Objective
We conducted a systematic review and meta-analysis to summarize and quantitatively pool evidence on cardiometabolic health disparities between Black and White women with polycystic ovary syndrome in the United States in response to the call for further delineation of these disparities in the international evidence-based guideline for the assessment and management of polycystic ovary syndrome.
Data Sources
Databases of MEDLINE, Web of Science, and Scopus were searched initially through March 05, 2020, and confirmed on September 11, 2020.
Study Eligibility Criteria
Observational studies documenting cardiometabolic risk profile (glucoregulatory, lipid profile, anthropometric, and blood pressure status) in Black and White women with polycystic ovary syndrome were included. Studies on children (<17 years old) and pregnant or menopausal-aged women (>50 years) were excluded. The primary outcome was fasting glucose. Furthermore, data on major cardiovascular events (stroke, coronary heart disease, heart failure) and mortality rate (cardiovascular death, total mortality) were evaluated.
Methods
Data were pooled by random-effects models and expressed as mean differences and 95% confidence intervals. Studies were weighted based on the inverse of the variance. Heterogeneity was evaluated by Cochran Q and I 2 statistics. Study methodologic quality was assessed by the Newcastle-Ottawa scale.
Results
A total of 11 studies (N=2851 [652 Black and 2199 White]) evaluated cardiometabolic risk profile and all had high quality (Newcastle-Ottawa scale score of ≥8). No studies reported on cardiovascular events and mortality rate. Black women had comparable fasting glucose (−0.61 [−1.69 to 2.92] mg/dL; I 2 =62.5%), yet exhibited increased fasting insulin (6.76 [4.97–8.56] μIU/mL; I 2 =59.0%); homeostatic model assessment of insulin resistance (1.47 [0.86–2.08]; I 2 =83.2%); systolic blood pressure (3.32 [0.34–6.30] mm Hg; I 2 =52.0%); and decreased triglyceride (−32.56 [−54.69 to −10.42] mg/dL; I 2 =68.0%) compared with White women (all, P ≤.03). Groups exhibited comparable total cholesterol, low-density lipoprotein cholesterol, high-density lipoprotein cholesterol, and diastolic blood pressure (all, P ≥.06).
Conclusions
Black women with polycystic ovary syndrome have a greater tendency for an adverse cardiometabolic risk profile (increased insulin, homeostatic model assessment of insulin resistance, and systolic blood pressure) despite lower triglycerides than White women. Our observations support the consideration of these disparities for diagnostic, monitoring, and management practices in Black women and for future guideline recommendations. Given the heterogeneity among studies, future research should address the relative contributions of biologic, environmental, socioeconomic, and healthcare factors to the observed disparities. Furthermore, longitudinal research is required to address patient-pressing complications, including cardiovascular events and mortality rate in Black women with polycystic ovary syndrome as a high-risk yet understudied population.
Introduction
Women with polycystic ovary syndrome (PCOS) exhibit an adverse cardiometabolic risk profile compared with their counterparts without PCOS. Poor glucoregulatory status, characterized by abnormal glucose and insulin levels secondary to insulin resistance (IR), is the key pathophysiological factor associated with a wide range of metabolic disruptions in PCOS. , , Although the true risk of cardiovascular mortality remains unclear in this clinical population, the theoretical risk is substantial and warrants consideration of treatment and prevention strategies given both short- and long-term implications of cardiometabolic aberrations on pregnancy outcomes, lifelong health, and wellness.
Why was this study conducted?
We aimed to evaluate cardiometabolic health disparities between Black and White women with polycystic ovary syndrome (PCOS) in the United States.
Key findings
Black women with PCOS have a greater tendency for adverse cardiometabolic risk profiles, evidenced by increased insulin, insulin resistance, and systolic blood pressure, despite lower triglyceride levels than White women.
What does this add to what is known?
