Cutaneous Opportunistic Mycobacterial and Fungal Infections in Adult-Onset Immunodeficiency Due to Anti-Interferon-Gamma Autoantibodies—Decoding Skin Involvement Patterns Authors: Rujira Rujiwetpongstorn, Phichayut Phinyo, Nayanunt Prinyaroj, Romanee Chaiwarith, Salin Kiratikanon, Napatra Tovanabutra, Siri Chiewchanvit, Charoen Choonhakarn, Suteeraporn Chaowattanapanit, Mati Chuamanochan, Atibordee Meesing Published: 2025 ABSTRACT AbstractBackgroundAdult-onset immunodeficiency (AOID) due to anti-interferon-gamma autoantibodies (AIGA) predisposes patients to multiple opportunistic infections (OIs), including cutaneous involvement. However, data specifically addressing cutaneous OIs in AOID remain limited. This study aimed to characterize and compare clinical and laboratory features, with a focus on cutaneous OIs caused by mycobacteria and fungi.MethodsA retrospective chart review was conducted on AOID patients with cutaneous OIs in Chiang Mai and Khon Kaen University Hospital from January 2000 to December 2020. Cases with infection due to Mycobacterium species—both Mycobacterium tuberculosis complex and non-tuberculous mycobacteria—and fungal pathogens were included. Clinical characteristics and laboratory parameters were analyzed, and a multivariate predictive model was developed to differentiate between cutaneous mycobacterial and fungal infections.ResultsA total of 61 AOID patients with 70 cutaneous infectious episodes were identified. Among these, fifty episodes were due to Mycobacterium infection. Lesions involving the neck with concurrent lymphadenopathy were more suggestive of Mycobacterium infection. In contrast, fungal infections were associated with generalized or truncal involvement, anemia, neutrophilia, lower monocyte percentage, hyperglobulinemia, and higher levels of aspartate aminotransferase. A multivariate model incorporating these variables achieved excellent discriminatory performance (area under the receiver operating characteristic curve: 0.94; 95% confidence interval: 0.87–0.99).ConclusionsCutaneous lesions involving the neck area and lymphadenopathy are clinical clues that suggest cutaneous Mycobacterium OIs. Nevertheless, integration of clinical features with laboratory findings enables the development of a predictive model that can effectively differentiate between cutaneous mycobacterial and fungal infections. This predictive model may aid in selecting appropriate empirical antimicrobial therapy, particularly when microbiological confirmation is pending or inconclusive despite high clinical suspicion.Keywords: adult-onset immunodeficiency, anti-interferon-gamma autoantibodies, cutaneous opportunistic infections, fungal skin infection, mycobacterial skin infection Interferon-gamma (IFN-γ), a crucial cytokine produced mainly by activated T Helper 1 lymphocytes, plays an essential role in the host defense mechanism against intracellular infection, including mycobacteria, certain viruses, and fungi [1, 2]. Autoantibodies to this cytokine result in an emerging form of adult-onset immunodeficiency (AOID) [3]. Adult-onset immunodeficiency due to anti-interferon-gamma autoantibodies (AIGA) is characterized by the presence of high-titer neutralizing autoantibodies against IFN-γ. These autoantibodies impair IFN-γ signaling, leading to increased susceptibility to opportunistic intracellular infections, particularly with nontuberculous mycobacteria (NTM), Salmonella spp., specific viral pathogens, and certain fungal pathogens. This syndrome, often occurring in previously healthy adults without known primary immunodeficiencies or human immunodeficiency virus (HIV) infection, has been increasingly reported in Southeast Asia and other regions [3]. Cutaneous manifestations in AOID have been described as both infective and reactive skin disease [4–6]. However, data specifically focusing on cutaneous opportunistic infections (OIs) remain limited. Previous reports have shown that cutaneous OIs can present as either localized or disseminated cutaneous infections, most caused by herpes viruses, followed by mycobacterial and fungal pathogens [5, 6]. Herpes infection