Skip to main content

Trial Summaries & Critiques

SNAP - Benzylpenicillin

The Staphylococcus aureus Network Adaptive Platform (SNAP) Trial Group. Benzylpenicillin versus flucloxacillin or cloxacillin for the treatment of penicillin-susceptible Staphylococcus aureus bacteraemia (SNAP): an international, multicentre, open-label, non-inferiority randomised controlled trial. Lancet. 2026; published online June 17

Image: Shutterstock / labelmydata

Staphlococcus Aureus on gram stain

Publication

Context & Rationale

  • Background
    • Staphylococcus aureus bacteraemia is common, high-risk, and still supported by a surprisingly thin randomised evidence base.
    • Global practice before SNAP was highly variable, including antibiotic choice, oral switch practice, imaging, treatment duration, and definitions of persistent bacteraemia.1
    • Penicillin-susceptible S aureus (PSSA) had re-emerged internationally after decades in which most S aureus isolates were assumed to be penicillin-resistant.
    • Anti-staphylococcal penicillins such as flucloxacillin, cloxacillin, oxacillin, and nafcillin were historically favoured for serious PSSA infection because of concern that laboratories might falsely classify low-level penicillinase-producing isolates as susceptible.
    • Benzylpenicillin is pharmacologically attractive for true PSSA: lower MICs, lower protein binding than flucloxacillin, a narrower spectrum, and an easy oral step-down option with amoxicillin.
    • Pre-randomised evidence consisted mainly of observational studies, including cohorts suggesting similar or better outcomes with benzylpenicillin compared with anti-staphylococcal penicillins, but with confounding by indication, immortal-time bias, and heterogeneity in penicillinase testing.34
  • Research Question/Hypothesis
    • The trial asked whether intravenous benzylpenicillin was non-inferior to intravenous flucloxacillin or cloxacillin for 90-day all-cause mortality in adults with PSSA bacteraemia.
    • It also tested whether the narrower drug reduced clinically important toxicity, especially acute kidney injury.
    • The question was embedded within the antibiotic-backbone domain of the broader Bayesian adaptive SNAP platform, designed to answer several S aureus bacteraemia questions in parallel.2
  • Why This Matters
    • If benzylpenicillin is effective and safer for true PSSA, then many adults can avoid broader, more nephrotoxic anti-staphylococcal penicillins.
    • The result directly tests a long-standing clinical dogma: that penicillin should be avoided for serious S aureus infection despite confirmed in-vitro susceptibility.
    • The trial also addresses whether modern confirmatory penicillin-susceptibility testing can safely drive bedside treatment decisions.

