It is of interest to predict the in vivo behavior of an oral extended-release drug product based on its in vitro dissolution profile. In some cases a suitable convolution-based prediction model can be found. We present a methodology for developing statistical models of in vitro-in vivo relationships under the framework of the mixed-effects nonlinear model and discuss methods for assessing the validity and strength of the relationship. These methods are illustrated and contrasted with a level A in vitro-in vivo correlation using data from a study involving four different formulations of an oral extended-release drug product.