Association Between Topical Calcineurin Inhibitor Use and Keratinocyte Carcinoma Risk Among Adults With Atopic Dermatitis

Key Points

Question  Is exposure to topical calcineurin inhibitors among adult patients with atopic dermatitis associated with increased keratinocyte carcinoma risk?

Findings  In this large, health plan–based cohort study of 93 746 adults with atopic dermatitis, there was no increased risk of keratinocyte carcinoma overall among topical calcineurin inhibitor–exposed patients compared with topical corticosteroid–exposed patients or patients unexposed to topical calcineurin inhibitors or topical corticosteroids. In addition, no increased risk was noted when data were examined by subtype of keratinocyte carcinoma (basal cell carcinoma or squamous cell carcinoma).

Meaning  The findings of this study suggest that use of topical calcineurin inhibitors may not increase the risk of keratinocyte carcinoma risk among adults with atopic dermatitis.

Importance  Topical calcineurin inhibitors (TCIs), primarily used to treat atopic dermatitis (AD), carry a black box label warning users about the potential for increased skin cancer risk. The risk associated with keratinocyte carcinoma (KC), the most common cancer, defined as basal cell carcinoma (BCC) and squamous cell carcinoma (SCC), remains poorly defined because findings from large-scale postmarketing surveillance studies have not been reported.

Objectives  To examine KC risk overall and by subtype (BCC and SCC) among adults with AD exposed to TCIs compared with those exposed to topical corticosteroids (primary comparator group) and those unexposed to TCIs or topical corticosteroids (alternative comparator group) as well as alterations in risk with TCI dose, frequency, and duration of exposure.

Design, Setting, and Participants  A retrospective cohort study was conducted at Kaiser Permanente Northern California, a large, integrated health care delivery system, of adults 40 years or older (n = 93 746) with a physician-rendered diagnosis of AD or dermatitis. Patients who were diagnosed from January 1, 2002, to December 31, 2013, were included, with follow-up through December 31, 2017. Data analysis was conducted from June 1, 2016, to October 1, 2018.

Exposures  Time-varying pharmacy-dispensed TCI exposure (n = 7033) over the study period was compared with topical corticosteroids (n = 73 674) and no TCI or topical corticosteroid exposure (n = 46 141).

Main Outcomes and Measures  Electronic pathologic testing–validated incident KCs (n = 7744).

Results  Among a cohort of 93 746 members, the mean (SD) age was 58.5 (12.7) years, and 55 023 patients (58.7%) were women. Multivariable Cox proportional hazards regression revealed no association between TCI exposure and KC risk (adjusted hazard ratio [aHR], 1.02; 95% CI, 0.93-1.13) compared with topical corticosteroid exposure. Similarly, there were no significant differences in BCC risk (aHR, 1.01; 95% CI, 0.90-1.14, TCI vs topical corticosteroids) or SCC risk (aHR, 0.94; 95% CI, 0.82-1.08, TCI vs topical corticosteroids). Changing the comparator group to unexposed individuals yielded similar findings (aHR, 1.04; 95% CI, 0.91-1.19, TCI vs unexposed for basal cell carcinoma). There were no associations between TCI dose, frequency, and duration of use and BCC, SCC, or overall KC risk.

Conclusions and Relevance  The results of this postmarketing surveillance study of adult health plan members with AD revealed no apparent association between TCI exposure and overall KC, BCC, or SCC risk. Secondary analyses examining dose, frequency, and duration of TCI exposure revealed no associations. These findings suggest that use of TCIs may be safe with respect to KC risk among adults with AD.

Atopic dermatitis (AD) is a chronic, pruritic, inflammatory skin disease that affects between 2% and 10% of the adult population in industrialized nations.¹ Treatment primarily consists of application of topical emollients and anti-inflammatory agents, such as topical corticosteroids and topical calcineurin inhibitors (TCIs), although more severe cases often require systemic immunosuppressive therapy.² Two TCIs, tacrolimus and pimecrolimus, were approved by the US Food and Drug Administration (FDA) for the treatment of AD in 2001 and 2002.^3,4 Understanding the long-term safety profile of this newer class of topical agents is important for patients, clinicians, and regulators.

