Statistical issues related to semiquantitative food-frequency questionnaires.

Semiquantitative food-frequency (SQFF) questionnaires have been used by many epidemiological studies to collect dietary data. These methods are especially attractive for use in large-scale studies because they are often self-administered and/or completed with minimal time or effort. This paper discusses several statistical issues related to the use of dietary data collected by frequency methods. In particular, the paper addresses 1) whether SQFF methods are well-validated dietary methods as suggested by many investigators and 2) whether data based on SQFF methods can be used to compare mean dietary intakes among different populations and to estimate the distribution of usual nutrient intakes in a national probability population. Data from several validation studies are used to evaluate the impact on statistical analyses of measurement error associated with certain SQFF methods. These examples help to illustrate the potential problems and clarify the interpretation of the results. In addition, concerns about the misuse of dietary data derived from SQFF questionnaires are also discussed.