Measurement error in total energy intake in the United Kingdom National Diet and Nutrition Survey (2008-2015)

Measurement error in the estimation of energy intake (EI) from self-reported dietary assessment is a common phenomenon, which can lead to implications for epidemiological studies and surveys that aim to determine relationships between diet and health.

The aim of this study was to examine measurement error in estimated EIs within the UK National Diet and Nutrition Survey (NDNS, 2008-2015), to determine the degree of error, if measurement error has changed over the course of data collection periods, and to identify any related associations.

Data from 770 (383 males and 387 females) participants from the UK NDNS recruited between 2008 and 2011 and 2013 and 2015 who had height and weight measurements, ≥3 of 4 diet diary days, and total energy expenditure estimations using the doubly labeled water (DLW) method were used for this secondary analysis study. Data were obtained via the UK Data Service (https://ukdataservice.ac.uk/).

Overall, when compared with DLW total energy expenditure measurements, EI from self-report dietary data was underestimated by 27% [95% confidence interval (CI): 25, 28%], and did not differ significantly in combined age/sex strata between survey years 2008-2011 and survey years 2013-2015 (25%; 95% CI: 23, 27% compared with 28%; 95% CI: 26, 31%, respectively). Likewise, measurement error in reported EI in the majority of age/sex strata generally remained stable. Measurement error in reported EI was associated with body mass index (-0.02 ratio units per kg/m²; 95% CI: -0.02, -0.01 ratio units per kg/m²), mean difference (56 kcal/kg/m²; 95% CI: 47.1, 64.3 kcal/kg/m²), age (mean difference: -8.4 kcal/y; 95% CI: -1.2, -4.8 kcal/y), and sex (mean difference in females: -173 kcal; 95% CI: -252.7, -93.9 kcal) and not associated with ethnicity, education, equivalized income or geography.

We show that although measurement error in estimation of EI is evident within the UK NDNS, it has remained similar. Measurement error was associated with body mass index, age, and sex.