The study population: general characteristics and potential confounding factors

Abstract

Objective: To test the representativity of the study population and to identify potential selection bias.

Design: Non-response analysis and comparison with other representative samples from the Dutch population. Description of physical and clinical biochemical variables as well as health and lifestyle characteristics among low vitamin B6 groups and reference groups.

Results and conclusions: The non-response rate for the food frequency questionnaire was 54%; subsequently, non-response for the detail study was 42%. A greater proportion of respondents to the FFQ and of participants in the reference group of the study followed a special dietary regimen or a therapeutic diet or used dietary supplements compared with non-respondents, indicating that we may have selected a somewhat more health conscious reference group. There were no differences in vitamin B6 intake (absolute or per g protein) between participants and the total group of non-respondents for the detail study. Selection of a low B6 group did not result in important bias regarding physical, biochemical or general health and lifestyle characteristics. Exceptions were a difference in the proportion of dieters (younger women) and supplement users (older women) and a difference in creatinine clearance and supplement use between reference and low B6 groups. The differences found, however, were small. Nonetheless, these factors have to be taken into account when analysing and interpreting the results.