Signs of Potential Androgen Excess Across the Lifespan in a US-based Digital Cohort Study

Abstract

Context

Androgen excess (AE)-related symptoms can vary widely and may appear across the life course.

Objective

We assessed the prevalence of signs of potential AE and heterogeneity by demographic/health characteristics.

Methods

We used data of 24 435 participants who consented and enrolled during November 2019 to December 2022 in a US digital cohort to evaluate the prevalence and heterogeneity of self-reported signs of potential AE: possible hirsutism (having thick coarse hair on ≥4 of 8 body locations), hair level on the chin, hair loss on top of the head, and moderate to severe acne.

Results

The prevalence of possible hirsutism, having several/a lot of hair on the chin, significantly reduced hair/visible scalp on top of the head, and moderate to severe acne were 6.9%, 12.6%, 1.7%, and 31.8%, respectively. While possible hirsutism and moderate to severe acne decreased with age (range: 18-86 years), hair on the chin and hair loss on the head increased with age. Participants who self-identified as Hispanic or South Asian reported a higher prevalence of possible hirsutism (11.2%, 16.9%, vs 6.3% among non-Hispanic White participants). Participants with higher body mass index had a higher prevalence of possible hirsutism. Moderate to severe acne was more common among those with polycystic ovary syndrome. Possible hirsutism and hair loss were less common among participants using hormones for contraception.

Conclusion

In this large cohort, signs of potential AE varied by demographic and health factors. These results could provide a new understanding of how potential AE may appear differently in diverse groups, informing future work to develop more inclusive evaluation at a population level.

Androgen excess (AE)-related symptoms include hirsutism, acne, alopecia, and, in extreme settings, virilization (1-3). Clinical signs of AE may be multifactorial including familial/genetic contributions, diseases promoting androgen production, and other factors. Variation of these symptoms in females may occur during developmental time windows of puberty, at reproductive age due to ovulation disorders such as polycystic ovary syndrome (PCOS), and during menopause (4). Symptoms related to AE often are associated with biochemical hyperandrogenism, defined by elevated levels of androgen above the 95th percentile (5).

Clinic-based studies on hyperandrogenism postmenopause showed mixed results: some suggested improvement (3), while others indicated persistent elevated androgen levels, particularly among those with PCOS (6, 7). Symptoms that have been used to evaluate AE are known to vary by ethnicity, with higher levels of terminal hair observed among females of Hispanic, South Asian, and Middle Eastern backgrounds (8). Recently, a large, multinational (East Asia/Middle East/Europe/Africa/America) study of 10 919 participants recommended taking ethnicity into consideration when defining hirsutism, based on variations observed in the modified Ferriman-Gallwey score (mFG) (9). However, it remains important to analyze the prevalence of symptoms relevant to AE by race and ethnicity in the United States, recognizing that race is a social construct (10, 11) and that socioeconomic and environmental determinants (12-14), in addition to ancestral or geographic origins, may influence AE signs. Studies have also found a higher body mass index (BMI) is associated with a higher likelihood of AE-related symptoms (15). Apart from the study by Kiconco et al, most studies that estimated the prevalence of AE and/or self-reported clinical signs of AE were case control/cohort studies based on small or moderate-sized, homogenous populations or from clinical settings (5, 16-20). One study that included 369 Black or White females in the United States found no significant difference in hirsutism by this dichotomized measure of race (18). Another study of 156 Mexican Americans found a high prevalence of PCOS (13%) or AE (46%) (19). A cross-sectional analysis among 720 patients with PCOS found that up to 75% had AE (21). Another community-based study of 6986 individuals from Australia found a self-assessed mFG score ≥ 10 as indicative of hirsutism (20). One study utilized the National Health and Nutrition Examination Survey database and estimated the prevalence of hyperandrogenism among 7561 participants, highlighting 10% with hyperandrogenemia based on total testosterone and 4% based on free androgen index, but the data was focused on biochemical assessment of hyperandrogenism only (22).

Studies aiming to describe AE for aiding the diagnosis of PCOS may use populations referred to care, which can be subject to referral bias influenced by severity of symptoms, awareness of PCOS, and access to healthcare. In addition to hirsutism, other symptoms that can be used to evaluate potential AE such as alopecia or acne were less extensively studied, especially among unselected populations over the life course. Thus, a study is needed that investigates a broad spectrum of self-reported signs of potential AE without selecting for a diagnosis of PCOS or other health conditions.

The objective of our study was to assess the prevalence of self-reported signs of potential AE in a large, mobile-application-based cohort in the United States. We also evaluated heterogeneity in these patterns by age groups, race and ethnicity, Asian ancestry, BMI, PCOS status, and hormone use indications. Of note, our evaluation was restricted to self-reported symptoms that may indicate potential AE. Consequently, throughout the remainder of this manuscript, we describe key outcomes of interest as signs of potential AE instead of using any clinical terminology (eg, “hirsutism” or “alopecia”). Rationales of selecting the most relevant outcomes of interest (possible hirsutism defined as having thick coarse hair on ≥4 of 8 body locations, hair levels on the chin, severity of hair loss on top of the head, and moderate to severe acne) are further described in Materials and Methods.

