Polycystic Ovary Syndrome, Metabolic Syndrome, and Inflammation in the Hispanic Community Health Study/Study of Latinos

Abstract

Context

Polycystic ovary syndrome (PCOS) is a multifaceted endocrine disorder with reproductive and metabolic dysregulation. PCOS has been associated with inflammation and metabolic syndrome (MetS); however, the moderating effects of inflammation as measured by C-reactive protein (CRP) and menopause on the PCOS-MetS association have not been studied in Hispanic/Latinas with PCOS who have a higher metabolic burden.

Objective

We studied the cross-sectional association between PCOS and (1) MetS in 7316 females of the Hispanic Community Health Study/Study of Latinos (HCHS/SOL), (2) subcomponents of MetS including impaired fasting glucose (IFG) and elevated triglycerides (TGL), and (3) effect modification by menopausal status and CRP.

Design

The HCHS/SOL is a multicenter, longitudinal, and observational study of US Hispanic/Latinos. Our study sample included females from visit 2 with self-reported PCOS and MetS (ages 23-82 years).

Results

PCOS (prevalence = 18.8%) was significantly associated with MetS prevalence [odds ratio [odds ratio (OR) = 1.41 (95% confidence interval: 1.13-1.76)], IFG and TGL (OR = 1.42 (1.18-1.72), OR = 1.48 (1.20-1.83), respectively]. We observed effect modification by menopausal status (OR_pre = 1.46, P_int= .02; OR_post = 1.34, P_int= .06) and CRP (OR_elevated = 1.41, P_int= .04; OR_normal = 1.26, P_int= .16) on the PCOS-MetS association. We also observed a superadditive interaction between CRP and PCOS, adjusting for which resulted in an attenuated effect of PCOS on MetS (OR = 1.29 [0.93-1.78]).

Conclusion

Hispanic/Latino females with PCOS had higher odds of MetS, IFG, and elevated TGL than their peers without PCOS. Interaction analyses revealed that the odds of MetS are higher among PCOS females who have premenopausal status or high inflammation. Interventions in Hispanic/Latinas should target these outcomes for effective management of the disease.

Polycystic ovary syndrome (PCOS) is characterized by chronic anovulation, hyperandrogenism, and polycystic ovaries. PCOS starts during puberty, and women who are predisposed to PCOS demonstrate many abnormal endocrine characteristics during puberty (1-3). These characteristics persist through adulthood and interfere with other physiologic processes in the body that pertain to insulin resistance, metabolic syndrome (MetS), cardiovascular risk, and liver function (4, 5). PCOS is recognized as a disorder of the hypothalamus-pituitary-gonadal axis (6). It has been found that, due to hyperinsulinemia in PCOS patients, fatty acids from the central adipose tissues enter the blood circulation and supply more substrate to the liver for triglyceride production, resulting in elevated triglycerides (7, 8). Our study aims to investigate the association between PCOS and MetS and examine MetS subcomponents and possible effect modifiers (eg, menopausal status and inflammation).

PCOS affects 1 in 10 women in the United States (9, 10). and twice as many US Hispanic/Latinas (11, 12). Hispanic/Latino adults also experience a disparate burden of obesity and metabolic abnormalities (13, 14). Studying this population therefore is crucial and is made more germane given that the Hispanic/Latino population reached 62.1 million in 2020 (18.7% of total US population), ranking as the second largest race or ethnic group in the United States after non-Hispanic White Americans (15). Many US Hispanic/Latino adults face disparities in the social and environmental determinants of health. As a result, psychosocial stress, obesity, diabetes, and mental illness have been drastically increasing in this population (compared to non-Hispanic/Latino individuals), and this population is also faced with low health insurance coverage rates (16). Hispanic/Latino individuals exposed to chronic stressors have 1.5 times the odds of obesity compared to those who are not exposed to chronic stressors (14). The metabolic disparities faced by Hispanic/Latino adults include high central adiposity and dyslipidemia, as compared to non-Hispanic White adults (17). Given that PCOS is an endocrine malfunction that results in subfertility and MetS, it is unknown if all subcomponents of MetS are equally related to PCOS among Hispanic/Latinas. Building on prior work in this population (18), we analyze not only the association between PCOS and MetS in this high-risk population of Hispanic/Latina females but also the association with all MetS subcomponents.

Both PCOS and MetS have components of insulin resistance (19-21), but whether other factors such as inflammation levels and menopause status moderate the effect of PCOS on MetS in this high-risk population is not known. Given that metabolic dysfunction progresses after menopause, postmenopausal females with PCOS have been found to have additional MetS and cardiovascular risks (8, 22, 23). With the onset of menopause, physiological changes such as decreased estrogen levels, increased blood pressure, and increased body weight, among others, place females with PCOS at a higher risk for MetS (22, 24, 25). However, the presence of impaired glucose metabolism, gonadotropin and estrogen dysregulation, and thereby the polycystic ovarian morphology during the reproductive years have been associated robustly with MetS and mark the formative years for PCOS pathophysiology (18). Studying the postmenopausal period could provide information that leads to better interventions for patients with PCOS later in life to mitigate cardiovascular risk. Therefore, this study aims to characterize the moderating effect of menopausal status in the association between PCOS and MetS.

In addition to studying the moderating effect of menopausal status, we also aim to examine the moderating effect of inflammation in the PCOS-MetS relationship. A study found that TH1 (proinflammatory cytokine) to TH2 (anti-inflammatory cytokine) ratio is higher in the follicular fluid of females with PCOS compared to those without (26). A meta-analysis of 31 articles showed that PCOS was associated with higher circulating C-reactive protein (CRP), specifically, 96% higher levels compared to controls (27). Obesity and insulin resistance, which are features prevalent in PCOS patients, have also been associated with chronic inflammation (28-30). Further, it has been shown that PCOS is associated with chronic low-grade inflammation (27, 31). Hence, we study here the moderating role of inflammation as measured by CRP in the PCOS-MetS association.

In this study, we sought to analyze the association between PCOS and MetS, and its subcomponents, in the Hispanic Community Health Study/Study of Latinos (HCHS/SOL) and assess whether this effect is modified by menopause status and CRP. Although PCOS and MetS have been previously studied in premenopausal females of this cohort (18), our study aims to replicate these associations and assess the influence of menopause to provide insight into how these relationships differ across the pre- and postmenopause periods of the female life course in Hispanic/Latinas. Additionally, we further assessed the effect modification by inflammation, as measured by elevated CRP, in the association between PCOS and MetS. We also hypothesized that the use of metformin (an antidiabetic medication that lowers blood glucose levels and increases insulin sensitivity) may affect the association between PCOS and MetS given that insulin resistance plays a huge role in PCOS. We also hypothesized that menopausal status and CRP may interact additively with PCOS to elevate MetS risk in our study population (32, 33)—US Hispanic/Latinas—who bear a disparate burden of both PCOS and MetS (18, 34-36).

