Patients' informed acceptance of chronic medical therapy hinges on communicating the potential benefits of drugs in quantitative terms. In a hypothetical scenario of treatment initiation, the authors assessed how three different formats of the same data affected the willingness of 100 outpatients to take what were implied to be three different lipid-lowering drugs. Side-effects were declared negligible and costs insured. Subjects make a "yes-no" decision about taking such a medication, and graded the decision on a certainty scale. Advised of a relative risk reduction--"34% reduction in heart attacks"--88% of the patients assented to therapy. All other formats elicited significantly more refusals (p < 0.0001): for absolute risk difference--"1.4% fewer patients had heart attacks"--42% assented; for inverted absolute risk--"treat 71 persons for 5 years to prevent one heart attack"--only 31% accepted treatment. When the data were extrapolated to disease-free survival--"average gain of 15 weeks"--40% consented. Similar responses were obtained for descriptions of an antihypertensive drug: 89% assented to therapy when given relative risk reduction but only 46% when given absolute risk reduction. The subjects were confident in both acceptance and refusal: 93% of the decisions were rated "somewhat certain" to "completely certain." The authors conclude that patients' views of medical therapy are shaped by the formats in which potential benefits are presented. Multiple complementary formats may be most appropriate. The results imply that many patients may decline treatment if briefed on the likelihood or extent of benefit.