Positive association between actinic keratosis and internal malignancies: a nationwide population-based cohort study

Little is known about the comorbidities in actinic keratosis patients. To evaluate the association of actinic keratosis with certain malignancies. All patients with actinic keratosis (n = 61,438) and age- and sex-matched control subjects (n = 307,190) at a 5:1 ratio were enrolled using data from the Korean National Health Insurance Service between the years 2007 and 2014. In subjects with actinic keratosis, overall cancer incidence was higher than that for controls after income level, habitat, diabetes, hypertension, and dyslipidemia were adjusted (Hazard Ratio [HR] = 1.43 [95% confidence interval 1.38–1.47]). The positive association of specific cancers were observed in the following order: skin cancer (HR = 3.43 [2.47–4.75]), oral cavity and pharyngeal cancer (HR = 1.99 [1.57–2.52]), lymphoma (HR = 1.59 [1.28–1.96]), leukemia (HR = 1.35 [1.03–1.77]), prostate cancer (HR = 1.35 [1.21–1.51]), renal cancer (HR = 1.29 [1.02–1.63]), liver cancer (HR = 1.21 [1.09–1.35]), thyroid cancer (HR = 1.20 [1.05–1.38]), and gastric cancer (HR = 1.13 [1.03–1.23]). Although further research on pathologic mechanism is needed, the implications of a positive correlation between actinic keratosis and internal organ malignancies has great significance.

www.nature.com/scientificreports/ of AK were IRF4, MC1R, and TYR 11 . However, the mechanism of progression from AK to SCC is still poorly understood.
In a systematic review, the progression rate of AK to SCC was 0-0.08% per lesion-year 12 . The rate of conversion of individual AK to skin cancer is variously reported from 0.03 to 16% 2,13-17 . Criscione et al. reported that 0.6% of patients developed a SCC in the AK field during the first year, rising to 2.6% after 4 years 18 . In a 5-year follow-up study, 65% of SCCs were found in previous AK lesions 18 . In an additional study, the probability of malignant transformation to SCC within 10 years in patients with an average of 7.7 AKs was approximately 10% 14 . Foote et al. reported that the incidences of basal cell carcinoma and SCC in adults with AK in the USA were 0.04 and 0.03 per person-year, respectively 19 . Chen et al. reported a six-fold increase in risk for skin cancer in people with AK compared with people without AK in the United States 20 .
AK is known to occur in elderly and associated with UV radiation, but there has been no research that conducted on whether to increase risks of other cancers except SCC and basal cell carcinoma. A previous study reported that the risk of internal malignancies was significantly higher in the patients with non-melanoma skin cancer patients compared with controls, including bone, nasal cavity and larynx, oral cavity and pharynx, anal, and cervical cancers 21 . We considered the possibility that AK patients might be associated with incidence of other internal malignancy due to their lifestyle or genetic susceptibilities. Studies based on large populations are needed to evaluate the association between cancers and AK. Therefore, the following study was a large-scale cohort investigation of the real incidence of cancer in AK using the Korea Nationwide Health Insurance System (KNHIS) database.

Results
Baseline characteristics of study population. We enrolled 61,438 AK patients and 307,190 age-, sex-, and index year-matched controls in this study; the mean age was 62 years in both groups, and the percentage of men in both groups was also the same, 38.8%. The percentage of Korean citizens in the lowest income level was higher in the control group (25.5%) than it was in the AK group (22.5%). There was no difference between groups in habitat, and there was no significant difference in the proportions of individuals with DM; however, hypertension and dyslipidemia were more prevalent among the AK patients than in the control group (Table 1).
Occurrence of all kinds of cancers in AK patients and controls. The occurrence of cancer was significantly higher in the AK group (4420 cancers, 7.19%) than among the controls (15,755 cancers, 5.13%) during the same observation time. The statistically significant differences were for skin, oral cavity, and pharyngeal cancer; lymphoma; leukemia; prostate, renal, liver, thyroid, gastric, and pancreatic cancer; and cancer of the central nervous system (P value less than 0.05, Table 2).  [1.14-5.12]) was markedly higher than the rate among AK patients older than age 40. AK patients aged 40-64 showed lower multiple myeloma rate (HR = 0.31 [0.10-0.98]) than did other age groups (Table 4).

