Polycystic Ovaries Are Common in Women withHyperandrogenic Chronic Anovulation but Do NotPredict Metabolic or Reproductive Phenotype

Richard S. Legro, Percy Chiu, Allen R. Kunselman, Christina M. Bentley, William C. Dodson, andAndrea Dunaif

Departments of Obstetrics and Gynecology (R.S.L., P.C., W.C.D.) and Health Evaluation Sciences (A.R.K., C.M.B.),Pennsylvania State University College of Medicine, Hershey, Pennsylvania 17033; and Division of Endocrinology,Metabolism, and Molecular Medicine (A.D.), Feinberg School of Medicine, Northwestern University, Chicago, Illinois60611-3008

Polycystic ovary syndrome (PCOS) is a heterogeneous disor-der of unexplainedhyperandrogenic chronic anovulation. Ex-perts have recommended including the morphology and vol-ume of the ovary in the diagnostic criteria for PCOS. Weperformed this study to determine whether there was an as-sociation between themorphology and size of the ovaries andmarkers of insulin sensitivity as determined by dynamic test-ing within women with PCOS or compared with a group ofcontrol women. We then examined reproductive parameters.We studied 88 unrelated PCOS women and 21 control women,aged 1745 yr. All were in the early follicular phase or itsequivalent (no follicle with > 10 mm diameter and anovula-tory serumprogesterone level< 3 ng/ml). Subjects underwenton the same day a phlebotomy for baseline hormones, a 2-horal glucose tolerance test, and transvaginal ultrasound todetermine the morphology and volume of the ovaries. Ninety-five percent (84 of 88) of women with PCOS and 48% (10 of 21)of the control womenhad polycystic ovaries using the criteriaof at least one ovary greater than 10 cm3 (PCOV) and/or poly-cystic ovary morphology (PCOM) using the criteria of 10 or

more peripheral follicular cysts 8 mm in diameter or less inoneplanealongwith increased central ovarian stroma.PCOMwas a better discriminator than PCOV between PCOS andcontrol women. The odds of women with PCOS having PCOMwere elevated 50-fold compared with controls (odds ratio, 50;95% confidence interval, 10240;P< 0.0001), whereas the oddsofPCOVwereelevated5-fold inwomenwithPCOS (odds ratio,4.6; 95% confidence interval, 1.712.6; P 0.003). Neither theinsulin sensitivity index, fasting or 2-h values, or any inte-grated measures of glucose and insulin varied in women ac-cording to either morphology or volume, nor was there anassociation with circulating androgen levels. Women withPCOS and PCOM had lower FSH levels than women withPCOS and non-PCOM. Women with PCOS and PCOV had ahigher LH to FSH ratio thanwomenwithout PCOVandPCOS.These data support the hypothesis that polycystic ovaries arean abnormal finding. However, neither the morphology northe volume of the ovaries is associated with distinctive met-abolic or reproductive phenotypes in women with PCOS.(J Clin Endocrinol Metab 90: 25712579, 2005)

POLYCYSTIC OVARY SYNDROME (PCOS) was first di-agnosed by the presence of enlarged ovaries by pelvicexam combined with a history of amenorrhea and hirsutism(1). A characteristic abnormalmorphologywas subsequentlyconfirmed by examining histologic sections of the ovary (1).The syndrome was considered to have only reproductiveconsequences (2) until the discovery that insulin resistancewas frequently found in women with PCOS (3, 4). This find-ing led to the recognition that PCOS was a multisystemendocrinopathy characterized by hyperandrogenism andchronic anovulation (57). Nevertheless, debate continuesabout the significance of polycystic ovary morphology, inboth apparently normal women and women with the endo-crinopathy of PCOS (8, 9).

