Although most clinical trials encompass aspects of safety, methods for assessing the safety of a drug by recording adverse events have been poorly studied. It has been suggested that adverse events rather than adverse drug reactions should be monitored, since a reliable determination of which events were caused by the drug and which were not is only possible after analysing data from a substantial number of clinical trials. In the present study adverse events were monitored to see the extent to which events recorded on the case record forms were reported as adverse events. Data from omeprazole and felodipine programmes were used, comprising altogether 143 clinical trials from eight different projects, and encompassing 12,069 patients in whom 11,812 events were recorded. The first project was started in 1982 and the last in 1988. Overall, 74% of recorded events were entered on a special adverse event form used in the trials, and 26% were not. Initially, about 35% of adverse events were not reported as such, as opposed to 13% towards the end of the study period. Serious adverse events were reported less frequently than non-serious events, but in the most recent project all serious adverse events were reported. Adverse events in women were reported more often than adverse events in men, and reporting was more complete for the middle-aged than for the oldest and the youngest persons. Certain types of adverse events were reported more completely than others. In conclusion, the transition from registering adverse reactions to registering adverse events has been a gradual one in spite of intensive educational efforts when the projects were started.(ABSTRACT TRUNCATED AT 250 WORDS)