Materials and Methods

Study Population

This study included participants of the Apple Women's Health Study. Eligibility for the Apple Women's Health Study was described previously (23). Briefly, eligible participants were users of the Apple Research app on their iPhone (sole usage, including their iCloud account) who have ever menstruated, live in the United States, were at least 18 (19 in Alabama and Nebraska, 21 in Puerto Rico) years old, and were able to communicate in English. Enrollment began November 2019 and is ongoing. Written informed consent of participation was provided at enrollment. For this analysis, we included a total of 24 435 participants who consented and enrolled from November 14, 2019, to December 13, 2022. Inclusion criteria for this analysis was further restricted to those who reported female sex assigned at birth and completed the Hormonal Symptoms Survey (at baseline) and provided baseline demographics, medical, and reproductive history data (flowchart in Supplemental Fig. S1) (24).

Signs of Potential AE

We evaluated the following signs of potential AE based on the participant's response to relevant survey questions. Details of the survey questions used are shown in Supplemental Table S1 (24). For all questions, responses of “I don’t know” or “I prefer not to answer” were considered missing and excluded from the analysis.

Possible hirsutism

While the mFG score is widely regarded as the clinical gold standard for diagnosing hirsutism, it is important to recognize that the score was developed from a limited sample of 430 White women in England (25), and its generalizability to US individuals with heterogeneous demographic and health backgrounds may be limited. Although our study is constrained by the absence of data necessary to calculate an mFG score, we derived “possible hirsutism” based on self-reported symptoms. This approach aims to inform future research on developing more inclusive tools for assessing clinical signs of potential AE. Based on the question about having thick, coarse, and dark hair (yes/no) on 8 body locations (upper lip, chin, breasts, chest between the breasts, back, abdomen, upper arms, and upper thighs), we derived a summary score, calculated as the sum of body locations with thick, coarse, and dark hair (range: 0-8). We examined the distribution of this discrete summary score and used the score value (= 4) that corresponded to the internal 95th percentile as the cutoff value for defining “possible hirsutism” within this population. This approach aligns with general recommendations for defining hyperandrogenism based on androgen levels exceeding the 95th percentile (5, 26, 27).

Hair levels on the chin

Participants were given diagrams (Supplemental Table S1) (24) to aid their selection (none, a few, several, a lot) for hair on their chin and on their upper lip. Previous research suggested either the chin area or a combination of upper/lower abdomen and chin is a relatively accurate predictor of the mFG score (28-30). Thus, we used the severity level measure on the chin as another potential indicator of hirsutism. The choice of this location (chin) was constrained by the availability of data within this cohort.

Hair loss severity on top of the head

Participants were given diagrams (Supplemental Table S1) (24) to aid their selection (thick, thin hair with widening part, thin hair with widening part and some scalp showing, significantly reduced hair and more scalp showing, or scalp mostly visible) for hair loss on top of their head. We used this severity level measure on top of the head (typically representing female pattern hair loss) (31) as an indicator for evaluating alopecia that may originate from potential androgen excess.

Moderate to severe acne

Participants were given the options of no acne, small/raised pimples, red/irritated pimples, pimples that have pus, or scars from acne to describe acne on their face or back. We defined moderate to severe acne as reported “red, irritated pimples” and/or “pimples that have pus.”

Secondary outcome measures of potential AE

Other survey questions relevant to clinical signs of potential AE were evaluated as secondary outcomes (survey questions in Supplemental Table S1) (24): hair on the body relative to female family members (less, more, about the same); hair levels on the upper lip (none, a few, several, a lot); experience of hair loss (yes/no) from 5 locations of the head (temples or sides of the forehead, top of the head, hairline, random/small patches across the head, or across the whole head); hair loss relative to female peers (less hair loss, more hair loss, about the same); responses to the 4 different characteristics of acne (small/raised pimples, red/irritated pimples, pimples that have pus, or scars from acne, exploring yes/no responses separately for each characteristic); and acne compared to female peers (less, more, same).

Demographic and Health Characteristics

Race and ethnicity were self-identified and grouped as non-Hispanic White (n = 18 233), non-Hispanic Black (n = 1169), Asian (n = 695), Hispanic (n = 1521), multiple races (n = 2273), or other races (n = 470). The “other races” group (2% of this analytic sample) included American Indian or Alaska Native, Middle Eastern or North African, Native Hawaiian or Pacific Islander, or “none of these categories can fully describe me.” Among those who self-identified as Asian, participants could further identify Asian ancestries, which we categorized as South (if selected Indian or Pakistani), East (if selected Chinese, Korean, or Japanese), or Southeast (if selected Cambodian, Filipino, Hmong, or Vietnamese) Asian ancestry. These 3 groups may partially overlap since participants were given the option to choose all ancestries that applied to them. Age was calculated as year of enrollment minus self-reported birth year (range 18-86 years), and was further categorized as 18 to 19, 20 to 29, 30 to 39, 40 to 49, 50 to 59, 60 to 69, and 70 to 86 years (ie, 10-year intervals for those aged 20-69 years; those aged 70-79 and 80-86 were aggregated due to sparse data for the 80-86 years category).

