Profile comparability
Results depend on collection, stabilisation, extraction, technology, database, pipeline and reference population. Profiles produced through different processes are not automatically equivalent [1–4].

Targeted qPCR
Measures selected microorganisms or genes and may support validated panels or longitudinal monitoring; it is not equivalent to profiling the entire community and does not detect what lies outside the panel [1,5].

Validity and utility
Analytical validity concerns the reliability of the measurement; clinical validity and utility concern the meaning of the result and the benefit of the decision in the intended context. These three levels are not interchangeable [1,13,14].

Biomarkers and outcomes
A change in a taxon, diversity or another biological signal does not automatically amount to a benefit. The outcome must be defined before the analysis and must be relevant to the individual or to the stated wellness purpose [8–11].

Longitudinal programme
Repeating a test adds value only when the change exceeds technical variability, the interval is biologically plausible and the result changes a predefined action [1,3,5].

A microbiome test produces a measurement; a nutraceutical recommendation requires a further step. The sample, analytical method and pipeline define what is observed, while its meaning for the individual depends on the reference population, context and interpretative rules. Even when an association is robust, it remains to be demonstrated which intervention changes a relevant outcome.

This Insight examines the entire chain, from profile comparability to endpoint selection. It distinguishes sequencing from targeted qPCR, separates biomarkers from benefit, critically assesses personalisation studies and defines the boundaries between research, wellness and clinical use. The aim is not to diminish the microbiome's potential, but to clarify which evidence and rules allow it to become a credible and verifiable nutraceutical programme.

The microbiome lends itself to an immediate commercial promise. The gut ecosystem varies between individuals, is linked to diet and metabolism, and can be observed through a sample collected at home. The next step seems natural: from the profile to an 'imbalance', then to the choice of foods, prebiotics, probiotics or other supplements.

The underlying biology is real, but it does not resolve the entire process. Associations between diet, the microbiome and metabolic phenotypes are numerous and partly reproducible. Three questions remain separate: which microorganisms are detected in the sample; what the data mean for that individual; and which intervention produces a relevant benefit. The first concerns analytical measurement. The second requires clinical validity; the third, clinical utility or at least a predefined and measurable wellness outcome.

This gap is what guides the international consensus published in 2025. The document recognises the potential of microbiome tests, but considers the evidence still insufficient to recommend their routine use in clinical practice and discourages providers from including post-test therapeutic advice in reports [1]. It does not mark the end of the sector: it clarifies which methodological and professional conditions are needed for it to mature.

The test does not read a stable object. The observed profile incorporates what happens before and after the analytical phase: collection timing and procedure, transport, temperature, stabilising agent, DNA extraction, amplified region, sequencing depth, database and bioinformatics pipeline. Medication, antibiotics, recent diet, geography, age and intestinal transit add to this technical variability.

In the Microbiome Quality Control project, 15 laboratories received blinded sample sets; the data were then analysed using nine bioinformatics protocols. Sample type and origin explained much of the variability, but extraction, the sample-handling environment and computational method also mattered [2]. The result explains why reproducibility must be built along the entire chain, not merely into the final screen of a report.

A check even closer to the market came in 2026. A homogeneous faecal material developed by NIST was sent, in triplicate, to seven direct-to-consumer services. The profiles showed substantial discrepancies between providers and, in some cases, between replicates; variability between services was of the same order as that observed between different donors. The study measures precision and comparability rather than establishing which service returns the true composition, but it demonstrates how risky it is to turn methodological differences into biological differences [3].

STORMS organises this complexity into a checklist to make microbiome study reporting more complete and comparable [4]. It improves transparency, but does not automatically validate the test or the commercial recommendation.

The same applies to popular concepts such as diversity and 'dysbiosis'. Higher diversity is not always synonymous with health; expected composition varies with population, method and conditions. The 2025 consensus considers the evidence insufficient to include a dysbiosis index or rigid relative-abundance ranges in reports. If a provider uses a proprietary reference, it must disclose the cohort, method and confounders. Without this context, even a very precise percentile communicates a certainty the data do not possess [1,3].

The technologies do not observe the same object. 16S sequencing mainly describes bacterial composition, with resolution limits and biases linked to the amplified region and pipeline. Shotgun metagenomics can offer greater taxonomic and functional resolution, with greater complexity; metatranscriptomics and metabolomics add other dimensions. None of these measurements, on its own, captures 'the microbiome' in its entirety.

qPCR, by contrast, answers a targeted question. It is suitable for detecting the presence or quantity of selected microorganisms or genes, monitoring a validated panel or testing a hypothesis in a trial. The international consensus notes that multiplex PCR and culture can be useful, but are not equivalent to profiling the entire community and do not constitute a proxy for it [1]. What is not included in the panel remains outside the observation; a set of targets does not thereby become an exhaustive map of the ecosystem.

MIQE 2.0 requires documentation of the sample, assay design and optimisation, controls, efficiency, limits, dynamic range, normalisation, analysis and reporting [5]. This transparency supports qPCR quality and reproducibility, but does not replace clinical validation. In a nutraceutical programme, speed and instrument precision are not enough: if the apparent variation arises from preparation or quantification, the feedback given to the user remains noise, even when it is well presented.

Research on personalised nutrition shows that different individuals can respond differently to the same food. Zeevi and colleagues integrated continuous glucose monitoring, clinical parameters, habits, activity and the microbiome to predict postprandial glycaemic responses; the predictions were tested in an independent cohort of 100 people and used in a short dietary intervention [6]. PREDICT 1 then linked habitual diet, the microbiome and numerous cardiometabolic markers in 1,098 deeply phenotyped participants [7].

