Mark Hopkins

Ph.D

University of Leeds

Lecturer in Nutritional Physiology

IDATA (Learn more about this study)

IDATA-38

Aug 10, 2020

Errors and uncertainty in the estimation of dietary mis-reporting in studies using doubly-labelled water to estimate energy intake

Assessment of dietary intake in human poses a major challenge. Energy intake is most commonly measured through self-report, which is established to be invalid due to misreporting. It introduces error in dietary assessment and includes both, under- and over-reporting of energy intake (Livingstone and Black, 2003). Although the prevalence of this phenomenon is still not well-established, the most recent review reported that approx. 30% of subjects can be classified as under-reporters and on average, 15% of energy intake is underestimated (Poslusna et al., 2009). Misreporting of food intake may distort conclusions made between diet and health outcomes (Garden et al., 2018) therefore better understanding of this phenomenon is needed.

There is no biomarker of EI per se. The most common proxy biomarker of EI has become the measurement of EE using doubly labelled water (DLW) (Schoeller 1995; Black et al. 1996; Black & Cole 2000; Hill & Davies 2001; Livingstone & Black 2003). However, the DLW technique is not a biomarker of nutritional exposure or of EI it is a biomarker of EE. The DLW method is precise, accurate and non-invasive, but is associated with considerable analytical and isotope cost which limits its applicability to large cohort studies. It can only be used periods of 10-14 days and measured EE is then compared with reported EI to evaluate magnitude of mis-reporting (Castro-Quezada I et al, 2015). Usually, the assumption is made that energy balance over this period is maintained, so expenditure reflects true intake. This assumption itself depends on the assumptions that 1) the measure of EE reflects habitual EE and 2) that measured EE reflects the actual intake of subjects over the course of measurement or that the correlation between intake and expenditure is extremely high. Those assumptions are not necessarily always correct. Another source of error in the estimation of dietary mis-reporting in DLW studies is that the measure of EI rarely covers the full period during which EE was measured. Intake measures range from a single 24-hour recall through a 7 day diet record to a food frequency questionnaire that estimates either the previous month or even year. The 95% confidence intervals over which a 7 day intake can be deemed to relate to EE (assuming a correlation between intake and expenditure of 0.425, and that the measures of intake and expenditure are concurrent) is ±18% (Black 2000). The same confidence interval for a 24 hour recall is ±40% (Livingstone & Black 2003). Thus, DLW may not as secure a biomarker of EI as is commonly believed. Only one study has directly measured mis-reporting of dietary intake and measured EE with DLW (Stubbs et al. 2014).

We are requesting access to the IDATA database in order to collate and analyse individual-level data from studies that include individual daily estimates of energy intake (EI) and estimates daily energy expenditure (EE) using doubly labelled water (DLW). These data will be used to:

1. Examine the relationship between EI:BMR ratios (BMR measured or estimated form predictive equations) and EI: EE (EE measured by a gold standard method – DLW) to establish if cut-offs can be used to adequately detect misreporting phenomenon
2. Establish estimates of measurement error and statistical uncertainty when using different approaches to estimate misreporting in the context of energy balance.
3. Examine if specific variables, including BMI, sex and age, influence the association between EI and EE.

We also aim to collect information on weight change or energy stores in order to:

4. Use the intake:balance method to estimate errors and uncertainty in calculating degree of misreporting using EI:EE estimates.
5. Examine the extent to which change in body weight recues errors and uncertainty in this estimate.