Keratinocyte carcinoma (KC), consisting of squamous cell carcinoma (SCC) and basal cell carcinoma (BCC), is the most common cancer in the US, with a nationwide annual estimated incidence of 5.4 million tumors.⁵ Risk factors for KC include lighter pigmentation phenotype, older age, male sex, UV light exposure, and immunosuppression.^6-8 Systemic immunosuppressive medications, including calcineurin inhibitors, have been associated with increased risk of cancer, particularly KCs.⁹ This increased KC risk associated with oral calcineurin inhibitor use prompted concern about the potential for systemic absorption of these agents with topical use. This theoretical concern, compounded by published reports of skin cancer associated with TCI use,¹⁰ prompted the FDA to implement a black box warning for TCIs, citing lack of long-term safety data.¹¹

Previously published epidemiologic studies of TCI exposure and KC risk have reported conflicting results.^4,11-18 Whereas some large-scale epidemiologic studies have shown that exposure to TCIs in patients with a history of AD does not appear to be associated with an increased risk of KC,^15-17 a recently published cohort study found increased KC risk.¹⁸ Additional large cohort studies among adults with long-term longitudinal follow-up and complete capture of both exposure and outcome are needed and have been mandated by the FDA to further clarify the role of TCIs in KC risk.

Using a large cohort of adult health plan members 40 years or older with a clinician-rendered AD or dermatitis diagnosis, we examined the association between TCI exposure and risk of subsequent KC over a 15-year study period. Our primary objective was to assess KC risk overall and by KC subtype (BCC and SCC) among individuals with AD exposed to TCIs compared with those exposed to topical corticosteroids and those unexposed to TCIs or topical corticosteroids. The secondary objective was to assess TCI dose, frequency, and duration of use in relation to KC risk.

A retrospective observational cohort study was conducted of adult health plan members diagnosed with AD or dermatitis by a clinician between 2002 and 2013 while enrolled in Kaiser Permanente Northern California, an integrated health care delivery system. Kaiser Permanente Northern California serves approximately 4.3 million members, representing about 30% of the insured population in 14 Northern California counties, with membership demographic characteristics closely resembling those of the underlying census population. Members receive comprehensive care at Kaiser Permanente Northern California facilities and the electronic health records allow tracking of patient enrollment, diagnoses, and prescriptions.

All study procedures were reviewed and approved by the Kaiser Permanente Northern California institutional review board with waiver of informed consent because the study was deemed low risk with a limited data set and would not be feasible if nearly 100 000 health plan members needed to provide consent. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cohort studies.

Kaiser Permanente Northern California health plan members 40 years or older with a clinician-recorded diagnosis of AD (International Classification of Diseases, Ninth Revision, code 691.8) or dermatitis (code 695.9) as identified in the medical record from January 1, 2002, through December 31, 2013 (n = 123 222), were included. Cohort entry date was defined as the first recorded diagnosis of AD or dermatitis. Patients were required to have at least 2 years of continuous Kaiser Permanente Northern California health plan membership with gaps in membership not exceeding 3 months before cohort entry to enable acquisition of relevant baseline covariates (n = 22 885 patients excluded). Individuals were also excluded if they had a cancer diagnosis (excluding KC) within 2 years before cohort entry as identified by linkage to the Kaiser Permanente Northern California Cancer Registry (n = 1914), were infected with HIV at any time during Kaiser Permanente Northern California membership as identified by linkage to the Kaiser Permanente Northern California HIV Registry (n = 484), were exposed to UV light–based treatments within 2 years before cohort entry (n = 469), had a solid organ transplant any time during Kaiser Permanente Northern California membership (n = 82), were exposed to tacrolimus any time before cohort entry (n = 2775), were diagnosed with rare genetic syndromes associated with increased KC risk as listed in eTable 1 in the Supplement (n = 821), were diagnosed with cutaneous T-cell lymphoma at any time during Kaiser Permanente Northern California membership given that it often clinically mimics AD (n = 116), or were missing sex information (n = 1). The final cohort consisted of 93 746 patients (Figure). Participants were censored at the earliest of the following time points: (1) diagnosis of KC, (2) disenrollment from the health care system for greater than 3 months, (3) death, or (4) December 31, 2017 (end of the study). Data analysis was conducted from June 1, 2016, to October 1, 2018.

Prescriptions for TCIs (tacrolimus and pimecrolimus) and topical corticosteroids (eTable 2 in the Supplement) were captured through December 2017 by linkage with the Kaiser Permanente Northern California pharmacy database using National Drug Code and text-string searches of topical formulations that contained the trade or generic names. Individuals were considered exposed on the date of the first dispensation. Patients were classified into 3 time-varying exposure groups to allow for potential movement across exposure categories through time, defined as (1) TCI-exposed (with or without previous topical corticosteroid exposure), (2) topical corticosteroid only-exposed, and (3) unexposed (defined as never exposed to TCIs or topical corticosteroids). The main comparison of interest was TCI-exposed compared with topical corticosteroid only–exposed individuals (given presumed similar levels of disease severity), while the unexposed group was used an additional comparator group. Other exposure variables ascertained included dosing, frequency of use, and duration of exposure as specified by the prescribing clinician.

