Association Between Topical Calcineurin Inhibitor Use and Risk of Cancer, Including Lymphoma, Keratinocyte Carcinoma, and Melanoma: A Systematic Review and Meta-analysis

Key Points

Question  Is topical calcineurin inhibitor (TCI) use associated with an increased risk of cancer?

Findings  This systematic review and meta-analysis of 11 studies revealed no association between TCI use and risk of cancer overall or skin cancer. Lymphoma risk was elevated with TCI use.

Meaning  Although this study found a positive association between TCI use and lymphoma, the low absolute risk of lymphoma makes the potential increased risk attributable to TCI use for any individual patient very small.

Importance  Topical calcineurin inhibitors (TCIs) are commonly used as second-line treatment for atopic dermatitis. In 2006, the US Food and Drug Administration issued a black box warning against TCI use, citing data from case reports and animal studies indicating a potential risk of cancer.

Objective  To evaluate the association between TCI use and risk of malignant neoplasms compared with nonactive and active comparator groups.

Data Sources  Electronic searches were conducted in MEDLINE via Ovid, Embase via Ovid, and Web of Science from database inception to August 21, 2020.

Study Selection  Observational studies investigating the association between treatment with TCIs (ie, tacrolimus and pimecrolimus) and the development of cancer with nonactive or active comparators were included. The population of interest was not limited to any specific disease state, age, or sex. All articles were assessed independently and in duplicate by 2 reviewers. Risk of bias was assessed using the Newcastle-Ottawa scale. Of 2464 nonduplicate records retrieved from the search, 11 studies met the inclusion criteria.

Data Extraction and Synthesis  Data extraction was conducted independently by 2 reviewers according to Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines. Random-effects meta-analyses were used to derive pooled relative risk (RR) estimates. Data were analyzed from July 25 to October 25, 2020.

Main Outcomes and Measures  Risk of cancer overall and risk of specific cancer types (lymphoma, melanoma, and keratinocyte carcinoma).

Results  Eight unique cohort studies (408 366 treated participants [55.1% female], 1 764 313 nonactive comparator controls, and 1 067 280 controls using topical corticosteroids) and 3 unique case-control studies (3898 cases [55.0% male] and 14 026 cancer-free controls [52.4% male]) were included. There was no association between TCI use and cancer overall compared with nonactive comparators (RR, 1.03; 95% CI, 0.92-1.16). Lymphoma risk was elevated with TCI use with both nonactive (RR, 1.86; 95% CI, 1.39-2.49) and topical corticosteroid comparators (RR, 1.35; 95% CI, 1.13-1.61). No significant association was found between TCI use and increased skin cancer (melanoma and keratinocyte carcinoma).

Conclusions and Relevance  The findings of this systematic review and meta-analysis suggest an association between TCI use and risk of lymphoma but not other cancers. Combined with the low absolute risk of lymphoma, the potential increased risk attributable to TCI use for any individual patient is likely very small.

Topical calcineurin inhibitors (TCIs), tacrolimus ointment and pimecrolimus cream, are indicated as second-line treatment of atopic dermatitis when first-line topical corticosteroids (TCSs) are ineffective or contraindicated. Quiz Ref IDThe US Food and Drug Administration issued a black box warning in 2006 indicating a potentially elevated risk of cancer with TCI use based on the findings of case reports (primarily of lymphomas and skin cancers), animal carcinogenicity studies, and studies on systemic tacrolimus use in organ transplant recipients.^1-6 Of the more than 6.7 million patients receiving TCI treatment, the black box warning was based on approximately 25 case reports, without systematic analysis supporting causation between TCI use and malignant neoplasms.⁴

Given the long potential latency for cancer development, postmarketing studies with long follow-up are necessary to determine whether there is an association between TCIs and cancer. A clinically important association would indicate the need for caution with TCI use, particularly for patients with chronic atopic dermatitis requiring long-term topical anti-inflammatory therapy. Conversely, if there is no clinically meaningful association, unwarranted worry stemming from regulatory safety warnings could lead to nonadherence and undertreatment of atopic dermatitis.⁷ In the 15 years since the US Food and Drug Administration issued their warning, several observational studies investigating the risk of malignant neoplasms with TCI use have been conducted,^8,9 but even individual large studies with rare outcomes may fail to accurately identify the true association. We conducted this systematic review and meta-analysis of observational studies to determine the association between treatment with TCIs and the risk of cancer, and specifically the risk of lymphoma, keratinocyte carcinoma, and melanoma.

