Design: multicentre randomised, double-blind, placebo-controlled trial with mechanistic sampling at baseline and month 3
Population: 44 children with oligoarticular or rheumatoid-factor-negative polyarticular juvenile idiopathic arthritis, enrolled 2017–2022, all on standard therapy
Primary outcome: proportion achieving ACR Pedi 30 response at three months
Effect: 47% with probiotic vs 63% with placebo (P = 0.33); worst-case imputation 36% vs 68% (P = 0.03); faecal Toll-like receptor 4 agonist activity rose significantly, with no change in diversity, permeability or systemic cytokines

Gut dysbiosis is a plausible contributor to juvenile idiopathic arthritis, and probiotics are the obvious thing to try. The PERMAJI trial — multicentre, randomised, double-blind, placebo-controlled — gave 44 children with oligoarticular or rheumatoid-factor-negative polyarticular disease either a multi-strain probiotic or placebo for three months alongside standard therapy, and measured both clinical response and what was happening in the gut.

Clinically it did not work. ACR Pedi 30 response at three months was 47% on the probiotic against 63% on placebo (P = 0.33), and under conservative worst-case assumptions for missing data the gap widened to 36% against 68% (P = 0.03) — that is, the sensitivity analysis pointed towards the probiotic doing worse, not better. The mechanistic findings explain why that is not merely a null result. Faecal Toll-like receptor 4 agonist activity rose significantly on the probiotic, while gut microbiota diversity, intestinal permeability and systemic cytokines were unchanged. In other words the supplement increased exposure to a pro-inflammatory microbial stimulus without shifting the community composition it was supposed to shift.

For clinic this is worth having, because probiotics are widely taken by children with juvenile arthritis on parental initiative and are generally assumed to be at worst neutral. This trial is small, and a 44-patient study cannot establish harm. But it does remove the 'can't hurt' argument, and it gives a mechanism for why a microbiome-targeted intervention in an autoimmune disease might not be neutral at all.

Ask about probiotic supplements at review; parents often do not consider them medication.
Do not recommend probiotics for disease activity in juvenile idiopathic arthritis.
Do not describe them as harmless either — this trial found a pro-inflammatory microbial signal.
Keep standard disease-modifying therapy as the intervention; both arms received it.
Read the worst-case imputation result as a reason for caution, not as evidence of harm.

Why it matters

It supplies a mechanism by which a supplement assumed to be neutral might not be, in a disease where families reach for it routinely.

The statistics, in plain English

The headline comparison (47% vs 63%, P = 0.33) is a null result in 44 children, which is far too few to exclude a real effect in either direction. The worst-case imputation result (36% vs 68%, P = 0.03) is the more interesting number and the more fragile one: assigning every missing outcome unfavourably to the treatment arm is a deliberately pessimistic assumption, and a significant P value under that assumption is not the same as a demonstrated harm. The mechanistic finding is the sturdier part, because it is a measured laboratory change rather than a clinical endpoint subject to dropout.