We evaluated self-reported clinician-diagnosed PCOS status (yes/no) from the Medical History Survey. Those who selected “I don’t know” or “prefer not to answer” for the list of gynecological conditions were considered missing. BMI was calculated from self-reported weight and height at baseline and further categorized as <18.5, 18.5 to 24.9, 25 to 29.9, and ≥30 kg/m² (32).

We also evaluated self-reported current hormone use indications. Specifically, participants reported if they had ever used hormones. Among those who had used hormones at least once, they may further provide their age at first use and last use. If the age at last use was within the same year of study enrollment, we considered it as current hormone use. If participants never used hormones or their age at last use was before the year of enrollment, they were grouped as not currently using hormones. Among those currently using hormones, participants could additionally report indications of use, which we categorized as for contraception, for managing menstrual cycles, for transgender medication (female to male), for 2 or more reasons, for other reason(s), or reason(s) not reported. We differentiated the indications because hormones typically used for female-to-male transgender medication (which usually include androgens) are fundamentally different from hormones used for contraception or managing cycles (which generally include progesterone and/or estrogen).

Statistical Analysis

We summarized the distributions of all signs of potential AE with mean ± SD and median (interquartile range) for continuous variables, or n (%) for categorical variables among all 24 435 participants. The 95% confidence intervals of the percentages for binary outcomes were further calculated as a measure of degree of uncertainty for the estimated population proportions. Then, to determine the heterogeneity of these symptoms by demographic and health characteristics, we evaluated the distributions of signs of potential AE, stratified by age groups, race and ethnicity, Asian ancestry, BMI categories, PCOS status, and current hormone use indications. Differences in the proportion of people with each symptom across categories of these factors were evaluated using Chi-square tests.

We conducted several sensitivity analyses. First, because not all hair loss is attributable to AE, we further explored the severity of hair loss on top of the head among individuals without other conditions that may cause hair loss (anemia, vitamin D deficiency, hyper- or hypothyroidism, anorexia, anxiety, or depression) (33, 34). Second, we evaluated the prevalence of possible hirsutism, stratified by hair compared to females in the family. Third, considering that the menopausal transition is often accompanied by substantial hormonal changes, we explored the distributions of the 4 key signs of potential AE, stratified simultaneously by age and self-reported menopausal status at baseline (menopause defined as those who either reported menopausal or reported a history of hysterectomy or bilateral oophorectomy). We also performed stratifications simultaneously by PCOS status and self-reported menstrual regularity (irregular cycles defined as those reported being “troubled by unpredictable periods,” cannot “usually predict the exact day when your next period will start,” or reported a typical menstrual cycle length of <21 days, ≥ 40 days, or “too irregular to estimate” from relevant survey questions). We also further evaluated the differences in signs of potential AE by subgroups of clinical relevance, stratified simultaneously by both PCOS diagnosis and age and also by BMI and age.

Data processing and statistical analyses were conducted in Python (version 3.6) and R (version 4.1.2). All statistical tests were 2-sided. Results with P-values < .05 were considered statistically significant and were presented in the results.

Results

Participants baseline characteristics are shown in Table 1. Of the 24 435 participants, the average age was 35.8 years (interquartile range: 27-43 years; range: 18-86 years). The majority of participants identified as non-Hispanic White (75%), followed by multiple races (9%), Hispanic (6%), non-Hispanic Black (5%), Asian (3%), and other races (2%). Among those who further provided Asian ancestry data, 238 participants self-identified as South Asian, 531 as East Asian, and 282 as Southeast Asian. Nearly all participants identified as woman (97%). Of note, participants could select all that apply when describing their gender identity, and the distribution does not exactly match with reported sex assigned at birth. The average BMI was 29.3 kg/m², and 3% of participants were <18.5 kg/m², 34% were 18.5 to 24.9 kg/m², 26% were 25 to 29.9 kg/m², and 38% were ≥30 kg/m² (18% with BMI of 30-<35, 11% with BMI of 35-<40, and 10% with BMI of 40 or higher). A clinician diagnosis of PCOS was reported by 12% of participants, and 7% reported a family history of PCOS. There were 20% of participants with current hormone use (for 2 or more reasons, contraception, or managing menstrual cycles were the commonly reported reasons).

Table 1.