Indeed, a recent jointly sponsored American Society ofReproductiveMedicine/European Society of Human Repro-duction and Endocrinology consensus conference on diag-nostic criteria and sequelae of PCOS included the polycysticovary as one of the key features of the syndrome (10, 11). Asubsequent publication defined a polycystic ovary on bothmorphometric and volume criteria (12). However, the diag-nostic criteria for PCOS, in the absence of evidence-basedcriteria, such as the predictive value for current or long-termpathology, rest primarily on expert opinion. Ultimately suchcriteria should be substantiated by data in the same way, forexample, that the cutoff in blood glucose values for diag-nosing diabetes is based on ongoing scientific scrutiny ofexisting data (13).Whereas many of the womenwith polycystic ovaries have

stigmata of the syndrome, including hyperandrogenism,chronic anovulation, and insulin resistance, some affectedwomen appear to be endocrinologically normal (14). How-ever, this latter group of women often have subtle repro-ductive endocrine abnormalities that may become apparentonly after provocative testing (15). Furthermore, it has beenproposed that inwomen inwhom the syndrome is diagnosedbased on endocrine criteria (unexplained hyperandrogenic

First Published Online February 15, 2005Abbreviations: ANCOVA, Analysis of covariance; BMI, body mass

index; CI, confidence interval; DHEAS, dehydroepiandrosterone sulfate;ISI, insulin sensitivity index; OGTT, oral glucose tolerance test; PCOM,polycystic ovary morphology; PCOS, polycystic ovary syndrome;PCOV, polycystic ovary volume; T, testosterone; uT, non-SHBG-boundtestosterone.JCEM is published monthly by The Endocrine Society (http://www.endo-society.org), the foremost professional society serving the en-docrine community.

0021-972X/05/$15.00/0 The Journal of Clinical Endocrinology & Metabolism 90(5):25712579Printed in U.S.A. Copyright 2005 by The Endocrine Society

doi: 10.1210/jc.2004-0219

2571

chronic anovulation), the presence of polycystic ovaries is amarker for insulin resistance (16). Moreover, in women witholigomenorrhea, a positive correlation between ovarian vol-ume and insulin sensitivity has been reported (17).In contrast, other studies have not found an association

between polycystic ovaries and cardiovascular risk factors,such as circulating insulin and lipid levels in women withPCOS (18). Thus, there are conflicting data about the signif-icance of polycystic ovary morphology in women with theendocrine syndrome of PCOS. We performed this study todetermine whether there was an association between themorphology or size of the ovaries and parameters of insulinsensitivity or reproductive hormone levels in women withPCOS.

Patients and MethodsPatients

We studied 88 unrelated women with PCOS and 21 control womenaged 1745 yr. All studies were approved by the Institutional ReviewBoard of the Pennsylvania State University College of Medicine (PennState), and all subjects gave written informed consent before study. Allwomen were in good health, euthyroid, and, for at least 1 month beforeeach study, were not taking any medication (except for oral contracep-tive agents, which were stopped for 3 months before study) known toaffect sex hormone or carbohydrate metabolism. The diagnosis of PCOSwas made by the presence of chronic anovulation as defined by six orfewermenses per year in association with elevated circulating androgenlevels (19). Nonclassical 21-hydroxylase deficiency, hyperprolactinemia,and androgen-secreting tumors were excluded by appropriate tests be-fore the diagnosis of PCOS was made (5, 20).

Twenty-one reproductively normal, control women with 27- to 35-dmenstrual cycles were studied as a reference population. These womendid not have a history of hypertension or diabetes mellitus, personallyor in a first-degree relative. All control women were examined by oneof the study investigators and had no hirsutism (Ferriman-Gallweyscore 8) (21). Control women had normal glucose tolerance based ona 75-g 2-h oral glucose tolerance test applying World Health Organi-zation (WHO) criteria (22).

Study protocol

All studies were performed between 0800 and 1100 h after a 3-d 300-gcarbohydrate diet and an overnight fast of 1012 h. Studies were per-formed in PCOS women without regard to the last episode of vaginalbleeding, and all subjects were anovulatory or in the follicular phase atthe visit with a serum progesterone level less than 3 ng/ml. No womanhad diagnosed diabetes mellitus before study. Height and weight wereobtained on all subjects and body mass index (BMI) was calculated bydividing weight by height squared (kilograms per square meter). An ivcatheter was inserted, the vein was kept open with an infusion of 0.9%normal saline at 30 cc/h, an oral glucose tolerance test (OGTT) wasperformed, and blood was sampled through the catheter.

