Loading today’s stories
Science1 distinct publisher2 min readUpdated
Stool samples linked to prescription records for more than 2,500 people showed lasting signatures from antidepressants, beta-blockers, acid reducers and benzodiazepines. Current-medication controls miss those.
The Scientist · Science desk
Compiled by The ScientistSomething wrong?How this is made
A stool sample is partly a record of a pharmacy counter, and the awkward part of this work is how the study handles drug classes. Diazepam and alprazolam are prescribed for overlapping reasons and sit in the same family, yet they differed in how strongly they tracked with disturbed gut communities [6]. Microbiome papers routinely lump drugs into classes when they adjust for medication [12]. A covariate labelled "benzodiazepine" that averages a strong actor with a weak one does not remove the confounding; it leaves a residual, and that residual will be attributed to whatever else the analysis is testing, diet included.
Lead author Oliver Aasmets puts the practice problem plainly: most microbiome studies consider only current medications, while past use turns out to be a surprisingly strong explanation of individual differences [7]. That is a statement about study design, not about biology, and it is the part that has consequences for anyone holding a dataset. Self-reported current use is a different variable from a linked prescription history with dates, which is what the Tartu group had [1].
Direction of causation is the standing weakness of any cross-sectional design here, because people are put on proton pump inhibitors for a reason and that reason is itself a candidate microbiome exposure. The repeat-sampling arm is what does the causal work: in a smaller group with a second stool sample, starting and stopping drugs was followed by shifts in the expected direction [8]. It also had few participants, so the confirmed-by-follow-up list is short rather than the whole catalogue [9].
What the announcement withholds is magnitude. The release, dated August 24, 2026 from the Estonian Research Council [10], carries a single number, the cohort of more than 2,500 [1], and no variance explained for any individual drug and no participant count for the follow-up arm [11]. Benzodiazepines being comparable to broad-spectrum antibiotics is the headline finding [5], and "comparable" needs a coefficient before anyone rebuilds a cohort protocol around it. Until the paper is available, the defensible reading is the narrow one: most of the drugs examined were associated with microbiome differences [2], several classes beyond antibiotics left detectable traces [4], and the exposure that produced them may have ended before anyone thought to record it.
Follow any of these and your For You feed starts watching them — no settings page required.
Ranked by verification strength, evidence, and original report placement.
Researchers led by the University of Tartu Institute of Genomics analysed stool samples and prescription records from more than 2,500 participants in the Estonian Biobank who were part of the Estonian Microbiome cohort.
For a substantial number of drugs, microbiome differences could still be detected years after people had stopped taking the medication.
Lead author Dr Oliver Aasmets said most microbiome studies only consider current medications, but the results show past drug use can be just as important and is a surprisingly strong factor in explaining individual microbiome differences.
The second time-point analysis involved a relatively small number of participants but confirmed persistent effects linked to proton pump inhibitors, selective serotonin reuptake inhibitors and antibiotics such as penicillins in combination and macrolides.
The announcement is dated August 24, 2026 and is sourced to the Estonian Research Council.
Most of the medications examined were associated with differences in the gut microbiome.
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
Peer-reviewed study behind it, but the only supplied account is quantity-free
The underlying work is a named peer-reviewed mSystems paper using real-world prescription records for more than 2,500 biobank participants, with a longitudinal second time point that reportedly reproduced persistence for PPIs, SSRIs, penicillin combinations and macrolides — a stronger design than cross-sectional association alone. Against that, the only supplied source is an institutional press release relayed by one outlet: it reports a single number, gives no effect size or variance explained for any drug, does not state the follow-up sample size, and includes no independent comment or discussion of confounding by indication.
No uptake evidence in supplied material
The supplied source only expresses a hope that researchers and clinicians will factor medication history into microbiome interpretation. There is no evidence of any study, cohort, guideline, tool or clinical workflow actually adopting prescription-history adjustment, and no usage, deployment or release data is present, so adoption cannot be scored without guessing.
Consumer-facing framing outruns the quantity-free, association-level results
The headline 'Common medications may change your gut for years' and the 'particularly striking' benzodiazepine comparison to broad-spectrum antibiotics assert magnitude and causal direction that the supplied text never quantifies; the release also slides from association to 'changes caused by medications taken in the past'. Hedging is present ('may', 'associated with', explicit acknowledgement that the second time point was small), and the actual methodological claim — that medication history is a confounder — is modest and well matched to the design, which keeps the gap moderate rather than severe.
Institutional promotion of its own researchers' study, relayed with light editing
The item is explicitly sourced to the Estonian Research Council and consists of materials provided by that funder-adjacent body about work led by the University of Tartu Institute of Genomics, with quotes only from the paper's lead and corresponding authors and a note that content may be edited for style and length. That is a clear promotional pathway for the institution and cohort, and it plausibly explains the striking-comparison framing and the absence of limitations. There is no evidence of commercial sponsorship, product interest or undisclosed conflicts in the supplied material, which bounds the score below the high range.
One publisher, one press release, no numbers to audit
Confidence is limited by structure rather than plausibility: a single-source, single-publisher cluster whose sole item is an institutional release. The existence of a named, DOI-identified peer-reviewed paper and an internally consistent account of design and findings supports a moderate floor, but no effect sizes, no follow-up sample size, no independent replication and no discussion of confounding by indication are available for verification.
science
Astronaut constipation gets a blood marker, and a Mars-length problem attached1 distinct publisher
Distinct publishers with included, body-backed reporting in this cluster.
1 article · August 24, 2026