Bias & Precision for Hemodynamic Parameters Resulting from best Model & weighted Model Strategies Based on the Akaike Information Criterion

Robert Luypaert¹, Steven Pieter Sourbron², Johan de Mey¹

The Akaike Information Criterion, which can rank pharmacokinetic models on the basis of goodness-of-fit and number-of-parameters, was applied to the exchange model (general) and the modified Tofts model (valid for negligible plasma mean transit time). Data with fixed noise level and varying validity of the Tofts model were simulated using the exchange model. Bias and precision for parameter estimates obtained by selecting the model with highest Akaike score (best model) and by calculating weighted averages based on the Akaike scores (weighted model) were studied. Although both approaches led to optimized estimates, unexpected properties that could hamper their usefulness were detected.