Coronary Heart Disease Risk Factors in Women With Polycystic Ovary Syndrome
Abstract The goal of the study was to compare cardiovascular heart disease risk factors in women with polycystic ovary syndrome (PCOS) and matched control subjects. Women with PCOS have risk factors, including anovulation, hyperandrogenism, and insulin resistance, that suggest a male coronary heart disease risk-factor profile. A total of 206 women with PCOS were recruited by using records from a large reproductive endocrinology practice. A clinical diagnosis of PCOS was made if there was a history of chronic anovulation in association with either clinical evidence of androgen excess (hirsutism) or if total testosterone level was >2 nm/L or the luteinizing hormone/follicle-stimulating hormone ratio was greater than 2. The overall response rate for cases was 76%. A control population was obtained by using a combination of area voters’ registration tapes and directories of households. A control subject was matched to each case subject by age±5 years, race, and neighborhood. The response rate for recruitment of the first or second eligible control subject was 83.6%. The average age at initial interview was 35.9±7.4 years for case and 37.2±7.8 years for control subjects. Women with PCOS had significantly increased cardiovascular disease risk factors compared with control women. These included increases in body mass index, insulin, and triglyceride levels (P<.001), decreased total HDL and HDL2 levels (P<.01), and increased total cholesterol and fasting LDL levels, waist/hip ratio, and systolic blood pressure (P<.05). After controlling for age, body mass index, and other confounding variables, differences in total cholesterol, total HDL, HDL2, LDL cholesterol, and triglycerides were still significant between case and control subjects. These risk factors were especially elevated in PCOS women in the early premenopausal years compared with control women, indicating that women with PCOS should be monitored for early detection and considered for appropriate clinical interventions.
- Received October 27, 1994.
- Accepted April 4, 1995.
Significant controversy exists about the primary pathophysiological defect in polycystic ovary syndrome (PCOS).1 There is now a broad consensus, however, that women with this condition are characterized by chronic anovulation in association with androgen excess, hyperinsulinemia, and alterations in gonadotropin secretion.1 2 The correlation between insulin and ovarian vein androgens is in the range of .88 to .99,3 and evidence suggests that the direction of causation is from insulin to androgens rather than the reverse: when androgens are reduced by a gonadotropin-releasing hormone analogue, insulin resistance remains in PCOS women,4 whereas pharmacological reduction of hyperinsulinemia leads to a reduction in androgen levels.5 Furthermore, reduction of hyperinsulinemia by weight loss in obese PCOS women is associated with a reduction in androgens and restoration of ovulatory menstrual cycles despite an unchanged gonadotropin profile as measured by luteinizing hormone (LH) level, pulsatility, and amplitude.6 Thus, insulin resistance appears to play a critical role in the pathophysiology of PCOS.1
Women with chronic anovulation, hyperandrogenism, hyperinsulinemia, and obesity have an increase in cardiovascular disease (CVD) risk factors.7 8 9 10 11 There have been few studies of the epidemiology of PCOS in free-living populations, however. Most investigations have used patients seen in outpatient clinics for infertility or irregular periods. Studies have generally used small numbers of women, varying definitions of PCOS, and different age criteria.
In 1985, Wild and colleagues7 found that women with PCOS had lower HDL levels, higher LDL/HDL ratios, and higher triglyceride levels than regularly menstruating women. More recently, Slowinska-Srzednicka et al8 have drawn attention to the role of insulin in the lipid abnormalities observed in hyperandrogenic women with PCOS. These investigators compared 27 women with PCOS and 22 eumenorrheic control subjects stratified by weight (obese and nonobese). Women with PCOS had significantly lower levels of HDL2 and higher levels of apoB and triglycerides. Multiple regression analysis within PCOS patients, after adjustment for age, body mass index (BMI), and sex steroids, revealed that fasting insulin was a significant explanatory variable for total triglyceride and apoA-1 levels. These results, which suggest that hyperinsulinemia independent of obesity may play a role in the lipid disturbances in PCOS, are supported by a 1992 study by Wild et al9 in which 31 women with evidence of androgen excess were treated for 3 months with a gonadotropin-releasing hormone agonist that suppressed ovarian estradiol and testosterone. Lipid profiles remained aberrant despite the sex steroid suppression and remained correlated with insulin resistance. It was concluded9 that lipoprotein abnormalities appeared to be associated more with insulin than with endogenous androgens or estrogens.