typically presents with a group of vesicles, which are distinctive and relatively easy to diagnose. In contrast, cutaneous Mycobacterium and fungal infections often lack specific clinical clues, making differentiation between these 2 groups of pathogens challenging in real-world clinical practice. This diagnostic uncertainty can significantly impact the selection of appropriate antimicrobial therapy, particularly in cases where mycobacterial or fungal infections are suspected. Our study aims to evaluate the clinical characteristics of cutaneous OIs caused by mycobacteria and fungi in patients with AOID associated with AIGA, and to develop a prediction model to aid in distinguishing between these 2 organisms for skin infections in this unique patient population. MATERIAL AND METHODS Subject and Study Design This retrospective chart review included all patients diagnosed with AOID associated with AIGA, presenting with cutaneous OIs caused by mycobacteria and fungi. These patients were followed up at Maharaj Nakorn Chiang Mai Hospital and Srinagarind Hospital, Thailand, between January 2000 and December 2020. The diagnosis of AOID associated with AIGA was based on the following criteria: (1) detection of AIGA using either dot enzyme-linked immunosorbent assay (ELISA) at Maharaj Nakorn Chiang Mai Hospital [7] or an inhibitory ELISA at Srinagarind Hospital [8]; (2) exclusion of other causes of immunodeficiency, including human immunodeficiency virus (HIV) infection, malignancies, and immunosuppressive therapies; and (3) presence of characteristic opportunistic infections that indicate impaired cell-mediated immunity (CMI). In all cases, specimens for microbiological confirmation were obtained primarily through tissue biopsy of cutaneous lesions or aspiration of subcutaneous abscess. The microorganisms were identified using culture or molecular methods, including line probe assays and polymerase chain reaction. For infections involving other sites, specimens were collected as clinically indicated, typically by excisional biopsy for lymph nodes, and by blood culture for bloodstream infections. These specimens were processed using the same microbiological and molecular methods as those applied to skin specimens. In some cases, with bone and joint or hepatic and splenic involvement, diagnosis and treatment response were assessed using imaging studies performed before and after therapy. The primary aim was to identify clinical characteristics of cutaneous OIs, specifically focusing on infections caused by mycobacteria and fungi in AOID associated with AIGA patients. The secondary aim was to develop a predictive model to help differentiate between mycobacterial and fungal skin infections. The study protocol was approved by the Research Ethics Committee, Faculty of Medicine, Chiang Mai University (Approval No. MED-2563-07705) and the Khon Kaen University Ethics Committee for Human Research (Approval No. HE641146). Data Collection Medical records of all AOID associated with AIGA patients and cutaneous OIs by mycobacteria and fungi were reviewed. Demographic data, cutaneous OIs including rash morphology and distribution, type of microorganism, associated systemic findings, and relevant laboratory results were documented. For patients who experienced multiple episodes of cutaneous OIs, all episodes were documented. Data collection encompassed both static demographic variables and dynamic clinical data corresponding to each episode of infection. Statistical Analysis We used descriptive statistics to summarize baseline demographic and clinical characteristics. Categorical variables were reported as counts and percentages, and continuous variables as means ± standard deviations or medians with interquartile ranges. Group comparisons were performed using the χ² test or Fisher's exact test for categorical variables, and the Student's t-test or Wilcoxon rank-sum test for continuous variables. Statistical significance was set at P < .1. Prior to statistical modeling, we performed k-nearest neighbor imputation (using the VIM package in R) to address missing values in the predictor