One of the fundamental factors determining the success of a new drug candidate is the identification of a dose range that is both efficacious and safe. This is achieved via the use of pharmacokinetic (PK) and pharmacodynamic (PD) data to establish dose-(concentration)-response relationships for the various drug effects. Modelling and simulation (M&S) techniques are now commonly used to analyse clinical PKPD data as part of a model-based drug development framework, however, their use in the pre-clinical stages is limited and within the safety sciences is virtually non-existent. Application of such techniques to pre-clinical safety pharmacology data may help to improve the predictability of adverse effects compared to conventional methods. The aim of this project is to develop PKPD models using safety pharmacology data, which can be used to predict cardiovascular (CV) effects in man.Non-linear mixed effects modelling was used to analyse the PKPD relationships for the haemodynamic effects of NG-nitro-L-arginine methyl ester (L-NAME), milrinone and doxazosin in three species (rat, dog and guinea-pig). PK data were described by compartmental models and PD data by either direct, effect compartment or indirect response models with stimulatory or inhibitory effects characterised by non-linear (Emax) or linear (slope) functions. Baseline mean blood pressure (MBP) and heart rate (HR) either displayed circadian rhythms, which were described by cosine functions for dog or a biorhythm model specifically developed for rats, or an anaesthetic effect in guinea-pigs, which was described using a linear function. A sequential modelling approach was taken where PK and baseline values were fixed when fitting the PD data. Human responses to the 3 compounds were predicted using the PD values determined from each animal model. Human PK and baseline were fixed, pharmacological values were set as the animal values and physiological values were scaled according to body weight using an exponent of -0.25.Adequate fits to PK, baseline and PD data were generally observed, although issues with sampling times and dose ranges restricted the model options in a number of cases. For L-NAME and milrinone, where a delay in effect was observed, indirect models gave better fits than effect-compartment models. Doxazosin had direct effects in all species. Some issues with model fit were due to the fixed baseline, therefore a simultaneous fit of baseline and PD data would be recommended. However, the complexity of the rat baseline model could lead to unusual profiles with this method and thus it is not recommended for standard use. Generally values agreed with literature observations although no consistency in the PD values was observed across species or between MBP and HR. In most cases human response was under-predicted but there was no consistent pattern regarding the level of under-prediction across compounds. The extent of human effect on HR by L-NAME and milrinone was successfully predicted using dog values. Unfortunately dog was not predictive for doxazosin or MBP for any compound. Time course of effect was not predicted successfully in any of the cases. It was theorised that more complex models containing the CV feedback mechanisms may be required for some of the unsuccessful predictions.Overall PKPD modelling can be used to describe CV safety pharmacology data in different species and if study design and modelling issues were addressed, greater accuracy could be obtained. Dog showed the best potential for prediction of CV measures and predictions might be improved with the use of CV models that incorporate accurate feedback mechanisms. If greater model complexity is restricted to physiological aspects the use of such models in safety pharmacology may also help to give more insight into mechanism of action.