Key Points
Question
What are the incidence rate and risk factors for keratoacanthoma?
Findings
In this cohort study of 40 438 residents of Queensland, Australia, the person-based incidence rate for keratoacanthoma was 409 individuals per 100 000 person-years. Older age, male sex, UV radiation exposure–sensitive phenotypes, indications of high sun exposure (such as previous keratinocyte cancer excisions), smoking, and high alcohol use were independently associated with the development of keratoacanthoma.
Meaning
This analysis is, to date, the first large prospective cohort study to quantify the incidence rate and risk factors for keratoacanthoma, many of which are shared with other keratinocyte cancers.
Abstract
Importance
Keratoacanthoma (KA) is a common and generally benign keratinocyte skin tumor. Reports of the incidence rates of KA are scant. In addition, the risk factors for KA are not well understood, although associations with UV radiation exposure and older age have been described.
Objective
To investigate the incidence rate of KA and the risk factors for developing KA.
Design, Setting, and Participants
The study included data from 40 438 of 193 344 randomly selected residents of Queensland, Australia, who participated in the QSkin Sun and Health (QSkin) prospective population-based cohort study. All participants completed a baseline survey between 2010 and 2011 and were ages 40 to 69 years at baseline. Histopathologic reports of KA were prospectively collected until June 30, 2014, through data linkage with pathologic records. Cox proportional hazards models were used to identify risk factors associated with KA while controlling for potential confounding variables. Data were analyzed from January 2 to April 8, 2020.
Exposures
Demographic characteristics, phenotypes, UV radiation exposure, medical history, and lifestyle.
Results
Among 40 438 participants (mean [SD] age, 56 [8] years; 18 240 men [45.1%]), 596 individuals (mean [SD] age, 62 [6] years; 349 men [58.6%]) developed 776 KA tumors during a median follow-up period of 3.0 years (interquartile range, 2.8-3.3 years). The person-based age-standardized incidence rate for KA in the age-restricted cohort was 409 individuals per 100 000 person-years (based on the 2001 Australian population). Risk factors after adjustment for potential confounders were older age (age ≥60 years vs age <50 years; hazard ratio [HR], 6.38; 95% CI, 4.65-8.75), male sex (HR, 1.56; 95% CI, 1.33-1.84), fair skin (vs olive, dark, or black skin; HR, 3.42; 95% CI, 1.66-7.04), inability to tan (vs ability to tan deeply; HR, 1.69; 95% CI, 1.19-2.40), previous excisions of keratinocyte cancers (ever had an excision vs never had an excision; HR, 6.28; 95% CI, 5.03-7.83), current smoking (vs never smoking, HR, 2.02; 95% CI, 1.59-2.57), and high alcohol use (≥14 alcoholic drinks per week vs no alcoholic drinks per week; HR, 1.42; 95% CI, 1.09-1.86).
Conclusions and Relevance
This is, to date, the first large prospective population-based study to report the incidence rate and risk factors for KA. The high person-based incidence rate (409 individuals per 100 000 person-years) highlights the substantial burden of KA in Queensland, Australia. Furthermore, the study’s findings suggest that older age (≥60 years), male sex, UV radiation–sensitive phenotypes, indications of high sun exposure (eg, previous keratinocyte cancer excisions), smoking, and high alcohol use are independent risk factors for the development of KA.
This cohort study uses data from the QSkin Sun and Health study to assess the incidence rate and risk factors for keratoacanthoma among residents of Queensland, Australia.
Introduction
Keratoacanthoma (KA) is a common rapidly growing skin tumor.1,2,3,4,5 Although KA has some similarities to cutaneous squamous cell carcinoma (cSCC) with regard to histopathologic characteristics and clinical appearance, it is considered a separate diagnostic entity.6 It has long been debated whether KA tumors represent benign or malignant lesions.4,7 Several case reports have noted metastasis8,9; however, a 2014 systematic review described no distant metastases in 445 KA tumors and spontaneous resolution of 52 of these tumors.1 Previous studies on the incidence of KA are scarce and out of date.10,11,12,13,14,15 The most current data originate from a study conducted in Hawaii in the 1980s, which used a small sample of patients with histologically confirmed KA to calculate an incidence rate of 104 individuals per 100 000 person-years.10
The etiologic factors of KA are understudied, with relatively few epidemiologic studies and no large prospective studies that have captured information on exposures at baseline. The extant knowledge derives largely from case series and case-control studies, and UV radiation exposure and older age are the risk factors most frequently implicated.10,14,16,17 Although inconclusive, the suggested association with exposure to UV radiation is based on data indicating that KA commonly occurs on the limbs,10,14,16 that patients with fair vs darker skin have a higher frequency of tumors,12 that multiple KA tumors have developed after medical phototherapy,18,19 and that people with xeroderma pigmentosum have a high prevalence of KA.20,21 Other risk factors that have been associated with KA development include male sex,10,14,16 smoking,16,22 immunosuppression,23 receipt of drugs that have consequences for the cell cycle,24 human papillomavirus,5,25 trauma to the skin,26 and exposure to carcinogens, such as tar.22
To bridge the knowledge gap regarding the incidence and etiologic factors of KA, we aimed to calculate the incidence rate and investigate the risk factors associated with KA in a large prospective study of participants from the general population.
Methods
Study Cohort and Data Collection
We used data from the QSkin Sun and Health (QSkin) study, which was initiated in 2010 to investigate the development of skin cancer among a prospective population-based cohort of 193 344 randomly sampled residents of Queensland, Australia, which had a population of 4.5 million people at the time of enrollment. A description of the cohort has been published previously.27 At inclusion (2010-2011), 43 794 people aged 40 to 69 years completed a baseline survey. Of those, 40 438 participants provided written informed consent for prospective linkage of their data to pathology laboratories, public hospital databases, and health administration data stored by Medicare Australia. Medicare Australia is the universal national health insurance program, and its database includes information about all medical services provided outside of the public hospital system. All participants provided written informed consent, and the study was approved by the human research ethics committee of the QIMR Berghofer Medical Research Institute in Brisbane, Australia. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cohort studies.