This is the first systematic review to show a disproportionate cardiometabolic risk burden among Black women with PCOS in the United States. Our findings support the need to address health disparities in current clinical practice, future guideline development, and longitudinal research evaluating cardiovascular events and mortality rate in Black women with PCOS as a high-risk understudied population.
The degree to which cardiometabolic aberrations may disproportionately impact diverse populations of women with PCOS is important yet understudied. We and others have shown differences in cardiovascular risk factors, metabolic syndrome (MetS) rate, prediabetes, and type 2 diabetes status across diverse populations of women with PCOS in the United States and worldwide, supporting the potential for cardiometabolic health disparities in Indigenous, South Asian, Middle Eastern, and Black women compared with White or Caucasian counterparts. The recent international evidence-based guideline for the assessment and management of PCOS emphasizes the need to further delineate these disparities to inform targeted diagnosis, monitoring, and management of high-risk clinical populations and subsequent adaptations of the guideline recommendations across geographic regions and clinical settings. Although the guideline recommends that health professionals consider presentation and manifestations of PCOS across diverse populations, our knowledge about the presence and magnitude of cardiometabolic health disparities is far from complete and remains to be elucidated.
In the United States, our knowledge about health disparities in the context of PCOS is limited to a few cross-sectional analyses that have yielded inconsistent results. Black or African Americans are a large underrepresented group in the United States. According to the Census Bureau (2019) data, approximately 44 million people in the United States are Black alone, representing 13.4% of the total population. However, Black women with PCOS are understudied, and their cardiometabolic characteristics have yielded conflicting results compared with their White counterparts. Some studies reported a worse cardiometabolic profile in Black compared with White women with PCOS, as evidenced by increased fasting glucose, increased fasting insulin, , , , , , elevated systolic blood pressure (SBP) , , and diastolic blood pressure (DBP), , decreased high-density lipoprotein cholesterol (HDL-C), , and increased MetS rate. , Conversely, others showed no differences between these groups in fasting glucose, , , , , , fasting insulin, , , SBP, , DBP, , total cholesterol (TC), , , low-density lipoprotein cholesterol (LDL-C), , , HDL-C, , , waist circumference (WC), , , and waist to hip ratio. , To complicate matters, there is even evidence to support that Black women with PCOS have a more favorable cardiometabolic profile than White women, as assessed by decreased TC, , increased HDL-C, , decreased LDL-C, and triglyceride (TG). , , , , ,
The inconsistency may stem, in part, from the small sample sizes of individual studies, which have limited their statistical power to capture health disparities in cardiometabolic outcomes. Difficulties in establishing accurate PCOS status owing to reliance on self-reported diagnosis and unreliable measures of androgen status and ovarian morphology may have also contributed. The use of retrospective data and failure to account for variations in age, adiposity, lifestyle, and socioeconomic status across studies may have led to inconsistent findings. Furthermore, data regarding cardiovascular events (eg, nonfatal stroke, myocardial infarction, and heart failure), cardiovascular mortality, or total mortality are sparse in Black women with PCOS. Collectively, whether Black and White women with PCOS exhibit cardiometabolic risk differences remains unknown yet is critical for sustainable and successful management strategies that set the foundation for achieving health equity in this clinical population.
To address this knowledge gap, we conducted a systematic review and meta-analysis to characterize cardiometabolic health disparities between Black and White women with PCOS in the United States. Our primary aim was to assess whether Black women with PCOS exhibited worse glucoregulatory status, as assessed by increased fasting glucose, than White women. As our secondary aim, we evaluated whether other cardiometabolic risk factors (glucoregulatory [insulin, IR], anthropometric, lipid profile, and blood pressure status), major cardiovascular events (stroke, coronary heart disease, and heart failure), and the mortality rate (cardiovascular death and total mortality) were exacerbated in Black women with PCOS than their White counterparts. Our focus on these cohorts is relevant because data from the general population have shown a substantial and persistent risk for cardiovascular disease and type 2 diabetes morbidity and or mortality in Black vs White cohorts, , which could manifest early in reproductive-aged women with PCOS with underlying metabolic complications, warranting early preventative and management modalities. We hypothesized that Black women with PCOS would exhibit increased glucoregulatory dysregulation (increased glucose, IR, and hyperinsulinemia) and an overall worse cardiometabolic risk compared with White women with PCOS.
