Nonmelanoma Skin Cancer Risk in Patients

469 rohibited. p s i e 3 l – c 1 i t r a s i h t f o n o i t c u d o r p e r d e ee of risk varies depending z i r o mmunosuppressive therapies † h t u a n U . c n I , h t l ], IL-12/17, and alpha-4/beta-7 integrin). c therapeutic agent used and the dose a e a ﬁ H r e w u l K s r e s Web site (www.dermatologicsurgery.org). The malignancy type and degr on the speci received. Among the commonly used immune suppressive agents in the treatment of IBD,6-mercaptopurine azathioprine, constitute and the thiopurine alpha [TNF Patients who have received i such as those listed above have been noted toincreased present risk with of various liquid and solid malignancies. t ’ l o W y b d e h Department of Dermatology, University of Pittsburgh Medical Center, s DOI: 10.1097/DSS.0000000000001455 i † l b · u P . 480 c – n I , and Melissa A. Pugliano-Mauro, MD y r † e g r u S c i g ammation localized to the s to tumor necrosis factor o ﬂ l o t (2) ulcerative colitis (UC) a ch inhibit generalized m r e D r o Dermatol Surg 2018;44:469 f y · t e i c o S

n a c i ammatory disease of the gastrointestinal tract that r ammatory bowel disease (IBD) is a chronic

e fl fl ammatory pathways (e.g., sulfasalazine, mesalazine, m n in fl A © 2018 by theISSN: American 1076-0512 Society for Dermatologic Surgery, Inc. Published by Wolters Kluwer Health, Inc. All rights reserved.

*Forefront Dermatology, Wexford, Pennsylvania; I The primary authorindicated (J.W.H.) no was significant funded interest through with an commercial ACGME-approved supporters. training program. The authors have therapy, although additionalpurines data seem are to neededpurine be therapy to at should support greater help risk this direct of skin trend. NMSC. cancer Younger screening Appreciating patients recommendations NMSC with and risk sun IBD in protective using patients measures. thio- with IBD undergoing thio- LIMITATIONS of Heterogeneity study designs limited direct comparisons of thiopurine exposureCONCLUSION and NMSC risk. Patientsproportional with to IBD therapy using thiopurines duration. seem Risk to have of a NMSC moderately seems increased risk to of decrease NMSC that or is return to baseline after discontinuing were included. Publications were evaluated using PRISMARESULTS guidelines. The systematic review yielded 67 articles; 18 met final inclusion criteria. METHODS Systematic“ulcerative review colitis,” of“non-melanoma,” “Crohn’s PubMed “squamous disease,” was cell performed “thiopurine,” carcinoma,” with and “azathioprine,” “basal keywords “6-mercaptopurine,” cell “inflammatory carcinoma.” “skin bowel All disease,” cancer,” available publication years thiopurine use in IBD. OBJECTIVE Thewith IBD aim using of thiopurine this therapy. study was to systematically review current literature on NMSC risk in patients BACKGROUND Azathioprineinflammatory bowel and diseaseincrease (IBD). 6-mercaptopurine nonmelanoma skin Although cancers (thiopurines) thiopurine (NMSCs), dermatologic therapy literature are in yields less organ common data regarding transplant NMSC adjunct recipients risk of is treatments known to for Joshua W. Hagen, MD, PhD* Therapy: A Systematic Review of the Literature Nonmelanoma Skin Cancer Risk inInflammatory Patients Bowel With Disease Undergoing Thiopurine Pittsburgh, Pennsylvania Supplemental digital content isin available the for this full article. text Direct and URL PDF citations versions appear of in the this printed article text on and the are journal provided newer monoclonal antibodie conditions, most of whi in systemic corticosteroids, thiopurines, methotrexate, and gastrointestinal tract, and which is characterized by in colon. An array of treatments has been described for both is loosely divided into 2 majordisease categories: (CD) (1) which Crohn occurs at any location along the e h t y b

1 0 2 ©

Downloaded from https://journals.lww.com/dermatologicsurgery by uj5n6iU8d2wpALGLKKm7RBfhB/TStktufUfUIBd3Kxr54oBVlg1o2dw7YdZUCTPzItwZ11Clr7YP1jILjGqDdSTK1E5Ce+T8K/TGVXmRsFwq4ZjAM34DnNZ5lFW30fUyykS250k7frs= on 04/19/2018 Downloaded from https://journals.lww.com/dermatologicsurgery by uj5n6iU8d2wpALGLKKm7RBfhB/TStktufUfUIBd3Kxr54oBVlg1o2dw7YdZUCTPzItwZ11Clr7YP1jILjGqDdSTK1E5Ce+T8K/TGVXmRsFwq4ZjAM34DnNZ5lFW30fUyykS250k7frs= on 04/19/2018 NONMELANOMA SKIN CANCER RISK IN IBD PATIENTS

