Skip to main content

Trial Summaries & Critiques

SNAP - Cephazolin

Lee TC, Barina LA, Walls G, et al; Staphylococcus aureus Network Adaptive Platform (SNAP) Trial Group. Cefazolin for methicillin-susceptible Staphylococcus aureus bacteremia. N Engl J Med. 2026;394:2329-2339

Image: Shutterstock / Nandani-Bridglal

Staphlococcus Aureus on gram stain

Publication

Context & Rationale

  • Background
    • Staphylococcus aureus bacteraemia is common, invasive, and associated with substantial early and 90-day mortality.
    • Despite its frequency and severity, the management of S aureus bacteraemia has historically rested on a small randomised evidence base and substantial international practice variation.1
    • For methicillin-susceptible S aureus (MSSA), definitive therapy has usually meant either an anti-staphylococcal penicillin, such as flucloxacillin, cloxacillin, nafcillin, or oxacillin, or cefazolin.
    • Older expert guidance, particularly for MSSA endocarditis, favoured anti-staphylococcal penicillins over cefazolin because of concern that cefazolin could be hydrolysed by staphylococcal beta-lactamase at high bacterial burden, the so-called cefazolin inoculum effect.23
    • Cefazolin remained attractive because it is easier to administer, has less protein binding, is generally better tolerated, and has repeatedly looked at least comparable to anti-staphylococcal penicillins in observational studies.
    • A contemporary systematic review and meta-analysis found cefazolin associated with lower 30-day mortality than anti-staphylococcal penicillins in observational data, but the evidence remained vulnerable to confounding, selection bias, and differences between individual comparator drugs.4
  • Research Question/Hypothesis
    • The trial asked whether intravenous cefazolin was non-inferior to intravenous flucloxacillin or cloxacillin for 90-day all-cause mortality in adults with penicillin-resistant MSSA bacteraemia.
    • The trial also tested whether cefazolin reduced clinically important harms, especially acute kidney injury and treatment discontinuation due to adverse events.
    • The comparison was embedded within the antibiotic-backbone domain of the broader Bayesian adaptive SNAP platform, which was designed to answer multiple S aureus bacteraemia treatment questions efficiently and concurrently.56
  • Why This Matters
    • MSSA bacteraemia is a daily clinical problem across infectious diseases, intensive care, acute medicine, nephrology, haematology, vascular surgery, orthopaedics, and outpatient parenteral antimicrobial therapy services.
    • The treatment decision is not marginal: anti-staphylococcal penicillins are associated with renal, hepatic, electrolyte, and infusion-related toxicity, while cefazolin is simpler to dose and often easier to deliver safely.
    • A large randomised trial was needed because observational studies could not reliably separate antibiotic effect from severity, source control, immortal time, drug discontinuation, and clinician selection.
    • The answer could change first-line MSSA bacteraemia therapy across health systems where anti-staphylococcal penicillins remain the default.

