Aplastic anemia (AA) is a rare, severe disease of mainly unknown origin. Numerous case history reports have incriminated drugs in the etiology of this disease. Because those reports were questionable, a case-control study was conducted in France between 1985 and 1988. Cases selected from the national register were eligible for inclusion when at least two blood lineages were depressed (hemoglobin < or = 10 g/100 mL and reticulocytes < or = 50 x 10(9)/L, granulocytes < or = 1.5 x 10(9)/L, platelets < or = 100 x 10(9)/L) and when the bone marrow biopsy was compatible with the disease. Using a standardized questionnaire, trained investigators interviewed one AA patient and two groups of controls (two hospitalized patients and one neighbor of the AA patient) matched for age, sex, and interviewer. One hundred forty- seven AA patients, 287 hospitalized controls, and 108 neighbors were interviewed. The occurrence of AA was analyzed by matched design with relation to medical history and drug use during the last 5 years, and specifically during the last year. Three times as many AA patients reported having suffered from clinical hepatitis during the last 6 months than either type of control. Similarly, a higher proportion of AA patients reported a history of chronic immune disorder, mainly rheumatoid arthritis (odds ratio of 6.8), and a previous use of gold salts and D-penicillamine in the 5 previous years (odds ratio of 4.9 for each drug). An excess of colchicine and allo/thiopurinol intake in the 5 previous years was observed among the AA patients (odds ratio equal to 4.1 and 3.6, respectively). These results for gold salts, D- penicillamine, and colchicine were confirmed when looking for drug use within the last year. A moderate risk was associated with acetaminophen or salicylate intake during the 5 previous years or during the last year (odds ratio between 1.8 and 2.0). The frequent use of salicylates within the last year was associated with a high risk of AA (odds ratio of 5.0). A high risk was also associated with indolic derivative intake but only when comparing AA patients to neighbor controls. No association could be evidenced with diclofenac intake, whatever the control group. Differences observed with recently published studies suggest that targeted studies on each category of drugs according to the treated pathologies should be initiated.