Introduction

Staphylococcus aureus is a leading cause of bloodstream infection (BSI) in both community and hospital settings1, with a mortality rate up to 30%2. The advent of penicillin quickly led to the emergence of penicillin-resistant S. aureus (PRSA) in the 1940s3, eventually constituting the vast majority of isolates4. In the search for antibiotics that are resistant to penicillinase, a new class of isoxazolyl penicillins was developed. These anti-staphylococcal penicillins (ASPs), such as cloxacillin, in addition to cephalosporins such as cefazolin, are now the antibiotics of choice in susceptible S. aureus infections5.

In recent years however, the incidence of penicillin-susceptible S. aureus (PSSA) has increased worldwide6,7,8,9, including in Asia10. At Tan Tock Seng Hospital (TTSH), a 1700-bed tertiary-care institution in Singapore, the proportion of PSSA among methicillin-susceptible S. aureus (MSSA) blood culture isolates averaged approximately 25% between 2011 and 2014, and rose to almost 30% in 2021. A possible reason for the resurgence of PSSA may be reduced selection pressure for penicillin resistance over the last decades9. This trend has drawn attention to the re-introduction of benzylpenicillin as definitive management for PSSA BSI11. Potential advantages of penicillin compared with the isoxazolyl penicillins as definitive treatment include more favorable pharmacokinetics/pharmacodynamics (PKPD) and fewer adverse events12. However, there are historical concerns of inducible penicillinase encoded by the blaZ gene7 and consequently, the possible suboptimal sensitivity of routine phenotypic methods in detecting benzylpenicillin resistance6,7,8,13,14,15,16.

Limited data is currently available on antibiotic choice and associated outcomes of PSSA BSI. All published studies are observational and have generally found no difference in treatment failure or mortality between benzylpenicillin and ASPs or cephalosporins12,17,18,19,20,21. However, two studies observed better treatment outcomes with benzylpenicillin when compared to ASPs22,23, and one other when compared to cefuroxime18. Additionally, an increased incidence of adverse drug events (ADEs) was reported with ASPs in comparison to benzylpenicillin12,17. Only three studies examined the association between clinical outcomes and potential risk factors for mortality and treatment failure, such as age, comorbidity, severity of infection, and sites of infection12,18,21. Reflecting this clinical equipoise, the Staphylococcus aureus bacteraemia Network Adaptive Platform Trial (SNAP) clinical trial is currently investigating treatment options for PSSA as one of its domains, but results are not expected for some time.

Therefore, this study aimed to: (i) compare clinical outcomes associated with benzylpenicillin versus non-benzylpenicillin beta-lactams (including ASPs and cephalosporins), and (ii) investigate risk factors influencing treatment failure in PSSA BSI.

Results

Inclusion and exclusion of patients

A total of 486 patients had first episode PSSA BSI from 1 January 2012 to December 2021. After evaluation of medical records, 151 patients were excluded based on the exclusion criteria. Thus, the final analysis included 335 patients (Fig. 1).

Fig. 1
figure 1

Inclusion and exclusion of patients.

Full size image

Demographic, clinical, and treatment characteristics of patients stratified by treatment groups

Table 1 details the demographic, clinical, and treatment characteristics of the 335 patients included in the final analysis, stratified by treatment group.

Table 1 Demographic, clinical, and treatment characteristics of patients stratified by benzylpenicillin and non-benzylpenicillin beta-lactams treatment group.
Full size table

Seventy-four (22.09%) patients received benzylpenicillin, while 261 (77.91%) patients received a non-benzylpenicillin beta-lactam as the predominant antibiotic. Within the non-benzylpenicillin beta-lactams group, 161 (48.06%) received cloxacillin, 75 (22.39%) received cefazolin, 11 (3.28%) received piperacillin/tazobactam, 7 (2.09%) received meropenem, 3 (0.90%) received amoxicillin/clavulanic acid, 2 (0.60%) received ertapenem, 1 (0.30%) received ceftriaxone, and 1 (0.30%) received cefepime. Overall, patients received a median duration of 14.00 days (IQR, 8.00–19.00) of predominant antibiotic therapy and 19.00 days (IQR, 16.00–43.00) of total active parenteral antibiotic therapy (data not shown).