therapeutic family that is frequently used as adjunctive setting of thiopurine use were also excluded. Of the therapy. Pharmacologically, these act as prodrugs that initial 67 publications, 18 met the final inclusion cri- are metabolized to deoxythioguanine-59-triphosphate, teria (Figure 1). The reference sections from the which serves as a purine analogue for DNA poly- selected publications were reviewed thoroughly for merases, preferentially blocking DNA synthesis in cells evidence of any additional relevant articles that were lacking a nucleotide salvage pathway (i.e. activated T- not captured by the initial search strategy, but no and B-cells).4 Both azathioprine and 6-mercaptopurine additional relevant publications were discovered. The are known photosensitizers that reduce the minimal complete list of publications included for systematic erythema dose for UV-A radiation and also play a role review and their characteristics are provided in in generation of reactive oxygen species.5–7 The risk of Table 1.11–28 The selected studies were evaluated solid and liquid malignancies seems to be increased in according to PRISMA guidelines.29 patients with IBD who have undergone thiopurine therapy.8 Similar patterns have been frequently Results observed for solid organ transplant patients who have used thiopurine therapy as part of their immunosup- Article Selection Process pressive regimen, but these patients are often main- tained at a higher level of immune suppression The Pubmed search strategy (see Supplemental Digital compared with well-controlled patients with IBD.9,10 Content 1, Appendix I, http://links.lww.com/DSS/A86) produced a total of 67 articles that made some mention The dermatology and dermatologic surgery literature has of IBD, thiopurine therapy, and skin cancer. The search devoted significant recent attention to the increased skin produced a partial internal validation in the form of cancer risk in solid organ transplant recipients, but few a meta-analysis conducted in 2014, which had analyzed publications exist in the authors’ literature to address the all relevant publications up through 2012. The authors’ extent to which skin cancer risk extends to patients with search strategy identified all the articles that were 20 IBD using similar therapeutic agents, albeit at lower doses. included in that meta-analysis (8 total). Case reports This review aims to address the question of the degree of (10 total) and reviews (30 total) were excluded from the risk thiopurine therapy confers for development of non- final analysis due to low level of evidence (LOE) or lack melanoma skin cancer (NMSC) in patients with IBD. of novel data synthesis, respectively. One meta-analysis was included for final review due to its novel synthesis of data from previous studies.20 Review of the Materials and Methods abstracts and/or full text of the remaining 27 articles A systematic review was performed using PubMed/ Medline to collect publications on risk of developing NMSC in patients with IBD undergoing thiopurine therapy. The keywords used in the search were “inflammatory bowel disease,”“ulcerative colitis,” “Crohn’s disease,”“thiopurine,”“azathioprine,” “6-mercaptopurine,”“skin cancer,”“non-melanoma,”“squamous cell carcinoma,” and “basal cell carcinoma”. The actual search language string is included in Supplemental Digital Content 1, Appendix I, http://links.lww.com/DSS/A86. The search yielded 67 potential publications. Case reports/series and review articles were excluded from final inclusion due to low level of clinical evidence. Remaining publications that were deemed by abstract or full-text review to not address the topic of NMSC risk in the Figure 1. Search strategy for systematic review.

470 DERMATOLOGIC SURGERY

© 2018 by the American Society for Dermatologic Surgery, Inc. Published by Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited. TABLE 1. Summary of Publications Included in Systematic Review

LOE Mean (USPSTF/ Reference Study Type Year Country Drug(s) Exposure Cases Controls Follow-Up CEBM) Armstrong and Nested case control 2010 United Kingdom AZA >1 Rx 43 15,398 6.4 yrs II/2b colleagues11 Long and Retrospective cohort; 2010 USA AZA/6MP $1 Rx, <90 d from 742 2,968 1.32 yrs II/2b colleagues12 nested case control NMSC Setshedi and Retrospective cohort 2011 South Africa AZA/6MP TP use (any duration) 123 691 9.9 yrs II/2b colleagues13 Singh and Retrospective cohort; 2011 Canada AZA/6MP 2 + Rx 170 680 11.7 yrs II/2b colleagues14 case control Peyrin-Biroulet Prospective 2011 France AZA/6MP TP use (any duration) 8,676 10,810 2.55 yrs II/2b and colleagues15 observational cohort van Schaik and Retrospective cohort 2011 Holland AZA/6MP $50 mg AZA/6MP 819 2,068 6.46 yrs II/2b colleagues16 over 6 mo Camus and Retrospective cohort; 2012 France AZA 2.25 mg/kg AZA · 1 yr 220 440 12.17 yrs II/2b colleagues17 case control Long and Retrospective cohort; 2012 USA AZA/6MP $1 Rx 3,288 12,945 2 yrs II/2b colleagues18 nested case control Go´ mez-Garcıa´ and Retrospective 2013 Spain AZA/6MP Any use 429 383 12.19 yrs II/2b colleagues19 observational cohort Ariyaratnam and Meta-analysis (of 8 2014 N/A (United AZA/6MP Per included studies 14,081 46,000 3.81 yrs II/2a Subramanian20 previous studies) Kingdom) (weighted avg.) Abbas and Retrospective cohort; 2014 USA AZA/6MP Person-years of any 3,346 14,527 8.1 yrs II/2b colleagues21 nested case control TP exposure Beigel and Retrospective cohort 2014 Germany AZA/6MP 6 TNFa 2.0–2.5 mg/kg AZA 262 404 5.66 yrs II/2b colleagues22 PUGL AND HAGEN McKenna and Database inquiry 2014 USA AZA/6 MP and TP use in TNFa trial 241 752 N/A III/2c colleagues23 (adverse event) TNFa (FAERS) Osterman and Retrospective cohort; 2014 USA AZA/6 MP + ADA ADA alone, ADA + TP 563 900 1.5 yrs II/2b colleagues24 nested case control vs ADA alone or MTX, ADA + TP (median) only

44ARL21 471 2018 44:4:APRIL Kopylov and Nested case control 2015 Canada AZA/6MP, MTX #12 mo, 13–36 mo, or 474 4,680 No reported II/2b

colleagues25 and TNFa >36 mo exposure mean IANO-MAURO (>1 yr, all patients)

© 2018 by the American Society for Dermatologic Surgery, Inc. Published by Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited. NONMELANOMA SKIN CANCER RISK IN IBD PATIENTS

identiﬁed an additional 9 that did not provide data regarding the question of NMSC risk in patients with IBD using thiopurines. In total, there were 18 articles LOE

CEBM) that met criteria for inclusion in the ﬁnal review (USPSTF/ (Figure 1).

