Background: Although different methods for the evaluation of energy intake (EI) misreport have been described, it is unclear which one is the most appropriate.
Aim: To assess the performance of these methods in the prevalence of EI misreports and accuracy of nutrient intake estimates.
Methods: Reports of 3,639 adults from the Portuguese National Food, Nutrition and Physical Activity Survey 2015-2016 were classified using univariate (Willett; interquartile range) and multivariate (Goldberg; predicted total energy expenditure [pTEE], testing different standard deviations [SD]) methods. Self-reported intakes were compared to their respective estimates by urinary excretion in a sub-sample of 80. The effect of the exclusion of misreporters on nutrient estimates was assessed by the differences in linear regression coefficients between plausible and total sample.
Results: The highest prevalence of EI misreport was observed using pTEE 1SD (63.9%). Differences in the associations between nutrient self-reported intake and estimated intake using urinary biomarkers were verified with misreporters' exclusion by pTEE 1SD method (β-protein = 0.209; 95% CI = 0.074-0.529; β-potassium = 0.276; 95% CI = 0.060-0.560) and Goldberg 2SD (β-protein = 0.080; 95% CI = 0.025-0.235; β-potassium = 0.106; 95% CI = -0.048-0.246).
Conclusions: Multivariate methods lead to a higher prevalence of misreports and larger differences in nutrient estimates. The application of the pTEE 1SD and Goldberg 2SD methods resulted in more accurate nutrient estimates.
Keywords: Misreport; Portugal; energy intake; national dietary survey; urinary biomarkers.