Reproducibility and relative validity of dietary glycaemic index and glycaemic load assessed by the food-frequency questionnaire used in the Dutch cohorts of the European Prospective Investigation into Cancer and Nutrition.

Limited information is available on the reproducibility and validity of dietary glycaemic index (GI) and glycaemic load (GL) estimated by habitual diet assessment methods such as FFQ, including the FFQ used in the Dutch cohorts of the European Prospective Investigation into Cancer and Nutrition study. To examine the reproducibility and relative validity of GI and GL, we used data from 121 Dutch men and women aged 23-72 years. They completed the FFQ three times at intervals of 6 months and twelve 24-h dietary recalls (24HDR) monthly during 1991-2. GI and GL were calculated using published values. Intra-class correlation coefficients of the three repeated FFQ were 0.78 for GI and 0.74 for GL. Pearson correlation coefficients between the first FFQ and the weighted average of the 24HDR were 0.63 for both GI and GL. Weighted kappa values between the first FFQ and the average of the 24HDR (in quintiles) were 0.40 for GI and 0.41 for GL. Bland-Altman plots showed a proportional bias in GI (beta = 0.46), but not in GL (beta = 0.06). In conclusion, this FFQ can be used in epidemiological studies to investigate the relationship of GI and GL with disease risks, but the proportional bias should be taken into account when using this FFQ to assess the absolute GI values.