Study Characteristics and Level of Evidence Mean

Follow-Up Fourteen of the selected studies were population or

rosis factor; TP, thiopurine. patient database cohort studies (13 retrospective design and 1 prospective design). Six of the 14 cohort studies also provided a nested case-control analysis of the data12,14,17,18,21,24; 2 additional studies were designed as nested case-control analyses without 381 2,407 2.24 yrs II/2b a population cohort.11,25 In addition to these, there was 1 adverse event database inquiry23 and 1 meta- analysis that were included20 (Table 1).

There was signiﬁcant heterogeneity in the deﬁnition

Any TP use 1,084 874 9.8 yrsof thiopurine II/2b exposure used in each of the studies, least 1 Rx after incident NMSC

2 Rx in 4 mo, with at with some only requiring history of any thiopurine $ use during the study period, whereas others deﬁned speciﬁc dose and duration of use as inclusion crite-

a a ria. Some studies were designed to compare use of

TNF TNF biologic immunotherapy with and without thiopurines, and therefore did not include a thiopurine- only group (Table 1). The number of cases (or cohort size) is listed along with control population size and the mean follow-up (ranging 1.5–12.19 years) in Table 1.

The LOE was assessed by scales developed by the United States Preventive Services Task Force and the Oxford Center for Evidence-Based Medicine. The LOE ranged from II/2a to III/2c, with most of the included articles assigned to II/2b LOE (n = 16, 89%) (Table 1).

Outcome Measures and Statistical Risk

Most of the studies included in the authors’ analysis Retrospective cohort 2016Retrospective cohort 2016 USARetrospective cohort the Netherlands 2017 AZA/6MP/TG AZA/6MP and Ireland >365 d of use AZA/6MP and 456sought 2,496 to record 8.1 yrs any II/2b incidence of NMSC during any

27 portion of the follow-up period (n = 15, 83%). A

26 28 smaller number of studies set time parameters for occurrence of NMSC within the assessment period21,24 (n = 2, 11%) and 1 study sought to determine the time colleagues colleagues and colleagues to incidence of a second NMSC after a primary inci- TABLE 1. (Continued) ReferenceScott and Study Type Year6MP, 6-mercaptopurine; ADA, adalimumab; Country AZA, azathioprine; LOE, level of evidence; NMSC, nonmelanoma skin cancer; TG, thioguanine; TNF, tumor nec Drug(s) Exposure Cases Controls Clowry and van den Heuvel dent NMSC was identiﬁed26 (Table 2).

472 DERMATOLOGIC SURGERY

© 2018 by the American Society for Dermatologic Surgery, Inc. Published by Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited. TABLE 2. Outcomes, Statistical Measures, and Limitations of Reviewed Publications

Reference Outcome Statistical Measure(s) Limitations/Risk of Bias/Comments Armstrong and NMSC incidence in AZA users OR: 0.99 (95% CI, 0.35–2.81) Selection bias in the form of community provider– colleagues11 only database in areas where secondary care cen- ters provide significant portion of AZA Rx’s; ascertainment bias is likely; study only analyzes AZA users and not 6MP Long and First NMSC Dx and Tx (claims based) after Cohort study, IRR: 1.64 (95% CI, 1.51–1.78) Selection criteria limit study population to those 12 colleagues IBD diagnosis Case-control study, recent use OR: 3.56 (95% CI, aged <64 yrs; retrospective cohort controls were 2.81–4.50); persistent use adjusted OR: 4.27 non-IBD (nested case study had IBD controls); (3.08–5.92) exposure assessment was pharmacy claim–based but without dose information; detection/ascertainment bias is not controlled for in cohort analysis; relatively short mean duration of follow-up Setshedi and Any cancer (with NMSC subset) identified by RR, 4.9 (95% CI, 1.1–21.8); white subset, RR 1.4 Small study population likely to affect risk colleagues13 chart review in IBD database (2.3–47.7) calculations; study population is mix of white, mixed race, and African patients who have very different NMSC risk susceptibility; small number of patients with NMSC prevented dose–response effect analysis; UV exposure levels do not corre- spond well to other studies reviewed Singh and SCC and BCC cases occurring in IBD Cohort study, SCC HR: 5.40 (95% CI, 2.00– Database did not include use of colleagues14 database 14.56); BCC HR: 1.12 (0.68–1.85) immunosuppressant medications before 1995; Case-control study, SCC OR: 20.52 (2.42– coding and capture of NMSC had not been 173.81); BCC OR: 2.07 (1.10–3.87) previously validated in the database used; this study was the only study to differentiate between SCC and BCC risk profiles; examined the effect of dose on NMSC risk Peyrin-Biroulet NMSC Dx during cohort entry period Ongoing TP, HR: 5.9 (95% CI, 2.1–16.4; CESAME cohort included a younger patient 15

and colleagues p = .0006) population than many other studies reviewed, PUGL AND HAGEN Past TP, HR: 3.9, (1.3–21.1; p =.02) which may overestimate the HR due to low incidence of NMSC in age-matched controls van Schaik and NMSC Dx during/after TP use Unadjusted HR: 0.94 (95% CI, 0.58–1.50); Small study sample size limits conclusions; colleagues16 adjusted HR: 0.85 (0.51–1.41) inclusion criteria more strict than other studies, limiting direct comparisons 44ARL21 473 2018 44:4:APRIL IANO-MAURO

