Osteosarcopenia in Reproductive-Aged Women with Polycystic Ovary Syndrome: A Multicenter Case-Control Study

Abstract

Context

Osteosarcopenia (loss of skeletal muscle and bone mass and/or function usually associated with aging) shares pathophysiological mechanisms with polycystic ovary syndrome (PCOS). However, the relationship between osteosarcopenia and PCOS remains unclear.

Objective

We evaluated skeletal muscle index% (SMI% = [appendicular muscle mass/weight (kg)] × 100) and bone mineral density (BMD) in PCOS (hyperandrogenism + oligoamenorrhea), and contrasted these musculoskeletal markers against 3 reproductive phenotypes (i): HA (hyperandrogenism + eumenorrhea) (ii); OA (normoandrogenic + oligoamenorrhea) and (iii), controls (normoandrogenic + eumenorrhea). Endocrine predictors of SMI% and BMD were evaluated across the groups.

Design, Setting, and Participants

Multicenter case-control study of 203 women (18-48 years old) in New York State.

Results

PCOS group exhibited reduced SMI% (mean [95% confidence interval (CI)]; 26.2% [25.1,27.3] vs 28.8% [27.7,29.8]), lower-extremity SMI% (57.6% [56.7,60.0] vs 62.5% [60.3,64.6]), and BMD (1.11 [1.08,1.14] vs 1.17 [1.14,1.20] g/cm²) compared to controls. PCOS group also had decreased upper (0.72 [0.70,0.74] vs 0.77 [0.75,0.79] g/cm²) and lower (1.13 [1.10,1.16] vs 1.19 [1.16,1.22] g/cm²) limb BMD compared to HA. Matsuda index was lower in PCOS vs controls and positively associated with SMI% in all groups (all Ps ≤ 0.05). Only controls showed associations between insulin-like growth factor (IGF) 1 and upper (r = 0.84) and lower (r = 0.72) limb BMD (all Ps < 0.01). Unlike in PCOS, IGF-binding protein 2 was associated with SMI% in controls (r = 0.45) and HA (r = 0.67), and with upper limb BMD (r = 0.98) in HA (all Ps < 0.05).

Conclusions

Women with PCOS exhibit early signs of osteosarcopenia when compared to controls likely attributed to disrupted insulin function. Understanding the degree of musculoskeletal deterioration in PCOS is critical for implementing targeted interventions that prevent and delay osteosarcopenia in this clinical population.

Polycystic ovary syndrome (PCOS) is a complex endocrine disorder affecting up to 18% of reproductive age women (1). Traditionally considered a condition of impaired fertility (2), it is now accepted that PCOS imparts significant and broad-spectrum health risks, including compromised cardiometabolic health, disordered sleep, depression, and anxiety, as well as a heightened risk for certain cancers (2). Some of the symptoms and health risks in PCOS are characteristics of aging (3-5), which could imply an increased likelihood of reduced musculoskeletal health (6-8). The obesity, insulin resistance (IR), sex hormone abnormalities, chronic inflammation, altered vitamin D status, and sedentary lifestyle seen in PCOS (4,9-15) are consistent with pathophysiological mechanisms that drive disorders of musculoskeletal health, such as osteosarcopenia, which can begin to manifest early in mid-life (16-19). Osteosarcopenia is the combination of sarcopenia, with either osteopenia or osteoporosis. Sarcopenia refers to the progressive and generalized loss of lean body mass (LBM) accompanied by impaired muscle function, associated with frailty and falls (18,20). Osteopenia and osteoporosis are identified by a loss of bone mineral density (BMD) and deteriorated bone microarchitecture (21), both of which contribute to frailty, falls, fractures, and increased mortality (16,18). Sarcopenia and osteopenia/osteoporosis are considered a highly “hazardous duet” (22) given the interconnectedness of the physical, mechanical, and biochemical cross-talk of the bone–muscle unit through endocrine and paracrine mechanisms (18,23–25). These concurrent alterations in bone and muscle are compensated by visceral and intramuscular fat infiltration that accelerate the deterioration in musculoskeletal composition (17,18). Understanding the degree to which musculoskeletal health is compromised in PCOS would better delineate any need for targeted interventions that effectively prevent and delay the development of osteosarcopenia in this clinical population.

Despite biological plausibility, the relationship between osteosarcopenia and PCOS remains unclear, owing, in part, to the relatively new status of osteosarcopenia as a musculoskeletal condition (20,23). McBreairty et al recently reported an increased prevalence of sarcopenic obesity, as defined by a concurrent decrease in the percentage of appendicular lean mass (ALM) and an increase in total body fat (>35%), in obese women with PCOS compared to lean controls (26). However, the observation was limited by their use of self-report, rather than clinical verification, to establish control cases, as well as the lack of accounting for body mass index (BMI) differences between the PCOS and control groups. Further, previous reports using dual-energy X-ray absorptiometry (DXA) and bioelectrical impedance analysis methods have yielded conflicting results on the impact of PCOS on the isolated body composition predictors of osteosarcopenia. Some studies have reported negative influences of PCOS on LBM (27,28) or BMD (6,7,29, 30), while others showed no such associations (LBM (31–35) or BMD (8, 33–41). To complicate matters, there is even evidence to support positive associations between PCOS and LBM (41,42) or BMD (43,44). The contradictory evidence may stem, in part, from difficulties in collecting reliable measures of body composition (45) and androgen status (46), the use of retrospective data, and failure to account for regional variations of body composition, age, and adiposity. Collectively, whether women with and without PCOS exhibit differences in their body composition and endocrine determinants of osteosarcopenia remains unknown.

To address this research gap, we evaluated LBM and BMD status as early predictors of osteosarcopenia in a well-defined cohort of women with PCOS as defined by the National Institutes of Health (NIH) criteria (hyperandrogenism and oligoamenorrhea) (47) as described previously (48). As such, we contrasted the LBM and BMD of women with PCOS against women without PCOS across 3 reproductive phenotypes (1): HA (hyperandrogenism and eumenorrhea) (2) OA (normoandrogenic oligoamenorrhea), and (3), controls (normoandrogenic and eumenorrhea). By adopting the NIH definition, we were able to investigate key pathological factors that are known to drive musculoskeletal degeneration in women per se, including altered androgen status and menstrual irregularity (6,49–52). Further, endocrine predictors of LBM and BMD were evaluated across groups. We hypothesized that reproductive-aged women with PCOS would exhibit early signs of osteosarcopenia, as evidenced by decreased LBM and BMD and that loss of LBM and BMD would be attributed to endocrine disruptions in PCOS.