Participants completed the baseline survey in writing or online; this survey included questions about sociodemographic characteristics, medications received, phenotypes, previous exposure to UV radiation, sun protection behavior, lifestyle (eg, alcohol use and smoking status), and medical history. A copy of the survey can be found online.28 Items related to medical history included the self-reported number of skin cancers treated by excision and the self-reported number of actinic keratoses or skin cancers treated by destruction (ie, burnt or frozen off). We assumed that the predominant share of lesions treated by excision were keratinocyte cancers and that lesions treated using other modalities were predominantly actinic keratoses. In a test-retest setting, the measures of agreement for self-reported survey items regarding phenotype and past medical history were good to excellent.29 We used health administration data from Medicare Australia to identify participants who were treated for skin cancer excisions, and we reviewed histologic diagnoses through linkage with pathologic records. In all, 5 pathology laboratories provided reports for our study. Our outcome for the primary analysis was a confirmed histologic diagnosis of KA, which was identified through manual review of the pathologic report. In the pathologic reports, KA was commonly described as a crateriform well-differentiated squamoproliferative lesion with signs of regression, similar to the diagnostic criteria in the WHO Classification of Skin Tumours.6 Our analysis included 8 KA tumors with a cSCC occurring from within the lesion. Histopathologic reports of KA were prospectively collected until June 30, 2014.
Although only a small proportion (<3%) of data were missing for most variables, information about sunburns in early life and educational level were missing for approximately 9% and 7% of the participants, respectively. To avoid bias in the estimate owing to missing data, we imputed all missing values using a multiple imputation model. We assumed data were missing at random. In the imputation step, we included all phenotypic, UV radiation exposure, and lifestyle variables as well as the outcome variable,30 and we used logistic regression analysis to impute ordinal variables. For the multiple imputation model, we used the PROC MI procedure in SAS software, version 9.4 (SAS Institute). Imputations were performed for 50 cycles, generating 50 imputed data sets.
Statistical Analysis
We estimated unadjusted and age-standardized person-based incidence rates for KA using the direct standardization method. In this method, the incidence rates of each 5-year age band in the study sample were weighted by the distribution of the general Australian population in 2001 and the general US population in 2000, respectively. We used a time to event analysis, with the primary outcome being the first incident KA. For participants with KA, person-time was calculated as the time from consent at baseline (2010-2011) to the time of the first KA diagnosis. For participants without KA, person-time was calculated as the time from consent at baseline to the date of death (obtained through data linkage with the National Death Index at the Australian Institute of Health and Welfare) or the date of the last follow-up (June 30, 2014), whichever occurred first. We used Cox proportional hazards models to estimate hazard ratios (HRs) and 95% CIs for each imputed data set. We then obtained pooled estimates for each risk factor from 50 imputed models using the Rubin rule to identify factors associated with KA while accounting for potential confounding variables.31 The proportional hazards assumption was assessed by examining Kaplan-Meier curves and was met for all identified risk factors.
To select potential risk factors for KA and their confounders, we used directed acyclic graphs.32 The directed acyclic graph for the association between lifestyle variables and KA is provided as an example in the eFigure in the Supplement. We broadly grouped variables by (1) phenotype (skin color, sunburn tendency, tanning ability, eye color, hair color, and presence of freckles at age 21 years), (2) exposure to UV radiation (sunburns before age 10 years, sunburns between ages 10 and 20 years, sunburns after age 20 years, and previous actinic keratoses or keratinocyte cancers as proxies for chronic sun exposure), and (3) lifestyle (smoking and alcohol use). First, we assessed each variable within the phenotype group as a risk factor, adjusting for age, sex, and all other phenotypic factors. Only those phenotypic variables that changed the effect estimate by 10% or more were retained in this model. We then adjusted for exposure to UV radiation to calculate the association of phenotype with the risk of KA (ie, not mediated through sun exposure). The UV radiation exposure variables were only retained in the final models if they changed the effect estimate of the risk factor under investigation by 10% or more. As a result, the set of adjusted variables was different for each risk factor. For each lifestyle factor, we adjusted for age, sex, and educational level.
To examine the consequences of potential case misclassification at baseline, we conducted a sensitivity analysis that excluded participants (n = 16 359) who reported 1 or more excision for skin cancer or sunspots before baseline. For all analyses, we used SAS software, version 9.4 (SAS Institute). Data were analyzed from January 2 to April 8, 2020.
Results
Among 40 438 participants in the cohort, the mean (SD) age at baseline was 56 (8) years, and 18 240 participants (45.1%) were men. Of those, 596 participants (1.5%; mean [SD] age, 62 [6] years; 349 men [58.6%]) developed 776 KA tumors during a median follow-up period of 3.0 years (interquartile range, 2.8-3.3 years). The unadjusted person-based incidence rate per 100 000 person-years of KA was 494 individuals, and the corresponding age-standardized rates were 409 individuals (based on the Australian population in 2001) and 398 individuals (based on the US population in 2000). After mutual adjustment, the risk estimates for sex and age indicated an HR of 1.56 (95% CI, 1.33–1.84) for male vs female sex, an HR of 6.38 (95% CI, 4.65-8.75) for individuals 60 years or older vs individuals younger than 50 years, and an HR of 2.79 (95% CI, 2.00-3.90) for individuals aged 50 to 59 years vs individuals younger than 50 years.
