Validity and reliability of surrogate information for controls in a case-control study on Alzheimer's disease

Abstract

This study is aimed to investigate proxy respondent's information usefulness in retrospective studies by comparing information obtained with a questionnaire (with a total of 171 items) from controls and their proxy respondents in a case-control study on Alzheimer's disease. Kappa indices and intraclass correlation coefficients were calculated to assess reliability, and bias factor and mean differences were calculated to assess validity. Proxy respondent's personal features (gender, age and relationship with the index subject) were also studied. Kappa indices and intraclass correlation coefficients were in general good or very good (>0.6), specially regarding control's personal and family data (ranges 0.45-1 for Kappa and 0.86–0.99 for intraclass correlation coefficient) and occupational exposures (range for Kappa 0.48–1). No systematic biases were found (range for bias factor 0.65–4.12 and range for mean differences −1.81–1.30, none of them statistically significant). Proxy respondent's individual features were not found to systematically affect reliability. The use of surrogate information for controls in etiologic case-control studies of Alzheimer's disease may be useful without unacceptable loss of information or systematic biases.