Methods
The work presented herein was performed according to the Preferred Reporting Items for Systematic Reviews and Meta-analyses guidelines. The study protocol is available at the International Prospective Register of Systematic Reviews (registration identification: CRD42020183485).
Review question (Population, Exposure, Outcome)
The Population, Exposure, and Outcome (PEO) criteria were defined before the literature search and are detailed in Supplemental Table 1 . Concisely, our study question was, in women with PCOS in the United States (P), do women who self-identify as Black or African American compared with White or Caucasian (E) exhibit worse cardiometabolic risk factors/outcomes (O)?
Primary and secondary outcomes
Our primary outcome was fasting glucose as a marker of glucoregulatory status. Glucose levels, unlike insulin levels, are routinely measured in practice and are most commonly used to establish cardiometabolic abnormalities. Increased glucose levels, secondary to exhaustion of pancreatic β cell function, may represent a more robust predictor for the secondary cardiometabolic abnormalities in PCOS , evaluated in this study. Secondary outcomes included (1) other cardiometabolic risk profiles (fasting insulin, the homeostatic model of IR [HOMA-IR], lipid profile [TC, LDL-C, HDL-C, TG, WC], blood pressure [SBP, DBP]) and (2) major cardiovascular events (stroke, coronary heart disease, heart failure) and mortality (death from any cardiovascular cause, total mortality) as defined with greater details in Supplemental Table 1 .
Data sources and search strategy
A systematic search of published literature was conducted in electronic databases of MEDLINE, Institute for Scientific Information Web of Science, and Scopus from January 1, 1990, to March 5, 2020, and updated on June 28, 2020, using a search strategy based on the PEO framework ( Supplemental Table 1 ). Details of the search strategy are available ( Supplemental Table 2 ). Furthermore, manual searches of reference lists from included studies supplemented the electronic database searches. Non-English studies, studies on animals, or published before 1990 were excluded. The last adjustment was chosen to allow the inclusion of studies where a PCOS diagnosis complied with the 1990 National Institutes of Health (NIH), 2003 Rotterdam, or 2006 Androgen Excess and PCOS Society (AE-PCOS) criteria. The search was continuously updated to identify the newest relevant studies until September 11, 2020.
Inclusion and exclusion criteria
Studies were included if they met the PEO criteria described in Supplemental Table 1 . Briefly, observational (cross-sectional, case-control, cohort) studies or cross-sectional analyses of baseline measures from randomized controlled trials on women with PCOS in the United States were included wherein the influence of population (Black and White subtypes) as an exposure variable was evaluated on our study outcomes of interest ( Supplemental Table 1 ). Self-identification as Black or White, as reported in individual studies, was used. , , Nonpeer-reviewed studies; studies without the design of interest; studies wherein our outcomes of interest were not compared between Black and White women with PCOS; studies that were not conducted on women with PCOS in the United States; studies wherein the diagnosis of PCOS did not comply with the NIH, Rotterdam, or AE-PCOS criteria; studies in children (<17 years old), pregnant women, or menopausal-aged women (>50 years old); and studies where the study data were irretrievable after contacting their corresponding authors were excluded ( Supplemental Table 1 ).
The screening processes were completed by 2 investigators (M.K. and J.Y.K.) independently using the double-blind coding assignment function of the online Covidence systematic review platform ( Covidence.org , Alfred Health, Melbourne, Australia) and EndNote version X9.2 (Thomson Reuters, Carlsbad, CA). All discrepancies were resolved by consensus with a third investigator (M.E.L.).
