Maximum likelihood estimation via the expectation maximization algorithm (EML) was implemented into ADAPT 5 to support population pharmacokinetic/pharmacodynamic studies. Two detailed simulation studies were conducted using a one compartment pharmacokinetic model with sparse data and a complex pharmacokinetic/pharmacodynamic model to evaluate this algorithm. Additionally, a population analysis of the minimal model for the intravenous glucose tolerance test in mice was performed as an example of biological study using real data. In the simple pharmacokinetic model, less than 1% mean prediction errors were obtained for the estimated population mean and covariance parameters. For the complex model, the EML estimation resulted in prediction error less than 8% for the mean parameters and these results were not influenced by different initial guesses for model parameters. The EML estimation for the animal study resulted in a correlation coefficient between predicted and measured values of glucose concentration in plasma of 0.97.