Management and Genetics of Alopecia Areata within the USA: A Cross-Sectional Study of All of Us Open Access

Alopecia areata (AA) is a common autoimmune disease that causes hair loss with up to 2% of the population experiencing AA during their lifetime [1, 2]. This disorder can vary greatly between patients, with some only experiencing small circumscribed patches of hair loss that resolve, while others will experience much more extensive hair loss which is difficult to treat [3]. This disorder is not merely cosmetic; it has a significant impact on patients’ quality of life with increased anxiety and depression [4, 5].

The goal of AA treatment is to halt hair loss and possibly promote hair regrowth; however, treatment can be difficult due to the high rate of relapse and unpredictable disease course [1, 6]. Pharmacological treatment typically includes intralesional, topical and oral corticosteroids, as well as other topical and oral immunosuppressive agents, and the newer Janus Kinase inhibitors [7]. Additionally, AA patients may present with comorbid conditions, particularly autoimmune disorders such as systemic lupus erythematosus (SLE), type 1 diabetes mellitus (T1 DM), and atopic dermatitis (AD) with which AA shares genetic risk factors; these can impact treatment decisions in this population [5]. On top of comorbid conditions, the impact of various demographic factors needs to be considered as these will contribute to inequities in access to care. Previously, it has been shown that the risk of AA is higher in those that are of non-European ancestry, and it is known that these populations tend to face more difficulties in accessing effective care [3, 5]. Moreover, many genetic studies investigating AA have been performed using only subjects from one ethnicity, which is usually European ancestry, limiting our knowledge of all contributing risk factors.

In 2017, the National Institute of Health (NIH) began recruiting participants for the All of Us research program, a large-scale health dataset that will enable research on diseases and disorders on a scale never seen before in a diverse population [8‒10]. The shear amount of data available within the All of Us dataset will allow us to not only study the influence of comorbid and demographic factors on AA but also examine the association of genetic factors in a diverse population that represents the real-world clinical picture. The goal of the present study was to determine the impact of comorbid conditions and socioeconomic factors on the management of AA, as well as uncover the influence of previously determined genetic risks, specifically previously reported single nucleotide polymorphisms (SNPs), within differing ethnic groups.

All studies were conducted using the controlled tier dataset (version 7) within the NIH’s All of Us research program (AoU). The AoU contains detailed health data on a diverse group of voluntary participants from underrepresented demographics within the USA. Specific details regarding the AoU research program can be found in publications by Denny et al. [8], Mayo et al. [9], and Ramirez et al. [10]. We extracted data from 287,012 participants, which had registered between the summer of 2017 up until July 1, 2022.

Study specific ethics approvals were not required as the AoU is a public resource where all ethics approvals have been secured by the NIH. The NIH has implemented data usage agreements with participating institutions, where all researchers wishing to access data must undergo data security and privacy training before access can be approved for the registered and controlled tier datasets.

We included only those participants who had electronic health records (EHRs) available for all studies. For genomic studies, we also required that participants have genomic data available, resulting in 206,173 participants being available for these studies. The AoU dataset makes use of the Observational Medical Outcomes Partnership (OMOP) framework, where all concept ID codes are searchable within the Athena OHDSI Vocabularies Repository (Odysseus Data Services, Inc.). All concept ID codes used within this study are contained within online supplementary Table S1 (for all online suppl. material, see https://doi.org/10.1159/000545194).

We determined the year-over-year trend for AA diagnoses in EHR in both males and females. We investigated the risk of AA diagnosis within the following autoimmune/inflammatory conditions: AD, autoimmune thyroid disease, psoriasis, rheumatoid arthritis (RA), SLE, T1 DM, and vitiligo.

We determined the usage trends of prescribed medications, namely, corticosteroids (betamethasone, dexamethasone, triamcinolone, and prednisone) and other immunosuppressive drugs in AA patients with several of the comorbid conditions listed above, namely, AD, psoriasis, SLE, and vitiligo due to their impact on the skin. In addition to studying comorbid conditions, we also investigated the impact of demographic factors on the rate of AA diagnosis and prescribing rates.

