Screening for Androgen Excess in Women: Accuracy of Self-Reported Excess Body Hair Growth and Menstrual Dysfunction

Abstract

Context

Epidemiologic studies of polycystic ovary syndrome (PCOS) are limited, especially in populations where diagnostic resources are less available. In these settings, an accurate, low-cost screening tool would be invaluable.

Objective

To test the use of a simple questionnaire to identify women at increased risk for PCOS and androgen excess (AE) disorders.

Study Design

Prospective cohort study from 2006–2010.

Setting

Community-based.

Participants

Women aged 14 to 45 years.

Intervention

A screening telephone questionnaire consisting of 3 questions was tested, where participants were asked to self-assess the presence/absence of male-like hair and menstrual irregularity. Participants were then invited to undergo a direct examination, including completing a medical history and undergoing a modified Ferriman-Gallwey (mFG) hirsutism score, ovarian ultrasound, and measurement of circulating total and free testosterone, DHEAS, TSH, prolactin and 17-hydroxyprogesterone levels.

Main Outcome Measure

Accuracy of questionnaire in predicting PCOS, AE, and irregular menses.

Results

Participants with self-assessed irregular menses and/or excess hair were labeled “Possible Androgen Excess (Poss-AE)” and those self-assessed with regular menses and no excess hair were labeled “Probable Non-Androgen Excess (Non-AE).” The study was completed in 206/298 (69%) of the Poss-AE and in 139/192 (73%) of the Non-AE. Of Poss-AE and Non-AE subjects, 82.5% and 15.8%, respextively, presented with PCOS. The calculated sensitivity, specificity, positive predictive value, and negative predictive value of the 3-question telephone survey to predict PCOS was 89%, 78%, 85%, and 83%, respectively.

Conclusions

A simple telephone questionnaire, based on self-assessment of body hair and menstrual status, can be used with a high predictive value to identify women at risk for AE disorders, including PCOS, and to detect healthy controls. This approach could be an important tool for needed epidemiologic studies.

Polycystic ovary syndrome (PCOS), characterized by menstrual irregularity, hyperandrogenism, and polycystic ovarian morphology, affects approximately 5% to 15% of reproductive-aged women, depending on the criteria used (1–4). Epidemiologic studies are critical not only for assessing prevalence, but also for determining public health burden and fostering the development of associated preventive and diagnostic policies and strategies. Furthermore, comparative epidemiologic studies are critical for determining the role that ethnicity/race and environmental factors play in the development and phenotype of the disorder (5, 6).

However, epidemiologic studies of PCOS have been few, especially in populations where diagnostic resources are less available. In part, this relates to the diagnostic complexity of the assessment, and in part, to the associated costs. Current methods for screening of populations for PCOS are labor-intensive, requiring a physical examination, obtaining a medical history, and performing blood testing and ovarian ultrasonography in large numbers of women, 80% to 90% of whom will not have the disorder (7, 8).

Low-cost tools that would facilitate the prediction of PCOS are needed to allow for better targeting and utilization of scant resources in epidemiologic studies. Furthermore, the advent of such tools would allow for more effective identification and prescreening of affected and unaffected individuals for inclusion in genetic studies (1) and clinical trials (9) of the disorder. The present study was designed to determine the predictive value of a simple telephone questionnaire to identify women at increased risk for PCOS and other androgen excess (AE)-related symptoms, and to identify healthy women that could be recruited as healthy controls.

Materials and Methods

Subjects

The study was conducted in Los Angeles, California. Women were recruited through advertisements, one calling for women with irregular menses and/or unwanted body/facial hair growth, and the other calling for women who felt they were healthy and normal. All individuals were then screened by telephone using a 3-question questionnaire (see below). To be eligible for further study, subjects had to be from 14 to 45 years of age and not have a history of hormonal medication or contraception use in the 3 months prior to the study. Approval for this study was provided by the Cedars-Sinai Medical Center Institutional Review Board and consent was provided by all subjects, including assent at the time of the telephone survey and written consent at the time of their full evaluation.

Study protocol

Participants who called in response to the advertisements were interviewed by telephone after providing study assent and asked to self-assess the presence or absence of excessive/unwanted body/facial hair and menstrual irregularity. The telephone questionnaire included the following 3 questions:

1. Do you have infrequent or irregular cycles; either more than every 35 days between the beginning of one period and the next, or 8 or less cycles per year?

