Ten-Year Follow-up of Persons With Sun-Damaged Skin Associated With Subsequent Development of Cutaneous Squamous Cell Carcinoma

Key Points

Question  What is the risk of cutaneous squamous cell carcinoma after diagnosis of actinic keratosis (AK) during long periods of follow-up?

Findings  In this cohort study, in a community-based population in California, the risk of cutaneous squamous cell carcinoma was 1.92% per year after diagnosis of 1 or more AKs and 0.83% per year without an AK diagnosis, with older age being a stronger risk factor than AK diagnosis.

Meaning  These results can be used to develop recommendations for risk-based patient education, treatment, and surveillance.

Importance  Risk of cutaneous squamous cell carcinoma (cSCC) after the diagnosis of actinic keratosis (AK) has not been studied during long follow-up periods.

Objective  To estimate the risk up to 10 years and identify risk factors for cSCC development.

Design, Setting, and Participants  This longitudinal cohort study, performed from January 1, 2009, to February 29, 2020, examined Kaiser Permanente Northern California patients with AK and control patients matched 1:1 on age, sex, race/ethnicity, medical center, and date of the initial diagnosis plus 30 days in the patients with AK.

Exposures  Patients with AK and control participants were followed up for up to 10 years for incidence of cSCC.

Main Outcomes and Measures  Incident cSCC was obtained from pathologic data, and subdistribution hazard ratios (HRs) and 95% CIs were estimated using Cox proportional hazards regression analysis, accounting for competing risks, calendar year, demographic factors, and number of AKs.

Results  The study included 220 236 patients with AK and 220 236 matched control patients (mean [SD] age, 64.1 [12.2] years; 231 248 [52.5%] female). After losses to follow-up were accounted for, risk of cSCC increased with each year of follow-up by 1.92% (95% CI, 1.89%-1.95%) in patients with AK and 0.83% (95% CI, 0.81%-0.85%) in matched control patients (subdistribution HR, 1.90; 95% CI, 1.85-1.95). However, among patients 49 years or younger, those diagnosed with AK were nearly 7 times more likely to be diagnosed with cSCC than those without AK (HR, 6.77; 95% CI, 5.50-8.32). At 10 years, the cumulative incidence of cSCC reached 17.1% (95% CI, 16.9%-17.4%) in patients with AK and 5.7% (95% CI, 5.5%-5.9%) in control patients. Increased numbers of AKs were modestly associated with increased cSCC risk (≥15 AKs vs 1 AK: subdistribution HR, 1.89; 95% CI, 1.75-2.04). Older patients had much higher risk of cSCC than younger patients (compared with those ≤49 years of age at AK diagnosis; ≥80 years of age: subdistribution HR, 8.18; 95% CI, 7.62-8.78). Other than AK, risk factors for cSCC included older age, White race (a proxy for skin type), history of basal cell carcinoma, and male sex. Risk decreased between 2009 and 2019 (2018-2019 vs 2009-2010: subdistribution HR, 0.67; 95% CI, 0.63-0.72).

Conclusions and Relevance  The results of this longitudinal cohort study can be used to develop recommendations to increase early detection of cSCC. Additional research is needed to understand the effect of AK treatment on cSCC risk and outcomes of cSCC.

Actinic keratoses (AKs) develop from sun exposure and affect more than 10% of the US population.^1-3 A key risk factor for cutaneous squamous cell carcinoma (cSCC) is cumulative sun exposure.⁴ Although AKs are easily detected and treated, the extent to which they serve as a proxy for sun-damaged skin has not been adequately established.³ Estimates of the risk of cSCC after AK diagnosis have been based on short-term studies, and further investigation is needed to establish guidelines to standardize the frequency of skin surveillance to reduce cSCC morbidity and mortality.⁵ We conducted a large, community-based, matched longitudinal cohort study to assess cSCC risk for 10 years after the diagnosis of AK.