All subjects underwent a 75-g oral glucose load and blood was ob-tained for glucose determinations at 0, 30, 60, 90, and 120min.Additionalblood was obtained at time 0 for analysis of testosterone (T), non-SHBG-bound T (uT), dehydroepiandrosterone sulfate (DHEAS), and gonado-tropins. Glucose tolerance was assessed by the WHO criteria (22). Nor-mal glucose tolerance is defined as a 2-h glucose stimulated value lessthan 140 mg/dl (7.71 mmol/liter), impaired glucose tolerance is definedas a 2-h value from 140 to 199 mg/dl (7.7110.96 mmol/liter), and type2 diabetes mellitus is defined as a 2-h value of 200 mg/dl or greater(11.02 mmol/liter). We used the insulin sensitivity index (ISI0,120)based on the fasting (0 min) and 120-min OGTT insulin and glucoseconcentrations as developed by Gutt et al. (23). The ISI0,120 has beenhighly correlated with the rate of whole-body glucose disposal duringthe euglycemic insulin clamp as originally developed by DeFronzo et al.(24).

Upon completion of the OGTT, a transvaginal ultrasound was per-

formed of the pelvis. The following measures were obtained: endome-trial thickness, ovarian size in three dimensions, the size of the largestovarian follicle, and ovarian morphology. Polycystic ovary morphology(PCOM or non-PCOM) was determined by the criteria of Adams et al.(25), which includes the presence of 10 ormore peripheral follicular cysts8 mm or less in diameter in one plane along with increased centralovarian stroma. We did not use the criteria of Balen et al. (12) of 12 ormore follicles because we commenced this study before the publicationof these guidelines. No subject studied had a developing follicle (definedas largest follicle with mean diameter 10 mm) or an ovarian cyst.

Ovarian sizewas obtained bymeasuring the largest plane of the ovaryin two dimensions and then turning the vaginal probe 90 degrees andobtaining a third measurement. Volume of the ovary was calculatedusing the formula for an ellipsoid [length height width (/6)](26). In all subjects both ovaries were visualized on ultrasound, allowingfor the calculation of total ovarian volume (right left ovarian volume).Polycystic ovary volume (PCOV or non-PCOV) was determined by thecriteria of Balen et al. (12), which is defined as at least one ovary havinga volume greater than 10 cm3 with no cysts or follicles greater than 10mmmean diameter. Endometrial thickness was determined as the larg-est anterior-posterior measurement of the endometrium in the sagittalplane.

Assays

Assays for total T, DHEAS, and gonadotropinswere performed usingDiagnostic Products Corp. (Los Angeles, CA) Coat-A-Count kits (19).Free and weakly bound T was measured using a modification of theprocedure of Tremblay and Dube (27). Plasma glucose levels were de-termined by the glucose oxidase technique (28). Insulin was determinedwith a double-antibody method using reagents obtained from LincoResearch, Inc. (St. Charles, MO) (29). All assays had intra- and interassaycoefficients of variation less than 10%.

Data analysis

For continuous variables, comparisons between groups (PCOM vs.non-PCOM, PCOV vs. non-PCOV, and PCOS vs. control) were assessedusing analysis of covariance (ANCOVA)models adjusting for the effectsof age andBMI. Logarithmic transformations of the continuous variableswere taken to meet ANCOVA modeling assumptions. The data arereported as the model-based mean estimate, adjusted for age and BMIand the 95% confidence interval (CI). The association between totalovarian volume and reproductive and metabolic parameters was as-sessed after adjusting for age and BMI using the Spearman partialcorrelation coefficient within PCOS and control women separately.All analyses were performed using either SAS statistical software(SAS Institute Inc., Cary, NC) or S-Plus software (Insightful Corp., Se-attle, WA).

Results

Baseline characteristics of the two groups are found inTable 1. BMI, 2-h postchallenge glucose, and insulin levels aswell as integrated OGTT glucose and insulin responses weresignificantly higher in PCOS than the control women. TheISI0,120 was significantly higher in control women thanwomen with PCOS (Fig. 1). By design, all of the controlwomen had normal glucose tolerance. Based on WHO cri-teria applied to 2-h postchallenge glucose levels, 25 womenwith PCOS and PCOM had impaired glucose tolerance andfourwomenhad type 2 diabetesmellitus, comparedwith fiveand two women, respectively, with normal ovaries. PCOSwomen had a higher total ovarian volume than the controlwomen. For all hormonal parameters (sex steroids and go-nadotropins), with the exception of FSH, PCOS women hadsignificantly higher levels than the control women.Ninety-five percent of women (84 of 88) with PCOS and