Design & Methods

  • Research Question: In adults with confirmed PSSA bacteraemia, is benzylpenicillin non-inferior to flucloxacillin or cloxacillin for all-cause mortality at 90 days, and does it reduce treatment-related harm?
  • Study Type: Investigator-initiated, international, multicentre, open-label, randomised, Bayesian adaptive-platform, non-inferiority trial conducted within the PSSA silo of SNAP’s antibiotic-backbone domain.
  • Population:
    • Adults aged ≥18 years admitted to 67 hospitals in Australia, New Zealand, Canada, Israel, the Netherlands, the UK, Singapore, and South Africa.
    • Eligible patients had at least one blood culture positive for PSSA.
    • PSSA required confirmatory testing: disc diffusion with zone diameter and zone-edge interpretation by CLSI or EUCAST methods, or absence of blaZ by PCR.
    • Key platform exclusions were anticipated enrolment >72 h after index blood-culture collection, polymicrobial bacteraemia judged clinically significant, prior SNAP participation, recent same-silo S aureus bacteraemia within 72 h to 180 days, imminent inevitable death, and goals of care excluding antibiotic therapy.
    • Key domain exclusions were type I hypersensitivity or severe delayed reaction to penicillin or cephalosporin, non-severe penicillin rash unless de-labelled, maintenance dialysis, or need for an additional systemic antibacterial agent with S aureus activity that could not be stopped or substituted.
    • Adults with PSSA bacteraemia screened for the domain: 493.
    • Randomised adults: 281; benzylpenicillin 156 and flucloxacillin or cloxacillin 125.
  • Intervention:
    • Intravenous benzylpenicillin.
    • Standard recommended dose: 1.8 g IV every 4 h, or 2.4 g IV every 6 h, at the treating clinician’s discretion.
    • For critical illness, endocarditis, or central nervous system infection, the recommended high dose was benzylpenicillin 2.4 g IV every 4 h.
    • Renal dose adjustment was specified for reduced glomerular filtration rate and continuous renal replacement therapy.
    • Continuous infusion at 75-100% of the total daily dose was allowed according to local practice.
    • Allocated antibiotics were intended for the full parenteral treatment duration: minimum 14 days for uncomplicated bacteraemia and 28-42 days for complicated bacteraemia, unless patients entered the early-oral-switch domain.
  • Comparison:
    • Intravenous flucloxacillin or cloxacillin.
    • Flucloxacillin standard dose: 2.0 g IV every 6 h.
    • Flucloxacillin high dose for critical illness, endocarditis, or central nervous system infection: 2.0 g IV every 4 h.
    • Cloxacillin was used where flucloxacillin was unavailable, specifically Canada, Israel, Singapore, and South Africa.
    • Cloxacillin dose: 2.0 g IV every 4 h.
    • Renal dose adjustment was not recommended for cloxacillin.
    • Additional anti-staphylococcal antibiotics were discouraged, except clindamycin in patients also enrolled in the adjunctive-treatment domain.
  • Blinding: Open-label. Trial staff, treating clinicians, and participants knew allocation. Aggregate outcome analyses remained restricted to the unblinded analytical team and DSMC until closure of the relevant silos. This matters because mortality and AKI were objective, but decisions such as changing antibiotics for perceived inefficacy were vulnerable to clinician expectation.
  • Statistics: There was no conventional fixed-sample alpha/beta power calculation. The Bayesian adaptive design used simulations with a maximum platform sample size of 7000; benzylpenicillin non-inferiority was defined as an adjusted OR <1.20 for 90-day mortality, corresponding to about a 2.5% absolute mortality difference if control mortality was 15%; the prespecified stopping threshold for the adult PSSA silo was posterior probability of non-inferiority >99%; scheduled analyses occurred after every 500 platform participants reached 90-day follow-up. The primary analysis followed the intention-to-treat principle using all participants with available 90-day mortality data, and a protocol-adherent analysis was prespecified.2
  • Follow-Up Period: 90 days after platform entry, with data collection at platform days 7, 14, 28, 42, and 90.

Key Results

This trial was stopped early. After the fourth planned interim analysis, recruitment was paused because of an acute kidney injury safety signal in the flucloxacillin or cloxacillin group across the PSSA and MSSA silos; on Aug 7, 2024 the PSSA silo was closed before the prespecified PSSA non-inferiority stopping threshold had been met.