Data extraction
The following data were extracted using a standardized protocol: (1) first author’s name; (2) study publication year; (3) participants’ characteristics, including total sample size and the sample size of participants in the Black and White groups; (4) study design and setting and type of data analysis and collection (prospective or retrospective); (5) participants’ age; (6) participants’ body mass index (BMI); and (7) reported cardiometabolic outcomes of interest.
Additional information, including the presence of other underlying conditions (besides PCOS), tobacco use, and medication use, including metformin and hormonal contraceptive use, were recorded where data were available. In the case of any missing or unclear data, 2 attempts were made to contact the corresponding author by email to request data or clarify methods. Data extraction was completed by 2 investigators independently (J.Y.K. and M.K.) and was reviewed by all other authors (M.E.L., S.A.P., and R.A.) for any potential extraction error.
Assessment of study quality
We used the Newcastle-Ottawa scale (NOS) to evaluate the quality of all included studies in the present work as previously described. Notably, 2 investigators (J.Y.K. and M.K.) completed the assessments independently. All discrepancies were resolved by consensus or discussion with a third investigator (M.E.L.).
Data synthesis and analysis
Effect sizes for each outcome measure were expressed as the weighted mean difference (MD) and 95% confidence interval (CI) between the cases (Black) and control (White) groups, and the studies were weighted based on the inverse of the variance for the evaluated measure with a random-effects model. Mean differences and standard deviations (SDs) of cardiometabolic outcome measures were collected to estimate pooled effects across the evaluated measures. Where standard errors of the mean (SEMs) were only reported in the cases and control groups, SDs were calculated using a formula SD=SEM×square root (n), where n is the number of participants. , When medians and interquartile ranges were reported instead of means and SDs, we used the median in place of mean and formula of SD=(third quartile-first quartile)/1.35 to calculate SD. ,
The chi-square test was used to evaluate heterogeneity, and the Cochran Q and I 2 statistics were reported. The I 2 value describing the percentage variation between studies was calculated as 100%×([Q–df])/Q, Q being the χ 2 value and df corresponding to degrees of freedom. Low, moderate, and high heterogeneities were defined according to the cutoffs of 25%, 50%, and 75%, respectively, using the results of the I 2 test values. Tau-square was estimated using the restricted maximum likelihood (REML) method and used to evaluate between-study variance.
Sources of heterogeneity were explored using meta-regression analysis and sensitivity analyses. We performed the random-effects meta-regression by the REML method based on a priori hypotheses given the biologic plausibility to explore any influence of (1) participants’ mean age (continuous), (2) mean BMI (continuous), (3) PCOS diagnostic criteria (NIH, Rotterdam), and (4) ethnicity (non-Hispanic only, mixed Hispanic, and non-Hispanic) parameters where sufficient data were available (to allow ≥2 studies per subgroup in the categorical variables). Our meta-regression for ethnicity was performed to account for the small number of studies , that did not delineate their Black and White cohorts exclusively as non-Hispanic. Notably, none of the included studies used AE-PCOS criteria to define PCOS. Each parameter was analyzed in a univariate meta-regression to determine their relationship with the cardiometabolic outcome variable. Furthermore, sensitivity analyses were performed by removing 1 study each time and recalculating the effect size to detect the influence of each study on the overall effect size and determine whether an individual study exerted undue influence. Publication bias was assessed by visual inspection of funnel plots and formal testing by Begg’s rank correlation tests and Egger’s regression asymmetry. , M.K. performed all analyses with an experienced statistician (S.A.P.) using R version 3.6.1. (R Foundation for Statistical Computing, Vienna, Austria). Results were considered significant at P ≤.05.