© 2018 by the American Society for Dermatologic Surgery, Inc. Published by Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited. 474 OMLNM KNCNE IKI B PATIENTS IBD IN RISK CANCER SKIN NONMELANOMA EMTLGCSURGERY DERMATOLOGIC TABLE 2. (Continued)

Reference Outcome Statistical Measure(s) Limitations/Risk of Bias/Comments Camus and Incident cancers (w/NMSC subset) occurring No statistical measure provided; 1.8% of TP Only analyzes AZA users and only patients with CD colleagues17 in registry of patients with CD responders and 0% of controls developed (not UC); data collected from single-center tertiary NMSC care facility’s database; high percentage of patients lost to follow-up (19%); very small number of NMSCs in exposed group and none in control group (limited power); no statistical analysis specific to NMSC risk was provided by authors Long and First NMSC Dx and Tx (claims based) Cohort study, IRR: 1.46 (95% CI, 1.40–1.53); Selection criteria limit study population to those colleagues18 after IBD diagnosis adjusted HR: 1.34 (1.28–1.40) aged <64 yrs; study has 6 mo lead time or Case-control study, any TP use, adjusted OR: screening time which may underestimate affected 1.85 (95% CI, 1.66–2.05); TP use with anti-TNF cases; study uses administrative claims data that is agent, OR: 3.89 (2.33–6.46) subject to misclassification of exposure/outcomes; no data on TP dosing are obtainable; relatively short mean duration of follow-up; study is large and has robust statistical power and was controlled for health care utilization Go´ mez-Garcıa´ and Any malignancy (included NMSC as subset) Not provided; post hoc calculation of RR: 1.19 Small study, overall study population, and small colleagues19 in TP user (95% CI, 0.42–3.40, p = .74) number of controls; no power calculation performed; imprecise definition of TP exposure; incidence rates on study population may be influenced by UV exposure of study location (Granada, Spain) Ariyaratnam and NMSC Dx while on TP Pooled-adjusted HR: 2.28 (95% CI, 1.50–3.45) Significant heterogeneity of studies included in Subramanian20 meta-analysis (population characteristics and duration of follow-up contributed heavily); HR was heavily influenced by hospital-based populations and shorter duration of follow-up; variations in study design within the meta-analysis cannot be controlled for; provides the largest collective dataset and robust external validity Abbas and NMSC Dx (and Tx) at least 6 mo from study HR: 2.1 (95% CI, 1.6–2.6) (HR 0.7, 95% CI, 0.5– Patient population limited to VA health care system colleagues21 start; also measured risk after stopping TP 1.0, after stopping TP) (older, white, male patient bias) which impacts and cumulative incidence by increasing IRR: first year, 1.6 (p = .09); second year, 2.1 external validity; did account for ascertainment years of exposure (p = .005); third year, 2.2 (p = .007); fourth year, bias which decreases magnitude of HR; 2.1 (p = .04); fifth year, 3.6 (p < .0001); 5 + year, socioeconomic status of study population better 2.9 (p < .0001) matches general population than tertiary care center populations; controlled for relative UV exposure levels across the study population

Reference Outcome Statistical Measure(s) Limitations/Risk of Bias/Comments Beigel and Any malignancy (with NMSC subset) 8 cases of NMSC in thiopurine subset (vs zero Very small study size limits the statistical power of colleagues22 occurring during study period in anti-TNF group); no statistical calculation this analysis; there were no NMSC events in the provided, post hoc OR: 27.0 (1.55–470) comparison group (anti-TNF), which does not match trends noted in other studies; no statistical analysis is provided by the authors (our post hoc OR is provided for ease of comparison; confidence interval is wide) McKenna and Incident NMSC while on anti-TNF therapy PRR, augmented odds of NMSC with Adverse event database (FAERS) is limited by colleagues23 plus TP compared with anti-TNF alone concomitant TP and anti-TNF (N = 51, p < .001) a spontaneous reporting bias; underreporting and vs with anti-TNF alone (N =93,p = .036) overreporting are not controlled for; does not specify dose or duration of exposure; database was heavily skewed toward patients with CD (not characteristic of IBD population as a whole); statistical measures difficult to compare with other studies Osterman and Incident NMSCs on ADA alone, ADA + any TP + ADA vs ADA alone, unadjusted RR: 3.63 Crohn disease only; no immunomodulator-only colleagues24 immunomodulator, or ADA + TP, (up to 70 (95% CI, 1.14–11.58); adjusted RR: 4.01 group (i.e., TP only); short median follow-up time d after last ADA dose), Crohn disease only (1.24–13.00) compared with other studies which may underestimate NMSC in ADA only group; immunomodulator duration was not factored into risk calculations; did not account for relative UV exposure; combined randomized clinical trial data with prospective observational data; study had 2 comparison groups (general population and patients with CD) Kopylov and NMSCs after specified duration of therapies $3 yrs of TP, OR 1.41 (95% CI, 1.11–1.79) Younger patients are underrepresented in this study 25 colleagues selected >5 yrs of TP, OR: 2.07 (95% CI, 1.36–3.7) due to database characteristics; no information on After stopping TP, OR: 1.04 (0.69–1.55) disease severity; UV exposure was not controlled PUGL AND HAGEN for Scott and Second NMSC, 1 + year after incident NMSC HR: 1.49 (95% CI, 0.98–2.27) (HR 1.35, 95% CI, Older patient population (Medicare only); limited colleagues26 0.87–2.70 for short-term TP use) comparability with other studies because outcome was second NMSC 1 + years after incident NMSC. Exclusion criteria may have removed high incidence outliers that would further elevate HR 44ARL21 475 2018 44:4:APRIL and improve significance calculation IANO-MAURO van den Heuvel Any cancer Dx (with NMSC subset) obtained Total IBD cohort, SCC, SIR: 3.83 (95% CI, 1.83– Medication analyses were based on prescription and colleagues27 from chart review and Dutch cancer 7.04), TP-only exposure group, SCC SIR: 3.88 data only; HRs were not calculated which limits registry cross-reference. (1.04–9.93) comparison with other studies; highly reliable cancer diagnoses due to cross-reference strategy

