4CPS-165 Estimation of precision and accuracy of five population pharmacokinetics models of infliximab in patients with inflammatory bowel diseases

Background Infliximab is a monoclonal antibody approved for the treatment of inflammatory diseases. The best approach in adjusting the dose of infliximab because of the high interindividual variability in serum concentrations is with a Bayesian approach. Purpose Our aim was to estimate the precision and accuracy of five pharmacokinetic models (PopPKmodel) in patients with inflammatory bowel disease. Material and methods An observational retrospective study was designed. Patients with ulcerative colitis or Crohn’s disease treated with infliximab during 2014 were included. Trough blood samples for determining infliximab were drawn. Five PopPKmodels were implemented in NONMEM: PopPK in ulcerative colitis and Crohn’s disease. The Infliximab concentrations were estimated from five models at the sample times, through the empirical Bayesian of estimates (EBEs) of the pharmacokinetic parameters. To validate these models, bias of estimated concentrations were calculated as the mean residual predictive error (MRPE) and the precision was calculated as the root mean square predictive error (RMSPE) in our population. Results Two hundred and seventy-three serum infliximab concentrations from 160 patients (54% males and 46% females) were included. The mean age was 36 years (CI 95%: 31 to 41), weight 73.1 kg (CI 95%: 71.1 to 75.2) and 3.92 mg/dL (CI 95%: 3.86 to 3.98) baseline serum albumin concentration. 62.5% of patients were diagnosed with Crohn’s disease and 36.3% for ulcerative colitis. The mean trough serum concentration of infliximab was 4.1 mg/L (CI 95%: 3.6 to 4.6). 68.1% of patients were treated with infliximab and 31.9% with biosimilar. Bias of estimated concentrations (MRPE) and precision (RMSPE): Conclusion In our study, neither PopPKmodels overestimate infliximab concentrations in the population, although Model 4 was better, (i.e. closer to zero) in terms of bias and accuracy. Reference and/or Acknowledgements 1. Fasanmade, et al. 2009: Model 1; 2. Fasanmade, et al. 2011: Model 2; Buurman y col, 2015: Model 3; Dotan y col, 2014: Model 4 and Brandse y col, 2017: Model 5. No conflict of interest