Materials and Methods

Study Design and Population

The HCHS/SOL is a large multisite longitudinal cross-sectional study of cardiovascular, pulmonary, and psychosocial determinants of health across the lifespan (37). The study recruited 16 415 self-identified Hispanic/Latino participants (18-74 years) between March 2008 and June 2011 from 4 sites (Bronx, NY; Chicago, IL; Miami, FL; San Diego, CA). The baseline study design and sampling methods have been published previously (37, 38). Briefly, the participants were chosen from households that were selected randomly by census block groups from its 4 sites to mimic the general composition of Hispanic/Latino backgrounds (or heritages) in the United States. Participants represent various backgrounds including Cuban, Dominican, Mexican, Puerto Rican, Central American, and South American. For visit 2 (V2; 2014-2017), 11 623 adults returned to the clinic for additional measurements.

Only adults identified as female who completed questionnaires on reproductive histories and pregnancy complication histories at V2 (n = 7342) were included in this study. We excluded participants who had indicated that they were pregnant at baseline (n = 10) and those who did not have values for PCOS or MetS variables (n = 16). Our final analytic sample size was 7316 and did not differ significantly on any observable study characteristics from the full study sample. All participants provided informed consent, and the HCHS/SOL sites and coordinating center obtained institutional review board approval.

Study Measurements

Self-reported PCOS

At V2 (n = 11 623), the Reproductive Medical History questionnaire was administered to all females and included a question on menstrual cycles (“How many days did your typical menstrual cycle last, that is, how many days were between the beginning of one menstrual period to the beginning of bleeding of the next period?”), which provided 4 response categories: “less than 24 days,” “24 to 35 days,” “greater than 35 days,” or “too irregular to report.” While answering the question, participants were asked to think back to when they were at the age of 20 to 40 years and not on any birth control or hormone medications, not pregnant, and not breastfeeding. Having PCOS signs was defined by the presence of irregular menstrual cycles (cycles that last longer than 35 days or cycles that are “too irregular to report”) or taking hormones to regulate periods. Responses to a self-report question on PCOS (“Has your healthcare provider ever told you that you have polycystic ovary syndrome or PCOS?”) were coded as self-reported PCOS. All females from V2 were characterized as having PCOS if they had either the signs of PCOS (n = 889) or had a self-reported PCOS diagnosis (n = 350). Both characteristics were combined into a single variable for PCOS.

Other reproductive measures

Questions on whether a participant ever used birth control and, if yes, in which form (pills, ring, shot, implant, or intrauterine device), were asked at V2 as part of the reproductive questionnaire. We used this information to derive a birth control use variable. Additionally, the questionnaire collected information regarding age at natural menopause, reasons for why periods stopped, whether the participant experienced infertility for greater than 1 year, and menstrual cycle lengths.

Metabolic outcomes

Metabolic syndrome at V2 was defined by National Cholesterol Education Program Adult Treatment Panel III guidelines (39). Subcomponents of MetS were derived from laboratory measures collected at V2. According to the criteria, MetS was characterized as having 3 or more of the following: (1) abdominal obesity measured as a waist circumference (WC) greater than or equal to 88 cm, (2) triglycerides (TGL) greater than or equal to 150 mg/dL, (3) high-density lipoprotein (HDL) less than 50 mg/dL, (4) blood pressure (BP) greater than or equal to 130 mm Hg systolic and/or greater than or equal to 85 mm Hg diastolic and/or on antihypertensives, and (5) fasting glucose (FG) (>10 hours) greater than or equal to 100 mg/dL and/or on drug treatment or diagnosed as diabetic (40). Diabetes status was derived from laboratory measures of FG, self-report, and use of antidiabetic medication. Homeostatic model assessment for insulin resistance (HOMA-IR) was derived from laboratory measures.

Measures of effect moderators

We operationalized the menopause status variable to include both natural and surgical menopause indicators. We categorized females who answered that their menstrual periods ended naturally for over a year before the V2 examination into the natural menopause category. Those who answered that their periods stopped due to surgery such as hysterectomy, oophorectomy, endometrial ablation, radiation/chemotherapy, or some other reproductive surgery such as to treat ectopic pregnancies were classified into the surgical menopause category. We classified females who reported menopause due to either natural or surgical reasons into the postmenopausal category. In addition, we classified females who said they had experienced menopause and their periods stopped due to birth control (eg, hormonal intrauterine devices) or other hormonal treatment to stop periods into the postmenopausal category. For analyses including the menopause status variable, we excluded those who reported their age at menopause at V2 examination as greater than their age at V2 (n = 2), who were unsure about their menopause status and then skipped follow-up questions on reasons for menopause (n = 31), or who were missing all menopause variables (n = 191). The analytic sample size for analyses involving menopause variables was 7098 and did not differ significantly on any observable study characteristics from the full study sample.

Information from baseline examination on CRP was analyzed dichotomously (CRP lesser than 3.0 mg/L as normal levels, and greater than or equal to 3.0 mg/L as elevated CRP). The concentration of 3.0 mg/L was selected as the threshold for being at elevated risk of cardiovascular disease based on the American Heart Association (AHA)/Centers for Disease Control and Prevention (CDC) scientific statement for healthcare professionals (41). Unfortunately, CRP was not measured at V2 but was used in our analyses of V2 to adjust for past inflammation (∼6 years earlier).

Variables in sensitivity analyses

Metformin use was derived from the Medication Use Questionnaire administered at V2 of the HCHS/SOL where participants were asked to bring prescribed or over-the-counter medications taken in the 4 weeks preceding the baseline visit. Metformin use was extracted from the variables reported on medication use and then operationalized as a dichotomous variable (0/1). Body mass index (BMI) at V2 was derived by the HCHS/SOL using anthropometric measures of height and weight (BMI = weight in kg/[height in meters]²). Standing height was measured in centimeters using a wall-mounted stadiometer with a backboard and headboard. Weight was measured using a digital scale in kilograms.

Statistical Analyses

All models to test the association between PCOS, MetS, and its subcomponents considered the complex sampling design of the HCHS/SOL by accounting for sampling weights, primary sampling units, and strata. V2 sampling weights account for nonresponse adjustments and calibration to known population totals, which were taken from the 2010 decennial census (37).

Using directed acyclic graphs and model fit statistics (not shown here), we determined an a priori set of covariates to include in our models. Akaike information criteria (AIC), calculated from maximum likelihood estimate of the model, to explain the greatest amount of variance using the least number of independent variables was used to determine model fit. A lower AIC score corresponded to greater model fit, and models with lower AIC scores were used to determine which covariates to include. In our models, we observed that including the following variables as covariates yielded the best fit: age, study site (Chicago, IL; Bronx, NY; Miami, FL; San Diego, CA) by Hispanic/Latino background (Dominican, Central American, Cuban, Mexican, Puerto Rican, South American, more than 1/other heritage) categories, education (1—no high-school diploma or GED; 2—at most a high-school diploma or GED; 3—greater than high school (or GED) education), age at immigration (US born or immigrated between the ages of 0-12, 12 to 25, 25 to 35, 35 to 55, or more than 55), and health insurance coverage. All models adjusted for these covariates and all categorical levels of the covariates had more than 100 individuals at baseline. All analyses were 2-sided with a significance set at the level of .05. All statistical analyses were performed in Statistical Analysis Software, SAS 9.4 (SAS Institute, Cary, NC).

