Purpose. To explore how medical doctors judge risk reducing drug therapies when the benefit is presented in terms of number needed to treat (NNT) or postponements of adverse events and whether they use numerical information when explaining risk reductions to patients. Methods. A questionnaire was mailed to 160 general practitioners and 160 internists. A clinical vignette presented a male patient at high risk for cardiovascular disease. Drug therapy would reduce his risk of heart attack. Using a simple statistical model of statin efficacy we estimated that after five years of treatment, an NNT of 15 to prevent one heart attack was roughly equivalent to postponement of heart attack by 6 months, whereas an NNT of 30 was equivalent to postponement by 3 months. In the vignette effectiveness of therapy was presented either as NNT=15, NNT=30, postponement of 3 months or postponement of 6 months. The doctors were randomly allocated to one of these risk reduction measures and asked to evaluate the drug on a zero (“a very poor choice”) to ten (“very good choice”) scale. Subsequently, we asked whether they usually inform patients about risk reductions in numerical terms or qualitative terms. Finally, they were asked to indicate on a zero to ten scale whether they perceived benefits of preventive drug therapies as a certain, but small benefit (zero) or as a lottery with a small probability of a large benefit (ten). We used t-tests to compare means. Results. The response rate was 38%. Doctors rated therapies presented in terms of NNT higher than those presented in terms of postponements (mean 5.4 versus 3.1, p<0.001). On the “lottery” versus “certain benefit” scale the mean scores for thiazids and statins were 5.8 and 4.2 respectively. Only 35% of the doctors stated that they use numerical terms when informing patients about risk reductions. Of these, the majority used NNT (83%) or absolute risk reduction (77%), whereas less than half reported using survival data (49%) or relative risk reductions (38%). Conclusions. Medical doctors rated benefits in terms of NNT higher than equivalent postponements of heart attacks. Neither the small but certain benefit nor the lottery model was endorsed by the respondents. The majority of medical doctors seem to avoid numerical terms when informing patients about risk reducing therapies.