A probiotic in juvenile arthritis that moved the wrong marker

Design

Multicentre randomised, double-blind, placebo-controlled trial

Population

44 children with oligoarticular or RF-negative polyarticular juvenile idiopathic arthritis on standard therapy

Primary outcome

ACR Pedi 30 response at 3 months

Effect

47% VSL#3 vs 63% placebo (P=0.33); worst-case analysis 36% vs 68% (P=0.03); faecal TLR4 agonist activity significantly increased

Forty-four children with oligoarticular or rheumatoid-factor-negative polyarticular juvenile idiopathic arthritis were randomised to VSL#3 or placebo for three months alongside standard therapy, with stool and serum sampling at baseline and month three.

There was no clinical benefit. ACR Pedi 30 response was 47 per cent on the probiotic and 63 per cent on placebo, a difference that did not reach significance — but a conservative worst-case analysis for missing data put it at 36 versus 68 per cent, which did.

The mechanistic result is why this matters beyond one product. VSL#3 significantly raised faecal Toll-like receptor 4 agonist activity — that is, it increased the pro-inflammatory microbial stimulus reaching the gut — while leaving microbial diversity, intestinal permeability and systemic cytokines unchanged. A supplement given to calm an immune-mediated disease measurably increased innate immune stimulation instead.

• Advise against probiotic supplementation as an adjunct in juvenile idiopathic arthritis.
• Ask about supplements directly; families often do not consider them medication.
• Do not generalise to other strains or products, but do not assume they are inert either.
• Forty-four children is small — this shows no benefit rather than proving harm.
• The TLR4 finding is a reason for caution in any immune-mediated disease, not just JIA.

Why it matters

It turns a harmless-sounding supplement into something with a measured direction of effect on the immune system.

Don't overread it

A 44-child trial of one probiotic preparation; the apparent disadvantage rests on a worst-case missing-data assumption.

The statistics, in plain English

The primary comparison (47 versus 63 per cent, P=0.33) is underpowered with 44 children and should be read as no evidence of benefit. The worst-case imputation result (36 versus 68 per cent, P=0.03) is a sensitivity analysis designed to be pessimistic, so its significance is a warning flag rather than a finding: it says the conclusion is sensitive to who dropped out, which is exactly what a small trial cannot resolve.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.