Kaiser Permanente Northern California provides care to one-third of the population of Northern California. Patients are encouraged to seek initial treatment for dermatologic conditions from their primary care practitioners, who can choose from among several modalities to consult with dermatologists.^6,7 Continuing education of primary care practitioners is a priority for the dermatology department, and dermatologists provide immediate access via teledermatology to assist primary care practitioners with diagnosis and treatment while the patient is in the examination room. A clinical practice guideline for AK diagnosis and treatment is located within a web-based clinical library, and UpToDate is integrated within the electronic medical record. Most patients with relatively few AKs receive cryotherapy, whereas those with a larger number receive field therapy or both cryotherapy and field therapy. The study was approved by the local institutional review board. Informed consent was not required. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

The retrospective, longitudinal cohort study included adults with an AK diagnosis (International Classification of Diseases, Ninth Revision [ICD-9] code 702.0 or International Statistical Classification of Diseases and Related Health Problems, Tenth Revision [ICD-10] code L57.0) recorded during an in-person visit with a primary care practitioner or dermatologist from January 1, 2009, to December 31, 2019, with follow-up through February 29, 2020 (eFigure in the Supplement). We defined the index date as 30 days after the date of AK diagnosis to identify and exclude patients whose lesion was diagnosed as cSCC within 30 days. We also excluded patients without 2 years of enrollment before their AK diagnosis or who had missing information on sex or on the location where they received primary care. Finally, we excluded the small percentage who had a history of cSCC, melanoma, organ transplant, or HIV before their index date.

General health plan members without a diagnosis of AK were eligible as control patients. Control patients were selected at random and were individually matched on the index date, sex, race/ethnicity, and age of the patients with AK. We also matched on primary care medical center, a proxy for current sun exposure. Like cases, control patients were required to have continuous health plan membership for 2 years before their index date and were ineligible if they had an AK diagnosis in the 2 years preceding their index date or a history of cSCC, melanoma, HIV, or organ transplant.

All clinical information was obtained from the electronic medical record from 2 years before the index date to the end of follow-up. The exposure variable was the severity of sun-damaged skin. The outcome variable was the first pathologic diagnosis of cSCC.

The severity of sun-damaged skin was estimated using a proxy: the number of AKs treated within 30 days of the AK visit. For most patients (the first group), the practitioner recorded the number of AKs treated with cryotherapy as a procedure and recorded the number of lesions removed in a drop-down menu that uses 1, 2, 3 to 14, and 15 or more as response categories. A second group of patients received an AK diagnosis, but the number of lesions removed was not coded. For these patients, we used natural language processing of the clinical note, using terms linked to liquid nitrogen and cryotherapy and accounting for negation. For validation, 50 notes positive for cryotherapy and 50 notes negative for cryotherapy were manually reviewed and search terms were edited until all 100 notes were assigned correctly. We refer to this group as cryotherapy, number unspecified. A third group was diagnosed with AKs and received field therapy (fluorouracil, imiquimod, photodynamic therapy, and aminolevulinic acid) and were classified as receiving field therapy. A fourth group received both cryotherapy and field therapy and were classified as such. A fifth group was diagnosed with AKs but did not receive therapy. For this group, we reviewed a convenience sample of 10 medical records, noting that the reason for not receiving therapy was typically patient preference or a decision to wait (eg, until after a wedding). In situ and invasive cSCCs were identified using pathologic data with SNOMED codes M8070 to M8074 and M8076.^8,9 To assure that the pathologic specimen was from the skin, we used natural language processing¹⁰ using the terms skin, actinic, keratosis, shave biopsy, and derm to rule out specimens from the tongue and mucosa, for example. Covariate information included age, sex, race/ethnicity, and history of a basal cell carcinoma (BCC).

