Malignancy Risk in Patients withInflammatory Eye Disease Treated withSystemic Immunosuppressive TherapyA Tertiary Referral Cohort Study

William B. Yates,1,2 Claire M. Vajdic, PhD,3 Renhua Na, MD,3 Peter J. McCluskey, MD,1,2,4

Denis Wakefield, MD, PhD1,2

Objective: To ascertain whether patients on long-term systemic immunosuppressive therapy for inflamma-tory eye disease (IED) are at increased risk of malignancy.

Design: A single-center, retrospective cohort study.Participants: We included 190 adults with IED treated with corticosteroids only (n ¼ 58) or systemic

immunosuppression (n ¼ 132) for �6 months between 1985 and 2007. Immunosuppressed patients were treatedwith antimetabolites, T-cell inhibitors, and/or alkylating agents.

Methods: Incident malignancies were ascertained by self-report and confirmed by medical record review.Multiple malignancies in a single patient were counted, except for nonmelanoma skin cancer (NMSC), where onlythe first was counted. Standardized incidence ratios (SIRs) were calculated by malignancy type. Cox regressionmodels were used to compare malignancy incidence by treatment type.

Main Outcome Measures: Risk of malignancy relative to the general population and within the cohort.Results: During a median 7.34 years of follow-up, 25 malignancies were observed in 17 patients, namely, 2.10

per 100 person-years and 0.43 per 100 person-years in the immunosuppressed and corticosteroid only groups,respectively. In the immunosuppressed group, the most common malignancies were NMSC (n ¼ 11) and non-Hodgkin’s lymphoma (NHL; n ¼ 4) and malignancy risk was significantly increased compared with the generalpopulation for any malignancy (SIR, 4.39; 95% CI, 2.78e6.59) and for any malignancy excluding NMSC (SIR, 4.16;95% CI, 1.67e8.57). Significantly elevated SIRs were observed for NMSC and NHL in those treated with immu-nosuppressive agents. Compared with the corticosteroid treatmenteonly group, the immunosuppressed groupwas at an increased risk of any malignancy (adjusted hazard ratio, 4.36; 95% CI, 1.02e18.7), but not first malig-nancy (n ¼ 17; adjusted hazard ratio, 2.56; 95% CI, 0.57e11.5). No cancer-related deaths were observed.

Conclusions: Our findings suggest that patients with IED treated with systemic immunosuppressive therapyare at increased risk of malignancy; however, the increase in absolute risk was modest. The types of malignanciesobserved at excess risk are similar to those observed in solid organ transplant recipients and patients withautoimmune diseases treated with systemic immunosuppression. Immunosuppressive therapy remains animportant treatment modality in IED; however, patients may benefit from targeted malignancy-prevention stra-tegies and long-term clinical follow-up. These findings require validation by a prospective, long-term, population-based cohort study. Ophthalmology 2014;-:1e9 ª 2014 by the American Academy of Ophthalmology.

Inflammatory eye disease (IED), including uveitis, scleritis,and mucous membrane pemphigoid, is responsible for 10%of blindness in the United States.1 Therapy for IED aimsto control the disease, maintain vision, and ensure qualityof life. The long-term use of systemic corticosteroids hasproved a double-edged sword, effective at reducing theinflammation, but associated with significant ocular andsystemic side effects.2 Immunosuppressive drugs are used ascorticosteroid-sparing agents and when the ocular inflam-mation is severe, refractory, or a manifestation of a specificdisease, such as Behçet’s disease.3

� 2014 by the American Academy of OphthalmologyPublished by Elsevier Inc.

The observation that systemic immunosuppression maypredispose to malignancy has been a major concern sincethe early findings of increased malignancy risk in immu-nosuppressed solid organ transplant recipients.4e6 It isestimated that 10% to 15% of patients who receive a kidneyallograft will develop a malignancy within 10 years ofinitiating immunosuppressive therapy.6e8 The overall riskof developing a malignancy in the solid organ transplantpopulation is 3- to 4-fold greater than the general popula-tion, with nonmelanocytic skin cancer (NMSC) and non-Hodgkin’s lymphoma (NHL) being the most commonly

1http://dx.doi.org/10.1016/j.ophtha.2014.08.024ISSN 0161-6420/14

Ophthalmology Volume -, Number -, Month 2014

observed malignancies.9,10 Particularly relevant in Australia,NMSC is associated with high levels of ambient solar ra-diation and correspondingly high rates of NMSC.11 Inaddition, NHL is the most common malignancy in patientswith rheumatoid disease who require long-term immuno-suppressive therapy.12 Similarly, the risk of NHL develop-ment in patients with systemic lupus erythematous wasestimated to be 7-fold greater than the general population.13

In contrast with solid organ transplant recipients and pa-tients treated for rheumatoid arthritis, patients with IED willnot necessarily be on lifelong immunosuppression ther-apy.3,14 Furthermore, although the typical dosages ofimmunosuppressive agents are one half to one third lower inpatients with IED compared with transplant recipients,depending on the therapeutic agent, they are similar to thosewith autoimmune disease.3

The potential mechanisms of carcinogenesis owingto immunosuppression have been discussed previously.Immunosuppressive agents may increase susceptibility toinfection by oncogenic viruses, impair immune surveillance,and have a direct deleterious effect on DNA (e.g., alkylatingagents).15

