Parameter Identifiability and Predictive Reliability Assessment of Kinetic Models for PLGA-Based Drug Delivery Systems

Poly(lactic-co-glycolic acid) (PLGA)-based drug-delivery systems necessitate kinetic models that can not only fit well with the data but also give a stable estimation of the model parameters and a reliable prediction.In this simulation-based study, the study analyzes the ability of four models (zero-order, first-order, Higuchi, Korsmeyer–Peppas) to accurately describe a Simulatedrelease profile for PLGA under varying sampling densities and observation-noise conditions.Nonlinear least-squares fitting was used to estimate the model parameters, and practical identifiability and uncertainty and withheld-data predictive analysis were conducted.The Korsmeyer–Peppas model was best (=0.969, RMSE=0.050, MAE=0.044 and 95% prediction coverage) for the nominal C3 condition with 20 sampling points and 5% observation noise, but with a parameter CV of 18%. The first-order model achieved (=0.962), RMSE = 0.056, MAE = 0.045, and 92% coverage. Maximum value of the parameters CV was equal to 25%, 28%, 24 and 35% for the zero order, first order, Higuchi and Korsmeyer–Peppas model respectively under the most difficult C5 condition.The results demonstrate that predictive fitting is not necessarily related to good parameter identifiability.