Dahlgren et al10 11 studied a retrospective cohort of 33 older women (mean age, 50 years) with ovarian histopathology typical of PCOS at wedge resection 22 to 31 years previously and 132 age-matched control subjects (mean age, 51.7±5.3 years). A total of 30% of PCOS women and 56% of control women had reached menopause. Compared with control subjects, PCOS patients had a higher prevalence of central obesity, diabetes, and hypertension and had higher basal serum insulin concentrations. The authors concluded that the increase in CVD and diabetes among PCOS patients should be viewed not only as a condition requiring treatment for anovulation and infertility but as a metabolic syndrome requiring ongoing medical surveillance.
Another approach to the study of hyperandrogenism and coronary artery disease (CAD) was conducted by Wild et al,12 who used a case series of 102 women with CAD consecutively evaluated between July 1986 and December 1987 via cardiac catheterization. Mean ages were 63.6 and 60.5 years for case and control subjects, respectively. A self-administered questionnaire was given to the study group concerning the distribution of body hair and prevalence of acne. No hormonal assays were obtained, nor was there a specific diagnosis of PCOS. A total of 52 women with and 50 women without CAD were noted, and twice as many case subjects as control subjects reported excess body hair.
In the present study, we seek to elaborate on these findings concerning the association between PCOS and CVD risk factors by using a matched case-control design with a larger sample than previously reported.
Subjects with PCOS were recruited from two sources: previous (1970 through 1986) and current (1987 through 1990) academic practice records. Initially, the charts of 2777 women seen at Magee-Womens Hospital between 1970 and 1986 were reviewed. Magee-Womens Hospital is the regional center for gynecological and obstetric care. These records were obtained from the previous director of reproductive endocrinology (Dr David Archer) and reviewed by the authors (E. Talbott, D. Guzick, and A. Clerici). During the 1970s and early 1980s, the diagnosis of PCOS was mainly based on clinical findings, as biochemical parameters (ie, hormone assays) were not in widespread use. Therefore, a presumptive clinical diagnosis of PCOS was made if there was a history of chronic anovulation in association with either clinical evidence of androgen excess (hirsutism) or if total testosterone concentration was >2 nmol/L or the LH/follicle-stimulating hormone (FSH) ratio was greater than 2. Based on our review of the infertility clinic records in the charts of 2777 women in the 1970 through 1986 cohort, 10% (274) met the criteria for PCOS. Of these 274, 190 had hirsutism; 84 did not have hirsutism but had elevated testosterone (n=24), increased LH/FSH ratio (n=49), or both (n=11). The average number of years of follow-up was 14, and the current age of these women was 40.2 years. Detailed tracing efforts were employed for this population. A total of 235 (85.7%) were interviewed by telephone, 21 refused to be interviewed, 15 were not located, and 3 were deceased. Vital status was also verified through the National Death Index. It was determined that 152 (64.7%) lived within the greater Pittsburgh area (<50 miles), and 112 of these women (73.7%) participated in the clinical phase of our study.
The current group (1987 through 1990) comprised 204 women with PCOS from the academic practice of the Division of Reproductive Endocrinology at Magee-Womens Hospital. In general, the diagnosis of PCOS was based on hormone parameters in this population. A total of 188 (92.2%) were interviewed by telephone, 12 refused to be interviewed, and 4 could not be located at the time of analysis. Of these women, 140 (74.5%) lived in the greater Pittsburgh area, and 99 (70.7%) of these participated in the clinical phase. The final case population is shown in the Figure⇓. For both current and previous cohorts, a total of 278 women lived within a 50-mile radius who were eligible to participate in the clinical phase of the study; of these, 211 actually participated.