2. Do you have male-like hair growth on your upper lip, chin, chest, abdomen, buttocks, or back?

3. Are you on any hormonal medications (birth control pills, hormone medications, etc.) or any other medications in the past 3 months?

Based on the responses given, subjects were divided into 2 cohorts: (a) women with self-assessed irregular menses and/or male-like body hair were designated as having “Possible AE (Poss-AE)”; and (b) women self-assessed as having regular menses and no male-like body hair were designated as “Probably not having AE (Non-AE).” Respondents were invited for further study if they responded “Yes” to either question #1 or #2, and responded “No” to question #3 (Poss-AE subjects), or if they responded “No” to all 3 questions (Non-AE subjects). Further evaluation of eligible subjects who agreed to participate included undergoing a complete history and physical exam, including a modified Ferriman-Gallwey (mFG) hirsutism scoring (29) and ovarian ultrasonography, and blood sampling. Participants were asked not to remove any body hair for 1 week prior to their evaluation.

To minimize the number of visits, appointments were scheduled according to the participants’ menstrual cycle, if they had any, as follows. Women who reported menstrual cycle lengths of 26 to 34 days were seen between cycle days (CDs) 2 to 10 for their examination and measurement of baseline hormone levels. No measurement of a luteal phase progesterone (P4) level was performed. Subjects with menstrual cycle lengths of 35 to 60 days and a negative pregnancy test (Quick Vue One-Step hCG Combo Test; Quidel Pacific Biotech, Inc.) underwent baseline hormonal testing on CDs 2 to 12. Prometrium 200 mg at bedtime (10) was offered to subjects with cycles longer than 60 days and a negative urine pregnancy test, to induce vaginal bleeding and consequently schedule their evaluation on CDs 2 to 10 following the start of their bleed.

Blood testing included serum levels for total testosterone (TT), free testosterone (FT), dehydroepiandrosterone sulfate (DHEAS), prolactin (PRL), TSH, and 17-hydroxyprogesterone (17-HP). If perimenopausal/menopausal (e.g., vasomotor) symptoms were described by the subject, serum follicle-stimulating hormone (FSH) and estradiol (E2) were measured on menstrual CDs 2 to 3 before further hormonal evaluation (11).

Ovarian and uterine morphology was evaluated using either transvaginal ultrasonography (TV-US) or abdominal ultrasonography for those patients not tolerating or nondesirous of TV-US (Philips EnVisor Ultrasound System, with 6.26 MHz endovaginal transducer). Sonographic criteria for polycystic ovaries included: (a) an ovarian volume of ≥ 11 cm³ in at least 1 ovary, and/or (b) the presence of 12 or more follicles 2 to 9 mm in diameter in the ovarian cortex of at least 1 ovary (Rotterdam criteria 2003) (12, 13). The sonographic criteria of 20 or more follicles per the updated International Consortium guidelines were secondarily considered in the analysis (4). Ovarian volume was calculated using the simplified formula for the prolate ellipsoid volume (0.52 × length × width × thickness in cm) using the maximum longitudinal, antero-posterior and transverse diameters.

Hyperandrogenism was defined as the presence of hirsutism and/or hyperandrogenemia. The mFG method for determining the extent of unwanted hair growth and the presence of hirsutism includes assessment of 9 areas of the body (upper lip, chin, neck, chest, upper and lower abdomen, thighs, upper and lower back) and each area is scored on a scale of 0 (no terminal hair growth) to 4 (male-like hair growth). The total mFG score is the sum of all the body area scores. Women with an mFG score ≥ 4 were considered to be hirsute, as previously described and consistent with the current international guidelines (4). Hyperandrogenemia was defined either as a TT > 35.0 ng/dL, a FT > 3.5 pg/mL (14, 15), and/or elevated age-specific DHEAS levels (16), as previously determined.

Eumenorrhea was defined as regular menstrual periods with 26 to 34 CD intervals. Oligo-ovulation was defined by oligo-amenorrhea (menstrual cycles ≥ 35 days in length, or ≤ 8 bleeding episodes per year) (17, 18).

