Big Data Research on Keloids in Atopic Dermatitis Patients with Allergic Comorbidities: Increased Prevalence of Keloids in Atopic Dermatitis

Atopic dermatitis (AD) is associated with allergic comorbidities, such as asthma, allergic rhinitis (AR), and allergic contact dermatitis (ACD). The etiology of keloid is largely unknown; however, AD and keloid share inammatory pathways characterized by T-helper cell 2 cytokines and increased dermal broblast activity. The prevalence of keloids is known to increase in patients with AD. This study aimed to estimate the prevalence of keloids in patients with AD, and compare it with the prevalence of other comorbidities of AD. We assessed the Korean National Health Information Database and medical records including coexisting asthma, AR, and ACD. Single and multiple logistic regression models were created for keloids and each allergic disease. The prevalence of keloids was higher in the AD group than in the control group. Among patients with AD, adolescents and adults had a higher prevalence of keloids than infants and children. The risk of keloids was high with AD alone, and coexisting asthma signicantly increased the risk. Similarly, the risk of keloids was higher in AR associated with AD and ACD associated with AD than in AD alone. Thus, among Koreans, patients with AD have a higher risk of keloid development, with coexisting allergic diseases increasing the risk.


Introduction
Atopic dermatitis (AD) is a chronic in ammatory skin disease associated with numerous atopic and nonatopic comorbidities [1]. The prevalence of AD is estimated to be 15-20% in children and 1-3% in adults, and the incidence has increased two-to three-fold during the past few decades in industrialized countries [2].
Keloid is a broproliferative dermal tumor resulting from skin trauma and subsequent abnormal scar formation [3]. The mechanism underlying the pathological development of keloids is largely unknown; however, it is mainly characterized by an overproduction of extracellular matrix and high activity of broblasts during the in ammatory and remodeling phases of scar formation, resulting in the excessive accumulation of collagen [4,5]. These processes are also observed in broblasts among patients with AD [6]. Although the association between AD and keloids is still unclear, several possible links have been hypothesized.
AD and keloid share common in ammatory pathways that are characterized by T-helper cell 2 (Th2) cytokines. An impaired in ammatory response of brogenesis involving Th2 cytokines secreted by CD4 + T cells such as interleukin (IL)-4, IL-5, IL-10, and IL-13, leads to the imbalance between pro-brotic and anti-brotic signals [7]. Additionally, in AD, upregulation of Th2 cytokines is triggered, and IL-4, IL-5, IL-10, and IL-13 are secreted, resulting in epidermal thickening, in ammation, pruritus, and decreased expression of barrier proteins such as laggrin and loricrin, all of which are clinical characteristics of AD [8]. The reason why prurigo nodularis is frequently observed in patients with AD, with licheni cation being one of the most common clinical manifestations of AD, is presumed to be due to this process.
It is well established that patients with AD have a higher chance of developing allergic comorbidities such as asthma, allergic rhinitis (AR), and allergic contact dermatitis (ACD) [1]. Furthermore, the severity of AD was associated with a higher prevalence of comorbid allergic disorders, as well as increased severity of comorbidities. Based on these facts, the authors hypothesized that patients with both AD and allergic comorbidities might have a higher prevalence of keloids.
The prevalence of keloids throughout life is approximately 10% in Africa [9]. The prevalence is much lower in Asia, with the annual keloid incidence rate being 0.15% among the general Taiwanese population [10]. In a large-scale retrospective cohort study in Taiwan, the probability of keloid development in patients with AD is 3.19 times higher than in the healthy population [11]. However, to date, there are no known reports on the prevalence of keloids in AD accompanied by allergic diseases.
To the best of our knowledge, this study rst analyzed big data from the Korean National Health Information Database (NHID) and aimed to 1) estimate the prevalence of keloids in patients with AD and 2) compare the prevalence of keloids with those of other comorbidities of AD.

Data sources
The data used in this study were obtained from the Korean NHID, in which diseases are coded according to the International Classi cation of Diseases, Tenth Revision, Clinical Modi cation (ICD-10-CM). This database contain nationwide claims data, and is linked at an individual level with the Korean registration number that is assigned to all Korean residents at birth or immigration for practical purposes. This study was approved by the Institutional Review Board of Kyung Hee University (IRB number 2019-07-076) and a waiver of informed consent was approved from IRB due to the retrospective nature of the big datasets. All methods were performed in accordance with the relevant guidelines and regulations.