Control Subject Recruitment
Age (±5 years)– and race-matched neighborhood control subjects were selected by using a combination of voters’ registration tapes for 1992 for the greater Pittsburgh area and Cole’s Cross Reference Directory of households.13 The major sources of control candidates were the Allegheny County and Westmoreland County voters’ registration tapes. A nonrepeating random selection process was used that matched on date of birth, sex, race, and zip code of the case subject. For each PCOS case subject, five control subjects were randomly drawn, and a letter was then sent to the first person listed. This letter explained the nature of the study and invited the person to participate. A preaddressed, stamped postcard was included in the letter. After 2 weeks, a follow-up letter was sent to the same address.
All participants completed a telephone questionnaire conducted by a trained interviewer. Demographic information was collected regarding age at interview, date of birth, current marital status, occupation, and highest level of education attained. A detailed reproductive and gynecological history was obtained, including total number of pregnancies, live births, age at menarche, age at menopause (if applicable), reason for menopause (natural, surgical, drug-induced, etc), previous use of fertility medication (including type and duration), and use of oral contraceptives or hormone replacement therapy. A disease history was also obtained, including history of heart disease, cancer, and diabetes and treatment for hypertension. A positive history of myocardial infarction, stroke, angina, and other circulatory problems was documented. Surgical history included reproductive organ–related surgeries (hysterectomy, oophorectomy, or ovarian wedge resection) as well as nonreproductive organ–related surgeries. Self-reported height and weight were collected in the anthropometric section of the interview.
Upon completion of the telephone questionnaire, those participants residing within the greater Pittsburgh area were asked to participate in the clinical phase of our study by coming to Magee-Womens Hospital. Subjects were evaluated after a 12-hour fast. An additional questionnaire was administered on-site that included a repeat medical history, review of medication use, current medical practices, and family history of PCOS. Height was measured and rounded to the nearest half inch; weight was measured and rounded to the nearest half pound. Two measures each of waist and hip were collected, and the average value was recorded. Waist/hip ratio was calculated as the waist circumference divided by the hip circumference. Wrist circumference was measured for estimation of frame size. The subject’s blood pressure was measured by using a random-zero sphygmomanometer, and the average of two values was recorded. A fasting blood sample was obtained for lipid and hormone assays.
Lipid and Lipoprotein Measurement
Serum concentrations of total cholesterol, total HDL cholesterol (HDL-C), subfractions HDL2 and HDL3, triglycerides, LDL, and VLDL were measured in the Heinz Lipid Laboratory at the University of Pittsburgh. Total cholesterol was determined by using the enzymatic method of Allain et al.14 Duplicate samples with standards, control sera, and serum calibrators were included in each run. The coefficient of variation (CV) between runs was 1.3%.
HDL-C was determined after selective precipitation by heparin/manganese chloride and the removal by centrifugation of VLDL and LDL.15 The cholesterol was measured as described above for total cholesterol. Duplicate samples, standards, and control sera were included in each run. The CV between runs was 2.1%. After precipitation and removal of VLDL and LDL by heparin/manganese chloride as described above, the supernatant was mixed with dextran sulfate.16 The HDL2 precipitates and the HDL3 in the supernatant were then measured using the methods described above for cholesterol measurement. The HDL2 content was estimated by subtracting the HDL3 level from the total HDL content. Duplicate samples, standards, and control sera were included in each run. The CV between runs was 6.0%. LDL concentration was estimated by using the Friedewald formula.17 Triglycerides were determined by using the enzymatic procedure of Bucolo and David.18 The CV between runs was 1.7%. Blood for glucose and insulin determinations was obtained when the subjects were fasting. Plasma glucose was analyzed by using an enzymatic assay (Yellow Springs Glucose Analyzer, Yellow Springs Instruments) and plasma insulin by radioimmunoassay.