Design & Methods

  • Research Question: In adults with penicillin-resistant MSSA bacteraemia, is intravenous cefazolin non-inferior to intravenous flucloxacillin or cloxacillin for all-cause mortality at 90 days, and is it safer?
  • Study Type: Investigator-initiated, international, multicentre, pragmatic, open-label, randomised, Bayesian adaptive-platform, non-inferiority comparison within the MSSA silo of SNAP’s antibiotic-backbone domain.
  • Population:
    • Adults aged ≥18 years admitted to hospital with S aureus bacteraemia.
    • The adult analysis was conducted at 91 sites in eight countries: Australia, Canada, Israel, the Netherlands, New Zealand, Singapore, South Africa, and the United Kingdom.
    • Participants had to enter the platform within 72 h after collection of the index blood culture.
    • The analysed silo was restricted to penicillin-resistant, methicillin-susceptible S aureus to avoid bias towards cefazolin non-inferiority from including penicillin-susceptible isolates that could not have a blaZ-mediated cefazolin inoculum effect.
    • Susceptibility was determined locally using automated testing, phenotypic disc diffusion, molecular testing for mecA and blaZ, or combinations of these methods.
    • Key platform exclusions included anticipated platform entry >72 h after index blood-culture collection, polymicrobial bacteraemia, previous SNAP participation, a recent same-silo S aureus bacteraemia episode, imminent and inevitable death, death since index blood-culture collection, or goals of care excluding active antibiotic therapy.
    • Key domain-specific exclusions included active allergy to penicillin or cefazolin, maintenance haemodialysis or peritoneal dialysis, and ongoing systemic antibacterial therapy active against S aureus that could not be stopped or substituted.
    • Of 8546 screened patients, 2602 adults were enrolled in the platform, 1757 had MSSA, and 1341 underwent randomisation in the MSSA silo: 671 to cefazolin and 670 to flucloxacillin or cloxacillin.
  • Intervention:
    • Intravenous cefazolin.
    • Standard recommended dose: 2 g IV every 8 h.
    • Recommended high-dose regimen for critical illness, endocarditis, or central nervous system infection: 2 g IV every 6 h.
    • Renal adjustment: estimated GFR >40 mL/min, 2 g every 8 h; GFR 20-40 mL/min, 2 g every 12 h; GFR <20 mL/min, 1 g every 24 h; continuous renal replacement therapy, 2 g every 12 h.
    • Continuous infusion at 75-100% of the total daily dose was allowed according to local practice.
    • The assigned antibiotic was recommended for the full duration of intravenous antibiotic therapy, with minimum treatment durations of 14 days for uncomplicated bacteraemia and 28-42 days for complicated bacteraemia.
  • Comparison:
    • Intravenous flucloxacillin or cloxacillin.
    • Flucloxacillin standard recommended dose: 2 g IV every 6 h.
    • Flucloxacillin high-dose regimen for critical illness, endocarditis, or central nervous system infection: 2 g IV every 4 h.
    • Cloxacillin was used where flucloxacillin was unavailable, specifically Canada, Israel, Singapore, and South Africa.
    • Cloxacillin recommended dose: 2 g IV every 4 h.
    • Cloxacillin renal dose adjustment was not recommended.
    • Post-enrolment additional antibiotics active against S aureus were discouraged unless participants were also allocated to another relevant SNAP domain, such as adjunctive antibiotics.
    • Specialist infectious diseases consultation was advised.
    • Subsequent randomisation to SNAP’s early-oral-switch domain was permitted for eligible participants at 7 days for uncomplicated bacteraemia and 14 days for complicated bacteraemia.
  • Blinding: Open-label. Participants, treating clinicians, and trial staff knew the assigned antibiotic after MSSA-silo eligibility was confirmed. Aggregate outcome analyses remained restricted to the unblinded analytic team and the DSMC until the silo was closed. The primary outcome was objective, but clinician-driven decisions such as antibiotic switching and diagnostic/source-control intensity remained exposed to open-label behaviour.
  • Statistics: There was no conventional fixed-sample frequentist power calculation with a single alpha, beta, and final sample size. The Bayesian adaptive platform used simulations to set operating characteristics, with a maximum sample size of 6000 platform patients for this comparison, scheduled analyses after every 500 platform participants reached 90-day follow-up, and an approximate frequentist type I error of 2%. The non-inferiority margin was an adjusted OR <1.20 for 90-day mortality, corresponding to an absolute mortality difference of less than approximately 2.5 percentage points if comparator mortality was 15%. The adult MSSA silo could stop for cefazolin non-inferiority if the posterior probability of non-inferiority exceeded 99%, or for futility if the posterior probability of superiority was less than 1%. The primary analysis followed the intention-to-treat principle using a complete-case primary-outcome population; a protocol-adherent analysis and missing-data sensitivity analyses were also performed.
  • Follow-Up Period: 90 days after platform entry, with interim trial data collection through platform days 14, 28, 42, and 90.

Key Results

This trial was stopped early. At the fourth planned interim analysis, recruitment in the MSSA silo was paused because of an acute kidney injury safety signal; on August 5, 2024 the DSMC recommended closure because the cefazolin non-inferiority trigger had been met and the safety signal persisted, and the trial steering committee accepted this on August 7, 2024.