Statistical risk assessments were very heterogeneous in the collection of reviewed studies. Seven studies (39%) reported hazard ratios (HRs) for NMSC risk associated with thiopurine treatment of IBD. These HRs ranged from 0.85 to 5.9 for recent or ongoing use of thiopurines and from 0.7 to 3.9 for past use of thiopurines (see Table 2 for specific HRs and confidence intervals). Most of the studies reporting HR indicated a ratio greater than 1 and 5 of the 7 reported confidence intervals that did not intercept 1 (n = 5 [of 7], 71% of studies reporting HR). A smaller set of studies BD, inflammatory bowel disease; IRR,

noma; SIR, standardized incidence ratio; TNF, (n = 4, 22%) reported odds ratios (ORs) for NMSC with thiopurine use in a range from 0.99 to 5.26, with 3 of 4 having signiﬁcantly increased odds

was available in the85 database; yrs patients were older excluded; than limitedand by use small of sample size aand hospital-based ascertainment database bias); (selection notexposure stratified by UV (Table 2).11,14,25,28 Incidence rate ratios (IRRs)were reported by 3 studies (17%), with a range from 1.46 to 2.9, and all 3 showed confidence intervals that fell above the significance cutoff.12,18,21 Relative risk calculation was provided by 1 study (4.9 [95% CI, 1.1–21.8])13; it was also calculated post hoc by the authors of this review for 1 study without statistical calculations provided specifically for NMSC risk (1.19 [95% CI, 0.42–3.40, p = .74]).19 One study provided a proportional reporting ratio for NMSCs identified in an adverse event database,23 and 1 study reported

OR: 5.26 (95% CI, 2.15–12.93) No dose or duration of immune suppressive therapy standardized incidence ratios (SIRs) for SCC in an IBD cohort with all treatment types (3.83 [95% CI, 1.83–7.04]) as well as a thiopurine-only IBD treatment group (3.88 [95% CI, 1.04–9.93]), both of which were statistically signiﬁcant.27

Limitations and Bias

Study-speciﬁc limitations and potential for bias are reviewed in Table 2. Of note, 2 studies were set up in a manner that only analyzed patients with CD, but not other IBD subtypes.17,24 Several studies were performed in databases that forced arbitrary age cut- national cancer registry offs.12,18,26 Two studies only analyzed azathioprine 11,17

cross-reference of IBD database and users but not 6-mercaptopurine. The potential

First NMSC of any subtype appearing in impact of these limitations and biases are discussed below.

28 Discussion

Through the collaborative efforts of dermatologists

colleagues and transplant surgeons, the authors now appreciate TABLE 2. (Continued) ReferenceClowry and 6MP, 6-mercaptopurine; AZA, azathioprine;incidence BCC, rate basal ratio; cell NMSC, Outcome nonmelanoma carcinoma;tumor skin CD, cancer; necrosis Crohn OR, factor; odds disease; TP, ratio; CI, PRR, thiopurine; conﬁdence proportional Tx, reporting interval; treatment; ratio; Dx, RR, UC, diagnosis; relative ulcerative risk; HR, colitis. SCC, hazard squamous ratio; cell I carci Statistical Measure(s) Limitations/Risk of Bias/Comments the degree to which immune suppression confers an