Study population
We assessed the Korean NHID from 2017 to 2018, and a survey of 2,517,401 subjects with AD and 3,132,283 non-AD subjects matched for age and sex. Medical records were analyzed according to the ICD-10-CM codes.

Inclusion criteria of patients and classi cation of data
We assessed 2,517,401 patients with AD from 2017 to 2018. A control group of 3,132,283 non-AD patients was extracted from a group that was not diagnosed with AD. All subjects in both groups were classi ed into two subgroups, with or without keloids. Demographic information (age and sex) and comorbidities of allergic diseases identi ed by ICD-10-CM codes were also collected. The associated allergic diseases in this study included asthma (ICD-10 code J304), AR (ICD-10 code L209), and ACD (ICD-10 code L239). All patients and the control group were classi ed by age: infant and early childhood, 0-5 years; late childhood, 6-10 years; adolescence, 11-20 years; adulthood, 21-60 years; and the elderly, >61 years.

Statistical analysis
We analyzed the chi-square test and Odds Ratio (OR) to see if there are differences in the frequency of occurrence of AD group and control group by age and gender. In the age group, the OR of late childhood, adolescence, adulthood, elderly was compared with infancy and early childhood as a reference. Also, in the sex, the OR of female was compared as male as a reference. In addition, we analyzed the variable that affect the risk of keloids according to the presence or absence of AD, age, and gender through the logistic regression. Furthermore, we analyzed the OR of the risk of keloids between the AD patients without asthma (or allergic rhinitis, allergic contact dermatitis) and with asthma (or allergic rhinitis, allergic contact dermatitis), and was compared to the control group through the multiple logistic regression adjusted by age, gender. A p-value of <0.05 was considered statistically signi cant. All data processing and statistical analyses were performed using SAS Enterprise Guide 7.1 (SAS Institute Inc., Cary, NC).

Baseline characteristics of patients
Baseline characteristics of patients in the AD and control group are shown in Table 1. All personnel in this study were Koreans. A total of 5,649,684 personnel were included in this study, and 44.56% (2,517,401) had AD (see Supplementary Table. S1 online). In patients with AD, 26.01% of patients were younger than 5 years, and 51.88% were younger than 20 years of age. The rate of AD tended to decrease with age. Of the patients with AD, 53.01% were females. Among all age groups, the prevalence of AD was higher only among infants and children but not among the remaining age groups. Baseline characteristics of patients in the AD and control group. All personnel in this study were Koreans.
The p-value was obtained through the chi-square test. CI; con dence interval Increased prevalence of keloids in the AD group The prevalence of keloids was higher in the AD group than in the control group (  The effect of presence or absence of AD on the risk of keloids was evaluated using the simple logistic regression.

