Cancer risk by the subtype of alopecia

The cancer risk in patients with alopecia areata (AA) or alopecia totalis (AT)/alopecia universalis (AU) remains unknown. In this study, national statistical data were used to study the association between these forms of alopecia and the risk of cancer. We enrolled 668,604 patients who were treated for alopecia from 2007 to 2014, and age- and sex-matched control subjects. AA and AT/AU patients had slightly higher overall cancer risks (hazard ratio (HR), 1.043; 95% confidence interval (CI), 1.022–1.065 and HR, 1.07; 95% CI, 1.013–1.129, respectively) than controls, after adjusting for confounding factors. The risks of oral cavity, esophagus, liver, biliary tract, pancreas, larynx, lung, kidney, breast, cervix, ovary, uterus, testis, nerve, and skin cancers; and lymphoma, multiple myeloma, and leukemia, were not increased in alopecia patients. In AA or AT/AU patients, the only increased risk was that of thyroid cancer. In AA patients alone, the risks of bladder and prostate cancers were increased. Thus, the cancer risks varied by the alopecia subtype. Careful monitoring is needed to explore if the actual risks of thyroid, bladder, and prostate cancers are increased in alopecia patients.


Discussion
To our knowledge, this is the first study to investigate the total cancer risk in alopecia patients using a population-based approach. We found that the total cancer risk was increased in alopecia patients compared with  age-and sex-matched controls. This was attributable to an increase in the risks of thyroid, prostate, and bladder cancer depending on the alopecia subtype. In particular, the risk of thyroid cancer was higher in the AT/AU group than in the age-and sex-matched control group or the AA group. Alopecia is a hair-loss disorder associated with inflammation and an autoimmune response 5 , affecting an estimated 4.5 million people in the United States 6 . AA is thought to be a TH1-mediated autoimmune disease in which the hair follicle loses its status as an immunoprivileged site, resulting in perifollicular CD8+ cytotoxic T cell infiltration followed by elevated IFN-γ production 7 . Genetic polymorphisms in genes encoding cytokines affect their transcriptional levels, associated with inter-individual variations in cytokine production, thus influencing the outcomes of cancers and autoimmune diseases 8 . Lew et al. reported that the single-nucleotide polymorphism IL17RA (rs879577) was significantly enriched in AA patients 9 . In addition, a significant association between the intron 3VNTR polymorphism of the IL-4 gene and AA susceptibility was reported in a Turkish population 10 13 . We found that the risk of thyroid cancer was higher in AA patients than the general population. Chu et al. performed a nationwide population-based study in Taiwan and found a relationship between AA and thyroid diseases 14 . In addition, recent studies have shown that AA is associated with autoimmune thyroiditis, but the causative AA subtype was not identified, and thyroid autoantibodies were absent 15,16 . Sun et al. reported that the risk of thyroid cancer was increased in females with alopecia, but found no association with AA in general, or a particular AA subtype 4 . Although no association between AA and thyroid cancer was noted in previous studies, it is significant that our present large-scale study revealed such a relationship, most notably in the AT/AU group.
The association between AA and thyroid cancer may be attributable to shared pathological features. Chronic inflammation increases the long-term risk of cancer of the thyroid, prostate, and bladder 17 . FAS and FASLG are pro-apoptotic proteins playing major roles in both cancer development and that of various diseases of the immune system 18 . Reduced FAS expression and/or increased FASLG expression facilitate tumor development and progression by inhibiting tumor cell apoptosis or by inducing immune cell apoptosis. Recently, FAS and FASLG polymorphisms have been reported in various cancers and alopecia 11,19 . Additional mechanism studies including FAS and FASLG are needed to define the risk of thyroid cancer in AT/AU patients.
We found that AA was associated with an increased risk of prostate cancer (HR, 1.259; 95% CI, 1.139-1.391) compared with AT/AU. Few previous reports have sought associations between AA and prostate cancer. Even the suggested association between androgenic alopecia and prostate cancer remains unclear [20][21][22][23] . However, androgen levels is suggested influence the pathogenesis of prostate cancer 24 . Taylor et al. suggested a relationship between a vitamin D receptor gene polymorphism and prostate cancer 25 . Recently, Conic et al. found that AA was associated with vitamin D deficiency and increased androgen levels, as are thyroid disease, anemia, and eczema 26 . Thus, increased androgen levels and a vitamin D deficiency may contribute to the pathogenesis of both diseases. Another shared mechanism may be the pathway involving Th17 cells and IL-17. Zhang et al. suggested that IL-17 promoted the initiation and growth of prostate cancer, and that the IL-17-MMP7 pathway was involved in prostatic intraepithelial neoplasia prior to the development of prostate cancer 27 . However, these mechanisms do not adequately explain the relationship between AA and prostate cancer.
Moreover, we found that AA was associated with an increased risk of bladder cancer (HR, 1.216; 95% CI, 1.059-1.397). Androgen-mediated signaling by the androgen receptor is known to play roles in bladder and urothelial cancer 28 . Th17 and Treg cells (especially the former) were involved in the development and progression of bladder cancer 29 . These mechanisms partially overlap with those of prostate cancer (described above). However, it remains difficult to completely explain the associations between alopecia and prostate/bladder cancer; further studies are needed.
In addition, we found that the incidence of skin cancer was not increased in AA patients, consistent with previous studies 30,31 . Miller et al. found that the risks of non-melanoma skin cancers (basal cell carcinoma and squamous cell carcinoma), and melanoma, were not increased among AA patients in Cleveland (OH, USA) 30 .
Our results also showed that the risk of skin cancer was not increased in AT/AU patients.
Our study had certain limitations. We lacked information on the numbers of alopecia lesions and their treatment, lifestyles, smoking and drinking status, and the levels of physical activity. Also, the ages at onset of alopecia and various cancers differ; our follow-up duration may have been inadequate. Despite these limitations, we found that a large-scale analysis of national data revealed that the risks of thyroid, prostate, and bladder cancer increased depending on the alopecia subtype. Further basic and clinical studies are required. Regular follow-up in terms of thyroid and urinary tract cancers (prostate and bladder cancers) may be necessary for alopecia patients.

