Calibration of citrus intake assessed by food frequency questionnaires using urinary proline betaine in an observational study setting.

BACKGROUND: Whether observational study can be employed to establish calibration equations for self-reported dietary intake using food biomarkers is unknown.

OBJECTIVE: This study aims to demonstrate the feasibility of obtaining calibration equations based on food biomarkers and 7-day diet records (7DDRs) to correct measurement errors of food frequency questionnaires (FFQ) in an observational study setting.

METHODS: The study population consisted of 669 male and 749 female from the Women and Men's Lifestyle Validation Study. In the training set, the biomarker-predicted intake derived by regressing 7DDR-assessed intake on urinary proline betaine level was regressed on the FFQ-assessed intake to obtain the calibration equations. The regression coefficients were applied to the test set to calculate the calibrated FFQ intake. We examined total citrus as well as individual citrus fruits/beverages.

RESULTS: Urinary proline betaine was moderately correlated with orange juice intake [Pearson correlation (r): 0.53 for 7DDR and 0.48 for FFQ] but only weakly correlated with intakes of orange (r: 0.12 for 7DDR and 0.15 for FFQ) and grapefruit (r: 0.14 for 7DDR and 0.09 for FFQ). The FFQ-assessed citrus intake was systematically higher than the 7DDR-assessed intake, and after calibrations the mean calibrated FFQ measurements were almost identical to 7DDR assessments. In the test set, the mean intake levels from 7DDRs, FFQ, and calibrated FFQ were 62.5, 75.3, 63.2 g/d for total citrus, 41.6, 42.5, 41.9 g/d for orange juice, 11.8, 24.3, 12.3 g/d for oranges, and 8.3, 9.3, 8.6 g/d for grapefruit. We observed larger differences between calibrated FFQ and 7DDR assessments at the extreme ends of intake, although on average good agreements were observed for all citrus except grapefruit.

CONCLUSION: Our two-step calibration approach has the potential to be adapted to correct systematic measurement error for other foods/nutrients with established food biomarkers in a cost-effective way.