Outcome Benzylpenicillin Flucloxacillin or cloxacillin Effect p value / 95% CI Notes
90-day all-cause mortality, primary analysis 21/152 (14%) 26/121 (21%) Adjusted OR 0.67 95% CrI 0.35 to 1.28; no p value calculated Posterior probability of benzylpenicillin non-inferiority 96.1%; superiority 88.9%; prespecified >99% non-inferiority stopping threshold not met.
90-day all-cause mortality, protocol-adherent population 9/122 (7%) 19/103 (18%) Adjusted OR 0.43 95% CrI 0.19 to 0.93; no p value calculated Posterior probability of non-inferiority 99.6%; superiority 98.4%.
Acute kidney injury within 14 days 17/153 (11%) 27/124 (22%) Adjusted OR 0.50 95% CrI 0.26 to 0.94; no p value calculated Posterior probability of non-inferiority 99.8%; superiority 98.4%.
AKI severity Stage 1: 10/153 (7%); Stage 2: 5/153 (3%); Stage 3: 2/153 (1%) Stage 1: 18/124 (15%); Stage 2: 6/124 (5%); Stage 3: 3/124 (2%) Not modelled as ordinal in main table Not reported Most AKI episodes were mild, but more occurred with flucloxacillin or cloxacillin.
Treatment failure by day 90 32/148 (22%) 32/113 (28%) Adjusted OR 0.75; adjusted risk difference -5.5% OR 95% CrI 0.42 to 1.33; risk difference 95% CrI -14.1% to 6.1%; no p value calculated Composite of death, microbiological treatment failure, or new infectious focus; posterior probability of non-inferiority 99.9% using OR <1.85 margin.
14-day mortality 10/156 (6%) 8/125 (6%) Adjusted OR 1.14 95% CrI 0.47 to 2.76; no p value calculated Early mortality was similar.
28-day mortality 13/156 (8%) 16/124 (13%) Adjusted OR 0.77 95% CrI 0.36 to 1.62; no p value calculated Direction favoured benzylpenicillin.
42-day mortality 18/156 (12%) 17/123 (14%) Adjusted OR 1.00 95% CrI 0.49 to 2.00; no p value calculated No clear difference.
Change of antibiotic due to perceived inefficacy 11/156 (7%) 1/125 (1%) Adjusted OR 5.82 95% CrI 1.74 to 30.28; no p value calculated Likely influenced by open-label clinician concern about penicillin susceptibility.
Microbiological treatment failure after day 14 6/141 (4%) 3/106 (3%) Adjusted OR 1.30 95% CrI 0.42 to 4.00; no p value calculated Low event rates.
New focus of infection after day 14 7/143 (5%) 6/113 (5%) Adjusted OR 0.90 95% CrI 0.33 to 2.40; no p value calculated Similar rates.
Any serious adverse reaction related to study drug 6/156 (4%); 7 events 9/125 (7%); 10 events Adjusted OR 0.79 95% CrI 0.29 to 2.02; no p value calculated Flucloxacillin/cloxacillin events included AKI, hepatic failure, hyponatraemia, and one SUSAR of pyroglutamic acidosis.
New renal replacement therapy within 90 days 3/156 (2%) 5/125 (4%) Adjusted OR 0.57 95% CrI 0.17 to 1.79; no p value calculated No survivors required ongoing renal replacement therapy at day 90.
Hepatotoxicity within 14 days 18/145 (12%) 17/107 (16%) Adjusted OR 0.76 95% CrI 0.37 to 1.50; no p value calculated ALT or GGT >2.5 times upper limit of normal.
Clostridioides difficile infection 0/156 (0%) 1/123 (1%) Adjusted OR 0.39 95% CrI 0.05 to 2.07; no p value calculated Rare.
Central catheter complication requiring removal 7/156 (4%) 6/125 (5%) Adjusted OR 0.84 95% CrI 0.31 to 2.18; no p value calculated No clear separation.
90-day mortality, no endocarditis subgroup 20/143 (14%) 23/106 (22%) Adjusted OR 0.69 95% CrI 0.34 to 1.33; no p value calculated Direction consistent with primary analysis.
90-day mortality, endocarditis subgroup 1/8 (13%) 3/15 (20%) Adjusted OR 0.99 95% CrI 0.17 to 5.40; no p value calculated Only 23 patients; no meaningful subgroup conclusion.
  • Benzylpenicillin did not cross the trial’s stringent prespecified PSSA non-inferiority stopping threshold for 90-day mortality, but the posterior probability of non-inferiority was high at 96.1%.
  • The most clinically persuasive separation was safety: AKI occurred in 11% with benzylpenicillin versus 22% with flucloxacillin or cloxacillin, with an adjusted OR 0.50; 95% CrI 0.26 to 0.94.
  • Separate descriptive results for the comparator agents were similar: 90-day mortality was 21/152 (14%) with benzylpenicillin, 18/86 (21%) with flucloxacillin, and 8/35 (23%) with cloxacillin; AKI was 17/153 (11%), 20/87 (23%), and 7/37 (19%), respectively.

