Can the Xpert MRSA/SA BC assay be used as an antimicrobial stewardship tool? A prospective assay validation and descriptive impact assessment study in a South African setting

We aimed to assess the performance and role of the Xpert in facilitating more appropriate antibiotic use in patients with S. aureus bloodstream infection, and reducing inappropriate antibiotic therapy in those with S. aureus and with coagulase-negative staphylococci. We performed a prospective observational study to evaluate the Xpert against the reference method of culture-based techniques, and to describe the potential impact of implementation of this assay in our setting by combining clinical history with attending clinician survey.

The study took place from January to June 2016 at the National Health Laboratory Service (NHLS) Microbiology laboratory at Tygerberg Hospital in the Western Cape, South Africa. Tygerberg Hospital is a 1384-bedded referral hospital serving the northern and eastern subdistricts of the Cape Metro region, and the surrounding rural districts, including 4 regional hospitals and 17 district hospitals.

Blood culture bottles were submitted from facilities throughout this drainage area at the clinicians’ discretion. The automated blood culture system in use is the BacT/Alert 3D Microbial Detection System (bioMérieux, Marcy L’Étoile, France) which includes anaerobic (FN Plus), aerobic (FA Plus) and paediatric bottles (PF Plus). These were kept at room temperature during transport, and were incubated as soon as possible after specimen receipt. After flagging positive, Gram stains are performed on an aliquot of the broth. All patients with positive blood cultures exclusively showing GPCC on Gram stain, were included. Blood culture specimens showing mixed morphology on Gram stain, and duplicate blood culture samples from the same patient, were excluded.

Paediatric patients were defined as patients below the age of 13 years.

Patients with positive blood cultures (indicating suspected sepsis) were defined as having community-acquired (CA) infection if the culture was collected < 72 h after admission, and hospital acquired (HA) infection thereafter; admission within the preceding 6 weeks was also considered a criterion for HA infection [4] .

Regional hospital guidelines in our drainage area advise a targeted approach of a semi-synthetic beta-lactamase stable penicillin (cloxacillin) for MSSA BSI, with vancomycin advised for MRSA BSI.

As per standard practice, clinicians were contacted with the Gram stain result if the available clinical information or previous results suggested S. aureus sepsis, or if the patient was admitted in an Intensive Care Unit (ICU). In all other cases, clinicians were contacted with the culture result, whether S. aureus or CoNS.

Basic clinical and demographic information were obtained, including an antibiotic history. We surveyed the attending clinicians to assess whether knowledge of the culture and susceptibility result on the day the bottle flagged positive would have impacted antibiotic choice. This was categorised as:

We regarded cephalosporins (with the exception of ceftazidime), beta-lactam–beta-lactamase inhibitors, carbapenems and vancomycin as being active against MSSA. Penicillin, ampicillin, amoxicillin, and ceftazidime were the beta-lactam agents regarded as being ineffective against MSSA.

CoNS were considered significant only if reported as the likely cause for the patient’s clinical condition, as judged by the attending clinician.

Days of antibiotic therapy saved were also evaluated during clinician interview, based on the empiric antibiotic regimen commenced and on clinician opinion as to whether an earlier result would have impacted antibiotic choice or duration (i.e. if the Xpert had been utilised as soon as the bottle flagged positive, leading to a “final” result being available shortly after Gram stain availability).

Microsoft Excel, VassarStats (www.​vassarstats.​net) and EpiCalc 2000 v1.02 (Brixton Books, South London, UK) were used, with an α-level of 0.05 chosen. Continuous numerical variables were summarised using means and standard deviations for normally distributed variables, or medians and interquartile ranges (IQR) for variables not following a normal distribution. Binary categorical variables were summarised using proportions and 95% confidence intervals (CI). Comparison of selected independent proportions were evaluated using the z-test.

A valid Xpert result was initially obtained from 225/231 bottles (97.4, 95% CI 94.5–98.8%). Of the remaining 6 assays, 2 (2.6, 95% CI 1.2–5.6%) were invalid (failure of internal control) and were not repeated. Four were resulted as “error” with signal loss after initial amplification; two provided a valid result after repeat testing. The conservative failure rate was 1.7% (95% CI 0.7–4.4%), with 227 bottles yielding an interpretable result.