Materials and Methods
Ethics approval. This study was approved by the Institutional Review Board of the Korean NHIS (no. NHIS-2017-1-138). The study design was approved by the Ethics Committee of Seoul St. Mary's Hospital, the Catholic University of Korea (approval no. KC17ZESI0312) and followed all relevant principles of the Declaration of Helsinki.
Data sources. We used the NHIS database, which contains medical information on almost all Koreans. The NHIS manages all health-related data, including checkup results, treatment details, long-term care of the elderly, institutional data, and the frequencies of cancers and rare diseases. The NHIS uses the comprehensive database of the Health Insurance and Review Agency (HIRA) 32 . The database contains patient demographics, outpatient histories, diagnoses and comorbidities based on the International Classification of Disease (ICD)−10 codes, and prescription and procedural details 33 . All patients are assigned identification numbers and all data are managed anonymously.
Study population. We retrieved data on patients who visited clinics or hospitals and received the diagnostic code (ICD-10) for alopecia (L63) more than once in any year from January 2007 to December 2014. Of these, subjects under 20 years of age or with a history of cancer during a 2005-2006 washout period were excluded. Finally, alopecia patients older than 20 years were included (n = 668,604) (Fig. 1). The control group (n = 3,343,020) was randomly selected at a 1:5 ratio; this group contained matched age-and sex-stratified subjects who were not treated for alopecia during the same period. In addition, we divided alopecia patients into two groups, as follows: 1) an AA group (ICD-10 codes L63.8 and L63.9) including all types of AA except AT/AU (ICD-10 L63); and, 2) an AT/AU group (ICD-10 codes L63.0 and L63.1).
Cancer information was also extracted from the NHIS database; we sought incident cancer reports to 2015. The Korean government maintains records of patients with cardiovascular and cerebrovascular diseases, cancers, and rare incurable diseases (RIDs), and supports such patients financially. We used the cancer code (ICD-10 C00-C96).
Statistical analysis. Baseline characteristics are presented as means with standard deviations or as numbers with percentages. We calculated cancer incidence rates (in 1,000 person-years) by dividing the number of incident cases by the total follow-up period. Cox's proportional hazards regression analysis was used to evaluate the association between alopecia presence and subtype, and cancer incidence. Overall and site-specific cancer risks of AA or AT/AU patients compared to those of the age-and sex-matched control population were expressed as hazard ratios (HRs) with 95% confidence intervals (CIs). Model 1 was adjusted for age, gender, and income level; and diabetes mellitus, hypertension, and dyslipidemia status. To avoid temporal bias, we matched the date of alopecia diagnosis with the dates of registration of control subjects. The cumulative cancer incidences by alopecia presence and subtype were calculated using Kaplan-Meier curves, and the log-rank test was employed to analyze differences among the groups. All data were analyzed with the aid of SAS version 9.4 (SAS Institute, Cary, NC, USA) and R version 3.1.0 (the R Foundation for Statistical Computing, Vienna, Austria, http://www.R-project. org) software.