Demographic characteristics and regression analyses

Means, frequencies, and SEs for demographic and reproductive characteristics were calculated by the presence or absence of PCOS (signs and self-report). Weighted frequencies and SEs for MetS and its subcomponents were calculated using survey frequency procedures by PCOS status. Associations between PCOS and MetS were analyzed in a series of multivariate logistic regression models. In a series of linear regression models, we also sought to describe the relationship between PCOS, age, menopause status, high sensitivity CRP, metformin use, and BMI on continuous cardiometabolic subcomponents of MetS including total cholesterol (TCL), HDL, low-density lipoprotein (LDL), TGL, systolic and diastolic BP, WC, and FG. We also analyzed study variables by stratifying them into categories of CRP and menopausal status. Weighted means and frequencies and their corresponding SEs for all study demographic characteristics and cardiometabolic health characteristics, including MetS and PCOS status, were reported by stratifying the sample by menopause status and CRP. For these stratification analyses, CRP was further categorized into 4 levels: low (0-1.0 mg/L), moderate (>1.0 and ≤3.0 mg/L), high CRP (>3.0 and ≤10.0 mg/L), and very high or chronic inflammation (>10.0 mg/L). We further evaluated the associations between select study characteristics and metformin use in minimally adjusted models and fully adjusted models. Minimally adjusted models accounted for age, study site by self-reported Hispanic/Latino background, and education. Fully adjusted models accounted for all the covariates included in the minimally adjusted models plus age at immigration and health insurance status. We evaluated the effect of BMI on select cardiometabolic health characteristics in a similar way.

Interactions with PCOS on MetS

Interactions between exposures and outcomes can be modeled intentionally to measure deviation from either perfect additivity or multiplicativity. From a biological and public health standpoint, however, it is important to estimate interaction effects on an additive scale. To estimate interactions between exposures and MetS odds on an additive scale, we calculated the relative excess risk due to interactions (RERI) (42). For the same, we created disjoint indicator variables for all levels of exposures (PCOS—yes, no; CRP—low, high; menopausal status—premenopausal, postmenopausal) and calculated RERI for the effect of interaction between these exposures on the association with MetS odds. MetS and its subcomponents were analyzed in logistic regression models with PCOS, CRP, and menopausal status, adjusting for covariates and complex survey sampling measures (37, 43). The association of PCOS with MetS was evaluated in a series of models: model 1: with just PCOS as the exposure; model 2: with PCOS and menopause status as exposures; model 3: with PCOS and menopause status as exposures and an interaction term for PCOS and menopause status; model 4: with PCOS and CRP as exposures; model 5: with PCOS and CRP as exposures and an interaction term for PCOS and CRP; model 6: with all 3 variables (PCOS, menopause status, and CRP) as exposures and terms for interaction between PCOS and CRP, and CRP and menopause status. All models were adjusted for the covariates described earlier.

Sensitivity analyses

In addition to describing the associations between select cardiometabolic characteristics and metformin, regression models for MetS and subcomponents by PCOS status were also evaluated by adjusting for metformin use. Similar sensitivity analyses were also run for BMI. Additionally, we assessed the interaction between PCOS and CRP on MetS after adjusting for metformin use, BMI, and both, iteratively, along with our minimally adjusted model covariates.

Results

Demographic Characteristics

Our final sample had a weighted mean age of 47.9 years, ranging from 23 to 82 years. The overall prevalence of PCOS (either signs or self-report) was 18.8% in our sample (N_PCOS = 1161) and 44.3% were postmenopausal (N_POST = 4675). Roughly one-fifth of those who had MetS were PCOS cases, and 22.8% of premenopausal females were PCOS cases. The average age at natural menopause was 46.2 years, and average CRP concentration was 4.71 mg/L.

Descriptive demographics of our sample and cardiovascular health characteristics are provided in Table 1. Participants’ self-reported Hispanic/Latino background indicated that there were 11.2% Dominican, 7.7% Central American, 18.3% Cuban, 38.3% Mexican, 15.1% Puerto Rican, 5.1% South American, and 3.8% self-identified as having more than 1 or other Hispanic/Latino heritage. Sixty percent of the sample had achieved greater than a high school or GED education. Roughly 55% of the sample had less than 30 000 USD of annual household income and 72% reported that they had health insurance at V2. Prevalence of MetS in the sample was 38.1%, females with PCOS had higher levels of CRP (5.0 vs 4.7 mg/L), and females with PCOS had lower age at natural menopause compared to those without (43.9 vs 46.6 years). Demographic and cardiovascular characteristics were also summarized by menopausal status and 4 CRP levels [Supplementary Tables S1 and 2 (44)]. Overall, cardiometabolic characteristics (BMI, WC, LDL, TGL, BP, HOMA-IR, FG) were elevated and HDL concentration was lower in postmenopausal females as compared to premenopausal females. When assessing CRP [Supplementary Table S2 (44)], we noticed a dose-response relationship where the greater impairment of these cardiometabolic characteristics were associated with higher concentrations of CRP. For example, the mean concentration of TGL at low, moderate, high, and chronic inflammation were 98.2, 118.4, 121.3, and 123.9 mg/dL, respectively. Following a similar trend, the WC of the participants increased with the concentration of CRP: 89.5 cm, 96.6 cm, 102.8 cm, and 109.3 cm for those with low, moderate, high, and very high CRP, respectively.

Table 1.

Demographic and health characteristics of women from the HCHS/SOL study, overall and by PCOS exposure (self-reported and signs of PCOS) (overall n = 7316)

