When economic evaluations are conducted alongside clinical trials estimates of resource utilization (RU) are likely to be inaccurate. This problem arises due to presence of protocol-mandated visits and forced compliance. It is rare to find two databases that may be compared to determine the scale of this problem. OBJECTIVE: 1) to compare the resource utilization associated with rheumatoid arthritis from a clinical trial database to a cross-sectional database, and 2) to understand advantages and disadvantages of using each database in economic analyses. METHODS: In this novel study, two databases were examined: one a pragmatic, cross-sectional database (CSD: with 6-month retrospective recall of RU) of rheumatoid arthritis patients (RA) and the other a database from a phase III clinical trial program (CTD) in RA matched on several demographic variables. Subjects were age/gender matched to control for differences in the database populations. Hospitalizations, GP and specialist visits were compared between the databases. RESULTS: When the CTD is examined for just protocol visits, a total of 8 specialist consultations were recorded. When the CTD is examined for non-protocol visits, few additional visits are recorded: 0.12 GP visits, 0.11 specialist consultations, and 0.01 hospitalizations. The addition of the non-protocol visits to the protocol total visits increase RU to 8.12 GP visits and 8.11 specialist consultations. Based on data from the CSD, over a 6-month time frame, randomly matched on age and gender, an average of 2.1 GP visits, 2.21 specialist consultations, and 1.3 hospitalizations were reported by patients. CONCLUSIONS: The inclusion of protocol-mandated visits substantially overstates true RU while their omission understates thereby leaving researchers in a quandary. The use of CTDs for modeling purposes raises the issues of generalizability and accuracy and suggests that naturalistic rather than piggyback trials should be used for PE evaluations.