Outcome Cefazolin Flucloxacillin or cloxacillin Effect p value / 95% CI Notes
90-day all-cause mortality, primary complete-case analysis 97/645 (15.0%) 109/642 (17.0%) Adjusted OR 0.81 95% CrI 0.59 to 1.12; no p value calculated Posterior probability of non-inferiority 99.2%; superiority 89.8%. Prespecified non-inferiority trigger crossed.
90-day mortality, all missing primary outcomes assumed dead 123/671 (18.3%) 137/670 (20.4%) Adjusted OR 0.83 95% CrI 0.63 to 1.10; no p value calculated Posterior probability of non-inferiority 99.6%; superiority 90.7%.
90-day mortality, all missing primary outcomes assumed alive 97/671 (14.5%) 109/670 (16.3%) Adjusted OR 0.81 95% CrI 0.59 to 1.11; no p value calculated Posterior probability of non-inferiority 99.3%; superiority 90.5%.
90-day mortality, protocol-adherent population 81/589 (13.8%) 73/520 (14.0%) Adjusted OR 0.88 95% CrI 0.61 to 1.26; no p value calculated Posterior probability of non-inferiority 95.4%; superiority 76.0%.
Acute kidney injury within 14 days 92/660 (13.9%) 127/648 (19.6%) Adjusted OR 0.67 95% CrI 0.50 to 0.89; no p value calculated Posterior probability of non-inferiority >99.9%; superiority 99.7%.
AKI severity Stage 1: 66/671 (9.8%); Stage 2: 14/671 (2.1%); Stage 3: 12/671 (1.8%) Stage 1: 89/666 (13.4%); Stage 2: 19/666 (2.9%); Stage 3: 19/666 (2.9%) Not modelled as ordinal in main table Not reported Missing AKI-stage data: 11/671 (1.6%) versus 18/666 (2.7%).
Any serious adverse reaction related to trial drug 12/671 (1.8%) 33/670 (4.9%) Adjusted OR 0.41 95% CrI 0.21 to 0.77; no p value calculated Posterior probability of non-inferiority 99.9%; superiority 99.8%. Supplement: 14 versus 37 serious adverse reaction events.
Acute liver injury within 14 days 75/574 (13.1%) 78/562 (13.9%) Adjusted OR 0.96 95% CrI 0.68 to 1.35; no p value calculated ALT or GGT above 2.5 times upper limit of normal.
Death by day 14 25/667 (3.7%) 37/665 (5.6%) Adjusted OR 0.67 95% CrI 0.39 to 1.11; no p value calculated Posterior probabilities not reported in the table.
Death by day 28 47/665 (7.1%) 70/665 (10.5%) Adjusted OR 0.61 95% CrI 0.41 to 0.91; no p value calculated Direction favoured cefazolin.
Death by day 42 63/663 (9.5%) 86/662 (13.0%) Adjusted OR 0.66 95% CrI 0.46 to 0.94; no p value calculated Direction favoured cefazolin.
Change of antibiotic due to perceived inefficacy 14/671 (2.1%) 16/670 (2.4%) Adjusted OR 1.05 95% CrI 0.51 to 2.10; no p value calculated Low and similar rates, despite open-label design.
Microbiological treatment failure after day 14 18/587 (3.1%) 12/566 (2.1%) Adjusted OR 1.41 95% CrI 0.71 to 2.78; no p value calculated Low event rates; secondary outcome not adjusted for multiplicity.
New focus of infection after day 14 28/623 (4.5%) 30/609 (4.9%) Adjusted OR 0.87 95% CrI 0.53 to 1.45; no p value calculated No clear separation.
Change of antibiotic due to adverse event 11/671 (1.6%) 61/670 (9.1%) Adjusted OR 0.21 95% CrI 0.11 to 0.38; no p value calculated Large safety and tolerability separation favouring cefazolin.
New renal replacement therapy within 90 days 17/668 (2.5%) 27/657 (4.1%) Adjusted OR 0.61 95% CrI 0.33 to 1.11; no p value calculated Posterior probability of lower risk with cefazolin 94.6%.
Ongoing renal replacement therapy at 90 days 4/662 (0.6%) 3/643 (0.5%) Adjusted OR 1.00 95% CrI 0.31 to 3.16; no p value calculated Rare event.
Clostridioides difficile infection 14/664 (2.1%) 10/661 (1.5%) Adjusted OR 1.28 95% CrI 0.60 to 2.66; no p value calculated No clear separation.
Intravenous catheter complication leading to removal 22/668 (3.3%) 33/665 (5.0%) Adjusted OR 0.68 95% CrI 0.39 to 1.16; no p value calculated Direction favoured cefazolin.
Endocarditis subgroup: 90-day mortality 8/51 (15.7%) 13/53 (24.5%) Adjusted OR 0.54 95% CrI 0.20 to 1.43; no p value calculated Posterior probability of non-inferiority 94.4%; superiority 88.9%. Subgroup was underpowered.
Post hoc timing subgroup: enrolled 0-48 h after index blood culture 44/255 (17.3%) died by 90 days 49/273 (17.9%) died by 90 days Not reported Not reported Descriptive post hoc subgroup only.
Post hoc timing subgroup: enrolled 48-72 h after index blood culture 53/390 (13.6%) died by 90 days 60/369 (16.3%) died by 90 days Not reported Not reported Descriptive post hoc subgroup only.
  • Cefazolin crossed the prespecified Bayesian non-inferiority threshold for 90-day mortality: 15.0% versus 17.0%; adjusted OR 0.81; 95% CrI 0.59 to 1.12; posterior probability of non-inferiority 99.2%.
  • The clinically clearest difference was safety: AKI was 13.9% with cefazolin versus 19.6% with flucloxacillin or cloxacillin, and antibiotic discontinuation due to adverse events was 1.6% versus 9.1%.
  • The endocarditis subgroup was directionally reassuring but too small to settle the cefazolin inoculum-effect question in high-inoculum infection; only seven participants had central nervous system infection, precluding meaningful comparison.