Results
Literature search
The systematic search resulted in 3327 records ( Figure 1 ), of which 11 studies , , , , , , were eligible and included to evaluate cardiometabolic risk profile between Black and White women; however, we identified no studies that evaluated cardiovascular events or mortality rates. Selected studies comprised a total of 2851 participants, and 22 experimental arms (n=652 in the Black or African American [11 arms] and 2199 in the White or Caucasian [11 arms]) were included in the systematic review to evaluate cardiometabolic risk profile.
Notably, 10 of the 11 studies , , , , , provided sufficient data for pooling meta-analyses and were included in the qualitative assessment, albeit not all of these studies reported on every measure, as specified in the meta-analyses sections below.
Of the total of 11 studies included in our work, 1 study lacked sufficient data for pooling meta-analyses, and another did not provide sufficient data for the quantitative assessment of a portion of their data (lipid profile); thereby, their data were included only in the qualitative assessment. The reasons for excluding studies at each stage of the literature screening are reported ( Figure 1 ). The main reasons for excluding the studies at the first and second stages of data screening were unrelated topics and irrelevant study designs for the current work, respectively.
Study characteristics
General characteristics of included studies are presented in the Table . Studies were published between 2006 and 2020 and were all conducted in the United States by design , , , , , , and conducted in medical centers. The mean age and BMI of women ranged from 20 to 41 years and 28.2 to 41.0 kg/m 2 , respectively, across the case and control arms. Of a total of 11 studies, 9 , , , , included only non-Hispanic Black and White women, and the remaining 2 , included Black and White groups who had both Hispanic and non-Hispanic ethnicities. Most studies (7 of 11, 63.6%) , , , used the NIH, whereas 4 , , , used the Rotterdam criteria. In addition, most studies (8 of 11, 72.7%) , had prospective, whereas 3 , , had retrospective data collection. Of a total of 11 studies, 2 (18.2%) had a cohort design, , 2 (18.2%) had a case-control design, , and 8 (72.7%) had a cross-sectional design or included a cross-sectional analysis of baseline data from a randomized controlled trial. , , , ,
| Author, y (reference) | Participants’ n (total and across racial groups) | Study design, setting, data analysis | PCOS definition | Age (y) | BMI (kg/m 2 ) | Reported outcomes of interest | Total quality score |
|---|---|---|---|---|---|---|---|
| Chang et al, 2016 | Total, 94 | Cohort study | Rotterdam | NH Black, 41 (37–42) a | NH Black, 32.3 (27.6–38.0) a | ↔fasting glucose, ↑fasting insulin, ↔HOMA-IR, ↓TC, ↔LDL-C, ↔HDL-C, ↔TG, ↑SBP, ↔DBP | 8 |
| NH Black, 62 | Academic medical center | NH White, 41 (37–43) a | NH White, 28.2 (23.4–32.6) a | ||||
| NH White, 32 | Prospective | ||||||
| Hillman et al, 2014 | Total, 311 | Cohort study | Rotterdam | NH Black, 20–34 | NR | ↑fasting glucose, ↓HDL-C, ↓TG, ↑(either SBP/DBP) | 8 |
| NH Black, 67 | Academic medical center | NH White, 20–34 | NR | ||||