Characteristics of the 24 435 AWHS participants

Characteristicsa	Distributions
Age at enrollment, years
Mean ± SD	35.8 ± 11.4
Median (IQR)	34 (27-43)
Age at enrollment, n (%)
18-19	983 (4.0)
20-29	7090 (29.0)
30-39	8015 (32.8)
40-49	5428 (22.2)
50-59	2026 (8.3)
60-69	668 (2.7)
70-79	214 (.9)
80-86	11 (.0)
Race and ethnicity, n (%)
Non-Hispanic White	18 233 (74.8)
Non-Hispanic Black	1169 (4.8)
Asian	695 (2.9)
Hispanic	1521 (6.2)
Multiple races	2273 (9.3)
Other races	470 (1.9)
Asian ancestry: South Asianb	238 (1.0)
Asian ancestry: East Asianb	531 (2.2)
Asian ancestry: Southeast Asianb	282 (1.2)
Self-identified gender,c n (%)
Woman	23 705 (97.0)
Man	39 (.2)
Trans Woman	10 (.0)
Trans Man	73 (.3)
Genderqueer/Nonbinary	643 (2.6)
Another gender identity	89 (.4)
Prefer not to answer	70 (.3)
BMI categories, n (%)
<18.5 kg/m2	597 (2.5)
18.5-24.9 kg/m2	8034 (33.6)
25-29.9 kg/m2	6150 (25.7)
30 kg/m2 or above	9149 (38.2)
Self-reported PCOS diagnosis, n (%)	2921 (12.0)
Family history of PCOS, n (%)	1759 (7.3)
Current hormone use by indication, n (%)
Not currently using hormones	18 845 (80.1)
Yes, for contraception	359 (1.5)
Yes, for managing menstrual cycles	273 (1.2)
Yes, for transgender medication	12 (.1)
Yes, for 2 or more reasons	2628 (11.2)
Yes, for other reason(s)	128 (.5)
Yes, reason(s) not reported	1294 (5.5)
Thick, coarse, and dark hair on the upper lip, n (%)	4442 (18.3)
Thick, coarse, and dark hair on the chin, n (%)	6603 (27.2)
Thick, coarse, and dark hair on the breasts, n (%)	3661 (15.1)
Thick, coarse, and dark hair on the chest between the breasts, n (%)	1065 (4.4)
Thick, coarse, and dark hair on the back, n (%)	427 (1.8)
Thick, coarse, and dark hair on the abdomen, n (%)	2844 (11.7)
Thick, coarse, and dark hair on the upper arms, n (%)	922 (3.8)
Thick, coarse, and dark hair on the upper thighs, n (%)	2652 (10.9)
Summary score of thick, coarse, and dark hair for the 8 locations
0 location, n (%)	14 218 (58.2)
1 location, n (%)	3888 (15.9)
2 locations, n (%)	2892 (11.8)
3 locations, n (%)	1616 (6.6)
4 locations, n (%)	896 (3.7)
5 locations, n (%)	460 (1.9)
6 locations, n (%)	179 (.7)
7 locations, n (%)	78 (.3)
8 locations, n (%)	74 (.3)
Median (IQR)	0 (0-2)
5th-95th percentile	0-4
Possible hirsutism (summary score ≥ 4)	1687 (6.9)
Hair on the body compared to females in the family, n (%)
More hair	4263 (19.3)
About the same	14 365 (64.9)
Less hair	3500 (15.8)
Hair level on the upper lip, n (%)
None	15 189 (62.5)
A few	6939 (28.5)
Several	1848 (7.6)
A lot	339 (1.4)
Hair level on the chin, n (%)
None	13 936 (57.2)
A few	7347 (30.2)
Several	2230 (9.2)
A lot	835 (3.4)
Hair loss on temples or sides of forehead, n (%)	2934 (12.1)
Hair loss on top of head, n (%)	1961 (8.1)
Hair loss on hairline, n (%)	2671 (11.0)
Hair loss on random, small patches across the head, n (%)	919 (3.8)
Hair loss across whole head, n (%)	2806 (11.5)
Hair loss compared to female peers, n (%)
Less hair loss	2817 (18.7)
About the same	8718 (55.9)
More hair loss	3948 (25.3)
Thickness of hair on top of head, n (%)
Thick	17 792 (73.6)
Thin hair with widening part	4394 (18.2)
Thin hair with widening part and some scalp showing	1572 (6.5)
Significantly reduced hair with more scalp showing	305 (1.3)
Scalp mostly visible	101 (.4)
Experience acne as an adult, n (%)	16 315 (67.7)
Acne compared to female peers, n (%)
Less acne	6557 (32.4)
About the same	7778 (38.4)
More acne	5929 (29.3)
Small, raised pimples on the face or back, n (%)	11 325 (46.5)
Red, irritated pimples on the face or back, n (%)	5596 (23.0)
Pimples that have pus on the face or back, n (%)	5382 (22.1)
Scars from acne on the face or back, n (%)	5706 (23.4)
Moderate-to-severe acne (having red irritated pimples and/or pimples that have pus), n (%)	7750 (31.8%)

Abbreviations: AWHS, Apple Women's Health Study; BMI, body mass index; IQR, interquartile range; PCOS, polycystic ovary syndrome.