Discussion
This large-scale population-based study showed overall and site-specific cancer incidence rate in Korean patients with AK. Our results showed an increased incidence rate of overall cancer, leukemia, lymphoma, oral cavity and pharyngeal cancer, prostate cancer, and thyroid cancer in AK patients compared with controls after adjusted for confounding factors.
Several researchers have reported increased skin cancer risk in AK patients 20,[22][23][24] . Although AK is considered a premalignant lesion to skin cancer, many AKs persist in the same stage or regress, while only a few cases progress into skin cancer. The key pathomechanism of AK's progression to skin cancer has not been fully understood. Until now, the number of AKs and their areas and severity indices are considered predictors of advancement to SCC. Green et al. reported that patients with more than 15 AKs had a 10-15 times greater risk of SCC than controls 25 . An AK area and severity index over 7 also suggests an increased risk of SCC transformation 26 . In this study, our findings of increased risk of skin cancer in AK patients (HR, 3.43) are consistent with previous results. A recent study reported that non-melanoma skin cancer patients has increased risk of internal malignancies, including bone, nasal cavity and larynx, oral cavity and pharynx, anus, cervix, thorax, esophagus, breast, lung, stomach cancers, thyroid gland and non-Hodgkin's lymphoma 21 .
In this study, the proportions of oral cavity and pharyngeal cancer were significantly higher in AK patients, which could be because oral SCC is related to AK of the mouth, and oral SCC is the most common oral or pharyngeal cancer 27 . Premalignant oral lesions are associated with smoking habit, oral lichen planus, oral submucous fibrosis, discoid lupus erythematosus, and AK of the mouth 28 , which is also caused by actinic damage on the www.nature.com/scientificreports/ lower lip (actinic cheilitis) and oral mucosa 29 . That common cause of UV exposure in both AK and oral SCC might explain the increased risk of oral cavity and pharyngeal cancer in AK patients.
In this study, we observed a markedly higher incidence rate of lymphoma among AK patients. A few prior investigators reported increased risk for non-Hodgkin's lymphoma in skin cancer patients 30,31 . Hu et al. found in large population-based studies this increased risk of subsequent non-Hodgkin's lymphoma in skin cancer patients and an increased risk of skin cancer in patients with non-Hodgkin's lymphoma 32 . The association between skin cancer and lymphoma is related to immunosuppression, as suggested by associations of human immunodeficiency virus (HIV) infection 33 and solid organ transplantation with skin cancer 34 . Recently, Engels et al. investigated 55 medical conditions to identify new associations with Non-Hodgkin's lymphoma and found links with HIV, solid organ transplantation, hepatitis virus, autoimmune conditions, and, interestingly, nonmelanoma skin cancer and AK 35 . The authors reported that only 3 of the 55 medical conditions were associated with increased risk for all five subtypes of non-Hodgkin's lymphoma, skin cancer, AK, and hemolytic anemia. In this study, we also found increased risk of lymphoma in AK patients.
Although an association between leukemia and AK has not been reported yet, increased skin cancer risk in chronic lymphocytic leukemia was reported, with rates approximately double those in the general population 36 . Therefore, Mulcahy et al. recommended skin cancer monitoring for patients with chronic lymphocytic leukemia 37 . The increased risk of skin cancer in leukemia patients was explained by their immunocompromised state 38 , but we found in this study an increased risk of leukemia in patients with AK compared with controls vice versa. So far, researchers have reported on several cases of leukemia cutis on AK lesions and on one case of leukemic cells observed in AK lesions and diagnosed as acute myeloid leukemia 39 . Investigators also observed infiltration of leukemic cells in AK lesions of leukemic patients 40 . In this study, it is unknown whether leukemic cell was present in the skin of AK patient who were later diagnosed with leukemia. There might have been diagnostic confusion between AK and leukemia cutis in previous reports, but we enrolled AK patients after excluding previous diagnoses of leukemia, and we still identified increased risk of leukemia in AK patients. These associations may reflect a shared predisposition to both AK and leukemia.
In this study, we observed increased incidence rate of prostate, thyroid, renal, liver, and gastric cancers, but until now, neither AK nor skin cancer has been reported to be associated with these specific cancers in the English literature. One possible explanation is that there might be a link of genetic mutations between AK and these specific cancers. Recently, Yousef et al. reported threefold lower expression of human kallikrein gene 10 (KLK10) www.nature.com/scientificreports/ in AKs compared with normal skin from the same patient, and this decreased expression was found in carcinomas of the skin and prostate. The authors observed overexpression of KLK10 in pancreatic and gastric cancers 41 . Age, male gender, skin type, and cumulative sun exposure are considered independent risk factors for AK 42 , and recently, an association between tall stature and AK has been reported in a Korean population 43 . In fact, human height has been revealed to have an 80% additive genetic contribution 44 ; several single nucleoside variations that are associated with the human height trait are reported in genome-wide association studies [45][46][47] , and tall individuals have been revealed to be susceptible to cancer including malignant melanoma, leukemia, non-Hodgkin lymphoma, rectal cancer 48 , pancreatic cancer 49 , kidney cancer 50 , prostate cancer 51,52 , and breast cancer [53][54][55][56] . A possible explanation of a relationship between height and cancer was that the several genetic variants related to the insulin-growth factor signaling pathway are also related to height 57,58 . The previous reports and the present study showed consistent results that the high-risk cancers in tall individuals accorded with high risks of those cancers in AK patients. However, no researchers have elucidated exact pathomechanisms of any links between AK and specific cancers.
There are some limitations in this study. First, it was an epidemiologic study, and therefore, we could not determine any cause-effect relationships. In addition, we did not include in our analyses all the medications or therapies that might influence the development of cancers. We also lacked information on smoking status, drinking habits, obesity, and family history of cancer and thus could not include these variables in the analyses. Despite these limitations, in this large cohort study, we found that AK patients had a higher overall incidence rate of cancer and higher specific incidence rate of leukemia, lymphoma, oral cavity and pharyngeal cancer, prostate cancer, and thyroid cancer. This study has the implications of a positive correlation between actinic keratosis and internal organ malignancies, but studies are needed on any shared pathogenesis of AK and specific cancers.

