There’s still one reason to request extended incubation of blood cultures in infective endocarditis: Cutibacterium acnes. I was taught that asking for prolonged incubation of blood cultures was an unnecessary holdover from the old days when culture media lacked proper growth factors and was generally lousy at growing viable microorganisms in blood. It seems there’s an exception to that teaching, though. The authors reviewed the results of blood cultures in patients with infective endocarditis at the Mayo clinic over a two-year period, specifically focusing on the results of extended culture incubation when requested by the ID team. Extended incubation was requested for a total 116 blood cultures from 53 patients, and yielded additional results in five cultures from three patients, all with growth of Cutibacterium acnes from the anaerobic bottle. One patient had a single culture deemed to be a contaminant; however, in the other two patients, multiple cultures were positive, the patients had prosthetic valves, and histopathology was consistent with C. acnes prosthetic valve endocarditis.
That’s a positivity rate of 4% (2/53 patients), and I imagine would be substantially higher if the additional incubation requests had only been made in patients with prosthetic valves. So, I think when I see a culture-negative prosthetic valve endocarditis (or endovascular graft infection?), I’ll be routinely requesting extended incubation of the blood cultures. 30293204
How much does a pan-CT add to the management of infective endocarditis? Very little. I don’t think of the chest/abdomen/pelvis CT as being particularly valuable for the workup of endocarditis, and this study supports that sentiment. The authors performed a retrospective review of all patients with possible or definite endocarditis by Duke’s criteria seen at one of several French hospitals over a three-year period who received a CT of the chest, abdomen, and pelvis. The researchers identified whether the patients had any specific documented symptoms indicating the CT scan, whether anything was found on the CT, and whether the CT findings influenced the overall management of the case.
Five hundred and twenty-two patients were included in the final analysis; the mean age was 69, 72% of patients had a vegetation on echocardiography, the most common organisms were staphylococci (36%) and streptococci (39%), and 61% of patients had an extracardiac focus or complication of infection. Pan-CT revealed an endocarditis-related lesion in 42% of patients, of which the most common types were splenic (in 25%), renal (in 12%), and vertebral (in 11%). Patients for whom lesions were versus were not found received similar durations of antibiotic therapy (difference in length of therapy -0.6 days; 95% CI -8 to 7, p=0.87). Of the 217 patients who had a lesion identified on CT, 19% received a specific treatment (e.g. drainage of a deep abscess, addition of a second antibiotic “for better bone diffusion,” or treatment of a mycotic aneurysm). However, only nine of these patients were asymptomatic. Put another way, you have to pan-scan about 58 patients with endocarditis to discover one patient with an asymptomatic CT lesion needing some specific additional treatment. The authors end by pointing out that 17% of all patients receiving the scan developed AKI, and that this was associated with increased mortality (OR 4.47; p<0.001). This line of argument doesn’t do much for me; correlation is not causation, and the other thing that gives you AKI is sepsis from inadequately controlled infection. Still, I’m fully onboard with not getting pan-CTs in infective endocarditis on the grounds that these data show how little value the scans offer. 30615098
How well do culture and patient data predict antibiotic susceptibilities in gram-negative bacteremia? The authors conducted a multicenter retrospective cohort study across two large health networks on the US-Canada border, evaluating patients hospitalized between 2010 and 2015 with monomicrobial gram-negative bacteremia. Researchers collected demographic and clinical data as well as each patient’s culture data from the preceding 12 months. The primary outcome of interest was the antibiotic susceptibility of the index blood culture isolate, and the authors assessed the predictive value of multiple clinical decision-making models based on the data collected. Specifically, the models used were as follows: A) gram stain-guided antibiogram-based predictions, B) Model A adjusted for the patient’s prior culture results, prior ICU and/or hospital admission, recent receipt of antibiotics, and location in the hospital, C) Model A, but using the specific pathogen isolated rather than the gram stain to inform predictions, and D) Model B, but using the specific pathogen rather than the gram stain to inform predictions.
A total 1,832 patients with gram-negative bacteremia were included in the cohort. About half of cases were due to E. coli and about a fifth were due to K. pneumoniae; most patients were over 65 years of age, more than half had received antibiotics in the preceding 90 days, and more than a third had been hospitalized on the floor or in an ICU in the preceding 90 days. The authors found that models incorporating patient data versus simply recommending antibiotics based on the antibiogram resulted in substantial (25% to >50%) reductions in inappropriate therapy, and that recommendations based on the specific pathogen isolated were more accurate than recommendations based on the gram stain alone. As one might expect, decreasing the model’s tolerance for error (i.e. increasing the treatment threshold from 80% to 90% or 95%) significantly reduced the proportion of patients for whom inadequate empiric therapy would be recommended. However, this approach comes at the cost of models that recommend meropenem for everyone (21-100% of patients) versus 0-8% of patients in the lower treatment threshold models. The results of these models are all presented in Figure 2 and Table 3; your mileage may vary, but I thought the most practical model was the pathogen-guided model including patient variables, which resulted in meropenem prescriptions for 2-21% of patients, ceftriaxone for 50-60%, and undertreatment rates of only 6-7%.
But again, the real question that should be in all our minds is whether these clinical decision models offer any value in the era of rapid PCR-based identification and susceptibility testing of blood isolates, wherein you can tell whether your patient’s K. pneumoniae harbors the CTX-M or KPC genes within half an hour of the blood culture bottle turning positive. 29705558
Nasal MRSA colonization has fair predictive power for whether a patient’s S. aureus infection (at any site) will be methicillin-resistant. The high negative predictive value of a nasal MRSA screen for MRSA pneumonia is clearly established, and this makes sense given the anatomic proximity of the upper and lower respiratory tracts. But what about infections at other sites? The authors reviewed the records of adults hospitalized at the Cleveland Clinic over a seven-year period, looking at the initial hospitalizations of all patients screened for MRSA by nares PCR or culture, extracting those patients who went on to have an S. aureus infection (defined as any non-nares culture positive for S. aureus), and then stratifying them by the nares MRSA colonization status.
The authors found 90,891 initial hospitalizations among patients screened for MRSA. Of these, 1999 went on to involve an S. aureus infection; 30% were MRSA carriers, 26% MSSA carriers, and 44% non-carriers. The baseline characteristics were similar across groups, except that the non-carriers were more frequently identified at the Cleveland Clinic’s main versus peripheral sites, and were more heavily recruited in the later years of the study.
A total 1024 (51%) of the S. aureus isolates were methicillin resistant. Compared with being an MRSA carrier, being either a non-carrier (OR 0.07) or an MSSA carrier (OR 0.008) was protective against infection with MRSA (p<0.0001 for both). These associations held true for the subset of patients with S. aureus bacteremias (OR 0.07 for non-carriers and 0.0006 for MSSA carriers, respectively; p<0.0001 for both) and in multivariate analysis. Figure 2 has a nice breakdown of the prevalence of MRSA infections among MSSA carriers stratified in various ways (e.g. by age, gender, hospital, type of infection etcetera); notably, the prevalence of MRSA among MSSA carriers with S. aureus infection was consistently in the 5-10% range.
Bottom line? Be it a pneumonia, bacteremia, or some other type of infection, your patient colonized with MSSA is fairly unlikely to have an MRSA infection. Interestingly, being a non-carrier was far less protective versus being an MSSA carrier. I suspect that non-carriers represent unfilled ecologic niches into which MRSA can immigrate once the patient gets to the hospital. This could be a good reason to use a nasal PCR screen that can detect the presence of S. aureus as well as recognize the MecA gene for methicillin resistance. 29649598