We prospectively registered our study protocol in PROSPERO.¹⁰ Research ethics board approval was not sought for this project because it did not involve participant-level data. We adhered to the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) reporting guideline¹¹ and the Meta-analysis of Observational Studies in Epidemiology (MOOSE) reporting guideline.¹²

We searched MEDLINE via Ovid, Embase via Ovid, and Web of Science from inception to August 21, 2020, using a search strategy with terms for cancer, topical, and calcineurin inhibitors (detailed search strategy in eTables 1-3 in the Supplement). The search strategy was reviewed by a librarian with expertise in systematic reviews. We originally planned to restrict our search to English-language articles, but our final search was performed without language restriction. We manually searched the citations of included studies and relevant review articles for additional studies that were not included in our electronic searches.

We included observational studies investigating the association between treatment with TCIs (tacrolimus or pimecrolimus) and the incidence of cancer or cancer mortality. The population of interest was not limited to any specific disease state (eg, atopic dermatitis), age, or sex. Studies were included in the meta-analyses if they reported an effect estimate for the risk of cancer in individuals receiving TCIs compared with a control group (including active and nonactive comparators). Conference abstracts and unpublished studies were excluded. Outcomes were any cancer, when studies reported the risk of cancer overall, and the specific cancer types lymphoma (including systemic and cutaneous lymphoma), melanoma, and keratinocyte carcinoma.

Two investigators (M.L. and J.W.Z.) screened titles, abstracts, and full texts independently and in duplicate. Discrepancies were resolved by consulting the senior author (A.M.D.). The following data were extracted independently and in duplicate by 2 investigators (M.L. and J.W.Z.) using a standardized form: study characteristics (author, year, study design, country, participant source, inclusion and exclusion criteria, method for identifying cancer, and follow-up period), population characteristics (number of participants, age, sex, and purpose of treatment), treatment factors (type of calcineurin inhibitor, strength, dosing schedule, other concurrent treatment, and duration and amount of use), and outcomes (type of cancer, number and proportion of participants, effect estimate if reported, and factors adjusted for). After protocol registration, we also decided to extract effect estimates on the risk of cancer stratified by TCI strength, cumulative dose, and frequency of use.

We assessed the quality of included studies using the Newcastle-Ottawa scale for cohort and case-control studies.¹³ Funnel plots were generated and visually inspected to evaluate risk of publication bias.

Data were analyzed from July 25 to October 25, 2020. We used random-effects models to account for potential heterogeneity between studies. We intended to pool cohort studies and case-control studies separately, and ultimately only conducted meta-analyses of cohort studies because there were insufficient case-control studies to pool for each outcome. Primary analyses were performed using the inverse variance method, pooling study-reported effect estimates for the risk of cancers with TCIs for a summary relative risk (RR) estimate. Adjusted effect estimates were used preferentially when available. Incidence ratios, odds ratios, and hazard ratios with corresponding SEs were considered approximate measures of RR and pooled as previously described,^14-18 given an underlying assumption of the rarity of outcome events. We used the I² statistic to assess statistical heterogeneity across studies in the pooled analysis.

We performed subgroup analyses separating topical tacrolimus and pimecrolimus and for studies limited to children. Secondary analyses using the Mantel-Haenszel method were performed if included studies reported zero groups. We performed subsequent sensitivity analyses removing zero groups from the initial secondary analyses. Quantitative analyses were conducted using Review Manager, version 5.3 (The Nordic Cochrane Centre, The Cochrane Collaboration).¹⁹ Unpaired 2-sided P < .05 indicated statistical significance.