variables [9]. The group variable was excluded from the imputation process to prevent data leakage. Predictors with a P-value < .1 in the univariable analysis were included in the multivariable Firth's logistic regression, which is appropriate for sparse outcome data [10, 11]. We used stepwise backward elimination to exclude non-significant predictors at a P-value threshold of 0.1. The final reduced model was evaluated for discrimination and calibration. We reported the area under the receiver operating characteristic curve (AuROC) and a calibration plot to represent the model's apparent performance. Internal validation was then performed using bootstrap resampling with 200 replicates. We estimated and reported the optimism values for both the AuROC and calibration slope, which can inform external validation. All statistical analyses were conducted using R (version 4.4.2; R Core Team 2021) and Stata 18.5 (StataCorp, College Station, Texas, USA). RESULTS Baseline Characteristics A total of 399 cases diagnosed with AOID associated with AIGAs were identified; of these, 61 patients (15.3%) with 70 cutaneous OIs episodes (50 episodes caused by mycobacteria and 20 episodes caused by fungi) were included in the study. A younger age of onset (48.5 ± 10.4 years) and a slight female predominance were observed in cutaneous OIs caused by mycobacteria compared to those caused by fungi (Table 1). In cutaneous OIs caused by mycobacteria, M. abscessus was the most common pathogen, followed by Mycobacterium avium complex, while Talaromyces marneffei was the most common associated causative fungal pathogen (Figure 1). Table 1.Baseline Characteristics of Patients With AOID Associated With AIGA and Cutaneous Opportunistic Infections Caused by Mycobacteria and FungiParameterMissingn (%) Mycobacterium (n = 50)Fungus(n = 20) P Valuen (%)n (%)Demographic dataAge, mean (SD)0 (0)48.5 ± 10.453.4 ± 8.6.052Sex, female0 (0)27 (54)7 (35).190Cutaneous OIs morphologya Papule0 (0)20 (40)7 (35).790 Plaque0 (0)6 (12)3 (15).708 Nodule0 (0)13 (26)7 (35).560 Patch0 (0)3 (6)2 (10).619 Vesicle0 (0)0 (0)1 (5).286 Pustule0 (0)5 (10)4 (20).265 Abscess0 (0)3 (6)3 (15).343 Otherb0 (0)11 (22)2 (10).322 Location a Head0 (0)14 (28)5 (25)1.000 Neck0 (0)20 (40)2 (10).021 Upper extremities0 (0)13 (26)7 (35).560 Lower extremities0 (0)8 (16)1 (5).430 Trunk0 (0)2 (4)4 (20).051 Generalized0 (0)0 (0)2 (10).079 Otherc0 (0)4 (8)1 (5)1.000 Extracutaneous involvement 38 (76)10 (50).047 Organ involvementa Lymph node0 (0)32 (64)1 (5)<.001 Lung0 (0)8 (16)4 (20).732 Bone0 (0)6 (12)5 (25).274 Blood0 (0)7 (14)4 (20).717 Liver0 (0)0 (0)1 (5).286 Spleen0 (0)1 (2)1 (5).493AIGA, anti-interferon gamma autoantibodies; AOID, adult-onset immunodeficiency. aOne episode could have more than one morphology and location for the lesion, as well as other organ involvement. bOther morphology included an ulcer, crust, and mass. cOther locations included the genitalia, groin, and buttocks. Baseline Characteristics of Patients With AOID Associated With AIGA and Cutaneous Opportunistic Infections Caused by Mycobacteria and Fungi AIGA, anti-interferon gamma autoantibodies; AOID, adult-onset immunodeficiency. aOne episode could have more than one morphology and location for the lesion, as well as other organ involvement. bOther morphology included an ulcer, crust, and mass. cOther locations included the genitalia, groin, and buttocks. Types and proportions of causative organisms. MTB complex, Mycobacterium tuberculosis complex; NTM, nontuberculous mycobacteria*. *NTM refers to cases in which the non-tuberculous mycobacterium was identified only to the genus level. Lesional Morphology and Distribution Between Mycobacterial and Fungal Skin Infections Papules and nodules were the common presentation in both cutaneous infections with Mycobacterium and fungi (Table 1). There was no statistically significant difference between the morphology of the 2 groups. Interestingly, we found that the location of the rashes was significantly different. The neck was the most reported site for patients with cutaneous OIs caused by mycobacterial infection (P-value .021) (Figure 