Abstract
Background: Specific metabolites detected in human biofluid after the ingestion of food have been proposed as objective markers of dietary intake. Objective markers are desirable, as they have the potential to address limitations within traditional dietary assessment methods that rely on self-reporting. To date, hundreds of metabolites related to single foods or food groups in blood and urine have been identified, however, beyond identification, there has been relatively limited work evaluating the relationships between proposed biomarkers and diet.
Objective: To test a range of proposed metabolite biomarkers of food intake reflecting major food groups to then compare biomarker values in fasted whole blood, 24-hour urine, dried blood spots and plasma with known and self-reported dietary intakes.
Methods: A systematic review and meta-analysis was conducted to assess 23 proposed metabolite biomarkers of food intake for potential dose response relationships with their associated foods and food groups and comment on their suitability for real-world application. Three dietary intervention studies were then conducted. The first was a randomised crossover feeding trial of three interventions each including a day of known foods consumed under observation. Each feeding day was of three food groups (e.g. whole grains, dairy, fish) with optional snacks (e.g. fruit, chicken, legumes) but entirely absent of three other food groups (e.g. meat, vegetables, nuts and seeds). Fasted whole blood and 24-hour urine samples were analysed by liquid chromatography mass spectrometry (LC-MS) to detect previously proposed biomarkers of food intake. The second and third studies were randomised parallel trials of whole grain, healthy fats and high fibre groceries provided to participants for 12 weeks. Biological samples were collected in the form of dried blood spots or plasma pre- and post-intervention and analysed by gas chromatography mass spectrometry (GC-MS) to detect previously proposed biomarkers of food intakes. A wide range of statistical methods was applied to the three studies, including correlation coefficients (Pearson’s and Spearman’s rank), linear regression, generalised linear modelling and Cohen’s kappa statistics of quartiles of agreements.
Results: The systematic review (Chapter 1) identified 113 publications of dietary interventions that reported on 23 proposed biomarkers of food intake and their corresponding food groups. The most promising biomarkers exhibiting dose-repose relationships with food dose were for whole grain, soy foods and chocolate. The acute feeding intervention study (Chapter 2) and the 12 weeks parallel study of whole grain foods (Chapter 3) identified relationships between all 11 food groups assessed and at least one biomarker of food intake. In contrast, the 12-week parallel study of health fats and high fibre foods (Chapter 4) identified associations between eight of the foods and food groups and metabolites, with no biomarkers positively related to vegetables, chocolate and dairy. One biomarker, anserine, showed associations with the intake of animal meats in all three dietary interventions (GLM rate ratio (95% confidence interval) 0.21(0.05, 0.37); linear regression coefficient (95% CI) 1.45 (0.00, 2.90) and 10.03 (1.62, 18.44)). All three intervention studies showed inverse relationships between alkylresorcinols and whole grain intake, except AR C19:0 detected in dried blood spots which showed a modest Spearman’s correlation (r 0.22) when compared with baseline whole grain intake (Chapter 3). Of metabolites detected only through LC-MS (Chapter 2), urinary trimethylamine N-oxide was associated with fish intake (GLM coefficient (95% CI) 67.13 (41.98, 92.29)) and choline from whole blood was associated with egg intake (GLM rate ratio (95% CI) 0.14 (0.06, 0.21)). Of the metabolites detected primarily in GC-MS (Chapters 3 and 4), docosahexaenoic acid showed a relationship with fish intake (r 0.13), pentadecanoic acid (C15:0) with dairy intake (linear regression coefficient (95% CI) 84.20 (14.15, 154.25)), and linoleic acid with fruit intake (r 0.16). Other biomarkers, such as 1-methylhistidine emerged as a marker for red meat intake only when assessing metabolites against self-reported long-term intake (r 0.13 to 0.16) (Chapters 3 and 4), while 3-methylhistidine showed as association with chicken intake only when compared with acute known dietary intake (GLM rate ratio (95% CI) 0.26 (0.11, 0.41) (Chapter 2). Among relatively well-researched biomarkers of fruit and vegetables, s-methylcysteine in blood and urine was associated with vegetable intake and fructose in blood was associated with fruit intake, while sulforaphane was associated with vegetable intake only when compared with intake of foods consumed under observation (Chapter 2). Lastly, among previously well-established markers of intake, 24 hour urinary excretion of proline betaine and sodium did not show associations with known dietary intake in our findings (Chapter 2).
Conclusion: Overall findings from this thesis indicate that while associations between proposed biomarkers of food intake and food groups exist, the relationships are modest, and not reproducible across the three studies undertaken. Some of the inconsistencies observed between studies are due to different analytical assessment methods, different biospecimens as substrates, and different dosing amounts of each food or food group. While the work of this thesis does move the biomarker of food intake field incrementally forward, at present, biomarkers of food intake cannot replace traditional dietary assessment methods. Further research to develop standardised assessment methods are required, as well as to better understand within-individual interactions to better understand how the food consumed corresponds to the circulating biomarker concentration.