Internal Validity

  • Randomisation and Allocation:
    • Participants were randomised 1:1 within the MSSA silo at platform entry.
    • Randomisation used a web-based interactive randomisation system.
    • Simple randomisation without stratification was used.
    • Allocation was concealed until eligibility for the MSSA silo of the backbone domain had been confirmed.
    • The final allocation was almost exactly balanced: 671 cefazolin and 670 flucloxacillin or cloxacillin.
  • Dropout or Exclusions:
    • Of 8546 patients screened, 2602 adults were enrolled in the platform.
    • Of 1757 adults with MSSA, 1341 underwent randomisation in the MSSA silo.
    • Reasons adults with MSSA were not randomised included no consent in 111, site not participating in the relevant trial in 9, and ineligibility in 296.
    • Common ineligibility reasons included >72 h since index blood-culture collection in 119, maintenance dialysis in 115, non-severe cefazolin or penicillin reaction in 62, exclusion in the patient’s interest in 58, type I hypersensitivity in 35, severe delayed reaction in 16, death in 6, and inability to discontinue another systemic antibacterial agent in 4.
    • Primary outcome data were missing for 54/1341 participants (4.0%): 26/671 in the cefazolin group and 28/670 in the flucloxacillin or cloxacillin group.
    • In the cefazolin group, 5 withdrew from data collection, 3 were lost to follow-up before day 42, and 18 were lost to follow-up at or after day 42.
    • In the flucloxacillin or cloxacillin group, 5 withdrew from data collection, 2 were lost to follow-up before day 42, and 21 were lost to follow-up at or after day 42.
    • Patients with missing primary-outcome data were younger and more likely to have recent injection drug use than those with complete data, but missing-data sensitivity analyses gave posterior probabilities of non-inferiority above 99%.
  • Performance/Detection Bias:
    • The trial was open-label.
    • Open-label management could influence diagnostic intensity, source-control decisions, antibiotic switching, and adverse-event attribution.
    • The primary outcome, all-cause mortality at 90 days, was objective and minimally susceptible to detection bias.
    • The key safety outcome, AKI, was based on serum creatinine criteria and was also largely objective.
    • More interpretive outcomes, such as new foci of infection and antibiotic change due to perceived inefficacy, were more vulnerable to clinician judgement.
  • Protocol Adherence:
    • The protocol-adherent primary-outcome population included 589/671 (87.8%) assigned to cefazolin and 520/670 (77.6%) assigned to flucloxacillin or cloxacillin.
    • Protocol-adherent mortality was 81/589 (13.8%) with cefazolin versus 73/520 (14.0%) with flucloxacillin or cloxacillin; adjusted OR 0.88; 95% CrI 0.61 to 1.26.
    • Protocol-adherent posterior probability of non-inferiority was 95.4%, lower than the primary complete-case estimate.
    • Protocol deviations were similar: 82/671 (12.2%) with cefazolin and 77/670 (11.5%) with flucloxacillin or cloxacillin.
    • Incorrect administration, duration, route, cessation, or non-administration of study drug occurred in 13/671 (1.9%) versus 17/670 (2.5%).
    • Randomisation of an ineligible participant occurred in 9/671 (1.3%) versus 11/670 (1.6%).
  • Baseline Characteristics:
    • Median age was 66 years in both groups: 66 (IQR 53-76) with cefazolin and 66 (IQR 52-77) with flucloxacillin or cloxacillin.
    • Female sex was 215/671 (32.0%) versus 206/670 (30.7%).
    • Median weight was 82.0 kg (IQR 69.0-96.8) versus 80.0 kg (IQR 68.0-95.0).
    • Diabetes mellitus was present in 231/671 (34.4%) versus 217/670 (32.4%).
    • Chronic kidney disease was present in 96/671 (14.3%) versus 93/670 (13.9%).
    • ICU admission at recruitment was 74/671 (11.0%) versus 72/670 (10.7%).
    • Median Pitt bacteraemia score was 0 (IQR 0-1) in both groups.
    • Median C-reactive protein was 192.0 mg/L (IQR 113.5-266.5) versus 203.1 mg/L (IQR 122.0-283.8).
    • Endocarditis was present in 55/671 (8.2%) versus 57/670 (8.5%).
    • Recent injection drug use was 31/671 (4.6%) versus 52/670 (7.8%).
  • Heterogeneity:
    • The trial included 91 sites across eight countries, improving pragmatic relevance but introducing site-level variation.
    • Country distribution was not identical but broadly balanced: Australia 282/671 (42.0%) versus 310/670 (46.3%); Canada 191/671 (28.5%) versus 181/670 (27.0%); New Zealand 99/671 (14.8%) versus 102/670 (15.2%).
    • Comparator heterogeneity was clinically important: flucloxacillin was used where available, while cloxacillin was used in countries where flucloxacillin was unavailable.