There was 100% concordance between the systems for the identification of MSSA, MRSA and CoNS. The Xpert assay had an overall sensitivity of 100% (57/57; 95% CI 93.7–100%), and specificity of 100% (170/170, 95% CI 97.8–100%). The positive and negative predictive values of the assay were both 100% (95% CIs 93.7–100% and 97.8–100% respectively). Analysis by MSSA, MRSA and CoNS is summarised in Supplementary Table 1.

We performed a subanalysis of mecA and methicillin resistance in CoNS, for the 169 isolates in which cefoxitin susceptibility on culture was known. The gene target, mecA, was detected in 85 of 88 CoNS isolates that were cefoxitin-resistant on culture (96.6%; 95% CI 90.5–98.8%) giving a positive predictive value of 86.7% (95% CI 78.6–92.1%). The specificity of detection of mecA in CoNS on Xpert was 84.0% (95% CI 74.5–90.4%) with a negative predictive value of 95.8% (95% CI 88.3–98.6%).

The median time between blood culture positivity and final result authorisation was 31.3 h (interquartile range (IQR) 20.7–42.5 h).

Of the 227 patients included, 50.7% (115/227, 95% CI 44.0–57.3%) were males. The median age of the adults included was 43.0 years (IQR 31.0–58.0); the median age of the paediatric population was 71.5 days (IQR 17.8–325.0). Twenty-five adults (25/151, 16.6, 95% CI 11.2–23.7%) and 10 children (10/76, 13.2, 95% CI 6.8–23.3%) were admitted to an ICU at the time of blood culture collection. Sepsis was CA in 51.1% (116/227, 95% CI 44.4–57.8%), HA in 38.3% (87/227, 95% CI 32.0–45.0%); and unknown in the remaining 10.6% (24/227, 95% CI 7.0–15.5%). Only 1/11 with MRSA BSI reportedly had CA onset of infection (9.1, 95% CI 0.5–42.9%).

Adequate history regarding source of sepsis was obtained for 195 patients. Sixty-two patients (31.8, 95% CI 25.4–38.9%) had an unclear source of sepsis. Respiratory tract infection was the most commonly identified suspected source overall (56/195, 28.7, 95% CI 22.6–35.7%) and in CA-infection (33/101, 32.7, 95% CI 23.9–42.8%). For HA-infection, 29 patients (36.3, 95% CI 26.0–47.8%) had no clear source of sepsis; of those with known primary sources, the major foci were the respiratory tract and skin or skin structures (both 18/80 patients, 22.5, 95% CI 14.2–33.5%).

Information on the suspected source of sepsis was obtained in 46 patients with S. aureus BSI. The most commonly identified suspected source of sepsis (in isolation or as part of two potential sources) was skin or skin structure infection (n = 16, 34.8%), followed by suspected respiratory tract infection (n = 11, 23.9%). Source was undetermined at the time of clinician interview in 8 patients with S. aureus BSI (17.4%).

Characteristics of the included patients can be found in Supplementary Table 2.

Methicillin resistance was detected with 100% accuracy in S. aureus; the small MRSA sample size precludes firm conclusions. Methicillin resistance in S. aureus is chiefly mediated by mecA at present, although mecC-mediated resistance has been reported in 0.004% of human isolates [6]. The assay also performed well for MSSA BSI, showing 100% concordance with culture-based testing. The error rate of 1.7% is similar to previous reports [3, 7] and is acceptable. Although this assay has not been rigorously tested for the detection of methicillin resistance in CoNS to date, the negative predictive value of 95.8% in this context may be useful in selected cases of potentially significant CoNS bloodstream infection.

We recommend conventional culture to be performed at least limitedly in parallel, for mixed cultures, as a backup in case of an unsuccessful Xpert result and for further antimicrobial susceptibility testing or surveillance. Genotype-phenotype mismatch, with the Xpert assay incorrectly reporting a susceptible methicillin result due to insertions in the SCCmec-orfX junction region, has been reported in 3 S. aureus isolates from the United States [8], and misclassification of S. aureus as CoNS in 2 isolates from Australia [9]. Studies investigating the prevalence of mutations that may affect the targets of this assay in a South African setting are needed to define the role of this assay more accurately.