There were more males in the benzylpenicillin group [75.68% vs. 61.69%; P = 0.03]. Patients in the benzylpenicillin group had a lower Charlson Comorbidity Index (CCI) score24 [2.00 (IQR, 1.00–4.00) vs. 4.00 (IQR, 3.00–6.00); P < 0.01] and a lower Pitt Bacteraemia Score (PBS)25 [0.00 (IQR, 0.00–1.00) vs. 0.00 (IQR, 0.00–1.00); P = 0.05] than those in the non-benzylpenicillin beta-lactams group. The proportion of patients with a PBS of 0 was 68.92% vs. 58.23% (data not shown). Between the benzylpenicillin and non-benzylpenicillin beta-lactams groups, the duration of predominant antibiotic therapy received [13.50 (IQR, 8.25–19.00) vs. 14.00 (IQR, 8.00–18.00)] as well as total active parenteral antibiotic therapy [17.50 (IQR, 16.00–39.75) vs. 20.00 (IQR, 16.00–44.00)] were similar (data not shown). The rest of the variables were not statistically different between the two groups.

Outcomes

Fifty-five (16.42%) out of the 335 patients experienced overall treatment failure, which comprised of 46 (13.73%) cases of 30-day all-cause mortality and 9 (2.69%) cases of 90-day relapse (Table 2). Overall treatment failure (13.51% vs. 17.24%; P = 0.45), 30-day all-cause mortality (9.46% vs. 14.94%; P = 0.23), and 90-day relapse (4.05% vs. 2.30%, P = 0.42) were all comparable between the definitive treatment groups (Table 2). The use of benzylpenicillin as the predominant definitive antibiotic was also not significantly associated with overall treatment failure in the univariate [odds ratio (OR) 0.75, 95% confidence interval (CI) 0.36–1.57; P = 0.45] and multivariate analyses [adjusted OR (aOR) 1.28, 95% CI 0.52–3.17; P = 0.60] (Table 3). However, time to microbiological clearance was shorter [3.00 days (IQR, 2.00–4.00 days) vs. 4.00 days (IQR, 3.00–5.00 days); P = 0.03] (Table 2) and incidence of persistent BSI was lower (22.97% vs. 36.02%; P = 0.04) in the benzylpenicillin group (Table 2). Additionally, a marginally lower incidence of ADEs (6.76% vs. 7.66%; P = 0.79) was observed in the benzylpenicillin group, but this difference was not significant (Table 2).

Table 2 Outcomes of patients in benzylpenicillin and non-benzylpenicillin beta-lactams treatment groups.
Full size table
Table 3 Univariate and multivariate analyses of association between various variables and overall treatment failure.
Full size table

In the multivariate analysis, older age (aOR 1.07, 95% CI 1.04–1.11; P < 0.01), higher CCI score (aOR 1.23, 95% CI 1.08–1.40; P < 0.01), and higher PBS (aOR 1.54, 95% CI 1.26–1.89; P < 0.01) were significantly associated with overall treatment failure (Table 3). Notably, an unknown site of infection (aOR 4.63, 95% CI 1.47–14.64; P < 0.01) was identified as an independent risk factor for overall treatment failure, whereas review by an Infectious Disease (ID) specialist (aOR 0.30, 95% CI 0.12–0.73; P = 0.01) was protective (Table 3). Because there were more patients with end-stage renal disease requiring haemodialysis in the non-benzylpenicillin beta-lactams group (23.37% vs. 13.51%; P = 0.07) (Table 1), we also performed a supplemental analysis which included haemodialysis status in the multivariate analysis. This did not change the results of the multivariate analysis and haemodialysis status was not associated with treatment failure (data not shown).

Discussion

We found no difference in overall treatment failure rates between benzylpenicillin and non-benzylpenicillin beta-lactams in the treatment of PSSA BSI, consistent with most previous studies12,17,18,19,20,21. Henderson et al., the most extensive study so far with over 900 patients, observed an increased 30-day mortality with flucloxacillin compared to benzylpenicillin22. Moreover, Aldhaeefi et al. noted the superiority of benzylpenicillin over nafcillin and oxacillin in terms of 60-day clinical failure23, and Nissen et al. found comparatively higher 30-day mortality with cefuroxime18. While we did not find evidence that benzylpenicillin was protective against overall treatment failure, faster microbiological clearance and reduced incidence of persistent BSI were observed. However, this interpretation should be taken cautiously as physicians’ preference for non-benzylpenicillin might be influenced by perceived greater infection severity or when factors causing prolonged BSI were present. This was evidenced by the lower median CCI score and PBS in the benzylpenicillin group, and the relatively smaller proportion of patients with abscesses or complicated sites of infection such as bone and joint or cardiovascular system.