Internal Validity

  • Randomisation and Allocation:
    • Participants were randomised 1:1 using a central electronic data-capture system.
    • Allocation was revealed only after PSSA silo and domain eligibility were confirmed.
    • Simple randomisation without stratification contributed to numerical imbalance: 156 allocated to benzylpenicillin versus 125 to flucloxacillin or cloxacillin.
  • Dropout or Exclusions:
    • Of 493 adults with PSSA bacteraemia screened, 281 were eligible and consented.
    • Exclusions among screened PSSA adults were substantial: 42 did not consent, 83 were at sites not participating in the PSSA backbone domain, and 87 were domain-ineligible.
    • Eight participants had missing 90-day primary outcome data: 4 in each treatment group.
    • Primary complete-case analysis included 273 patients: 152 benzylpenicillin and 121 flucloxacillin or cloxacillin.
    • Sensitivity analysis assuming all missing patients died gave mortality 25/156 (16%) versus 30/125 (24%); adjusted OR 0.68; 95% CrI 0.30 to 1.22.
    • Sensitivity analysis assuming all missing patients survived gave mortality 21/156 (13%) versus 26/125 (21%); adjusted OR 0.70; 95% CrI 0.37 to 1.32.
  • Performance/Detection Bias:
    • The trial was open-label, which is unavoidable for pragmatic antibiotic-allocation trials but increases risk of clinician-behaviour effects.
    • The primary outcome, all-cause mortality at 90 days, is objective and minimally susceptible to ascertainment bias.
    • The key safety outcome, AKI, was based on serial serum creatinine measurements and is also largely objective.
    • Subjective downstream decisions were clearly affected: antibiotic change due to perceived inefficacy was 11/156 (7%) with benzylpenicillin versus 1/125 (1%) with flucloxacillin or cloxacillin.
  • Protocol Adherence:
    • The protocol-adherent population comprised 122/156 (78%) in the benzylpenicillin group and 103/125 (82%) in the flucloxacillin or cloxacillin group.
    • Protocol-adherent results favoured benzylpenicillin more strongly than the primary analysis: 9/122 (7%) versus 19/103 (18%); adjusted OR 0.43; 95% CrI 0.19 to 0.93.
    • Protocol deviations were similar: 16/156 (10%) with benzylpenicillin and 14/125 (11%) with flucloxacillin or cloxacillin.
  • Baseline Characteristics:
    • Median age was 65.5 years in the benzylpenicillin group and 69.0 years in the flucloxacillin or cloxacillin group.
    • Female sex at birth was 46/156 (29%) versus 41/125 (33%).
    • Diabetes was present in 48/156 (31%) versus 43/125 (34%).
    • Chronic kidney disease was present in 20/156 (13%) versus 14/125 (11%).
    • ICU admission at recruitment was 23/156 (15%) versus 11/125 (9%), favouring the comparator group at baseline.
    • Endocarditis was recorded in 8/156 (5%) versus 15/125 (12%), favouring the benzylpenicillin group at baseline.
    • Median Pitt Bacteraemia Score was 0 in both groups.
  • Heterogeneity:
    • The trial was geographically broad, including eight countries, but nearly all patients came from high-income health systems.
    • Comparator heterogeneity was handled pragmatically: flucloxacillin was used where available and cloxacillin where flucloxacillin was not available.
    • Analysing flucloxacillin and cloxacillin separately did not materially alter the mortality or AKI pattern.
    • The hierarchical Bayesian model adjusted for age, country, temporal epoch, and eligibility and assignment in other SNAP domains.
  • Timing:
    • Enrolment had to occur within 72 h of index blood-culture collection.
    • Median time from index blood culture to platform entry was 54.4 h with benzylpenicillin and 49.0 h with flucloxacillin or cloxacillin.
    • Median time from index blood culture to domain entry was 63.4 h versus 63.7 h.
    • Almost all patients had received antibiotics before platform entry: 155/156 (99%) versus 124/125 (99%).
  • Dose:
    • Benzylpenicillin dosing was clinically credible and allowed renal and high-severity adjustment.
    • Flucloxacillin was dosed at 2 g every 6 h as standard, with 2 g every 4 h for critical illness, endocarditis, or central nervous system infection.
    • Cloxacillin was dosed at 2 g every 4 h, giving 12 g/day, with no renal dose adjustment; this may have contributed to nephrotoxicity and limits extrapolation to lower-dose cloxacillin regimens.
  • Separation of the Variable of Interest:
    • Before platform entry, 84/156 (54%) assigned to benzylpenicillin and 66/125 (53%) assigned to flucloxacillin or cloxacillin had already received flucloxacillin or cloxacillin.
    • Median estimated total antibiotic duration was 39 days (IQR 21-58) with benzylpenicillin and 42 days (IQR 18-52) with flucloxacillin or cloxacillin.
    • Days alive and free of antibiotics between platform entry and day 90 were similar: median 43 days (IQR 5-61) versus 44 days (IQR 12-59).
    • Specialist infectious diseases consultation was nearly universal: 153/156 (98%) versus 122/125 (98%).
    • Change of antibiotic due to adverse event was 8/156 (5%) versus 10/125 (8%).
  • Key Delivery Aspects:
    • Clinical management other than allocated backbone antibiotic was left to treating clinicians, consistent with a pragmatic effectiveness trial.
    • Formal infectious diseases consultation was advised and available at all sites.
    • The trial required modern confirmation of penicillin susceptibility, which was central to safety and interpretability.
  • Adjunctive Therapy Use:
    • The proportions allocated to ongoing SNAP domains, such as adjunctive clindamycin or early oral switch, were not reported to preserve blinding and trial integrity.
    • Those domains used 1:1 randomisation and were adjusted for in the main statistical model.
  • Outcome Assessment:
    • Mortality and AKI were clearly defined, objective outcomes.
    • New infectious foci and antibiotic changes were more clinically interpretive and more exposed to open-label bias.
    • Site staff assessed participant-level outcomes with knowledge of allocation.
  • Statistical Rigor:
    • The analysis followed a prespecified Bayesian hierarchical model and prespecified decision thresholds.
    • The PSSA silo was closed before the prespecified non-inferiority threshold was met for the primary outcome.
    • No p values were calculated, and there was no prespecified multiplicity adjustment for secondary outcomes.
    • The treatment-failure composite was added after PSSA and MSSA recruitment closed but before unblinding and analysis, to align with prior S aureus bacteraemia drug-registration trial outcomes.