Weighted means or frequenciesa			Overall	PCOS (self-reported or signs of PCOS)
			(n = 7332)	No (n = 6156)	Yes (n = 1160)
	Unweighted n	Weighted na	Mean (SE) or weighted % (SE of %)a	Mean (SE) or %a	Mean (SE) or %a
Demographic characteristics
Age (years)	7316	6049	47.94 (0.30)	48.99 (0.33)	43.36 (0.59)
Hispanic background	7314	6035
Dominican	710	676.36	11.21 (0.85)	81.54 (2.17)	18.46 (2.17)
Central American	793	462.5	7.66 (0.68)	84.17 (1.95)	15.83 (1.95)
Cuban	919	1107	18.34 (1.51)	87.47 (1.50)	12.53 (1.50)
Mexican	3102	2309	38.26 (1.64)	79.47 (1.15)	20.53 (1.15)
Puerto Rican	1098	912.17	15.11 (0.85)	78.49 (1.74)	21.51 (1.74)
South American	495	307.71	5.09 (0.40)	81.30 (2.54)	18.70 (2.54)
More than one/other heritage	177	231.69	3.84 (0.40)	72.10 (4.21)	27.90 (4.21)
Missing	20	28.03	0.46 (0.15)	81.36 (13.13)	18.64 (13.13)
Study site
Bronx, NY	1718	1826	30.26 (1.55)	80.26 (1.20)	19.74 (1.20)
Chicago, IL	1815	885.77	14.68 (0.96)	81.73 (1.40)	18.27 (1.40)
Miami, FL	1751	1721	28.51 (2.03)	85.46 (1.28)	14.54 (1.28)
San Diego, CA	2032	1602	26.55 (1.68)	77.42 (1.49)	22.58 (1.49)
Education
No high school diploma or GED	1177	993.11	16.46 (0.65)	82.78 (1.67)	17.22 (1.67)
At most a high school diploma or GED	1733	1398	23.16 (0.77)	82.12 (1.45)	17.88 (1.45)
Greater than high school (or GED) education	4405	3643	60.37 (0.83)	80.42 (0.89)	19.58 (0.89)
Income
Yearly household income less than USD 30 000	4272	3347	55.46 (1.02)	81.67 (0.90)	18.33 (0.90)
Yearly household income more than USD 30 000	2424	2145	35.55 (1.05)	80.75 (1.16)	19.25 (1.16)
Missing	620	542.71	8.99 (0.52)	80.12 (2.28)	19.88 (2.28)
Age at immigrationb
0-12	1571	1752	29.03 (1.03)	76.09 (1.47)	23.91 (1.47)
12 to 25	1885	1640	27.17 (0.83)	80.27 (1.40)	19.73 (1.40)
25 to 35	1718	1262	20.91 (0.74)	82.61 (1.47)	17.39 (1.47)
35 to 55	1791	1064	17.64 (0.70)	86.55 (1.08)	13.45 (1.08)
55+	309	273.73	4.54 (0.37)	92.68 (2.42)	7.32 (2.42)
Missing	42	43.34	0.72 (0.16)	78.58 (9.41)	21.42 (9.41)
Health insurance coverage
Yes	5442	4366	72.35 (0.94)	80.91 (0.83)	19.09 (0.83)
No	1597	1342	22.23 (0.87)	82.92 (1.38)	17.08 (1.38)
Missing	277	327.22	5.42 (0.44)	78.14 (3.45)	21.86 (3.45)
Cardiometabolic health characteristics
BMI (kg/m2)	7098	5787	30.39 (0.14)	30.24 (0.14)	31.02 (0.34)
Waist circumference (cm)	7087	5782	98.30 (0.32)	97.97 (0.32)	99.64 (0.74)
Average systolic blood pressurec (mmHg)	7307	6028	119.01 (0.36)	119.75 (0.41)	115.68 (0.78)
Average diastolic blood pressurec (mmHg)	7313	6031	71.32 (0.21)	71.41 (0.23)	70.89 (0.47)
HDL-C(mg/dL)	7267	5981	53.81 (0.29)	54.01 (0.29)	52.94 (0.76)
LDL-C (mg/dL)	7185	5925	113.92 (0.69)	114.22 (0.72)	112.63 (1.46)
Triglycerides (mg/dL)	7266	5980	115.04 (1.21)	113.81 (1.32)	120.36 (3.42)
Glucose, fasting (mg/dL)	7245	5967	105.42 (0.63)	105.66 (0.74)	104.41 (1.42)
Diabetes (lab or self) or atidiabetic medication use
No	5216	4691	77.73 (0.69)	81.14 (0.85)	18.86 (0.85)
Yes	2100	1344	22.27 (0.69)	81.42 (1.19)	18.58 (1.19)
HOMA IR	7242	5964	4.14 (0.07)	4.03 (0.07)	4.62 (0.22)
Prevalence of metabolic syndrome
No	3912	3733	61.86 (0.84)	81.04 (0.94)	18.96 (0.94)
Yes	3404	2302	38.14 (0.84)	81.48 (1.17)	18.52 (1.17)
Age at natural menopause	4677	6049	46.22 (0.19)	46.63 (0.19)	43.86 (0.65)
High-sensitivity C reactive protein (mg/dL)b	7272	5982	4.71 (0.18)	4.65 (0.21)	5.00 (0.30)

All categories with fewer than 10 observations have been suppressed.

Abbreviations: BMI, body mass index; HCHS/SOL, Hispanic Community Health Study/Study of Latinos; HDL-C, high-density lipoprotein cholesterol; HOMA IR, Homeostatic Model Assessment for Insulin Resistance; LDL-C, low-density lipoprotein cholesterol; PCOS, polycystic ovary syndrome.

^a Weighted n, weighted means, frequencies, and SEs in this table were calculated using primary sampling units, sampling weights, stratification variables, or cluster identifiers to capture complex sampling in HCHS/SOL. All weighted for categorical outcomes percentages stratified by PCOS case control status are presented row-wise.

^b Measures (age at immigration and C-reactive protein were only derived from questionnaire/measured at visit 1).

^c Average of visit 1 and visit 2 blood pressures.

We assessed reproductive characteristics such as birth control use (hormonal, implants, or pill form), use of hormones for period regulation, menopause status, menopause reason, reasons why periods stopped, and infertility (Fig. 1) and report these characteristics stratified by PCOS status (Table 2). In our sample, 2.1% of females reported having menstrual cycle durations more than 35 days long and 11.7% had cycle lengths “too irregular” to report. Additionally, 31.4% were postmenopausal by natural causes and 12.7% by surgical causes. Less than 1% reported having undergone endometrial ablation or radiation. Around 11% reported that they had experienced infertility. Around 22.0% of those who reported surgical menopause had PCOS, and 11.5% reporting natural menopause had PCOS. We further analyzed our study characteristics separately in those who self-reported PCOS, those who had PCOS signs, and those who had both these subdefinitions. We observed that females who have both subdefinitions tended to have more impaired cardiometabolic characteristics followed by those with PCOS self-report (BMI 37.5 kg/m² compared to 31.3 in those with PCOS signs and 31.9 in those with PCOS self-report) [Supplementary Table S3 (44)]. Similar trends were also seen for fasting glucose, HOMA-IR, triglycerides, MetS, and infertility.

Figure 1.

Prevalence of select reproductive characteristics in the Hispanic Community Health Study/Study of Latinos sample. Weighted percentages of the reproductive characteristics are shown here along with their standard errors as error bars. Missing categories have been suppressed for the figure. Complete distributions for these characteristics can be found in Table 2.

Table 2.

