years or age unknown.

Subjects who completed the study on-site had their blood pressure, height and weight measurements performed by clinical research staff. Some of the subjects were studied off-site. Off-site subjects self-reported their height and weight measurements. We have previously validated self-reported height and weight (Sam et al., 2006). All subjects had blood samples, which were obtained in the morning after an overnight fast, and levels of total cholesterol and high-density lipoprotein (HDL) cholesterol were assayed (Legro et al., 2002a). Data on reproductive and metabolic phenotypes, including fasting glucose, lipid and lipoprotein levels, have been reported on some of the probands and mothers and fathers (Sam et al., 2006; Stewart et al., 2006; Coviello et al., 2009).

Reference population

In order to determine the prevalence of CVD events and risk factors in a reference group, we used a US population-based sample of comparable gender, mean age and mean BMI data abstracted from the nationwide National Health and Nutrition Examination Survey (NHANES) 2001–2002 database (www.cdc.gov/nchs/nhanes.htm). All records from the NHANES database were selected such that they were gender appropriate, within ± 1 year in age, and within ± 0.5 kg/m² in BMI of subjects in each of our subsets (probands, mothers and fathers) (Sam et al., 2006; Coviello et al., 2009). NHANES 2001–2002 contains data on anthropometric measurements, blood pressure and fasting glucose and lipid levels and was sampled from all U.S. households during this time frame.

Framingham CHD score

A CVD risk score was calculated for subjects aged 20 years and older who had no cardiovascular events or diagnosis of diabetes using the online National Cholesterol Education Program calculator (http://hp2010.nhlbihin.net/atpiii/calculator.asp). We chose this method because it is commonly used clinically and has been validated in other populations in the USA (D'Agostino et al., 2001). The algorithm predicts the 10-year risk of coronary events using the following measurements: age, gender, total and HDL cholesterol levels, smoking status, current use of anti-hypertensive medications and systolic blood pressure. These parameters were measured at the time of the subject's visit to the center and from information noted in their completed medical history.

Statistical analysis

Many parents did not participate in the study and thus were not phenotyped. Exclusion of non-studied parents could potentially bias the results, for example, a non-studied parent may have been unable to participate in the study due to death or incapacitation from a CVD event. The inclusion of non-studied parents in the analysis to reduce this potential bias required the use of proband reporting of non-studied parental CVD events. Validation for using proband reporting of parental CVD events was undertaken by calculating the kappa statistic and 95% confidence interval to assess the agreement between the probands and their studied parents' reporting of parental CVD events.

The selected NHANES data as described under the Reference Population section was weighted appropriately as per NHANES study instructions to provide population parameters of heart attack, stroke and 10-year Framingham CHD risk score. The one-sample binomial exact test was used to compare heart attack and stroke prevalence in the PCOS families with the reference population from NHANES. The one-sample t-test was used to compare the Framingham 10-year CHD risk of the PCOS families to the NHANES reference population.

Results

Validation of proband reports of parental CVD history

Proband parental history reports were compared with the surveys completed by their respective parents. Two hundred eleven mothers (52%) completed surveys and were included in the validation process. One hundred eighty fathers (44%) completed surveys and were included in the validation process. Ninety-eight percent (379/387) of the parents were in agreement with the proband's report of parental heart attack history [Kappa = 0.82; 95% CI: (0.69, 0.94)]. Ninety-nine percent (371/376) were in agreement with the proband's report of parental stroke history [Kappa = 0.79; 95% CI: (0.62, 0.97)].

Cardiovascular events

Demographic characteristics of the PCOS probands, mothers and fathers and their respective NHANES reference groups displayed in Tables I and II show the prevalence of cardiovascular events in the different subject populations. We found only one self-reported CVD event in a PCOS proband: a stroke at age 12. Given that an exponential increase in CVD events occurs for men at age 55 years, 56% of our fathers qualified as being in the higher-risk age group. The prevalence of heart attack and stroke in fathers of women with PCOS was significantly increased compared with the NHANES population. All other occurrences of CVD events in PCOS probands and mothers were not significant compared with our reference population. However, the exponential increase in CVD events occurs for women at age 65 years; only 7% of our mothers were in the higher-risk age group.

Table I.

Demographics for PCOS families and NHANES reference population.