    • The statistical model adjusted for age, country, temporal epoch, and eligibility and assignment in other platform domains.
    • Post-randomisation care such as source-control intensity and additional diagnostic testing was left to clinicians and not fully collected, which limits mechanistic interpretation.
  • Timing:
    • Participants had to be enrolled within 72 h of index blood-culture collection.
    • Median time from index blood culture to platform entry was 52.3 h (IQR 43.5-65.5) with cefazolin and 50.9 h (IQR 42.3-64.7) with flucloxacillin or cloxacillin.
    • Median time from index blood culture to MSSA silo entry was 62.2 h (IQR 50.2-67.9) versus 61.1 h (IQR 50.0-67.5).
    • The timing was clinically appropriate for definitive therapy after organism and susceptibility information became available, but most participants had already received active empirical antibiotics.
  • Dose:
    • Cefazolin 2 g every 8 h, or 2 g every 6 h for critical illness or endocarditis, is a clinically credible adult MSSA bacteraemia regimen.
    • Flucloxacillin 2 g every 6 h, or 2 g every 4 h for critical illness or endocarditis, is a standard high-intensity anti-staphylococcal penicillin strategy.
    • Cloxacillin 2 g every 4 h delivered 12 g/day and was not renally adjusted, which may affect interpretation of toxicity in settings using lower cloxacillin exposure or different renal-adjustment practice.
    • Renal dose adjustment for cefazolin was clearly specified.
  • Separation of the Variable of Interest:
    • Before platform entry, cefazolin had been administered to 249/671 (37.1%) subsequently assigned to cefazolin and 239/670 (35.7%) subsequently assigned to flucloxacillin or cloxacillin.
    • Before platform entry, flucloxacillin had been administered to 323/671 (48.1%) subsequently assigned to cefazolin and 314/670 (46.9%) subsequently assigned to flucloxacillin or cloxacillin.
    • Before platform entry, vancomycin had been administered to 303/671 (45.2%) in each group.
    • Median IV antibiotic duration was 26.0 days (IQR 14.0-42.0) with cefazolin and 26.0 days (IQR 14.0-41.0) with flucloxacillin or cloxacillin.
    • Median oral antibiotic duration was 7.0 days (IQR 0.0-28.0) in both groups.
    • Antibiotic change due to adverse event was 11/671 (1.6%) with cefazolin versus 61/670 (9.1%) with flucloxacillin or cloxacillin.
    • Antibiotic change due to perceived inefficacy was 14/671 (2.1%) versus 16/670 (2.4%).
  • Key Delivery Aspects:
    • More than 99% of participants received specialist infectious diseases consultation.
    • Allocated treatment was embedded in usual clinical care rather than delivered by a tightly controlled efficacy-trial protocol.
    • Source control, echocardiography, imaging, and other clinical decisions were at clinician discretion.
    • The pragmatic design increases relevance but reduces control over post-randomisation management.
  • Crossover:
    • Formal crossover counts beyond captured antibiotic changes were not reported.
    • The protocol-adherent definition allowed no more than one day of the alternative study antibiotic during the adherence assessment period.
    • The main measurable separation issue was drug discontinuation due to adverse events, which was much more common with flucloxacillin or cloxacillin.
  • Adjunctive Therapy Use:
    • Participants could be enrolled in other SNAP domains.
    • Eligibility and assignment in other platform domains were included in the Bayesian model.
    • Detailed concurrent-domain results were not reported to preserve ongoing platform integrity.
  • Outcome Assessment:
    • Death at 90 days was determined through hospital records, community healthcare-provider follow-up, patient or carer contact, or linkage with death registries.
    • AKI used prespecified serum creatinine criteria and systematically collected renal-function data.
    • Acute liver injury used prespecified ALT or GGT thresholds.
    • New infectious foci and antibiotic change for inefficacy remained more subjective and were assessed by site staff aware of treatment allocation.
  • Statistical Rigor:
    • The Bayesian hierarchical model matched the adaptive platform structure.
    • The stopping threshold for cefazolin non-inferiority was prespecified and crossed.
    • Model diagnostics in the supplement reported satisfactory convergence for the primary model, including all R-hat values <1.01.
    • No p values were calculated.
    • No adjustment for multiplicity was prespecified for secondary outcomes; therefore, non-safety secondary credible intervals should not be treated as definitive.
    • The planned Bayesian-Weibull time-to-event model was not used because assumptions were not satisfied; unadjusted Kaplan-Meier curves were presented instead.