Table 1, Table 2, and Table 3 present the distribution of phenotypes, exposure to UV radiation, and lifestyle variables, respectively, along with the risk estimates for the association between each variable and KA. Tendency to sunburn, eye color, hair color, and self-reported sunburns after age 10 years were associated with KA in the age- and sex-adjusted analysis but were not statistically significant after final adjustment (Table 1 and Table 2). In our final models, we found associations between KA and the following variables: phenotype (skin color, tanning ability, and the presence of freckles at age 21 years) (Table 1), UV radiation exposure (sunburns before age 10 years and treatment for previous actinic keratoses or keratinocyte cancers) (Table 2), and lifestyle (smoking status and number of alcoholic drinks per week) (Table 3). Fair skin compared with olive, dark, or black skin was associated with a more than 3-fold increase (HR, 3.42; 95% CI, 1.66-7.04) in the risk of KA, and the inability to tan compared with the ability to tan deeply was associated with a more than 1.5-fold increase (HR, 1.69; 95% CI, 1.19–2.40) in the risk of KA. Participants with many freckles compared with no freckles at age 21 years also had a higher likelihood of developing KA (HR, 1.44; 95% CI, 1.06-1.95). Individuals who reported 11 or more sunburns compared with those who reported no sunburns before age 10 years had a higher risk of KA (HR, 1.35; 95% CI, 1.01-1.81), and those who had previous excisions of keratinocyte cancers were more than 6 times as likely to develop KA than those who did not have previous excisions (HR, 6.28; 95% CI, 5.03-7.83). Participants who currently smoked had a 2-fold increased risk of developing KA compared with those who never smoked (HR, 2.02; 95% CI, 1.59–2.57), and those who consumed 14 or more alcoholic drinks per week had an increased risk of KA compared with those who consumed no alcoholic drinks per week (HR, 1.42; 95% CI, 1.09–1.86).
Table 1. Association Between Phenotypic Variables and Keratoacanthoma.
| Variable | Participants, No. (%) (N = 40 438)a | Hazard ratio (95% CI) | |||
|---|---|---|---|---|---|
| Without keratoacanthoma (n = 39 842) | With keratoacanthoma (n = 596) | Adjusted for age and sex | Adjusted for age, sex, and phenotypeb | Adjusted for age, sex, phenotype, and UV radiation exposurec | |
| Skin color | |||||
| Olive, dark, or black | 3168 (8.0) | 8 (1.3) | 1 [Reference] | 1 [Reference] | 1 [Reference] |
| Medium | 12 736 (32.0) | 109 (18.3) | 3.38 (1.65-6.94) | 3.00 (1.45-6.19) | 2.45 (1.19-5.07) |
| Fair | 23 716 (59.5) | 477 (80.0) | 8.09 (4.02-16.26) | 4.69 (2.29-9.64) | 3.42 (1.66-7.04) |
| Missing | 222 (0.6) | 2 (0.3) | NA | NA | NA |
| Sunburn tendency | |||||
| Not burn | 3600 (9.0) | 31 (5.2) | 1 [Reference] | 1 [Reference] | 1 [Reference] |
| Burn a little | 17 149 (43.0) | 190 (31.9) | 1.41 (0.97-2.07) | 1.07 (0.73-1.57) | 0.93 (0.63-1.37) |
| Burn moderately | 13 339 (33.5) | 224 (37.6) | 2.35 (1.61-3.43) | 1.33 (0.90-1.97) | 1.04 (0.69-1.54) |
| Burn badly | 5521 (13.9) | 147 (24.7) | 4.15 (2.81-6.12) | 1.67 (1.09-2.55) | 1.25 (0.81-1.92) |
| Missing | 233 (0.6) | 4 (0.7) | NA | NA | NA |
| Tanning ability | |||||
| Tan deeply | 9172 (23.0) | 72 (12.1) | 1 [Reference] | 1 [Reference] | 1 [Reference] |
| Tan moderately | 19 529 (49.0) | 250 (41.9) | 1.61 (1.24-2.10) | 1.22 (0.93-1.60) | 1.18 (0.90-1.55) |
| Tan a little | 8289 (20.8) | 175 (29.4) | 2.78 (2.11-3.66) | 1.60 (1.19-2.15) | 1.41 (1.05-1.90) |
| Not tan | 2543 (6.4) | 93 (15.6) | 4.82 (3.54-6.58) | 2.05 (1.44-2.91) | 1.69 (1.19-2.40) |
| Missing | 309 (0.8) | 6 (1.0) | NA | NA | NA |
| Eye color | |||||
| Brown or black | 9375 (23.5) | 84 (14.1) | 1 [Reference] | 1 [Reference] | 1 [Reference] |
| Green or hazel | 14 793 (37.1) | 221 (37.1) | 1.67 (1.30-2.15) | 1.23 (0.95-1.59) | 1.16 (0.90-1.50) |
| Blue or gray | 15 138 (38.0) | 287 (48.2) | 2.00 (1.57-2.55) | 1.27 (0.98-1.64) | 1.22 (0.94-1.57) |
| Missing | 536 (1.3) | 4 (0.7) | NA | NA | NA |
| Hair color | |||||
| Dark brown or black | 17 162 (43.1) | 199 (33.4) | 1 [Reference] | 1 [Reference] | 1 [Reference] |
| Light brown | 14 670 (36.8) | 223 (37.4) | 2.76 (2.10-3.62) | 1.04 (0.85-1.27) | 1.02 (0.83-1.24) |
| Blonde | 5542 (13.9) | 99 (16.6) | 1.74 (1.36-2.22) | 1.18 (0.91-1.52) | 1.12 (0.87-1.45) |
| Red or auburn | 2242 (5.6) | 71 (11.9) | 1.36 (1.12-1.65) | 1.05 (0.78-1.43) | 1.05 (0.77-1.42) |
| Missing | 226 (0.6) | 4 (0.7) | NA | NA | NA |
| Freckles at age 21 y | |||||
| None | 18 496 (46.4) | 222 (37.2) | 1 [Reference] | 1 [Reference] | 1 [Reference] |
| A few | 12 517 (31.4) | 163 (27.3) | 1.32 (1.08-1.62) | 1.04 (0.85-1.28) | 0.90 (0.73-1.12) |
| Some | 6207 (15.6) | 131 (22.0) | 2.29 (1.84-2.86) | 1.48 (1.18-1.87) | 1.18 (0.93-1.49) |
| Many | 2402 (6.0) | 77 (12.9) | 3.71 (2.85-4.83) | 1.91 (1.41-2.58) | 1.44 (1.06-1.95) |
| Missing | 220 (0.6) | 3 (0.5) | NA | NA | NA |
Abbreviation: NA, not applicable.
Percentages may not total 100% because of rounding.