Descriptive statistics for reproductive characteristics by exposure (self-reported and signs of PCOS) among women from the HCHS/SOL study (unweighted overall n = 7332)

Weighted means (*) or frequency and SEs	Unweighted frequency	Weighted frequency	Overall	PCOS (self-reported or signs of PCOS)
Reproductive characteristics			(n = 7332)	No (n = 6171)	Yes (n = 1161)
Birth control use (ever), %
Yes	4543	3834	63.39 (0.89)	79.88 (0.88)	20.12 (0.88)
No	2617	2069	34.20 (0.87)	82.47 (1.19)	17.53 (1.19)
Unsure/don't know	15	9.69	0.16 (0.06)	96.29 (3.85)	3.71 (3.85)
Missing	154	135.02	2.23 (0.30)	100 (0.00)
Refused to answer	3	0.82	0.01 (0.01)	51.36 (30.73)	48.64 (30.73)
Use of hormones for period regulation, %
Yes	470	500.84	8.28 (0.53)	49.62 (3.19)	50.38 (3.19)
No	4008	3270	54.05 (0.87)	84.32 (0.84)	15.68 (0.84)
Unsure/don't know	2	1.08	0.02 (0.01)	100 (0.00)
Don't use hormones at all	2700	2143	35.42 (0.88)	82.74 (1.18)	17.26 (1.18)
Missing	152	134.55	2.22 (0.30)	100 (0.00)
Cycle length, %
<24 days	451	437.19	7.23 (0.50)	89.63 (2.51)	10.37 (2.51)
24-35 days	5754	4536	74.98 (0.75)	94.38 (0.48)	5.62 (0.48)
More than 35 days	137	128.27	2.12 (0.25)	.	100 (0.00)
Too irregular to say	718	704.46	11.65 (0.59)	.	100 (0.00)
Refused to answer	2	1.28	0.02 (0.02)	100 (0.00)
Unsure/don't know	104	91.46	1.51 (0.21)	97.79 (1.33)	2.21 (1.33)
Missing	166	150.54	2.49 (0.30)	100 (0.00)
Menopause status, %
Premenopausal	2433	3168	52.37 (0.93)	77.18 (1.14)	22.82 (1.14)
Postmenopausal	4675	2679	44.28 (0.89)	85.44 (0.78)	14.56 (0.78)
Missing	224	202.11	3.34 (0.35)	89.26 (3.52)	10.74 (3.52)
Menopause reason, %
Premenopausal	2433	3168	52.37 (0.93)	77.18 (1.14)	22.82 (1.14)
Natural menopause	3434	1898	31.38 (0.75)	88.49 (0.75)	11.51 (0.75)
Surgical menopause	1223	768.52	12.71 (0.54)	77.98 (1.93)	22.02 (1.93)
Reason unknown	18	12.32	0.20 (0.06)	80.24 (12.06)	19.76 (12.06)
Missing	224	202.11	3.34 (0.35)	89.26 (3.52)	10.74 (3.52)
Age when periods stopped (years)	7332	6048.73	46.22 (0.19)	46.64 (0.19)	43.87 (0.65)
Reasons why periods stopped, %
Unsure/don't know	16	10.51	0.17 (0.05)	91.30 (6.43)	8.70 (6.43)
They stopped naturally	3434	1897	31.36 (0.74)	88.38 (0.75)	11.62 (0.75)
Surgery to remove uterus or ovaries	1210	746.08	12.33 (0.51)	79.10 (1.93)	20.90 (1.93)
Endometrial ablation	15	7.79	0.13 (0.04)	69.20 (14.30)	20.80 (14.30)
Radiation or chemotherapy	27	14.87	0.25 (0.06)	71.01 (10.28)	28.99 (10.28)
Other	102	97.36	1.61 (0.26)	67.71 (8.21)	32.29 (8.21)
Missing	152	133.46	2.21 (0.29)	99.71 (0.29)	0.28 (0.29)
Period didn't stop	2376	3142	51.94 (0.93)	77.11 (1.13)	22.89 (1.13)
Infertility (>1 year), %
No	6431	5255	86.87 (0.58)	82.97 (0.69)	17.03 (0.69)
Yes	723	640.05	10.58 (0.50)	63.19 (2.70)	36.81 (2.70)
Unsure/don't know	4	2.87	0.05 (0.03)	43.91 (29.25)	56.09 (29.25)
Missing	173	151.07	2.50 (0.30)	98.14 (1.16)	1.86 (1.16)
Refused to answer	1	0.18	0.003 (0.003)	100 (0.00)

*All percentages and SEs of percentages were adjusted by strata, cluster identifiers, primary sampling units, and weights to account for complex sampling.

Abbreviations: HCHS/SOL, Hispanic Community Health Study/Study of Latinos; PCOS, polycystic ovary syndrome.

Supplementary Table S4 (44) shows the descriptive statistics of MetS subcomponents stratified by PCOS status in our sample. As shown, 72.9% of females in the HCHS/SOL had elevated WC of which 18.7% had PCOS. Roughly one-fifth of those with high TGL and low HDL had PCOS, whereas 14.3% of those with hypertension had PCOS.

Regression Analyses

In logistic regression models estimating the association between PCOS and MetS (and its subcomponents) (Table 3), we found that PCOS was significantly associated with higher odds of MetS [odds ratio (OR) = 1.41% and 95% confidence interval (CI) = 1.13-1.76], with higher odds of IFG and elevated TGL [OR = 1.42 (95% CI = 1.18-1.72); OR = 1.48 (95% CI = 1.20-1.83), respectively, P < .01]. Upon assessing MetS and its subcomponents by PCOS subdefinitions (PCOS self-report and PCOS signs), we observed that PCOS signs were still significantly associated with higher odds of MetS [OR = 1.53 (95% CI = 1.17-2.01)], IFG [OR = 1.48 (95% CI = 1.17-1.88)], and elevated triglycerides [OR = 1.41 (95% CI = 1.11-1.96)], while PCOS self-report was significantly associated with only elevated WC [OR = 1.63 (95% CI = 1.01-2.63)] [Supplementary Table S5 (44)]. Overall, these sensitivity analyses show that the association between PCOS subdefinitions and MetS and its subcomponents are directionally consistent.

Table 3.

Association between PCOS and MetS and between PCOS and subcomponents of MetS, presented in the form of parameter estimates, SE, ORs of the effect and their 95% confidence intervals, and P-value of the association (overall n = 7316)

Outcome/association with PCOS	n used in the model	b (SE)	OR [LLCI-ULCI]	P-value	Model fitting criteria AIC
MetSa	6993	0.34 (0.11)	1.41 [1.13-1.76]	<.01	6699.35
MetS subcomponents
1. Elevated waist circumference (wc ≥ 88 cm)	6968	0.21 (0.12)	1.24 [0.99-1.55]	.07	5826.55
2. Hypertension (SPB > 130 mmHg, DBP > 85 mmHg) or on antihypertension medication	6993	−0.07 (0.11)	0.93 [0.75-1.16]	.53	5124.42
3. Impaired fasting glucose (>100 mg/dL) or metformin use or diabetic	6993	0.35 (0.10)	1.42 [1.18-1.72]	<.01	6668.53
4. Low HDL-cholesterol (≤50 mg/dL)	6993	0.16 (0.10)	1.17 [0.96-1.44]	.12	7729.98
5. Elevated triglycerides ≥150 mg/dL or lipid-lowering drug use	6993	0.39 (0.11)	1.48 [1.20-1.83]	<.01	6300.34

All models were adjusted for age, study site by Hispanic background, education, age at immigration, and health insurance coverage. All models were also adjusted for complex sampling design using strata, sampling weights, and clustering using primary sampling units. Bolded significant with P-value < .05.

The differences in overall n for each outcome are attributed to the missingness in the outcome variables.