Conclusion on Internal Validity: Internal validity is strong for the primary conclusion that cefazolin is non-inferior to flucloxacillin or cloxacillin for 90-day mortality in adult MSSA bacteraemia. The main caveats are open-label management, incomplete capture of post-randomisation source-control and diagnostic decisions, missing primary outcome data in 4.0%, and a protocol-adherent analysis that was directionally compatible but less statistically decisive than the primary analysis.

External Validity

  • Population Representativeness:
    • The study population was typical for adult S aureus bacteraemia by age and sex: median age 66 years and 421/1341 (31.4%) female sex at birth.
    • The most common focus was osteoarticular infection: 430/1341 (32.1%).
    • Skin and soft-tissue infection was present in 381/1341 (28.4%).
    • Intravascular catheter-associated infection was present in 205/1341 (15.3%).
    • Endocarditis was present in 112/1341 (8.4%).
    • Central nervous system infection was rare: 7/1341 (0.5%).
    • People who inject drugs were under-represented: 83/1341 (6.2%), compared with approximately 10% in recent S aureus bacteraemia cohorts.
    • Most participants came from high-income countries: 1337/1341, with 4 participants from South Africa.
    • Children were not included in this adult analysis; the paediatric SNAP trial was ongoing.
    • Patients on maintenance dialysis were excluded, limiting applicability to a high-risk bacteraemia subgroup.
  • Applicability:
    • The findings apply directly to adults with penicillin-resistant MSSA bacteraemia managed in hospitals where cefazolin and anti-staphylococcal penicillins are available.
    • The results do not apply to MRSA bacteraemia, penicillin-susceptible S aureus bacteraemia, children, or patients receiving maintenance dialysis.
    • The comparator drugs were flucloxacillin and cloxacillin, not nafcillin or oxacillin; extrapolation to nafcillin or oxacillin is biologically plausible but indirect.
    • The endocarditis subgroup was clinically important but underpowered.
    • The trial is highly relevant to antimicrobial stewardship because cefazolin is generally easier to administer and safer, but it may have broader Gram-negative activity than anti-staphylococcal penicillins and local ecological effects were not assessed.
    • Generalisability to resource-limited settings is constrained by the small number of participants from upper-middle-income countries and none from low-income countries.

Conclusion on External Validity: External validity is strong for adult hospitalised MSSA bacteraemia in high-income health systems with infectious diseases input. It is more limited for children, maintenance dialysis, low-resource systems, central nervous system infection, endocarditis-specific decisions, and centres using nafcillin or oxacillin rather than flucloxacillin or cloxacillin.