It should be noted that the inconsistent results might be partly due to heterogenous methods of penicillinase detection across the previous studies, which have been shown to differ significantly in sensitivity. However, this was not a crucial consideration for our study as our microbiology laboratory used penicillin zone-edge testing with the CLSI P10 method and prior investigation by our institution had found a 99.1% accuracy in penicillinase detection with this method against blaZ PCR of 112 PSSA isolates21. An alternative explanation for better outcomes with benzylpenicillin could be its longer duration of free plasma concentration above the minimum inhibitory concentration after dosing (fT > MIC), attributed to a lower protein-bound fraction as well as an expected lower MIC compared to both ASPs and cephalosporins7,11,26.

We found that benzylpenicillin was not significantly associated with a lower incidence of ADEs compared to non-benzylpenicillin beta-lactams, contrary to our initial expectations and the existing studies on PSSA BSI. Hagstrand et al. reported a significantly higher number of ADEs, particularly diarrhoea and non-invasive Candida infections, with cloxacillin12, while Shah et al. found a significantly higher incidence of ADEs, mainly gastrointestinal intolerance, with nafcillin17. Other studies on S. aureus BSI have also frequently reported on ADEs associated with ASPs and cefazolin27,28,29. Nonetheless, benzylpenicillin offers several other advantages such as a narrower spectrum of activity compared to cephalosporins, as well as less evolutionary pressure for developing methicillin resistance11, which we did not examine in this study. Overall, this study adds to the evidence supporting the use of benzylpenicillin in PSSA BSI. However, given the small, retrospective nature and inconsistent results of the previous studies, there is a pressing need for more robust data. The ongoing SNAP trial will hopefully provide a definite answer regarding the optimal treatment for PSSA BSI.

Our study identified several independent risk factors for overall treatment failure in PSSA BSI, which agrees with the existing literature. Studies on S. aureus BSI have consistently found older age30 and having more comorbidities as indicated by a higher CCI score31 to be associated with increased mortality in S. aureus BSI. These were in line with our findings and the study by Hagstrand et al.12. Greater severity of illness, as measured by PBS, was also an independent risk factor for treatment failure, as was the case in the study by Nissen et al.18, despite some conflicting evidence against the use of PBS in S. aureus BSI32. Interestingly, Intensive Care Unit (ICU) admission, a surrogate marker for infection severity, was not an independent risk factor.

Notably, we found that the lack of review by an ID specialist and the absence of an identified site of infection were strong predictors of overall treatment failure. These two factors go hand in hand: unlocalised deep foci may have required longer therapy durations and possible surgical intervention for source eradication, while ID specialist review often improves localisation of sites of infection33. Both general S. aureus BSI literature33,34 and studies on PSSA BSI18,21 have identified unknown site of infection as an independent risk factor for treatment failure and mortality. Although our study was the first to find ID specialist review to be protective in PSSA BSI specifically, it has been consistently observed in S. aureus BSI33. The findings highlight the importance of having a low threshold for ID specialist review for S. aureus BSI and aggressive identification of sites of infection.

The main limitation of our study was its retrospective nature as we might not have identified all possible confounding factors. As previously mentioned, physicians might be biased against benzylpenicillin in more severe infections or in patients with more comorbidities. Comparative data was further constrained in the latter half of our study. We observed changes in our institutional ID blood culture advisory service recommendation for PSSA BSI, shifting from benzylpenicillin to cloxacillin starting in 2016. This led to a dramatic reduction in the number of use cases of benzylpenicillin, compounded by the relatively low prevalence of PSSA and physicians likely being accustomed to using ASPs for MSSA BSI 9. Additionally, we did not evaluate the appropriateness of dosing, and the duration of antibiotic therapy received could have affected the treatment outcomes. However, all instances of S. aureus BSI in our hospital were reviewed by the ID blood culture advisory service which should have accounted for variations in treatment regimes. The short median duration of total parenteral treatment in our study may suggest a higher proportion of uncomplicated cases. Therefore, the findings from this study may not fully generalise to complicated infections. We also did not assess if patients had other concomitant infections which might have affected the choice of antibiotic for PSSA BSI and treatment outcomes. Lastly, we did not adjust for multiple hypothesis testing; statistical significance in the secondary outcomes may not indicate clinical significance.

In conclusion, our study suggests that benzylpenicillin is a viable and potentially preferred option for the management of PSSA BSI. The findings support our institutional guidelines for clinicians to use benzylpenicillin for the management of PSSA BSI based on sensitivity reporting using the CLSI P10 method. We also identified older age, higher CCI score and higher PBS as independent risk factors for overall treatment failure in PSSA BSI. Importantly, unknown site of infection and lack of review by an ID specialist contributed independently to overall treatment failure, emphasising the need for prompt ID specialist consultation and thorough identification of sites of infection. With high mortality and morbidity, optimal treatment for PSSA BSI represents a crucial avenue for further research. We await the results from the SNAP trial to validate our findings.