We cross-tabulated demographic characteristics between the AK and control groups to assess the closeness of matching. We used Cox proportional hazards regression with Fine and Gray competing risks to test the hypothesis that AK was associated with cSCC risk.¹¹ Models were adjusted for 2-year period, age, sex, race/ethnicity, and history of BCC. The first proportional hazards regression model compared patients with AK (exposed) with matched control patients without AK (unexposed). A second model was similar but included a term for the interaction of age and AK status. A third model used the number of AKs treated and use of field therapy as a proxy for severity of sun-damaged skin, and patients with 2 or more treated AKs or with field therapy were compared with patients with 1 AK removed using cryotherapy. Patients were followed up from their index date to the date of cSCC diagnosis (outcome) or on the end-of-study date of February 29, 2020, death, disenrollment from the health plan, diagnosis of melanoma or HIV, or date of organ transplant. We report the subdistribution hazard ratios (HRs) and 95% CIs for the Fine and Gray model.¹² The subdistribution HRs represent the relative change in the subdistribution hazard function associated with the given covariate, which can be interpreted as the relative change in the instantaneous event rate among individuals who are event free or who have experienced a competing event. The incidence of cSCC per 100 person-years (ie, percentage per year) and 95% CIs were computed by year, age, sex, race/ethnicity, and history of BCC in the AK and control cohorts using an Aalen-Johansen estimator to account for competing risks^13,14 and stratifying by 10-year age group and severity of sun-damaged skin.

A total of 220 236 patients with AK and 220 236 control participants were included in the study (mean [SD] age, 64.1 [12.2] years; 231 248 [52.5%] female). Inclusion and exclusion criteria are shown in the eFigure in the Supplement, and baseline characteristics are listed in Table 1.

The median length of follow-up was 4.2 years (interquartile range [IQR], 1.9-7.1 years) in the patients with AK and 3.2 years (IQR, 1.2-6.1 years) in the control patients (Table 2), and control patients were more often censored because of death or disenrollment. A total of 19 620 patients with AK (8.9%) developed cSCC compared with 7095 control patients (3.2%), for mean incidence rates of 1.92 (95% CI, 1.89-1.95) for patients with AK and 0.83 (95% CI, 0.81-0.85) for control patients per 100 person-years. After multivariable adjustment, the HR for the overall association of AK with cSCC was 1.90 (95% CI, 1.85-1.95) (Table 2). After accounting for AK status, the subdistribution HR decreased in recent years, increased sharply with age, was somewhat higher in men than women, was most pronounced among White individuals, and was increased for those with a history of BCC.

We generated a second model to examine whether the association of AK with cSCC was different in younger and older persons by introducing an interaction term for AK status and age group (results not tabulated). Younger persons with an AK diagnosis had a greater HR of cSCC than older persons with an AK diagnosis. Thus, patients 49 years or younger who had AKs were nearly 7 times more likely than those without AKs to develop cSCC during the study period (HR, 6.77; 95% CI, 5.50-8.32). For those 50 to 59 years of age, the subdistribution HR was 3.20 (95% CI, 2.96-3.64), whereas the subdistribution HR was less than 2.00 for older patients (P < .001 for interaction).

A third model restricted to patients with AK examined the association of severity of sun-damaged skin with risk of cSCC (Table 3). Approximately one-third of patients who received treatment for AK had 1 AK removed, whereas more than half of cSCC cases occurred in patients with 1 or 2 AKs. In those who received cryotherapy, cSCC risk increased somewhat in relation to the number of lesions removed (≥15 AKs vs 1 AK: subdistribution HR, 1.89; 95% CI, 1.75-2.04), whereas risk in those who received field therapy was intermediate.

The cumulative incidence of cSCC associated with the severity of sun-damaged skin is given in Table 4 and the Figure after age stratification. Among those with AK, the cumulative incidence of cSCC reached 10% at 7 to 8 years for patients in their 50s, 5 to 6 years for patients in their 60s, 3 to 4 years for patients in their 70s, and 1 to 2 years for patients in their 80s. At 10 years, the cumulative incidence of cSCC reached 17.1% (95% CI, 16.9%-17.4%) in patients with AK and 5.7% (95% CI, 5.5%-5.9%) in control patients. Table 4 reports the cumulative incidence of cSCC for patients with AK compared with patients without AK in every age stratum.

