Reproducibility and validity of a food frequency questionnaire in a case-control study on breast cancer.

A 110-item food frequency questionnaire was tested among 152 community controls of the Kuopio Breast Cancer Study. They completed the questionnaire twice and kept two 7-day diet records at 3-month intervals. The intraclass correlations for nutrients varied from 0.49 (thiamine with supplements) to 0.81 (lactose), and for foods from 0.52 (poultry) to 0.84 (alcoholic drinks). The Pearson correlations between the first food frequency questionnaire and the 14-day diet records, after adjustment for energy, varied for nutrients from 0.18 (thiamine without supplements) to 0.80 (alcohol), and for foods from 0.30 (inner organs) to 0.90 (coffee). Comparison of quintile classification between the two methods is reasonably accurate when observed restrictions concerning some nutrients and foods are taken into account. A low association (r = 0.12) was observed between toenail selenium and dietary selenium intake, indicating the difficulty of estimating selenium intake in the Finnish diet.