Conclusion on Internal Validity: Internal validity is moderate-to-strong. Randomisation, concealed allocation, objective primary and key safety outcomes, and consistency with the protocol-adherent analysis support credibility, but early stopping before the PSSA non-inferiority trigger, open-label management, small sample size, and baseline/allocation imbalance prevent a fully definitive interpretation.

External Validity

  • Population Representativeness:
    • The age and sex distribution was typical of adult S aureus bacteraemia: median age 67 years, 87/281 (31%) female and 194/281 (69%) male.
    • Osteoarticular infection was the most common source: 85/281 (30%).
    • People who inject drugs were under-represented: 12/281 (4%) reported injection drug use within 6 months, whereas contemporary cohorts often include approximately 10%.
    • Maintenance dialysis patients were excluded.
    • Paediatric patients were not reported in this analysis.
    • Endocarditis was uncommon: 23/281 (8%); the endocarditis subgroup cannot drive practice alone.
  • Applicability:
    • Findings apply best to adult patients with PSSA bacteraemia in hospitals capable of reliable confirmatory penicillin-susceptibility testing and close infectious diseases oversight.
    • The results should not be applied to isolates labelled penicillin-susceptible by automated antimicrobial susceptibility testing alone.
    • The results do not apply to MRSA, penicillin-resistant MSSA, children, patients on maintenance dialysis, or patients with clinically important beta-lactam allergy.
    • Nearly all participants were from high-income countries: 279/281, with only 2 participants from South Africa.
    • Generalisability to low-resource laboratories is constrained by the need for disc-edge interpretation or blaZ PCR.
    • Generalisability to nafcillin or oxacillin is biologically plausible but indirect, because the comparator agents were flucloxacillin and cloxacillin.

