Alopecia areata in underrepresented groups: preliminary analysis of the all of us research program

Alopecia areata (AA) is an autoimmune condition characterized by patchy, nonscarring hair loss. Few studies of AA have adequately included participants from underrepresented groups when evaluating the burden of AA in the United States. We conducted a cross-sectional study of personal/demographic factors and AA using the ongoing All of Us (AoU) Research Program. AoU enrolls adults over 18 years either as direct volunteers or through participating Health Care Provider Organizations by prioritizing recruiting underrepresented groups. We linked data from surveys and electronic health records (EHRs) to estimate the prevalence of AA by race/ethnicity, physical disability, sexual orientation/gender identity (LGBTQIA +), income, and education. The latest AoU release (version 5) includes 329,038 participants. Average age was 51.8 years (standard deviation, SD 16.7), and 60.2% of participants were female. Of these, 251,597 (76.5%) had EHR data and 752 were diagnosed with AA (prevalence, 0.30%; 95% CI 0.28–0.32). We used multivariate logistic regression adjusted for age and other factors to estimate the odds ratio (OR) and 95% confidence intervals (CIs) for prevalence of AA. Compared to Whites, Blacks and Hispanics had higher odds of AA (OR, 1.72; 95% CI 1.39–2.11 and OR, 2.13; 95% CI 1.74–2.59, respectively). Lower odds of AA were observed in participants with less than a high school degree (OR, 0.80; 95% CI 0.59–1.08), household income ≤ $35,000 (OR, 0.67; 95% CI 0.54–0.83), and no health insurance (OR 0.35; 95% CI 0.20–0.56). In this diverse population of US adults, participants with skin of color had higher prevalence of AA. Lower prevalence of AA among individuals with lower education and income levels and those lacking health insurance may reflect limited access to dermatologic care and potentially higher levels of undiagnosed AA in these groups.


Introduction
Alopecia areata (AA) is a T cell-mediated autoimmune disorder targeting the hair follicles that usually results in patchy, nonscarring hair loss [1]. AA is divided into subtypes based on disease severity, which ranges from a single patch of hair loss to total loss of scalp hair (alopecia totalis [AT]) or total scalp and body hair loss (alopecia universalis [AU]). AArelated hair loss may be an isolated incident, recurrent, or persistent [1]. The majority of small, solitary AA patches resolve spontaneously; however, widespread and recurrent cases often fail to respond to treatment, and patients experience significant physical and psychologic burden of the disease. Moreover, when compared to patients with other skin conditions, AA patients report lower health-related quality of life, emphasizing the need for additional research [2].
Accumulating evidence suggests that underrepresented groups defined by race and ethnicity may experience a disproportionate burden of AA [3][4][5]. Underrepresented groups defined by features other than race (i.e., gender identity, sexual orientation, age, physical disability) may also be at increased risk of AA; however, to our knowledge, these potential disparities have never been evaluated. The purpose of the current study was to determine potential disparities of AA in underrepresented groups using the All of Us (AoU) database.

Study design
The AoU Research Program protocol has been published elsewhere [6]. The AoU protocol and materials have been approved by the AoU Institutional Review Board (IRB). In short, AoU represents a precision medicine initiative to build a database of at least 1 million participants that reflects the diversity of the United States (US), with special attention to marginalized groups who have historically been left out of health research. Inclusion criteria include age over 18 years, capacity to consent, and current residence in the US or a US territory. The only exclusion criterion is being a prisoner at the time of enrollment. All study participants gave written informed consent. We used the most recent release dataset (version 5), which includes updated data from all participants who enrolled from May 30, 2017 (program inception) to April 1, 2020.

Baseline data
Enrolled participants complete a baseline health survey on socio-demographics, overall health, lifestyle, and health care access/utilization at a direct volunteer or participating health care provider organization (HPO) site. As of December 2021, there are participating HPOs in 20 states, including Alabama, Arizona, California, Connecticut, Florida, Georgia, Hawaii, Illinois, Louisiana, Massachusetts, Michigan, Minnesota, Mississippi, New York, Pennsylvania, South Carolina, Tennessee, Texas, Virginia, and Wisconsin. The National Direct Volunteer Program allows those who do not have access to participating HPOs to enroll online or at additional community-based enrollment sites across the country. Participants are also asked to authorize sharing of data from their EHR if available to AoU. Longitudinal data collection is performed via continuous extraction of EHR data, including demographics, clinic visits, disease diagnoses, procedures, medications, laboratory examinations, vital signs, and physician notes.
The AoU database provides definitions of underrepresented groups used in AoU research, which were adopted for the purposes of this study (see Figure S1). Baseline survey data were used to ascertain self-reported race/ethnicity in the following categories: Hispanic/Latino/a/x participants (regardless of race), non-Hispanic White, non-Hispanic Black, non-Hispanic Asian, another single population, none of these, and non-Hispanic > 1 race. We combined non-Hispanic Asian, another single population, none of these, and non-Hispanic > 1 race into a single "Other race" category due to the modest number of participants with AA in the individual categories. Other underrepresented groups were identified, defined by gender identity/ sexual orientation (LGBTQIA +), age (> 75 years), educational attainment (less than a high school degree), and annual household income (≤ $35,000). Participants with physical disabilities were identified as those who answered, "A Little" or "Not At All" to the question: "To what extent are you able to carry out your everyday physical activities such as walking, climbing stairs, carrying groceries, or moving a chair?".