Dose was categorized dichotomously for tacrolimus as low dose (0.03% formulation) and high dose (0.1% formulation) based on the most recently dispensed prescription. Pimecrolimus dosing was not categorized because it is available in only a 1% formulation. Frequency of use was categorized based on the directions for application of the most recent prescription as follows: once daily or less (defined as low frequency), twice daily or more (defined as high frequency), or unspecified (defined as unknown frequency).

Length of use was defined as a categorical, time-varying variable as follows: less than 2 years (short-term), 2 to less than 4 years (moderate-term), and 4 or more years (long-term). A gap in treatment was defined as not having at least 1 prescription dispensed within a consecutive 2-year period. The rationale for selecting 2 years as the treatment gap window was based on the fact that TCIs have a 2-year shelf life.¹⁹ To test the validity of this assumption, a sensitivity analysis was performed altering this gap period to 1 year. To test the robustness of our duration definition, the cumulative number of filled prescriptions was also examined as a time-varying covariate.

Incident KC was identified from the Kaiser Permanente Northern California electronic pathology database, using Systematized Nomenclature of Medicine codes.^20,21 All pathologic test reports with basal cell carcinoma and squamous cell carcinoma Systematized Nomenclature of Medicine morphology codes were reviewed and assessed for study inclusion by a board-certified dermatologist (M.M.A.) blinded to exposure status.

The following variables were assessed: age (categorical, 5-year intervals); sex; race/ethnicity; cohort entry year; number of visits with a dermatologist during the previous year (time varying); autoimmune diseases (including rheumatoid arthritis, systemic lupus erythematosus, ankylosing spondylitis, dermatomyositis, and inflammatory bowel disease) during follow-up (time varying); exposure to prescription topical corticosteroids within 2 years before cohort entry; KC diagnosis within 2 years before cohort entry; exposure to systemic immunosuppressive medications, including corticosteroids, chemotherapy, radiotherapy, and phototherapy; and cohort entry year.

Crude incidence rates and 95% CIs for incident KC, BCC, and SCC were calculated for the 3 exposure groups with allocation of at-risk person-time across time-varying levels of exposure. Cox proportional hazards regression models were used to estimate KC, BCC, and SCC hazard ratios (HRs) with 95% CIs for each contrast of interest, adjusting for relevant covariates. We did not use a new user design because most (93%) patients using TCI had previous topical corticosteroid exposure, and exclusion of this large number of patients would limit study power and generalizability. Three sets of models were examined: unadjusted, adjusted for demographics (age, sex, and race/ethnicity) and calendar year, and fully adjusted. For analyses examining dose, frequency, and duration of TCI use, the reference group was no TCI or topical corticosteroid exposure. In primary analyses comparing TCI with topical corticosteroid exposure, we had sufficient power (80%) to detect HRs of at least 1.12 for any KC, 1.14 for BCC, and 1.17 for SCC (2-sided test, α = .05).

Sensitivity analyses included (1) altering lag time to KC occurrence by excluding tumors that developed early during exposure (6-, 12-, and 24-month lag times) to account for possible biologic plausibility of tumor initiation and progression, (2) stratification on KC status before cohort entry to assess potential heterogeneity in effect, (3) stratification by race/ethnicity, and (4) change in assumed duration of exposure of a topical medication from 2 years to 1 year to assess the validity of the exposure duration assumption. Statistical analysis was conducted with SAS, version 9.4 (SAS Institute Inc).

The final cohort included 93 746 persons with AD, of whom 7033 received at least 2 prescriptions for TCIs and 73 674 received at least 2 prescriptions for topical corticosteroids during follow-up (mean, 7.70 years). The mean (SD) age at cohort entry of this racially and ethnically diverse cohort was 58.5 (12.7) years, with more women than men (55 023 [58.7%] vs 38 723 [41.3%]) (Table 1). The mean cohort entry year was 2006 with follow-up through December 31, 2017, amounting to 722 271 person-years of follow-up. Approximately 5% of the cohort had a history of KC verified through pathologic testing. Less than 0.1% of the patients had exposure to systemic therapies used to treat severe AD, including cyclosporine, tacrolimus, methotrexate, and tumor-necrosis factor inhibitors (adalimumab, etanercept, and infliximab). Linkage of the cohort to the Kaiser Permanente Northern California pathology database revealed that 8.3% had at least one pathologic test report with a Systematized Nomenclature of Medicine code indicating possible KC. Electronic pathology reports from 28 599 skin biopsies that met inclusion criteria were reviewed, revealing 5478 incident BCCs and 3773 incident SCCs.

