Evaluation of Tobramycin Exposure Predictions in Three Bayesian Forecasting Programmes Compared with Current Clinical Practice in Children and Adults with Cystic Fibrosis.

Abstract

Bayesian forecasting (BF) methods for tobramycin dose individualisation has not seen widespread clinical adoption, despite being endorsed by clinical practice guidelines. Several freeware and commercial programmes using BF methods are available to support personalised dosing. This study evaluated exposure estimates, dose recommendations, and predictive performance compared with current clinical practice. Data from 105 patients (50 adults and 55 children) with cystic fibrosis who received intravenous tobramycin treatment and had paired concentration-time measurements were analysed using (1) log-linear regression analysis, and (2) three BF programmes: TDMx, InsightRX, and DoseMe. Exposure estimates and dose recommendations were compared using the Wilcoxon signed-rank test and Bland-Altman analysis. Predictive performance of BF programmes was compared based on bias and imprecision. Median estimated tobramycin exposure with current clinical practice was significantly lower (87.8 vs. 92.5, 94.0 and 90.3mg h l-1; p≤0.01), hence median subsequent dose recommendations were significantly higher (10.1 vs. 9.4, 9.4 and 9.2mgkg-1; p≤0.01) compared with BF programmes. Furthermore, median relative dose-adjustment differences were higher in adults (>10%) compared with children (4.4-7.8%), and differences in individual dose recommendations were >20% on 19.1-27.4% of occasions. BF programmes showed low bias (<7%) and imprecision (<20%), and none of the programmes made consistently significantly different recommendations compared with each other. On average, the predictions made by the BF programmes were similar, however substantial individual differences were observed for some patients. This suggests the need for detailed investigations of true tobramycin exposure.