Abbreviations: AIC, Akaike information criteria; DBP, diastolic blood pressure; HCHS/SOL, Hispanic Community Health Study/Study of Latinos; HDL, high-density lipoprotein; LLCI, lower limit of 95% confidence interval; MetS, metabolic syndrome; OR, odds ratio; PCOS, polycystic ovary syndrome; SBP, systolic blood pressure; ULCI, upper limit of 95% confidence interval.

^a Operationalized by National Cholesterol Education Program Adult Treatment Panel III, 2001. Updated by American Heart Association and the National Heart Lung and Blood Institute in 2005. According to this criteria, MetS is present if 3 or more subcomponents of the 5 are elevated.

We conducted a series of multivariate linear regression models to estimate the relationships among variables, which revealed that PCOS was significantly associated with TCL, TGL, and WC [Supplementary Table S6 (44)]. Age was significantly associated with all cardiometabolic risk factors, whereas menopausal status was significantly associated with TCL, LDL, and WC (P < .05). In our adjusted models, continuous high-sensitivity CRP was significantly associated with HDL [b = −5.14 (SE = 0.50)], TGL [b = 9.65 (2.29)], diastolic BP [b = 2.08 (0.36)], WC [b = 10.32 (0.50)], and FG [b = 10.74 (1.42)]. Metformin use was significantly associated with all cardiometabolic factors except BP; BMI was also associated with all risk factors except TCL.

Menopause and CRP Interactions With PCOS on MetS

To estimate the deviations of our observed interactions from an additive scale, we estimated RERI (42) and present the effect of cross-classification of PCOS status by CRP levels (PCOS × CRP) on MetS, PCOS status, and menopausal status levels (PCOS × MENO), and menopausal status and CRP levels (MENO × CRP) in Table 4. Estimated RERI for PCOS × CRP interaction and MENO × CRP interaction were 0.69 and 0.29, respectively, indicating that both exposures in each combination interact in a superadditive (or positive, when the RERI >0) manner to exacerbate their individual effects. There was a weak negative interaction between PCOS and menopausal status on odds of MetS when modeled on an additive scale (RERI = −0.03), indicating that there is a subadditive interaction. This, paired with the insights from the interaction between PCOS and menopausal status on the multiplicative scale, show that while the PCOS-MetS effect may be stronger in premenopausal females, it may not have a strong effect in postmenopausal females. This indicates that menopausal status is a weak effect modifier in this relationship and therefore also of less import to public health interventions to lower MetS burden compared to CRP.

Table 4.

Modeling interactions on the ratio scale between exposures and odds MetS and using combined indicator variables to estimate RERI method on an additive scale

Outcome	Interaction	Indicators for RERI estimation	b (SE)	OR [95% CI]	P-value	RERI	Model fit (AIC)
MetS	PCOS × CRP	low CRP, PCOS control	REF	REF	REF	0.69	6500.94
		high CRP, PCOS control	0.84 (0.08)	2.32 [1.97-2.73]	<.01*
		low CRP, PCOS case	0.23 (0.16)	1.26 [0.92-1.73]	.16
		high CRP, PCOS case	1.18 (0.16)	3.27 [2.37-4.50]	<.01*
	PCOS × Meno	premenopausal, PCOS control	REF	REF	REF	−0.03	6435.10
		postmenopausal, PCOS control	0.24 (0.14)	1.28 [0.97-1.68]	.08
		premenopausal, PCOS case	0.36 (0.15)	1.43 [1.06-1.92]	.02
		postmenopausal, PCOS case	0.52 (0.20)	1.68 [1.14-2.48]	<.01*
	CRP × Meno	premenopausal, low CRP	REF	REF	REF	0.29	6261.90
		premenopausal, high CRP	0.33 (0.09)	2.34 [1.83-3.00]	<.01*
		postmenopasual, low CRP	−0.33 (0.08)	1.22 [0.91-1.63]	<.01*
		postmenopausal, high CRP	0.53 (0.09)	2.85 [2.07-3.94]	<.01*

Results for RERI represent the interaction effect on an additive scale, alternatively interpreted as relative excess risk due to interaction on an OR scale modeled by measuring the effect of the cross-classification of PCOS (yes or no) levels with menopause and CRP levels on MetS using disjoint indicator variables for the exposures. A RERI above 0 indicates a superadditive interaction between the 2 exposures whereas a RERI below 0 indicates a subadditive interaction. Estimating interaction effects on an additive scale provides insight into the mechanisms underlying the statistical model in a biological context, where a multiplicative interaction may not always be at play and may be more important for implementation of results for public health (42). Bolded significant with P-value < .05.

Abbreviations: AIC, Akaike information criteria; CI, confidence interval; CRP, C-reactive protein; Meno, menopause status; MetS, metabolic syndrome; OR, odds ratio; PCOS, polycystic ovary syndrome; REF, reference; RERI, relative excess risk due to interactions.

In subsequent modeling (Table 5), we estimated the association between PCOS and MetS when considering their interactions with menopausal status and CRP. We observed that PCOS remained significantly associated with MetS odds even after adjusting for menopausal status [OR = 1.38 (95% CI = 1.11-1.76)], although the effect was weakened. We also observed that the association between PCOS and MetS remained significant after adjusting for CRP [OR = 1.34 (95% CI = 1.07-1.67)] (Fig. 2). The interactions between PCOS presence and premenopausal status and PCOS presence and high-CRP were statistically significant (Fig. 3). The interactions between PCOS status and menopause status and between PCOS status and CRP status on a multiplicative scale were not statistically significant (P_INT = .89; P_INT = .06, respectively). Based on the additive interactions seen in analyses from Table 4, we also estimated the effect of PCOS on MetS when adjusting for both CRP and menopause status, the interaction between PCOS and CRP, and the interaction between CRP and menopause status in model 6. We observed that this model had the best model fit (AIC = 6254.5). The effect of PCOS on MetS was no longer statistically significant in this model, although the effect remained directionally consistent (Table 5).

Table 5.

Effect of PCOS (signs and self-report) on MetS among females of HCHS/SOL study (unweighted n = 7316), adjusted by covariates, presented in the form of parameter estimates, SE, ORs of the effect and their 95% confidence intervals, and P-value of the association

