Background/Purpose: For rheumatoid arthritis (RA) patients (pts) with inadequate response to a TNF inhibitor (TNFi), limited evidence exists from observational studies and indirect comparisons of randomized trials to support switching to a nonTNFi vs another TNFi. These exploratory analyses evaluate clinical outcomes and healthcare resource use (HCRU) after switching from a TNFi to another TNFi or nonTNFi biologic across 2 de-identified real world data sources.

Methods: Pts (≥18 years) with ≥2 outpatient or 1 inpatient visit for RA (ICD-9: 714.xx) and (1) HCP-reported TNFi discontinuation and switch to a biologic or conventional DMARD (cDMARD) within 180 days in Humedica EHR database (1/07-7/13) or (2) switched from TNFi to another biologic (or 1 pt to tofacitinib) in Truven Marketscan^® claims database (2010-2013) were included. Pts had continuous enrollment/follow-up ≥6 months (mo) before and ≥12 (claims) or 18 (EHR) mo after discontinuation. EHR cohort was followed for 18-mo all-cause HCRU and change in patient-reported pain scores (0-10 on provider-determined scales; ≥30 days pre-/post-switch). Claims cohort was evaluated for subsequent biologic switching and RA-related HCRU costs. Multivariable analyses evaluated the impact of switching to a TNFi vs nonTNFi on RA‑related costs.

Results: Of 2799 pts who discontinued a TNFi (47% etanercept, 28% adalimumab, 21% infliximab, 4% other) in the EHR cohort, reasons were lack of efficacy (14%), AE/other clinical reason (16%), cost (6%), or unknown (65%). Following discontinuation, 21% switched to another biologic (67% TNFi) and 11% to cDMARD. In the claims cohort, 68% switched to another TNFi, of whom 43% switched again vs 28% of those first switched to a nonTNFi biologic. Among EHR pts with pre-/post-switch pain scores, TNFi (n=58), nonTNFi biologic (n=19), or cDMARD (n=55) switchers had mean (standard deviation [SD]) pain reductions of 0.72 (3.12), 1.11 (2.87), and 0.42 (3.88), respectively. Office visits comprised the largest HCRU category in the 18 mo after switching; mean (SD) number: 42.7 (34.5), 57.6 (47.3), and 38.4 (33.3) for TNFi, nonTNFi, or cDMARD switchers, respectively. Nearly half (n=7.2) the difference in office visits with TNFi vs nonTNFi was associated with biologic administration procedures. Unadjusted mean (SD) RA‑related costs were $27544 ($21100) for TNFi vs $44742 ($30743) for nonTNFi switchers. Including biologic administrations without RA diagnosis (TNFi, $389; nonTNFi, $1126), most (79%) of the mean cost differences were attributed to outpatient visits, followed by prescriptions (18%). Biologic administration visits accounted for 89% of outpatient cost differences. NonTNFi therapy was associated with 32% higher total RA-related costs in adjusted analyses (p<0.0001).

Conclusion: NonTNFi switchers were less likely to switch again and had greater pain reduction vs TNFi switchers (in a small subset). NonTNFi switchers had higher all‑cause HCRU and RA-related costs, largely due to in‑office administration. NonTNFi therapies that do not require in-office administration may reduce switching and costs. Disease outcomes and financial consequences of clinical decision-making warrant future prospective structured research.