The crude incidence rates for the association between TCI exposure and KC risk are reported in Table 2. There was little variability in the incidence rates of KC, BCC, and SCC across the exposure categories of TCI (KCC: 9.43, BCC: 6.46, and SCC: 4.52 cases/1000 person-years), topical corticosteroids only (KCC: 10.71, BCC: 7.47, and SCC: 5.17 cases/1000 person-years), and unexposed (KCC: 11.16, BCC: 7.76, and SCC, 4.84 cases/1000 person-years). Patients in the TCI-exposed group had the lowest incidence rates for all 3 skin cancer outcomes across all exposure groups.

In examining risk, the HR for overall KC, BCC, and SCC was similar across exposure categories (TCI vs topical corticosteroids and TCI vs unexposed) in both adjusted and unadjusted models (Table 3). Participants exposed to TCIs had a borderline significant decreased unadjusted risk of KC, which was attenuated and nonsignificant with adjustment for potential confounders. In adjusted analyses, patients in the TCI-exposed group had a similar KC risk compared with patients exposed to topical corticosteroids (fully adjusted HR [aHR], 1.02; 95% CI, 0.93-1.13) and unexposed patients (aHR, 1.03; 95% CI, 0.92-1.14). Examination of risk by KC subtypes suggested that patients exposed to TCIs had a similar BCC risk compared with patients exposed to topical corticosteroids (aHR, 1.01; 95% CI, 0.90-1.14) and unexposed patients (aHR, 1.04; 95% CI, 0.91-1.19). With respect to SCC, patients exposed to TCIs had a nonsignificant decrease in their risk estimate compared with patients exposed to topical corticosteroids (aHR, 0.94; 95% CI, 0.82-1.08) and unexposed individuals (aHR, 0.91; 95% CI, 0.78-1.06). Exploratory analysis to determine which variables were associated with attenuating the risk estimates in the fully adjusted model revealed that the annual number of dermatology visits was a notable factor in the attenuation of risk in all 3 outcomes.

Dose (tacrolimus only), frequency, and duration of use analysis did not reveal any significant changes in overall KC risk, BCC risk, or SCC risk with TCI dose in adjusted or unadjusted analyses. In examining an alternative definition of duration (ie, number of TCI dispensed prescriptions), we found no statistically significant associations with KC, BCC, or SCC risk (Table 4).

In this large, FDA-mandated health plan–based retrospective cohort study including 93 746 adults with AD with 722 271 person-years of follow-up, we found no statistically significant risk of KC overall or by BCC or SCC tumor subtype among patients ever exposed to TCIs compared with those exposed to topical corticosteroids and patients with no unexposure to either of these drug classes. Furthermore, we found no statistically significant associations with dose, frequency, or duration of exposure to TCIs. To our knowledge, this study includes the highest number of person-years of follow-up reported in the literature and curated KC outcome data gathered from validated pathology records for a cancer that is difficult to capture. This study was conducted to address safety concerns related to the risk of KC after treatment with TCIs and the findings may provide some reassurance regarding the safety profile to patients, clinicians, and regulators.

Controversy has surrounded the association between TCI exposure and KC risk since the black box warning was issued by the FDA.¹¹ A hypothesized mechanism of action for TCIs increasing KC risk includes a direct effect of calcineurin inhibition on DNA repair and apoptosis,⁴ which could influence keratinocyte carcinogenesis. However, epidemiologic research assessing this association has yielded conflicting results. Whereas some large-scale epidemiologic studies have shown that exposure to TCIs in patients with a history of AD is not associated with an increased risk of KC,^15-17 a recently published cohort study found increased KC risk.¹⁸ A prospective, Dutch population-based study of 4375 participants with AD who underwent full-body skin examinations did not show a greater risk of BCC (adjusted odds ratio, 0.71 for BCC and 1.54 for SCC).²² However, a recently published multicenter European cohort study showed increased KC risk among adults exposed to TCI with AD with an adjusted incidence rate ratio of 1.12 (95% CI, 1.01-1.23) but the association was attenuated and became nonsignificant after applying quantitative bias analysis for unmeasured confounders (corrected incidence rate, 1.08; 95% CI, 0.98-1.19).¹⁸ While the investigators in the European cohort study had substantial follow-up time, with 190 412 person-years in the TCI group and 777 075 person-years in the topical corticosteroids group, their mean duration of follow-up ranged from 2.2 to 5.6 years, compared with the mean follow-up time in our study of 7.70 years. Given that KCs are often slow growing, our longer follow-up time would allow for examination of the association between long-term TCI use and KC risk. The incidence rates of KC in our study were approximately 2-fold higher than those in the European cohort, which may suggest incomplete ascertainment of KCs in the European cohort population or different disease risk based on risk factors, such as sun exposure. Furthermore, whereas all cases of KC in our study were confirmed by expert review, only a random sample of outcomes were confirmed in the European study, which could also lead to misclassification of outcomes. Nevertheless, our findings and their corrected risk estimates were similar.