The common point is not that the microbiome alone is enough for personalisation. PREDICT 1 describes associations in an observational cohort; Zeevi's study assesses a model built with many features. The two studies have different designs, but both place the microbiome within a multivariable framework that includes phenotype, behaviour and outcomes. They do not demonstrate that an isolated gut test can automatically select a supplement.

More recent trials show the same incomplete maturity. In 121 people with irritable bowel syndrome, a personalised diet assisted by microbiome data and a low-FODMAP diet both reduced symptoms over six weeks; the between-group difference in the primary outcome was not significant (p=0.29). The microbiome analysis was performed by a company in which two authors were shareholders and another two were employees: a disclosed conflict that does not invalidate the result, but strengthens the need for independent replication [8].

In the PROMOTe trial, 36 pairs of older twins — 72 people — received a prebiotic or placebo for 12 weeks; both groups also undertook resistance exercise and branched-chain amino acid supplementation. The prebiotic changed microbial composition and showed a secondary cognitive signal, but did not improve the primary outcome of chair-rise time. The article should be read in its updated version incorporating the 2025 correction [9,10].

A 2024 proof-of-concept study reported metabolic improvements with a microbiota-based nutritional programme. The study was open-label, lasted 90 days and included 30 participants, 15 per group; the programme and algorithm were proprietary and most authors belonged to the company developing them [11]. It is a signal worth investigating further, not sufficient evidence to generalise efficacy or commercial transferability.

Increasing an expected taxon does not demonstrate that a supplement has improved a symptom or reduced a risk. The opposite can also occur: a benefit emerges without an interpretable change in the panel. Reading the result therefore requires three distinct levels: analytical response, biological change and a relevant outcome.

An intervention trial should define in advance which outcome matters: a validated symptom measure, adherence, a metabolic marker, quality of life or another appropriate endpoint. The microbiome may be an exploratory endpoint, a hypothesised mediator or a stratification criterion; its role must be stated. If all markers are explored after the result, the probability of finding chance associations and turning them into a causal narrative increases.

Follow-up has value only when it makes a decision better informed. Before repeating the test, it is necessary to know whether technical variability is lower than the expected change, whether the interval between measurements is biologically plausible and which action will change as a result. Without these conditions, retesting adds data but does not guide the programme.

The one-shot report is the most exposed model: it translates relative abundances into absolute judgements and links the result to a catalogue of supplements. When information, interpretation and sales coincide, it becomes difficult to tell whether the recommendation follows the evidence or product availability. The differences already observed between direct-to-consumer services make this overlap even riskier [1,3].

A more solid offering first declares its intended purpose. In research, the test can stratify participants and measure predefined targets. In wellness, it may support awareness and monitoring if its purpose, language and actions remain non-medical. In clinical use, analytical validity, clinical validity and utility must be demonstrated; professional governance must be consistent with the stated purpose.

The most defensible opportunities are often less conspicuous: standardised collection and biobanks, targeted panels for trials, longitudinal monitoring, datasets linked to outcomes, quality and interoperability tools, follow-up studies and real-world evidence. In these services, data gain value because they remain traceable and comparable over time.

For a nutraceutical company, the starting point should not be the number of taxa to display. It is the decision to improve: which action will change thanks to the programme, and against which outcome will it be assessed? Defining this relationship reduces promotional risk and strengthens the product proposition.

In the European Union, nutrition and health claims in commercial communications for foods are governed by Regulation (EC) No 1924/2006. Claims must be supported by generally accepted scientific evidence and, for health claims, used in compliance with the applicable authorisation regime [12]. Measuring the microbiome does not automatically confer a health claim on the supplement.

When a test is intended by its manufacturer to provide, for medical purposes, information on a physiological or pathological state, its qualification must be assessed within the scope of Regulation (EU) 2017/746. The intended purpose and actual claims matter more than the commercial label; calling an output 'wellness' does not neutralise communication that leads the user to diagnose or treat a condition [13].

In the United States, the FDA includes microbiome tests among direct-to-consumer tests and distinguishes analytical validity, clinical validity and claims. Its information page also notes that users should not make dietary or healthcare decisions based solely on the result without consulting a professional. This is a US reference, not a European rule, but it makes the same communication risk apparent [14].

There is no need to bring every offering into the medical sphere. Evidence, purpose and commercial language need to remain consistent, ensuring that communication does not attribute to the test a decision-making power unsupported by the evidence. A transparent programme states what it measures and what lies outside its scope, the decision it intends to support and who takes responsibility for it.

Before integrating a test into a nutraceutical offering, it is worth checking the complete chain. Is the sample stable under real conditions? Is the method reproducible between batches and sites? Is the reference compatible with the target population? Has the algorithm been tested on independent data? Can the recommendation be separated from the commercial inventory? Has the outcome been predefined and measured against an appropriate comparator?

Communication is part of the quality system. The report must set out uncertainty, limitations and confounders, allowing the user to distinguish an observed association, a general recommendation and a clinical decision. Where an algorithmic model is involved, its version, inputs, changes, monitoring and human oversight must remain traceable.

In the long term, the most useful metric will be the proportion of recommendations that lead to a measurable and reproducible change without generating inappropriate decisions. Building this evidence is difficult, but creates a stronger competitive barrier than the number of kits sold.

The microbiome can support personalised nutrition, provided it is treated as a measurement to be governed rather than a shortcut from the test tube to the shelf. Value depends on comparable samples, disclosed methods, targets consistent with the purpose, verified interpretative rules, measurable outcomes and proportionate communication. This is the chain that makes the programme credible.

The most interesting commercial phase is also the most demanding: building the market while building the evidence. Longitudinal programmes and pragmatic trials allow laboratories, nutraceutical companies and platforms to connect recommendations, measurements and planned actions. Personalisation becomes credible when data change a decision in a justified way and can also indicate when they should not change it.