A stool transplant changed the gut microbiome and changed nothing else

The gap between showing that a gut microbiome differs and showing that changing it helps is the central difficulty in microbiome medicine. Sequencing can demonstrate the first cheaply and repeatedly. Only a controlled trial can demonstrate the second, and controlled trials of microbiome interventions remain comparatively scarce.

Two randomised trials published in 2026 show what happens when the question is put properly.

The null result

A randomised, double-blind, sham-controlled trial published in JAMA Internal Medicine tested whether FMT could clear multidrug-resistant organisms from the guts of patients with gastrointestinal disease [s1]. Gut colonisation by such organisms is a risk factor for subsequent infection with them, and there are no approved therapeutic interventions against it [s1].

The trial randomised 114 patients in a gastroenterology ward and intensive care unit at a tertiary care centre in India, recruited between July 2022 and June 2024 [s1]. Mean age was 40.6 years (SD 12.5) and 80 participants (70.2%) were male; 52 (45.6%) had pancreatitis, 43 (37.7%) cirrhosis, and 19 (16.7%) other gastrointestinal disorders [s1]. Fifty-eight received FMT by colonoscopy and 56 received a sham procedure — sigmoidoscopy with a saline injection [s1]. Nearly all were colonised at baseline with carbapenem-resistant or extended-spectrum β-lactamase-producing Enterobacteriaceae [s1].

The co-primary outcomes were decolonisation rate and reduction in antimicrobial resistance genes at four weeks [s1].

Neither moved. In the intention-to-treat analysis, 18 patients (31.0%) in the FMT group and 17 (30.4%) in the sham group were decolonised — an absolute difference of 0.6%, with a confidence interval running from −16.2% to 17.6% (P = .94) [s1]. Resistance gene counts were a median of 2.5 (IQR 1.2–3.0) in the FMT group against 2.0 (IQR 1.0–3.0) in the sham group (P = .68) [s1]. Adverse events were comparable [s1].

The part that did work

The trial's mechanistic measurements did not fail. Among the 71 patients who underwent 16S ribosomal RNA sequencing at four to six weeks, the FMT group showed enrichment of bacteria capable of producing short-chain fatty acids, and the sham group did not [s1]. Viral diversity was unchanged after FMT, and the gut mycobiome showed only modest, transient alterations [s1].

So the intervention did what it was designed to do at the level of composition, and that composition change produced no clinical benefit on either co-primary endpoint. The authors' conclusion is exactly that: a single session of FMT did not significantly enhance decolonisation or reduce resistance genes, while it did modulate gut microbiome diversity and composition [s1].

An accompanying commentary in the same journal framed the open question as whether benefit requires disruption [s2] — that is, whether a resident microbial community must first be destabilised, by antibiotics or bowel preparation, before transplanted organisms can displace what is already there.

This is the microbiome field's characteristic result, and it is worth naming plainly. A shift in 16S sequencing output is a biological effect. It is not, by itself, a clinical one, and the two are routinely conflated in coverage of microbiome research.

The scalability trial

The second paper addresses a different obstacle. FMT is an effective therapy for recurrent Clostridioides difficile infection, but it has undefined composition and poor scalability [s3] — because every dose originates in a different donor.

A phase 1b randomised trial published in Nature Medicine on 2 June compared the same bacterial strains delivered two ways: as donor-sourced FMT, and as MTC01, an in vitro manufactured 15-strain live biotherapeutic product [s3]. The authors describe it as the first direct comparison of its kind, and published regulatory documentation and manufacturing protocols alongside it to lower the barrier for other groups [s3].

Of 20 patients screened, 18 were eligible and were randomised evenly across four arms: low-dose FMT (n = 4), high-dose FMT (n = 5), low-dose MTC01 (n = 4) and high-dose MTC01 (n = 5), with a 5:1 female-to-male ratio [s3].

The primary outcome was safety, and it was met: 10 adverse events across eight patients, split evenly between MTC01 and FMT recipients at five events each, with no treatment-related adverse events in any arm [s3].

On the secondary efficacy outcome, recurrent infection was prevented at eight weeks in seven of nine patients who received the manufactured product, against eight of nine who received FMT [s3]. Strain engraftment was high and durable for both, with a dose effect for the biotherapeutic [s3].

Why nine-versus-nine is not a result

Seven of nine against eight of nine looks like equivalence and is not evidence of it. With nine patients per intervention, a single different outcome moves the proportion by 11 percentage points, and the trial was neither designed nor powered to detect a difference in efficacy — safety was the primary endpoint [s3].

The contribution here is not that MTC01 works as well as FMT. It is that a defined, manufacturable, 15-strain product engrafted durably in humans and was safe at both doses, and that the manufacturing route is now documented for others to use [s3]. That is a platform result, and platform results are how a field stops depending on donors.

What the pair suggests

Read together, the two trials describe a discipline whose tools have outrun its indications. It is now possible to change a human gut microbiome deliberately, to verify the change by sequencing, to manufacture defined consortia to specification, and to demonstrate durable engraftment.

What remains difficult is showing that any of this alters a clinical outcome outside the single indication where it was first demonstrated. The sham-controlled trial is the more instructive of the two precisely because it was properly controlled: without a sham arm, a 31% decolonisation rate would have read as a treatment effect rather than as what the sham group also achieved [s1].

What to watch

The disruption hypothesis raised in the accompanying commentary [s2] is testable — trials that condition the gut before transplant would answer whether engraftment resistance explains the null. For live biotherapeutics, the meaningful next step is a properly powered efficacy trial of a defined product against standard care, in an indication chosen because the mechanism predicts benefit rather than because the correlational literature is large.

The trials are registered as CTRI/2022/07/043847 [s1] and NCT05911997 [s3].