476 DERMATOLOGIC SURGERY

estimated 65- to 250-fold increase in NMSCs, as well using thiopurines.20 Among the publications as the significant morbidity and mortality associated reviewed, studies with smaller case cohorts were more with them.30 This knowledge has translated into likely to report lack of statistical significance for widespread aggressive, longitudinal screening pro- reported risk.11,16,19 grams and increased awareness at both the physician and patient level. By comparison, the dermatologic By asking whether thiopurine therapy (vs anti-TNF literature has directed much less attention to the agents) increases the risk of a second incident NMSC, potential risk of NMSC in other autoimmune disease Scott and colleagues sought to address the danger of states that require similar prolonged immune sup- maintaining patients on these treatments (for short or pression using steroid-sparing agents. Skin cancer longer courses) once they have identified themselves as screening guidelines are lacking for these patients, as is sensitized individuals by virtue of a primary, incident consensus on the relative risk attributable to various NMSC. Their analysis, however, indicated that short- therapeutic agents. term thiopurine therapy (HR 1.53; 95% CI, 0.87– 2.70) and 1 year or more of thiopurine therapy (HR Risk in Inflammatory Bowel Disease Versus 1.49; 95% CI, 0.98–2.27) was not statistically asso- Noninflammatory Bowel Disease ciated with increased risk of a second incident NMSC, although a trend toward increased risk was noted.26 The risk of NMSC in the IBD patient population This conclusion may have been influenced by exclu- compared with unaffected control patients was eval- sion criteria that removed patients with new NMSCs uated in a subset of health care database studies during the first 6 months after the primary incident included in this review. Confirming what has been NMSC. The authors’ experience (authors J.W.H. and suggested in previous reports on the risk of general M.A.P.-M.) in treating patients with IBD using thio- malignancies, the risk of NMSC within these pop- purine therapy would suggest that there is a subset of ulations was increased with an IRR of 1.46 to patients with rapid and frequent development of non- 1.64.12,18 A large French cohort reported the SIR of NMSC in their patients with IBD at 2.71 (95% CI, preexisting NMSC that might be missed by applica- 1.83–3.87; p < .0001), which includes those who have tion of this stringent exclusion criterion. and have not received thiopurine therapy.15 Data Dose and Duration Effect derived from a Canadian cohort suggested that increased risk of NMSC in patients with IBD as From a theoretical and practical perspective, the a whole was only present for those younger than 50 degree to which the immune system is suppressed years of age, beyond which the risk returned to base- seems to correlate with frequency and severity of 14 line. These data are all consistent with reports of suppression-related sequelae. Solid organ transplant NMSC risk in patients with IBD predating the era of recipients who receive high levels of immune sup- 31 common use of immunosuppressants as treatment. pression present with an elevated risk profile for NMSC compared with the level of suppression typi- Risk Attributable to Thiopurine Use cally seen in patients with IBD.30 In so far as it was Large health care database cohorts have provided the specifically addressed in a subset of the reviewed arti- most robust data on the NMSC risk attributable to cles, however, azathioprine and 6-mercaptopurine did thiopurine use. The VA health care system analysis not show a daily dose-dependent risk at the doses that showed an adjusted HR of 2.1 for development of are typically used to treat patients with IBD.21 This NMSC while on thiopurine therapy.21 Similarly, may indicate the presence of a threshold effect, increased risk was noted in other large health care wherein typical daily doses of thiopurine immuno- database cohorts12,15,18,25 (Table 2). The largest col- suppression used in IBD supercede the dose required to lection of patient data is provided by a meta-analysis of observe sensitization but do not alter the order of 8 early studies, which indicated an adjusted NMSC magnitude of NMSC risk (as seen in transplant HR of 2.28 (95% CI, 1.50–3.45) in patients with IBD patients).

44:4:APRIL 2018 477

However, several studies suggested that overall Residual Risk duration of therapy (i.e. total cumulative dose) did The question of whether NMSC risk is modifiable in affect risk profile, with longer use of thiopurines affected patients with IBD by virtue of discontinuing translating to escalating NMSC risk.22 Incidence rate thiopurine therapy has been addressed by a handful ratio rose from 1.6 (p =.09)toashighas3.6 of studies. From a biochemical perspective, incorpo- (p < .0001) by the fifth year and remained high (IRR ration of modified purine analogues (6-thioguanine, 2.9, p < .0001) for treatment extending beyond peak absorbance 342 nm) into the DNA of replicat- Year 5.21 Similarly, a 5-year treatment duration was identified as a threshold for further increasing NMSC ing cells (e.g. basal keratinocytes) renders their DNA – 6,7 risk in a Canadian cohort.25 Comparing recent (#90 sensitive to UV-A induced mutagenesis. Although days) versus persistent (>365 days) thiopurine use, the authors traditionally recognize UV-B radiation as Long and colleagues12 demonstrate a similar trend the primary driver of mutagenic events and counsel toward increased risk with longer therapy (OR 3.56 patients regarding UV-B protective habits, UV-A vs 4.27, respectively). A Canadian cohort demon- wavelengths pass through glass and constitute a more strated that a cumulative dose of >5.7 g AZA/6-MP chronic, less-heralded source of mutagenesis in sen- could be used as a threshold to stratify those with sitized patients (e.g. thiopurine users). The process of significantly increased odds of developing BCC or cumulative UV-A–mediated mutagenesis during thi- SCC.14 In contrast to this, some authors find no sig- opurine use has been suggested to affect the tumor nificant risk from beginning to end of an extended suppressor, patched, which has a well-established therapeutic regimen.16 Additional high-quality role in NMSC development.32 A French, prospective cohort analyses may help to better define the effect of observational cohort of patients with IBD showed therapy duration on NMSC risk, but current evidence potential evidence of a residual effect of past thio- weighs heavy on the side of cumulative dose influ- purine exposure, with NMSC HR reported at 3.9 encing NMSC risk. (95% CI, 1.3–12.1; p = .02). This risk is reduced, however, from the reported HR for ongoing thiopurine therapy (5.9, 95% CI, 2.1–16.4).15 In com- Impact of Age parison, a VA cohort indicated that relative risk of Common knowledge supports that risk of NMSC NMSC returned to baseline after stopping thiopurine increases with advancing age, a phenomenon that is use (adjusted HR 0.7, p = .07), even after up to 5 years echoed by the various studies included in this review. of continual use.21 Similar findings were reported in Of note, however, is the fact that studies specifically a Canadian cohort.25 Although it seems clear that evaluating the risk of NMSC in IBD regarding patient removing thiopurines from the therapeutic approach age often identified exposure windows that more sig- reduces NMSC risk, the degree of residual risk fi 28 ni cantly impacted risk. Clowry and colleagues remains a question that requires additional study to suggest that thiopurine exposure between the ages of further clarify the true benefit of therapeutic sub- 30 to 50 and also those over 70 were age groups with stitution/discontinuation. greatest observed impact on NMSC risk. It is also interesting to note that population-based studies Risk of Alternative Therapies drawing from predominantly younger pop- ulations12,15,18 produced more robustly increased This analysis did not seek specifically to address NMSC risk than those studies cataloguing a decidedly a comparison of the NMSC risk between alternative older population.21,26 This suggests that beyond therapeutic options for IBD. However, several of the a certain age, the impact of thiopurine use on NMSC individual studies that the authors evaluated did risk is lessened due to accumulating age-related risk. It compare the relative risk of thiopurine therapy versus also highlights the importance of early preventative newer biologic therapies. In a study that failed to show screening for young patients who are not otherwise a statistical increase in second incident NMSC on thought of as high risk for NMSC. thiopurine therapy, anti-TNF therapy (i.e., infliximab,