Twenty-one-hydroxylase–deficient non-classic adrenal hyperplasia (NCAH) was excluded by a basal follicular phase 17-HP level of ≤ 2.0 ng/mL performed in the morning (19). If the basal 17-HP exceeded this cut-off value, the subject underwent an acute 60 minute 0.25 mg Cortrosyn (Amphastar Pharmaceuticals Rancho Cucamonga, California) stimulation test; a 17-HP level after acute adrenocorticotropic hormone (ACTH) stimulation of ≥ 10.0 ng/mL was diagnostic of NCAH.

Hyperprolactinemia and hypothyroidism were excluded by normal levels of PRL and TSH, respectively. Women on treatment for thyroid abnormalities and hyperprolactinemia were included in the study if TSH or prolactin levels had been normal and stable for 3 months prior to evaluation. Cushing syndrome was excluded, if clinically indicated, by a 24-hour urine free cortisol level < 100.0 mg.

If symptoms requiring medical consultation were detected prior to complete evaluation, the subjects were referred for medical evaluation and follow-up as appropriate. Likewise, if the endometrial stripe was > 6 mm after a vaginal bleed, if there was no response to a P4 challenge, or if ovarian abnormalities were found on ultrasonography.

Assays

Serum levels of TT and 17-HP were determined by liquid chromatography tandem mass spectrometry, and FT determined based on these results using equilibrium dialysis (Quest Diagnostics, San Juan Capistrano, CA) (15). The levels of PRL, DHEAS, and TSH were obtained on the IMMULITE (Diagnostic Products Corp., Los Angeles, CA) automated, random-access, chemiluminescent immunoassay system.

Statistical analysis

Numerical variables satisfying normality were compared using the Student t test to obtain group mean differences and data is presented as mean ± standard deviation (SD). Numerical variables exhibiting skewedness were log-transformed in order to achieve normality. Sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) were calculated using standard methods. The PCOS analysis was performed using both Rotterdam 2003 criteria (12, 20) and updated ultrasound criteria based on the International Consortium guidelines (4). A subanalysis was performed on women under the age of 40, given the potential impact of older age on regularity of menses, serum androgen levels, antral follicle count, and ovarian volume on ultrasound. An additional subanalysis was performed on the impact of ethnicity/race in the Poss-AE and Non-AE ethnic cohorts.

Results

Of the women who met inclusion criteria by the telephone interview, 490 women presented for at least 1 study visit. Of these, 298 women were considered as Poss-AE (i.e., yes to either questions #1 or #2, and no to question #3) and 192 women were considered Non-AE (i.e., no to all 3 questions). The full evaluation was completed in 206 of 298 (69.1%) and 139 of 192 (72.4%) of the Poss-AE and Non-AE cohorts, respectively. The other 145 women did not complete the evaluation.

Study population demographics are presented in Table 1. Poss-AE were slightly but significantly younger and more obese than the non-AE subjects. The mean (± SD) ages of the 206 Poss-AE and 139 Non-AE subjects were 29.8 ± 6.3 (range, 14-45) years and 31.9 ± 6.9 (range, 20-44) years, respectively (P = 0.003), and the mean body mass index (BMI) of the cohorts were 31.8 ± 8.3 kg/m² and 27.0 ± 6.5 kg/m², respectively (P ≤ 0.001). The ethnicities of the subjects were representative of the population seen at the study institution (Table 1).

Table 1.

Study Population Demographics

	Poss-AE (n = 206)	Non-AE (n = 139)
Age (y) mean ± SD (range)	29.8 ± 6.3 (14–45)	31.9 ± 6.9 (20–44)a
BMI (kg/m2) mean ± SD	31.8 ± 8.3	27.0 ± 6.5b
Mean mFG score mean ± SD (range)	6.7 ± 4.3 (0–22)	1.3 ± 2.0 (0–10)b
mFG ≥ 4	155/206 (75.2%)	20/139 (14.4%)
Race/ethnicity
Non-Hispanic White	77/206 (37.4%)	41/139 (29.5%)a
Hispanic White	66/206 (32.0%)	32/139 (23.0%)
Black	36/206 (17.5%)	37/139 (26.6%)
Asian	16/206 (7.8%)	20/139 (14.4%)
Other	11/206 (5.3%)	9/139 (6%)

Abbreviations: BMI, body mass index; mFG, modified Ferriman-Gallwey; Non-AE, probably no androgen excess disorder; Poss-AE, possible androgen excess disorder; SD, standard deviation.

