Imagine, the look on the face of that ambitious PhD student, each day screening six hospitals for patients with S. aureus endocarditis, opening the NEJM and seeing that the Danes randomized 400 patients with Infective endocarditis (IE). And then his supervisor rubbing in that these 400 all underwent two extra transoesphageal echocardiograms for study purposes, that there were zero losses to follow-up and telling him how many samples of blood were collected to analyze antibiotic concentrations. Luckily, he was scheduled for our Journal club, which allowed him to apply the “trias scientifica”.

Endocardiis is a rare, deadly and heterogeneous disease and evidence for treatment is based mainly on small cohort studies and an exceedingly rare randomized trial. The established paradigm is that IE should be treated for 4-6 weeks with high doses of IV antibiotics (usually 12 mil. units penicillin, 12 grams of cloxacillin or 12 grams of amoxicillin) to ensure successful penetration in cardiac vegetations. In the POET trial, 400 patients with left-sided IE caused by staphylococci, streptococci and enterococci were – after at least 10 days of IV treatment – randomized to either continued IV or oral consolidation treatment, with relatively low doses of oral antibiotics (e.g. 4x1gram cloxacillin + 2x600mg rifampicin) for the last weeks of treatment. And for clarity: ~25% of the patient population had prosthetic valve endocarditis and ~40% had undergone cardiac surgery for IE in the weeks before. For the composite outcome of all-cause mortality, unplanned cardiac surgery, embolic events and relapse of bacteraemia, oral treatment was found to be non-inferior to IV treatment, with 18/201 failures occurring in the oral group and 24/199 in the IV group (between group risk difference 3.1% [95% CI –3.4 to 9.6]). A sledgehammers’ blow to the conventional wisdom practiced by generations of medical doctors.

The “trias scientifica” is the evaluation of the validity, precision and generalizability of a study. Nothing non-valid with an RCT and the confidence interval was convincingly narrow (although some things were said about the chosen non-inferiority margin for sample size calculation, see).

But what about the generalizability of the study results? Should we immediately offer oral treatment to all patients who fit the inclusion criteria to this trial? Not everyone was convinced. In several subgroups, the included number of patients was small and there were many different treatment regimens. For S. aureus IE alone 13 different antibiotic regimens were used. So, how sure can you be that your next patient with S. aureus prosthetic valve endocarditis does well on oral antibiotics, if that subgroup in the trial consisted of only 7 patients?

Consensus was not reached. Yet, a new trial seems unlikely (it took all of Denmark 6 years to screen nearly 2000 patients). All we know is that in this heterogeneous group of patients, exposed to a variety of antibiotic regimens, partial oral treatment is as safe is IV-treatment. Whether this is enough to convince clinicians to abandon their old textbook knowledge (originating from long before the evidence-based era) and join the Danes in this brave new world remains to be seen.

Thomas van der Vaart does a PhD on the recognition of endocarditis in Staphylococcus aureus bacteraemia.