2). Generalized lesions as well as truncal involvement were preferred for fungal infection (P-value .079 and .051, respectively). Multiple erythematous papules and crusted plaques on the submandibular and neck area in a patient with cutaneous mycobacterial infection. Seventy-six percent of patients with Mycobacterium skin infections had not only cutaneous OIs but also concomitant systemic involvement, with lymph nodes being the most commonly affected organ. In contrast, extra-nodal involvement, particularly in the bones and joints, bloodstream, and lungs, was more common in the fungal group. Laboratory Characteristics of Mycobacteria and Fungal Skin Infections Anemia, higher neutrophil percentage, monocytosis, elevated aspartate aminotransferase (AST) levels, as well as hypoalbuminemia with hyperglobulinemia, were significantly associated with cutaneous OIs caused by fungi compared to those caused by the mycobacteria group (Table 2). Table 2.Laboratory Characteristics of Patients With AOID Associated With AIGA and Cutaneous Opportunistic Infections Caused by Mycobacteria and FungiCharacteristicsMissingn (%)Mycobacteria (n = 50)Fungus (n = 20) P ValueMean ± SDMean ± SDHemoglobin (g/dL)13 (16.1)10.4 ± 1.98.6 ± 1.3<.001Hematocrit (%)13 (16.1)32.4 ± 5.427.5 ± 4.5.001Platelet count (×103/μL)Median ± IQR13 (16.1)381.5 (315.0, 441.0)562.5 (203.5, 670.0).104White blood cell count(×103 cells/μL)Median ± IQR13 (16.1)16.3 (11.0, 22.3)16.9 (14.8, 22.3).469Neutrophil (%)13 (16.1)71.5 ± 13.577.7 ± 9.7.054Lymphocyte (%)Median ± IQR13 (16.1)13.7 (8.8, 25.4)11.9 (8.6, 18.5).334Monocyte (%)Median ± IQR13 (16.1)4.4 (3.3, 5.8)3.3 (2.4, 4.2).014Eosinophil (%)Median ± IQR13 (16.1)4.0 (2.4, 5.3)4.7 (2.4, 5.8).723Basophil (%)Median ± IQR14 (17.3).5 (.3, 0.8)0.4 (0.3, 0.6).438Creatinine (mg/dL)16 (19.8)1.0 ± 0.31.0 ± 0.3.980AST (U/L)Median ± IQR18 (22.2)21.5 (15.0, 25.5)27.5 (18.0, 37.5).046ALT (U/L)Median ± IQR18 (22.2)19.5 (12.5, 26.0)17.0 (12.5, 32.0).815ALP (U/L)Median ± IQR20 (24.7)191.0 (126.0, 287.0)241.5 (163.5, 472.5).271GGT (U/L)Median ± IQR71 (87.7)190.5 (94.0, 236.5)482.0 (239.0, 725.0).116Albumin (g/dL)20 (24.7)3.6 ± .72.4 ± .6<.001Globulin (g/dL)20 (24.7)4.4 ± 1.05.8 ± 1.2<.001ESR (mm/h)42 (51.9)83.3 ± 29.4109.1 ± 24.8.038CRP52 (64.2)57.6 ± 39.451.8 ± 18.9.642AIGAs, anti-interferon gamma autoantibodies; ALP, alkaline phosphatase; ALT, alanine transaminase; AOID, adult-onset immunodeficiency; AST, aspartate aminotransferase; CRP, C-reactive protein; ESR, erythrocyte sedimentation rate; GGT, gamma-glutamyl transferase. Laboratory Characteristics of Patients With AOID Associated With AIGA and Cutaneous Opportunistic Infections Caused by Mycobacteria and Fungi AIGAs, anti-interferon gamma autoantibodies; ALP, alkaline phosphatase; ALT, alanine transaminase; AOID, adult-onset immunodeficiency; AST, aspartate aminotransferase; CRP, C-reactive protein; ESR, erythrocyte sedimentation rate; GGT, gamma-glutamyl transferase. Development of a Predictive Model to Distinguish Between Mycobacterial and Fungal Cutaneous OIs The significant parameters associated with pathogenic organisms, as identified through univariable analysis, were further examined using multivariable Firth's logistic regression analysis, both in full and reduced models (Table 3). The final reduced model comprised of lesion at the neck (OR 5.69; 95% CI 1.06, 30.67), lymph node involvement (OR 137.71; 95% CI 4.94, 3840.45), hemoglobin level (OR 2.10; 95% CI 1.02, 4.32), percentage of neutrophil (OR 1.10; 95% CI 1.00, 1.20), percentage of monocyte (OR 2.64; 95% CI 1.10, 6.34), and AST levels (OR 0.98; 95% CI 0.95, 1.00). The entire formulation model is available at https://bioce.shinyapps.io/mtb-app/ Table 3.Multivariable Firth's Logistic Regression AnalysisCharacteristicFull ModelReduced ModelβmOR95%CI P-valuemOR95%CI P-valueAge0.980.88, 1.10.772Location of lesions Neck3.580.54, 23.73.1875.691.06, 30.67.0431.74 Trunk0.740.45, 1.23.247………… Generalized0.840.52, 1.33.450…………Lymphadenopathy77.592.39, 2521.25.014137.714.94, 3840.45.0044.93Hemoglobin (g/dL)1.880.81, 4.33.1402.101.02, 4.32.0430.74Neutrophil (%)1.090.98, 1.22.1131.101.00, 1.20.0550.09Monocyte (%)2.430.98, 6.04.0552.641.10, 6.34.0290.97AST (U/L)0.980.96, 1.01.1480.980.95, 1.00.073−0.02Globulin (g/dL)0.790.32. 