The model included additional measures of phenotype. Skin color was adjusted for age, sex, sunburn tendency, tanning ability, and the presence of freckles at age 21 years. Sunburn tendency was adjusted for age, sex, skin color, tanning ability, eye color, hair color, and the presence of freckles at age 21 years. Tanning ability was adjusted for age, sex, skin color, sunburn tendency, eye color, hair color, and the presence of freckles at age 21 years. Eye color was adjusted for age, sex, skin color, sunburn tendency, tanning ability, hair color, and the presence of freckles at age 21 years. Hair color was adjusted for age, sex, skin color, sunburn tendency, tanning ability, eye color, and the presence of freckles at age 21 years. The presence of freckles at age 21 years was adjusted for age, sex, skin color, sunburn tendency, tanning ability, eye color, and hair color.
The model included additional UV radiation exposure variables. Skin color was adjusted for age, sex, sunburn tendency, tanning ability, the presence of freckles at age 21 years, and previous actinic keratoses or keratinocyte cancers that were destroyed or excised. Sunburn tendency was adjusted for age; sex; skin color; tanning ability; eye color; hair color; the presence of freckles at age 21 years; sunburns before age 10 years, sunburns between ages 10 and 20 years, and sunburns after age 20 years; and previous actinic keratoses or keratinocyte cancers that were destroyed or excised. Tanning ability was adjusted for age, sex, skin color, sunburn tendency, eye color, hair color, the presence of freckles at age 21 years, and previous actinic keratoses or keratinocyte cancers that were destroyed or excised. Eye color was adjusted for age, sex, skin color, sunburn tendency, tanning ability, hair color, the presence of freckles at age 21 years, and previous actinic keratoses and keratinocyte cancers that were destroyed or excised. Hair color was adjusted for age; sex; skin color; sunburn tendency; tanning ability; eye color; the presence of freckles at age 21 years; sunburns before age 10 years, sunburns between ages 10 and 20 years, and sunburns after age 20 years; and previous actinic keratoses or keratinocyte cancers that were destroyed or excised. The presence of freckles at age 21 years was adjusted for age, sex, skin color, sunburn tendency, tanning ability, eye color, hair color, sunburns before age 10 years, sunburns between ages 10 and 20 years, and sunburns after age 20 years; and previous actinic keratoses or keratinocyte cancers that were destroyed or excised.
Table 2. Association Between UV Radiation Exposure Variables and Keratoacanthoma.
| Variable | Participants, No. (%) (N = 40 438)a | Hazard ratio (95% CI) | ||
|---|---|---|---|---|
| Without keratoacanthoma (n = 39 842) | With keratoacanthoma (n = 596) | Adjusted for age and sex | Adjusted for age, sex, and phenotypeb | |
| Sunburns before age 10 y, No. | ||||
| Never | 7792 (19.6) | 80 (13.4) | 1 [Reference] | 1 [Reference] |
| 1-5 | 16 326 (41.0) | 217 (36.4) | 1.36 (1.06-1.76) | 1.08 (0.83-1.39) |
| 6-10 | 6316 (15.9) | 115 (19.3) | 1.93 (1.45-2.57) | 1.34 (1.00-1.79) |
| ≥11 | 5701 (14.3) | 120 (20.1) | 2.24 (1.69-2.96) | 1.35 (1.01-1.81) |
| Missing | 3707 (9.3) | 64 (10.7) | NA | NA |
| Sunburns between ages 10-20 y, No. | ||||
| Never | 2016 (5.1) | 21 (3.5) | 1 [Reference] | 1 [Reference] |
| 1-5 | 16 633 (41.7) | 210 (35.2) | 1.46 (0.94-2.27) | 1.04 (0.67-1.63) |
| 6-10 | 9471 (23.8) | 147 (24.7) | 1.91 (1.22-3.01) | 1.21 (0.77-1.92) |
| ≥11 | 10 342 (26.0) | 198 (33.2) | 2.45 (1.57-3.82) | 1.39 (0.89-2.20) |
| Missing | 1380 (3.5) | 20 (3.4) | NA | NA |
| Sunburns after age 20 y, No. | ||||
| Never | 6311 (15.8) | 100 (16.8) | 1 [Reference] | 1 [Reference] |
| 1-5 | 20 110 (50.5) | 263 (44.1) | 0.90 (0.72-1.14) | 0.85 (0.67-1.07) |
| 6-10 | 6310 (15.8) | 93 (15.6) | 1.02 (0.77-1.35) | 0.90 (0.68-1.20) |
| ≥11 | 5332 (13.4) | 101 (16.9) | 1.25 (0.95-1.65) | 1.03 (0.78-1.36) |
| Missing | 1779 (4.5) | 39 (6.5) | NA | NA |
| Previous destroyed actinic keratoses or keratinocyte cancersc | ||||
| Never | 18 082 (45.4) | 66 (11.1) | 1 [Reference] | NA |
| Ever | 21 550 (54.1) | 526 (88.3) | 1.62 (1.36-1.88) | NA |
| Missing | 210 (0.5) | 4 (0.7) | NA | NA |
| Previous excised keratinocyte cancersc | ||||
| Never | 23 983 (60.2) | 96 (16.1) | 1 [Reference] | NA |
| Ever | 15 576 (39.1) | 494 (82.9) | 6.28 (5.03-7.83) | NA |
| Missing | 283 (0.7) | 6 (1.0) | NA | NA |
Abbreviation: NA, not applicable.
Percentages may not total 100% because of rounding.
The model included other measures of phenotype. Sunburn before age 10 years was adjusted for age, sex, skin color, sunburn tendency, tanning ability, eye color, hair color, and the presence of freckles at age 21 years. Sunburn between ages 10 and 20 years was adjusted for age, sex, skin color, sunburn tendency, tanning ability, eye color, hair color, and the presence of freckles at age 21 years. Sunburn after age 20 years was adjusted for age, sex, skin color, sunburn tendency, tanning ability, eye color, hair color, and the presence of freckles at age 21 years.
The original variable was split into 2 variables to calculate the separate risk estimates for previous destroyed lesions (predominantly actinic keratoses) and previous excised lesions (predominantly keratinocyte cancers). The risk estimates changed minimally (<10%) when adjusted for other measures of phenotype; hence, these estimates were not adjusted.