	Proband, n (%)	Female NHANES controls for probanda (%)	Mother, n (%)	Female NHANES controls for mothera (%)	Father, n (%)	Male NHANES controls for fathera (%)
Race
Hispanic	27 (7%)	19%	7 (3%)	11%	7 (3%)	11%
White	327 (81%)	61%	248 (91%)	74%	209 (91%)	78%
Black	32 (8%)	15%	13 (5%)	11%	6 (3%)	9%
Other	19 (5%)	5%	5 (2%)	5%	8 (3%)	3%
Smoker	90 (24%)	29%	38 (17%)	23%	26 (14%)	23%
	Mean (SD)	Mean	Mean (SD)	Mean	Mean (SD)	Mean
Age (years)	28 (5)	28	54 (8)	52	57 (8)	55
BMI (kg/m2)	36 (9)	29	31 (7)	28	31 (6)	29
Gravidity	0.6 (1.1)	2.8	3.8 (2.3)	3.6

NHANES, The National Health and Nutrition Examination Survey.

^aNHANES data were appropriately weighted to obtain a population parameter for the proportion or the mean; therefore, no sample size or estimate of variability is provided.

Table II.

Prevalence for CVD events in PCOS families and NHANES reference population.

Heart Attack				Stroke
Relation	PCOS family data fraction (%)	NHANES dataa (%)	P-valueb	PCOS Family Data Fraction (%)	NHANES dataa (%)	P-valueb
Proband	0/407 (0%)	0.2	0.89	1/408 (0.2%)	1.0	0.20
Proband mother	12/405 (3.0%)	2.4	0.50	11/405 (2.7%)	2.7	1.00
Proband father	45/405 (11.1%)	5.3	<0.0001	12/405 (3.0%)	1.0	0.002

^aNHANES data were appropriately weighted to only obtain a population parameter for the proportion of events; therefore, no sample sizes are provided.

^bOne-sample binomial exact test.

10-year calculated CHD risk

The factors shown in Table III were used to estimate the Framingham 10-year CHD risk for the different subject populations. Subjects were only included in this analysis if we had complete information on all risk factors included in the Framingham calculation. The proband fathers had an increased risk profile compared with the NHANES population (P = 0.03; Table IV). The 10-year CHD risk for both probands and their mothers did not differ from their respective NHANES population.

Table III.

Factors used in the Framingham 10-year CHD risk assessment.

	Proband (n = 263)	Female NHANES controls for probanda	Mother (n = 120)	Female NHANES controls for mothera	Father (n = 85)	Male NHANES controls for fathera
	Mean (SD)	Mean	Mean (SD)	Mean	Mean (SD)	Mean
Age	28 (5)	28	53 (7)	52	56 (8)	54
Total cholesterol (mg/dl)	186 (36)	188	211 (42)	213	196 (34)	213
HDL cholesterol (mg/dl)	44 (11)	54	57 (21)	58	43 (11)	46
Systolic blood pressure (mmHg)	121 (13)	109	128 (18)	124	132 (16)	124
Smoker, n (%)	87 (25%)	27%	30 (17%)	21%	20 (15%)	21%
Hypertensive medications, n (%)	14 (4%)	2%	52 (28%)	21%	47 (33%)	16%

^aNHANES data were appropriately weighted to only obtain a population parameter for the mean; therefore, no sample size or estimate of variability is provided.

Table IV.

10-year calculated CHD risk rate using the Framingham model.

Relation	10-year CHD risk (%)
	PCOS family data		NHANES dataa
	n	Mean (SD)	Mean	P-valueb
Proband	263	0.7 (1.0)	0.6	0.15
Proband Mother	120	2.7 (2.6)	2.5	0.45
Proband father	85	11.5 (6.5)	9.9	0.03

^aNHANES data were appropriately weighted to only obtain a population parameter for the mean; therefore, no sample size or estimate of variability is provided.

^bOne-sample t-test.

Discussion

Our study noted an increased prevalence of heart attack (11.1 versus 5.3%) and stroke (3.0 versus 1.0%), as well as an increased 10-year CHD risk in fathers of women with PCOS compared with population parameters obtained from men in the NHANES 2001–2002 database of similar age and BMI. There was no statistically significant increased prevalence of events in women with PCOS or their mothers, nor was there an increased 10-year CHD risk compared with the women in the NHANES 2001–2002 database of similar age and BMI. Our study thus suggests that first-degree male members of PCOS families have both increased risk and corresponding event rates, whereas females are relatively normal.