^a  P < 0.05 as compared with the Poss-AE group

^b  P < 0.0001 as compared with the Poss-AE group

Comparing the accuracy in self-assessing the presence of hirsutism and menstrual dysfunction

Among women with Poss-AE, 75.2% had an mFG ≥ 4 (i.e., hirsutism). In the non-AE cohort, 14.4% had hirsutism. The mean mFG scores in the Poss-AE and Non-AE cohorts were 6.7 ± 4.3 (range, 0–22) and 1.3 ± 2.0 (range, 0–10), respectively (P ≤ 0.001). On evaluation, hirsutism (i.e., mFG ≥ 4) was present in 141 of 171 (82.5%) participants who had stated that they had excess or unwanted body/face hair growth. Of the 31 women who stated that they did not have male-like hair growth, 14 (45%) were found to have an mFG score ≥ 4. (Table 2).

Table 2.

Study Population Phenotypes and Diagnoses

	Poss-AE (n = 206)	Non-AE (n = 139)
Oligo-amenorrhea only (%)	35/206 (17.0%)	0/139 (0%)
Hirsutism only (%)	46/206 (22.3%)	19/139 (13.7%)
PCOM only (%)	9/206 (4.4%)	41/139 (29.4%)
Oligo-amenorrhea +  Hirsutism + PCOM (%)	83 (40.2%)	1/139 (0.7%)
Oligo-amenorrhea  + Hirsutism (%)	11/206 (5.3%)	0/139 (0%)
Oligo-amenorrhea  + PCOM (%)	24/206 (11.7%)	0/139 (0%)
Hirsutism + PCOM (%)	34/206 (16.5%)	10/139 (7.2%)
PCOS by Rotterdam criteria (2003)c (%)	170/206 (82.5%)	22/139 (15.8%)
NCAH (%)	1/206 (0.5%)	0/139 (0%)
Hyperprolactinemia (%)	3/206 (1.5%)	2/139 (1.4%)
Thyroid abnormality (%)	4/206 (1.9%)	2/139 (1.4%)
Free testosterone (pg/mL) mean ± SD	5.2 ± 3.7	2.4 ± 1.9b
Total testosterone (pg/mL) mean ± SD	40.9 ± 25.8	26.5 ± 14.3b
DHEAS (ug/dL) mean ± SD	241.1 ± 118.6	193.3 ± 98.9a

Abbreviations: DHEAS, dehydroepiandrosterone sulfate; NCAH, non-classic adrenal hyperplasia; Non-AE, probably no androgen excess disorder; PCOM, polycystic ovarian morphology; Poss-AE, possible androgen excess disorder; SD, standard deviation.

^a  P < 0.05 as compared with the Poss-AE group

^b  P < 0.0001 as compared with the Poss-AE group

^cTwo of the following 3 criteria: oligo-amenorrhea, hyperandrogenism, or polycystic ovarian morphology

Of the 168 women who answered the telephone questionnaire stating that they had irregular menses, oligo-amenorrhea was observed in 69.4% (143/206) of the total Poss-AE cohort. In the Poss-AE group, 177/206 (85.9%) accurately self-assessed the presence of either oligo-amenorrhea or normal menses. In the Non-AE group, 98.6% accurately self-assessed regular menses. There was no statistically significant difference in the accuracy to predict menstrual status between women in the Non-AE and Poss-AE cohorts. Overall, correct assessment of both menstrual and body hair status was made in a higher percentage of the Non-AE cohort (85.6%) compared with the AE cohort (67.0%), P < 0.001.

