Sensitivity Analysis and Uncertainty Quantification in Pulmonary Drug Delivery of Orally Inhaled Pharmaceuticals

Abstract

In spite of widespread use of modeling tools in inhalation dosimetry, it remains difficult to quantify the output uncertainties when subjected to various sources of input variability. This study aimed to develop a computational model that can quantify the input sensitivity and output uncertainty in pulmonary drug delivery by coupling probabilistic analysis package NESSUS with ANSYS Fluent. An image-based mouth-lung model was used to simulate the transport and deposition of drug particles and variability in particle size, density, and inhalation speed were considered. Results show that input variables have different importance levels on the delivered doses to lungs. For a given level of variability, the delivered dose is more sensitive to the variance of particle diameter than that of the inhalation speed and particle density. The range of input scatters has a profound impact on the outcome probability of delivered efficiencies, while the input distribution type (normal vs. log-normal) appears to have an insignificant effect. Despite normal distributions for all input variables, the output exhibits a non-normal distribution. The proposed model in this study allows easy specification of input distributions to conduct multivariable probabilistic analysis of inhalation drug deliveries, which can facilitate more reliable treatment planning and outcome assessment.