From a case-control study of the relation between oral contraceptives and breast cancer carried out in East Germany during 1982-1986, the authors obtained information on oral contraceptive use through interviews of study subjects and from the records of prescribing gynecologists. The degree of agreement regarding information from these two sources was assessed for 234 breast cancer cases and 524 controls who had ever used oral contraceptives. Agreement between information obtained from medical records and that from interviews on total duration of use, number of episodes of use, and time since first and last use was reasonably good, and levels of agreement did not differ appreciably between cases and controls. Lower levels of agreement were observed for individual brand names and the duration of use of specific brands. Attempts should be made to obtain information on specific brands from medical records when investigating the effects of individual preparations.
PIP: Researchers used the results of interview data on oral contraceptive (OC) use in both breast cancer cases and hospital controls to compare the accuracy of their recall with data from gynecologic records. The 234 breast cancer cases were patients at the Central Institute of Cancer Research in Berlin, the German Democratic Republic. The 524 age=matched controls were patients in the ear, nose, and throat; eye; orthopedic; and skin wards at Klinikum Berlin-Buch, the district hospital. All the data were collected between November 1982 and June 1986. Interviewers used a calendar and samples of OC brands to help the women recall periods of use and brand names used. The availability of gynecologists' records was significantly associated with use of medical care, including cervical smears, duration of OC use, and the time since last OC use (p .01). When the interval since first and last OC use within 12 months, the agreement between patients and gynecologists was essentially the same for both cases and controls (first OC use, 72.4% vs. 67.8%; weighted kappa = .81 vs. .79; p .001) (last OC use, 80.4% vs. 68.7%, weighted kappa = .86 vs. .89; p .001). The highest percentages of agreement on ever use of OCs occurred with the most recently introduced brand name (Minisiston, 98.7% for cases and 99% for controls) and phasic OCs (Deposiston, 97.9% and 96.4%, respectively; Sequenz-Ovosiston, 92.7% and 89.9%, respectively). These findings showed that use of a calendar and brand samples results in rather accurate recall of overall OC use. They also indicated that having breast cancer does not affect recall of OC use. This study's investigators suggested that in studies examining the effects of individual substances or doses, researchers should try to obtain data from medical records.