Estimates of sensitivity, specificity and predictive values of the telephone questionnaire for prediction of PCOS

The diagnoses of the women in our study population are depicted in Table 2. PCOS by Rotterdam criteria was observed in 82.5% (170/206) of Poss-AE subjects, similar to the prevalence observed if PCOS was identified by the updated International Consortium 2018 guidelines (168/206 or 81.6%) (4). In women who answered “yes” to both of the first questions, 101/136 (74.3%) also had polycystic ovary morphology (PCOM) on ultrasound. Interestingly, 15.8% (22/139) of the non-AE cohort was found to also have PCOS by Rotterdam criteria. The prevalences of the 4 phenotypes of PCOS are also reported in Table 2. The proportion of PCOS phenotype A (hyperandrogenism, oligo-amenorrhea, and PCOM) was 40.2%. The proportion of phenotype B (hyperandrogenism and oligo-amenorrhea) was 5.3%, phenotype C (hyperandrogenism and PCOM) was 16.5% and phenotype D (oligo-amenorrhea and PCOM) was 11.7%.

The specific values for sensitivity, specificity, NPV, and PPV are denoted in Table 3. Considering the entire population studied, the sensitivity, specificity, PPV and NPV of the telephone questionnaire to predict a diagnosis of PCOS as defined by the Rotterdam 2003 criteria was 89%, 78%, 85% and 83%, respectively. We found similar test characteristics to predict PCOS as defined by International Consortium guidelines: sensitivity 88%, specificity 77%, PPV 84%, NPV 81%. The sensitivity, specificity, PPV and NPV of the telephone questionnaire to predict male-like body hair status was 81%, 80%, 81%, and 80% respectively and to predict oligo-amenorrhea was 98%, 87%, 84%, and 98%, respectively.

Table 3.

Estimates of Sensitivity and Specificity of the Questionnaire*

	Sensitivity	Specificity	PPV	NPV
Total Population
Accuracy of self-reported male-like body hair for predicting hirsutism (i.e., mFG ≥ 4)	81%	80%	81%	80%
Accuracy of self-reported oligo-amenorrhea	98%	87%	84%	98%
Accuracy to detect PCOS by Rotterdam 2003a	89%	78%	85%	83%
Women <40 years old
Accuracy of self-reported male-like body hair for predicting hirsutism (i.e., mFG ≥ 4)	81%	80%	82%	79%
Accuracy of self-reported oligo-amenorrhea	97%	87%	86%	97%
Accuracy to detect PCOS by Rotterdam 2003a	76%	86%	90%	67%

Abbreviations: mFG, modified Ferriman-Gallwey; NPV, negative predictive value; PCOS, polycystic ovary syndrome; PPV, positive predictive value.

*Irrespective of accuracy of self-assessment

^aTwo of the following three criteria: oligo-amenorrhea, hyperandrogenism, or polycystic ovarian morphology

Subanalysis of women under 40 years of age

Given the potential impact that older age may play in androgen values, regularity of menses, and appearance of ovaries on ultrasound, we performed a subanalysis including only women under the age of 40 (n = 311). The sensitivity, specificity, PPV, and NPV of the questionnaire for predicting hirsutism, oligo-amenorrhea, and PCOS by Rotterdam is demonstrated in Table 3. Sensitivity, specificity, PPV, and NPV were approximately similar to the total cohort for accuracy of predicting hirsutism and oligo-menorrhea; however, both sensitivity and NPV to detect PCOS by Rotterdam criteria were lower, at 76% and 67%.

Racial/ethnic differences

A subanalysis was performed on the different race/ethnicities in the Poss-AE and Non-AE ethnic cohorts, categorizing subjects as Non-Hispanic White (NHW), Hispanic Whites (HW), Black, Asian, and Other. NHW was the most prevalent race/ethnicity in both the Poss-AE and Non-AE cohorts (Table 1). Asian women had the lowest mean BMI compared with the other groups (Asian 24.2 ± 5.5 kg/m², NHW 29.3 ± 8.5, HW 31.4 ± 7.6, Black 31.4 ± 7.6, Other 30 ± 7.7; P = 0.0001) and the lowest mFG score (Asian 2.8 ± 4.2, NHW 4.6 ± 4.5, HW 5.2 ± 4.2, Black 4.5 ± 4.7, and Other 4.3 ± 4.9; P = 0.04).

