Objective: This study was conducted to define an evaluation modelto estimate changes in the co-prescription of gastroprotective agents(GPAs) induced by rofecoxib in the treatment of osteoarthritis (OA).Methods: On the basis of a cross-linking information, which werestored in different administrative and clinical databases, a multivariateregression analysis was used to develop the model. Data were collectedby 30 general practitioners of the Local Health Unit of Ravenna (middle-northof Italy). Results: The study population consisted of 2,944 patientstreated with non-steroidal anti-inflammatory drugs (NSAIDs) and 487treated with rofecoxib. Patients treated with rofecoxib generallypresented a higher number of gastrointestinal damage risk factorsand also a lower level of GPAs co-prescription compared to thosetreated with NSAIDs. Including in the model variables such as typeof anti-inflammatory treatment (NSAIDs or rofecoxib), gender, ageby class, previous hospital admissions due to gastrointestinal complications,number of different NSAIDs used, and prescription of corticosteroids,the regression equation and its coefficients were identified. A non-linearrelationship between the percentage of patients treated with rofecoxiband the relative reduction of GPAs co-prescription was found. Ithas been estimated the basis of the registered percentage of patientstreated with rofecoxib (17,6%) adjusting for gastrointestinal demagerisk factors, and on a 63% (IC95%; 55%-70%) relative reduction ofGPA use with rofecoxib with respect to NSAIDs was estimated. Conclusions:Based on data collected in the clinical practice after the introductionof rofecoxib, a model evaluating the relationship between the frequencyof its use in the OA population and the expected reduction of GPAs,has been developed.