The aim of this study was to develop a population pharmacodynamic (PPD) model to describe uric acid (UA)–lowering effects in patients treated with febuxostat based on electronic medical records in 2 independent hospitals (university and city hospitals). Interhospital differences in the PPD model were also evaluated. We conducted the following 2 approaches to build the PPD models. A PPD model was developed separately using individual hospital data, and structural models and covariates between the two hospitals were compared (approach A). Another PPD model was developed using all available data from both hospitals, and differences between the 2 hospitals were evaluated by performing a covariate analysis on all PPD parameters (approach B). PPD analyses were performed by NONMEM using data from 358 patients. In both approaches, one indirect response model was established. In approach A, 2 diuretics (loops and thiazides) and renal function tests (Scr or BUN) were selected as covariates for the UA baseline level (serum UA levels just before the febuxostat treatment), whereas 2 diuretics and BUN were selected in approach B. A covariate analysis indicated that loops and thiazides increased UA baseline levels by 7%–14% and 6%–11%, respectively. In approach B, “hospital” was identified as a significant covariate for the UA baseline level
the baseline level was 7% higher in the city hospital. A PPD analysis may provide a precise description of the time course of the UA-lowering effects of febuxostat and quantitatively detect an interhospital difference in the UA baseline level.