| NH White, 244 | Retrospective | ||||||
| Ladson et al, 2011 | Total, 120 | Case-control study | NIH | Black, 27.9 (5.0) | Black, 39.0 (9.3) | ↔fasting glucose, ↑fasting insulin, ↑HOMA-IR, ↔TC, ↑HDL-C, ↔LDL-C, ↓TG, ↔SBP, ↔DBP | 8 |
| Black, 43 | Academic medical centers | White, 26.0 (6.9) | White, 37.7 (6.3) | ||||
| White, 77 | Prospective | ||||||
| Welt et al, 2006 | Total, 216 | Case-control study | NIH | NH African American, 28.4 (6.7) | NH African American, 36.3 (7.9) | ↔dasting glucose, ↑fasting insulin, ↑HOMA-IR, ↔TC, ↔LDL-C, ↔HDL-C, ↔TG, ↑WC, ↔SBP, ↔DBP | 8 |
| NH African American, 44 | Medical centers | NH Caucasian, 28.8 (5.5) | NH Caucasian, 30.7 (9.2) | ||||
| NH Caucasian, 172 | Prospective | ||||||
| Chan et al, 2017 | Total, 284 | Cross-sectional study | Rotterdam | NH Black, 29 (25.5–34) a | NH Black, 37.5 (9.4) a | ↔fasting glucose, ↓HDL-C, ↓TG, ↑SBP, ↑DBP | 8 |
| NH Black, 100 | Medical centers | NH White, 29 (25–32) a | NH White, 30.6 (8.1) a | ||||
| NH White, 184 | Retrospective | ||||||
| Ehrmann et al, 2005 | Total, 354 | Cross-sectional study | NIH | NH Black, 27.6 (0.8) b | NH Black, 37.1 (1.2) b | ↔fasting glucose, ↑fasting insulin, ↑HOMA-IR, ↔WC | 9 |
| NH Black, 51 | Academic medical center | NH White, 28.8 (0.3) b | NH White, 36.3 (0.5) b | ||||
| NH White, 303 | Prospective | ||||||
| Engmann et al, 2017 | Total, 574 | Cross-sectional study | Rotterdam | NH Black, 28.7 (4.9) | NH Black, 35.7 (7.9) | ↔fasting glucose, ↔fasting insulin, ↔HOMA-IR, ↔LDL-C, ↔HDL-C, ↓TG, ↔WC, ↔SBP, ↔DBP | 9 |
| NH Black, 98 | Medical centers | (modified) | NH White, 28.8 (4.2) | NH White, 35.1 (9.8) | |||
| NH White, 476 | Prospective | ||||||
| Ezeh et al, 2020 | Total, 181 | Cross-sectional study | NIH | NH African American, 31.1 (1.5) b | NH African American, 34.7 (1.7) b | ↔fasting glucose, ↑fasting insulin, ↑HOMA-IR | 9 |
| NH African American, 24 | Academic medical center | NH White, 26.4 (0.5) b | NH White, 29.0 (0.7) b | ||||
| NH White, 161 | Prospective | ||||||
| Koval et al, 2010 | Total, 126 | Cross-sectional study | NIH | NH African American, 30.6 (7.6) | NH African American, 41.0 (9.6) | NR fasting glucose, NR fasting insulin, ↑HOMA-IR, ↔TC, ↑HDL-C, ↓TG, NR SBP, NR DBP, | 8 |
| NH African American, 32 | Academic medical center | NH Caucasian, 30.5 (6.8) | NH Caucasian, 37.0 (7.1) | ||||
| NH Caucasian, 94 | Retrospective | ||||||
| Legro et al, 2006 | Total, 544 | Cross-sectional analysis | NIH | African American, 27.9 (4.3) | African American, 36.0 (8.4) | ↔fasting glucose, ↔fasting insulin, ↔HOMA-IR, NR WC, NR SBP, NR DBP | 9 |
| African American, 109 Caucasian, 435 | of baseline data of RCT | Caucasian, 28.2 (3.9) | Caucasian, 35.4 (8.8) | ||||
| Academic medical centers | |||||||
| Prospective | |||||||
| Sam et al, 2015 | Total, 43 | Cross-sectional study | NIH | Non-Hispanic Black, 28 (6) | Non-Hispanic Black, 36.8 (10.6) | ↔fasting glucose, ↔fasting insulin, ↔HOMA-IR, ↔LDL-C, ↔HDL-C, ↔TG, ↔SBP, ↔DBP | 9 |
| NH Black, 22 | Academic medical center | Non-Hispanic White, 29 (7) | Non-Hispanic White, 33.8 (10.9) | ||||
| NH White, 21 | Prospective |
a Values are presented as median and interquartile range
b Variations are presented as the standard error of the mean.