478 DERMATOLOGIC SURGERY

adalimumab, certolizumab, golimumab, or eta- mean follow-up duration, which limits interpretation nercept) showed a HR of 1.36 (95% CI, 0.76–2.44) of long-term risk estimates. Studies that only evaluated and slightly lower (but statistically insignificant) azathioprine as an exposure cannot be reliably used to decreased incidence of a second NMSC while on assess risk on 6-mercaptopurine users. The 2 studies therapy.26 The authors argued that anti-TNF therapy which examined only patients with CD limit the ability does not demonstrate a clear benefit regarding NMSC to extend risk assessment to patients with UC, who risk because HR significantly overlap. In a larger have slightly different malignancy risk profiles. Addi- health care database population, patients with IBD tional limitations specific to each study reviewed are treated with recent anti-TNF agents did not show an noted in Table 2. increased risk of NMSC (adjusted OR 1.14; 95% CI, 0.95–1.36) but persistent use of biologic therapy alone Conclusion increased NMSC OR to 1.63 (95% CI, 1.12–2.36).18 This was still below the risk level of persistent thio- Nonmelanoma skin cancer risk seems to be moder- purine use or persistent thiopurine use combined with ately increased in patients with IBD undergoing thio- a biologic agent (OR, 2.72 and 3.89, respectively).18 purine therapy. This increased NMSC risk does not Osterman and colleagues24 showed that adalimumab seem to be daily dose-dependent but some studies therapy alone did not increase NMSC risk (SIR 1.20; show a duration-dependent (cumulative dose) trend 95% CI, 0.39–2.80), whereas addition of a thiopurine for increasing risk. Although there is a relative paucity to this regimen did increase NMSC risk significantly. of data directly addressing NMSC risk profile after As more data accumulate on the use of various bio- discontinuing therapy, there is evidence to suggest that logic agents in IBD treatment, the authors will be able residual NMSC risk decreases and/or returns to to make more confident assessments on risk profiles of baseline after stopping thiopurine therapy. There alternative therapies. However, sufficient data exist to seems to be disproportionately increased risk in suggest a trend for decreased overall risk of NMSC younger patients who, by virtue of their age, are not with single-agent biologic treatments. commonly targeted for routine skin screening. Dermatologists should educate patients with IBD Limitations using thiopurines on their increased NMSC risk, the importance of daily sun protective measures, and The publications included in this review represent benefits of regular skin cancer screening starting at a range of study designs that introduce significant a younger age. heterogeneity at the level of population selection (e.g., age, nationality, and care setting), definition of thio- Acknowledgments The systematic review was purine exposure, duration of follow-up, and approach performed and written solely by the authors. to statistical analysis of attributable risk. All studies, excluding one, were retrospective in nature and are References therefore subject to selection and misclassification 1. Kandiel A, Fraser AG, Korelitz BI, Brengsinger C, et al. Increased risk of bias. Some, but not all, of the studies made attempts to lymphoma among inflammatory bowel disease patients treated with address these potential biases. Several of the studies azathioprine and 6-mercaptopurine. Gut 2005;54:1121–5. described comparably small IBD patient populations 2. Madanchi M, Zeitz J, Barthel C, Samaras P, et al. Malignancies in fl and their conclusions are often as odds with the con- patients with in ammatory bowel disease: a single-center experience. Digestion 2016;94:1–8. sensus provided by meta-analysis and larger inde- 3. Pereira R, Faria R, Lago P, Torres T. Infection and malignancy risk in pendent studies. Larger, nationwide health care patients treated with TNF inhibitors for immune-mediated database inquiries have provided robust statistical inflammatory diseases. Curr Drug Saf 2017;12:162–170. power but the data are sometimes limited to specific 4. Karran P. Thiopurines, DNA damage, DNA repair and therapy-related cancer. Br Med Bull 2006;79–80:153–70. patient populations that individually limit their gen- eralizability (e.g., age group limitations). Several of the 5. Hemmens VJ, Moore DE. Photochemical sensitization by azathioprine and its metabolites-II. Azathioprine and nitroimidazole metabolites. large health care database studies provided shorter Photochem Photobiol 1986;43:257–62.

44:4:APRIL 2018 479

6. O’Donovan P, Perret CM, Zhang X, Montaner B, et al. Azathioprine 21. Abbas AM, Almukhtar RM, Loftus EV Jr, Lichtenstein GR, et al. Risk and UVA light generate mutagenic oxidative DNA damage. Science of melanoma and non-melanoma skin cancer in ulcerative colitis 2005;309:1871–4. patients treated with thiopurines: a nationwide retrospective cohort. Am J Gastroenterol 2014;109:1781–93. 7. Zhang X, Jeffs G, Ren X, O’Donovan P, et al. Novel DNA lesions generated by the interaction between therapeutic thiopurines and UVA 22. Beigel F, Steinborn A, Schnitzler F, Tillack C, et al. Risk of malignancies light. DNA Repair (Amst) 2007;6:344–54. in patients with inflammatory bowel disease treated with thiopurines or anti-TNF alpha antibodies. Pharmacoepidemiol Drug Saf 2014;23: 8. Lichtenstein GR, Rutgeerts P, Sandborn WJ, Sands BE, et al. A pooled 735–44. analysis of infections, malignancy, and mortality in infliximab- and immunomodulator-treated adult patients with inflammatory bowel 23. McKenna MR, Stobaugh DJ, Deepak P. Melanoma and non-melanoma disease. Am J Gastroenterol 2012;107:1051–63. skin cancer in inflammatory bowel disease patients following tumor necrosis factor-a inhibitor monotherapy and in combination with 9. Marcen R. Immunosuppressive drugs in kidney transplantation: impact thiopurines: analysis of the Food and Drug Administration Adverse on patient survival, and incidence of cardiovascular disease, malignancy Event Reporting System. J Gastrointestin Liver Dis 2014;23:267–71. and infection. Drugs 2009;69:2227–43. 24. Osterman MT, Sandborn WJ, Colombel JF, Robinson AM, et al. 10. Jiyad Z, Olsen CM, Burke MT, Isbel NM, et al. Azathioprine and risk Increased risk of malignancy with adalimumab combination therapy, of skin cancer in organ transplant recipients: systematic review and compared with monotherapy, for Crohn’s disease. Gastroenterology meta-analysis. Am J Transpl 2016;16:3490–503. 2014;146:941–9.

