On the use of surrogate respondents for controls in a case-control study of Alzheimer's disease

To examine the presence and extent of bias introduced by using surrogate respondents for healthy controls in a case-control study of Alzheimer's disease (AD).

Comparative study of matched responses to questionnaire ascertaining lifestyle issues.

University Hospitals/Case Western Reserve University Alzheimer Center.

Controls (n = 50) were identified through the Research Registry. Surrogates (n = 50) were their healthy relatives or friends.

Answers in the areas of demographic and occupational history, smoking habits, medical history, dietary intake, and leisure and work activities were recorded. The analysis was based on methods for paired data. Continuous variables were analyzed, focusing on paired differences between self and surrogate responses.

For occupations and exposures, over 80% of the surrogates agreed with the subjects on over 80% of the questions. On smoking history, over 90% of the surrogates agreed with the subjects on over 70% of the questions. On leisure and work activities, over 70% of the surrogates agreed with the subjects on over 50% of the questions. There was less agreement regarding medical history. For continuous variables, most paired t-tests of zero mean difference between self and surrogate responses resulted in nonrejection of this hypothesis. Computed mean differences were not always positive or always negative.

We did not find systematic under- or overreporting by the surrogates of the controls. Therefore, if there are biases in the responses of surrogates of the AD cases in our case-control study, they would not be canceled out by using surrogates for the controls.