Table 3. Association Between Lifestyle Variables and Keratoacanthoma.
| Variable | Participants, No. (%) (N = 40 438)a | Adjusted for sex and educational level, hazard ratio (95% CI)b | |
|---|---|---|---|
| Without keratoacanthoma (n = 39 842) | With keratoacanthoma (n = 596) | ||
| Smoking status | |||
| Never | 21 861 (54.9) | 282 (47.3) | 1 [Reference] |
| Past | 14 123 (35.4) | 219 (36.7) | 1.02 (0.85-1.22) |
| Current | 3700 (9.3) | 92 (15.4) | 2.02 (1.59-2.57) |
| Missing | 158 (0.4) | 3 (0.5) | NA |
| Alcoholic drinks, No./wk | |||
| None | 7647 (19.2) | 101 (16.9) | 1 [Reference] |
| <1 | 6638 (16.7) | 80 (13.4) | 1.04 (0.78-1.40) |
| 2-13 | 18 713 (47.0) | 271 (45.5) | 1.18 (0.93-1.48) |
| ≥14 | 6633 (16.6) | 140 (23.5) | 1.42 (1.09-1.86) |
| Missing | 211 (0.5) | 4 (0.7) | NA |
Abbreviation: NA, not applicable.
Percentages may not total 100% because of rounding.
The model included age, sex, and educational level per the directed acyclic graph in the eFigure in the Supplement. Smoking status was adjusted for age, sex, and educational level. The number of alcoholic drinks per week was adjusted for age, sex, and educational level. There was no change in the risk estimates when they were also adjusted for phenotype or UV exposure measures; hence, these estimates were not adjusted.
In the cohort of participants who did not report having previous excisions for skin cancer or sunspots before baseline, 96 were diagnosed with 1 or more KA tumor during the follow-up period, and our risk factor findings were essentially unchanged. The association between KA and phenotypic variables, UV radiation exposure variables, and lifestyle variables among this cohort are shown in eTable 1, eTable 2, and eTable 3 in the Supplement, respectively.
Discussion
To our knowledge, we have conducted the first large-scale analysis of the incidence and risk factors for KA within a population-based prospective study with almost complete follow-up data. In this predominantly fair-skinned population who reside in an environment of high year-round sun exposure, we observed a high incidence of KA (409 individuals per 100 000 person-years). Furthermore, we found that risk factors (such as sun-sensitive skin and indications of high exposure to the sun) that are known to be associated with other forms of skin cancer, particularly cSCC, are also risk factors for KA.
The incidence of KA in our study is 2.5 times higher than that documented in a 1979 study conducted in 3 Australian pathology departments, which found an incidence of approximately 150 cases per 100 000.14,15 The most current data on the incidence of KA originate from a study performed between 1983 and 1987 in Hawaii, in which 53 people had histologically confirmed KA tumors; the person-based incidence rate (which was age and sex standardized to White individuals in the US in 1970) in the ethnically mixed population of all ages was 104 individuals per 100 000 person-years.10 However, these incidence rate estimates are not directly comparable with ours because of the different sampling frame, ethnic makeup, and age range of the study populations. The incidence of KA that we have reported (409 individuals per 100 000 person-years) is approximately one-third of the person-based incidence of cSCC (1270 individuals per 100 000 person-years, age standardized to the Australian population in 2001) observed in the QSkin cohort (N. Pandeya, MMedSc, PhD, QIMR Berhofer Medical Research Institute, email, March 31, 2020). The incidence of keratinocyte cancer in the QSkin study is slightly higher than that of the general population of Queensland, which is likely owing to self-selection (ie, individuals with a higher innate risk of skin cancer were more likely to participate in the study); however, the internal validity of the study is high, and our findings highlight the substantial burden of KA compared with cSCC.
Our findings indicate that older age and male sex are risk factors for KA development, as is the case for other keratinocyte cancers.33,34 We are not aware of any previous literature that has reported the risk of KA by age or sex. Compared with those younger than 50 years, participants in our study who were 60 years or older had a 6-fold greater risk of developing KA. Other studies have reported the age at diagnosis to range between 64 and 71 years.10,14,16,17 In our study, men had an approximately 1.5-fold higher risk of developing KA compared with women. Other studies have reported male to female ratios ranging from 1.2:1 to 2:1.10,14,16
High exposure to UV radiation is known to be the primary factor in the development of keratinocyte cancers.33,34,35 In cSCC carcinogenesis, for example, exposure to UV radiation is associated with DNA mutations in the transformation-related protein 53 (TP53; OMIM 191170) tumor suppressor gene.35 Furthermore, people with UV radiation–sensitive phenotypes have an increased risk of developing basal cell carcinoma and cSCC.35,36 The present study provides data indicating that exposure to UV radiation and phenotype are also important etiologic factors of KA, which is consistent with previous studies.10,14,18,37,38 After final adjustment, we found that phenotypic characteristics, such as the inability to tan and the presence of many freckles, as well as the 3-fold increase in risk among people with fair skin compared with people with olive, dark, or black skin, were risk factors. A meta-analysis of risk factors for basal cell carcinoma found associations with skin phenotype, color, and freckling.39 In cSCC, increased risk among people with sun-sensitive skin types and freckles has also been reported.36 We found that a tendency to sunburn is associated with the risk of developing KA, but the lack of association after final adjustment suggests that this risk is mediated by exposure to UV radiation. The dominant factors of previous UV radiation exposure included sunburns before age 10 years and a history of excised keratinocyte cancers, with the latter indicating a more than 6-fold increase in risk compared with no history of excised keratinocyte cancers. A European study36 reported somewhat higher risk estimates for childhood sunburns (6-10 sunburns vs no sunburns) in the development of basal cell carcinoma (odds ratio, 2.33; 95% CI, 1.62-3.36) and cSCC (odds ratio, 2.32; 95% CI, 1.46-3.70).