Ascertainment of AA
This analysis utilized the Systematized Nomenclature of Medicine (SNOMED) to identify participants with AA (concept ID 141933, SNOMED code 68225006). SNOMED connects the terminology, medical codes, synonyms, and definitions used among all the different EHR systems. For example, if one EHR system uses ICD9 codes while another uses ICD10 codes, SNOMED allows the same data point from the two systems to be matched. SNOMED codes were also utilized to identify patients with AU (concept ID 4312756, SNOMED code 86166000) and AT (concept ID 4056343, SNOMED code 19754005). AU was categorized as a subset of AA, while AT was categorized separately as a subset of alopecia. Patients with ophiasic alopecia areata were also included to AA (concept ID 4239312, SNOMED code 5860009).

Statistical analysis
We estimated AA prevalence as the number of participants with AA divided by the total number of participants with EHR data available and calculated the 95% confidence interval (CI) for the proportion. Age-adjusted and multivariate logistic regression were used to estimate the odds ratios (ORs) of an AA. Multivariate analyses adjusted for age, race/ethnicity, gender identity, sexual orientation, household income, educational attainment, physical disability, health insurance status, and history of autoimmune disease comorbidities simultaneously. A 2-tailed P value of < 0.05 was considered statistically significant. This same procedure was repeated for both AU and AT individually and then for AU and AT combined. Analyses were conducted using R version 4.0.5 in the Jupyter Notebook environment.

Results
The current release of AoU includes data from 329,038 participants. Of these, 251,597 (76.5%) had EHR data. Average age at consent was 51.8 years, and 197,949 (60.2%) were women ( Table 1). The overall prevalence of AA was 0.30%; (n = 752; 95% CI 0.28-0.32) among those with EHR. Of these, 25 (3.3%) had AU. There were 35 participants with AT, giving it an overall prevalence of 0.01% (95% CI 0.01-0.02). The AT/AU cohorts were combined for analysis, but the small sample size yielded insufficient power to detect any significant associations with demographic factors.

