The recommendation of up to six weeks of antibiotics for left-sided infective endocarditis has rested on consensus, not trials. POET II tested it. This international open-label trial randomised 508 adults with endocarditis caused by Staphylococcus aureus, Enterococcus faecalis or streptococcal species. Crucially, everyone had already received 2 to 4 weeks of therapy and met prespecified criteria for clinical stabilisation before randomisation. Those assigned to tailored therapy then stopped; those assigned to standard therapy completed a total of 4 to 6 weeks.
On the efficacy endpoint, tailored therapy won: a median of 183 days alive without antibiotics in six months versus 169, a difference of 13 days (95% CI 12 to 13, p<0.001). The safety composite of death, unplanned cardiac surgery or symptomatic embolism occurred in 8.2% versus 10.7%, difference -2.4 percentage points (95% CI -7.7 to 2.7), meeting the prespecified noninferiority margin of 7.5 points.
The secondary endpoint is where judgement is needed. Relapse of bacteraemia or endocarditis occurred in 13 patients (5.1%) on tailored therapy and 4 (1.6%) on standard therapy, p=0.04. That is a tripling, from a small base, and it did not translate into more deaths or surgery within six months — but six months is a short window for a disease that recurs.
For a non-specialist, the useful frame is this. Two weeks less of intravenous antibiotics is a real gain: fewer line days, fewer line infections, earlier discharge, less cost. It is bought with a roughly 3.5 percentage point higher chance of relapse. That trade is reasonable for many patients and not for others — a prosthetic valve, a difficult organism, or a patient who will be hard to follow up all argue for the longer course.
- Only consider shortening if the patient is genuinely stabilised after at least 2 to 4 weeks of appropriate therapy, as the trial required
- Counsel explicitly on relapse: about 5% within six months on the shorter strategy versus about 2%
- Arrange definite follow-up before stopping early — the strategy depends on catching relapse, not on avoiding it
- Do not extrapolate to right-sided disease, prosthetic valves, or organisms outside S. aureus, E. faecalis and streptococci
- Weigh the line: prolonged intravenous access carries its own infection and thrombosis risk, which is part of what the shorter course avoids
The statistics, in plain English
This trial tested two different things in two different ways, and the distinction matters. The efficacy endpoint was tested for superiority and won outright. The safety endpoint was tested for noninferiority against a 7.5 percentage point margin, meaning the trial set out to show the shorter course is not more than 7.5 points worse — a wide allowance for a disease that kills. The observed difference actually favoured the short course, but the confidence interval reaches 2.7 points in the other direction. The relapse result, at p=0.04 on a secondary endpoint with 17 events in total, is fragile as a statistic and clinically credible as a mechanism: less antibiotic, more relapse.
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.
Scan to keep reading on your phone. No account needed to start.
Tomorrow morning, before your first patient
One edition a day for top clinical updates, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free