Materials and Methods

Study design and setting

The present case-control study represents a secondary data analysis of 212 women of reproductive age that participated in 5 studies involving the prospective evaluation of body composition. Participants were recruited between January 2013 and July 2018 using paper and electronic advertisements at (i) the Human Metabolic Research Unit, Division of Nutritional Sciences, Cornell University (Ithaca, NY, US); (ii) Strong Fertility Center, Department of Obstetrics and Gynecology, University of Rochester Medical Center (Rochester, NY, US); and (iii) Ronald O. Perelman and Claudia Cohen Center for Reproductive Medicine, Weill Cornell Medicine (New York, NY, US).

Ethical approval

The institutional review boards at Cornell University, University of Rochester, and Weill Cornell Medicine approved the study protocols (ClinicalTrials.gov Identifiers: NCT1410015577, NCT01927432, NCT01927471, NCT01859663, and NCT01785719). All procedures were conducted in compliance with the World Medical Association Declaration of Helsinki, and the Guidelines of the International Conference on Harmonization on Good Clinical Practice. All participants provided written informed consent at study enrollment.

Study participants

Women were eligible to participate if they were 18 to 48 years old and exhibited no symptoms of the menopausal transition (ie, no recent changes in menstrual patterns or abnormal elevations in follicle-stimulating hormone [FSH]). Exclusion criteria for the present analysis were a medical history of conditions known to interfere with reproductive or metabolic function (besides PCOS) or bone and muscle metabolism, including diabetes mellitus, hyperprolactinemia, untreated thyroid dysfunction, premature ovarian failure, hyperparathyroidism, rheumatoid arthritis, renal disease, chronic obstructive pulmonary disease, and osteoporosis. None of the included participants were involved in fertility therapy, used medications that were confirmed or suspected to affect bone and muscle metabolism, including hypolipidemic agents, antihypertensive medications, or antiaggregants/anticoagulants. Of the 212 women deemed eligible to be included in the present study, 7 were ultimately excluded due to the recent (≤2 months) use of oral hormonal contraceptives (OCP) or insufficient data to determine the reproductive phenotype, resulting in a final study population of 203 women.

Definition of study groups

Four groups were included in the present study: (i) PCOS (hyperandrogenism and oligoamenorrhea), (ii) HA (hyperandrogenism and eumenorrhea), (iii) OA (normoandrogenic and oligoamenorrhea), and (iv) controls (normoandrogenic and eumenorrhea). Participants were evaluated either during the early follicular phase of the menstrual cycle (in women reporting regular menstrual cycles) or at a random time when no dominant follicles or corpora lutea were present (in women reporting irregular menstrual cycles). PCOS was diagnosed according to the NIH criteria (47) of the simultaneous presence of HA and OA. HA was defined by either (i) a modified hirsutism score ≥ 6 (2) and/or (ii) hyperandrogenemia based on internally developed thresholds (2). Hirsutism was evaluated by visual and self-reported scoring of usual hair growth on nine regions of the body (53). Biochemical evidence of hyperandrogenism was defined by a fasting serum total testosterone concentration ≥ 61.5 (ng/dL), free androgen index (FAI) ≥6 (2), free testosterone ≥0.815 (ng/dL), or bioavailable testosterone ≥19.06 (ng/dL). Thresholds reflected the 95th percentiles of androgen concentrations in an internal reference cohort. OA was defined by a self-reported average menstrual cycle length ≥36 days or <8 cycles per year (2). Menstrual cycle length was recorded as the average interval between menstrual cycles in the 12 months before study enrollment. Women who presented with only one isolated feature of PCOS were placed in the HA and OA groups, as appropriate. No women in the OA group presented with OA due to extreme athletic activities or eating disorders based on self-report or medical history. The control group consisted of women that had neither OA nor HA.

Study procedures

Clinical assessments.

A standardized health history and physical examination was completed for all women to assess demographics, anthropometry, vitals, and, endocrine, metabolic, and menstrual status of groups as described previously (54) to enable phenotyping groups and characterizing determinants of their body composition. BMI was calculated as weight in kilograms divided by height in meters squared. Waist and hips circumference measurements were used to determine the waist-to-hip ratio as described previously (55). Blood pressure was measured using standard blood pressure monitors.

Biochemical assessments.

Fasting serum concentrations of total testosterone were measured by liquid chromatography-tandem mass spectrometry at a clinical chemistry lab participating in the Centers for Disease Control and Prevention Hormone Standardization Program (Brigham Research Assay Core, Boston, MA, US), as previously described (56). Fasting serum concentrations of total insulin-like growth factor 1 (IGF-1), total insulin-like growth factor 2 (IGF-2), and insulin-like growth factor-binding protein 2 (IGFBP-2) were measured by enzyme-linked immunosorbent assay (Ansh Labs, Webster, TX, US) as nontraditional markers of glucoregulatory status (57-59). Plasma glucose was measured by commercial glucometer test strips (Accu-Chek Aviva Plus, Roche Diagnostics, Indianapolis, IN, US). Remaining analytes, including insulin, sex hormone-binding globulin (SHBG), estradiol, luteinizing hormone (LH), FSH, thyroid-stimulating hormone were measured using chemiluminescence immunoassays (Siemens Medical Solutions Diagnostics, Deerfield, IL, US).

Blood samples were collected in the early follicular phase in women with regular menstrual cycles or the absence of a dominant follicle (≥10 mm) or corpus luteum in those with irregular or absent menstrual cycles. Insulin and glucose responses to a standard 75-g oral glucose tolerance test were measured at 30, 60, 90, and 120 min. Free and bioavailable testosterone (60), FAI (61), the homeostasis model assessment of IR (HOMA-IR) (62), and the Matsuda index (63) were calculated using validated formulae, as previously described. Samples were processed for serum and stored at −80ºC until the time of analyses. All inter- and intra-assay coefficients of variation were ≤10.5%, consistent with good assay performance.