48% control women (10 of 21) had polycystic ovaries using

2572 J Clin Endocrinol Metab, May 2005, 90(5):25712579 Legro et al. PCO Morphology in PCOS

the criteria of at least one ovary greater than 10 cm3 and/orPCOM. By morphology alone the majority of women withPCOS (74 of 88 or 84%) had polycystic ovaries. Of the 21control women, the vast majority (19 of 21 or 90%) hadnormal-appearing ovaries by morphology alone on ultra-sound. Only two control women had polycystic ovaries onultrasound by morphology. Thus, the odds of women withPCOS having PCOM on ultrasound were elevated 50-fold,

compared with controls (odds ratio 50; 95% CI 10240; P 0.0001). PCOV was present in 81% of our group of womenwith PCOS (71 of 88) but was more common in the controlwomen (48%) than PCOM. Despite this, the odds of havinga PCOV was significantly elevated in women with PCOS,compared with controls (odds ratio 4.6, 95% CI 1.712.6, P0.003). Thus, PCOM was a better discriminator than PCOVbetween PCOS and control women.

TABLE 1. Model-based estimates adjusting for age and BMI for indices of glucose metabolism, reproductive hormones, and otherultrasound parameters according to diagnosis (PCOS/control)

Variable PCOS (n 88)[mean (95% CI)]Control (n 21)[mean (95% CI)]

Ratio of PCOS tocontrol (95% CI) P value

a

Age (yr)b 27.7 (26.4, 29.0) 29.5 (26.8, 32.4) 0.9 (0.8, 1.0) 0.24BMI (kg/m2)b 35.1 (33.5, 36.9) 25.8 (23.4, 28.5) 1.4 (1.2, 1.5)

We compared reproductive and metabolic parameters be-tween women with PCOS plus PCOM with women withPCOS minus PCOM (Table 2). Women with PCOS plusPCOM had lower FSH levels than women with PCOS minusPCOM. However, there were no differences in the ISI0,120,fasting or 2-h postchallenge glucose or insulin values orintegrated OGTT glucose and insulin responses. Further-more, there were no differences in circulating androgen orgonadotropin levels between these groups. We did not per-form these analyses within the control women given the lowprevalence of PCOM (n 2).Furthermore, we compared these same reproductive and

metabolic parameters between women with PCOV withwomen without PCOV within the PCOS and controls sep-arately (Table 3).Womenwith PCOSplus PCOVhad ahigherLH to FSH ratio than women with PCOS minus PCOV.Control women with PCOV had a greater endometrial thick-ness than control women without PCOV. There were nodifferences in the ISI0,120, fasting or 2-h postchallenge glucoseor insulin values or integrated OGTT glucose and insulinresponses or circulating androgen or gonadotropin levelsbetween PCOV and non-PCOV within the women withPCOS and the control women.We investigated the relationship between these same pa-

rameters and total ovarian volumeboth in the controlwomenand the women with PCOS, without regard to morphology(Table 4). There were no significant correlations betweentotal ovarian volume and any parameter of glucose metab-olism or reproductive hormone levels, with the exception ofFSH, in either the women with PCOS or the control women.In women with PCOS, total ovarian volume was modestly

correlated with FSH (R 0.25; P 0.02). Furthermore,there were no differences in the correlations between thegroups (PCOS vs. control) for total ovarian volume and anyparameter of glucose metabolism (including ISI0,120 in Fig. 2)or reproductive hormone levels (including T in Fig. 3), withthe exception of the LH to FSH ratio (P 0.02) (Table 3).

