On longitudinal versus cross-sectional studies of obesity: possible artefacts.

Within the framework of a large-scale research project on overweight, relationships between body mass index (BMI; weight/height) and two types of eating behaviour (restrained eating and emotional eating) were determined. All results obtained on one assessment date in a longitudinal study within this project were found to differ significantly from those of a cross-sectional study. The present study aimed to replicate these results on another assessment date of the same longitudinal study, and in another cross-sectional sample. The two types of studies were found to differ systematically. This difference may well be attributable to greater prevalence in the longitudinal study of the response set, acquiescence in the women of normal weight, and social desirability in the overweight women. These response biases may have resulted from sensitization of the subject to the questionnaires, and also from the intensive contact between subject and investigator due to repeated measurements. Thus in further studies use of independent response-set marker scales is recommended to improve identification and control of response sets.