Malignancy risk in patients with IED treated with sys-temic immunosuppressive therapy has been assessed;however, outcomes varied significantly between stud-ies.16e18 A retrospective cohort study based in the UnitedStates of 2340 patients with IED treated with immuno-suppressive therapy found no excess risk of death or deathfrom malignancy.16 However, this study was unable toquantify the risk of nonfatal malignancies and relied on theattribution of underlying cause of death to assess malig-nancy risk.16 The single prior study to ascertain incidentmalignancies reported no difference in incidence betweenIED patients prescribed corticosteroids (n ¼ 207) and thoseon �1 immunosuppressive agents (n ¼ 330).17 This studyfollowed patients for a median of 1.34 years, which may beinsufficient, with studies suggesting �5 years of follow-upis necessary.19e21 It is possible that these studies did notadequately address the risk of neoplasia and we hypothe-sized that patients with IED who receive long-term sys-temic immunosuppressive therapy have an increased riskof malignancy. To address this risk, we performed aretrospective cohort study examining the incidence of denovo invasive malignancy in patients with IED managedwith and without systemic immunosuppression at a singletertiary hospital. We computed malignancy risk relative tothe general population (adjusted for age and sex), andmalignancy risk in those treated with immunosuppressiveagents compared with those treated with corticosteroidsalone.

Methods

Study Design and Population

We performed a single-center, retrospective, cohort study at StVincent’s Clinic, Sydney, New South Wales, Australia. Patientswith IED diagnosed and treated with systemic corticosteroids and/or immunosuppressive agents at the clinic between 1985 and 2007were identified. All patients were assessed by an ophthalmologist

2

and physician at presentation and during follow-up. Patients >15years of age at diagnosis and treated for �6 months were eligiblefor this study. Patients reviewed only for a second opinion andthose treated by a specialist outside the Clinic were ineligible.Patients with rheumatoid arthritis, Sjögren’s syndrome, psoriasis,systemic lupus erythematosus, or hepatitis B/C infection wereexcluded, because these diseases are associated with an increasedrisk of malignancy. Patients with a history of malignancy beforetherapy (n ¼ 4) were not excluded, but they did not contributeperson-years at risk for that malignancy type.

Patients consented to participate in the study and to the retrievalof additional medical records if required. The study was conductedin accordance with the tenets of the Declaration of Helsinki. Hu-man Research Ethics Committee approval was obtained from StVincent’s Hospital.

Data Collection

Medical records were reviewed and information regarding IEDdiagnosis, type, and duration of treatments were obtained. Thedate of diagnosis, nature, and anatomic location of the IED andany underlying autoimmune or syndrome features were recorded.The date that corticosteroid and immunosuppressive drugs werestarted and stopped, the type of drug, and their dosage werecollected.

A questionnaire was given to all patients at the clinic or mailedto their home address. For patients whose eyesight was poor, theycould opt to answer the questions by telephone. Telephone contactwas attempted if the patient did not return the questionnaire. Anaverage of 5 phone calls was made and if there was no forwardingnumber the referring doctor was contacted to ensure current contactinformation. Patients were classified as “lost to follow-up” if theycould not be contacted to complete the questionnaire. Informationobtained from the questionnaire included personal history of ma-lignancy, history of malignancy in any first-degree relatives,smoking history, and skin complexion.

Self-reported malignancy diagnoses were confirmed by medicalrecord review and clinical examination. All relevant histopathologyand immunohistochemistry reports were obtained from the treatingdoctor. The date of diagnosis, site, and type of each malignancywere recorded. Solid (nonhematopoietic) malignancies were clas-sified according to the International Classification of Disease(ICD), 10th revision, and hematopoietic malignancies were clas-sified according to the ICD for Oncology, 3rd edition. Whereapplicable, the date of death was recorded.

Malignancy incidence rates for the New South Wales popula-tion were obtained from the New South Wales Central CancerRegistry by 5-year age group, calendar year, and sex for 1985 to2007; incidence rates for 2007 were applied to the 2008 through2012 follow-up years. The Cancer Registry is population based andall cases of primary invasive malignancies, except NMSC, must bereported by statute. Because NMSCs are not notifiable malig-nancies in New South Wales, a large NMSC incidence survey (57215 participants) conducted in 2002 was used to estimate theincidence of cutaneous basal cell carcinoma (BCC) and squamouscell carcinoma (SCC) in the general population by sex and agegroup (<40, 40e49, 50e59, 60e69, �70 years).11 These rateswere applied to all calendar years of follow-up.

Statistical Analyses

Treatment regimen was classified as corticosteroids only or corti-costeroids and immunosuppression; IED was classified as idio-pathic or associated with an underlying systemic disease or IEDsyndrome.

Yates et al � Malignancy Risk with Systemic Immunosuppression

Standardized Incidence Ratios

Person-years of follow-up accrued from the start of corticosteroidtherapy for the corticosteroid group or from the start of immuno-suppressive therapy for immunosuppressed group until the date ofmalignancy, death, or the date of last follow-up (December 31,2012), whichever occurred first. Malignancy incidence rates in theIED cohort were compared with the New South Wales populationusing the standardized incidence ratio (SIR), defined as the ratio ofthe observed and the expected numbers of malignancies. The ex-pected numbers of incident malignancies were calculated bymultiplying the cohort person-years at risk by the corresponding 5-year age, sex, and calendar year-specific incidence rates of ma-lignancies for the New South Wales population. The exact methodwas used to calculate 95% CI.22 For any malignancy, any solidmalignancy, and specific malignancies, we calculated SIRs for theentire cohort and the immunosuppressed subgroup. Only the firstSCC or BCC in a single patient was counted in these analyses; allother higher-order primary neoplasms in the same individual werecounted. A sensitivity analysis was performed by excluding SCCsand BCCs from the risk estimate for any malignancy and any solidmalignancy. In addition, a sensitivity analysis was performed byincluding all patients lost to follow-up and assuming they did notdevelop a malignancy.