1.97.617…………Constanta………………−18.31Full and reduced models are based on the imputed dataset (n = 70).CI, confidence interval; mOR, multivariable odds ratio. aShown in log odds scale. Multivariable Firth's Logistic Regression Analysis Full and reduced models are based on the imputed dataset (n = 70). CI, confidence interval; mOR, multivariable odds ratio. aShown in log odds scale. The final model demonstrated strong discriminative ability, with an apparent AuROC of 0.94 (95% CI: 0.87–0.99) for distinguishing between mycobacterial and fungal cutaneous infections in AOID patients with AIGA. Apparent calibration was acceptable, with a calibration-in-the-large (CITL) of 0.072 and a calibration slope of 1.370 (Figure 3). After internal validation, the optimism-adjusted AuROC, CITL, and calibration slope were 0.916, −0.148, and 0.497, respectively. A, Apparent discrimination performance of the model in differentiating between cutaneous mycobacterial and fungal infections. B, Model apparent calibration. DISCUSSION Skin is one of the most affected organs in AOID associated with AIGA. While reactive dermatoses have been more frequently reported and extensively described in the literature, infective dermatoses—particularly those caused by OIs—have received relatively limited attention. In our study, we found that approximately 15% of AOID-associated AIGA patients experienced at least one episode of cutaneous OIs caused by pathogenic organisms, particularly NTM and fungi. This study is the first to try to define and compare data on Mycobacterium and fungal skin infections in AOID associated with AIGA patients. The most common pathogen reported as OIs in AOID associated with AIGA patients was NTM [3, 12], which can be classified into 2 groups: rapid-growing mycobacteria and slow-growing mycobacteria [13]. Their natural habitats are ubiquitous, encompassing both saprophytic and aquatic environments. Cutaneous Mycobacterium infection can occur as a primary inoculation to the skin or as regional spread from an internal organ, as well as through hematogenous spread [13, 14]. Clinically, cutaneous mycobacterial infections may present either as a cutaneous limited form or as part of disseminated disease involving internal organs [15–17]. We observed that cutaneous mycobacterial infections in our AOID-associated AIGA patients were frequently accompanied by systemic involvement, most notably lymphadenopathy. Our results are consistent with previous reports in immunosuppressed patients, who mainly have disseminated cutaneous as well as internal organ involvement [15]. The high rate of internal organ involvement in mycobacterial infections in our cohort emphasizes the importance of early recognition and thorough systemic evaluation when assessing cutaneous manifestations in this unique immunodeficient population. In our study, M. abscessus was identified as the most frequently isolated species among OIs followed by Mycobacterium avium complex (MAC). This pattern aligns with findings from previous reports, particularly from the Central part of Thailand and other regions in Asia [6, 16–19]. The predominance of M. abscessus in these settings remains incompletely understood; however, this observation may be influenced by a combination of biological and genetic factors that vary by geographic region [19, 20]. To date, only a few studies have identified specific genetic variants associated with infections by particular NTM species. Most genetic risk factors broadly affect host immunity, rather than targeting specific species, such as M. avium, M. abscessus, or M. kansasii. Single-gene immune pathway defects, such as mutations in the IL-12/IFN-γ axis, including IFNγR1, IFNγR2, IL12β, IL12Rβ1, STAT1, RORC, and NEMO, are responsible for Mendelian susceptibility to mycobacterial disease, leading to severe NTM infections [21]. Autoantibodies against interferon-γ and primary immunodeficiencies like GATA2 deficiency or CD4+ T-cell defects predispose to