We retrieved 2854 records from our literature search. After removing 390 duplicate records, we assessed 2464 titles and abstracts and excluded 2304. We assessed full texts of 160 studies for eligibility. We ultimately included 8 cohort studies^8,9,20-25 and 3 case-control studies^26-28 (Figure 1). Six studies^{8,22-24,27,28} were conducted in the United States, 3 studies^9,21,26 were based in Europe (Denmark and United Kingdom), 1 study²⁰ was conducted in Asia (Singapore), and 1 study²⁵ included participants from centers across North America and Europe.

Five studies^8,9,20,23,24 included a nonactive comparator or untreated control group. Two studies^22,25 used expected or standardized incidence rates from the Surveillance, Epidemiology, and End Results program of the National Cancer Institute as a comparator. Four studies^9,21,23,24 included an active comparator group treated with TCSs.

A total of 408 366 participants were treated with TCIs in cohort studies, with a mean age of 17.1 years and a mean percentage of female participants of 55.1% and male participants of 44.9%. Of these participants, 151 772 were treated with tacrolimus and 214 640 with pimecrolimus. In the 5 studies^8,9,20,23,24 with a nonactive comparator group, a total of 1 764 313 untreated controls were reported. Of the 4 studies using a TCS comparator,^9,21,23,24 a total of 1 067 280 TCS-treated participants were reported. Four cohort studies^9,21,22,25 included a children-only group, with 93 120 children included. A detailed overview of the characteristics of included cohort studies and participants treated with TCIs can be found in the Table. Of the 3 case-control studies,^26-28 3898 participants with cancer and 14 026 cancer-free controls were reported, with a mean of 45.0% female and 55.0% male cases and 47.6% female and 52.4% male controls (eTable 4 in the Supplement).

A summary of scoring distribution for quality assessment using the Newcastle-Ottawa scale for nonrandomized studies can be found in eTable 5 in the Supplement. One study²² was assigned 3 stars, indicating a high or unclear risk of bias, owing to insufficient description of study cohorts and self-report of malignant neoplasm; 6 studies^{8,9,21,23,25,28} were assigned 4 to 6 stars, indicating a moderate risk of bias; and 4 studies^20,24,26,27 were assigned 7 to 9 stars, indicating a low risk of bias.

Four studies^8,20,22,25 reported effect estimates for risk of any cancer with TCI treatment compared with untreated cohorts (nonactive comparator). Primary analysis using the inverse variance method yielded a pooled RR of 1.03 (95% CI, 0.92-1.16; I² = 10%) (eFigure 1 in the Supplement). Heterogeneity was decreased (I² = 0%) when sensitivity analyses were performed with participants treated with tacrolimus^8,20,25 (RR, 0.93; 95% CI, 0.81-1.06) and pimecrolimus^8,20,22 (RR, 1.16; 95% CI, 1.01-1.32). Two studies^22,25 reported risk of any cancer in child-only cohorts, with a pooled RR of 1.10 (95% CI, 0.59-2.06, I² = 0%) (Figure 2). No studies reported overall risk of any cancer with TCI compared with TCS treatment.

Five studies^8,9,20,22,23 reported effect estimates for risk of lymphoma. Primary analysis using the inverse variance method showed a pooled RR of 1.86 (95% CI, 1.39-2.49; I² = 27%) (eFigure 1 in the Supplement). With subgroup analysis of participants treated with tacrolimus,^8,20,23 pooled RR and heterogeneity increased (RR, 2.20; 95% CI, 0.96-5.07; I² = 70%). Subgroup analysis of participants treated with pimecrolimus^8,20,22,23 resulted in a decreased pooled RR and heterogeneity (RR, 1.82; 95% CI, 1.27-2.60; I² = 0%) (Figure 3A).

