NSAIDs are considered minor drugs in that their clinical benefit is limited to reduction of symptoms. Although they may be responsible for _ 25% of side effects, some of them fatal, most NSAIDs are prescribed for controversial indications. An intervention was prospectively undertaken to rationalize prescribing in an Emergency Department (ED). METHODS. The annual census of our ED is 50 000 pts., 40% of whom receive prescription drugs. Prescriptions written by a total of 57 physicians were collected during 5 d before (P1) a 45-d intervention, then 5 d immediately after (P2) and 12 d, 3 years later (P3). The intervention was an educational program with emphasis on contra-indications and adverse effects. Conforming indications were either established (E) (mainly rheumatological) or admitted(A) but controversial (trauma, ENT). For the latter, NSAIDs may be withheld. Between P1 and P2 all E&A indications were shown. Between P2 and P3, only the E indications were available. Prescribing errors were the main endpoint. RESULTS. NSAIDs were prescribed for 37/434 pts. (8.5%), 42/414 (10%) and 12/583 (2%) respectively (p <0.0001). Between P1 and P2, prescribing errors decreased from 19% to 14% (p = 0.76). In P3, 1 was found. Between P1 and P3, conforming prescriptions increased from 81% to 92% (p = 0.66) : NSAIDs decreased for trauma (50% to 8%; p = 0.02) and increased for rheumatology (19% to 42% ; p = 0.14). Duration of treatment decreased from 8.0 ñ 4.0 d to 6.0 ñ3.0 d (p = 0.07). CONCLUSIONS. Overall NSAIDs as well as incorrect prescribing decreased after an intervention. However 1) "Soft indications" stimulated prescribing ; 2) Its curbing was obtained by a limited list between P2 and P3. This approach may be applied to other controversial drugs.