Discussion
This cross-sectional study evaluated disparities in the prevalence of AA in a racial/ethnically diverse population. A higher burden of AA was observed among skin of color. Additionally, participants with lower education and household income and no health insurance were less likely to be diagnosed with AA.
The higher prevalence of AA among skin of color participants demonstrated in our study is consistent with findings of other US studies [3][4][5]. Prior studies have found increased odds of AA among Blacks and Hispanics, consistent with our findings [3][4][5]. For example, an analysis of 14,400 visits for AA from the National Ambulatory Medical Care Survey between 2007 and 2016 showed increased likelihood of non-White individuals visiting for AA compared to White individuals (OR, 2.4; 95% CI 1.2-5.1; P = 0.02) [5].
Studies on the overall prevalence of AA are scarce, and estimates vary widely. The most frequently referenced study is the 1971-1974 First National Health and Nutrition Examination Survey, which estimated the period prevalence of AA in the US to be between 0.1% and 0.2% [7]. However, two recent studies in Greece and Japan reported period prevalence of 1.27% and 2.45%, respectively [8,9]. Recently in 2020, a large cross-sectional online survey of the prevalence of AA in the US reported a clinician-adjudicated point prevalence of 0.21% (95% CI 0.17%-0.25%), while we report an overall prevalence of 0.30% (95% CI 0.28-0.32) [10]. The increased prevalence of AA in individuals with skin of color and the high level of diversity in the AoU database may partly explain that our estimated prevalence of AA is slightly higher than that reported by other US-based studies, such as the aforementioned online survey, which included 77.1% white and 13.3% black participants [10].
The underlying cause of the increased prevalence of AA in skin of color has yet to be elucidated. Other immunemediated diseases known to be associated with AA, such as systemic lupus erythematosus (SLE), disproportionately affect Blacks and Hispanics and may have shared aspects of pathophysiology with AA [11]. Cohort studies evaluating racial disparities in SLE have identified specific genetic susceptibility loci that differ by racial or ethnic background. For example, higher IFN-α activity has been observed among Black patients with SLE [12]. Notably, several case reports have reported AA as a side effect of IFN-α therapy, and expression of type 1 interferon-related proteins in inflammatory lesions of AA has been demonstrated [12].
Similar to prior studies, we found a higher prevalence of AA in females compared to males. Two studies evaluating incidence of AA in the US reported similar higher incidence in women than in men [13,14]. However, a crosssectional analysis of an estimated 2.6 million AA outpatient visits from the National Ambulatory Medical Care Survey and National Hospital Ambulatory Medical Care Survey revealed that males made fewer clinic visits per 1,000 capita compared to females [15]. Additionally, in a US study of 481 Caucasian patients from AA patient conferences, there was a female predominance of AA (female to male ratio 2.3:1) [16], and a Turkish study reported that 53.4% of patients diagnosed as having AA, AT, or AU were female and 46.6% were male. [17] Associations of AA with LGBTQIA + status, physical disability, socioeconomic status, education level, or health insurance status have not been evaluated in prior epidemiological studies. While LGBTQIA + status and AA and physical disability were not associated with AA, we found LGBTQIA + , lesbian, gay, bisexual, transgender, queer, intersex, and asexual a The "other" category comprises the following categories from All of Us questionnaires: Another single population: participants self-reporting either Middle Eastern or North African or Native Hawaiian or other Pacific Islander (please note All of Us does not provide disaggregated data on these yet). None of these populations: participants self-reporting "None of these fully describe me" (options are White, Black, African American, or African, Asian, Middle Eastern or North African, Native Hawaiian or other Pacific Islander). > 1, non-Hispanic > 1 race selected b Income corresponds to annual household income c Disability indicates physical disability (participants who answered that they cannot carry out every day physical activities at all or only a little) decreased odds of AA in participants without health insurance and those with lower education and lower income. These findings may indicate limited access to dermatologic care in these populations. As a result, AA may be underdiagnosed in these populations, leading to increased morbidity and decreased health-related quality of life [18]. This is supported by the increased prevalence of AA in the AoU database, which is focused on the inclusion of underrepresented populations, compared to other AA prevalence studies [10]. As such, our findings reinforce the theorized lack of access to dermatologic care among populations at lower socioeconomic status and emphasize the need for improved healthcare access. [19] When participants lacking health insurance were excluded from analysis, all associations stayed similarly, indicating that lack of health insurance does not account for the decreased likelihood of AA diagnosis in participants with less education and lower annual income. Rather, these factors appear to be independently associated with decreased odds of being diagnosed with AA, pointing to underdiagnosis and lack of access to dermatologic care among these vulnerable populations.
Our study had several strengths. This study benefits from the diversity of the AoU database and its commitment to including individuals from underrepresented populations. As a result, we were able to evaluate the burden Table 2 Odds ratios of alopecia areata prevalence with demographic factors a Multivariate model adjusts for race, ethnicity, age, sex, household income, education, physical disability, and health insurance status simultaneously b The "other" category comprises the following categories from all of us questionnaires: another single population: participants self-reporting either middle eastern or north african or native hawaiian or other pacific islander (please note all of us does not provide disaggregated data on these yet). None of these populations: participants self-reporting "none of these fully describe me" (options are white, black, african american, or african, asian, middle eastern or north african, native hawaiian or other pacific islander). > 1, non-hispanic > 1 race selected c LGBTQIA + indicates lesbian, gay, bisexual, transgender, queer, intersex, and asexual; or, odds ratio d Income corresponds to annual household income e Disability indicates physical disability (participants who answered that they cannot carry out every day physical activities at all or only a little) of AA in a national cohort that reflects the diversity of the US population. We evaluated a large sample size of 251,597 participants with available EHR data and were able to adjust for numerous demographic factors that have been suggested as risk factors of AA risk in underrepresented racial and ethnic groups, such as socioeconomic status, educational attainment, physical disability, and health insurance status [4]. We also adjusted for co-occurrence of other autoimmune skin conditions and thyrotoxicosis, which are known comorbidities of AA [4]. Since our analysis takes notable potential confounders into account, our study is one of the first to demonstrate that there is an increased burden of AA in underrepresented racial and ethnic groups, regardless of socioeconomic differences.
Our study was subject to the limitations of a cross-sectional design. Additionally, opt-in databases such as AOU are subject to selection bias. Another limitation of our study was the small sample size of participants diagnosed with AT and AU, so we were unable to determine whether underrepresented groups experience more severe AA subtypes compared to the general population. This potential relationship merits evaluation in future studies with a larger number of AT and AU participants. Finally, we were not able to evaluate age at diagnosis of AA because AoU data provides only age at the time of enrollment.
In conclusion, our findings regarding the epidemiology of AA are generally consistent with prior studies, validating the scientific consistency of the AoU database and indicate its usefulness for evaluating the epidemiology of dermatologic diseases. Specifically, we confirmed that the increased prevalence of AA among underrepresented racial and ethnic groups persists, even after adjustment for socioeconomic status, educational attainment, physical disability, and health insurance status. Additionally, we found significantly decreased odds of AA diagnosis in participants without health insurance and those with low education and household income. These findings suggest that AA may be underdiagnosed in these populations, potentially mediated by limited access to dermatologic care. Future studies are needed to further replicate these associations and elucidate their underlying causes.