Body composition assessments.

Body composition, including fat and lean tissues, and BMD were evaluated by DXA (Discovery and Horizon QDR series, Hologic Inc., Bedford, MD, US) at the whole body and regional levels. Cross-calibration and quality control of DXA machines were performed using a phantom scan daily. Body composition data were analyzed by certified radiology technologists using a standardized protocol.

ALM was measured using the sum of the lean mass (without bone) of the upper and lower extremities, which are mainly muscle, consistent with a common approach in the study of sarcopenia (64-67). Given that many women with PCOS presented with overweight or obesity (2,4), and that lean mass is associated with body weight, we used proxies for sarcopenia that were adjusted for weight. Also, because of the lack of standardized criteria or a consensus definition, several indices of sarcopenia were used to comprehensively evaluate early signs of the condition and to identify any overlap with osteoporosis and adiposity consistent with the approaches used in previous reports (68,69), including (1), skeletal muscle index (SMI% = ALM [kg] / body weight [kg] × 100) (69,70) and (2) ALM/Ht² = ALM (kg) / body height (m)² (3,71) lower extremity skeletal muscle mass (LESM = sum of lean soft tissue of right and left lower extremity [kg]) (4); LESM index (LESMI% = LESM [kg]/lower extremity body weight [kg] × 100); and (5), LESM/Ht² = LESM (kg)/body height (m)² (69).

Total and regional BMD, including upper and lower limbs, thoracic and lumbar spine (L1-L4 vertebrae), and pelvis were measured. Upper and lower limb BMD were calculated as the sum of the BMD of the arms and legs, respectively. All coefficients of variation were ≤6.2% for the evaluated body composition measured by DXA.

Statistical analyses.

Statistical analyses were performed using SPSS version 25.0 (IBM Corp., Armonk, NY, US) and R version 3.6.1. (R Foundation for Statistical Computing, Vienna, Austria). Results were presented as mean (95% confidence interval [CI]) or frequency (percentage) for each group. Demographic, clinical, biochemical, anthropometric, and body composition data were compared between study groups by Chi-square analyses or 1-way analyses of variance after confirming the normality of residuals by visual inspection of histograms and probability plots where appropriate. Linear regression analyses were used to identify differences between anthropometric and body composition measures across study groups in crude models and also after accounting for the potential effects of age and BMI as covariates in the adjusted models. If a significant group effect was identified, post hoc pairwise comparisons were performed using Bonferroni corrections to determine where differences occurred.

Our statistical models evaluated the potential relationships between LBM and BMD with select markers of hyperandrogenism, glucoregulatory status, and female sex-hormone profile given the current pathophysiologic theories underpinning osteosarcopenia. Partial correlations were performed to identify the relationship between body composition markers of osteosarcopenia and endocrine factors across all groups after accounting for age and BMI differences between the groups, where appropriate. Multiple testing was corrected by the false discovery rate procedure to control the expected proportion of false positives. The standard R function p.adjust was used to adjust P-values for multiple testing using the Benjamini-Hochberg method (72). Results were considered significant at P ≤ 0.05.

Results

Demographic, anthropometric, clinical, and biochemical characteristics

The baseline characteristics of participants are presented in Table 1. Overall, women were young (mean age = 27.4 years) and primarily White (126/203 [62.1%]). There were age differences among the groups (P < 0.0001; Table 1). Results of Bonferroni post hoc pairwise analyses showed that women with PCOS (P < 0.01) and OA (P < 0.01) were statistically younger than controls.

Table 1.

Demographic, anthropometric, physiologic, clinical, and biochemical characteristics of women (n = 203)

