Diet is a major determinant of life course outcomes, yet the accurate measurement of an individual’s dietary intake remains a persistent challenge(1). Intake biomarkers measured in urine and blood using advanced analytical tools, have proven to be a robust and objective method of measuring the intake of particular dietary components(2-3). The “Standardised and Objective Dietary Intake Assessment Tool” (SODIAT)-1 study aimed to determine the effectiveness of using a combination of dietary assessment technologies, to accurately measure an individual’s dietary intake(4).
A randomised controlled crossover trial recruited 33 UK adults (Male; 14, Female;19) with a median age of 43 years (IQR; 29, 56), across two sites; University of Reading and Hammersmith Hospital, Imperial College London. Participants consumed two 4-day controlled diets, one designed to be compliant and one non-compliant with recommended UK dietary guidance. For each 4-day period participants consumed two separate menu plans, on alternating days. Participants collected nine spot urine samples; First Morning Voids on days 1-5 and Bed Time on days 1-4. Fasted capillary blood samples were self-collected by the participants on days 1, 2 and 4 using the OneDraw blood collection system. Urine samples were analysed by Ultra High Performance Liquid Chromatography UHPLC Triple Quadrupole Mass Spectrometry using a panel of previously validated intake biomarkers(2). Dried blood samples were analysed using UHPLC High Resolution Mass Spectrometry for lipid biomarkers of dietary intake. All biomarkers were measured as absolute concentrations. Supervised machine learning (Random Forest) was used to determine the extent of discrimination between study diets. Classification accuracies were calculated as mean values from 1000 randomised resamples.
Supervised models consisting of 83 urinary biomarkers yielded classification accuracies of 0.975 [95% CI; 0.973 - 0.978], the addition of 154 lipid biomarkers from dried blood samples enabled perfect classification; 1.000 [95% CI; 0.999 - 1.000]. Following recursive feature elimination, the model was reduced to a total of 47 biomarkers, with no loss in classification performance. Top ranked features in the optimal model consisted of triacylglycerols (TAG:53_2, TAG:50_2, TAG:49_1) which sufficiently discriminated between dairy and sugar containing components, and urinary markers of the intake of meat (3-Methyl-histidine, L-Anserine), wholegrains (DHPPA-3-Sulfate) and high anti-oxidant containing components (Protocatechuic acid, Hippuric acid).
Biomarkers from spot urine samples and capillary blood samples can provide objective measurements of the intake of dietary components commonly consumed in the UK. Combining urinary and lipid biomarkers into single models, improves performance while extending the coverage of detectable dietary components. These models can serve as a foundation for scalable and objective reporting of dietary intake in free living populations.
- Improving the accuracy of dietary assessment by integration of objective biomarkers: a dual site randomised controlled cross-over trial
- Thomas Wilson - Aberystwyth UniversityDamon Parkington - Cambridge University Hospitals NHS Foundation TrustPaulina Guevara-Domínguez - Cambridge University Hospitals NHS Foundation TrustLaura Lyons - Aberystwyth UniversityJuliet Vickar - Aberystwyth UniversityMichelle Weech - University of ReadingKaterina Petropoulou - Imperial College LondonEka Bobokhidze - University of ReadingJennifer Pugh - Imperial College LondonFrank P-W.Lo - 4Imperial College London, United KingdomRosalind Fallaize - University of ReadingAmanda Jane Lloyd - Aberystwyth UniversityJulie Lovegrove - University of ReadingGary Frost - Imperial College LondonManfred Beckmann - Aberystwyth UniversityAlbert Koulman - Cambridge University Hospitals NHS Foundation Trust
- Proceedings of the Nutrition Society, Vol.85(OCE1), E55
- Cambridge University Press on behalf of The Nutrition Society
- 1
- 991005896087107891
- © The Author(s), 2026.
- Centre for Computational and Systems Medicine
- English
- Abstract