Conclusion on External Validity: External validity is strong for high-resource adult PSSA bacteraemia services with validated penicillin-susceptibility testing. It is limited for children, dialysis patients, endocarditis-specific decisions, automated-testing-only laboratories, and health systems where confirmatory PSSA testing is unavailable.

Strengths & Limitations

  • Strengths:
    • First randomised controlled trial directly comparing benzylpenicillin with anti-staphylococcal penicillins for adult PSSA bacteraemia.
    • International, pragmatic, multicentre design across 67 hospitals.
    • Embedded in a Bayesian adaptive platform capable of answering multiple S aureus bacteraemia questions efficiently.
    • Objective primary outcome of 90-day all-cause mortality.
    • Objective key safety outcome of AKI using serial creatinine measurements.
    • High infectious diseases consultation rates in both groups.
    • Required confirmatory penicillin-susceptibility testing, directly addressing the central historical safety concern.
    • Protocol-adherent analysis and missing-outcome sensitivity analyses were directionally consistent with the primary analysis.
  • Limitations:
    • Stopped before the PSSA silo crossed the prespecified >99% posterior probability non-inferiority threshold.
    • Open-label design, with clear evidence of clinician behaviour differences in antibiotic switching for perceived inefficacy.
    • Small PSSA sample size for a mortality endpoint: 281 randomised and 273 in the primary analysis.
    • Simple randomisation produced group-size imbalance and some baseline imbalance.
    • Most participants were recruited in high-income settings.
    • Patients receiving maintenance dialysis and children were not represented in the adult analysis.
    • Endocarditis subgroup was too small for confident inference.
    • Does not answer benzylpenicillin versus cefazolin for confirmed PSSA bacteraemia.
    • Cloxacillin dosing at 12 g/day without renal adjustment may limit generalisability to lower-dose cloxacillin practice.
    • Most AKI was stage 1, although renal replacement therapy and serious AKI-related adverse reactions were numerically more common with flucloxacillin or cloxacillin.

Interpretation & Why It Matters

  • Clinical practice
    • For adults with true PSSA bacteraemia confirmed by reliable susceptibility testing, benzylpenicillin should generally be preferred to flucloxacillin or cloxacillin as the intravenous backbone.
    • The key practical reason is safety: AKI was 11% versus 22%, with no signal of worse 90-day mortality or treatment failure.
    • This should encourage laboratories and stewardship programmes to test and report penicillin susceptibility accurately rather than automatically treating all MSSA as penicillin-resistant.
  • Dogma tested
    • The trial challenges the assumption that serious S aureus infection requires a penicillinase-stable anti-staphylococcal penicillin even when the isolate is convincingly penicillin-susceptible.
    • The results support the safety of modern confirmatory PSSA testing as a treatment-enabling diagnostic step.
  • Methodological importance
    • SNAP demonstrates how an adaptive platform can address relatively uncommon pathogen-susceptibility subgroups that would be difficult to study in a conventional stand-alone trial.
    • The result is practice-shaping for confirmed PSSA bacteraemia, while the wider SNAP programme is likely to be more influential across S aureus bacteraemia as additional domains mature.2