Measures (unit)	PCOS (n = 55) (With HA and OA)	HA (n = 56) (With HA without OA)	OA (n = 32) (With OA without HA)	Control (n = 60) (Without HA and OA)	Reference	P-Value
Demographics
Age (year)	25.8 (24.5, 27.1) a	27.5 (26.0, 29.1)	24.6 (22.3, 26.9) a	30.3 (28.6, 32.1)	N/A	<0.0001
Race (n (%))
African American	5 (9.1)	11 (19.6)	1 (3.1)	10 (16.7)	N/A	0.16
Asian	5 (9.1)	5 (8.9)	5 (15.6)	7 (11.7)
White	41 (74.5)	32 (57.1)	22 (68.8)	31 (51.7)
Other	4 (7.3)	8 (14.3)	4 (12.5)	12 (20.0)
Current metformin use (n [yes %])	1 (1.8)	0 (0)	0 (0)	0 (0)	N/A	0.43
Anthropometrics and vitals
Weight (kg)	89.9 (81.8, 97.9) a	77.8 (72.4, 83.2)	79.5 (69.1, 90.0)	73.4 (68.2, 78.7)	N/A	<0.01
Height (m)	1.62 (1.61, 1.65)	1.65 (1.63, 1.66)	1.65 (1.62, 1.68)	1.64 (1.62, 1.66)	N/A	0.48
BMI (kg/m2)	33.3 (30.6, 36.1) a,b	28.7 (26.7, 30.6)	28.9 (26.7, 32.2)	27.2 (25.4, 29.1)	18.5-25.0	<0.01
BMI ≥ 30 kg/m2 (n [yes %])	33 (60.0) a	19 (33.9)	11 (34.4)	19 (31.7)	N/A	<0.01
WC (cm)	99.4 (92.7, 106.1) a,c	90.5 (86.0, 95.1)	86.4 (78.5, 94.4)	87.4 (83.4, 91.4)	<88	<0.01
WHR	0.86 (0.83, 0.89) c	0.84 (0.82, 0.86)	0.81 (0.78, 0.84)	0.83 (0.82, 0.85)	≤0.85	0.06
Systolic BP (mmHg)	116 (113, 120)	113 (108, 119)	112 (107, 118)	109 (105, 113)	<120	0.11
Diastolic BP (mmHg)	74 (72, 78)	72 (69, 76)	72 (68, 76)	70 (68, 73)	<80	0.27
Endocrine and metabolic
Total testosterone (ng/dL)	52.4 (45.7, 59.2) a,c	45.7 (39.4, 52.0) a	37.4 (33.9, 40.1)	32.0 (28.8, 35.2)	<61.5	<0.0001
Free testosterone (ng/dL)	0.976 (0.827, 1.125) a,b,c	0.745 (0.622, 0.868) a	0.534 (0.462, 0.607)	0.405 (0.366, 0.444)	<0.815	<0.0001
Bioavailable testosterone (ng/dL)	22.87 (19.38, 26.35) a,b,c	17.46 (14.57, 20.36) a	12.53 (10.84, 14.22)	9.50 (8.59, 10.42)	<19.06	<0.0001
Modified hirsutism scored	9 (8, 11) a	9 (8, 10) a,c	2 (2, 3) a	3 (2, 3)	<6	< 0.0001
FAI (%)	7 (5, 8) a,b,c	5 (4, 6) a	3 (2, 3)	2 (2)	<6	<0.0001
FPG (mg/dL)	96.2 (93.1, 99.3)	95.2 (92.1, 98.3)	92.1 (88.7, 95.5)	93.9 (90.9, 96.8)	<100.0	0.37
Fasting insulin (µIU/mL)	14.6 (11.2, 18.0) a	10.2 (7.6, 12.8)	9.9 (6.6, 13.3)	8.3 (6.5, 10.1)	<25	<0.01
SHBG (nmol/L)	37.3 (30.2, 44.4) a,c	45.3 (38.7, 51.9)	58.4 (41.2, 75.7)	59.70 (53.0, 66.4)	18.0-114.0	<0.0001
Total IGF-1 (ng/mL)	312.3 (255.2, 369.5)	301.0 (225.2, 376.7)	408.5 (316.2, 500.9) a	262.6 (215.7, 309.6)	N/A	0.02
Total IGF-2 (ng/mL)	532.8 (469.5, 596.1)	485.2 (432.8, 537.6)	540.4 (498.1, 582.6)	479.07 (445.9, 512.2)	N/A	0.14
IGFBP-2 (ng/mL)	22.5 (13.4, 31.6)	44.3 (22.6, 66.0)	25.9 (13.5, 38.2)	50.4 (34.6, 66.2)	N/A	0.04
HOMA-IR index e	3.6 (2.7, 4.6) a	2.5 (1.8, 3.3)	2.3 (1.5, 3.1)	2.0 (1.5, 2.4)	<3.0	0.01
Matsuda indexf	5.3 (3.4, 6.9) a	6.8 (5.2, 8.5)	7.2 (4.8, 9.6)	8.6 (6.8, 10.4)	N/A	0.05
Estradiol (pg/mL)	59.6 (47.7, 71.5)	42.1 (42.2, 56.0)	49. 5 (41.8, 57.1)	56.4 (42.9, 69.8)	N/A	0.46
LH (µIU/mL)	8.4 (7.3, 9.5) a	6.8 (5.7, 9.7) a	8.2 (5.2, 10.4) a	4.8 (4.2, 5.4)	N/A	< 0.0001
FSH (µIU/mL)	6.0 (5.6, 6.5)	7.3 (5.7, 8.9)	6.2 (5.4, 7.0)	6.7 (6.0, 7.4)	<20	0.28
LH/FSH ratio	1.46 (1.23, 1.68) a	1.11 (0.91, 1.33)	1.34 (0.96, 1.73)	0.88 (0.62, 1.14)	N/A	0.01
TSH (μIU/mL)	1.6 (1.4, 1.8)	1.7 (1.5, 2.0)	1.6 (1.3, 1.8)	1.8 (1.5, 2.1)	<5.0	0.52
Menstruation
Age at menarche (year)	12.7 (12.2, 13.2)	12.5 (12.0, 13.0)	12.6 (12.2, 13.1)	12.2 (11.8, 12.6)	N/A	0.40
Menstrual cycle length (day)	110.7 (80.7, 140.6) a,b	29.8 (28.9, 30.6) c	90.0 (61.0, 119.1) a	28.5 (27.7, 29.4)	21-35 day	<0.0001

Data are expressed as mean (95% confidence interval) and number (percentage). One-way analysis of variance and chi-squared tests were used for comparisons of means and proportions between groups, respectively.

Abbreviations: PCOS, polycystic ovary syndrome; HA, hyperandrogenism; OA, oligo-amenorrhea; BMI, body mass index; WC, waist circumference; WHR, waist-to-hip ratio; BP, blood pressure; FAI, free androgen index; FPG, fasting plasma glucose; SHBG; sex hormone-binding globulin; IGF; insulin-like growth factor; IGFBP, insulin-like growth factor-binding protein; HOMA-IR, homeostatic model assessment of insulin resistance; LH, luteinizing hormone; FSH, follicle-stimulating hormone; TSH, thyroid-stimulating hormone.

^aSignificantly different from the control group using the Bonferroni adjustment (P < 0.05).

^bSignificantly different from the HA group using the Bonferroni adjustment (P < 0.05).

^cSignificantly different from the OA group using the Bonferroni adjustment (P < 0.05).

^dDetermined using the Ferriman–Gallwey Index (53).

^eCalculated using the equation (fasting insulin (µIU/mL) × fasting glucose (mg/dL)/405 (normalizing factor) (62).

^fCalculated using the equation: 10 000/√(fasting glucose × fasting insulin) (mean glucose × mean insulin) (63).

Anthropometric characteristics, including BMI, waist circumference (WC), and waist-to-hip ratio, were different across groups (P < 0.01; Table 1). The mean BMI of our cohort was 29.5 kg/m². Forty percent (82/203) of all women had obesity (BMI ≥ 30 kg/m²). Results of post hoc comparisons showed an increased BMI and WC in the PCOS group compared to controls (all Ps < 0.01). Also, the PCOS group had an increased BMI (P = 0.02) compared to the HA group, and higher WC (P = 0.02) and waist-to-hip ratio (P = 0.05) than the OA group.

Groups differed in their hyperandrogenism indices by design (P < 0.0001) (Table 1). Women with PCOS and HA had increased total, free, and bioavailable testosterone concentrations, FAI levels, and hirsutism scores compared to controls (all Ps ≤ 0.01) as identified by post hoc analyses. Women with PCOS also had increased total, free, and bioavailable testosterone concentrations, and FAI levels than the OA group (all Ps ≤ 0.01). The HA group had a higher hirsutism score than the OA group (P < 0.01).

