Repeatability of sexual history in longitudinal studies on HPV infection and cervical neoplasia: Determinants of reporting error at follow-up interviews

Background Misclassification of sexual history due to faulty recall or reporting bias may be the reason for variability in the association between sexual history and human papillomavirus (HPV) infection seen in studies conducted in different geographical areas. This study aimed to assess the repeatability of questionnaire information on sexual-history variables and their correlates, using information from repeat interviews by six international prospective cohort studies. Methods. The pooled dataset included over 14 775 women interviewed on two separate occasions, of whom 5690 returned for a third interview. At each return visit women were re-asked questions on age at first inter-course and number of sexual partners. The six cohorts originated from studies in Denmark, Costa Rica, San Francisco, Toronto, Montreal and São Paulo. Results. Exact agreement between age at first intercourse recalled on separate occasions ranged from 60-85%, whereas exact recall rates for number of sexual partners were substantially lower and more study-dependent, varying between 20% and 77%. The intraclass correlation coefficients gauging the degree of repeatability in responses ranged from 0.68 to 0.97 for age at first intercourse and 0.08 to 0.94 for number of sexual partners. Age, ethnicity, education and cohort membership were the strongest predictors of reporting error for both sexual history markers, although study design characteristics also seemed to play a role. HPV infection status seemed to influence recall of number of partners, but not age at first intercourse. Conclusions. Information on sexual behaviours is not reliably collected in epidemiological studies of sexually transmitted diseases, which may influence the magnitude of relative risk estimates.