Controversies & Other Evidence

  • Primary non-inferiority threshold:
    • The PSSA comparison did not meet its prespecified non-inferiority trigger for mortality: posterior probability 96.1% versus required >99%.
    • The clinical conclusion therefore rests on a high, but not threshold-crossing, probability of non-inferiority plus a consistent and clinically relevant AKI reduction.
  • Early stopping and effect-size inflation:
    • Recruitment stopped for safety rather than for a PSSA efficacy or non-inferiority declaration.
    • Early stopping reduced sample size and precision, and may exaggerate treatment effects.
    • The accompanying Lancet Comment nevertheless judged the observations important because mortality, AKI, intention-to-treat, protocol-adherent, and missing-data sensitivity analyses were directionally consistent.8
  • AKI interpretation:
    • Most AKI was mild: stage 1 occurred in 10/153 (7%) with benzylpenicillin and 18/124 (15%) with flucloxacillin or cloxacillin.
    • The safety signal should not be dismissed as purely biochemical because new renal replacement therapy was 3/156 (2%) versus 5/125 (4%), and serious adverse reactions involving AKI were also numerically more common with flucloxacillin or cloxacillin.
  • Diagnostic gatekeeping:
    • The trial’s conclusion depends on accurate PSSA identification; it should not be extrapolated to automated-testing-only susceptibility calls.
    • Modern EUCAST/CLSI disc methods with zone-edge interpretation are increasingly validated, but require trained laboratory interpretation and are not universally available.5
  • Guideline tension:
    • Older endocarditis guidance favoured anti-staphylococcal penicillins for MSSA endocarditis and discouraged penicillin for PSSA because of concern about missed penicillinase production.6
    • More recent WikiGuidelines work emphasised that many infective endocarditis recommendations still lack high-quality prospective controlled evidence, which is directly relevant to the very small endocarditis subgroup in this SNAP comparison.7
  • Benzylpenicillin versus cefazolin remains unresolved:
    • This trial tested benzylpenicillin against flucloxacillin or cloxacillin, not against cefazolin.
    • The accompanying Lancet Comment identified the missing benzylpenicillin-versus-cefazolin comparison as a major reason the long-term practice impact is uncertain.8
    • For MSSA bacteraemia, the SNAP cefazolin domain and the CloCeBa trial both support cefazolin as at least non-inferior to anti-staphylococcal penicillins and safer with respect to kidney injury or adverse events.910
    • A contemporary systematic review and meta-analysis also supported cefazolin as non-inferior for mortality and probably safer than anti-staphylococcal penicillins for MSSA bacteraemia, although that evidence was largely observational.11
  • Further reading and related evidence:
    • PSSA observational evidence: Henderson et al. reported 30-day mortality 10.5% with benzylpenicillin versus 14.2% with flucloxacillin in 915 patients; Chua et al. reported comparable treatment failure but faster microbiological clearance with benzylpenicillin in a 335-patient ten-year cohort.34
    • Platform-methods evidence: the SNAP protocol explains the trial’s disease-syndrome platform structure, shared primary outcome, silo-specific backbone comparisons, and repeated Bayesian updating.2
    • Guidelines and consensus documents: AHA infective endocarditis guidance remains important historical context, while WikiGuidelines illustrates the limited prospective evidence base for several endocarditis treatment questions.67

Summary

  • SNAP compared intravenous benzylpenicillin with intravenous flucloxacillin or cloxacillin in 281 adults with confirmed PSSA bacteraemia.
  • The trial stopped early for an AKI safety signal in the flucloxacillin or cloxacillin group before the PSSA non-inferiority stopping threshold was met.
  • 90-day mortality was 21/152 (14%) with benzylpenicillin versus 26/121 (21%) with flucloxacillin or cloxacillin; adjusted OR 0.67; 95% CrI 0.35 to 1.28; posterior probability of non-inferiority 96.1%.
  • AKI was lower with benzylpenicillin: 17/153 (11%) versus 27/124 (22%); adjusted OR 0.50; 95% CrI 0.26 to 0.94.
  • The findings support benzylpenicillin as preferred treatment for adult PSSA bacteraemia when susceptibility is confirmed by reliable phenotypic testing or blaZ PCR, but do not answer benzylpenicillin versus cefazolin.

Overall Takeaway

SNAP is practice-shaping for a precisely defined group: adults with confirmed PSSA bacteraemia. It does not provide a definitive mortality declaration by its own stringent Bayesian threshold, but the combination of compatible mortality, less AKI, and modern confirmatory susceptibility testing makes benzylpenicillin the most defensible backbone over flucloxacillin or cloxacillin in this setting.

Overall Summary

  • Use benzylpenicillin rather than flucloxacillin or cloxacillin for adult PSSA bacteraemia when PSSA is confirmed by reliable testing.
  • Do not apply this result to automated penicillin-susceptibility calls without confirmatory testing.
  • The mortality result is probabilistically reassuring but not a formal crossing of the prespecified non-inferiority threshold.
  • The clearest patient-centred advantage is reduced AKI.

Bibliography

Added June 18th, 2026