Standardization of dietary intake measurements by nonlinear calibration using short-term reference data.

Statistical analysis of pooled dietary intake data in multicenter and multiethnic studies is often hampered by lack of comparability due to application of different food frequency questionnaires (FFQs). To remove this deficiency, dietary intake measurements should be standardized. This paper presents a standardization procedure based on nonlinear calibration, which aims to approximate the usual intake distribution estimated by reference measurements. The method can be applied in studies with repeated standardized reference measurements that can refer to time periods different from that of the FFQ. It was developed especially for short-term reference assessment methods, such as 24-hour recalls, diet records, and biomarkers. Similar to linear calibration, the proposed method does not change the rankings of subjects in each center or group; therefore, it maintains the within-center validity of the FFQ data. In contrast to linear calibration, the mixture of nonlinearly calibrated intake measurements from different centers or groups corresponds to the mixture of usual intake expected from the reference measurements. This paper illustrates this property of achieving high between-center validity by using macronutrient intake data from the 1995-1996 European Prospective Investigation into Cancer and Nutrition-Potsdam validation study. Here, the proposed method is compared with three linear calibration methods.