Quality assessments
The quality assessment of each study based on the NOS score is presented in the Table . All studies were considered high quality based on the overall NOS score of ≥8.
Systematic review
One study that was ineligible for pooled meta-analyses is included in the qualitative review herein. Hillman et al characterized differences in individual components and the overall prevalence of MetS between 67 Black and 244 White women (age range, 20–34 years) with PCOS. We were unable to conduct meta-analyses on the components of MetS that were relevant to the purpose of the current work (glucose, HDL-C, TG, SBP, and DBP) because only percentages of women presented with increased components of MetS were reported vs mean and SD. However, authors reported an increased overall MetS prevalence (40.0% vs 22.6%; P <.01) in all but 1 component (TG) of MetS in Black women compared with White women with PCOS. Black women exhibited a higher prevalence of increased fasting glucose (≥100 mg/dL; 18.8% vs 4.9%), decreased HDL-C (≤50 mg/dL; 76.6% vs 35.6%), increased blood pressure (≥130 and or ≥85 mm Hg for SBP and DBP, respectively; 45.5% vs 31.9%), and obesity (BMI of ≥30 kg/m 2 ; 72.7% vs 51.7%; all, P <.05), where BMI was used as a proxy for increased WC. Conversely, Black women had a lower prevalence of increased TG (≥150 mg/dL; 10.9% vs 24.6%) than White women. After adjusting for age and BMI differences between the groups, the prevalence of increased fasting glucose, decreased HDL-C, and increased TG persisted (all, P <.001). In a study by Sam et al, the mean and SD of LDL-C, HDL-C, and TG were not reported in their experimental arms; therefore, the data were insufficient for pooling results. However, the investigators reported comparable lipid profiles between groups without P values being provided.
Meta-analyses
Glucoregulatory status
Glucose
Black women had comparable fasting glucose compared with White women (MD, −0.61; 95% CI, −1.69 to 2.92 mg/dL; P =.56) ( Figure 2 A), using pooled data from 10 eligible studies. , , High heterogeneity was observed among the studies ( I 2 =62.4%; P =.004).
Results of univariate meta-regression analysis are presented in Supplemental Table 3 . Age, BMI, PCOS criteria, or participant’s ethnicity did not explain the heterogeneity ( Supplemental Table 3 ).
None of the individual studies influenced the overall effect sizes, as evidenced by sensitivity analyses. We observed no evidence of publication bias (funnel plot, Supplemental Figure 1 A) (Begg’s test, P =.53; Egger’s test, P =.70).
Insulin
Black women had increased fasting insulin compared with White women (MD, 6.76; 95% CI, 4.97–8.56 μIU/mL; P <.0001) ( Figure 2 B), as reflected by pooled data from 9 eligible studies. , , However, high heterogeneity was observed among the studies ( I 2 =59.0%; P =.01).
Meta-regression analyses based on age, BMI, PCOS criteria, or ethnicity did not explain heterogeneity ( Supplemental Table 3 ).
Similarly, none of the individual studies influenced the overall effect sizes, as evidenced by sensitivity analyses. We observed no evidence of publication bias (funnel plot, Supplemental Figure 1 B) (Begg’s test, P =.21; Egger’s test, P =.08).
Homeostatic model assessment of insulin resistance
Pooling data from 9 eligible studies , , , showed increased HOMA-IR in Black women compared with White women (MD, 1.47; 95% CI, 0.86−2.08 mg/dL; P =.0005) ( Figure 2 C). High heterogeneity was evident among the studies ( I 2 =83.2%; P <.001).
Meta-regression analyses based on age, BMI, PCOS criteria, or ethnicity did not explain heterogeneity ( Supplemental Table 3 ).
None of the individual studies influenced the overall effect sizes, as evidenced by sensitivity analyses. We observed evidence of publication bias (funnel plot, Supplemental Figure 1 C) (Begg’s test, P =.04; Egger’s test, P =.32).
