Important: The POPLINE website will retire on September 1, 2019. Click here to read about the transition.

Drs. Thomas and Noonan reply re "Comparison of Recalled and Validated Oral Contraceptive Histories" [letter]

Author: 
Thomas DB; Noonan EA
Source: 
AMERICAN JOURNAL OF EPIDEMIOLOGY. 1995 Apr 15; 141(8):791.
Abstract: 

Dr. Realini asserts in his letter that the results of the WHO Collaborative Study of Neoplasia and Steroid Contraceptives concerning oral contraceptives and breast cancer could have been biased as a result of better recall of prior oral contraceptive use by cases in comparison to controls. Although this is a possibility in any case-control study based on data collected by interviews, being able to validate positive oral contraceptive histories of more cases than controls does not necessarily mean more cases than controls who were users of oral contraceptives gave a history of such use. The issue of bias due to recall was discussed in the original paper. The medical records of women who claimed contraceptive use were checked for brand names and duration of use; the medical records of women who did not were not checked. This procedure did not alter their classification as users or nonusers. Since few combined oral contraceptives were available, and erroneous reports of use of combined or noncombined preparations were corrected, the estimated values of the relative risk of breast cancer in women who ever used combined oral contraceptives could not be appreciably influenced by any differences in the proportion of cases or controls whose oral contraceptive histories were supplemented by information from medical records. However, such differences could alter estimated values of the relative risk in relation to duration, latency, or recency. Information was obtained from the medical records of users in 27% of cases and 18% of controls. These percentages varied with center (0-94% of cases, 0-89% of controls). Information was most frequently obtained from the medical records of long-term and current or recent users in both groups. Similar results were obtained separately from countries in which information from medical records was obtained for relatively high and low proportions of users, and in individuals whose use was ascertained solely from interviews and from both interviews and medical records.

Language: 
Year: 
Document Number: 
103913
Add to my documents. Add to My Documents