Absolute Risk of Malignancy. The absolute risk of malignancywas calculated based on the number of observed malignanciesdivided by the person-time at risk, that is, the crude rates observedin each patient group. We derived 95% CIs using an exact Poisson-based method (Byar).23

Relative Risk of Malignancy within the Cohort. Within thecohort of all IED patients, the risk of any malignancy was esti-mated for those treated with immunosuppression relative tothose receiving corticosteroids alone, with adjustment for potentialconfounders. A marginal Cox regression model, the Wei-Lin-Weissfeld model, was applied to enable inclusion of multiplemalignancies in the same individual and the computation of hazardratios with 95% CIs.24 Covariates considered for inclusion in thefinal multivariable model were age at start of therapy (continuous),sex (male, female), systemic diseases other than those excluded(yes, no), family history of malignancy (yes, no), personal historyof smoking (ever, never), sun exposure per day (�2, >2 hours),and skin type (fair, light olive/olive, dark olive/dark). A covariatewas included in the final model if the univariable P was �0.25. Allvariables included in the final model satisfied the proportionalhazards assumption.

Analyses were performed using SAS software v9.2 (SAS Inc,Cary, NC). The% stratify macro was used to compute person-yearsand SIRs23 and the Wei-Lin-Weissfeld marginal model wascomputed using the PHREG program.

Results

A search of all available medical records identified 458 potentiallyeligible patients. A total of 202 patients met the study inclusioncriteria (Fig 1). We excluded 256 patients owing to systemic dis-ease associated with an increased the risk of malignancy, treatmentinitiated at another clinic, or <6 months of continuous therapy.Three patients declined to participate and 9 were lost to follow-up.Of the remaining 190 patients, 132 (69%) received �1 immuno-suppressive agents and 58 (31%) received corticosteroids only(Table 1). There were 128 females (67%) and 62 males, and themean age at the start of therapy was 44 years (interquartile range[IQR], 30e56)d40 years (IQR 29e54) for the corticosteroid-onlygroup, and 46 years (IQR, 31e57) for the immunosuppressedgroup.

The majority of patients had idiopathic ocular inflammationonly, 44 (75%) in the corticosteroid group and 88 (67%) in thegroup with immunosuppressive therapy. A similar proportion ofpatients in the corticosteroid group and the immunosuppressedgroup had systemic disease (Table 1; P ¼ 0.48). The most commonsystemic diseases in patients treated with immunosuppression wereBehçet’s syndrome (n ¼ 13); seronegative spondyloarthropathy(n ¼ 11), including ankylosing spondylitis (n ¼ 6), reactivearthritis (n ¼ 3), and undifferentiated spondyloarthritis (n ¼ 2); andsarcoidosis (n ¼ 5). In the corticosteroid-only group, a range ofsystemic diseases was observed including polychondritis (n ¼ 3),Susac syndrome (n ¼ 3), sarcoidosis (n ¼ 2), multiple sclerosis(n ¼ 2), Cogan syndrome (n ¼ 2), and polyarteritis nodosa (n ¼ 1).

The IED cohort was followed up for a total of 1562 person-yearsd469 person-years for the corticosteroid only group and1093 person-years for the immunosuppressed group. The medianduration of follow-up was 7.34 yearsd7.27 (IQR, 4.75e13.9) and7.34 (IQR, 4.33e11.0) for the corticosteroid and immunosup-pressed groups, respectively.

The median duration of exposure to immunosuppressive agentswas 4.0 years (IQR, 2.3e7.2). Sixty-five patients (55%) weretreated with 1 immunosuppressant only, 33 (28%) with 2 agents,16 (13%) with 3, and 5 (4%) with �4. A total of 31 patients (26%)received �2 immunosuppressive agents concurrently. At any timeduring follow-up, methotrexate was the most commonly usedimmunosuppressant (n ¼ 69), followed by mycophenolate (MMF;n ¼ 33), cyclosporine (CSA; n ¼ 31), azathioprine (AZA; n ¼ 24),5-aminosalicylic acid (n ¼ 17), oral cyclophosphamide (CYP; n ¼15), intravenous CYP (n ¼ 1), thalidomide (n ¼ 1), etanercept (n ¼1), infliximab (n ¼ 1), and adalimumab (n ¼ 1). Neither the typenor dosage of immunosuppression was included in the multivari-able models owing to the small number of incident malignancies.

During follow-up, 2 patients (1%) died (both from cardiovas-cular disease) and 25 incident primary malignancies were observedin 17 patients (9%). Two malignancies occurred in the corticoste-roid only group and 23 in the immunosuppressed group. Of the 17patients with �1 malignancy, the median age at first diagnosis was55 years (IQR, 50e72). The most frequent malignancies wereNMSC (n ¼ 12; BCC, n ¼ 7; SCC, n ¼ 5), NHL (n ¼ 4), ma-lignancy of the vulva (n ¼ 2), and lip malignancy (n ¼ 2; Table 2).A breast malignancy and a BCC were observed in the corticoste-roid only group (Table 2). The median time between the initiationof treatment and the first malignancy was 7.3 years (IQR. 4.7e9.9)for the corticosteroid group and 6.2 years (IQR, 3.4e9.6) for pa-tients on immunosuppressive therapy. The NHLs we observed didnot resolve on discontinuation of immunosuppressive therapy andrequired treatment.