In the Poss-AE cohort, self-assessed excessive body hair was reported by 90%, 85%, 94%, 75%, and 91% of NHW, HW, Black, Asian, and Other subjects, respectively; hirsutism (mFG ≥ 4) was found in 69%, 89%, 63%, and 73% of these racial/ethnic cohorts, respectively, with no statistically significant difference between groups (P = 0.11). Irregular menses was reported in 74%, 91%, 78%, 94%, and 73% of the NHW, HW, Black, Asian, and Other groups, respectively; oligo-amenorrhea was demonstrated in 62%, 82%, 61%, 81%, and 72% of these racial/ethnic groups, respectively, with no statistically significant difference noted between groups (P = 0.06). PCOS (by updated International guidelines) was seen in 81%, 85%, 88%, 93%, and 82% of NHW, HW, Black, Asian, and Other subjects, respectively, with no significant difference in prevalence between groups (P = 0.74).

In the Non-AE cohort, the mean mFG scores did not differ between the different ethnicities in the Non-AE cohort. Prevalence of PCOS was 18%, 30%, 11%, 21%, and 13% in NHW, HW, Black, Asian, and Other subjects, respectively, with no significant difference in prevalence between groups (P = 0.37). The sensitivities, specificities, PPV, and NPV of the questionnaire to detect any form of PCOS in the different ethnic cohorts was also examined. No significant difference in predictability of the questionnaire to detect PCOS between the difference ethnic groups was found (data not shown).

Discussion

The purpose of this study was to evaluate a simple telephone questionnaire, assessing self-reported menstrual and body hair status, to assist in the identification of women at increased risk for AE disorders, especially PCOS. In the Poss-AE cohort, 83% of participants were found to have PCOS according to Rotterdam criteria. We found similar test characteristics in predicting PCOS using the most recent International Consortium guidelines (4). When all participants in both the Poss-AE and Non-AE cohorts were considered, the sensitivity, specificity, PPV, and NPV of the questionnaire to predict a diagnosis of PCOS by Rotterdam criteria was 89%, 78%, 85%, and 83%, respectively. Overall, our data suggest that the use of this simple questionnaire, which can be administered by telephone, is highly useful as a screening tool for PCOS and other AE disorders.

A simple, questionnaire-based tool to distinguish women with PCOS was used in a large Finnish longitudinal cohort. The questionnaire used questions for oligo-amenorrhea (“Is your menstrual cycle often, over twice a year, more than 35 days?”) and hirsutism (“Do you have excessive hair growth?”), similar to the questionnaire used in this study. Studies arising from these data demonstrated that these questions are useful in identifying women with a PCOS endocrine, metabolic, psychologic profiles, as well as PCOM on ultrasound (21–23). In the 2004 study published on this cohort, the prevalence of PCOM was 70.4% in women who reported both hirsutism and oligo-amenorrhea (“yes” to both questions) (22). This is similar to what we found in our study (74%). Higher medians of serum testosterone were also observed among these symptomatic women compared with controls (21). Higher levels of TT and FT, as well as DHEAS, were found in our Poss-AE group. This data demonstrates that this screening questionnaire approach can be applied in different populations and establishes its use an important epidemiologic tool. When comparing the prevalences of the 4 PCOS phenotypes, we found that the proportion of PCOS phenotype A was higher in our study compared with unselected populations evaluated in a prior meta-analysis by Lizneva et al (40.2% vs 19.0%) (24), but similar to the prevalence of phenotype A in a referral population (50%). The other phenotypes in our study were also more similar in prevalence compared with the referral population (Phenotype B: 5.3% vs 13%; Phenotype C: 16.5% vs 14%; and Phenotype D: 11.7% vs 17%) compared with the unselected in that study (Phenotype B: 25%; Phenotype C: 34%; and Phenotype D: 19%). As noted in the discussion of the Lizneva et al meta-analysis, participants who did not complete the full medical and laboratory evaluation for PCOS may bias towards a higher prevalence of phenotypes B, C, or D, which may be the case for women in the unselected population. Since all of our participants completed the full evaluation, the phenotype prevalences are closer to a referral population (in which phenotype A was the highest).

Accuracy of reporting male-like body hair for predicting an mFG score of ≥4 was high in the overall cohort with a sensitivity, specificity, PPV, and NPV of 81%, 80%, 81%, and 80%, respectively. As an examiner, it can be difficult to determine the true extent of hirsutism; however if present, hirsutism has been shown to be a good predictor of PCOS (25).