Methods
Data source and study population. The Korean government mandates the KNHIS database, which covers almost 99% of the Korean population (about 51 million). The Health Insurance and Review Agency operates this comprehensive database of all health care utilization information in Korea 59 . The KNHIS database used is publicly available and identity of participants is not revealing. The KNHIS database includes all claims data as well as a registry of rare incurable diseases such as cancer. The database also contains demographics of subjects Table 4. Subgroup analysis of association between internal malignancy and actinic keratosis by sex and age. *Adjustment for diabetes mellitus, hypertension, dyslipidemia, habitat, and income status. Study design. The design of this study was similar to that of our previous studies 61,62 . Briefly, we enrolled all the 76,885 subjects who visited clinics or hospitals more than once with an ICD-10 code of AK (L570) in a given year from January 2007 to December 2014 in the KNHIS database. We considered the year in which subjects were first diagnosed with AK to be that subject's index year. We excluded individuals who already had diagnosed with all type of cancers before the index year (n = 5654) and within a year of the index year (n = 9789) to exclude any preexisting cancers. Four individuals were excluded because we could not find age-, sex-, and index year-matched controls. Ultimately, we analyzed the data of 61,438 subjects with AK and 307,190 age-, sex-, and index year-matched controls without AK, who were randomly selected at a 5:1 ratio, during the same period. Both groups were followed up for cancer development until December 31, 2017 (Fig. 1). We collected the cancer information from the KNHIS database for patients with ICD-10 codes, C00-C96.
Comorbidities. We needed to adjust for comorbidities such as diabetes mellitus (DM), hypertension, and dyslipidemia that have been reported to increase cancer risks. We considered the presence of DM, hypertension, and dyslipidemia to be established, respectively, by ICD-10 codes E11-14, I10-13 and I15, and E78 with medication.
Statistical analysis. We calculated the cancer incidence rates (IRs) by dividing the number of incident cases by the total follow-up period presented as 1000 person-years. The censor was defined as individuals who had no cancer until the last day of follow-up (this study is December 31, 2017), or who died before the last day of this study. We used multivariate Cox regression models to assess the cancer incidence rates after we adjusted for confounding factors including income level, habitat, and the comorbidities of DM, hypertension, and dyslipidemia. We defined incidence of overall cancer as a development of any type of cancers in control group and AK group during the follow-up period. And then we analyzed the difference of incidence of each type of sitespecific cancer between the AK group and control. In some cases, two types of cancer were registered in the same patient. The claim data have limitations of not being able to determine which cancer is primary or metastatic. In www.nature.com/scientificreports/ this reason, the competitive risk regression model was not adopted in this study. Overall and site-specific cancer incidences for AK patients and controls are expressed as hazard ratio (HRs) and 95% confidence interval (CIs). We analyzed the study data using SAS version 9.4 (SAS Institute, Cary, NC, USA). A P-value less than 0.05 is considered statistically significant.