disseminated (often non-pulmonary) NTM infections. While environmental factors play the main role behind this regional species distribution, the traditional host-risk variants (eg, the IL-12/IFN-γ axis, HLA, CFTR, etc.) are broadly studied but not specifically linked to regional populations [20, 22]. For fungal skin infections, half of our patients with cutaneous fungal manifestations also had evidence of systemic fungal infection, affecting other organs, such as the musculoskeletal system, lungs, and hematologic system. Interestingly, the most commonly identified fungal pathogen in our study was Talaromyces marneffei. This finding is consistent with previous epidemiological data from northern and northeastern Thailand, where Talaromyces is endemic and recognized as a leading opportunistic infection in immunocompromised hosts, particularly those with HIV infection. The predominance of Talaromyces in our cohort likely reflects regional environmental exposure and endemicity of the organism [23–25]. Papules and nodules predominantly characterized the cutaneous morphology observed in our cohort. This presentation is consistent with prior reports describing skin lesions in patients with skin and soft tissue infections caused by these organisms [15, 16, 18, 26]. Notably, we found no significant morphological differences between lesions caused by Mycobacterium species and those associated with fungal infections, as there is no previous report comparing these data. However, a key distinguishing feature in our study is the distribution of lesions. Specifically, neck involvement is markedly prominent in cases of mycobacterial infection. This pattern differs from earlier reports, where lesion distribution was more generalized or predominantly involved the extremities [16–19]. The increased localization to the neck in our study suggests a potential link between cutaneous mycobacterial infection and underlying or adjacent lymph node involvement, particularly in the cervical region. However, the lesions observed were not scrofuloderma-like in appearance, and the precise relationship between the neck lesions and regional lymphadenopathy remains unclear. Further investigation is warranted to explore the relationship between cervical lymph node involvement and the increased prevalence of cutaneous infections at the neck in AOID patients. Among the laboratory profiles, anemia, neutrophilia, higher levels of AST, and hyperglobulinemia were significantly associated with fungal infections. This observation may reflect a broader systemic inflammatory response, which is often more pronounced in disseminated or invasive fungal infections. The decision to include globulin rather than albumin in the multivariable model aimed to reduce potential confounding. Albumin levels can be influenced by multiple non-inflammatory factors such as nutritional status, liver function, and fluid balance, which may limit their specificity as a marker of infection-related inflammation [27, 28]. In contrast, serum globulin more accurately reflects immune activation and chronic inflammatory states, making it a more suitable variable for distinguishing between mycobacterial and fungal infections in this context. Informed by these considerations, we developed and evaluated a multivariable Firth's logistic regression model to differentiate between mycobacterial and fungal infections based on clinical and laboratory features. The reduced model identified key independent predictors, with cutaneous lesions involving the neck area, presence of lymphadenopathy, hemoglobin levels, neutrophil and monocyte percentages, and levels of serum AST contributing significantly to organism classification. Higher model-derived predicted probabilities were associated with an increased likelihood of mycobacterial infection. This predictive model may offer diagnostic utility in informing the selection of appropriate initial antimicrobial therapy, particularly during the interim period awaiting microbiological confirmation, or in cases where microbiological investigations are negative despite a high index of suspicion for chronic mycobacterial