Lastly, we examined the impact of previously reported gene variants on the risk of AA. We studied SNPs in genes that have been linked to AA but also to the comorbid conditions listed above. We specifically examined SNPs within HLA-DRB1, HLA-DQA1, HLA-DQB2, CTLA4, IL2Ra, and NOTCH4 and analyzed the impact of these variations within participants of black/African American, Hispanic/Latino, and white/Caucasian ancestry. Online supplementary Table S2 lists all SNPs examined in this work.

The AoU program has strict privacy regulations that must be followed by all researchers, including data presentation. To protect participant privacy, all data within graphs are presented as percentages where applicable. In tables where participant numbers are shown, we have replaced any value of less than 20 with “<20,” in accordance with AoU regulations.

The SAS application within the AoU’s researcher workbench environment was used to conduct all analyses. We used logistical regression, Wald’s chi‐square test, Fisher’s exact test in our analysis. N values are listed in the captions for each figure and table where applicable.

We defined AA diagnosis as the first mention of AA in a patient’s EHR. When examining AA diagnoses year-over-year (Fig. 1), we found that there was an increase in the rate of diagnoses for both males (starting in 2013) and females (beginning in 2009). In addition, the risk of AA diagnosis was lower in males compared to females (OR 0.56 [0.48–0.65], p < 0.0001) (Fig. 2). We also investigated the likelihood of AA diagnosis in comorbid conditions (Fig. 2). There was as significant association with all comorbid conditions that we analyzed: vitiligo, AD, SLE, autoimmune thyroid disease, psoriasis, RA, and type 1 diabetes mellitus.

We investigated whether there were differences in the use of prescription medications within AA patients with comorbid conditions versus those with no additional comorbidity (Fig. 3). We focused on corticosteroid treatments as the proportion of prescriptions of other immunosuppressive drug was small compared to corticosteroids (online suppl. Fig. S1). We found an increased likelihood of topical corticosteroids (Fig. 3a) being prescribed to AA patients with vitiligo, psoriasis, and AD. No difference in intralesional corticosteroids was observed in AA patients with these comorbidities (Fig. 3b).

We investigated the likelihood that AA is listed in a patient’s EHR according to different demographic factors (Fig. 4). We found that AA was more likely to be listed for those that have higher levels of education as well as those with an annual household income greater than USD 75,000. We also observed that those participants who are black/African American or other (non-black, non-white, non-Hispanic) ethnicity were more likely to have AA in their EHR (OR 1.72 [1.40–2.12], p < 0.0001; OR 2.23 [1.71–2.92], p < 0.0001, respectively). The likelihood in Hispanics was trending toward an increased risk of AA.

We also examined the prescribing patterns in patients of different ethnicities (Fig. 5). We found that Hispanic and other (non-black, non-white, non-Hispanic) ancestry AA patients were less likely to receive any type of corticosteroid treatment (Fig. 5a); furthermore, other ancestry (non-black, non-white, non-Hispanic) AA patients were also less likely to receive intralesional corticosteroids when compared to white/Caucasian AA patients (Fig. 5b).

We examined the association of AA with previously reported genes (Fig. 6) in multiple ethnicities. When we look at all ethnicities combined (Fig. 6a) as well as white/Caucasian (Fig. 6b), we found that HLA-DQB2 (rs2301271) was associated with an increased risk, while HLA-DRB1 (rs9270502) and NOTCH4 (rs379464) were protective. However, in those of black/African American ancestry (Fig. 6c), only HLA-DRB1 had a protective association. In Hispanics, we found that HLA-DQB2 was no longer is a risk, but HLA-DRB1 and HLA-DQA1 were both protective. Additionally, we observed that CTLA4 variants had an impact, with the variant rs3087243 being protective, while the variant rs1024161 was associated with increased risk.