11. Armstrong RG, West J, Card TR. Risk of cancer in inflammatory bowel 25. Kopylov U, Vutcovici M, Kezouh A, Seidman E, et al. Risk of disease treated with azathioprine: a UK population-based case-control lymphoma, colorectal and skin cancer in patients with IBD treated with study. Am J Gastroenterol 2010;105:1604–9. immunomodulators and biologics: a Quebec claims database study. Inflamm Bowel Dis 2015;21:1847–53. 12. Long MD, Herfarth HH, Pipkin CA, Porter CQ, et al. Increased risk for non-melanoma skin cancer in patients with inflammatory bowel 26. Scott FI, Mamtani R, Brensinger CM, Haynes K, et al. Risk disease. Clin Gastroenterol Hepatol 2010;8:268–74. of nonmelanoma skin cancer associated with the use of immunosuppressant and biologic agents in patients with a history of 13. Setshedi M, Epstein D, Winter TA, Myer L, et al. Use of thiopurines in autoimmune disease and nonmelanoma skin cancer. JAMA Dermatol the treatment of inflammatory bowel disease is associated with an 2016;152:164–72. increased risk of non-melanoma skin cancer in an at-risk population: a cohort study. J Gastroenterol Hepatol 2012;27:385–9. 27. van den Heuvel TR, Wintjens DS, Jeuring SF, Wassink MH, et al. Inflammatory bowel disease, cancer and medication: cancer risk in 14. Singh H, Nugent Z, Demers AA, Bernstein CN. Increased risk of the Dutch population-based ISDSL cohort. Int J Cancer 2016;139: fl nonmelanoma skin cancers among individuals with in ammatory 1270–80. bowel disease. Gastroenterology 2011;141:1612–20. 28. Clowry J, Sheridan J, Healy R, Deady S, et al. Increased non-melanoma 15. Peyrin-Biroulet L, Khosrotehrani K, Carrat F, Bouvier AM, et al. skin cancer risk in young patients with inflammatory bowel disease on Increased risk for nonmelanoma skin cancers in patients who receive immunomodulatory therapy: a retrospective single-centre cohort study. fl thiopurines for in ammatory bowel disease. Gastroenterology 2011; J Eur Acad Dermatol Venereol 2017;31:978–985. 141:1621–8.e1–5. 29. Moher D, Liberati A, Tetzlaff J, Altman DG. The PRISMA Group. 16. van Schaik FD, van Oijen MG, Smeets HM, van der Heijden GJ, et al. Preferred reporting items for systematic reviews and meta-analyses: the fl Risk of nonmelanoma skin cancer in patients with in ammatory bowel PRISMA statement. PLoS Med 2009;6:e1000097. disease who use thiopurines is not increased. Clin Gastroenterol Hepatol 2011;9:449–50.e1. 30. Euvrard S, Kanitakis J, Claudy A. Skin cancers after organ transplantation. N Engl J Med 2003;348:1681–91. 17. Camus M, Seksik P, Bourrier A, Nion-Larmurier I, et al. Long-term outcome of patients with Crohn’s disease who respond to azathioprine. 31. Pedersen N, Duricova D, Elkjaer M, Gamborg M, et al. Risk of extra- fl Clin Gastroenterol Hepatol 2013;11:389–94. intestinal cancer in in ammatory bowel disease: meta-analysis of population-based cohort studies. Am J Gastroenterol 2010;105: 18. Long MD, Martin CF, Pipkin CA, Herfarth HH, et al. Risk of 1480–7. melanoma and nonmelanoma skin cancer among patients with ’ inflammatory bowel disease. Gastroenterology 2012;143:390–9.e1. 32. Harwood CA, Attard NR, O Donovan P, Chambers P, et al. PTCH mutations in basal cell carcinomas from azathioprine-treated organ 19. Gómez-Garc´ıa M, Cabello-Tapia MJ, Sánchez-Capilla AD, De transplant recipients. Br J Cancer 2008;99:1276–84. Teresa-Galván J, et al. Thiopurines related malignancies in inflammatory bowel disease: local experience in Granada, Spain. World – J Gastroenterol 2013;19:4877 86. Address correspondence and reprint requests to: Joshua W. 20. Ariyaratnam J, Subramanian V. Association between thiopurine use Hagen, MD, PhD, Department of Dermatology, University and nonmelanoma skin cancers in patients with inflammatory bowel of Pittsburgh Medical Center, 11 Vassar Avenue, disease: a meta-analysis. Am J Gastroenterol 2014;109:163–9. Pittsburgh, PA 15229, or e-mail: [email protected]

480 DERMATOLOGIC SURGERY