Strengths & Limitations

  • Strengths:
    • Largest randomised comparison of cefazolin versus anti-staphylococcal penicillins for adult MSSA bacteraemia.
    • International enrolment across 91 sites in eight countries.
    • Embedded in a prespecified Bayesian adaptive-platform architecture.
    • Clinically meaningful primary endpoint: all-cause mortality at 90 days.
    • Stringent non-inferiority margin for mortality: adjusted OR <1.20, approximately <2.5% absolute difference if comparator mortality was 15%.
    • Objective key safety endpoint of AKI.
    • Clear and consistent safety advantage for cefazolin across AKI, serious adverse reactions, and antibiotic discontinuation due to adverse events.
    • High infectious diseases consultation rates.
    • Pragmatic design reflected routine clinical practice.
    • Missing-outcome sensitivity analyses were congruent with the primary analysis.
  • Limitations:
    • Open-label design.
    • Post-randomisation clinical management, diagnostic intensity, source control, and antibiotic additions were not fully standardised or collected.
    • Primary outcome analysis excluded 54/1341 (4.0%) participants with missing 90-day mortality data.
    • Protocol-adherent posterior probability of non-inferiority was 95.4%, below the primary-analysis posterior probability and below the stopping threshold.
    • No p values were calculated, consistent with the Bayesian design.
    • Secondary outcomes were not adjusted for multiplicity.
    • Central laboratory testing for the cefazolin inoculum effect had not yet been completed or reported.
    • Endocarditis subgroup inference was imprecise.
    • Central nervous system infection was too rare for inference.
    • Maintenance dialysis patients and paediatric patients were not represented in this adult analysis.
    • Most participants came from high-income countries.
    • The trial does not directly compare cefazolin with nafcillin or oxacillin.

Interpretation & Why It Matters

  • Clinical practice
    • For most adults with MSSA bacteraemia, cefazolin should be considered a preferred intravenous backbone over flucloxacillin or cloxacillin.
    • The mortality result met the trial’s prespecified Bayesian non-inferiority threshold.
    • The safety result is clinically compelling: AKI, serious adverse reactions, and adverse-event-related antibiotic discontinuation all favoured cefazolin.
    • In critical care and acute medicine, where renal vulnerability is common, a reduction in nephrotoxicity is not a minor tolerability issue; it affects fluid balance, drug dosing, contrast decisions, renal replacement therapy risk, and downstream morbidity.
  • Dogma tested
    • The trial directly challenges the long-standing preference for anti-staphylococcal penicillins as the default definitive therapy for MSSA bacteraemia.
    • The cefazolin inoculum effect remains biologically real, but SNAP shows that, across a broad adult MSSA bacteraemia population, it did not translate into excess 90-day mortality with cefazolin.
    • The endocarditis subgroup was directionally consistent with the overall result but too small to abolish uncertainty in the highest-inoculum syndromes.
  • Methodological importance
    • SNAP demonstrates that platform trials can produce large, practice-changing randomised evidence for infectious diseases questions that had previously been left to observational cohorts and expert preference.
    • The adaptive design allowed closure once a prespecified non-inferiority threshold and safety signal were established, while the wider platform continued to answer other S aureus bacteraemia questions.56