Discussion

We found no association between either the morphologyor size of the ovary and any parameter of insulin action orwith T levels. These data suggest that the morphology andsize of the ovaries are of little value in identifying distinctivemetabolic or reproductive abnormalities in women with theendocrine syndrome of PCOS. Nonetheless, polycystic ova-ries were highly and significantly prevalent (compared withour control group), although not invariably present inwomen with PCOS. In fact, abnormal ovarian morphologywas significantly less common in the majority of controlwomen selected on the basis of normal reproductive andmetabolic parameters.Our findings of no association between polycystic ovaries

and insulin resistance varies from prior studies, which founda positive association between either polycystic ovaries (16)or ovarian volume with insulin resistance (17). The formerstudy is limited by the small sample size (only five withpolycystic ovaries). The latter study by Pache et al. (17) foundonly a modest correlation between ovarian volume and in-sulin resistance (R 0.23, P 0.05). Furthermore, they usedonly fasting measures of glucose and insulin to calculateinsulin resistance. Our study used an ISI from a dynamic test

TABLE 2. Model-based estimates adjusting for age and BMI for indices of glucose metabolism, reproductive hormones, and otherultrasound parameters according to ultrasound morphology of the ovaries in women with PCOS

VariablePCOS plus PCOM

(n 74)[mean (95% CI)]

PCOS minus PCOM(n 14)

[mean (95% CI)]

Ratio of PCOS plusPCOM to PCOS minus

PCOM (95% CI)P valuea

Age (yr)b 27.2 (26.0, 28.5) 30.3 (27.3, 33.7) 0.9 (0.8, 1.0) 0.06BMI (kg/m2)b 35.4 (33.4, 37.4) 34.0 (29.9, 38.7) 1.0 (0.9, 1.2) 0.58OGTT parametersFasting glucose (mg/dl) 92 (89, 94) 96 (89, 103) 1.0 (0.9, 1.0) 0.29Fasting insulin (U/ml) 22 (20, 24) 27 (21, 34) 0.8 (0.6, 1.1) 0.15Fasting glucose to insulin ratio 4.2 (3.8, 4.6) 3.6 (2.8, 4.5) 1.2 (0.9, 1.5) 0.24Glucose 2 h (mg/dl) 129 (121, 137) 145 (126, 167) 0.9 (0.8, 1.0) 0.14Insulin 2 h (U/ml) 112 (94, 132) 143 (96, 213) 0.8 (0.5, 1.2) 0.27Integrated glucosec 16,546 (15,743, 17,379) 18,043 (16,201, 20,094) 0.9 (0.8, 1.0) 0.15Integrated insulinc 12,939 (11,379, 14,714) 15,105 (11,396, 20,021) 0.9 (0.6, 1.2) 0.33ISI0,120 from OGTT 50 (46, 54) 41 (34, 50) 1.2 (1.0, 1.5) 0.10

Hormonal parametersDHEAS (ng/ml) 1,980 (1,770, 2,216) 2,540 (1,953, 3,303) 0.8 (0.6, 1.0) 0.09T (ng/dl) 67 (61, 73) 68 (55, 85) 1.0 (0.8, 1.2) 0.85uT (ng/dl) 21 (19, 23) 24 (19, 31) 0.9 (0.6, 1.1) 0.30LH to FSH ratio 1.2 (1.1, 1.4) 1.2 (0.9, 1.6) 1.0 (0.7, 1.5) 0.80FSH (mIU/ml) 9.6 (9.1, 10.2) 11.9 (10.4, 13.5) 0.8 (0.7, 0.9) 0.005

Ultrasound parametersTotal ovarian volume (right leftovarian volume) (cm3)

26.4 (24.1, 29.0) 23.4 (18.7, 29.1) 1.1 (0.9, 1.4) 0.31

Endometrial thickness (mm) 7.6 (6.8, 8.5) 8.4 (6.5, 10.9) 0.9 (0.7, 1.2) 0.49

Conversion factors to SI units: T and uT 3.467 (nmol/liter), DHEAS 0.002714 (mol/liter), glucose 0.0551 (mmol/liter), insulin 7.175(pmol/liter).

a P value reported from ANCOVA. Boldface indicates a statistically significant finding.b Not adjusted for age and BMI.c Sample sizes are n 65 for PCOM and n 14 for non-PCOM. Only subjects with glucose and insulin data that were complete at all time

points from the OGTT were included.