Groups also showed different glucoregulatory status, as identified by fasting insulin, SHBG, total IGF-1, IGFBP-2, HOMA-IR, and Matsuda index levels (P ≤ 0.05) (Table 1). The PCOS group had elevated fasting insulin, HOMA-IR, and lower SHBG and Matsuda index levels compared to controls (P ≤ 0.04). Further, the OA group exhibited elevated IGF-1 compared to controls (P = 0.01). Women with PCOS had a tendency toward decreased IGFBP-2 concentrations compared to controls (mean [95% CI]; 22.5 [13.4, 31.6] vs 50.4 [34.6, 66.2] ng/mL; P = 0.06) (Table 1).

Differences were observed between groups in the LH/FSH ratio (P = 0.01) (Table 1). Women with PCOS had increased LH/FSH ratio compared to controls (P < 0.01) revealed by post hoc analyses. The average length of menstrual cycles was different among groups by design (P < 0.01; Table 1). The PCOS (P < 0.01) and OA groups (P < 0.01) had longer intermenstrual intervals than controls. Also, the length of intermenstrual intervals was longer in the PCOS group when compared to the HA group (P < 0.01).

Body composition characteristics

Body composition measures across the study groups are presented in Supplemental Table 1 (crude models; available in a digital data repository (73)) and Table 2 (adjusted models). Adjusted analyses showed SMI% and LESMI% were different across the groups (all Ps = 0.01) (Table 2). Results of post hoc comparisons showed lower SMI% and LESMI% in the PCOS group compared to controls (all Ps < 0.01).

Table 2.

Body composition characteristics of women (n = 203)

Measures (unit)	PCOS (n = 55) (With HA and OA)	HA (n = 56) (With HA Without OA)	OA (n = 32) (With OA Without HA)	Control (n = 60) (Without HA and OA)	P-Value (Adjusted)
Fat
Total fat mass (kg)	29.6 (28.0, 31.3)	28.5 (27.0, 30.0)	28.2 (26.0, 30.2)	29.1 (27.5, 30.6)	0.67
Total fat (%)	34.9 (33.4, 36.4)	34.4 (33.0, 35.8)	34.7 (32.8, 36.6)	35.2 (33.8, 36.6)	0.95
Trunk fat mass (kg)	14.1 (13.2, 15.1)	13.2 (12.3, 14.1)	13.0 (12.7, 14.2)	14.1 (13.2, 15.1)	0.40
Total lean mass (kg)	47.1 (45.4, 48.9)	48.4 (46.8, 50.0)	46.6 (44.4, 48.8)	46.5 (44.8, 48.1)	0.33
Lean
Lean to fat mass ratio	1.95 (1.73, 2.18)	2.00 (1.80, 2.21)	2.13 (1.86, 2.41)	1.97 (1.76, 2.18)	0.74
ALM (kg)	20.8 (19.9, 21.7)	21.7 (20.9, 22.6)	20.9 (19.8, 22.1)	21.0 (20.1, 21.8)	0.41
SMI (%)	26.2 (25.1, 27.3) a	28.2 (27.1, 29.2)	27.7 (26.2, 29.1)	28.8 (27.7, 29.8)	0.01
ALM/Ht2 (kg/m2)	7.9 (7.6, 8.1)	8.0 (7.7, 8.2)	7.6 (7.3, 8.0)	7.8 (7.5, 8.0)	0.31
LESM (kg)	16.3 (15.6, 17.0)	17.2 (16.5, 17.9)	16.6 (15.6, 17.5)	16.7 (15.9, 17.5)	0.37
LESMI (%)	57.6 (56.7, 60.0) a	60.7 (58.7, 62.6)	60.0 (57.4, 62.6)	62.5 (60.3, 64.6)	0.01
LESMI/Ht2 (kg/m2)	21.7 (20.7, 22.6)	22.4 (21.5, 23.3)	22.1 (20.8, 23.3)	23.3 (23.3, 24.3)	0.15
Bone
Total BMD (g/cm2)	1.11 (1.08, 1.14) a, b	1.16 (1.14, 1.19)	1.11 (1.07, 1.15)	1.17 (1.14, 1.20)	<0.01
Upper limb BMD (g/cm2)	0.72 (0.70, 0.74) b	0.77 (0.75, 0.79)	0.73 (0.71, 0.76)	0.73 (0.71, 0.76)	<0.01
Lower limb BMD (g/cm2)	1.13 (1.10, 1.16) b	1.19 (1.16, 1.22)	1.16 (1.12, 1.19)	1.15 (1.12, 1.18)	0.04
Thoracic spine BMD (g/cm2)	0.85 (0.82, 0.89)	0.87 (0.84, 0.90)	0.82 (0.78, 0.87)	0.84 (0.80, 0.88)	0.30
Lumbar spine BMD (g/cm2)	1.10 (1.04, 1.12)	1.12 (1.07, 1.16)	1.07 (1.00, 1.13)	1.11 (1.06, 1.17)	0.65
Pelvis BMD (g/cm2)	1.27 (1.22, 1.32)	1.28 (1.24, 1.33)	1.26 (1.19, 1.32)	1.23 (1.18, 1.27)	0.37

Data are expressed as mean (95% confidence interval). Regression analyses were used for comparisons. Adjusted models included age for ALM, LESMI%, and LESMI/Ht² and age and BMI as covariates for all other variables.

Abbreviations: PCOS, polycystic ovary syndrome; HA, hyperandrogenism; OA, oligo-amenorrhea; BMI, body mass index; WC, waist circumference; WHR, waist to hip ratio; ALM, appendicular lean mass; SMI, skeletal muscle mass; Ht, height; LESM, lower extremity skeletal muscle mass; LESMI, lower extremity skeletal muscle mass index; BMD, bone mineral density.

^aSignificantly different from the control group using the Bonferroni adjustment (P < 0.05).

^bSignificantly different from the HA group using the Bonferroni adjustment (P < 0.05).