This study has several strengths in addition to the large cohort size and long-term follow-up. These strengths include the ethnically diverse study population within an integrated health care delivery system with detailed pharmacy, covariate, and outcome data. Additional strengths include a clinician-rendered diagnosis of AD as the disease definition, compared with patient- or caregiver-reported AD diagnosis used in other epidemiologic studies, which can be more prone to disease misclassification,²³ and curated outcome measures of KC diagnoses that were physician-adjudicated. The study also has limitations. With an observational design, there exists the potential for unmeasured confounding, including AD disease severity and sun exposure history, although we attempted to reduce confounding by disease severity by controlling for systemic AD treatment. Although we did not have data on sun exposure, we do not anticipate differential sun exposure patterns by topical AD treatment exposure categories. We relied on pharmacy-dispensed medications rather than self-reported use to characterize exposure, which may not reflect actual use of medication by the patient and does not capture over-the-counter topical corticosteroid use, although the potency of over-the-counter corticosteroids is low and any association with KC risk might be less than that of prescription topical corticosteroids. We were unable to completely account for potential screening biases, although we adjusted for health care use (ie, dermatology visits) as a proxy for skin cancer screening.

Pharmacoepidemiologic studies comparing medications often use the new-user cohort design, examining outcomes in treatment-naive patients. In this study, the primary comparison of interest was TCI contrasted with an established treatment (topical corticosteroids) with almost all initiators of the new drug having had prior topical corticosteroid exposure. Implementation of the new-user design would diminish the applicability and generalizability of study findings given the elimination of patients for whom decision-makers are seeking comparative safety evidence and would also limit sample size and estimation precision.²⁴

This study not only provides estimates of the risk of KC associated with TCI exposure, but also estimates overall KC and KC subtype risk by topical corticosteroid use, as well as the risk of KC in untreated AD. Such estimates may be useful for future pharmacoepidemiology research not just in AD, but in other inflammatory skin disorders as well.

We noted no increased risk of KC overall or by tumor subtype with TCI use among a large cohort of adults with AD. Our findings appear to support those of smaller postmarketing surveillance studies of TCI and KC risk and may provide some reassurance about the safety profile of this class of topical agents in the treatment of AD.

Accepted for Publication: May 12, 2020.

Corresponding Author: Maryam M. Asgari, MD, MPH, Department of Dermatology, Massachusetts General Hospital, 50 Staniford St, Ste 230A, Boston, MA 02114 (pores@mgh.harvard.edu).

Published Online: August 12, 2020. doi:10.1001/jamadermatol.2020.2240

Author Contributions: Drs Asgari and Quesenberry had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Asgari, Quesenberry.

Acquisition, analysis, or interpretation of data: Asgari, Tsai, Avalos, Sokil.

Drafting of the manuscript: Asgari, Avalos, Sokil.

Critical revision of the manuscript for important intellectual content: Asgari, Tsai, Avalos, Quesenberry.

Statistical analysis: Tsai, Quesenberry.

Obtained funding: Asgari.

Administrative, technical, or material support: Sokil.

Supervision: Asgari.

Conflict of Interest Disclosures: Dr Asgari reported receiving grants from Valeant during the conduct of the study and grants from Pfizer Inc outside the submitted work. No other conflicts were reported.

Funding/Support: This work has been supported by a grant to Kaiser Permanente from Valeant Pharmaceuticals.

Role of the Funder/Sponsor: The sponsor provided feedback in study design and protocol development, but had no role in the conduct of the study including collection, management, analysis, and interpretation of the data or preparation of the findings. The sponsor reviewed and approved the manuscript and supported the decision to submit the manuscript for publication.

Additional Contributions: Jessica Feng, BS (Department of Dermatology, University of Massachusetts), helped in characterizing the study cohort, which was described in a published abstract submitted to the Society for Investigative Dermatology Annual Meeting in 2017. She received no compensation outside of salary.