Risk of Malignancy Relative to General Population

Risk of any malignancy was significantly greater in the entire IEDcohort than the general population (n ¼ 25; SIR, 2.74; 95% CI,1.77e4.04; Table 3). That risk remained elevated when NMSCswere excluded (n ¼ 13; SIR, 2.57; 95% CI, 1.37e4.40). The riskwas also increased when only the first malignancy in patients whodeveloped >1 malignancy was counted (n ¼ 17; SIR, 1.90; 95%CI, 1.11e3.04).

Relative to the general population, the risk of any malignancyfor the immunosuppressed subgroup was >4-fold greater (n ¼ 23;SIR 4.39; 95% CI, 2.78e6.59) and more than 5-fold when NMSCswere excluded (n ¼ 12; SIR, 5.28; 95% CI, 2.73e9.22; Table 4). Asignificant excess risk was observed for 6 malignancy types,namely, BCC (n ¼ 6; SIR, 2.96; 95% CI, 1.09e6.44), SCC (n ¼ 5;SIR, 5.36; 95% CI, 1.74e12.5), NHL (n ¼ 4; SIR, 9.52; 95% CI,2.59e24.4), vulva (n ¼ 2; SIR, 79.8; 95% CI, 9.67e288), lip (n ¼2; SIR, 45.5; 95% CI, 5.51e164), and renal pelvis (n ¼ 1; SIR,

3

Figure 1. The study cohort. IED ¼ inflammatory eye disease; C-S Rx ¼ corticosteroid therapy; IS Rx ¼ immunosuppressive therapy; SLE ¼ systemic lupuserythematosus; RA ¼ rheumatoid arthritis; SS ¼ Sjögren’s syndrome.

Ophthalmology Volume -, Number -, Month 2014

43.4; 95% CI, 1.10e242; Table 5). Inclusion of the 9 patients lostto follow-up resulted in a 3% reduction in SIR but no loss ofstatistical significance (data not shown).

Absolute Risk of Malignancy

The absolute risk of any malignancy was 1.60 per 100 person-years(95% CI, 1.04e2.36) for all IED patients, 0.43 per 100 person-years (95% CI, 0.05e1.54) for the corticosteroid group, and 2.10per 100 person-years (95% CI, 1.33e3.16) for the immunosup-pressed group, corresponding with an excess of 1.67 malignanciesper 100 person-years (95% CI, 0.63e2.72) in those with immu-nosuppressive therapy. Excluding NMSCs, the correspondingvalues are 0.83 per 100 person-years (95% CI, 0.44e1.42), 0.21per 100 person-years (95% CI, 0.005e1.19), and 1.10 per 100person-years (95% CI, 0.57e1.92), and an excess of 0.89 per 100person-years (95% CI, 0.14e1.63) malignancies for the immuno-suppressed group.

Risk of Malignancy within the Cohort

In univariable models, the risk of any malignancy was significantlygreater for patients treated with immunosuppressive drugscompared with those receiving corticosteroids alone, but only

4

when any malignancy was counted (hazard ratio, 5.55; 95% CI,1.25e24.6; Table 4). Malignancy risk was significantly and posi-tively associated with older age and self-reported family history ofmalignancy for both first and multiple malignancies (Table 6).There was no association between malignancy risk and type ofIED, the presence of systemic disease, a history of smoking, highsun exposure, or fairer skin type.

In multivariable models, there was no difference in risk of firstmalignancy between the 2 treatment groups (adjusted hazard ratio,2.56; 95% CI, 0.57e11.5), but the immunosuppressed patientgroup had an increased risk when multiple malignancies in a singlepatient were counted in the analysis compared with the cortico-steroid only group (adjusted hazard ratio, 4.36; 95% CI,1.02e18.7; Table 7).

Discussion

Over a long period of follow-up, we found a significantlygreater risk of malignancy in a small cohort of patients withmainly idiopathic IED treated with systemic immunosup-pressive agents for �6 months. The risk of any malignancywas increased 2-fold relative to the general population and

Table 1. Characteristics of Clinical Cohort of Patients with In-flammatory Eye Disease by Therapeutic Approach (1985e2007)

CharacteristicCorticosteroids

Only (%)Immunosuppressive

Agents (%)P

Value

Total 58 132Age, y (mean � SD) 40.6�16.2 44.8�16.6 0.11Sex 0.86Male 18 (31) 44 (33.3)Female 40 (69) 88 (66.7)

Type of ocularinflammation

0.35

Anterior uveitis 22 40Intermediate uveitis 7 7Posterior or panuveitis 26 58Scleritis 3 18Ocular pemphigoid 0 9

Systemic disease 0.23Absent 44 (76) 88 (61)Present 14 (24) 44 (39)Behçet’s syndrome 1 13Sarcoidosis 5 8Seronegative

spondyloarthropthy4 15

Granulomatouspolyangitis

3 1

Other 1 7Family history of

malignancy0.50

Yes 27 (46.5) 63 (47.7)No 31 (53.5) 69 (52.3)

History of smoking 0.26Never 40 (69) 81 (61.4)Former 8 (13.8) 18 (13.6)Current 10 (17.2) 33 (25)

Skin type 0.71Fair, light olive, or olive 45 (78) 94 (71)Dark olive or dark 7 (12) 20 (15)Missing 6 (10) 18 (13)

SD ¼ standard deviation.

