Endocarditis: stopping antibiotics on clinical response saved two weeks, but relapse tripled

Design

International, open-label randomised trial

Population

508 stable adults with left-sided endocarditis (S. aureus, E. faecalis, streptococci)

Primary outcome

Days alive without antibiotics at 6 months; safety composite (non-inferiority)

Effect

+13 days (95% CI 12 to 13); safety 8.2% vs 10.7%; relapse 5.1% vs 1.6%

POET II, an international open-label randomised trial in the New England Journal of Medicine (August 2026), enrolled 508 stable adults with left-sided infective endocarditis caused by Staphylococcus aureus, Enterococcus faecalis or streptococci. All had received at least 2 to 4 weeks of antibiotics and met stabilisation criteria. Half then stopped; half continued to a standard total of 4 to 6 weeks.

Response-tailored therapy gave a median 13 more days alive without antibiotics over six months (183 vs 169; 95% CI 12 to 13). The safety composite — death, unplanned cardiac surgery or symptomatic embolism — occurred in 8.2% against 10.7%, meeting non-inferiority (difference −2.4 points, −7.7 to 2.7). But relapse of bacteraemia or endocarditis was 5.1% against 1.6% (P = 0.04).

The non-inferiority margin was 7.5 points — generous for outcomes this serious — and relapse was a secondary endpoint. A three-fold rise in relapse is the result that will weigh with most clinicians.

This strengthens the case for individualised duration and follow-up, but it is not yet a licence to stop at two weeks as routine.

• Keep standard 4-6 week courses as the default for left-sided endocarditis
• Where shortening is considered, document stabilisation criteria and involve the endocarditis team
• Arrange repeat blood cultures and early review after antibiotics stop, whatever the duration
• Tell patients relapse symptoms to report — fever, rigors, new breathlessness
• Remember POET II excluded unstable patients and those needing urgent surgery

Why it matters

It is the first randomised test of stopping endocarditis antibiotics on response, and it shows the price of shorter treatment is relapse.

Don't overread it

The safety margin was wide, and relapse — the outcome that went the wrong way — was secondary.

The statistics, in plain English

Non-inferiority means the safety composite was not worse by more than 7.5 percentage points; the observed difference (−2.4, CI −7.7 to 2.7) met that. But an upper limit of 2.7 points still allows a modestly worse outcome, and the relapse difference (5.1% vs 1.6%) was statistically significant.

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.