Background: If women tend to forget and underreport their past oral contraceptive (OC) use, but the recall among cases is enhanced by the presence of disease, recall bias may explain some reported health effects of OC use.
Methods: Two different sources of information on lifetime OC use were compared for 427 (84%) of a community-based sample of 511 women aged 20-34: (i) structured interviews, using a life event calendar and picture display as memory aids, and (ii) a register of all prescriptions dispensed by pharmacies in the county since 1970.
Results: Interview data and pharmacy records showed high levels of agreement for any OC use, current use, time since first and last use, total duration of use, and for duration of use in different 'time windows'. But there was a tendency to under-report specific kinds of OC used in the past.
Conclusion: Underreporting of OC use among non-cases would usually introduce little or no bias (as compared to pharmacy records) for this kind of interview and women. However, it may be preferable to use interviews for current OC use, and pharmacy records for specific kinds of OC used in the past.
PIP: In studies where women are interviewed about oral contraceptive (OC) use only after the emergence of disease (e.g., breast cancer), the experience of disease may enhance recall and reporting of OC use, resulting in overestimation of relative risks. This community-based study compared two sources of information on ever-use of OCs--structured interviews and pharmacy prescription records--in 427 women 20-34 years of age who were long-term residents of Sweden's Jamtland County where registration of prescriptions has been virtually complete since 1970. A life-events calendar and photographs of OCs in use since 1970 were used in the interviews to facilitate recall. 153 women reported current OC use; the pharmacy register failed to document a current prescription for 37 (24%). Only 4 (1.5%) of the 274 women who denied current OC use had an active prescription on file. Of the 37 women identified as never-users by interview, 2 (5%) had records of OC prescriptions (for under 2 years of use). Of the 36 women identified as never-users by pharmacy records, only 1 reported ever-use. Compared to the pharmacy registry, 13% reported a shorter and 28% a longer duration of OC use. Of women who reported OC use within 5 years of interview, 5-10 years earlier, and more than 10 years ago, 2.6% (8/312), 4.6% (14/307), and 12.8% (26/203), respectively, had no prescriptions on file for that time period. Discrepancies between the two data sources in time since first OC use were usually less than 2 years; the tendency was for women to overestimate rather than understate the actual duration of OC use. These findings suggest that interview data on OC use introduce minimal--if any--bias. Agreement between interviews and records in terms of the specific brands of OCs taken was less satisfactory. 19% of users of high-dose OCs, for example, did not recall the type of pill dispensed by the pharmacy. Thus, use of pharmacy records is recommended over interview data in studies where the specific OC formulation is of interest.