Our current study provides novel insight into the current management of AA within the USA and demonstrates differences in genetic risk factors. Understanding the management of this disorder will help inform researchers and policy makers about differences in care and how affected patients are impacted. Our work compliments but also builds upon the work of Moseley et al. [3] and Joshi et al. [4] by providing insights into treatment patterns and genetic risks in different populations [11].

When examining the diagnostic trends in AA, we found that the recorded cases increased in both males and females, beginning in 2013 for males and 2009 for females (Fig. 1). This observed increase may be due to more awareness – particularly with increasing visibility on social media platforms – and with increasing availability of treatment options for this disorder [12]. However, one caveat of this result is that we have used the first mention of AA in EHR data as the initial diagnosis, which has the potential to skew our results as the initial diagnosis may be earlier than that recorded for patients whose healthcare provider transitioned to EHR at a later time.

It is known that AA shares genetic risk factors with other autoimmune disorders, which could lead to an increased risk of AA in those patients [2, 13]. Our work confirms that there is an association between AA and these comorbid conditions, particularly vitiligo and AD (Fig. 2). Considering that these conditions are associated with AA, we then wanted to determine their impact on AA management. We analyzed skin-related comorbid conditions, finding that patients with comorbid AD, psoriasis, and vitiligo where more likely to receive topical corticosteroids compared to AA patients that were otherwise healthy (Fig. 3a). However, we cannot directly determine if these medications were prescribed for AA as patients that have these particular comorbid disorders are often prescribed topical corticosteroids, which would contribute to the increased likelihood that we have observed in this study [14].

We also wanted to assess the impact of demographic factors on AA management. Our data show that AA is more likely to be recorded in the EHRs of participants that have health insurance, higher level of education and higher income, as well as in participants that are of black/African American decent, or of other ancestry (non-white, non-black, non-Hispanic) in our multivariate model. We do see that the association of AA with Hispanic patients is trending toward significance. This result is consistent with that shown by Moseley et al. [3], where the odds of AA was significantly higher in participants that had access to health insurance and had higher income. This observation may be due to differences in a population’s ability to access care due to resource availability. Additionally, Lee et al. [5] and Sy et al. [15] also demonstrate increased rates and an association of AA with people not of white/Caucasian ethnicity. It is unclear what may be contributing to this increased risk, but it is known that the risk is higher in these populations for other autoimmune disorders, particularly SLE, which may share similar disease pathways [16]. However, we are specifically interested in how AA patients from these ethnic populations access care. Examining the odds of patients of non-white/Caucasian ancestry receiving any type of corticosteroid, we find that Hispanic and other (non-black, non-white, non-Hispanic) ancestry patients are less likely to receive corticosteroids, while black/African American patients are trending toward significance. Additionally, a similar result is observed for other (non-black, non-white, non-Hispanic) ancestry and black/African American patients when we examine intralesional corticosteroids. Differences in insurance coverage or concerns surrounding possible hypopigmentation may contribute to these observed differences [17‒19]. This result highlights the disparities that these patients face when accessing care, especially as these populations are at higher risk of developing AA. These disparities can be caused by geographic access, cultural-focused implicit bias, lack of knowledge by physicians on the specific needs ethnically diverse skin, patient awareness/education, and research biases [20‒22].

We have demonstrated important differences in how treatment choices vary between comorbidities but also by ethnicity. Additionally, we have also highlighted differences in the genetic risk factors for AA in different ethnicities.