^a Numbers may not add up to the total number due to missingness.

^b Numbers may add up more than the number of Asians since participants were given the option to select more than 1 Asian ancestry.

^c Numbers may add up more than the total number since participants were given the option to select more than 1 gender identity. The survey question was “Which of the following best describes your gender identity? (select all that apply).”

Overall, 7% of participants had potential hirsutism, as reflected by a summary score of thick, coarse, and dark hair on ≥4 locations (Table 1). Around 13% of participants reported having several or a lot of hair on their chin. There were 8% of participants who reported having thin hair with widening part and some scalp showing, significantly reduced hair with more scalp showing, or scalp mostly visible, and 32% of participants had moderate to severe acne. Distributions of other secondary measures of potential AE are shown in Table 1.

Signs of Potential Hirsutism, Variations by Demographic and Health Characteristics

As age increased, the prevalence of possible hirsutism decreased (Fig. 1, 8% among those aged 18-19, compared to 1% among those aged 70-86), but hair level on the chin increased (Fig. 2, 7% reported several or a lot of hair on the chin among those aged 18-19, compared to 14% among those aged 70-89). Possible hirsutism was most prevalent among Hispanic individuals (11%, compared to the lowest prevalence of 6% among non-Hispanic White individuals) and among those who reported South Asian ancestry (17%) (Fig. 1). The prevalence of having several/a lot of hair on the chin was highest among non-Hispanic Black individuals (16%), followed by non-Hispanic White (13%) individuals, and was lowest among Asian individuals (9%), but those who reported South Asian ancestry still had a much higher prevalence of several/a lot of hair on the chin (20%) (Fig. 2). Both possible hirsutism and hair level on the chin increased with higher BMI categories and were much more prevalent among those with PCOS compared to those without PCOS (Fig. 1 and Fig. 2). Individuals who currently used hormones for contraception purposes had a lower prevalence of possible hirsutism (5%) compared to those not using hormones (7%), as well as a lower prevalence of several/a lot of hair on the chin (9% vs 13%). Percentage values are shown in Supplemental Tables S2 and S3 (24).

Figure 1.

Possible hirsutism, stratified by sociodemographic and health characteristics. (A) By age; (B) by race and ethnicity; (C) by Asian ancestry; (D) by body mass index; (E) by PCOS; (F) by hormone use. Numeric labels correspond to percentages. Error bars represent 95% confidence intervals. P-values from chi-square tests <.001 for all 6 panels.

Abbreviations: BMI, body mass index; PCOS, polycystic ovary syndrome.

Figure 2.

Hair levels on the chin, stratified by sociodemographic and health characteristics. (A) By age; (B) by race and ethnicity; (C) by Asian ancestry; (D) by BMI; (E) by PCOS; (F) by hormone use. Numeric labels correspond to percentages. P-values from chi-square tests <.001 for all 6 panels.

Abbreviations: BMI, body mass index; PCOS, polycystic ovary syndrome.

Severity of Hair Loss, Variations by Demographic and Health Characteristics

Hair loss severity on top of the head increased as age increased (Fig. 3, from 7% reporting the top 3 severity levels among those aged 18-19 to 22% among those aged 70-86). Hair loss on top of the head was most severe among Asian participants (Fig. 3, 13% reporting the top 3 severity levels, compared to the lowest prevalence of 8% among those who identified multiple races), and among Asians, those who reported South Asian ancestry had the most severe hair loss (15%). Hair loss severity increased as BMI increased although in smaller magnitudes (Fig. 3, from 6% reporting the top 3 severity levels among those with BMI <18.5 kg/m² to 9% among those with BMI ≥30 kg/m²). Compared to individuals without PCOS, those with PCOS reported higher severity of hair loss (Fig. 3). Individuals who used hormones for contraception or managing menstrual cycles had less severe hair loss compared to nonusers (Fig. 3). Percentage values are shown in Supplemental Table S4 (24).

Figure 3.

Hair loss severity on top of the head, stratified by sociodemographic and health characteristics. (A) By age; (B) by race and ethnicity; (C) by Asian ancestry; (D) by BMI; (E) by PCOS; (F) by hormone use. Numeric labels correspond to percentages. P-values from chi-square tests <.001 for all 6 panels.

Abbreviations: BMI, body mass index; PCOS, polycystic ovary syndrome.

Moderate to Severe Acne, Variations by Demographic and Health Characteristics

As age increased, moderate to severe acne substantially decreased (Fig. 4, from 44% among those aged 18-19 to 4% among those aged 70-86). Moderate to severe acne was most prevalent among non-Hispanic White individuals (Fig. 4, 33%, compared to the lowest prevalence of 19% among non-Hispanic Black individuals) and was similar across Asian ancestries. There were no significant differences in moderate to severe acne across BMI groups; only those who were underweight suggested a moderately higher prevalence (36%, compared to 32% among those with BMI 18.5-24.9 kg/m²). Moderate to severe acne was more prevalent among individuals with PCOS and among those using hormones for transgender medication (Fig. 4). Percentage values are shown in Supplemental Table S5 (24).