Model/association with PCOS	Interaction	Effect variable	n used in model	Effect estimates b (SE)	OR	LLCI	ULCI	P-value	R2	Model fitting criteria
										AICa
Model 1		PCOS	6993	0.34 (0.11)	1.411	1.132	1.758	.0022	0.1523	6699.35
Model 2		PCOS	6794	0.32 (0.11)	1.378	1.106	1.717	.0043	0.1574	6433.36
		Menopause status		0.23 (0.14)	1.257	0.96	1.644	.0958
Model 3	By menopausal status	PCOS, in postmenopausal females	6794	0.28 (0.14)	1.317	0.992	1.748	.0572	0.1574	6435.10
		PCOS, in premenopausal females		0.36 (0.15)	1.427	1.064	1.915	.018
	By PCOS status	menopause status, PCOS cases		0.16 (0.22)	1.178	0.770	1.804	.4504
		menopause status, PCOS controls		0.24 (0.14)	1.277	0.971	1.678	.0801
Model 4		PCOS	6993	0.29 (0.11)	1.337	1.068	1.673	.0113	0.1820	6499.48
		Elevated CRP		0.86 (0.08)	2.366	2.034	2.753	<.01 b
Model 5	By elevated CRP	PCOS, in high-CRP females	6993	0.34 (0.16)	1.411	1.022	1.949	.0368	0.1820	6500.94
		PCOS, in low-CRP females		0.23 (0.16)	1.259	0.916	1.730	.1567
	By PCOS status	High CRP, PCOS cases		0.95 (0.21)	2.598	1.709	3.948	<.01 b
		High CRP, PCOS controls		0.84 (0.08)	2.317	1.966	2.730	<.01 b
Model 6		PCOS	6794	0.25 (0.16)	1.289	0.934	1.778	.1225	0.1853	6254.54
		Elevated CRP		0.82 (0.14)	2.282	1.743	2.986	<.01 b
		Menopause status		0.19 (0.15)	1.212	0.909	1.615	.1915
	PCOS × CRP by elevated CRP	PCOS, in high-CRP females		0.29 (0.16)	1.334	0.967	1.839	.0798
		PCOS, in low-CRP females		0.25 (0.16)	1.289	0.934	1.778	.1225
	PCOS × CRP by PCOS status	High CRP, in PCOS cases		0.87 (0.21)	2.384	1.572	3.614	<.01 b
		High CRP, in PCOS controls		0.83 (0.09)	2.304	1.938	2.739	<.01 b
	CRP × Meno by elevated CRP	postmenopausal, in high-CRP females		0.21 (0.17)	1.235	0.889	1.717	.2081
		postmenopausal, in low-CRP females		0.19 (0.15)	1.212	0.909	1.615	.1915
	CRP × Meno by menopausal status	High CRP, in postmenopause		0.86 (0.13)	2.367	1.828	3.065	<.01 b
		High CRP, in premenopause		0.84 (0.14)	2.321	1.754	3.070	<.01 b

All models were adjusted for age centered, a squared term for centered age, study site by Hispanic background, age at immigration, education, and health insurance coverage. All models were also adjusted for complex sampling design using strata, sampling weights, and clustering using primary sampling units.

Associations were tested in the following models: model 1: with PCOS as the factor; model 2: with PCOS as the exposure and menopause status as a covariate in the model; model 3: with PCOS as the exposure, menopause status as a covariate, and with a term for the interaction between PCOS and menopause status; model 4: with PCOS as the exposure and elevated CRP as a covariate; model 5: with PCOS as the exposure, elevated CRP as a covariate, and with a term for the interaction between PCOS and elevated CRP; model 6 (FINAL MODEL): with PCOS as the main exposure with elevated CRP and menopause status as covariates in the model and interaction terms for exposures with superadditive interactions from Table 4, PCOS and CRP, and CRP and menopause status.

Abbreviations: AIC, Akaike inclusion criteria; CRP, C-reactive protein; HCHS/SOL, Hispanic Community Health Study/Study of Latinos; LLCI, lower limit of 95% confidence interval; MetS, metabolic syndrome; OR, odds ratio; PCOS, polycystic ovary syndrome; ULCI, upper limit of 95% confidence interval.

^a AIC for model fit in logistic regression, for intercept and covariates, rounded off to 2 decimal points.

^b Bolded significant with P-value < .05.

Figure 2.

Effect of PCOS on MetS in main regression models. ORs with 95% confidence intervals of outcomes in main effect regression models on odds of MetS are shown in the form of a forest plot. Model 1: with PCOS as the main independent variable; model 2: with PCOS as the exposure and menopause status as a covariate; model 4: with PCOS as the exposure and CRP as a covariate in the model. All models were adjusted for age, study site by Hispanic background, age at immigration, education, and health insurance coverage. All models were also adjusted for complex sampling design using strata, sampling weights, and clustering using primary sampling units.

Abbreviations: CRP, C-reactive protein; Meno, menopause; MetS, metabolic syndrome; OR, odds ratio; PCOS, polycystic ovary syndrome.

Figure 3.

Effect of PCOS on MetS in interaction models. ORs with 95% confidence intervals of outcomes in main effect regression models on odds of MetS are shown in the form of a forest plot. Model 3 and model 5 show effect estimation of PCOS by menopause status and CRP categories, respectively. Model 6 (FINAL MODEL) odds depict the effect of PCOS on MetS by CRP categories. Model 6 measured the effect of PCOS status on MetS, adjusted by menopause status, CRP categories, term for interaction between PCOS and CRP and between CRP and menopause status. All effect estimates can be found in Table 5. All models were adjusted for age, study site by Hispanic background, age at immigration, education, and health insurance coverage. All models were also adjusted for complex sampling design using strata, sampling weights, and clustering using primary sampling units.

Abbreviations: CRP, C-reactive protein; Meno, menopause; MetS, metabolic syndrome; OR, odds ratio; PCOS, polycystic ovary syndrome.

Supplementary Table S7 (44) shows the effect of all 3 exposures—PCOS, CRP, and menopause status—on the prevalence of MetS and its subcomponents. CRP level independently predicted higher odds of MetS prevalence and all 5 subcomponents, whereas menopause status independently predicted only IFG [OR = 1.30 (95% CI = 1.02-1.65)]. Models for the association between MetS subcomponents and PCOS, adjusting for CRP and/or menopause, were all significantly associated with higher odds of MetS, IFG, and elevated TGL.

Sensitivity Analyses

Since we observed that metformin use was significantly associated with several measures including BMI, WC, HDL, LDL, TGL, FG, and HOMA-IR in both minimally adjusted and fully adjusted models [Supplementary Table S6 (44)], we conducted a sensitivity analysis to quantify the effect of PCOS adjusted for metformin use on the association between PCOS and MetS [Supplementary Table S8 (44)]. We observed that adjusting for metformin attenuated the association of PCOS with MetS and its subcomponents, although the effect of PCOS on MetS, IFG, and elevated TGL remained significant and directionally consistent for WC, BP, and HDL. Similar trends were also seen for BMI-adjusted models [Supplementary Table S9 (44)]. Since we observed a significant interaction between PCOS and CRP, we sought to evaluate whether this effect remained robust after adjusting for metformin use or BMI. To assess this, we analyzed the effect of the PCOS-CRP interaction on MetS when adjusted for metformin use (model 1), BMI (model 2), and both (model 3) [Supplementary Table S10 (44)]. We observed that the interaction between high CRP status and PCOS case status on the multiplicative scale was still robust after adjusting for BMI, metformin use, and both.

