Retrospective versus original information on diet: implications for epidemiological studies.

The aim of this study was to evaluate the use of retrospective information on diet in relation to the information originally reported. Hospital controls (n = 131) who participated in a case-control study on diet and pancreatic cancer in Stockholm, Sweden, answered a second postal questionnaire four years later about their current and previous food consumption. The food consumption was dichotomized for the purpose of the analysis. In relation to the original information, the use of retrospective information overestimated previous food consumption among subjects with increased intake, while retrospective information underestimated previous food consumption among subjects with decreased intake. These tendencies were slightly more pronounced in the age group 60-79 years than in the age group 40-59 years. For subjects with unchanged food consumption, there was a high agreement between original and retrospective information. Considerable differential misclassification could be introduced in an epidemiological study if cases and non-cases differ with regard to changes in food consumption.