Figure 4.

Moderate to severe acne, stratified by sociodemographic and health characteristics. (A) By age; (B) by race and ethnicity; (C) by Asian ancestry; (D) by BMI; (E) by PCOS; (F) by hormone use. Numeric labels correspond to percentages. P-values from chi-square tests <.001 for variability by age, race and ethnicity, PCOS status, or hormone use. P-value = .942 for variability by Asian ancestry. P-value = .060 for variability by BMI.

Abbreviations: BMI, body mass index; PCOS, polycystic ovary syndrome.

Secondary and Sensitivity Analyses

Results from secondary and sensitivity analyses are shown in the supplemental materials (24). In detail, the mean summary score of hair on the body showed similar variability across demographic/health characteristics (Supplemental Fig. S2) as possible hirsutism (24). Hair level on the upper lip was most commonly reported among Hispanic or South Asian individuals, those with higher BMI, or those with PCOS (Supplemental Fig. S3) (24). As age increased, hair on the abdomen, back, chest, upper arms, and thighs decreased, while hair on the chin and upper lip became more prevalent (Supplemental Table S6) (24). The prevalence of having more hair on the body compared to females in the family decreased with age and was higher among those of South Asian ancestry, with BMI ≥30 kg/m², or with PCOS (Supplemental Fig. S4) (24). The prevalence of having possible hirsutism was much higher among those having more hair than females in the family, compared to those with similar or less hair than females in the family (Supplemental Table S7) (24). As age increased, hair loss on the top of the head increased, while hair loss as random, small patches across the head decreased (Supplemental Table S8) (24); all 4 characterizations of acne decreased with age (Supplemental Table S9) (24). Individuals with PCOS tended to report having more acne compared to female peers (Supplemental Table S10) (24). Those with other preexisting conditions that may cause hair loss had a higher prevalence of signs of potential alopecia compared to those without any of these conditions (Supplemental Table S11) (24); however, even among participants without any of these conditions, the relative variability pattern by demographic/health characteristics (Supplemental Fig. S5) remained similar to those seen in Fig. 3 (24). More hair loss compared to female peers was more prevalent among Hispanic participants or those with BMI <18.5 kg/m² (Supplemental Fig. S6) (24). When further stratified by menopausal status, those who were menopausal had a lower prevalence of possible hirsutism (Supplemental Fig. S7) (24). Among individuals without PCOS, those with irregular cycles were more likely to exhibit signs of potential AE than those with regular cycles (Supplemental Fig. S8) (24). Among individuals with PCOS, participants of older age tended to report less hirsutism on most body locations (Supplemental Fig. S9) (24). The differences in hair loss and acne over age by PCOS diagnosis or BMI categories were small (Supplemental Figs. S10-S14, Tables S12-S13) (24).

Discussion

This study investigated the distribution of signs relevant to potential AE in a population-based cohort. We found notable variation in possible hirsutism, hair loss, and moderate to severe acne by age, race and ethnicity, BMI, PCOS status, and hormone use. We created a conceptual model that illustrates factors impacting potential AE and its symptoms (Supplemental Fig. S15) (24). To our knowledge, this is the first large-scale cohort study to evaluate the prevalence of self-assessed symptoms relevant to 3 major characteristics of potential AE at the population level. Our survey also incorporated self-assessment of potential AE symptoms compared to female family members. While there is a strong familial component to hirsutism (35-37), having more body hair than other female relatives may point toward a divergence from hereditary patterns and indicate potential hirsutism. While studies have examined AE in relation to PCOS (17) and some specifically evaluate how certain AE characteristics (eg, mFG score) differ by race and ethnicity among individuals with or without PCOS (8, 9, 38), there is a lack of large-scale population-based studies evaluating composite signs of potential AE and whether they differ by demographic or health factors (16, 39). Our findings also suggested that the use of the ≥95th percentile of self-assessed symptoms as an indicator of possible hirsutism could potentially be applied to other similar cohorts as an alternative means to identify the potential risk of hirsutism at the population level. Understanding these patterns using self-reported data from a large cohort within the United States may help inform future work to develop more inclusive, population-wide evaluation tools.

