OBJECTIVES: This research investigates the use of longitudinal/cohort observational designs containing patient-specific data for the economic evaluation of pharmaceuticals in development or on the market, and appraises the value of such observational datasets, relative to the construction or use of RCT study designs. METHODS: A systematic review of a health economics literature database (HEED) was conducted to explore the use of observational and RCT data in applied pharmacoeconomic evaluations in the UK (1990-1999). A comparative analysis of longitudinal datasets (containing resource use and outcomes data) developed in the UK (e.g. DIN-LINK, GPRD) was undertaken with an assessment of the potential analytical approaches (multivariate techniques, cross-design synthesis) and methodological and practical considerations. RESULTS: A total of 71 studies were identified, 59 CEAs only, 6 CUAs only and 6 where both CEAs and CUAs were conducted. The table presents the type of data used in these studies for the three core data items of a pharmacoeconomic evaluation. RCT data represented the majority source for all three items, in particular for clinical event and disease outcome data. Observational data sources were more frequently used to provide resource use consequence data. None of the reviewed studies made use of any existing prospective or retrospective database. Number of studies using different types of data (review of 71 UK studies) Data source Core data for economic evaluations Observational RCT Observational & RCT Other Resource use consequences 21 27 3 20 Clinical events 19 37 6 9 Disease outcomes 19 33 6 13 All 3 sets of core data 9 14 1 6CONCLUSIONS: Regardless of the system of health care, there is an obvious and universal need for reliable data on the effectiveness and costs of clinical interventions. Observational datasets have implications for outcomes research conducted in pharmaceutical companies. There is little literature on the use of observational data in economic evaluation, and more debate on this subject is needed. As we gain a better understanding of the tradeoffs between internal and external validity, it is predicted that more pharmacoeconomic analyses will increasingly be based on observational data or a combination of observational and RCT data.