We observed an association between tobacco smoking and KA. There are several possible biological explanations for the association between smoking and the development of KA, including the suppression of the immune system by nicotine and the downregulation of the Notch tumor suppressor function that is associated with toxic tobacco components.40,41 In addition, 2 meta-analyses have indicated that smoking may be associated with an increased risk of cSCC.42,43 In a 2017 analysis of the QSkin cohort,44 individuals who currently smoked had more than twice the risk of developing cSCC compared with those who had never smoked (HR, 2.30; 95% CI, 1.46-3.62), which is consistent with the estimate we report for KA. Two small case-control studies have investigated the association between smoking and KA. One study compared individuals who had ever smoked with those who had never smoked, reporting an odds ratio of 9.1 (95% CI, 4.9-17.1; P < .01).16 In an age-matched case-control study conducted in 1963,22 the unadjusted odds ratio for the association between smoking and KA was 1.65 (95% CI, 1.07-2.53). We observed an association between high alcohol use and KA, which was consistent with a recent meta-analysis reporting that alcohol use was associated with the risk of both basal cell carcinoma and cSCC in a dose-dependent manner.45 Plausible biological factors underpinning the association include the tumorigenic consequences of the ethanol metabolite acetaldehyde and the general immunosuppressive consequences of ethanol.45
The QSkin study was designed to examine risk factors for skin cancer. This design enabled analysis of a large sample of people who are representative of the population of Queensland, Australia. Moreover, the QSkin cohort is a prospective study with almost complete follow-up data. Through the baseline surveys, we had access to information regarding a range of potential risk factors and their potential confounders. In the models, we carefully selected covariates, and we were informed by directed acyclic graphs.
Limitations
This study has several limitations. One is the use of self-reported survey data; however, we found high repeatability for most of the self-reported variables included.29 Participants in the study were ages 40 to 69 years. Restricting the cohort to this age group likely produced an underestimation of the overall incidence of KA despite standardization. We used histopathologic reports rather than less reliable self-reported data to confirm the diagnoses of KA. Although it is possible that some participants without a histologic diagnosis may have developed tumors that resolved spontaneously, we had no means of estimating the incidence of regression. Histopathologic examination cannot always accurately differentiate KA tumors from cSCC tumors.46 The challenge in diagnosing these tumors is reflected in a study from Ireland and Great Britain, which reported large regional disparities in the diagnosis ratio for cSCC to KA, which varied from 2.5:1 to 139:1.47 Other population-based studies have reported a ratio of 1:1 to 2:1.10,14 It is possible that some of the KA tumors in our study were misdiagnosed cSCC tumors (eg, the follicular infundibular variant). We may also have inadvertently excluded KA tumors that were misdiagnosed as cSCC tumors.48 Treatment of melanoma metastasis with BRAF gene (OMIM 164757) inhibitors, such as vemurafenib, can induce the development of KA.24 We investigated linked data for prescribed medications among the participants with KA in our study and did not find any listings for vemurafenib. Thus, it is unlikely that the incidence of KA was associated with this treatment.
We found that the etiologic factors of KA are similar to those of other keratinocyte cancers, especially to those of cSCC.35 Our findings indicate that UV radiation exposure is likely associated with KA. Furthermore, our study reports the importance of a sun-sensitive phenotype in the development of KA, which is comparable with known risk factors for cSCC.35 Smoking and alcohol use have been associated with an increased risk of developing cSCC, which is consistent with our findings for KA.
Conclusions
We report a high incidence of KA (409 individuals per 100 000 person-years) among residents of Queensland, Australia. We identified older age (≥60 years), male sex, phenotypic characteristics that indicate increased sensitivity to UV radiation, and indications of high sun exposure (eg, previous excisions of keratinocyte cancers) as independent risk factors for KA. In addition, we found that smoking and high alcohol use were associated with KA. To our knowledge, this is the first large prospective population-based study to report that many of the risk factors for KA are shared with other keratinocyte cancers.
eTable 1. Sensitivity Analysis of Phenotype Variables and Their Association With Keratoacanthoma Among Restricted Cohort of 24 079 Participants
eTable 2. Sensitivity Analysis of Ultraviolet Exposure Variables and Their Association With Keratoacanthoma Among Restricted Cohort of 24 079 Participants
eTable 3. Sensitivity Analysis of Lifestyle Variables and Their Association With Keratoacanthoma Among Restricted Cohort of 24 079 Participants
eFigure. Directed Acyclic Graph for the Association Between Lifestyle Variables and Keratoacanthoma
References