To reduce the possibility of a participant being included in both groups with B-cell lymphoma and T-cell lymphoma as reported in the study by Hui et al,⁸ we performed subgroup analysis excluding groups with T-cell lymphoma (RR, 1.56; 95% CI, 1.21-2.00; I² = 0%) and excluding groups with B-cell lymphoma (RR, 2.16; 95% CI, 1.52-3.06; I² = 21%). One study²⁰ reported zero groups for incidence of lymphoma; thus, a secondary analysis using the Mantel-Haenszel method to include outcomes from zero groups was performed. Meta-analysis of dichotomous outcomes found a similar risk of 1.87 (95% CI, 1.31-2.67; I² = 44%). Likewise, sensitivity analysis removing zero groups from secondary analysis yielded similar risk with increased heterogeneity (RR, 1.88; 95% CI, 1.29-2.75; I² = 56%) (eFigure 1 in the Supplement).

Quiz Ref IDThree studies^9,21,23 reported effect estimates for risk of lymphoma in TCI-treated compared with TCS-treated groups. The pooled RR in the primary analysis was 1.35 (95% CI, 1.13-1.61; I² = 0%) (eFigure 2 in the Supplement), for an elevated risk of 35% for TCIs. Two studies^9,21 reported effect estimates for children-only groups (RR, 1.78; 95% CI, 0.85-3.75; I² = 35%) (Figure 3B). Secondary analysis was not performed because no zero groups in lymphoma incidence were reported.

Quiz Ref IDFour studies^8,9,20,24 reported effect estimates for risk of skin cancer (melanoma and keratinocyte carcinoma). Primary analysis found a pooled RR of 0.72 (95% CI, 0.44-1.18; I² = 57%) for risk of all skin cancer (eFigure 1 in the Supplement). There was a lower risk of melanoma (RR, 0.54; 95% CI, 0.33-0.86) and no association with keratinocyte carcinoma (RR, 1.03; 95% CI, 0.93-1.15) (Figure 4A).

Secondary analysis including zero groups for risk of all skin cancer yielded a similar overall effect with substantial heterogeneity (RR, 0.70; 95% CI, 0.30-1.61; I² = 81%). Heterogeneity was reduced with subgroup analysis of melanoma (RR, 0.60; 95% CI, 0.26-1.38; I² = 51%) and keratinocyte carcinoma (RR, 0.67; 95% CI, 0.18-2.46; I² = 55%). Heterogeneity was further reduced (I² = 0%) with removal of zero groups in subsequent sensitivity analyses for both melanoma and keratinocyte carcinoma (eFigure 1 in the Supplement).

Three studies^9,21,24 reported effect estimates for risk of skin cancer (melanoma and keratinocyte carcinoma) with TCIs vs TCSs, with a pooled RR of 1.13 (95% CI, 0.99-1.28; I² = 72%) for all skin cancer. Subgroup analysis yielded a pooled RR of 1.03 (95% CI, 0.78-1.37) for melanoma and 1.16 (95% CI, 0.99-1.36) for keratinocyte carcinoma (eFigure 2 in the Supplement). Secondary analysis including zero groups resulted in a pooled RR of 1.04 (95% CI, 0.87-1.24; I² = 78%) for all skin cancer, and subgroup analyses resulted in risks of 1.02 (95% CI, 0.84-1.24) for melanoma and 1.05 (95% CI, 0.83-1.33) for keratinocyte carcinoma (eFigure 2 in the Supplement).

Pooled analysis for all skin cancer in children-only cohorts resulted in a pooled RR of 1.01 (95% CI, 0.22-4.62; I² = 29%).^9,21 Secondary analysis including zero groups resulted in a decreased association (RR, 0.74; 95% CI, 0.31-1.78; I² = 0%).

We included 3 case-control studies^26-28 and a fourth nested case-control analysis within the cohort study by Schneeweiss et al²³ (eTable 6 in the Supplement). The case-control study by Arellano et al²⁷ and the nested case-control analysis by Schneeweiss et al²³ examined lymphoma, with both having very wide CIs overlapping the null when comparing TCI-exposed with unexposed groups and exposure to different combinations of TCIs with each other. One study²⁸ examining the risk of keratinocyte carcinoma reported significantly decreased odds of keratinocyte carcinoma with tacrolimus exposure (odds ratio, 0.43; 95% CI, 0.30-0.60). The remaining case-control study²⁶ did not report effect estimates because no cases of lymphoma were found in any groups aside from the control group, where 3 cases were reported.