AA shares some genetic risk factors with other autoimmune disorders such as vitiligo, psoriasis, AD, T1 DM, and RA, potentially increasing the risk of developing AA [13, 23, 24]. We examined SNPs within genes that have been previously shown to be associated with AA through genome-wide sequencing studies [2, 13, 25‒27]. We examined how these genes vary by ethnicity, as most studies have been conducted using patients of white/Caucasian ethnicity, preventing inferences to other ethnic populations [2, 27]. We found that the SNPs that had significant association when we examined the white (Fig. 6b) and all ethnicity groups (Fig. 6a) do not necessarily have significant associations within the black/African American (Fig. 6c) and Hispanic groups (Fig. 6d). This result highlights that there are differing genetic risk factors that need to be considered in the care of these populations. The underlying disease mechanisms could vary slightly, which may impact treatment efficacy. To our knowledge, this is the first study to examine differences in genetic risk factors between different ethnic groups in the same population.

The important insights this work provides do however come with some limitations. First, we have accessed EHR data to conduct our analysis. This type of data source may not be perfectly accurate as shown by Cai et al. [28] where the overall incongruence rate of AoU data was found to be less than 1%. Even though this proportion may seem small, this still has the potential to impact studies of low occurrence conditions, potentially skewing results. Bell et al. [29] have also demonstrated the potential for errors within EHR data. Their work showed that 21.2% of patients that were surveyed had identified errors within their EHR data, with significant errors being reported in almost half of these cases [29]. This work by Cai et al. [28] and Bell et al. [29] highlight the risks of conducting retrospective studies using EHR data.

Additionally, due to the structure of the dataset, we cannot directly know if a medication is prescribed for AA or a related condition, which is evident in the analysis of prescriptions in AA patients with comorbid conditions. We also cannot easily ascertain the severity of AA and how that may impact treatment choices or be influenced by genetics. Patients with ophiasis, alopecia totalis, and alopecia universalis are coded within the dataset, but the number of patients is small and prevents meaningful statistical analysis from being performed. For patients coded with only AA, there is no indicator of the severity of the disorder within the dataset.

The AoU program also suppresses the healthcare provider’s specialty in an effort to protect patient privacy. By not having access to these data, we are unable to attribute the prescribing rates of certain medications to specialty but also not able to assess what the primary point of care for patients is for AA. Finally, not all potential treatments may be recorded within the AoU dataset, particularly those that are newer or procedure-based, such as laser therapies [1].

Detailed knowledge of current clinical picture of AA is essential for improving patient outcomes for this appearance altering disorder. Determining the impact of specific characteristics of the patient population on treatment choices will help improve patient outcomes. This work, as well as previous work by others, enables a greater understanding of the clinical landscape of AA, paving the way to better patient outcomes.

Institutional ethics approval was not required for this study as all ethics approvals have been obtained by All of Us program, who have put in place data usage and privacy agreements for researchers from registered partner institutions. Additionally, researchers using the All of Us program scientific resources are not considered to be conducting human subject research as there is not direct contact with participants, identifiable data, and no direct access to biospecimens. Further detail regarding ethics within the All of Us program can be found at the following links (https://support.researchallofus.org/hc/article_attachments/22307203770516, https://support.researchallofus.org/hc/en-us/articles/4415498292244-Overview-of-All-of-Us-Research-Program-Policies-for-Researchers).

The authors are employed by Mediprobe Research Inc. A.K.G. was a member of the journal’s editorial board at the time of submission.

No sources of funding were obtained for this work.

A.K.G.: conceptualization, methodology, writing – original draft, and writing – critical review. V.E.: conceptualization, methodology, data curation, data analysis, writing – original draft, and writing – critical review.

The data accessed in this study are part of the NIH’s All of Us research program. We are unable to share the underlying dataset used in this study due to privacy restrictions implemented by the NIH and detailed within the “Data and Statistics Dissemination Policy” (https://support.researchallofus.org/hc/en-us/articles/22346276580372-Data-and-Statistics-Dissemination-Policy) which prevents sharing of participant level information by registered users. However, the All of Us research program publishes aggregate data on its website (researchallofus.org), while individuals affiliated with partnered institutions are able to register with the program to access datasets. All codes that we have used within study available within the online supplementary data file and summary statistics can be made available upon request. Further inquiries can be directed to the corresponding author.