Yates et al � Malignancy Risk with Systemic Immunosuppression

4-fold compared with patients with IED treated with corti-costeroids alone when multiple malignancies in the samepatients were counted. The risk for first malignancies onlywas also significantly elevated compared with the general

Table 2. Types of Incident Malignancies Observed in a TertiaryCohort of Patients with Inflammatory Eye Disease

Malignancy TypeCorticosteroids

OnlyImmunosuppressive

Agents

Nonmelanocytic skin cancer 1 11Basal cell carcinoma 1 6Squamous cell carcinoma 0 5

Non-Hodgkin’s lymphoma 0 4Vulva 0 2Lip 0 2Breast 1 0Prostate 0 1Renal pelvis 0 1Melanoma 0 1Chronic myeloid leukemia 0 1Total 2 23

population. The risk estimates relative to the general pop-ulation remained substantial even if the lower limits ofthe CI were considered. The observed malignancy typesin this study were similar to those observed in solidorgan transplant recipients and patients with autoimmunedisease treated with systemic immunosuppressive agents,including NMSC, NHL, and malignancy of the lip andvulva.9e13,25e27 No malignancy-related deaths wereobserved and the excess absolute risk of malignancy inIED patients treated with immunosuppression was small,with <1 case for every 100 patients (excluding NMSC)followed up for a single year.

Previous studies have shown that patients with IEDtreated with immunosuppressive agents are not at increasedrisk of fatal malignancies. Kempen et al16 conducted thelargest study, examining the health records of 2340immunosuppressed patients with ocular inflammation over17 006 person-years and found no increased risk of overallor malignancy-related mortality compared with the generalUS population (standardized cancer mortality ratio, 1.03;95% CI, 0.80e1.31). There were no malignancy-associateddeaths observed in our cohort study; therefore, they wouldnot be evident in a mortality-based study. Lane et al17

conducted a retrospective cohort study in the United Statesof 543 patients with IED (n ¼ 330 immunosuppressed, n ¼207 corticosteroids only) and reported no increase in ma-lignancy between the 2 patient groups (P>0.90). Patientswere treated with immunosuppressive therapy for 1 to 2years, with AZA and CSA used for the longest duration(median, 1.6 years and 1.3 years, respectively). Further-more, this study had a median follow-up of only 1.34 years(range, 1 day to 18 years), which together may explain thediscordance with our findings.17 In our cohort, de novomalignancies typically developed around 6 years afterinitiation of therapy. Only 1 prior study found an increasedrisk of malignancy in IED patients. Murray et al18

observed 9 malignancies in a case series of 46 patientstreated with alkylating therapy, compared with 46 age- andsex-matched patients on corticosteroids only (P<0.05).However, alkylating agents such as CYP and chlorambucilare not commonly used as first-line therapy in IED.3,14

In transplant recipients, NMSC is the most commonlyobserved malignancy, with the risk estimated to be 28times greater than that of the general population.9 In ourcohort, the increased risk of BCC and SCC in the immu-nosuppressed group was 3- and 5-fold greater, respec-tively, compared with the general Australian population.This finding may be owing to a combination of immuno-suppression and the high ambient ultraviolet radiationlevels in Australia.11 For different immunosuppressedpopulations, SIRs cannot be compared statistically becauseof heterogeneity in the cohort age and sex distributionsand thus population incidence rates, but our estimatessuggest that immunosuppressed patients with IED maybe at a lesser risk of NMSC compared with transplantrecipients.

The majority of patients in this study (69%) had organ-specific ocular inflammation without associated systemicinflammatory disease. There is little, if any, current evidenceto suggest an increased risk of malignancy in association

5

Table 3. Standardized Incidence Ratios for All Patients with Inflammatory Eye Disease, Regardless of Treatment Type

Malignancy GroupNumber of

Observed CasesNumber of

Expected CasesStandardized Incidence

Ratio (95% CI)

Any cancer 25 9.14 2.74 (1.77e4.04)Any cancer excluding squamous and basal cell carcinoma 13 5.05 2.57 (1.37e4.40)Any solid cancer including squamous and basal cell carcinoma 20 8.33 2.40 (1.47e3.71)Any solid cancer excluding squamous and basal cell carcinoma 8 4.25 1.88 (0.81e3.71)Non-Hodgkin’s lymphoma 4 0.58 6.93 (1.89e17.7)

Ophthalmology Volume -, Number -, Month 2014

with these conditions.28e30 Furthermore, we found no dif-ference in the risk of malignancy in those with or without asystemic inflammatory disease. This is in contrast with pa-tients treated for rheumatoid arthritis and juvenile idiopathicarthritis, where the data suggest that disease-related,persistent immunologic stimulation accounts for the excessrisk of lymphoma.31,32 Whether an underlying systemicinflammatory disease such as sarcoidosis, Behçet’s syn-drome, or seronegative spondyloarthropathy confounds theincreased malignancy risk in patients with IED awaitselucidation in larger IED cohorts.

Our study has a number of strengths. The key strength isthe relatively long duration of follow-up, a median of 7years. The IED cohort was also relatively homogenous, themajority having idiopathic ocular inflammation. The num-ber of patients lost to follow-up was small (n ¼ 9) and asensitivity analysis including these patients did not mean-ingfully alter the SIRs. In addition, the same ophthalmolo-gist and physician managed all patients during the follow-upperiod.