Controversies & Subsequent Evidence

  • Cefazolin inoculum effect:
    • The central unresolved biological issue is whether a subset of blaZ-positive MSSA isolates with high-inoculum cefazolin hydrolysis causes worse outcomes in endocarditis, deep abscess, osteomyelitis, or other high-burden infections.
    • Earlier clinical work linked the cefazolin inoculum effect to increased mortality in MSSA bacteraemia, supporting the historical caution around cefazolin for high-inoculum disease.7
    • A 2024 systematic review concluded that the clinical significance of the cefazolin inoculum effect in serious MSSA infection remained uncertain, reflecting heterogeneity in testing methods, syndromes, and outcomes.8
    • A 2024 multicentre infective-endocarditis case series found that a beta-lactam inoculum effect to the beta-lactam received was associated with first-month mortality, keeping this issue clinically relevant for endocarditis even after SNAP.9
    • SNAP’s endocarditis subgroup did not show harm with cefazolin, but 104 patients with endocarditis cannot close the question definitively.
    • SNAP’s central isolate-level inoculum-effect analysis had not yet been reported in the index publication.
  • Non-inferiority versus superiority:
    • The primary declaration was non-inferiority, not formal superiority.
    • The posterior probability of superiority for 90-day mortality was 89.8%, which is suggestive but not a formal superiority declaration.
    • The early mortality time points favoured cefazolin numerically and probabilistically, but secondary mortality outcomes were not multiplicity-adjusted.
  • Protocol-adherent analysis:
    • The protocol-adherent population showed 90-day mortality of 13.8% with cefazolin versus 14.0% with flucloxacillin or cloxacillin.
    • The protocol-adherent posterior probability of non-inferiority was 95.4%, below the prespecified primary stopping threshold.
    • This does not overturn the intention-to-treat result, but it does temper any claim that the mortality result is unequivocally superior.
  • Comparator choice:
    • SNAP compared cefazolin with flucloxacillin or cloxacillin, not nafcillin or oxacillin.
    • Because anti-staphylococcal penicillins share antibacterial class activity but differ in pharmacokinetics, dosing, and toxicity, direct extrapolation to nafcillin and oxacillin should be made cautiously.
    • The clinical direction is nevertheless consistent with the broader observational literature comparing cefazolin with individual anti-staphylococcal penicillins.4
  • CloCeBa and convergent randomised evidence:
    • The French CloCeBa trial independently compared cefazolin with cloxacillin for MSSA bacteraemia and found cefazolin non-inferior for a 90-day therapeutic-success composite: 109/146 (75%) with cefazolin versus 108/146 (74%) with cloxacillin; treatment difference -1%; 95% CI -11 to 9; p=0.012 for non-inferiority.10
    • CloCeBa also found fewer serious adverse events by end of study treatment with cefazolin: 22/146 (15%) versus 40/146 (27%); p=0.010.10
    • AKI in CloCeBa occurred in 1/134 (1%) with cefazolin versus 15/128 (12%) with cloxacillin; p=0.0002, strongly reinforcing SNAP’s renal-safety signal.10
    • CloCeBa was smaller, excluded some high-risk subgroups, and used a composite efficacy endpoint, while SNAP was larger and used 90-day all-cause mortality as the primary outcome.
  • Guideline tension:
    • Older endocarditis guidance and expert practice patterns favoured anti-staphylococcal penicillins for MSSA endocarditis because of concern about cefazolin inoculum effect and historical experience with penicillinase-stable drugs.23
    • SNAP should shift the default for uncomplicated and many complicated MSSA bacteraemia syndromes towards cefazolin.
    • Endocarditis, central nervous system infection, retained prosthetic material, and isolates later shown to have clinically important inoculum-effect phenotypes remain areas where clinicians may reasonably want more granular evidence.
  • Further reading and related evidence:
    • Global practice variation and lack of a unified international standard of care for S aureus bacteraemia are documented by Westgeest et al.1
    • The SNAP master protocol and Bayesian statistical blueprint explain how the platform handles multiple domains, adult and paediatric subgroups, repeated interim analyses, and decision thresholds.56
    • The most relevant non-SNAP randomised comparator evidence is CloCeBa, which points in the same efficacy and safety direction as SNAP.10
    • The most relevant contemporary synthesis before SNAP is the 2025 cefazolin-versus-anti-staphylococcal-penicillin systematic review and meta-analysis.4

Summary

  • SNAP randomised 1341 adults with penicillin-resistant MSSA bacteraemia to cefazolin or flucloxacillin/cloxacillin in the antibiotic-backbone domain.
  • The MSSA silo was stopped after the fourth interim analysis because cefazolin crossed the prespecified non-inferiority threshold and an AKI safety signal persisted in the anti-staphylococcal penicillin group.
  • 90-day mortality was 97/645 (15.0%) with cefazolin versus 109/642 (17.0%) with flucloxacillin or cloxacillin; adjusted OR 0.81; 95% CrI 0.59 to 1.12; posterior probability of non-inferiority 99.2%.
  • AKI was lower with cefazolin: 92/660 (13.9%) versus 127/648 (19.6%); adjusted OR 0.67; 95% CrI 0.50 to 0.89.
  • Serious adverse reactions and antibiotic discontinuation due to adverse events were substantially less common with cefazolin.

Overall Takeaway

SNAP cefazolin is practice-shaping rather than merely confirmatory: it provides large, international, randomised evidence that cefazolin is non-inferior to flucloxacillin or cloxacillin for 90-day mortality in adult MSSA bacteraemia and is safer for the kidney. For most adult MSSA bacteraemia, cefazolin should now be the default intravenous backbone, with residual caution reserved for the small high-inoculum subgroups where isolate-level inoculum-effect data remain awaited.

Overall Summary

  • Cefazolin met the prespecified Bayesian non-inferiority threshold for 90-day mortality in adult MSSA bacteraemia.
  • Cefazolin caused less AKI and fewer serious drug-related adverse reactions than flucloxacillin or cloxacillin.
  • The result supports cefazolin as preferred treatment for most adults with MSSA bacteraemia.
  • Endocarditis and other high-inoculum infections remain important areas for further isolate-level and subgroup evidence.

Bibliography

Added June 19th, 2026