To our knowledge, this is the largest study assessing the potential impact of this assay to date. The median time saving to final result with use of this assay was approximately 30 h if the Xpert assay was performed immediately after microscopy, in line with previous estimates of a time saving of 24–48 h [2, 10]. This crude estimate can be influenced by factors such as workflow. Antibiotic therapy was optimised on availability of final result in 68.9% with S. aureus BSI. This impact was most clearly demonstrated in patients with MRSA BSI, where more than 50% received anti-MRSA therapy only in response to the final culture result. A more modest impact could be seen for patients with MSSA BSI (approximately 21%); however, it must be noted that this is a conservative estimate as all beta-lactam agents were considered to have activity against MSSA in this analysis, although they may not be equally suitable for the treatment of MSSA BSI and some are associated with a higher odds of death [11].

More rapid administration of appropriate therapy can significantly impact mortality. In one study, the mean mortality rate was 59% for patients with MRSA BSI who were initially started on a semisynthetic penicillin, as opposed to 23–36% for patients initially started on vancomycin [12]. The mean mortality rate was as low as 12% in patients empirically placed on a semisynthetic penicillin, who cultured MSSA [12]. Future studies should specifically assess the clinical impact of Xpert on more appropriate S. aureus therapy.

We initially expected this assay to be of most use in reducing antibiotic therapy in patients with CoNS, as has been reported previously [13], but this finding was not replicated. The vast majority of CoNS bloodstream isolates were regarded as being not clinically significant in this cohort, in keeping with the high rate of blood culture contamination in our setting [14]. Clinicians thus disregarded this “contaminated” result and continued the empiric antibiotic regimen that had been commenced based on the initial clinical assessment. Decisions regarding antibiotics are also governed by other factors not measured in this study, such as the clinical response to antibiotic therapy and additional test results. Although availability of relevant rapid diagnostic tests such as the Xpert is a core element to antimicrobial stewardship programmes in healthcare facilities in low-middle income countries [15], good quality microbiological sampling is also crucial and results in more appropriate use of laboratory resources. Ongoing educational interventions to optimise blood culture sampling (and reduce blood culture contamination) is a relatively cost-effective and easily implementable measure that can be instituted before turning to more costly solutions, and it is important that this remains a focus of antimicrobial stewardship programmes in resource-limited settings.

The potential benefits described must be weighed against the expense of the assay, including additional labour, infrastructure and consumable costs. The total cost of the test may be prohibitive in a resource-constrained setting such as our own.

Limitations to this study include that the S. aureus BSI rate was only 25.1% of all blood cultures with Gram positive cocci in clusters, and that methicillin resistance was detected in 26.3% of all S. aureus-containing blood cultures (similar to a contemporaneous study in our setting (27.1%) [16]); this resulted in modest absolute numbers of MSSA and MRSA BSI. This may confer a more moderate impact of test performance than would be seen in higher prevalence settings. CA-MRSA is also not common in our setting currently [17]. Secondly, complete clinical records were not available for all patients, prohibiting case-by-case assessment of whether antibiotic changes occurred as a result of the microbiological results or the evolving clinical picture. Furthermore, clinician recall bias may have influenced these results. Thirdly, other patient comorbidities, such as renal failure and severity of illness, may have contributed to the choice of agents used and the decision to modify antibiotic therapy; this could not be easily assessed. Fourthly, we sampled a proportion of those eligible for inclusion; although we compared these populations for all the key variables that would be available in a routine diagnostic laboratory setting (thereby limiting selection bias), other variables, such as source of sepsis, were not taken into account. Additionally, significant CoNS BSI may have been underestimated in this study, as clinicians in our setting infrequently adhere to recommendations for multiple blood cultures to be submitted in the investigation of sepsis. Finally, most beta-lactam agents were considered to have anti-MSSA activity for the purposes of this study, which may result in an underestimate of the clinical benefit of the assay for MSSA BSI.