The major limitation of our study is the small samplesize, resulting in imprecise risk estimates. A large, single-center cohort is difficult to obtain owing to the rarity ofIED and the proportion of patients with severe diseaserequiring systemic immunosuppressive therapy. The smallnumber of incident malignancies was a reflection of therelatively small cohort; nevertheless, it was sufficient toelucidate an excess risk relative to the general population.Patients were eligible for inclusion if treated with immu-nosuppressive therapy for �6 months; however, no statis-tical analysis could be conducted to correlate malignancyrisk with duration of therapy owing to the small samplesize. Similarly, the small number of observed malignanciesdid not allow an assessment of malignancy risk in relationto use of specific immunosuppressive agents. The abilityto study patients on long-term corticosteroid regimensalone was also limited by recommended clinical practice

Table 4. Standardized Incidence Ratios for the Imm

Malignancy GroupN

Obs

Any cancerAny cancer excluding squamous and basal cell carcinomaAny solid cancerAny solid cancer excluding squamous and basal cell carcinomaNon-Hodgkin’s lymphoma

6

guidelines; that is, immunosuppressive therapy iscommenced if corticosteroid dosages of <10 mg/d are noteffective in obtaining disease control. Observation bias ispossible in a medically supervised cohort, especially whencomparisons are made with cancer registryebased generalpopulation rates and NMSCs are included as an outcome.However, follow-up did not vary by treatment type. InNew South Wales, NMSCs are not mandatorily reported;consequently, a large, independent survey was used to es-timate SCC and BCC incidence.11 The available NMSCincidence data were for Australia as a whole, in 2002, andfor broad age groups, and as a result the estimates of SCCand BCC risk are not as reliable as those for notifiablemalignancies. Furthermore, ascertainment bias may exist forour SIR estimates because the incident cancers in the cohortand general population were not ascertained from identicalsources. Finally, no adjustments have been made for mul-tiple comparisons.

In our cohort, the most commonly utilized immuno-suppressive agents were methotrexate, MMF, CSA, AZA,and CYP. We were unable to examine whether certainagents conferred a higher malignancy risk owing to therelatively small sample size and the use of multiple agents,often concurrently, to control the ocular inflammation.However, expert panels established by the InternationalAgency for Research on Cancer have judged that long-termimmunosuppression is linked to an increased risk of can-cer.33 Furthermore, the International Agency for Researchon Cancer panels have determined that there is sufficientscientific evidence that CSA (a T-cell inhibitor), AZA (anantimetabolite), and CYP (an alkylating agent) are carci-nogenic to humans.33 The mechanism of action is impairedimmune surveillance owing to immunosuppression, butalso direct carcinogenic effects of the agents themselves,such as DNA double-strand breaks. In keeping with ourfindings, CSA and AZA are oncogenic for NHL andNMSC.33 Bladder malignancies and acute myeloid leukemia

unosuppressed Inflammatory Eye Disease Cohort

umber oferved Cases

Number ofExpected Cases

Standardized IncidenceRatio (95% CI)

23 5.23 4.39 (2.78e6.59)12 2.27 5.28 (2.73e9.22)18 4.64 3.88 (2.30e6.13)7 1.68 4.16 (1.67e8.57)4 0.43 9.38 (2.56e24.0)

Table 5. Standardized Incidence Ratios for the ImmunosuppressedInflammatory Eye Disease Cohort by Malignancy Type

Malignancy

Number ofObservedCases

Number ofExpectedCases

StandardizedIncidence

Ratio (95% CI)

Basal cell carcinoma 6 2.03 2.96 (1.09e6.44)Squamous cell carcinoma 5 0.93 5.36 (1.74e12.5)Non-Hodgkin’s lymphoma 4 0.42 9.52 (2.59e24.4)Vulva 2 0.03 79.8 (9.67e288)Lip 2 0.04 45.5 (5.51e164)Prostate 1 0.87 1.16 (0.03e6.44)Renal pelvis 1 0.02 43.4 (1.10e242)Melanoma 1 0.72 1.38 (0.03e7.69)Chronic myeloid leukemia 1 0.17 5.84 (0.15e32.5)

Yates et al � Malignancy Risk with Systemic Immunosuppression

are linked with use of CYP, but neither was observed in ourcohort. Currently, methotrexate and MMF are not consid-ered carcinogenic to humans.

The increased risk of malignancy observed in ourcohort study has important management implications forpatients with IED requiring long-term immunosuppressivetherapy. Systemic immunosuppressive therapy plays apivotal role in those with refractory IED; however, werecommend the minimum effective dose and duration beutilized and local and intraocular therapy be utilized wherepossible. Ultimately, the choice of treatment modality ismade on a case-by-case basis with the goal of IEDremission while minimizing ocular and systemic side

Table 6. Univariable Determinants of Malignancy Risk for Patients wiImmunosuppress

Characteristic

First Malignancy Only

n Hazard Ratio (95% CI)

TreatmentCorticosteroid only 2 1.00Immunosuppressant 15 3.61 (0.82e16.0)

Age (1-year increase) 17 1.06 (1.03e1.10)SexMale 7 1.00Female 10 0.58 (0.22e1.57)

Inflammatory eye disease typeAnterior/intermediate 7 1.00Posterior/panuveitis 8 0.95 (0.34e2.67)Other 2 0.66 (0.13e3.27)

Systemic diseaseNo 9 1.00Yes 8 1.84 (0.69e4.94)

Family history of malignancyNo 3 1.00Yes 14 4.27 (1.22e14.9)

SmokingNever 10 1.00Ever 7 1.57 (0.58e4.26)

Skin typeFair, light olive, or olive 15 1.00Dark olive or dark 2 0.65 (0.14e2.96)

CI ¼ confidence interval.

effects. Physicians in the fields of transplant and autoim-mune disease who utilize immunosuppressive therapyeducate patients regarding cancer prevention, and performregular and targeted long-term follow-up of patients tofacilitate early detection of malignancy. We believe thismay also be a suitable framework in the setting ofimmunosuppressive therapy for IED. Prevention strategiesto decrease the risk of malignancy would include immu-nization against oncogenic viruses (e.g., human papillo-mavirus, hepatitis), avoidance of excessive ultravioletradiation exposure, prompt treatment of premalignantcutaneous lesions (e.g., Bowen’s disease), regular skinchecks, cancer screening in accord with general population(e.g., cervix), and periodic careful physical examination(e.g., lymph nodes). Liaison with other specialists such asphysicians and dermatologists, when required, facilitatesthe management of any potential malignancies and opti-mize patient outcomes.21

