Early predictions of a potential drug candidate’s time-course of effect and side-effects, based on models describing drug concentrations, drug effects and disease progression, would be valuable to make drug development more efficient. Pharmacodynamic modeling can incorporate and propagate prior knowledge and be used for simulations of different scenarios. In this thesis three population pharmacodynamic models were developed to describe the antipsychotic effects and the side-effects prolactin elevation and Extra Pyramidal Symptoms (EPS) following administration of D2-receptor antagonists, commonly used in the treatment of schizophrenia. Model parameter estimates of prolactin elevating potencies of six compounds correlated with in vitro values of receptor affinities, and parameters related to diurnal prolactin variation and tolerance were similar for the different compounds. The developed prolactin model can thereby be used to predict the time-course of prolactin elevation in patients for a drug candidate using information on in vitro affinity to the D2-receptor. Furthermore, the clinical antipsychotic effect and the prolactin elevation was found to correlate on the individual level for the three antipsychotic compounds investigated and a quantitative relation between D2-receptor occupancy in the brain and prolactin elevation was established. These results support the use of prolactin concentrations as a biomarker in drug development or for individual dose adjustments in clinical care. The developed model for spontaneously reported EPS adverse events, following treatment with one of five antipsychotics drugs, characterized both the duration and severity of EPS. The model successfully described both the proportions and number of transitions between severity grades and was shown to adequately simulate longitudinal categorical EPS data. Complex pharmacodynamic models are often associated with long estimation times and non-normal distributions of individual parameters. A method for shortening computation times by substituting differential equations for difference equations was evaluated and shown to be valuable for some models. In addition, transformation of distributions allowed for non-normal distributions of between-subject variability to be better characterized and thereby simulation properties were improved. In conclusion, population pharmacodynamic models for a range of D2-receptor antagonists were developed and together with the investigated methods the models can facilitate prediction of effects and side-effects in drug development.