or fungal infection. There were some limitations to our study. First, the study is retrospective in nature, which may introduce recall bias and result in some missing data. Secondly, the relatively small sample size—especially in the group with cutaneous fungal infections—poses another limitation. Nonetheless, Firth's penalized likelihood approach was crucial in mitigating the potential bias caused by small sample sizes and rare events. This statistical approach enhances the robustness of the observed associations by reducing the issues of separation and model instability commonly encountered with standard logistic regression under these conditions. Finally, detailed histopathological data were not available, as histological features could provide additional insights for differentiating between these two infectious conditions. CONCLUSION This is the first study comparing the clinical characteristics of cutaneous opportunistic infections between mycobacterial and fungal infections in AOID associated with AIGA patients. The location of skin lesions, particularly neck involvement, and lymph node enlargement may provide important clinical clues suggesting a cutaneous mycobacterial infection in AOID patients with AIGA. In contrast, fungal infections may be linked with greater systemic inflammation and hematologic compromise. Integrating these clinical and laboratory parameters may enhance diagnostic accuracy and expedite appropriate management for this complex patient population. Notes Author contributions . All authors critically reviewed and contributed to the writing of the manuscript and gave final approval of the version to be published. R.R., P.P., M.C., and A.M. conceptualized the study; R.R., P.P., N.P., A.M. collected and extracted data; R.R., P.P. performed data analysis; R.R., P.P., A.M. wrote the manuscript; R.R. and P.P. have contributed equally to this work and share first authorship; R.C, S.K., N.T., Si.C., C.C., Su.C., and M.C. revised the manuscript; A.M. supervised the work. All co-authors contributed to the interpretation of the results and the approval of this final version for submission. Financial support. None. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Contributor Information Rujira Rujiwetpongstorn, Division of Dermatology, Department of Internal Medicine, Faculty of Medicine, Chiang Mai University, Chiang Mai, Thailand. Phichayut Phinyo, Center for Clinical Epidemiology and Clinical Statistics, Faculty of Medicine, Chiang Mai University, Chiang Mai, Thailand; Department of Biomedical Informatics and Clinical Epidemiology (BioCE), Faculty of Medicine, Chiang Mai University, Chiang Mai, Thailand. Nayanunt Prinyaroj, Division of Dermatology, Department of Internal Medicine, Faculty of Medicine, Chiang Mai University, Chiang Mai, Thailand. Romanee Chaiwarith, Division of Infectious Diseases and Tropical Medicine, Department of Internal Medicine, Faculty of Medicine, Chiang Mai University, Chiang Mai, Thailand. Salin Kiratikanon, Division of Dermatology, Department of Internal Medicine, Faculty of Medicine, Chiang Mai University, Chiang Mai, Thailand. Napatra Tovanabutra, Division of Dermatology, Department of Internal Medicine, Faculty of Medicine, Chiang Mai University, Chiang Mai, Thailand. Siri Chiewchanvit, Division of Dermatology, Department of Internal Medicine, Faculty of Medicine, Chiang Mai University, Chiang Mai, Thailand. Charoen Choonhakarn, Division of Dermatology, Department of Medicine, Faculty of Medicine, Khon Kaen University, Khon Kaen, Thailand. Suteeraporn Chaowattanapanit, Division of Dermatology, Department of Medicine, Faculty of Medicine, Khon Kaen University, Khon Kaen, Thailand. Mati Chuamanochan, Division of Dermatology, Department of Internal Medicine, Faculty of Medicine, Chiang Mai University, Chiang Mai, Thailand; Pharmacoepidemiology and Statistics Research Center (PESRC), Faculty of Pharmacy, Chiang Mai University, Chiang Mai, Thailand. Atibordee Meesing, Division of Infectious Diseases and Tropical Medicine, Department of Medicine, Faculty of Medicine, Khon Kaen University, Khon Kaen, Thailand. References