Reviews/Probiotics and Gut Microbiome/Probiotics prevent Clostridioides difficile infection
Clinical · PGM-GI-03 · v2
This corpus asks whether taking probiotics lowers the chance of a Clostridioides difficile infection, mostly in hospital patients on antibiotics. The largest, best-designed trials, in which neither patients nor staff knew who received the probiotic, found no reduction in infections [1][3][7]. The only studies showing a large benefit were much weaker: one where everyone knew who got the treatment [2], and one that simply compared before and after a policy change with no proper control group [4]. Several trials also happened to see very few infections in either arm, which left them too small to prove either way [1][8]. One large trial additionally flagged a safety concern, with the probiotic organism turning up in the bloodstream of some patients [3]. On balance the panel judged that the broad claim is not carried by the evidence, though a benefit tied to a specific strain or setting is not completely ruled out.
The blinded, adequately powered randomised trials in this corpus are null on CDI: [1] reports an incidence rate ratio of 0.92 (95% CI 0.68-1.23), [7] a relative risk of 0.71 (95% CI 0.34-1.47, p=0.35) that is favourable in direction but crosses no effect, and [3] finds no difference in CDI among 20 prespecified secondary outcomes. Further null or negative results appear in [5], [6] and [8]. The two positive signals come from the weakest designs: an open-label single-centre RCT [2] (0% vs 11%, p<0.01) and an uncontrolled before-and-after study [4] (0.1% vs 1.9%, relative risk 0.05 [0.01-0.39]). Where any effect is seen it is tied to specific strains, doses and populations rather than to probiotics as a class. [3] also records the probiotic organism in a sterile site in 15 (1.1%) probiotic vs 1 (0.1%) control patients (odds ratio 14.02; P<.001). The contested point: The division is between design and result: benefit appears only in the two highest-bias studies [2][4], while every blinded, well-powered RCT [1][3][7] lands null or non-significant. Whether the positive strain-specific signals in [2] and [4] reflect a real localised effect or simply bias is the open point, and low event rates in [1] and [8] mean the evidence base cannot fully exclude a modest strain-specific benefit.
All 12 papers behind this claim with their DOIs and what each does to it, the panel’s arbitration, the evidence gap map and the open research questions. Free, and sent to your inbox.
We’ll also send you other claims from Probiotics and Gut Microbiome, one at a time, and tell you when a release or a pivotal trial changes a grade. Unsubscribe by replying.
One reviewer's only job is to attack the finding. The challenges below were weighed before the grade was set.
Effect size tracks inversely with rigour: the only positive studies [2] and [4] are the weakest designs (open-label single-centre RCT; uncontrolled before-after), while the largest blinded RCTs [1][3][7] are null — a classic bias signature.
The claim is a blanket class-level statement, but the corpus effects are strain-, dose- and population-specific; a favourable strain-specific result does not license a general 'probiotics reduce CDI risk' claim.
Where it remains uncertain. The only large positive effects are in an open-label RCT [2] (0% vs 11%) and an uncontrolled before-after study [4] (0.1% vs 1.9%), so benefit tracks inversely with methodological rigour.
Aleatory
Baseline CDI incidence was low in [1] and diarrhoea incidence unexpectedly low in [8], so chance variation dominates and small trials cannot resolve a modest effect.
Epistemic
The positive results come from an open-label RCT [2] and an uncontrolled before-after study [4], both high risk of bias, so their effect estimates are not trustworthy.
Inferential
Extending a favourable strain-specific result to a blanket 'probiotics reduce CDI risk' claim is an inferential leap the corpus does not license.
The load-bearing papers from a corpus of 12, each classified by what it does to the claim. Every DOI is taken from the corpus the claim was graded on.
Largest cluster-RCT with CDI as primary outcome is null, undercutting a general prevention claim.
“there was no change in hospital-acquired CDI (incidence rate ratio 0.92; 95% CI 0.68-1.23)”
Panel auditInternal validity: medium · scope of inference: Hospitalized patients aged >=55 receiving antibiotics in acute-care settings with low baseline CDI incidence.Key limitations: low baseline CDI incidence · moderate adherence · cluster-level analysis
Positive signal but single-centre open-label design (high risk of bias).
“7/57 (11%) of patients treated only with antibiotics developed Clostridium difficile infection compared to 0% in the probiotic group (p<0.01)”
Panel auditInternal validity: low · scope of inference: Adult hospitalized patients on antibiotics at a single center receiving these two specific strains.Key limitations: open-label · single-center · small sample · zero events in one arm
Large ICU RCT shows no CDI benefit as a secondary outcome.
“None of the 20 prespecified secondary outcomes, including Clostridioides difficile infection and diarrhea, showed a significant difference.”
Panel auditInternal validity: high · scope of inference: Critically ill adults requiring mechanical ventilation across 44 ICUs.Key limitations: CDI is a secondary outcome · no CDI-specific numbers in abstract · probiotic detected in sterile sites (safety signal)
The abstract states L. reuteri did not significantly affect incidence or severity of AAD or CDI.
Panel auditInternal validity: low · scope of inference: Hospitalized children aged 3-12 on antibiotics with high asymptomatic C. difficile colonisation.Key limitations: small sample · very low event rate · underpowered · authors state no firm conclusion
4 of 12 papers shown. The full review lists all of them, with the panel’s arbitration →
Claims are graded at each quarterly release, and re-graded between releases only when a pivotal trial lands. Every change is a new version of the claim.
Rigorous adequately-powered RCTs are null and the positive signal is confined to the weakest designs, so the blanket claim is not supported though a strain-specific effect is not fully excluded.
Release 2026-q4 re-graded every claim from its corpus under protocol v0.19. An earlier release's grade is shown for the record; it came from an earlier protocol and is not a movement.
Assumptions that carry weight but are not themselves established.
The panel records the gap that would most move the claim, and the study that would close it.
Blinded RCT of the specific positive strains. The only positive signals [2][4] came from biased designs using particular strains; a rigorous test of those exact strains is missing. The panel rates this the highest-importance open gap for the claim, and names the study design that would close it: Double-blind placebo-controlled RCT of the L. reuteri/L. rhamnosus GG mix from [2] with CDI as the primary endpoint in antibiotic-exposed inpatients.
Open question: Do the specific strains that showed benefit in the weakest designs reduce CDI when tested under blinding? Why it is still open: The positive results in [2] and [4] are confounded by open-label and uncontrolled designs, while blinded RCTs of other strains [1][3][7] are null.
Priors is a sceptical first reader, not peer review, and not medical advice.
Release 2026-q4 · graded 18 Sep 2026 · protocol v0.19