Rebamipide was non-inferior to pantoprazole for gastroprotection on DOACs in a small open-label trial

Design

Randomised, open-label, single-centre, three-arm non-inferiority trial

Population

210 AF patients on DOACs (118 per protocol)

Primary outcome

Composite of GI bleeding, ulcers/erosions and severe dyspepsia at 24 weeks

Effect

13.9% pantoprazole vs 10.3% rebamipide vs 9.3% combination; non-inferior

REGATA, a single-centre open-label trial, randomised 210 patients with atrial fibrillation on direct oral anticoagulants to pantoprazole, rebamipide, or both for 24 weeks. Efficacy was analysed per protocol in 118 patients.

The composite of GI bleeding, ulcers or erosions and severe dyspepsia occurred in 13.9% on pantoprazole, 10.3% on rebamipide and 9.3% on the combination, meeting the 10% non-inferiority margin for both. Odds ratios against pantoprazole had wide intervals (rebamipide 0.71, 0.18–2.88). There was no major bleeding. Faecal calprotectin and zonulin fell substantially with rebamipide-containing regimens but not with pantoprazole.

Rebamipide is widely available and inexpensive in India and much of Asia, and its putative lower-GI protection is attractive for anticoagulated patients. But with per-protocol analysis of just over half the randomised patients, very few events and surrogate biomarkers, this is not enough to replace a PPI.

• Keep a PPI as the standard gastroprotective agent for high-risk patients on anticoagulants.
• Rebamipide may be an option where a PPI is not tolerated, but evidence is limited.
• Falls in faecal calprotectin and zonulin are surrogate markers, not proof of fewer bleeds.
• Assess GI bleeding risk and review concomitant antiplatelets and NSAIDs in every patient on a DOAC.

Why it matters

It raises lower-GI protection as a question for anticoagulated patients, where PPIs do little.

Don't overread it

Open-label, single-centre, per-protocol analysis with few events; the biomarker changes are surrogates.

The statistics, in plain English

The odds ratios have confidence intervals from about 0.2 to 2.9, meaning the data are consistent with large benefit or large harm. Per-protocol analysis of 118 of 210 randomised patients can bias results, and a 10% non-inferiority margin is wide for an outcome that occurred in about 10–14%.

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.