Several studies have investigated the risk of CVD in family members of women with PCOS (Atiomo et al., 2003; Yilmaz et al., 2005; Hunter et al., 2007). These studies have been limited by small sample sizes (<100 subjects). In addition, control populations for these studies were small and ascertained from disparate sources, for example, from infertility clinics. Thus, the control populations were subject to selection bias and were, perhaps, not representative of the larger population. Hunter et al. (2007) showed an increased CVD risk in brothers but not fathers of women with PCOS based on survey of the proband with PCOS. This study was limited by an inconsistent diagnosis of PCOS and by the fact that the investigators did not survey or phenotype the other first-degree relatives. Yilmaz et al. (2005) showed an increased risk of CVD in both fathers and mothers of women with PCOS. This study was designed similarly to ours, and our mothers’ ages were overlapping. Atiomo et al. (2003) showed an increased number of women with PCOS who had a positive family history of heart attacks. However, it is unclear what signified a parental history of heart attacks for the study, as it was not specified if the affected family member was a parent, sibling, aunt/uncle or grandparent.

Our study finds an increased risk for CVD events only in fathers. This is consistent with our previous studies noting clustering of CVD risk factors in first-degree male relatives of probands (Sam et al., 2008; Coviello et al., 2009). The mechanism for this is beyond the scope of the study. However, the increased prevalence of hypertension and use of hypertensive medications among the fathers who were phenotyped compared with the NHANES reference group suggests this significant CVD risk factor was a major contributor to our findings. Our failure to detect an increased risk of CVD events in affected women and their mothers may reflect their relatively young age since CVD events do not become prevalent in women until the seventh to eighth decades of life. Given that an exponential increase in CVD events occurs for men at age 55 years, 56% of our fathers qualified as being in the higher-risk age group. However, the exponential increase in CVD events occurs for women at age 65 years; only 7% of our mothers are in the higher-risk age group. It is, therefore, possible that our results may differ were we to enroll older female subjects or to observe our present cohort longer.

We also did not find an increased 10-year risk of CVD events in women with PCOS or their mothers, in contrast to previous studies (Yilmaz et al., 2005; Cheang et al., 2008). This observation may reflect the fact that while there were relative abnormalities in risk factors in women compared with a reference population, including in our own previous studies (Sam et al., 2008; Coviello et al., 2009), they were not substantial enough to increase the calculated CVD risk score. Alternatively, there may be some aspects of PCOS that are protective against CVD events. For example, prolonged ovarian reserve in women with PCOS may provide a longer window of sex steroid production and cardiovascular protection in women (Mulders et al., 2004; Tehrani et al., 2010).

Our study, like all of the others on the topic, was limited by the retrospective nature of the study and by the relatively small sample size. Though cardiovascular events are major life-changing occurrances, it is possible that recall bias may lead to lapses in the parent either directly documenting it or communicating it to their daughters. Although compared with similar studies of families of women with PCOS, ours is quite large, from an epidemiologic viewpoint, it is still a small study with a sample size of 410 families. Our study was strengthened by standardized phenotyping of subjects with PCOS and by systematic, prospective collection of CVD event history. Further, we have validated that a proband report of a parental CVD event had a high concordance with the parent report (≥98% for heart attack and stroke). This concordance reduced the likelihood of underestimating of event rates when a parent was lost to follow-up due to a fatal or incapacitating CVD event. The NHANES data allow for comparison with a large cross section of the U.S. general population. However, there were differences in the recruitment of subjects and the ascertainment of outcomes between NHANES and our study that could confound our findings. For example, the history of events was obtained by in-person home interview in NHANES and by standardized questionnaire in our study, potentially resulting in selective reporting of actual events. Further we chose a time point in the NHANES data that was roughly halfway between our start and completion date for this study, such that changing cardiovascular event rates could alter our findings.

Finally, it is impossible with such a study design to control for all confounding factors between the NHANES study and our case series. For example, we did not control for smoking, and we found higher rates of diabetes in our families than in the NHANES reference population. The prevalence of diabetes in our probands was 5%, in our mothers was 17%, and in our fathers was 20%. Whereas in the gender and age-relevant NHANES population it was 2% for probands, 6% for mothers and 11% for fathers. However, based on the pathophysiology of PCOS as a disorder of insulin resistance, we would expect this diabetes discrepancy.

As awareness and proper diagnosis of PCOS increases, larger studies of the long-term comorbidities of this syndrome will become more feasible. We conclude that there is an increase in CVD events and risk in fathers of women with PCOS. This is a distinct population who can be targeted for early screening and surveillance for CVD. Study of shared environment and genes may shed light on a common etiology to PCOS in daughters and CVD disease in fathers.

Authors’ roles

M.C.T., A.R.K., A.R.K., C.M.S., A.D., R.L. researched data, contributed to discussion, reviewed/edited manuscript. M.T. wrote manuscript.

Funding

This work was supported by U.S. National Institutes of Health(NIH) grants K24 HD01476, U54 HD034449, P50 HD044405; GCRC Grant M01 RR 10732; and construction grant C06 RR016499 to Pennsylvania State University. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.