This study must be replicated in a large, long-term, pro-spective cohort to determine the relative risk of differenttreatment regimens, and the specific type and severity ofIED. Such a study would have the statistical power to stratifyby systemic involvement and address the issue of whetherthe type of immunosuppressive agent, and the duration andintensity of immunosuppression, increases the risk of ma-lignancy in patients with IED. To avoid ascertainment bias,such a study should identify incident malignancies usingrecord linkage to national registries of notifiable malig-nancies. In the meantime, our data and a synthesis of

th Inflammatory Eye Disease Treated with Corticosteroids Only orive Therapy

Any Malignancy

P Value n Hazard Ratio (95% CI) P Value

2 1.000.09 23 5.55 (1.25e24.6) 0.02

<0.001 25 1.06 (1.03e1.09) <0.001

11 1.000.29 14 0.40 (0.15e1.07) 0.07

0.88 12 1.00 0.5210 0.65 (0.24e1.77)3 0.59 (0.11e3.30)

13 1.000.23 12 1.75 (0.62e4.92) 0.29

3 1.000.02 22 6.93 (2.04e23.5) 0.002

16 1.000.37 9 1.13 (0.41e3.14) 0.82

22 1.000.58 3 0.71 (0.22e2.38) 0.58

7

Table 7. Multivariable Determinants of Malignancy Risk for Patients with Inflammatory Eye Disease Treated with Corticosteroids Only orImmunosuppressive Therapy

Characteristic

First Malignancy Only Any Malignancy

n Adjusted Hazard Ratio* (95% CI) P Value n Adjusted Hazard Ratio* (95% CI) P Value

TreatmentCorticosteroid only 2 1.00 2 1.00Immunosuppressant 15 2.56 (0.57e11.5) 0.22 23 4.36 (1.02e18.7) 0.047

Age (1-year increase) 17 1.07 (1.03e1.11) 0.001 25 1.07 (1.04e1.10) <0.001Sexy

Male 11 1.00Female 14 0.48 (0.20e1.14) 0.10

Family history ofmalignancy

No 3 1.00 3 1.00Yes 14 5.05 (1.41e18.0) 0.01 22 7.57 (2.13e26.9) 0.002

CI ¼ confidence interval.*Model adjusted for treatment type, age, sex (all malignancies only), and family history of malignancy.ySex did not fulfill the criteria for inclusion in the multivariable model for risk of first malignancy.

Ophthalmology Volume -, Number -, Month 2014

evidence on the carcinogenicity of immunosuppressiveagents in humans33 support the development of targetedmalignancy prevention strategies and long-term vigilance tofacilitate early diagnosis and favorable outcomes.

References

1. Nussenblatt RB. The natural history of uveitis. Int Ophthalmol1990;14:303–8.

2. Stanbury R, Graham E. Systemic corticosteroid therapydsideeffects and their management. Br J Ophthalmol 1998;82:704–8.

3. Jabs D, Rosenbaum J, Foster S, et al. Guidelines for the use ofimmunosuppressive drugs in patients with ocular inflammatorydisorders: recommendations of an expert panel. Am J Oph-thalmol 2000;130:492–513.

4. Penn I. Malignancies associated with immunosuppressive orcytotoxic therapy. Surgery 1978;83:492–502.

5. Kelly GE, Mahony JF, Sheil AG, et al. Risk factors for skincarcinogenesis in immunosuppressed kidney transplant re-cipients. Clin Transplant 1987;1:271–7.

6. Penn I. Post-transplant malignancy: the role of immunosup-pression. Drug Saf 2000;23:101–13.

7. Lutz J, Heemann U. Tumours after kidney transplantation.Curr Opin Urol 2003;13:105–9.

8. Kauffman M, Cherikh W, McBride M, et al. Post-transplant denovo malignancies in renal transplant recipients: the past andpresent. Transpl Int 2006;19:607–20.

9. Grulich AE, van Leeuwen MT, Falster MO, et al. Incidence ofcancers in people with HIV/AIDS compared with immuno-suppressed transplant recipients: a meta-analysis. Lancet2007;370:59–67.

10. Vajdic CM, McDonald SP, McCredie MR, et al. Cancerincidence before and after kidney transplantation. JAMA2006;296:2823–31.

11. Staples MP, Elwood M, Burton RC, et al. Non-melanoma skincancer in Australia: the 2002 national survey and trends since1985. Med J Aust 2006;184:6–10.

12. Kamel OW, Holly EA, van de Rijn M, et al. A populationbased, case control study of non-Hodgkin’s lymphoma in

8

patients with rheumatoid arthritis. J Rheumatol 1999;26:1676–80.

13. Zintzaras E, Voulgarelis M, Moutsopoulos H. The risk oflymphoma development in autoimmune diseases: a meta-analysis. Arch Intern Med 2005;165:2337–44.

14. Hemady R, Tauber J, Foster S. Immunosuppressive drugs inimmune and inflammatory ocular disease. Surv Ophthalmol1991;35:369–85.

15. Kempen JH, Gangaputra S, Daniel E, et al. Long-term risk ofmalignancy among patients treated with immunosuppressiveagents for ocular inflammation: a critical assessment of theevidence. Am J Ophthalmol 2008;146:802–12.

