Background – The majority of epidemiological studies examining associations between glycemic load (GL) and risk of chronic diseases have used validated food frequency questionnaires with carbohydrate correlation coefficients ranging from 0.4 to 0.8. Few have reported the degree of agreement between the questionnaire and reference intake.
Objectives – To test the validity of a GL questionnaire (GLQ) by comparison with a detailed diet history.
Design –54 women aged 42 to 82 years were recruited from a cohort of 511 participants in the Longitudinal Assessment of Ageing in Women (LAW). Carbohydrate intake was assessed by a specially-designed GLQ; GL [carbohydrate (g) x glycemic index (%)] values were summed to provide the average daily GL. Data were validated against a diet history.
Outcomes – Mean + SEM intakes from the diet history were 6% higher than those from the GLQ for carbohydrate (216 + 6 versus 203 + 8 g/day, P<0.05) and the GL (110 + 4 versus 103 + 4, P=0.1), respectively. There were significant correlations between methods for carbohydrate (r=0.54, P<0.01) and GL (r=0.57, P<0.01). 95% limits of agreement determined by Bland Altman plots ranged from -111 to 83.7 g for carbohydrate, with almost half the subjects recording a difference of + 40 g; and -60.0 to 46.6 for GL, with a third of subjects recording a difference of + 25 units or more.
Conclusion – Our GLQ had acceptable validity in terms of correlation with the dietary history. From a clinical perspective however, substantial error existed in estimation of individual carbohydrate and GL intakes. We suggest that studies using food frequency questionnaires to estimate GL state limits of agreement instead of or as well as correlations when discussing validity. Failure to accurately assess carbohydrate intake could explain some of the discrepancies in results of dietary studies investigating associations between GL and the development of chronic diseases in individuals.