The goal of our study was to assess if empiric β-lactams with or without vancomycin compared to vancomycin alone was associated with differences in outcomes in patients with MSSA bacteremia. We found no differences in all-cause mortality at 28 and 90 days, or hospital LOS between these two groups. Clearance of bacteremia was delayed by a median of 1 day in the vancomycin group. However, this outcome may have been confounded by the earlier receipt of empiric antibiotics in the β-lactam group (median 2 h). When we analyzed a subset of patients from the vancomycin group (n = 43) whose median time to receipt of empiric antibiotics was 1.97 h, the duration of bacteremia was still longer compared to the β-lactam group, but just shy of statistical significance (95.1 vs 70.7 h, p = 0.06), likely because of the reduced sample size. Therefore, it does not appear that time to receipt of empiric therapy had a major impact on time to clearance of bacteremia in our study. Despite the high prevalence of MRSA at both of our institutions (25 and 38 %), only 67.2 % of patients in the β-lactam group received vancomycin empirically as well. Perhaps the awareness of MRSA was low among some treating clinicians or patients who did not receive empiric vancomycin were judged to be at low risk for MRSA infection.
Interestingly, the differential time delay in receipt of empiric antimicrobials was unexpected. Patients in the β-lactam group generally received antibiotics well before the blood culture became positive, while patients in the vancomycin group tended to receive antibiotics shortly after the blood culture turned positive. The reason for this observation is likely multifactorial. First, β-lactam patients generally had more identifiable sources of bacteremia (i.e. more skin and soft tissue infections). Second, the higher prevalence of community-onset bacteremia suggests these patients may have had their first medical contact with the emergency department where sepsis protocols facilitated timely administration of antibiotics.
In the β-lactam group, patients who received empiric β-lactam plus vancomycin had a higher Pitt bacteremia score, experienced more metastatic complications and stayed in hospital longer than those who received empiric β-lactam monotherapy. The greater severity of illness in this combination subgroup may explain the initial use of broad-spectrum β-lactams (ceftriaxone or piperacillin–tazobactam), with subsequent de-escalation to cloxacillin or cefazolin during the empiric period by the infectious diseases consultant when S. aureus was identified in the blood culture. De-escalation occurred within a median of 23.1 h, which follows the time to positivity of the first blood culture (median 20 h). Despite differences in baseline characteristics and antimicrobials prescribed, mortality rates and time to clearance of bacteremia were similar between these subgroups.
Our study outcomes were similar to those reported in the literature. The overall 28- and 90-day mortality in our study was low at 18 (7.06 %) and 36 (14.1 %) respectively, but is within the range of 3.6–51.7 % described in a meta-analysis of patients with MSSA bacteremia from catheter-related infections and infective endocarditis by Cosgrove et al. . Definite infective endocarditis was diagnosed in 36 (14.1 %) of our patients, which is similar to rates reported in previous studies [10, 14, 16, 23, 24].
The median duration of bacteremia was longer in the vancomycin compared to the β-lactam group (4 vs 3 days) in our study. In a similar study by Khatib et al. , clearance of bacteremia was delayed (duration ≥3 days) more often in patients who received empiric vancomycin (57.6 %) compared to those who received empiric β-lactams (37.5 %). They reported no difference in all-cause or attributable mortality between groups.
We did not find any differences in mortality between treatment groups in our study. In contrast, Lodise et al.  demonstrated that empiric β-lactam was associated with lower infection-related mortality than with empiric vancomycin monotherapy (11.4 vs 39.3 %, p = 0.005) among injection drug users with predominantly right-sided MSSA infective endocarditis. Even when patients were switched from vancomycin to a semi-synthetic penicillin within a median of 3 days, infection-related mortality remained high at 40.9 %. The overall mortality in Lodise’s cohort was unusually high at 22.2 % compared to a rate of 0–4 % described in a systematic review by Yung et al. . The largest study to date by McDanel et al. revealed that empiric β-lactam therapy (predominantly piperacillin–tazobactam and ceftriaxone) compared to vancomycin was not associated with differences in mortality in patients with MSSA bacteremia . However, this study excluded patients who received empiric vancomycin plus β-lactams, did not address microbiological cure, and evaluated empiric regimens independent of the definitive antimicrobial therapy prescribed.
We included a large proportion of patients who received empiric treatment with optimal anti-MSSA agents (cloxacillin or cefazolin), whereas previous observational studies have either failed to specify the β-lactam agents used or enrolled patients who received mostly broad-spectrum β-lactams. This is an important point because not all β-lactams have equal efficacy against MSSA. As demonstrated in one study, empiric cefazolin or cloxacillin was associated with improved short-term survival compared to empiric regimens containing other β-lactams . Therefore, the ideal study is one that compares a semi-synthetic anti-Staphylococcal penicillin or cefazolin to vancomycin.
Our study has several limitations. The reason for the lack of difference in the primary outcome is likely multifactorial. Because of the low event rate in both groups, our study was potentially underpowered to detect a significant difference in mortality. The lower than expected death rate may be partly due to the exclusion of patients who died within 24 h of the diagnosis of SAB and of patients who remained on broad-spectrum β-lactams. This group may have represented a sicker population and thus, we may have selected for less critically ill patients. The absence of matching with respect to baseline characteristics and the retrospective nature of the study may have also contributed to a lack of difference in the primary outcome. Obtaining subsequent blood cultures was often delayed, which may have led to an overestimation of the duration of bacteremia. However, this effect was likely balanced between both groups. Data regarding adverse effects were not collected due to the difficulty of establishing drug-related events in a retrospective study. A randomized controlled trial would be ideal to address our study question because it would provide better matching of patient baseline characteristics and control of antimicrobials prescribed, and permit prospective monitoring of adverse drug effects. In such a study, daily blood cultures would need to be collected to determine the exact date of clearance of bacteremia. As well, more accurate estimation of the expected mortality rates between groups would be needed when calculating the required sample size. Collaboration between the medical microbiology laboratory, infectious diseases service and antimicrobial stewardship team is essential to execute such a trial.
Until we have more concrete evidence from future prospective studies, the benefit of adding a β-lactam to empiric therapy for MSSA bacteremia remains unclear. Ultimately, the choice of empiric regimen will depend on patient factors, the prevalence of MRSA in the population, and the ability of the microbiology laboratory to rapidly differentiate MSSA from MRSA.
In conclusion, empiric therapy with β-lactams was associated with earlier clearance of bacteremia by a median of 1 day compared to vancomycin, but was not associated with differences in all-cause mortality or hospital LOS in patients with MSSA bacteremia. Our data should be interpreted with caution however, as major differences in the baseline characteristics between the groups may have overshadowed any potential treatment effect. Future prospective studies are needed to confirm our findings. For now, empiric treatment with vancomycin is reasonable if the prevalence of MRSA is significant. The addition of a β-lactam agent could be considered in critically ill patients.