Potential AE Variation by Age

Signs of potential hirsutism varied by age and by location of hair growth. For example, reports of overall possible hirsutism decreased with age, while hair growth on the chin and upper lip increased with age. There is no clear explanation for these trends as self-reported androgen excess and aging have not been examined in depth in the current body of literature. One study found an overall decline in serum androgen hormone levels [dehydroepiandrosterone sulfate (DHEAS), testosterone, and androstenedione] with increasing age and identified the steepest decline as occurring in the early reproductive years (4). Another study found a decrease in DHEAS levels with age, regardless of PCOS status (17). As a result, it is generally believed that signs of AE tend to resolve at the time of menopause, particularly for women with PCOS (3), although there is a growing body of literature studying signs of hyperandrogenism in postmenopausal women (7). Of note, our analysis is cross-sectional and does not evaluate individual-level change over time, but our findings of varying signs of potential AE by age and menopausal status provide a large-scale evaluation and may inform future, longitudinal research.

Potential AE Variation by Race and Ethnicity

Hispanic participants tend to have the highest prevalence of possible hirsutism. Asian participants had the highest prevalence of alopecia. Non-Hispanic White participants and participants who identified more than 1 race had the highest prevalence of moderate to severe acne. Within Asians, participants with South Asian ancestry had the highest prevalence of possible hirsutism and alopecia. This is consistent with a systematic review that assessed variations in clinical presentation in patients with PCOS by race that found that, compared to White women with PCOS, East Asian women with PCOS were less hirsute, while Hispanic, South Asian, and Middle Eastern women with PCOS were more hirsute (8). Another observational study of 395 participants determined the prevalence of adrenal androgen excess using age- and race-specific normative values (17) and found that DHEAS levels were significantly lower in Black women without PCOS compared to White women without PCOS. However, in our data, we found that non-Hispanic Black participants had the highest hair levels on the chin, indicative of potential hirsutism. Kazemi et al sought to find the area of hirsutism that correlated most closely between patient self-report and mFG score, a commonly used composite score to clinically evaluate hirsutism (25), and, consistent with other studies (28, 30), they found that the chin was the most concordant site (29).

Our evaluation of terminal hair compared to female relatives within each self-identified race and ethnicity and/or ancestry group may also add new information to the ongoing controversy about whether different mFG score cutoffs need to be implemented for different racial/ethnic groups. Here in our secondary analysis, we found that within each race, ethnicity, and/or ancestry group, there remained a sizable proportion of individuals (ranging between 17% and 36%) having more hair than females in their family. This variable was also strongly correlated with the “possible hirsutism” variable in this cohort. When resources are limited to conducting a full-scale evaluation (such as calculating an mFG score), a simple comparison of hair levels to female relatives may serve as an alternative self-evaluation tool, although further research and validations are needed.

Overall, our study differs from the existing studies as it includes a much larger sample size, represents more racial and ethnic diversity, and uses self-assessment of symptoms as well as a comparison of hirsutism to female family members as measures of potential AE. As the mFG score was developed from a sample of 430 White women in England (25), we hope our findings may provide new insight into the population-level distributions of possible AE among US individuals with heterogeneous demographic and health backgrounds and ultimately inform future research on developing inclusive assessment tools of clinical signs of potential androgen excess across the US population.

Potential AE Variation by BMI

Studies have found an association between increased BMI and AE (15, 40). Our results supported this with possible hirsutism and alopecia, which was most prevalent among individuals with BMI ≥ 30 kg/m² and least prevalent in individuals with BMI < 18.5 or 18.5 to 24.9 kg/m². However, we found that experiencing moderate to severe acne was most common among individuals who were underweight. This remained consistent across all characterizations of acne, where it was most prevalent among participants who were underweight and decreased with increasing BMI. This finding is consistent with Yang et al, who conducted a comparative study on cutaneous manifestations of hyperandrogenism and found that women who were obese reported acne less frequently than women who were nonobese (41). Although there was a hypothesis that increasing estrogen production from adipose tissues may contribute to alleviating acne in individuals with obesity, further studies on other mechanisms are needed.

Potential AE Variation by PCOS Status and Hormone Use

As AE is one of the diagnostic criteria of PCOS (clinical or biochemical hyperandrogenism, oligo and/or anovulation, and polycystic ovaries on transvaginal or transabdominal ultrasound) (42, 43), signs of potential AE were indeed more prevalent in those with PCOS. AE in those with PCOS is likely due to increased ovarian androgen production by follicular theca cells in the reproductive years (3, 44). Our results were consistent, as possible hirsutism, alopecia, and acne were all more prevalent in participants with PCOS compared to those without. Signs of potential AE also varied by hormone use indication. Hormone use for contraception was associated with a lower prevalence of possible hirsutism, hair on chin, and hair loss, which is in line with the fact that oral contraceptives are often the first pharmacologic therapy for hirsutism (45). Hormone use for transgender medication was associated with more moderate to severe acne compared to those not using hormones, consistent with studies that have found a correlation between testosterone therapy initiation and worsening acne among transgender men (46, 47). Given the close association between AE and PCOS, by investigating how AE presents in different populations, we hope to be able to identify those with underlying PCOS or at risk of PCOS in diverse populations with higher sensitivity and specificity. A goal for future research is creating better algorithms for earlier diagnosis of PCOS, which would allow for prompt management of PCOS-related comorbidities.