16. Kempen J, Ebenezer D, Dun J, et al. Overall and cancer relatedmortality among patients with ocular inflammation treatedwith immunosuppressive drugs: retrospective cohort study[report online]. BMJ 2009;339:b2480. Available at: http://www.bmj.com/archive/online/2009/06-29. Accessed August13, 2014.

17. Lane L, Tamesis R, Rodriguez A, et al. Systemic immuno-suppressive therapy and the occurrence of malignancy in pa-tients with ocular inflammatory disease. Ophthalmology1995;102:1530–5.

18. Murray NL, McCluskey PJ, Wakefield D. Increased risk ofmalignancy in patients with uveitis treated with systemicimmunosuppressive therapy. In: Nussenblatt RB,Whitcup SM, Caspi RR, Gery I, eds. Advances in OcularImmunology: Proceedings of the 6th International Symposiumon the Immunology and Immunopathology of the Eye.Amsterdam, Netherlands: Elsevier; 1994:421–4.

19. Ramsay HM, Fryer AA, Hawley CM, et al. Non-melanomaskin cancer risk in the Queensland renal transplant population.Br J Dermatol 2002;147:950–6.

20. van Leeuwen MT, Grulich AE, Webster AC, et al. Immuno-suppression and other risk factors for early and late non-Hodgkin lymphoma after kidney transplantation. Blood2009;114:630–7.

21. Yates WB, McCluskey PJ, Wakefield D. Are patients withinflammatory eye disease treated with systemic immunosup-pressive therapy at increased risk of malignancy? J OphthalmicInflamm Infect 2013;3:48–54.

22. Swift MB. Comparison of confidence intervals for a Poissonmean: further considerations. Commun Stat Theory Methods2009;38:748–59.

Yates et al � Malignancy Risk with Systemic Immunosuppression

23. Breslow NE, Day NE. Statistical Methods in Cancer Research.vol. II: The Design and Analysis of Cohort Studies. Oxford,UK: Oxford Univ. Press; International Agency for Researchon Cancer; 1987:III-82-92. IARC Sci Publ. no. 82.Available at: http://www.iarc.fr/en/publications/pdfs-online/stat/sp82/. Accessed August 13, 2014.

24. Wei LJ, Lin DY, Weissfeld LA. Regression analysis ofmultivariate incomplete failure time data by modeling mar-ginal distributions. J Am Stat Assoc 1989;84:1065–73.

25. Rostgaard K. Methods for stratification of person-time andeventsda prerequisite for Poisson regression and SIR esti-mation. Epidemiol Perspect Innov [serial online] 2008;5:7.Available at: http://archive.biomedcentral.com/1742-5573/5/7.Accessed August 13, 2014.

26. Barr BB, Benton EC, McLaren K, et al. Human papillomavirus infection and skin cancer in renal allograft recipients.Lancet 1989;1:124–9.

27. Bouvard V, Baan R, Straif K, et al. A review of human carcino-gensdpart B: biological agents. Lancet Oncol 2009;10:321–2.

28. Cengiz M, Altundag M, Zorlu A, et al. Malignancy in Behçet’sdisease: a report of 13 cases and a review of the literature. ClinRheumatol 2001;20:239–44.

29. Seersholm N, Vestbo J, Viskum K. Risk of malignant neoplasmsin patients with pulmonary sarcoidosis. Thorax 1997;52:892–4.

30. Feltelius N, Ekbom A, Blomqvist P. Cancer incidence amongpatients with ankylosing spondylitis in Sweden 1965-1995: apopulation based cohort study. Ann RheumDis 2003;62:1185–8.

31. Baecklund E, Iliadou A, Askling J, et al. Association ofchronic inflammation, not its treatment, with increased lym-phoma risk in rheumatoid arthritis. Arthritis Rheum 2006;54:692–701.

32. Beukelman T, Haynes K, Curtis J, et al. Rates of malignancyassociated with juvenile idiopathic arthritis and its treatment.Arthritis Rheum 2012;64:1263–71.

33. IARC Working Group on the Evaluation of CarcinogenicRisks to Humans. A Review of Human Carcinogens. Part A:Pharmaceuticals. vol 100A. Lyon, France: International Agencyfor Research on Cancer; 2012;319-230;337-9.

Footnotes and Financial Disclosures

Originally received: October 31, 2013.Final revision: July 30, 2014.Accepted: August 12, 2014.Available online: ---. Manuscript no. 2013-1829.1 Ocular inflammation Research Unit, School of Medical Sciences, Uni-versity of New South Wales, Sydney, Australia.2 Department of Ophthalmology, St Vincent’s Hospital, Darlinghurst,Sydney, Australia.3 Adult Cancer Program, Lowy Cancer Research Centre, Prince of WalesClinical School, University of New South Wales, Sydney, Australia.4 Save Sight Institute, Sydney Eye Hospital, Sydney Medical School,University of Sydney, Sydney, Australia.

Footnotes and Financial Disclosure(s):The authors have no proprietary or commercial interest in any materialsdiscussed in this article.

Abbreviations and Acronyms:AZA ¼ azathioprine; BCC ¼ basal cell carcinoma; CSA ¼ cyclosporine;CYP ¼ cyclophosphamide; ICD ¼ International Classification of Disease;IED ¼ inflammatory eye disease; IQR ¼ interquartile range; NHL ¼ non-Hodgkin’s lymphoma; NMSC ¼ nonmelanocytic skin cancer;SCC ¼ squamous cell carcinoma; SIR ¼ standardized incidence ratio.

Correspondence:William B. Yates, St Vincent’s Hospital, Department of Ophthalmology,Room 516, Level 5, Sydney, New South Wales 2052, Australia. E-mail:yates.willb@gmail.com.

9