One cohort study reported hazard ratios for risk of keratinocyte carcinoma with a nonactive comparator, stratified by dose of tacrolimus (0.03% vs 0.1%) and frequency of application.²⁴ The adjusted hazard ratios for risk of keratinocyte carcinoma by dose were 0.93 (95% CI, 0.65-1.33) for low-dose (0.03%) tacrolimus and 0.97 (95% CI, 0.84-1.12) for high-dose (0.1%) tacrolimus. Risk of keratinocyte carcinoma by frequency of application was 0.58 (95% CI, 0.39-0.86) for low frequency and 0.92 (95% CI, 0.82-1.02) for high frequency for any TCI.²⁴

Two cohort studies^21,23 reported effect estimates for risk of lymphoma, stratified by cumulative amount used, in comparison with TCS-treated groups. Schneeweiss et al²³ examined only pimecrolimus in comparison with a group treated with 60 g or less of TCSs and reported propensity score–adjusted odds ratios for pimecrolimus doses of 1.36 (95% CI, 0.68-2.70) for 60 g or less, 2.98 (95% CI, 0.85-10.5) for 61 to 100 g, and 4.17 (95% CI, 1.06-16.4) for more than 100 g. Castellsague et al²¹ reported adjusted incidence rate ratios of 1.40 (95% CI, 1.00-1.97) for tacrolimus doses of 0.05 g or less, 1.06 (95% CI, 0.59-1.92) for doses of more than 0.05 to 0.10 g, and 2.27 (95% CI, 1.39-3.69) for doses of more than 0.10 g of tacrolimus. The corresponding estimates for pimecrolimus were 0.89 (95% CI, 0.57-1.37) for doses of 0.05 g or less, 2.25 (95% CI, 1.19-4.26) for doses of more than 0.5 to 1.0 g, and 2.55 (95% CI, 1.27-5.12) for doses of more than 1.0 g.

One case-control study²⁸ reported odds ratios for risk of keratinocyte carcinoma, stratified by number of tubes of topical medication used, where risk decreased as the number of tubes used increased. The corresponding odds ratios for pimecrolimus were 0.49 (95% CI, 0.36-0.68) for less than 1 tube, 0.40 (95% CI, 0.24-0.67) for 1 to 2 tubes, and 0.37 (95% CI, 0.19-0.74) for at least 3 tubes; for tacrolimus, 0.36 (95% CI, 0.25-0.54) for less than 1 tube, 0.33 (95% CI, 0.19-0.55) for 1 to 2 tubes, and 0.18 (95% CI, 0.08-0.42) for at least 3 tubes.²⁸

Funnel plots were generated based on primary analysis of reported effect estimates in the inverse variance method (eFigures 3-7 in the Supplement). Although most plots did not suggest publication bias, interpretation was unclear owing to the small number of studies.²⁹

In this systematic review and meta-analysis, we found that TCI use was not associated with increased risk of cancer overall. Use of TCIs was associated with elevated risk of lymphoma compared with nonactive comparators and TCSs. Risk of skin cancer did not appear elevated with TCI use, and in 1 case-control study there were reduced odds of keratinocyte carcinoma associated with TCI use.

Several mechanisms have been proposed for the development of malignant neoplasms after calcineurin inhibitor use. The immunosuppressive effects of calcineurin inhibitors leading to decreased surveillance of cancerous cells may contribute to tumor promotion. In addition, evidence for direct tumor induction by calcineurin inhibitors exists.^30-32 Systemic penetration of topical preparations is low,^8,33 so although biologically plausible, the overall risk of malignant neoplasms attributable to these topical anti-inflammatory agents is likely low.

