A bivariate measurement error model for nitrogen and potassium intakes to evaluate the performance of regression calibration in the European Prospective Investigation into Cancer and Nutrition study.
European Journal of Clinical Nutrition 2009 ; 63 Suppl 4: S179-87.
Ferrari P, Roddam A, Fahey MT, Jenab M, Bamia C, Ocke M, Amiano P, Hjartåker A, Biessy C, Rinaldi S, Huybrechts I, Tjønneland A, Dethlefsen C, Niravong M, Clavel-Chapelon F, Linseisen J, Boeing H, Oikonomou E, Orfanos P, Palli D, Santucci De Magistris M, Bueno-de-Mesquita HB, Peeters PH, Parr CL, Braaten T, Dorronsoro M, Berenguer T, Gullberg B, Johansson I, Welch AA, Riboli E, Bingham S, Slimani N
DOI : 10.1038/ejcn.2009.80
PubMed ID : 19888273
Within the European Prospective Investigation into Cancer and Nutrition (EPIC) study, the performance of 24-h dietary recall (24-HDR) measurements as reference measurements in a linear regression calibration model is evaluated critically at the individual (within-centre) and aggregate (between-centre) levels by using unbiased estimates of urinary measurements of nitrogen and potassium intakes.
Between 1995 and 1999, 1072 study subjects (59% women) from 12 EPIC centres volunteered to collect 24-h urine samples. Log-transformed questionnaire, 24-HDR and urinary measurements of nitrogen and potassium intakes were analysed in a multivariate measurement error model to estimate the validity of coefficients and error correlations in self-reported dietary measurements. In parallel, correlations between means of 24-HDR and urinary measurements were computed. Linear regression calibration models were used to estimate the regression dilution (attenuation) factors.
After adjustment for sex, centre, age, body mass index and height, the validity coefficients for 24-HDRs were 0.285 (95% confidence interval: 0.194, 0.367) and 0.371 (0.291, 0.446) for nitrogen and potassium intakes, respectively. The attenuation factors estimated in a linear regression calibration model were 0.368 (0.228, 0.508) for nitrogen and 0.500 (0.361, 0.639) for potassium intakes; only the former was different from the estimate obtained using urinary measurements in the measurement error model. The aggregate-level correlation coefficients between means of urinary and 24-HDR measurements were 0.838 (0.637, 0.932) and 0.756 (0.481, 0.895) for nitrogen and potassium intakes, respectively.
This study suggests that 24-HDRs can be used as reference measurements at the individual and aggregate levels for potassium intake, whereas, for nitrogen intake, good performance is observed for between-centre calibration, but some limitations are apparent at the individual level.