AD, atopic dermatitis; CI, con dence interval
The effect of allergic comorbidities associated with AD on keloids The effect of having allergic comorbidities associated with AD on the risk of developing keloids (Table 3) was evaluated using multiple logistic regression. Patients with AD alone were at a higher risk of keloids (OR: 2.97; 95% CI: 2.91-3.03, p<0.001), and asthma coexisting with AD increased the risk of keloids signi cantly (OR 3.17; 95% CI 3.08-3.26, p<0.001). This effect was also observed in AR and ACD. In patients with AR associated with AD, the risk of keloids increased compared with that in AD alone (OR 3.  The effect of having allergic comorbidities associated with AD on the risk of keloids was evaluated using multiple logistic regression adjusted for age and sex.
AD, atopic dermatitis; AR, allergic rhinitis; ACD, allergic contact dermatitis; CI: con dence interval Discussion This is the rst study demonstrating that the risk of keloids is strongly associated with AD in the Korean population. Our results also demonstrate that the risk of keloids increases with coexisting allergic diseases, such as asthma, AR, and ACD. Previous data on the prevalence of keloids is based mainly on Caucasians and Taiwanese in Asia [10,11]. In Korea, where all citizens are obligated to subscribe to national medical insurance, big data research using electronic medical information is feasible. For the rst time, our ndings show that the prevalence of keloids increased in patients with AD in the Korean population. Hajdarbegovic et al. also suggested that keloids may be strongly associated with atopic asthma in European, African, and Asian populations [12]. The ndings in this study are consistent with those of previous studies wherein the risk of keloids is higher in AD patients with allergic comorbidities.
The molecular pathophysiology of keloid formation and progression is poorly understood. However, broblasts, one of the main cells responsible for most of the collagen and extracellular matrix deposition in both normal and abnormal wound healing, have been identi ed to contribute to keloid formation and dermal structural changes in AD skin lesions [13]. Recently, Shin et al. reported that keloidal broblasts treated with thymic stromal lymphopoietin (TSLP) produced signi cantly increased collagen I, collagen III, and transforming growth factor β [14]. TSLP, an IL-7-like cytokine, is thought to induce upregulation of Th2-in ammatory cytokines, and there is increased expression of the TSLP gene in AD skin lesions [15].
Another molecule that contributes to the collagen production of dermal broblasts in AD is Oncostatin M (OSM), a T lymphocyte/macrophage-derived proin ammatory signaling molecule similar to the IL-6 cytokine family [16]. Fibroblasts are target cells for OSM. OSM stimulates collagen and glycosaminoglycan production in dermal broblasts, and this process can be observed in both keloidal formation and prurigo nodularis in patients with AD [17].
Other than upregulation of Th2 cytokines, effector memory CD8 + T cells and CD103 + CD8 + resident memory T (TRM) cells are increased in keloid tissue. Therefore, increased CD8 + TRM in keloid tissues might contribute to local in ammation [18]. TRM cells have been shown to contribute to the recurrence of AD; however, the exact roles of skin TRM are still unclear [19].
There have been a few reports on the association between allergic diseases and keloid development; however, these results are still debated. Hajdarbegovic et al., for example, revealed no association between AD and keloids in the adjusted model of a case-control study. Additionally, they suggested that asthma was signi cantly, consistently, and strongly associated with keloids. However, there were no consistent associations found on keloids with atopic eczema or hay fever [12].  [11]. In our study, AD patients with ACD also had an increased prevalence of keloid risk, which is different from previous studies, suggesting that ACD may have the characteristic of AD-associated comorbidity. In the past, the higher permeation of contact allergens through the disrupted skin barrier was the main factor of ACD in patients with AD. Recently, potential shared immune pathways have been demonstrated for subsets of AD and ACD, including Th1, Th2, Th9, and/or Th17 [20,21].
Han et al. demonstrated the relationship between a disintegrin and metalloprotease 33 (ADAM33) polymorphism and keloid scars in an East Asian population. Q-1 SNPs in blood were signi cantly associated with keloid scars. ADAM33 protein is a zinc-dependent endopeptidase, characterized by a prodomain, metalloprotease domain, and disintegrin domain [22]. The ADAM33 gene was found to be associated with asthma and airway hyperresponsiveness [23]. It is thought that this may explain the increased keloid risk in AD associated with asthma. Considering the literature mentioned above, AD and keloid share various genetic and pathologic mechanisms.
One of the strengths of this study is the use of a large population-based claims dataset consisting of one ethnic group, which enabled the analysis of all cases of AD, keloids, and comorbid allergic diseases. Moreover, since the Korean NHID is one of the largest claims datasets including all age groups and the entire region, the possibility of selection bias is reduced to compare the relationships between the AD patients and control group. Second, it is the rst time that big data analysis has demonstrated the result of an increase in the prevalence of keloids in AD patients with allergic diseases. However, this study has several limitations. First, because the analyzed medical records were originally collected for national insurance purposes, not for research purposes, additional demographic or social histories were unavailable. Second, due to the drawbacks of big data analysis regarding control variables such as age and sex, we used multiple logistic regression models to match those of the two groups. This systematic problem may result in confounding factors.
In conclusion, this study demonstrated that in the Korean population, patients with AD have a higher risk of keloid development, and the risk is even higher with coexisting allergic diseases. Nevertheless, further exploration of this association is needed to con rm these relationships and the pathophysiological mechanisms common to AD and keloids.

Declarations
Availability of data and material All databases are available as Supplementary Information les.