Total, upper, and lower limb BMD were different across the groups (P < 0.01) (Table 2). Women with PCOS had lower total BMD compared to the HA (P = 0.03) and control (P = 0.02) groups. The OA group exhibited a tendency toward decreased total BMD than controls (mean [95% CI]; 1.11 [1.07, 1.15] vs 1.17 [1.14, 1.20] (g/cm²); P = 0.07) (Table 2). The PCOS group exhibited reduced upper BMD than the HA group (All: P < 0.01) and had a tendency toward decreased upper limb BMD when compared to controls (P = 0.06). Similarly, the PCOS group had decreased lower limb BMD than the HA group (P = 0.03), as evidenced by post hoc comparisons.

Of all (203) women, 165 (81.3%) were never users of OCP or had not used OCP in the 12 months before enrollment; 23/203 (11.3%) stopped using OCP 3 to 10 months before enrollment, whereas data on OCP cessation were not available for the remaining 15/203 (7.4%). The distribution of these data was comparable across the groups (P = 0.18). Results of our sensitivity analyses on the subset (165/203 [81.3%]) of women that were never users or had stopped use in ≥12 months before enrollment showed similar trends in BMD and LBM across the experimental groups when compared to total sample (all Ps ≤ 0.02; data not shown). Similarly, 201/203 (99.0%) of all women were never users of metformin or had ceased use ≥12 months before enrollment; only 1 (1/203 [0.5%]) was a current metformin user (Table 1), and another had stopped metformin use in preceding 12 months. Both metformin users were in the PCOS group, but metformin use was not different across groups when current and recent users were considered (P = 0.14). Results of sensitivity analysis by excluding the 2 metformin users showed that the observed trends in the BMD and LBM across the groups remained unchanged when compared to the total sample (all Ps < 0.01; data not shown).

Associations between LBM and BMD and endocrine measures

The partial correlations between LBM and BMD and endocrine factors across all 4 groups are shown in Table 3. Correlations are adjusted for (i) age for SMI% and LESMI% and (ii) age and BMI for total, upper, and lower limb BMD as covariates. Concerning androgens, there were inverse associations between FAI and LESMI% only in the OA group after adjusting for age and obesity (r = −0.55; P < 0.05).

Table 3.

Associations between endocrine factors and bone and muscle indices across the study groups (n = 203)

Measures (unit)	FAI (%)	Fasting insulin (µIU/mL)	Matsuda index	Total IGF-1 (ng/mL)	Total IGF-2 (ng/mL)	IGFBP-2 (ng/mL)	Estradiol (pg/mL)	LH/FSH ratio
PCOS (n = 55) (With HA and OA)
SMI (%)	−0.35	−0.55a	0.47 c	0.50	−0.06	0.09	−0.31	−0.19
LESMI (%)	−0.32	−0.45c	0.46 c	0.34	−0.29	0.13	−0.37	−0.17
Total BMD (g/cm2)	−0.22	−0.23	0.14	−0.11	−0.35	−0.14	−0.03	−0.04
Upper Limb BMD (g/cm2)	−0.22	−0.06	−0.20	0.48	−0.22	−0.05	0.19	0.11
Lower Limb BMD (g/cm2)	−0.10	−0.07	−0.08	0.58	−0.12	−0.43	−0.06	−0.03
HA (n = 56) (With HA without OA)
SMI (%)	−0.32	−0.36	0.60 b	−0.12	0.27	0.67 c	−0.02	−0.18
LESMI (%)	−0.17	−0.02	0.42	−0.26	−0.27	0.44	−0.12	−0.14
Total BMD (g/cm2)	−0.12	−0.07	−0.04	0.07	−0.59	0.03	−0.09	0.03
Upper Limb BMD (g/cm2)	0.22	0.13	0.01	−0.09	0.20	0.98 a	−0.21	0.15
Lower Limb BMD (g/cm2)	−0.05	−0.04	0.21	−0.32	0.52	0.37	−0.17	−0.08
OA (n = 32) (With OA without HA)
SMI (%)	−0.39	−0.29	0.65 c	−0.36	0.53	0.60	−0.37	−0.05
LESMI (%)	−0.55c	−0.22	0.74 b	−0.90b	0.40	0.91 b	−0.49	−0.16
Total BMD (g/cm2)	−0.44	0.47	0.19	0.18	0.24	0.27	0.38	0.20
Upper Limb BMD (g/cm2)	−0.27	0.20	0.17	0.15	0.47	0.16	0.64 b	0.54 c
Lower Limb BMD (g/cm2)	−0.39	0.41	0.28	−0.11	0.42	0.14	0.22	0.14
Control (n = 60) (Without HA and OA)
SMI (%)	−0.26	−0.55b	0.45 c	0.21	0.33	0.45 c	−0.19	0.22
LESMI (%)	−0.37	−0.44	0.26	−0.08	−0.10	0.50	−0.17	0.25
Total BMD (g/cm2)	−0.34	−0.01	0.02	0.06	−0.02	0.18	0.08	−0.04
Upper Limb BMD (g/cm2)	−0.08	0.13	0.24	0.84 a	0.21	0.40	0.09	0.07
Lower Limb BMD (g/cm2)	−0.13	−0.23	0.17	0.72 b	0.02	0.21	0.06	0.01

Data represent correlation coefficients. Partial correlation analyses were used for comparisons. Adjusted models included (1): age for SMI% and LESMI%; and (2), age and BMI for total, upper and lower limb BMD as covariates.

Abbreviations: FAI, free androgen index; IGF; insulin-like growth factor; IGFBP, insulin-like growth factor-binding protein; LH, luteinizing hormone; FSH, follicle-stimulating hormone; PCOS, polycystic ovary syndrome; HA, hyperandrogenism; OA, oligo-amenorrhea; SMI, skeletal muscle mass; LESMI, lower extremity skeletal muscle mass index; BMD, bone mineral density.

^a P < 0.0001 adjusted with false discovery rate corrections using the Benjamini-Hochberg method (72).

^b P < 0.01 adjusted with false discovery rate corrections using the Benjamini-Hochberg method (72).

^c P < 0.05 adjusted with false discovery rate corrections using the Benjamini-Hochberg method (72).