Accurate self-assessment of menstrual status was high in both the Poss-AE and Non-AE cohorts (86% and 99%, respectively). The sensitivity, specificity, PPV, and NPV of the questionnaire to correctly predict irregular menstrual status was 98%, 87%, 84%, and 98%, respectively. A higher number of women in the Non-AE cohort than those in the Poss-AE cohort correctly self-assessed both their menstrual and body hair status (86% vs 67%, respectively), suggesting that the questionnaire is also a good tool for recruiting healthy controls.

A number of studies have found ethnic differences in the prevalence and severity of AE symptoms (26–29). In the Poss-AE racial/ethnic cohorts PCOS was found in 81% to 93% of women, most prevalent in Black and Asian respondents (88% and 93%, respectively). A Taiwanese study from 2007 indicated a similar prevalence of PCOS in a patient cohort seen in the clinical setting, although the BMI of included women was much lower than in our study cohort (30).

The strength of this study is that it tested a simple screening questionnaire in a relatively large and medically unbiased population, combined with careful phenotyping of respondents. However, there are potential limitations to this study, the most significant of which is that participants may not be fully representative of women in the general population. For example, 16% of the non-AE group met criteria for PCOS, suggesting a degree of self-selection bias and highlighting the need to accurately phenotype all women responding to advertisements claiming to be “healthy” or “normal.” We recognize the potential impact of older age on regularity of menses, serum androgen levels, antral follicle count, and ovarian volume on ultrasound. No age-specific criteria for PCOM have been established and no age-specific cutoffs for testosterone levels were available in this study. Given this potential bias of our findings through inclusion of women > 40 years of age, we performed a subanalysis in women under the age of 40 (n = 311). This analysis of the performance of the questionnaire in women younger than 40 years of age demonstrated similar sensitivity, specificity, PPV, and NPV for accuracy prediction of hirsutism and oligo-amenorrhea, but found lower sensitivity and NPV for predicting PCOS by Rotterdam criteria. PCOM ovaries seen in older women older than 40 years is likely a more indicative finding for PCOS compared with PCOM ovaries seen in younger women, since younger women are expected to have a more robust antral follicle count (even in the absence of PCOS). Therefore, as expected, both the sensitivity and NPV for predicting PCOS in women younger than 40 are lower.

Identifying a simple and valid screening tool for PCOS is invaluable for epidemiologic studies of the disorder. Furthermore, it facilitates the liberal screening for PCOS, which has been shown to be a cost-effective preventive strategy, leading to earlier diagnosis, intervention and possibly the prevention of serious sequelae (31). While other large epidemiologic studies, such as the National Health and Nutrition Examination Survey (32) and the Nurses’ Health Study (33), have obtained menstrual histories, history of male-like hair growth is not well-elicited. Inclusion of both questions on irregular menses and excess/unwanted facial and body hair growth is crucial to our efforts to characterize PCOS in larger populational cohorts, and consequently better determine the epidemiology, public health and economic burden, and the value of preventive strategies for the disorder.

Glossary

Abbreviations

17-HP

17-hydroxyprogesterone

AE

androgen excess

BMI

body mass index

CD

cycle day

DHEAS

dehydroepiandrosterone sulfate

FT

free testosterone

HW

Hispanic White

mFG

modified Ferriman-Gallwey

NCAH

21-hydroxylase deficient non-classic adrenal hyperplasia

NHW

Non-Hispanic White

NPV

negative predictive value

P4

progesterone

PCOM

polycystic ovary morphology

PCOS

polycystic ovary syndrome

PPV

positive predictive value

PRL

prolactin

TSH

thyrotropin (thyroid-stimulating hormone

TT

total testosterone

Additional Information

Disclosure Summary: RA served as consultant to Ansh Labs, Medtronics, Spruce Biosciences, received grant support from Ferring Pharmaceuticals, and served on the Advisory Board of Martin PET Imaging; MDP received grant support from Ferring Pharmaceuticals. JLC, MP, UE, and RM have no competing interests to declare.

Data Availability

The dataset generated during and/or analyzed during the current study are not publicly available, but are available from the corresponding author on reasonable request.