Biomarkers of food intake: The key to precision nutrition

By Dr. Liji Thomas, MDReviewed by Benedette Cuffari, M.Sc.Jul 5 2024

The crucial role of nutrition in health necessitates the development of dietary assessment tools capable of accurately assessing causal relationships with various health-related consequences.

A recent study published in Nature Metabolism examines the potential utility of biomarkers of food intake (BFIs) on objective and accurate dietary assessments.

Study: Towards nutrition with precision: unlocking biomarkers as dietary assessment tools. Image Credit: Gorodenkoff / Shutterstock.com

What are BFIs?

BFIs are often used to evaluate dietary adherence in nutritional intervention and meal studies, assess the extent of misreporting, as well as validate epidemiologically-derived associations between food and disease risk. While food frequency questionnaires (FFQs) and dietary recalls are also useful assessment tools, their subjective nature can lead to biased reporting and poor compliance.

A BFI is a metabolite of ingested food and is defined as a measure of the consumption of specific food groups, foods, or food components. BFIs can be ranked based on their robustness, in which minimal interference from a varied dietary background affects the use of the BFI in research.

Reliability in BFIs implies that this marker is in qualitative and/or quantitative agreement with other biomarkers or dietary instruments. Plausibility depends on the specificity and chemical relationship of the metabolite to the nutrient in question, which limits the risk of misclassification due to other factors.

Biologic variability for BFIs depends on absorption, distribution, metabolism and elimination (ADME) of the food, as well as enzyme/transporter concentrations, genetic variation, and gut microbial metabolism. Importantly, this characteristic has not been reported for most BFIs.

Intra-class correlation (ICC) also reflects variability within a population or group in response to different factors. When ICC is low, the BFI may be associated with wrong sampling time, low frequency of consumption, or gross variation in the response over time within and between individuals and populations.

About the study

Following validated BFI reviews that met appropriate guidelines and methodologies, the researchers performed two systematic searches for experimental and observational studies. Thereafter, a four-level classification system was used to rank reported BFIs based on their robustness, reliability, and plausibility.

If all criteria were met, the BFI was classified as belonging to utility level one. At level two, the candidate BFI is plausible and robust but not known to be reliable. Level tjree BFIs are plausible but lack robustness and reliability, whereas level four BFIs have not been reported for the foods.

If these criteria are met, additional characteristics including time kinetics, which refers to the sampling window or time period for the BFI to be sampled after nutrient ingestion, analytical performance, and reproducibility are also assessed.

Level one and two BFIs

Utility level one or validated urine BFIs were found for total meat, total fish, chicken, fatty fish, total fruit, citrus fruit, banana, whole-grain wheat or rye, alcohol, beer, wine, and coffee. Level one blood BFIs exist for fatty fish, whole grain wheat and rye, citrus, and alcohol.

Level two candidate BFIs in urine include total plant foods and various plant foods including legumes and vegetables, dairy, and some specific fruits and vegetables. Blood BFIs at level two exist for plant foods, dairy products, some meat, and some non-alcoholic drinks; however, these BFIs comprise fewer foods with less validation.

Identification and validation of BFIs

The discovery and validation of BFIs requires discovery studies, followed by confirmation and prediction studies. Meal studies identify plausible BFIs; however, these may not be specific, unless other foods contain very low levels of the marker or are rarely consumed.

For example, betaine is present at high levels in oranges and is used to detect orange or citrus consumption, despite being found in many other foods at low levels. However, discovery studies may be very small or poorly representative.  

Observational studies can be used to identify associations between blood or urine metabolites and diet but are subject to confounding by lifestyle factors. When two types of foods are frequently consumed together, like fish and green tea in Japan, confounding occurs with the BFI of fish, as trimethylamine oxide (TMAO) can also be associated with green tea, thus making these foods not suitable for BFI discovery.

Endogenous metabolites are poorly robust BFIs, as they are produced both endogenously and from exogenous foods. These metabolites are also associated with significant variations with inter-individual genetic and microbial differences.

Prediction studies use models based on randomized controlled trials to identify the consumption of a given food. This approach outperforms correlation studies by identifying BFIs that may predict intake, but depends on the sampling window for accuracy.  

Several databases, such as Massbank, METLIN Gen2, mzCloud (Thermo Scientific), mzCloud Advanced, Mass Spectral Database, and HMDB, are available for metabolite search. The Global Natural Products Social Molecular Networking initiative is leading efforts to interconnect these databases and compare unknown compounds against known spectra, such as by the Global Natural Products Social Mass Spectrometry Search Tool (MASST).

BFI applications

BFI selection depends on the aim of the study. Qualitative BFIs are adequate for identifying non-compliance or conducting per-protocol analyses. Conversely, a combination of signature BFIs provides greater specificity and may even identify a whole meal or dietary pattern.

A stepwise approach could help identify actual consumers of a food of interest before assessing the amount consumed in a second step, allowing even less robust BFIs to play a role in these types of studies.

Habitual dietary patterns can be captured by multiple sampling, with the frequency and number dependent on the sampling window and frequency of consumption. Optimal sampling methods identified in the current study include spot urine samples such as first morning void or overnight cumulative samples, dried urine sports, vacuum tube stored samples, dried spot samples, and microsampling.

Remote sampling increases the number of possible participants and ability to monitor dietary patterns and changes over time. These methods can also improve epidemiological studies aiming to identify correlations between diet and disease risk.

Refining sampling and analytic methods may also improve the precision of nutrition research and establish trusted associations between dietary intakes and health consequences.

Future development

Future studies are needed to validate the development of single and multimarker BFI using different samples, food groups, and diets, as well as cooked and processed foods. Quantitative BFIs should also be characterized by dose-response studies, whereas BFI combinations should be established to predict and classify intake and dietary patterns.

Precision nutrition is of particular importance to curb obesity and cardiometabolic diseases for which a one-diet-fits-all approach doesn’t seem to work due to the highly varied individual response to diet. Personalized dietary interventions are good drivers of behaviour change, shown to improve diet quality.”