Clinical trials may lead to conflicting results. We studied how different ways of reporting results affected physicians' recommendations. A questionnaire distributed to 148 general practitioners presented results of a clinical trial where a reduction of cardiac events and an increase of mortality was reported. Results were shown in four different ways--relative risk reduction, absolute risk reduction, percentages of event-free patients, number needing to be treated to prevent an event--as if they derived from different trials. A fifth presentation was the reduced rate of cardiac events along with the increased rate of mortality. Physicians were asked to estimate how much they would be willing to prescribe each drug. The mean agreement of physicians' decisions was 77 (28)% for relative risk reduction, 24 (28)% for absolute risk reduction, 37 (37)% for different percentages event-free patients, 34 (34)% for number need to treat, and 23 (28)% for events reduction and mortality for increase (p < 0.001 relative risk vs others). The method of reporting trial results and the completeness of information in the case of controversial results affects physicians willingness to prescribe.