Quiz Ref IDOnly 2 cases of solid tumors were reported in 25 000 patients treated with pimecrolimus, and no lymphoma in almost 10 000 patients treated with tacrolimus in clinical trials.³³ From their approval to 2010, the rate of lymphoma in those prescribed TCIs reported in the US Food and Drug Administration’s adverse event reporting system was lower than the rate seen in the general population.^33-36 A 2015 systematic review³⁷ examining the risk of lymphoma associated with TCIs found that risk ratios were elevated for both topical tacrolimus and pimecrolimus use, but the review included only 2 cohort studies and resulted in wide CIs. We added 3 new studies with nonactive comparators,^9,20,22 which increased the precision of our effect estimates. Many individual studies did not show statistically significant associations with lymphoma, but there was a consistently positive direction of association as reflected in low I² values, and 2 studies^21,23 found increased risk of lymphoma with higher cumulative TCI exposure. The association with lymphoma was stronger in studies with a nonactive comparator, as opposed to those that compared TCI and TCS use, indicating that some of the association is likely a result of confounding by indication. Lymphoma is rare, with an annual worldwide incidence of 1.35 per 100 000 in children and 9.88 per 100 000 in adults.³⁸Quiz Ref ID The 35% increased relative risk found in our study for TCIs compared with TCSs would therefore result in estimated numbers needed to harm for lymphoma of more than 200 000 in children and almost 30 000 in adults.

Some included studies were small with relatively short follow-up periods, which limit their ability to determine the risk of malignant neoplasm induction with long latency periods. Although we examined lymphoma as a single entity in our primary analysis, it represents a heterogeneous group of diseases, which could bias our results toward the null if a true association exists for only one or some lymphoma subtypes. Our meta-analyses could have double-counted participants, for instance, those taking both tacrolimus and pimecrolimus in the same study; we performed sensitivity analyses to account for this possibility, and the results remained robust. Given the observational design of the included studies, unmeasured confounding limits interpretation of association and causation. Atopic dermatitis itself may be associated with increased risk of lymphoma^37,39 and keratinocyte carcinoma.⁴⁰ In addition, there may be a severity gradient with worse skin disease (and associated increased systemic inflammation) associated with further increased risk of cancer. Therefore, our results could be influenced by diagnostic bias and confounding by indication; our analyses examining the risk of malignant neoplasms with TCS comparators among patients with the same underlying clinical diagnosis (atopic dermatitis) mitigate this concern, but these biases are still possible because TCIs may be preferentially prescribed for more severe or refractory dermatitis. Studies have also suggested the possibility of reverse causation and protopathic bias, in which early cases of lymphoma, including cutaneous T-cell lymphoma, may be misdiagnosed as atopic dermatitis and treated with TCIs, leading to overestimation of the effect of TCIs.^8,21,37 Three cohort studies^8,20,23 used lag times between TCI exposure and lymphoma diagnosis to mitigate protopathic bias, but these lag times were relatively short (5 to 6 months).

Overall, our findings suggest an association between TCI use and risk of lymphoma but with no increased risk of other cancers, including skin cancers. Given that the absolute risk of lymphoma is low, particularly in children, the increase in relative risk translates to a very small increase in the absolute risk of lymphoma for a given patient.

Accepted for Publication: February 2, 2021.

Published Online: March 31, 2021. doi:10.1001/jamadermatol.2021.0345

Corresponding Author: Aaron M. Drucker, MD, ScM, Division of Dermatology, University of Toronto, 76 Grenville St, Toronto, ON M5S 1B2, Canada (aaron.drucker@wchospital.ca).

Author Contributions: Ms Lam 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: Lam, Tadrous, Drucker.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Lam, Tadrous.

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

Statistical analysis: Lam.

Supervision: Tadrous, Drucker.

Conflict of Interest Disclosures: Dr Tadrous reported consulting for the Canadian Agency for Drugs and Technologies in Health (CADTH). Dr Drucker reported receiving compensation from the British Journal of Dermatology (as a reviewer and section editor), the American Academy of Dermatology (as a guidelines writer), and the National Eczema Association (as a grant reviewer); consulting for CADTH; and receiving honoraria from CME Outfitters. No other disclosures were reported.

Additional Contributions: Stephanie Sanger, MLIS, and the research consultation team at McMaster University, Health Sciences Library, Hamilton, Ontario, Canada, assisted with the development of the search strategy, for which they were not compensated.