Reliability of dietary information from surrogate respondents.

A self-administered food frequency questionnaire was included as part of a case-control study of breast cancer in 1980-82. In 1986-87, a second food frequency questionnaire was sent to surviving cases and husbands of deceased cases; 30 spouses (86% response rate) and 263 surviving cases (88% response rate) returned questionnaires. The dietary questions concerned consumption of specific food items by the case before diagnosis of breast cancer. Missing values were less common in the second questionnaire; there was no significant difference in missing values between surviving cases and spouses of deceased cases. Kappa statistics comparing responses in the first and second questionnaires were significantly lower for spouses of deceased cases than for surviving cases. Reported level of confidence by the husbands regarding knowledge about their wives' eating habits did not influence the kappa statistics or the frequencies of missing values. The lack of good agreement has important implications for the use of proxy interviews from husbands in retrospective dietary studies.