2574 J Clin Endocrinol Metab, May 2005, 90(5):25712579 Legro et al. PCO Morphology in PCOS

TABLE

3.Model-based

estimates

adjustingforagean

dBMIforindicesof

glucose

metabolism,reproductivehormones,an

dother

ultrasoundparametersaccordingto

ovarian

volume(PCOV

atleastoneovary10

cm3)within

wom

enwithPCOSan

dwithin

controls

Variable

PCOS

Control

PCOSplusPCOV

(n

71)

[mean(95%

CI)]

PCOSminusPCOV

(n

17)

[mean(95%

CI)]

Ratio

ofPCOS

plusPCOVto

PCOSminusPCOV

(95%

CI)

Pvaluea

Control

plusPCOV

(n

10)

[mean(95%

CI)]

Control

minusPCOV

(n

11)

[mean(95%

CI)]

Ratio

ofcontrol

plusPCOVto

control

minus

PCOV(95%

CI)

Pvaluea

Age

(yr)b

27.4(26.1,28.7)

28.9(26.2,31.9)

0.9(0.9,1.1)

0.32

27.6(22.8,33.4)

31.2(26.1,37.4)

0.9(0.7,1.1)

0.34

BMI(kg/m

2)b

35.2(33.3,37.3)

34.8(31.0,39.1)

1.0(0.9,1.2)

0.86

25.6(22.8,28.8)

26.0(23.2,29.1)

1.0(0.8,1.2)

0.85

OGTTparameters

Fastingglucose

(mg/dl)

92(89,95)

94(88,100)

1.0(0.9,1.0)

0.49

83(79,88)

86(82,91)

1.0(0.9,1.0)

0.41

Fastinginsulin(U/m

l)22

(20,25)

24(19,30)

0.9(0.7,1.2)

0.58

13(11,16)

13(10,15)

1.0(0.8,1.3)

0.83

Fastingglucose

toinsulinratio

4.1(3.7,4.5)

3.9(3.2,4.8)

1.0(0.8,1.3)

0.72

6.5(5.3,7.9)

6.9(5.7,8.3)

0.9(0.7,1.2)

0.66

Glucose

2h(m

g/dl)

131(123,139)

135(119,154)

1.0(0.8,1.1)

0.63

83(71,98)

93(80,108)

0.9(0.7,1.1)

0.31

Insulin2h(U/m

l)110(92,130)

147(103,209)

0.7(0.5,1.1)

0.14

33(21,52)

34(22,52)

1.0(0.5,1.8)

0.94

Integrated

glucose

c16,724

(15,910,17,581)17,138

(15,451,19,009)

1.0(0.9,1.1)

0.67

12,517

(10,159,15,424)12,381

(10,697,14,331)

1.0(0.8,1.3)

0.92

Integrated

insulinc

12,710

(11,186,14,441)16,136

(12,380,21,031)

0.8(0.6,1.1)

0.11

6,357(3,703,10,911)

6,040(4,137,8,819)

1.1(0.5,2.1)

0.86

ISI 0,120from

OGTT

49(45,54)

45(37,53)

1.1(0.9,1.4)

0.34

100(83,120)

90(75,107)

1.1(0.9,1.5)

0.40

Hormonal

parameters

DHEAS(ng/ml)

2,034(1,811,2,285)

2,172(1,712,2,756)

0.9(0.7,1.2)

0.62

1,282(963,1,706)

1,496(1,140,1,965)

0.9(0.6,1.3)

0.43

T(ng/dl)

70(63,76)

58(48,70)

1.2(1.0,1.5)

0.08

30(23,39)

31(24,39)

1.0(0.7,1.4)

0.86

uT(ng/dl)

22(20,25)

18(14,22)

1.3(1.0,1.6)

0.07

6(4,9)

5(3,8)

1.1(0.6,2.1)

0.71

LH

toFSH

ratio

1.3(1.1,1.5)

0.9(0.7,1.2)

1.4(1.1,1.9)

0.02

0.5(0.3,0.8)

0.7(0.5,1.1)

0.7(0.4,1.3)

0.28

FSH

(mIU

/ml)

9.7(9.2,10.3)

11.1(9.8,12.5)

0.9(0.8,1.0)

0.05

9.7(7.9,11.9)

10.3(8.5,12.5)

0.9(0.7,1.2)

0.64

Ultrasoundparameters

Total

ovarianvolume

(right

leftovarian

volume)

(cm

3)

29.8(27.9,31.9)

14.5(12.6,16.7)

2.1(1.8,2.4)