- 1.Savage JA, Maize JC Sr. Keratoacanthoma clinical behavior: a systematic review. Am J Dermatopathol. 2014;36(5):422-429. doi: 10.1097/DAD.0000000000000031 [DOI] [PubMed] [Google Scholar]
- 2.Lai V, Cranwell W, Sinclair R. Epidemiology of skin cancer in the mature patient. Clin Dermatol. 2018;36(2):167-176. doi: 10.1016/j.clindermatol.2017.10.008 [DOI] [PubMed] [Google Scholar]
- 3.Kwiek B, Schwartz RA. Keratoacanthoma (KA): an update and review. J Am Acad Dermatol. 2016;74(6):1220-1233. doi: 10.1016/j.jaad.2015.11.033 [DOI] [PubMed] [Google Scholar]
- 4.Gleich T, Chiticariu E, Huber M, Hohl D. Keratoacanthoma: a distinct entity? Exp Dermatol. 2016;25(2):85-91. doi: 10.1111/exd.12880 [DOI] [PubMed] [Google Scholar]
- 5.Schwartz RA. Keratoacanthoma. J Am Acad Dermatol. 1994;30(1):1-19. doi: 10.1016/S0190-9622(94)70001-X [DOI] [PubMed] [Google Scholar]
- 6.Elder DE, Massi D, Scolyer RA, Willemze R, eds; WHO Classification of Tumours Group. Vol. 11 of WHO Classification of Skin Tumours. 4th ed. World Health Organization; 2018. [Google Scholar]
- 7.Sánchez Yus E, Simón P, Requena L, Ambrojo P, de Eusebio E. Solitary keratoacanthoma: a self-healing proliferation that frequently becomes malignant. Am J Dermatopathol. 2000;22(4):305-310. doi: 10.1097/00000372-200008000-00002 [DOI] [PubMed] [Google Scholar]
- 8.Hodak E, Jones RE, Ackerman AB. Solitary keratoacanthoma is a squamous-cell carcinoma: three examples with metastases. Am J Dermatopathol. 1993;15(4):332-342. doi: 10.1097/00000372-199308000-00007 [DOI] [PubMed] [Google Scholar]
- 9.Piscioli F, Boi S, Zumiani G, Cristofolini M. A gigantic, metastasizing keratoacanthoma. Report of a case and discussion on classification. Am J Dermatopathol. 1984;6(2):123-129. doi: 10.1097/00000372-198404000-00003 [DOI] [PubMed] [Google Scholar]
- 10.Chuang TY, Reizner GT, Elpern DJ, Stone JL, Farmer ER. Keratoacanthoma in Kauai, Hawaii. the first documented incidence in a defined population. Arch Dermatol. 1993;129(3):317-319. doi: 10.1001/archderm.1993.01680240057005 [DOI] [PubMed] [Google Scholar]
- 11.Reizner GT, Chuang TY, Elpern DJ, Stone JL, Farmer ER. Basal cell carcinoma and keratoacanthoma in Hawaiians: an incidence report. J Am Acad Dermatol. 1993;29(5 Pt 1):780-782. doi: 10.1016/S0190-9622(08)81701-X [DOI] [PubMed] [Google Scholar]
- 12.Reizner GT, Chuang TY, Elpern DJ, Stone JL, Farmer ER. Keratoacanthoma in Japanese Hawaiians in Kauai, Hawaii. Int J Dermatol. 1995;34(12):851-853. doi: 10.1111/j.1365-4362.1995.tb04420.x [DOI] [PubMed] [Google Scholar]
- 13.Whiting DA. Skin tumours in White South Africans. part I. patients, methods and incidence. S Afr Med J. 1978;53(3):98-102. [PubMed] [Google Scholar]
- 14.Sullivan JJ, Colditz GA. Keratoacanthoma in a sub-tropical climate. Australas J Dermatol. 1979;20(1):34-40. doi: 10.1111/j.1440-0960.1979.tb00122.x [DOI] [PubMed] [Google Scholar]
- 15.Sullivan JJ. Keratoacanthoma: the Australian experience. Australas J Dermatol. 1997;38(suppl 1):S36-S39. doi: 10.1111/j.1440-0960.1997.tb01007.x [DOI] [PubMed] [Google Scholar]
- 16.Miot HA, Miot LDB, da Costa ALB, Matsuo CY, Stolf HO, Marques MEA. Association between solitary keratoacanthoma and cigarette smoking: a case-control study. Dermatol Online J. 2006;12(2):2. [PubMed] [Google Scholar]
- 17.Vergilis-Kalner IJ, Kriseman Y, Goldberg LH. Keratoacanthomas: overview and comparison between Houston and Minneapolis experiences. J Drugs Dermatol. 2010;9(2):117-121. [PubMed] [Google Scholar]
- 18.Craddock KJ, Rao J, Lauzon GJ, Tron VA. Multiple keratoacanthomas arising post-UVB therapy. J Cutan Med Surg. 2004;8(4):239-243. doi: 10.1177/120347540400800407 [DOI] [PubMed] [Google Scholar]
- 19.Chuang TY, Heinrich LA, Schultz MD, Reizner GT, Kumm RC, Cripps DJ. PUVA and skin cancer. a historical cohort study on 492 patients. J Am Acad Dermatol. 1992;26(2 Pt 1):173-177. doi: 10.1016/0190-9622(92)70021-7 [DOI] [PubMed] [Google Scholar]
- 20.Bhutto AM, Shaikh A, Nonaka S. Incidence of xeroderma pigmentosum in Larkana, Pakistan: a 7-year study. Br J Dermatol. 2005;152(3):545-551. doi: 10.1111/j.1365-2133.2004.06311.x [DOI] [PubMed] [Google Scholar]
- 21.Baykal C, Topkarci Z, Polat Ekinci A. Management of keratoacanthoma in patients with xeroderma pigmentosum: a challenge for clinicians. J Eur Acad Dermatol Venereol. 2016;30(10):e91-e93. doi: 10.1111/jdv.13337 [DOI] [PubMed] [Google Scholar]
- 22.Ghadially FN, Barton BW, Kerridge DF. The etiology of keratoacanthoma. Cancer. 1963;16:603-611. doi: [DOI] [PubMed] [Google Scholar]
- 23.Walder BK, Robertson MR, Jeremy D. Skin cancer and immunosuppression. Lancet. 1971;2(7737):1282-1283. doi: 10.1016/S0140-6736(71)90602-7 [DOI] [PubMed] [Google Scholar]
- 24.Chapman PB, Hauschild A, Robert C, et al. ; BRIM-3 Study Group . Improved survival with vemurafenib in melanoma with BRAF V600E mutation. N Engl J Med. 2011;364(26):2507-2516. doi: 10.1056/NEJMoa1103782 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Conforti C, Paolini F, Venuti A, Dianzani C, Zalaudek I. The detection rate of human papillomavirus in well-differentiated squamous cell carcinoma and keratoacanthoma: is there new evidence for a viral pathogenesis of keratoacanthoma? Br J Dermatol. 2019;181(6):1309-1311. doi: 10.1111/bjd.18212 [DOI] [PubMed] [Google Scholar]
- 26.Pattee SF, Silvis NG. Keratoacanthoma developing in sites of previous trauma: a report of two cases and review of the literature. J Am Acad Dermatol. 2003;48(2)(suppl):S35-S38. doi: 10.1067/mjd.2003.114 [DOI] [PubMed] [Google Scholar]