Lipid profile
Total cholesterol
Black women had comparable fasting TC compared with White women (MD, −8.39; 95% CI, −33.46 to 16.68 mg/dL; P =.36) ( Figure 3 A) as evidenced by pooled data from 4 eligible studies. , , , High heterogeneity was observed among the studies ( I 2 =70.0%; P =.02).
Meta-regression analyses showed age and BMI did not explain heterogeneity ( Supplemental Table 3 ). We were unable to conduct meta-regression analyses based on PCOS criteria and ethnicity because only 1 study had subsets diagnosed with the Rotterdam criteria, and another had data for women with mixed Hispanic and non-Hispanic ethnicities.
None of the individual studies influenced the overall effect sizes, as evidenced by sensitivity analyses. We observed no evidence of publication bias (funnel plot, Supplemental Figure 2 A) (Begg’s test, P =.17; Egger’s test, P =.64).
Low-density lipoprotein cholesterol
Black and White women had comparable LDL-C (MD, −2.22; 95% CI, −6.42 to −10.85 mg/dL; P =.47) ( Figure 3 B), as evidenced by pooled data of 4 eligible studies. , , , The studies were homogenous ( I 2 =5.0%; P =.37).
The results of our a priori meta-regression analyses revealed no influence of age, BMI, and PCOS criteria on the overall effect estimate ( Supplemental Table 3 ). We were unable to conduct meta-regression analyses based on ethnicity because all but a single study included only non-Hispanic women in their White and Black cohorts.
None of the studies influenced the overall effect sizes, as evidenced by sensitivity analyses. We observed no evidence of publication bias (funnel plot, Supplemental Figure 2 B) (Begg’s test, P =.50; Egger’s test, P =.42).
High-density lipoprotein cholesterol
Black and White women had comparable HDL-C (MD, 1.80; 95% CI, −8.09 to 11.69 mg/dL; P =.66) ( Figure 3 C) using pooled data from 6 eligible studies. , , , , , High heterogeneity was observed among the studies ( I 2 =96.0%; P <.001).
Meta-regressions based on age, BMI, and PCOS criteria did not explain heterogeneity ( Supplemental Table 3 ). Meta-regressions based on ethnicity were not possible because all, but a single study, included only non-Hispanic women in their White and Black cohorts.
Sensitivity analyses showed excluding the study by Ladson et al (MD, −1.56; 95% CI, −8.46 to 5.34 mg/dL; P =.56) from the analyses altered the direction of the overall effect estimate, albeit results remained comparable groups. We observed no evidence of publication bias (funnel plot, Supplemental Figure 2 C) (Begg’s test, P =.57; Egger’s test, P =.78).
Triglyceride
Pooled data from 6 eligible studies , , , , , showed decreased fasting TG concentrations in Black women compared with White women (MD, −32.56; 95% CI, −54.69 to −10.42 mg/dL; P =.01) ( Figure 3 D). The studies were highly heterogeneous ( I 2 =68.0%, P =.008).
Meta-regression analyses showed that unlike age and PCOS criteria, BMI explained the heterogeneity ( Supplemental Table 3 ). BMI was associated with decreased TG ( P =.03) and reduced the tau 2 value from 262.19 to zero; the observation indicates that BMI likely explains all between-study variance observed on pooled analysis. Meta-regression analyses based on ethnicity were not possible because all, but a single study, included only non-Hispanic women in their White and Black cohorts.
The overall pooled effect estimate was not sensitive to any of the included studies. Furthermore, we observed no evidence of publication bias (funnel plot, Supplemental Figure 2 D) (Begg’s test, P =.85; Egger’s test, P =.95).
Anthropometrics
Waist circumference
Pooled data from 4 studies , , , showed comparable WC between groups (MD, −2.56; 95% CI, −4.92 to 10.04 cm; P =.36) ( Figure 4 ). High heterogeneity was observed among the studies ( I 2 =70.0%; P =.02).