Analyses of markers of glucoregulatory status revealed negative associations between fasting insulin concentrations and SMI% in both the PCOS (r = −0.55) and control groups (r = −0.55) and negative associations between insulin concentrations and LESMI% in the PCOS group (r = −0.45; all P < 0.05) after adjusting for age. The Matsuda index was positively associated with SMI% in all groups after adjusting for age (all Ps ≤ 0.05) (Table 3). Further, evaluating the nontraditional markers of glucoregulatory status showed negative associations between IGF-1 and LESMI% (r = −0.90; P < 0.01) only in the OA group after adjusting for age. Only controls showed positive associations between IGF-1 and upper (r = 0.84) and lower (r = 0.72) limb BMD after adjusting for BMI and age (all Ps < 0.01). Unlike the PCOS group, IGFBP-2 was positively associated with muscle or bone mass in other groups. Specifically, IGFBP-2 was associated with SMI% in control (r = 0.45) and HA (r = 0.67) groups and with LESMI% in the OA group (r = 0.91) after adjusting for age; further, IGFBP-2 was positively associated with upper limb BMD (r = 0.98) in the HA group after adjusting for age and obesity (all Ps < 0.05).

Analysis of female reproductive hormones showed that estradiol concentrations (r = 0.64) and LH/FSH ratio (r = 0.54) were positively associated with upper limb BMD only in the OA group (all Ps < 0.05) (Table 3).

Discussion

For the first time, we explored the relationship between LBM and BMD with endocrine aberrations in well-defined clinical cohorts using a multicenter design. The major finding of the current work was that women with PCOS exhibited early signs of osteosarcopenia, as evidenced by decreased total LBM and BMD compared to controls after adjustment for age and obesity. By contrast, our data showed that women who presented with isolated features of PCOS, including HA or OA alone, did not exhibit decreased LBM and BMD when compared to controls. We examined women with PCOS as defined by the NIH criteria, which enabled us to study pathogenic mechanisms of osteosarcopenia, including contributions by OA and/or HA (6,49-52), in a relatively homogeneous group. We acknowledge that some women in the OA and HA groups may exhibit milder variants of PCOS. However, our data do not support the development of osteosarcopenia in these milder phenotypes. Our data suggest that perturbations in insulin function may drive muscle and bone loss in PCOS. Our observations support and extend new evidence about PCOS as a risk factor of osteosarcopenia (26,74), albeit the actual development and clinical consequences of osteosarcopenia, including functional decline and fracture risk (7) in the long term remain largely unknown and need to be confirmed by longitudinal research. Understanding the biological mechanisms and risk factors of osteosarcopenia may yield management strategies that delay or prevent the condition, thereby improving the musculoskeletal health and associated long-term comorbidities in reproductive-aged women with endocrine and menstrual disruptions, as evidenced in PCOS.

Our observations of lower BMD (6,9,29,30,74,75) corroborate previous reports in women with PCOS. A recent systematic review and meta-analysis of 21 studies, with a total of 31 383 women with PCOS and 102 797 controls, showed the adverse effects of PCOS on select segmental BMD regions (spine and femur) (7). Our data coincide with previous reports of decreased total BMD in women with PCOS compared to controls and extend findings to support decreased upper and lower limb BMD in women with PCOS compared to isolated HA. There are few data on the impact of PCOS on upper and lower limb BMD. We are aware of only 1 study by Good et al that showed elevated upper limb and comparable total BMD in lean (BMI = 22.4 kg/m²) women with PCOS compared to controls where the authors attributed the findings to a protective effect of hyperandrogenism (38). However, we did not observe a propensity for increased upper limb BMD or comparable total BMD even in our lean subcohort of women with PCOS (data not shown). The discrepancy between studies may be explained by the milder metabolic phenotype studied by Good et al (38).. Our observation of lower total and upper and lower limb BMD in PCOS compared to the HA group may imply a protective effect of isolated hyperandrogenism on BMD in the absence of OA (76,77). Regular menstrual cyclicity is known to have protective effects on BMD through pre-established mechanisms (78,79). Accordingly, we observed a tendency toward lower BMD and positive associations between estradiol concentrations with total BMD in the OA group. Together, a combination of OA and HA, as seen in PCOS, seems to potentiate BMD loss to the degree that exceeds any potential protective impact of HA on bone health.

Data on LBM in PCOS are sparse. Our findings corroborate with 3 previous reports (27,28,74) about a negative impact of PCOS on LBM but contradict findings by Mario et al (8). and Carmina et al (41). wherein increased LBM was noted in women with PCOS compared to controls. The discrepancies in findings may be attributed to approaches used to account for adiposity. In contrast to Mario et al, we accounted for the impact of excessive body weight on LBM using conservative statistics. Further, Carmina et al (41). reported elevated LBM in a milder metabolic variant of PCOS compared to our study, as evidenced by BMI status (27.5 vs 33.3 kg/m²). Collectively, evidence on LBM in women with PCOS remains mixed, and further research is warranted to clarify the influence of PCOS on lean mass.

The biological mechanisms pertaining to the loss of LBM and BMD in reproductive-aged women with PCOS are not fully elucidated. A complex interaction between genetic and environmental determinants has been proposed to underlie the endocrine and paracrine derangements that exacerbate the loss of LBM and BMD in women with PCOS (7,80,81). Previous reports have proposed aberrations in insulin signaling and function, growth hormone, gonadotropin-releasing hormone, androgens, and estrogens, coupled with pro-inflammation, oxidative stress, and low vitamin D status as adversely affecting bone and muscle health (36,51,74,80-85). Our data support insulin as a key player in the derangement of skeletal muscle mass. We observed decreased insulin sensitivity and elevated insulin concentrations in women with PCOS consistent with the endocrine milieu of this clinical population (4,86). Also, we noted negative correlations between fasting insulin levels and skeletal muscle in both women with PCOS and controls and positive correlations between insulin sensitivity and skeletal muscle mass in all women. Our observations in controls are consistent with previous reports about the anabolic effects of insulin on increasing the rate of muscle protein synthesis and decreasing muscle protein degradation (81,87-89) at physiologic concentrations. However, IR and compensatory hyperinsulinemia may be drivers of muscle protein degradation (80,90) and/or bone resorption (81,91) in women with PCOS that compromise musculoskeletal health (51,81,92), albeit we did not note associations between glucoregulatory markers with total or limb BMD. We acknowledge specific questions remain about how and why IR and subsequent hyperinsulinemia specifically aggravate muscle protein degradation and/or bone resorption that remain to be addressed in future research. Importantly, a complementary mechanism may be the intrinsic programming of PCOS mesenchymal stem cells toward adipocyte enlargement and adipocyte formation (93) that may occur synchronously alongside altered proliferation and differentiation of myogenic (94,95) and/or osteogenic (96) cell lines. Further, the intrinsic insulin signaling dysfunction established in PCOS (86,97) is likely a contributor to the bone and/or muscle loss, consistent with emerging reports in non-PCOS studies (98-100).