LH and FSH were measured by using an immunofluorometric assay (Delfia, human LHSpec, human FSH, and prolactin; Wallac Inc). These methods were chosen for their high degree of sensitivity and low sample volume required (25 μL per determination). The intra-assay and interassay CVs ranged between 2% to 4% and 7% to 9%, respectively, for the three pituitary hormones. The sensitivity of both the LH and FSH assays was 0.05 IU/L. Total testosterone, androstenedione, and estradiol levels were determined by using a radioimmunometric assay (Coat-A-Count, Diagnostic Products Corp). Testosterone and estradiol levels were measured directly. Serum samples for the determination of androstendione were first extracted with an ethyl acetate–hexane (3:2) mixture, resuspended in the assay diluent, and then measured by radioimmunoassay. Sensitivities for total testosterone, androstendione, and estradiol were 0.04 ng/mL, 0.02 ng/mL, and 8 pg/mL, respectively. These immunoassays exhibited CVs ranging from 3% to 4% and 9% to 13% for the intra-assay and interassay, respectively.
All data collected for this study were entered and verified by using spss data-entry software. Univariate analyses were performed by using spss for windows. Mean levels of selected CVD risk factors were compared by using Student’s t tests in spss. Linear regressions were then performed on variables that remained significant (P<.05). To normalize the distribution of the triglyceride data, log transformations were performed before linear regression analyses were performed on the data by using spss for windows. sas for windows was used for matched-pair conditional logistic regression analyses, with case-control status as the dependent variable.
Of the 478 women in the final cohort, 278 (58.2%) lived within 50 miles of the research facility; of these, 211 (75.9%) participated in the clinical portion of the study, 206 of whom are included in the matched-pair analyses. One hundred thirty (of 220) of the first eligible control subjects consented to participate. Fifty-four control subjects were obtained from the second eligible name, and the remaining 36 were obtained after three or more attempts. The response rate for recruitment of first or second eligible control subjects was 83.6%.
The mean age of the total cohort was 35.1±7.9 years. The women in our study were white (n=388; 92.4%), black (n=17; 4.0%), or other races (n=15; 3.6%). Almost two thirds of the case subjects were currently married, 122 (29.0%) were never married, and 35 (8.4%) were separated or divorced. Thirty-one percent obtained a high school degree, and the remainder had some schooling beyond high school; almost 20% had received a college degree or higher. Approximately 23% of case subjects with PCOS were current smokers.
Table 1⇓ presents the age and race distribution of all PCOS cases by source. The characteristics of the women who were seen at Magee-Womens Hospital Clinic versus those who were interviewed only on the telephone were very similar. The proportion of individuals seen in the clinic on estrogen replacement or oral contraceptives was 41 (19.4%), compared with 33 (16.0%) in the telephone interview–only group. Thirteen (6.2%) of the cases evaluated clinically and 16 (7.6%) of those not seen reported a surgical menopause. Corresponding figures for natural menopause were 3 (1.5%) and 6 (2.8%), respectively, for case and control subjects.
A total of 206 control subjects who met the matching criteria also visited the clinic. Table 2⇓ shows selected sociodemographic and reproductive factors for case subjects and matched control subjects. Case subjects were 35.9±7.4 and control subjects were 37.2±7.8 years at the time of the initial interview; 33.2% of the case and 40.8% of the control subjects were over 40 years old at the time of the interview.
A similar proportion of case (88.4%) and control (86.0%) subjects reported that they were still menstruating. Surgical menopause occurred in 14 (7%) of PCOS patients and 21 (10%) of the control subjects. Similar rates of hysterectomy with bilateral oophorectomy were reported (case subjects, 5%; control subjects, 5.8%). Twenty percent of control and 18.1% of case subjects reported taking estrogen replacement or oral contraceptives. Presence of hirsutism or removal of unwanted hair from face or body was reported in 43% of the case and 13% of the control subjects (P<.01).