- 27.Olsen CM, Green AC, Neale RE, et al. ; QSkin Study Collaborators . Cohort profile: the QSkin Sun and Health study. Int J Epidemiol. 2012;41(4):929-929i. doi: 10.1093/ije/dys107 [DOI] [PubMed] [Google Scholar]
- 28.QSkin Sun and Health Study. QIMR Berghofer Medical Research Institute. https://www.qimrberghofer.edu.au/qskin2/
- 29.Morze CJ, Olsen CM, Perry SL, et al. ; QSkin Study Collaborators . Good test-retest reproducibility for an instrument to capture self-reported melanoma risk factors. J Clin Epidemiol. 2012;65(12):1329-1336. doi: 10.1016/j.jclinepi.2012.06.014 [DOI] [PubMed] [Google Scholar]
- 30.Moons KG, Donders RART, Stijnen T, Harrell FE Jr. Using the outcome for imputation of missing predictor values was preferred. J Clin Epidemiol. 2006;59(10):1092-1101. doi: 10.1016/j.jclinepi.2006.01.009 [DOI] [PubMed] [Google Scholar]
- 31.Rubin DB. Multiple Imputation for Nonresponse in Surveys. John Wiley & Sons; 1987. doi: 10.1002/9780470316696 [DOI] [Google Scholar]
- 32.Textor J, van der Zander B, Gilthorpe MS, Liskiewicz M, Ellison GT. Robust causal inference using directed acyclic graphs: the R package ‘dagitty.’ Int J Epidemiol. 2016;45(6):1887-1894. [DOI] [PubMed] [Google Scholar]
- 33.Nehal KS, Bichakjian CK. Update on keratinocyte carcinomas. N Engl J Med. 2018;379(4):363-374. doi: 10.1056/NEJMra1708701 [DOI] [PubMed] [Google Scholar]
- 34.Xiang F, Lucas R, Hales S, Neale R. Incidence of nonmelanoma skin cancer in relation to ambient UV radiation in white populations, 1978-2012: empirical relationships. JAMA Dermatol. 2014;150(10):1063-1071. doi: 10.1001/jamadermatol.2014.762 [DOI] [PubMed] [Google Scholar]
- 35.Green AC, Olsen CM. Cutaneous squamous cell carcinoma: an epidemiological review. Br J Dermatol. 2017;177(2):373-381. doi: 10.1111/bjd.15324 [DOI] [PubMed] [Google Scholar]
- 36.de Vries E, Trakatelli M, Kalabalikis D, et al. ; EPIDERM Group . Known and potential new risk factors for skin cancer in European populations: a multicentre case-control study. Br J Dermatol. 2012;167(suppl 2):1-13. doi: 10.1111/j.1365-2133.2012.11081.x [DOI] [PubMed] [Google Scholar]
- 37.Dufresne RG, Marrero GM, Robinson-Bostom L. Seasonal presentation of keratoacanthomas in Rhode Island. Br J Dermatol. 1997;136(2):227-229. doi: 10.1111/j.1365-2133.1997.tb14901.x [DOI] [PubMed] [Google Scholar]
- 38.Dogliotti M, Caro I. Keratoacanthoma in a Bantu child. Int J Dermatol. 1976;15(7):524. doi: 10.1111/j.1365-4362.1976.tb00722.x [DOI] [PubMed] [Google Scholar]
- 39.Khalesi M, Whiteman DC, Tran B, Kimlin MG, Olsen CM, Neale RE. A meta-analysis of pigmentary characteristics, sun sensitivity, freckling and melanocytic nevi and risk of basal cell carcinoma of the skin. Cancer Epidemiol. 2013;37(5):534-543. doi: 10.1016/j.canep.2013.05.008 [DOI] [PubMed] [Google Scholar]
- 40.Nicolas M, Wolfer A, Raj K, et al. Notch1 functions as a tumor suppressor in mouse skin. Nat Genet. 2003;33(3):416-421. doi: 10.1038/ng1099 [DOI] [PubMed] [Google Scholar]
- 41.Sopori M. Effects of cigarette smoke on the immune system. Nat Rev Immunol. 2002;2(5):372-377. doi: 10.1038/nri803 [DOI] [PubMed] [Google Scholar]
- 42.Leonardi-Bee J, Ellison T, Bath-Hextall F. Smoking and the risk of nonmelanoma skin cancer: systematic review and meta-analysis. Arch Dermatol. 2012;148(8):939-946. doi: 10.1001/archdermatol.2012.1374 [DOI] [PubMed] [Google Scholar]
- 43.Song F, Qureshi AA, Gao X, Li T, Han J. Smoking and risk of skin cancer: a prospective analysis and a meta-analysis. Int J Epidemiol. 2012;41(6):1694-1705. doi: 10.1093/ije/dys146 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Dusingize JC, Olsen CM, Pandeya NP, et al. ; QSkin Study Collaborators . Cigarette smoking and the risks of basal cell carcinoma and squamous cell carcinoma. J Invest Dermatol. 2017;137(8):1700-1708. doi: 10.1016/j.jid.2017.03.027 [DOI] [PubMed] [Google Scholar]
- 45.Yen H, Dhana A, Okhovat J-P, Qureshi A, Keum N, Cho E. Alcohol intake and risk of nonmelanoma skin cancer: a systematic review and dose-response meta-analysis. Br J Dermatol. 2017;177(3):696-707. doi: 10.1111/bjd.15647 [DOI] [PubMed] [Google Scholar]
- 46.Mandrell JC, Santa Cruz D. Keratoacanthoma: hyperplasia, benign neoplasm, or a type of squamous cell carcinoma? Semin Diagn Pathol. 2009;26(3):150-163. doi: 10.1053/j.semdp.2009.09.003 [DOI] [PubMed] [Google Scholar]
- 47.Carr RA, Houghton JP. Histopathologists’ approach to keratoacanthoma: a multisite survey of regional variation in Great Britain and Ireland. J Clin Pathol. 2014;67(7):637-638. doi: 10.1136/jclinpath-2014-202255 [DOI] [PubMed] [Google Scholar]
- 48.Carr RA, Taibjee SM, Turnbull N, Attili S. Follicular squamous cell carcinoma is an under-recognised common skin tumour. Diagn Histopathol. 2014;20(7):289-296. doi: 10.1016/j.mpdhp.2014.05.003 [DOI] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
eTable 1. Sensitivity Analysis of Phenotype Variables and Their Association With Keratoacanthoma Among Restricted Cohort of 24 079 Participants
eTable 2. Sensitivity Analysis of Ultraviolet Exposure Variables and Their Association With Keratoacanthoma Among Restricted Cohort of 24 079 Participants
eTable 3. Sensitivity Analysis of Lifestyle Variables and Their Association With Keratoacanthoma Among Restricted Cohort of 24 079 Participants
eFigure. Directed Acyclic Graph for the Association Between Lifestyle Variables and Keratoacanthoma