861 Mind the Gap: A Comparison of the Randomized Clinical Trial Efficacy and Real World Effectiveness of Tofacitinib for the Treatment of Inflammatory Bowel Disease

Abstract

INTRODUCTION: Randomized clinical trials (RCTs) represent the gold standard for clinical evidence. However, they have also raised concerns over generalizability due to unrealistic patient selection and endpoints. Real-world data (RWD) is receiving growing interest from the FDA as a complementary source of evidence. However, the suitability of these data for RCT-grade evidence and best practices in their use remain unclear. Here we compare the real-world effectiveness of Tofacitinib with efficacy assessed by RCTs of Inflammatory Bowel Disease (IBD). METHODS: Electronic health records (EHR) at the University of California San Francisco (UCSF) were reviewed for IBD patients prescribed Tofacitinib following a pre-protocol. The primary outcome was drug effectiveness measured by continued use and by RCT endpoints. Secondary outcomes assessed the completeness of RWD to measure RCT endpoints and the number of patients meeting RCT inclusion criteria. RESULTS: 68 Ulcerative Colitis (UC) and 18 Crohn’s Disease (CD) patients received Tofacitinib (Figure 1). Most of the data needed to calculate the RCT endpoints was available in the EHR (77% and 91%) with the remainder imputed. The baseline characteristics of the UCSF and trial cohorts were similar, except for greater disease duration and 100% TNFi failure in UCSF patients (Table 1). However, 0% of the real-world UC cohort met the RCT inclusion criteria (Table 2). The rate of continued treatment at 1 year was greater than 50% irrespective of disease subtype. Time-to-treatment-failure curves were nearly identical for UC and CD, with all drop-off occurring within the first 6 months. This rate of continued treatment in practice was significantly greater than the RCT primary endpoint and least stringent secondary endpoint (P-values 0.003 and 0.08) for UC, and for CD (P-value 0.02). However, the rates of meeting the primary endpoint for both subtypes using the Mayo score and CDAI were identical (P-values 0.5 and 1.0). CONCLUSION: Routinely collected data is sufficiently detailed to measure IBD trial endpoints. Analysis of RWD reproduced Tofacitinib efficacy measured by RCTs, despite trial selection criteria excluding most UCSF patients. However, real-world use patterns indicate much greater treatment durability than that suggested by trial endpoints, likely related to different treatment failure thresholds used in clinical care. These results support a role for JAK inhibition in CD, and overall highlight the value of real-world evidence to complement that obtained by RCTs.