USING SENSITIVITY ANALYSIS TO QUANTITATIVELY INVESTIGATE PARAMETERS IN A PK/PD MODEL

We conduct a global sensitivity analysis on an existing pharmacokinetic/pharmacodynamic (PK/PD) model that investigates the effect of different dosages of doripenem, a broad spectrum antibiotic drug, on different strains of bacteria. Our work investigates what parameters had the greatest effect on the PK/PD model outputs (including bacteria counts, drug amounts, and drug concentration). This kind of analysis allows an understanding of what parameters may need more precise data collection, what aspects of the model may vary drastically from one individual to another, and to what extent the model applies to different conditions. In order to investigate the contributions of all parameters to the model's outcomes, we used a sensitivity analysis that utilized Latin Hypercube Sampling focused on Partial Rank Correlation. The preliminary results show the most significant sensitivity in parameters were those affecting the amounts of active and inactive antibiotic; specifically the elimination rate of the drug and the rates at which antibiotics transition from being active to inactive (p-value < .05, PRCC > .95). Sensitivity analysis is an integral part of model evaluation and this work is critical to all areas model implementation and validation.