IGF-1 and IGF-2, have been established as key regulators of muscle and bone development, homeostasis, and metabolism (51,101,102) and were, therefore, evaluated in the present work. IGF-1 and IGF-2 bind IGFBP-2 (51,80). IGFBP-2 regulates the half-life of circulating IGFs and their availability and action in both stimulatory and inhibitory pathways (102-109). Associations between IGFs and IGFBP-2 with LBM and BMD remain mixed. Evidence from human studies has supported an anabolic effect of circulating IGF-1/IGF-2 on LBM and/or BMD across health status, skeletal regions, age, or sex categories (51,85,110-113) or has shown no association (114-119). Likewise, evidence from human and animal models has either supported (114,120,121) or opposed (106,107,122,123) a catabolic effect of circulating IGFBP-2 on LBM and/or BMD, or has shown no association (113,118,124). Our data do not support an anabolic effect of IGF-1 or catabolic effect of IGFBP-2 on LBM and BMD per se in all reproductive-aged women. Rather, our data support differential effects for these factors on muscle and bone in women across the HA and OA spectrum, implying a context-specific role (125) for these factors. With respect to muscle, IGF-1 levels were increased (1.5-fold) in the OA group when compared to controls and were negatively correlated with skeletal muscle in lower extremities (LESMI%). There is limited evidence to support an anabolic role for IGF-1 beyond physiologic concentrations (126). Instead, our observations likely indicate (i) muscle resistance to the anabolic action of IGF-1 (secondary to blocked access of IGF-1 receptor or declined sensitivity at the receptor level) or (ii) switching to catabolic action (secondary to excessive circulating levels and blocked local secretion of IGF-1) 126-129). The latter mechanism is supported by emerging evidence that has linked the anabolic effects of IGF-1 to local muscle secretion rather than circulating levels per se (128,130-132). The findings related to IGFBP-2 are harder to reconcile. Although levels were not statistically different across groups, our observations of a positive association of IGFBP-2 with skeletal muscle (SMI%) in the HA group and controls do not support a negative influence. While our data are consistent with an anabolic role for IGF-1 on BMD (115,133), a positive association of IGFBP-2 with upper limb BMD in controls and HA groups contradict a catabolic effect (110,114). The role of IGFBP-2 remains poorly identified in women, and more research is needed to elucidate any gender-specific effects on LBM and BMD (107,121). Further, evidence from knock-out mice supports a compensatory/potentiating role of the other 5 IGFBPs to balance the action of any single IGFBP in a finely tuned manner (105). Therefore, we acknowledge that the interrogation of other IGFBP members would have allowed us to better elucidate these relationships. This is especially true in the case of PCOS, where the true effects of the IGF axis remain largely undefined (91,134).

Our observations should be interpreted in light of limitations inherent to the case-control study design that does not allow causal inference. We used a prospective and comprehensive approach to obtain endocrine and body composition data across our recruiting centers and excluded women who used medications known to influence body composition. However, future research is required to account for other factors, including vitamin D status, inflammation, and lifestyle behaviors (eg, smoking, diet, and physical activity), which are known to independently impact bone and muscle health. Further, biochemical markers of bone turnover (eg, osteocalcin) and functional tests are needed to fully characterize osteosarcopenia. We controlled for the impact of BMI in our analyses using conservative statistical models. Obesity alone may accompany osteosarcopenia (17,18); therefore, we do not preclude the implications of obesity on decreased BMD and LBM despite our statistical adjustments. We also acknowledge that not matching our groups for obesity and insulin sensitivity was a major limitation. Having BMI- and insulin sensitivity–matched groups would have best accounted for the potential impacts of mechanical loading (135,136) and insulin signaling (137-139) on the effects of androgens and anovulation on BMD and LBM across groups. The inclusion of ovulatory normoandrogenic obese-insulin resistant women in future studies will help to clarify these associations. We used DXA as a gold standard method for measuring BMD. DXA is also a common tool for characterizing body composition (84,140); however, given its technical limitations for estimating body composition, we adopted protocols (141) that minimized biases of body size, adiposity, and hydration status (84,142). Further, the mean age of women with PCOS in the present study was less than controls (25.8 vs 30.3 years, respectively), and we adjusted for age in our analyses. Women were previously considered to achieve their peak bone mass between 25 and 30 years of age and skeletal muscle mass at 25 years of age (7,17,18). However, emerging evidence has shown that both BMD and LBM plateau in women after the second decade of life (143-145), which supports groups being at comparable stages. We acknowledge the accepted influence of race/ethnicity on the peak and distribution of BMD (146-150) and LBM (151-154), particularly increased BMD and LBM in African American women. We admit that the proportions of participants across racial/ethnic groups in this study were diverse—albeit not statistically different (P = 0.16) (Table 1). Results of a subanalysis after excluding African American women were consistent with decreased BMD and LBM in women with PCOS vs controls. Further studies are needed to fully address any musculoskeletal health disparity in women with PCOS as our study was limited by its inclusion of a predominantly White population.

Conclusions

Women with PCOS exhibited early signs of osteosarcopenia, as manifested by reduced BMD and LBM compared to controls. Perturbations in insulin function may drive muscle and bone loss in PCOS. Findings from the present work create a paradigm for researchers and allied healthcare providers to consider the potentially deleterious impacts of PCOS on musculoskeletal health and the relevance of adopting management strategies to prevent and delay the development of osteosarcopenia in this clinical population.

Additional Information

Disclosure Summary: The authors have nothing to disclose.

Data Availability: The data sets generated during and/or analyzed during the current study are not publicly available but are available from the corresponding author on reasonable request.