Methods

Study cohort

This retrospective cohort study was conducted at TTSH. We identified all patients in our blood culture database with PSSA BSI between 1 January 2012 and 31 December 2021. To increase the power of our study, we expanded our cohort by incorporating a historic cohort from TTSH, which included patients with PSSA BSI between 1 January 2012 to 31 October 2015. This historic cohort was previously analysed and the results have been published21. The inclusion criteria were patients (i) aged 21 years and above; (ii) with first episode of PSSA BSI (defined as at least one positive blood culture for PSSA). The exclusion criteria were patients (i) who received predominant oral antibiotic therapy; (ii) who received predominant non-beta-lactam antibiotic treatment; (iii) with polymicrobial BSI (defined as blood culture positive for PSSA and at least one other microbe, including PRSA and excluding coagulase-negative Staphylococcus spp. if clinicians treated them as contaminants); (iv) who developed BSI with other microbe(s) (including PRSA and excluding coagulase-negative Staphylococcus spp. if clinicians treated them as contaminants) with or without PSSA after the first PSSA-only positive blood culture within the treatment period; (v) who died before receiving definitive antibiotic treatment for PSSA BSI; (vi) who received treatment for first episode of PSSA BSI in hospitals other than TTSH; (vii) with missing outcome data.

Data collection

The following data were collected retrospectively from hospital medical records: age; sex; date of admission and discharge; date of death; haemodialysis status; CCI score24; location of care at time of first positive blood culture (ICU or general ward); PBS25; date of first positive blood culture and susceptibility results; sites of infection; presence of abscess; whether managed by ID specialists (either as primary team or by referral); antibiotics used and duration; adverse drug events attributed to antibiotic treatment; date of microbiological clearance; date of recurrence of S. aureus BSI.

Outcomes and Definitions

The composite primary outcome was overall treatment failure, which consisted of at least one of the following: (i) 30-day all-cause mortality, defined as death due to any reason within 30 days from date of first blood culture positive for PSSA; (ii) 90-day relapse, defined as a blood culture positive for S. aureus obtained within 90 days after the completion of parenteral antibiotics for PSSA BSI treatment. Secondary outcomes were time to microbiological clearance of blood cultures, incidence of persistent BSI, and occurrence of adverse drug events due to antibiotic treatment. Time to microbiological clearance of blood cultures was defined as the number of days between first positive blood culture and first negative blood culture. Persistent BSI was defined as at least 5 or more days of PSSA BSI from date of initiation of active therapy21. Adverse drug events were defined as any adverse effects during the course of treatment due to any antibiotic within the same treatment group as the predominant definitive antibiotic, to account for the combined safety profile of all antibiotics potentially contributing to adverse drug events within each comparator group (benzylpenicillin versus non-benzylpenicillin beta-lactams).

An active antibiotic was defined as any antibiotic with in vitro activity against the blood isolate of PSSA, following Clinical and Laboratory Standards Institute (CLSI) M100 guidelines35. Empiric therapy was defined as antibiotic treatment administered for at least 2 days between date of first positive blood culture and date of final blood culture result (including susceptibility profile). Definitive therapy was defined as antibiotic treatment administered after the final blood culture result was available and initiated no later than 5 days from date of first positive blood culture. The predominant antibiotic was defined as the antibiotic that was used for the longest duration within the first 7 days of definitive treatment, as it likely had the greatest influence on treatment outcomes and analysis of antibiotics given subsequently may be confounded by initial treatment21. If two or more antibiotics were used for an equal duration, the first antibiotic used was considered the predominant definitive antibiotic. ICU admission was defined as within 24 h of first positive blood culture.

Statistical analysis

The Pearson’s chi-square test or Fisher’s exact test was used to compare categorical variables while the Mann–Whitney U test was used for continuous variables. Univariate and multivariate analyses for overall treatment failure were performed using binary logistic regression. Variables which yielded P value < 0.05 in the univariate analysis were included in the multivariate analysis, with predominant antibiotic therapy forced into the model. 2-tailed P values < 0.05 were considered statistically significant in the multivariate analysis. The fit of the model was checked using the Hosmer–Lemeshow goodness-of-fit and Nagelkerke R2 tests. All statistical analyses were performed with IBM SPSS for Windows, version 29.0 (IBM Corp., Armonk, NY, USA).