Background/Purpose: The possible association between tumor necrosis factor inhibitors (TNFi) in the treatment of juvenile idiopathic arthritis (JIA) and an increased risk of malignancy remains uncertain. We combined and analyzed data from two large administrative claims sources to further evaluate malignancy rates in JIA.

Methods: Using national U.S. Medicaid claims from 2000-2010 inclusive and U.S. MarketScan claims from 2010-2014 inclusive, we identified cohorts of children with JIA using physician diagnosis codes and medication prescriptions. We identified non-JIA comparator cohorts of children diagnosed with attention-deficit hyperactivity disorder (ADHD). Children with any physician diagnosis code for malignancy prior to the start of follow-up were excluded, and all children had a ≥ 6 month baseline assessment period. JIA medication exposures included MTX (methotrexate or leflunomide), TNFi, and other systemic immunosuppressant agents (non-TNFi biologics or other non-biologics (e.g., cyclosporine)). All follow-up time following any medication exposure was considered exposed. Incident cancers were identified using a combination of diagnoses and treatment codes (claims for chemotherapy, radiation, or surgery). SEER cancer surveillance data were used to calculate expected cancer rates according to the age, sex, and race distribution of each cohort. Standardized incidence ratios (SIR) were calculated for the observed cancer outcomes compared to SEER estimates.

Results: We identified 2,657,899 children with ADHD and 27,621 children with JIA and observed 841 and 23 incident malignancies, respectively (Table). The outcome identification algorithm appeared specific and sensitive for incident cancer, since the large ADHD comparator had SIR of 1.18 [1.11-1.27]. SIRs were significantly increased among all JIA patients (2.7 [1.7-4.0]) and among those who did not receive any medications of interest (2.4 [1.1-4.5]). We observed 8 malignancies after 15,269 person-years of observation in children who received TNFi; the corresponding SIR (3.3 [1.4-6.6]) was similar to that for children who did not receive TNFi. Based on 7 cases, the SIR associated with any use of other systemic immunosuppression was markedly elevated (14.0 [5.6-28.9]), and included use of abatacept, cyclosporine, rituximab, tacrolimus, and tocilizumab each by 1 patient and anakinra by 2 patients.    

Conclusion: We did not observe a marked incremental increase in incident malignancies following treatment with TNFi compared to malignancy rates associated with the diagnosis of JIA. Receipt of non-MTX, non-TNFi systemic immunosuppressive therapies, a likely indication of severe or uncontrolled JIA, was strongly associated with an increased rate of malignancy.