Case subjects weighed more than control subjects (176.9±49 versus 153.76±35.7 pounds, respectively, P<.001). Almost 50% of the case and 27.8% of the control subjects weighed 170 pounds or more (P<.01). Wrist circumference, a surrogate for frame size, was 16 or greater in 62.8% of the case subjects and 34.4% of the control subjects (P<.01). Shown in Table 3⇓ are other anthropometric and hormonal parameters. Mean BMI and waist/hip ratio were significantly higher among case than control subjects (BMI, 30.5±8.3 versus 26.3±6.5, P<.001; waist/hip ratio, 0.82±.14 versus 0.76±.07, P<.001).
Mean total serum testosterone level was higher among case than control subjects (1.66 nmol/L versus 1.11, P<.0001). Serum samples from 28 case and 29 control subjects were selected randomly to compare the correlation of bioavailable testosterone to total testosterone; the results of the Spearman test indicated a highly significant correlation (r=.9241, P<.001). The mean LH value for PCOS case subjects under 40 years old was 12.5±1.4 mIU/mL compared with 7.7±1.3 mIU/mL for control subjects (P<.05).
BMI, waist/hip ratio, and insulin, which were significantly different between case and control groups in univariate analysis, were included in a multivariate matched-pair conditional logistic regression with case-control status as the dependent variable. All three variables were forced into the model. Waist/hip ratio was a significant predictor (P=.045), followed by insulin, which was almost significant (P=.075). BMI was highly correlated with these variables and did not show a significant independent contribution. One SD for each independent variable was used for the unit of risk for the odds ratios in Table 4⇓.
Serum total cholesterol, LDL cholesterol (LDL-C), and triglyceride levels were significantly higher for case than control subjects (respectively, 195.4±33.5 versus 185.6±37.8 mg/dL, P<.01; 118.4±31.5 versus 110.7±34.6 mg/dL, P=.03; and 129.0±88.8 versus 85.9±63.4 mg/dL, P<.001) (Table 5⇓). HDL-C was lower in case than control subjects (51.1±14.5 versus 57.8±14.5 mg/dL, P<.0001). Fasting insulin levels among case (23.5±17.9 μU/L) and control (13.6±8.7 μU/L) subjects differed significantly (P<.0001). Mean systolic blood pressure among case subjects was 113.6±14.3 mm Hg and among control subjects, 110.3±12.7 mm Hg (P=.025). Mean diastolic pressure was not significantly different. The prevalence of a mean diastolic blood pressure ≥90 mm Hg was 5.9% among case subjects and 3.3% among control subjects (NS).
Sixteen case (8%) and 6 control (3%) subjects reported a history of hypertension as diagnosed by their physicians. Nine of the 16 (5%) were currently taking blood pressure medication versus 3 (2%) of the control subjects. One case and one control subject reported a history of insulin-dependent diabetes, and non–insulin-dependent diabetes was noted in 8 case subjects (4%).
A comparison of CVD risk factors by cross-sectional age strata for case and control subjects (Table 6⇓) showed higher levels of CVD risk factors among PCOS patients at an earlier age than control subjects, with differences decreasing with age.
By using multiple regression analysis we determined whether lipid levels and blood pressure were higher among PCOS patients than control subjects after adjusting for other risk factors: BMI, fasting insulin, age, and use of exogenous hormones and/or oral contraceptives. PCOS women had significantly higher lipid levels than control subjects after adjusting for potentially confounding variables. There were no significant differences in blood pressure in the multiple regression analysis (Table 7⇓).
In this study, we have documented higher levels of total cholesterol, LDL-C, triglycerides, and insulin and lower levels of HDL-C, especially HDL2-C, among women with PCOS compared with matched control subjects. The PCOS patients were probably heterogeneous in terms of both etiology and clinical presentation. As noted, however, the risk-factor differences were robust and appear to have been little affected by variations in clinical presentation. There were also substantial differences in BMI and waist/hip ratio. However, even after adjusting for BMI and fasting insulin, the differences in risk factors between PCOS patients and matched control subjects persisted.
