Biochemical markers as additional measurements in studies of the accuracy of dietary questionnaire measurements: conceptual issues.

Studies of the accuracy of questionnaire measurements of habitual dietary intakes of individuals are generally based on a comparison with replicate weighed-food consumption records or 24-h recalls used as reference measurements. The correlation between the questionnaire measurements (Q) and the subjects' true long-term intake (T) of foods or nutrients (rho QT) is then estimated from the observed correlation with mean reference measurements, with correction for attenuating effects due to random errors in the reference measurements themselves. For this approach to be valid, it is crucial that these random errors are uncorrelated between questionnaire and reference measurements and between replicate reference measurements. In practice, however, these two criteria are not always fulfilled and rho QT may be systematically overestimated or underestimated. The advantage of using biochemical markers in dietary assessments is that there is often little reason to doubt that their random errors are truly independent of those of both questionnaire measurements and food-consumption records. Thus, the inclusion of such markers in dietary validity studies can make it more likely that the criteria of independent errors, crucial in validity studies, are met.