Strengths and Limitations

This study fills the gap for self-assessed signs of potential AE at a general population level in the United States. Our study also uniquely asks participants to evaluate their symptoms relative to female relatives or peers. While most studies on AE use biochemical measurements of androgen levels, which often requires a population with access to healthcare, higher socioeconomic status, and higher health literacy, our study utilizes participant-assessed data and graphical illustrations that might capture a broader population. Phenotyping of AE to aid the diagnosis of PCOS has been primarily based on findings from populations referred to care, where the study findings may be subject to referral bias when referral was usually heavily influenced by severity of symptoms, awareness of PCOS, socioeconomic status, and access to care (48). Understanding the patterns in symptoms that can be used to evaluate potential AE in a population-based cohort of individuals unselected by PCOS status or other health conditions may allow for more accurate representations of self-evaluated potential AE in the general population, including those with undiagnosed PCOS. Our study also includes the largest sample size to date for studying AE-related symptoms, which enhances the statistical power of our results. We believe our study makes an important contribution to the existing literature by adding to the limited understanding of population-level characteristics that may be associated with potential androgen excess—an analysis that has not previously been conducted for the demographically heterogeneous US population on a larger scale.

Our study is not without limitations. First and beyond the scope of this study, we did not determine the etiology of participants' symptoms. Second, PCOS diagnosis was ascertained by self-report and not confirmed with medical record verification. This could lead to either an over- or underrepresentation of the true proportion of participants with PCOS. Third, given the self-report nature of our data as opposed to a clinician assessment, the responses are subject to an individual's interpretation of the survey. For example, 1 person's definition of “coarse hair” could differ from another's. Furthermore, some survey questions only allowed for a binary “yes/no” response, such as for “dark, coarse body hair,” which does not allow for a gradation of symptoms or the ability to derive a mFG score from the data. Fourth, we do not have data on depilation and or its frequency for participants' female family members, so the question of hair compared to family members may be subject to potential misclassification. Fifth, while some survey questions, such as those evaluating hair loss severity on top of the head, were previously validated (49-51), other survey questions were not validated. However, the development of comprehendible self-evaluation tools at the population level holds the potential for mitigating potential disparities of access to healthcare. Last, our study population included a larger proportion of non-Hispanic White participants compared to the general US population and thus may limit our results' generalizability.

Summary

In this large, population-based cohort study, we estimated the prevalence of self-reported signs of potential AE and further identified variations of these symptoms by age, race and ethnicity, BMI, PCOS status, and hormone use. While further research is needed to identify factors contributing to these variations, our findings could provide new insights into the prevalence of potential AE at the population level and how it may appear differently in diverse groups, informing future work to develop more inclusive, population-wide evaluation tools.

Contributors

A.T.W. and Z.W. (contributed equally to this work) were responsible for the study design, data preparation, statistical analyses, interpretation of the results, and manuscript writing. S.M. oversaw all aspects of the work and contributed to developing the concept, analysis plan, interpretation of the results, manuscript writing, and revision. A.M.J. and D.D.B. provided critical reviews of the analysis and contributed to the interpretation of the results and manuscript revision. B.A.C., R.H., J.P.O., and M.A.W. contributed to the study design, interpretation of the results, manuscript writing, and revision. C.L.C. and T.F.C. contributed to review of the content in relation to the publication policy of Apple Inc. and did not participate in the analysis and interpretation of data. All the authors critically discussed the findings and provided feedback for the final draft. All authors read and approved the final manuscript.

Funding

This study received funding from Apple Inc. The funding source provided platforms and software for the collection and management of the data and participated in the review and approval of the manuscript. It played no role in the design and conduct of the study, analysis and interpretation of the data, preparation of the manuscript, or the decision to submit the manuscript for publication. This research was supported in part by the Intramural Research Program of the National Institutes of Health under award number Z01ES103333. Support for A.M.J. and D.D.B. was provided by the Intramural Research Program of the National Institute of Environmental Health Sciences, National Institute of Health.

Disclosures

C.L.C. and T.F.C. own Apple Inc. stock and are employed by Apple Inc. Other co-authors have no conflict of interest. There were no financial relationships with any organizations that might have an interest in the submitted work in the previous 3 years and no other relationships or activities that could appear to have influenced the submitted work.

Informed Patient Consent for Publication

All participants provided informed consent at enrollment.

Ethics Approval

This study was approved by the Institutional Review Board (IRB) at Advarra (CIRB #PRO00037562) and Harvard University (IRB #20-0384).

Prior Presentation

This work was presented in October 2023 at the Androgen Excess and PCOS Society Annual Meeting in Rotterdam, Netherlands, as a poster presentation.

Data Availability

Aggregated data that support the findings of this study may be available upon reasonable request from the corresponding author and senior author. Any request for data will be evaluated and responded to in a manner consistent with policies intended to protect participant confidentiality and language in the study protocol and informed consent form.