One of the most important observations is that the differences in risk factors between PCOS women and control subjects are generally stronger at earlier ages. This is probably due to an early onset of hormonal changes and obesity and possibly to the distribution of intra-abdominal fat among PCOS patients. The data presented, however, are cross-sectional; longitudinal follow-up will provide answers regarding the changes in risk factors over time, especially among the younger PCOS women.
One can argue that PCOS represents one of the best examples of syndrome X as defined by Reaven.19 According to our data, the risk factors in the PCOS women are probably elevated at an earlier age than among non-PCOS women. If the central obesity, hyperinsulinemia, and low HDL-C and high triglyceride levels noted in the PCOS cases are really a unique profile of risk factors of atherosclerosis and subsequent coronary heart disease in women, then women with PCOS should have much more atherosclerosis than control subjects, especially at younger ages. If the risk of atherosclerosis is primarily related to elevated LDL-C, then there is an alternative reason for a higher prevalence of atherosclerosis in PCOS patients. Comparison of the evaluation of the risk of atherosclerosis in relation to these different risk factors may provide an estimate of the independent contributions of LDL-C compared with the syndrome X profile.20 PCOS women in many ways present risk-factor characteristics found in younger men and also among older obese postmenopausal women. It will be important in the subsequent follow-up and evaluation to determine whether the clinical presenting characteristics of PCOS women are related to changes in risk factors or to the development of atherosclerosis.
The etiology of PCOS is still not fully understood. The possible association with insulin metabolism and genetic variation (5-α reductase)21 22 are interesting and potentially very important. The observation that at least some PCOS women represent an autosomal dominant genetic disorder and that the male counterpart is defined by premature baldness and central obesity is of great interest.23 24 25 Male-pattern baldness and central obesity are risk factors for CVD.19 23 It will be very important to evaluate both male and female siblings for risk factors for CVD in order to determine both familial aggregation of PCOS in women and the counterpart male syndrome. The extent of CVD and other aspects of atherosclerosis should be evaluated in these families.
This work was supported through a grant from the National Institutes of Health (R01,HL4664-01A2).
Barnes R, Rosenfield R. The polycystic ovary syndrome: pathogenesis and treatment. Ann Intern Med. 1989;110:386-397.
Slowinska-Srzednicka J, Zgliczynski S, Wierzbicki M, Srzednicki M, Stopinska-Gluszak U, Zgliczynski W, Soszynski P, Chotkowska E, Bednarska M, Sadowaki Z. The role of hyperinsulinemia in the development of lipid disturbances in non-obese and obese women with the polycystic ovary syndrome. J Endocrinol Invest. 1991;14:569-575.
Cole’s Cross Reference Directory. Allegheny County and Greater Pittsburgh Area. 1991-1993. Lincoln, Neb: Cole Publications and Information Services; 1993.
Allain CC, Poon LS, Chan CSG, Richmond W, Fu PC. Enzymatic determination of total serum cholesterol. Clin Chem. 1974;20:470-475.
Warnick GR, Albers JJ. A comprehensive evaluation of the heparin-manganese precipitation procedure for estimating high density lipoprotein cholesterol. J Lipid Res. 1978;19:65-76.
Gidez LI, Miller GJ, Burstein M, Slagle S, Eder HA. Separation and quantitation of subclasses of human plasma high density lipoproteins by a simple precipitation procedure. J Lipid Res. 1982;23:1206-1223.
Friedewald WT, Levy RI, Fredrickson DS. Estimation of the concentration of low-density lipoprotein cholesterol in plasma, without use of the preparative ultracentrifuge. Clin Chem. 1972;18:499-502.
Bucolo G, David H. Quantitive determination of serum triglycerides by the use of enzymes. Clin Chem. 1973;19:476-482.
Reaven GM. Role of insulin resistance in human disease. Diabetes. 1988;37:1595-1607.