Discussion

PCOS is a complex reproductive and endocrine disorder that manifests itself in females of reproductive age and is usually diagnosed after a woman has begun menstruating and monitoring her menstrual cycle duration (45, 46). It is characterized by oligo/amenorrhea, hyperandrogenism, and polycystic ovaries. Our study observed that ∼21% of females had irregular menses and ∼11% reported infertility. Females with PCOS may experience several other comorbidities such as insulin resistance (7, 19, 47) and chronic low-grade inflammation (48-50) and are at a higher risk for cardiovascular diseases (51-53). In this study, we examined the association between PCOS and MetS in Hispanic/Latino females—a high-risk population for cardiometabolic dysfunction. This ethnic group, with its great ancestral, cultural, and genetic diversity, has been underresearched in the field of PCOS and, therefore, in this study of HCHS/SOL communities, we analyzed the relationship between PCOS and MetS along with the effect modification by CRP and menopause status, which has not been explored in previous studies in this population.

In the HCHS/SOL, we observed that PCOS was significantly associated with MetS, and our results indicated a robust relationship between PCOS and MetS across CRP and menopause status, even after the adjustment for metformin use or BMI. We additionally observed that high CRP levels, defined as ≥3.0 mg/L, was an effect modifier of the relationship between PCOS and MetS, but we did not observe evidence of a statistically significant interaction between PCOS and menopause status in relation to odds of MetS prevalence. In Hispanic/Latinas, cardiovascular risk and metabolic risk is higher compared to non-Hispanic White females, and this risk exacerbates with age and menopause (17, 54, 55). In our sample, postmenopausal females had higher baseline/referent metabolic impairment and premenopausal females had higher prevalence of PCOS, and these may have restricted our ability to fully tease apart these interactions. Higher MetS risk in all postmenopausal females compared to premenopausal women in our sample may have overpowered the detrimental effect of PCOS in postmenopausal females and thereby concealed the PCOS-menopause interaction on MetS. Future studies could leverage electronic health record data and increase the sample size among pre- and postmenopausal females and study the development of MetS in postmenopausal females separately to test the interaction effects.

Our interaction findings are consistent with previous literature that suggests a strong inflammatory component to PCOS. Studies report elevated levels of CRP (56, 57), higher levels of leukocytes (50), independent of BMI, and elevated levels of other inflammatory markers in PCOS patients as well (58). It is unclear whether PCOS alone is associated with inflammation, but some studies have found that PCOS is associated with elevated serum IL-8 concentrations (29, 59, 60) and white blood cell counts (50, 61, 62), independent of obesity. Some studies have found that PCOS is no longer significantly associated with elevated CRP levels once adjusted for BMI (49, 63). In our study, we observed that CRP was significantly correlated to PCOS, MetS, and its subcomponents. We also observed that the presence of both PCOS and high CRP interacted to exacerbate the odds of MetS, and this association remained significant after adjusting for BMI and metformin use.

Due to the aberrant immune activation linked with PCOS, some researchers have also described PCOS as an autoimmune process (48, 58, 64) and associated chronic low-grade inflammation with PCOS (30, 31, 58). The thresholds used for high-sensitivity CRP in these studies are based on the AHA/CDC scientific statement for healthcare professionals, which was developed from studies such as the Women's Health Study that are mostly comprised of non-Hispanic White females (65-67). However, in 2004, a study called for studies to generate information on the CRP cut-offs in multiethnic populations as they may vary considerably due to differences in metabolic factors and may need validation for cardiovascular disease risk prediction (68). Therefore, we recognize the limitation of our study wherein the categorization of CRP was conducted based on AHA/CDC recommended values for CRP. Despite the limitation, we found associations between CRP and MetS, and CRP and all 5 MetS components (elevated WC, FG, BP, low HDL, and elevated TGL) and a high average CRP concentration (4.7 mg/L; Table 1). Additionally, we recognize that analyzing self-report of PCOS and signs of PCOS could be prone to misclassification and this is a limitation of our study, but self-report is nevertheless a good proxy for Rotterdam criteria in lack of hormonal panels and serum concentrations (69). We also acknowledge that our study did not measure androgens, hirsutism, or the observation of polycystic ovaries upon ultrasound, and the analyzed signs of PCOS may have resulted in a heterogenous group of females with some having impaired reproductive characteristics and some without PCOS morphology. Future research should measure other aspects of PCOS (eg, androgen levels, presence of polycystic ovaries, hirsutism, and anti-Müllerian hormone levels) to fully determine the association with MetS and effect modification by CRP and menopause status.

Overall, our findings are consistent with previous research in HCHS/SOL on PCOS and MetS, and we found that PCOS was also significantly associated in Hispanic/Latina females with impaired fasting glucose and elevated TGL, which are subcomponents of MetS (18). Our current cross-sectional study replicates some of the previous findings in Hispanic/Latinas with PCOS and prior research that has shown the increased burden of PCOS and MetS in this population compared to non-Hispanic White females (70-72). Although we did not find significant effect modification by menopause status on the PCOS-MetS association, we found that metabolic dysfunction may be more prevalent in postmenopause females with PCOS, which was not explored previously. Our study in the Hispanic/Latina community provides evidence for a robust association between PCOS and MetS independent of CRP, menopause status, metformin use, and BMI and suggests a superadditive interaction between PCOS and CRP on MetS. Our study demonstrates the strength of the association between PCOS and MetS across the female lifespan and reinforces the roles of inflammation, glucose impairment, and dyslipidemia in Hispanic/Latina females living with PCOS. The positive relationship between PCOS and MetS should inform future studies on the cardiovascular risks associated with PCOS, and interventions should target inflammation, glucose regulation, and dyslipidemia in Hispanic/Latinas living with PCOS.

Funding

The Hispanic Community Health Study/Study of Latinos (HCHS/SOL) is a collaborative study supported by contracts from the National Heart, Lung, and Blood Institute (NHLBI) to the University of North Carolina (HHSN268201300001I/N01-HC-65233), University of Miami (HHSN268201300004I/N01-HC-65234), Albert Einstein College of Medicine (HHSN268201300002I/N01-HC-65235), University of Illinois at Chicago (HHSN268201300003I/N01-HC-65236 Northwestern University), and San Diego State University (HHSN268201300005I/N01-HC-65237). The following institutes/centers/offices have contributed to the HCHS/SOL through a transfer of funds to the NHLBI: National Institute on Minority Health and Health Disparities, National Institute on Deafness and Other Communication Disorders, National Institute of Dental and Craniofacial Research, National Institute of Diabetes and Digestive and Kidney Diseases, National Institute of Neurological Disorders and Stroke, National Institute of Health (NIH) Office of Dietary Supplements. H.C.R. was supported by the National Center for Advancing Translational Sciences, Grant TL1 TR002016 and Grant UL1 TR002014. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH. H.C.R. is supported by American Heart Association grant # 24PRE1193934/HRIDYA RAO/2024. The authors thank Dr. David J. Vandenbergh and Dr. Linda Ann Wray for their help in framing this study's analyses.

Disclosures

The authors have nothing to disclose.

Data Availability

Availability of data and detailed policies for accessing HCHS/SOL study data can be found online (https://sites.cscc.unc.edu/hchs/). The HCHS/SOL study data are made available through the NHLBI BioLINCC repository (https://biolincc.nhlbi.nih.gov/studies/hchssol/).