Evidence Grading Framework
Microbiome science moves quickly, but not every finding is equally trustworthy. This page describes how we grade evidence and why uncertainty is common in this field.
Evidence hierarchy
Strong — replicated and clinical
Findings supported by multiple independent studies, including randomized or well-controlled human trials where relevant. Effect sizes are consistent enough that expert reviews cite them with confidence. Example: established roles of specific microbes in defined biochemical pathways under controlled conditions.
Moderate — associative
Repeated observational links in human cohorts, or strong animal data with some human corroboration. Direction of association is fairly consistent, but causation is not established. Example: correlations between certain taxa and metabolic markers across population studies.
Weak — exploratory
Single studies, small samples, post-hoc analyses, or novel claims awaiting replication. May include promising mechanistic work that has not been tested in diverse human populations. Treat as hypothesis-generating, not actionable.
Why microbiome research is uncertain
Variability between studies
- Different cohorts, diets, geographies, and health statuses
- Inconsistent sample collection, storage, and DNA extraction methods
- Changing sequencing platforms and bioinformatics pipelines
- Publication bias toward positive associations
Compositional data issues
Microbiome data are compositional: if one taxon increases as a fraction of the total, others must decrease. Standard statistics that ignore this structure can produce misleading associations. Labs also differ in detection limits, so “zero” or “low” often means below detection, not truly absent.
Claim grading system
We use a simple A–D scale for statements on this site. Grades describe evidence quality, not importance to your health.
| Grade | Meaning | Typical source |
|---|---|---|
| A | Strong evidence | Replicated human or mechanistic studies; clinical support where applicable |
| B | Moderate evidence | Consistent associations or robust models with limited clinical translation |
| C | Limited evidence | Early trials, conflicting cohorts, or indirect inference |
| D | Insufficient for claims | Speculation, single reports, or not valid for individual test interpretation |
How to use these grades
- Prefer acting on clinical advice and replicated outcomes over report flags
- When a species page shows grade B or C, treat it as context—not instruction
- Grade D items are included to flag common over-interpretations
- As new studies appear, grades may be updated; uncertainty is expected
Principle: Clear grading builds trust. We label evidence explicitly so readers can separate what is known, what is plausible, and what is not yet supported.