In 220 236 patients with AK and 220 236 general population control patients observed for up to 10 years, cSCC incidence rates were 1.92% per year after an AK diagnosis and 0.83% per year without an AK diagnosis. Age was more strongly associated with cSCC risk compared with AK diagnosis, and larger numbers of AKs were only modestly associated with increasing cSCC risk. However, age modified the association of AK with cSCC risk; therefore, among persons 18 to 49 years of age, those with an AK had a nearly 7-fold increased risk of cSCC compared with those who did not. The time from AK diagnosis to 10% risk of SCC was shortest in the oldest patients and in patients with more AKs or who received field therapy. In addition, this study found that after age and length of follow-up were carefully accounted for, the 2018-2019 cohort had only two-thirds the risk of the 2008-2009 cohort, suggesting improvements in primary prevention in more recent birth cohorts.

Limitations of the study included the use of clinical and not histologic diagnosis of AK and use of AK treatment information as a proxy for sun damage. In addition, some control patients may have had undiagnosed AK, which would have biased the subdistribution HR of 1.90 (95% CI, 1.85-1.95) for the association of any AK with cSCC risk downward, although the analysis of number of AKs (Table 3) would not be affected. To increase the feasibility of the study, AK diagnoses recorded after the start of follow-up were not assessed or solar elastosis identified, which is underrecorded in this setting. Information on skin type beyond race/ethnicity was lacking, which likely confounded the associations for Hispanic patients. Furthermore, some AK cases may be misdiagnosed. Regarding generalizability, Kaiser Permanente’s community-based population is well characterized and largely representative of the underlying population. In addition, although the study excluded patients with immune-mediated diseases, an analysis of these patients was recently published by Kaiser Permanente.¹⁵

This report is comparable to past reports,^16,17 which also observed higher risk with age, in men, and in relation to AK. It is not clear whether AKs transform to cSCC or merely indicate sun-damaged skin^3,17-20 with increased potential to develop cSCC. Studies have not adequately estimated AK regression and progression rates. In a 2013 systematic review,³ spontaneous regression of an AK lesion ranged from 15% to 63%, recurrence of AK from 15% to 53%, and progression of a single AK to an cSCC at the same anatomical location from 0% to 0.53% per year. This study treated AK as a marker of sun-damaged skin.

It is best practice to educate patients about the role of repeated sun exposure in damaging the skin and causing solar elastosis, AK, and skin cancer. Reducing sun exposure appears to be effective in that the incidence of skin cancer has decreased for more recent birth cohorts.²¹ Regarding skin examinations to detect cSCC, current recommendations by the US Preventive Services Task Force state that “there is insufficient evidence for any population that regular visual skin examination by a clinician can reduce skin cancer–related morbidity and mortality...”^{22(p 433)} The relative effectiveness and safety of watchful waiting, cryotherapy, and field therapy for treating AKs have not been adequately studied.^23-26 The American Academy of Dermatology is currently developing AK guidelines.²⁷ Patients who decide not to receive treatment for an AK should be informed about the signs of progression and when to contact a practitioner.

Accepted for Publication: February 4, 2020.

Published Online: March 24, 2021. doi:10.1001/jamadermatol.2021.0372

Correction: This article was corrected on June 23, 2021, to fix a typographical error in the Limitations section.

Corresponding Author: Lisa J. Herrinton, PhD, Division of Research, Kaiser Permanente, 2000 Broadway, Oakland, CA 94612 (lisa.herrinton@kp.org).

Author Contributions: Dr Herrinton had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Madani, Marwaha, Dusendang, Pham, Herrinton.

Acquisition, analysis, or interpretation of data: Madani, Marwaha, Dusendang, Alexeeff, Chen, Han, Herrinton.

Drafting of the manuscript: Madani, Marwaha, Dusendang, Herrinton.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Dusendang, Alexeeff, Chen, Herrinton.

Obtained funding: Marwaha, Herrinton.

Administrative, technical, or material support: Marwaha, Herrinton.

Supervision: Madani, Marwaha, Herrinton.

Conflict of Interest Disclosures: None reported.

Funding/Support: This research was funded solely by the Rapid Analytics Unit of the Permanente Medical Group (Drs Marwaha and Herrinton).

Role of the Funder/Sponsor: The funding source (or sources) had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.