Over the past decade, the human microbiome has become one of the most intensively studied areas in biomedical research. Advances in sequencing technologies and bioinformatics have enabled researchers to catalogue microbial communities across the body and link them to a growing list of conditions, from infectious diseases and inflammatory bowel disease to cancer, diabetes, and neurodegeneration.
We still do not have standardized, certified tests or reference datasets. We also lack pre- and post-therapy microbiome testing in clinical trials, which would be necessary to prove causality.
—Ingo Autenrieth, University of Heidelberg
However, despite this surge in research activity, microbiome analysis remains largely absent from routine clinical practice. Outside a handful of niche applications, clinicians rarely use microbiome data to guide diagnosis, prognosis, or treatment decisions. One fundamental reason, according to a new review published in Trends in Microbiology, is that medicine still lacks a clear, clinically useful definition of what constitutes a “healthy” versus “dysbiotic” microbiome.
Lead author Ingo Autenrieth told DDN that the challenge is structural, “We still do not have standardized, certified tests or reference datasets. We also lack pre- and post-therapy microbiome testing in clinical trials, which would be necessary to prove causality.”
Jessica Schneider, Chief Science Officer at Corundum Systems Biology, agreed, “The primary hurdle is that it is still yet to be proven actionable in clinical practice. While research has provided a wealth of data, this tends to stop short of prediction.
So, while a patient may have their microbiome profiled, there’s currently no standardized protocol or specific intervention that can be prescribed based on that data. In addition, current profiling methods still lack the specificity needed to provide a clear disease diagnosis.”
What does a “healthy” microbiome really mean?
While the field has been highly successful at identifying associations between microbial composition and disease, it has struggled to translate these findings into tools that work at the level of individual patients. In many cases, microbiome studies rely on group-level comparisons that reveal statistically significant differences between cohorts but offer little guidance on how to interpret a single patient’s microbiome profile in a clinical setting.
Part of the challenge lies in the extraordinary variability of the human microbiome. Microbial communities differ widely between individuals and can fluctuate over time in response to diet, medication, age, and environment. As a result, defining a single “normal” microbiome has proven elusive. What appears dysbiotic in one context may be benign — or even beneficial — in another.
Autenrieth argues that a universal definition of a healthy microbiome is unlikely. “Every individual is their own reference,” he said, “so longitudinal testing could reveal what is ‘normal’ for that person. If large cohorts were studied with certified tests and databases, then general reference indices might eventually become clinically useful.”
This variability has been compounded by a lack of standardization across microbiome studies. Schneider added, “Significant barriers include inconsistencies across workflows, specifically interlaboratory variability. Differences in how operators extract samples, prepare libraries, and sequence data results in inconsistencies in biological data production. Additionally, there’s a lack of consensus in the field on the precise definition of strains, which is compounded by a deficit of harmonized reference genome databases.”
Furthermore, many analyses still rely on genus- or species-level resolution, which may obscure functionally important differences between microbial strains. Adrian Egli, co-author of the study, told DDN, “With 16S sequencing, you only look at a single gene. You can’t tell what a bacterium is actually capable of — whether it carries toxin genes or antibiotic resistance genes. To get that information, you need whole-genome shotgun metagenomics, which is much more expensive.”
Translating research to clinical use
To move microbiome research closer to clinical utility, there needs to be a shift toward standardized, strain-level metagenomic analysis. The authors argue that strain-level resolution is often essential for distinguishing pathogenic from commensal microbes, understanding functional capacity, and identifying clinically relevant microbial signatures. As a result, strain-level data could determine whether a microbe is harmless or harmful — and therefore whether an intervention is needed.
In addition to improved resolution, there is an urgent need for large, well-curated reference datasets that capture microbiome variation across populations, disease states, and clinical contexts. Such datasets could enable the development of clinically meaningful indices that go beyond descriptive measures of diversity or abundance. Instead, these indices would aim to quantify microbiome states in ways that are interpretable and actionable for clinicians, helping to distinguish eubiosis from dysbiosis at the level of the individual patient.
The test would need to be clinically actionable, rather than just revealing microbial imbalance. It must either enable diagnosis, guide the clinician to prescribe a beneficial therapy, or lead to evidence-based recommendations for diet, lifestyle and preventative changes.
—Jessica Schneider, Corundum Systems Biology
Egli also emphasized the importance of aligning microbiome analysis with clinically relevant questions. Rather than asking whether a microbiome differs between healthy and diseased populations, future studies may need to focus on whether microbiome features can predict outcomes such as treatment response, disease progression, or risk of infection. This shift would bring microbiome research closer to the standards applied to other clinical diagnostics.
Egli said, “What medicine really needs is actionable diagnostics. A test where the doctor can actually change treatment based on the microbiome result. To prove that works, you need prospective, randomized controlled trials — and we’re not there yet.”
Schneider added, “The test would need to be clinically actionable, rather than just revealing microbial imbalance. It must either enable diagnosis, guide the clinician to prescribe a beneficial therapy, or lead to evidence-based recommendations for diet, lifestyle and preventative changes.”
What could clinical microbiology look like?
If clinical microbiology were integrated with modern metagenomics, microbiome data could improve diagnostics and targeted therapies. For example, it could support stratifying patients for targeted microbiome interventions, tracking disease-associated microbial changes over time, and identifying microbial factors that influence drug efficacy and toxicity.
“One clear use case is recurrent Clostridioides difficile infection. There’s an inverse relationship between microbiome diversity and recurrence risk, so measuring diversity can help guide clinical decisions,” said Egli. He also noted that, in cancer, certain bacteria can suppress or enhance immune responses. “There’s growing interest in using microbiome diagnostics to help predict or modulate treatment response.”
However, while technological capabilities continue to improve, significant work remains to validate microbiome-based metrics, establish regulatory standards, and demonstrate clinical utility. Without agreed-upon definitions of microbiome health and disease, microbiome testing risks remaining a research tool rather than a component of routine care.
