Study design and case definitions
This was a retrospective cohort study. Using our infection control team’s database, we identified 872 episodes of enterococcal bacteremia (positive blood cultures for Enterococcus species) at MD Anderson Cancer Center between September 2011 and December 2018. We included adult cancer patients with a first episode of enterococcal bacteremia diagnosed in the presence of a central venous catheter (CVC) that had been in place for at least 48 hours prior to the onset of bacteremia. We excluded 330 patients with polymicrobial bacteremia, with no indwelling CVC at the onset of index bacteremia, or with a CVC placed less than 48 hours before onset. We classified the patients with bacteremia into CLABSI or non-CLABSI groups according to the Centers for Disease Control’s definition of CLABSI (5). We further divided the CLABSI group into a CLABSI with mucosal barrier injury (MBI) subgroup and a CLABSI non-MBI subgroup. MBI was defined as the presence of either of the following criteria: 1) neutropenia with an absolute neutrophil count of < 500 cells/mm3 on 2 separate days within 3 days of bacteremia diagnosis; 2) in a patient who received a hematopoietic stem cell transplant (HSCT) within 1 year of the positive blood culture, the presence of either grade III or IV gastrointestinal graft-versus-host disease or severe diarrhea of ≥ 1 L in a 24-hour period within the 7 days prior to the positive blood culture (5).
In addition, we identified the cases that met the criteria for catheter-related bloodstream infections (CRBSI) according to the Infectious Diseases Society of America (IDSA) Clinical Practice Guidelines for the Diagnosis and Management of Intravascular Catheter-Related Infection (4). We analyzed 3 groups of patients: group 1 (G1) included patients whose bacteremia met the definition of CLABSI with MBI (considered as possible CLABSI); G2 included patients who had either CRBSI or CLABSI without MBI (considered as definite CRBSI); and G3 included patients with non-CLABSI who had a CVC in place but likely had bacteremia from another source.
Data extraction and study outcomes: The protocol was approved by our Institutional Review Board and a waiver of informed consent were obtained prior to the conduct of the study.
Patient data were extracted from our institution’s electronic medical records system (Epic). We collected data on patient demographics, underlying malignancy, neutropenic status, and risk factors for infection. Microbiological data collected included date and source of positive blood cultures, bacterial species, phenotypic susceptibility pattern, and status of colonization with vancomycin-resistant enterococci (VRE). We recorded all the antibacterial agents that were used to treat the bacteremia starting from the date of positive blood culture and for the subsequent 2-week period. CVC management (removal or retention) was evaluated at 2 time intervals: early (within 72 hours bacteremia onset and late (at 3 to 7 days). CVCs that were removed after 7 days were considered to have been retained. The three-day cut off was chosen to mimic the clinical scenario where some time is elapsed between blood collection and organism identification. Patients were followed for 3 months after the onset of the index bacteremia, until lost to follow-up, or until death, whichever occurred first.
Clinical and microbiological outcomes were determined as follows. Clinical resolution was defined as defervescence within 72 hours. Microbiologic eradication was defined as resolution of the bacteremia within 96 hours. Recurrence of the bacteremia during the follow-up period was identified by positive blood cultures with isolates that shared a similar phenotypic antimicrobial susceptibility pattern to that of the baseline isolate. Infection-related complications included the occurrence of deep-seated infection, such as infective endocarditis, thrombophlebitis, or osteomyelitis, during follow-up. We collected all-cause mortality and infection-related mortality data. Death was attributed to enterococcal bacteremia based on the clinical impression of the treating physicians and available clinical data. A successful overall outcome was a composite of: absence of infection-related complications, absence of infection-related mortality, and absence of microbiological recurrence within 90 days. Patients who died within 7 days of onset of index bacteremia and patients who received antibiotic catheter lock therapy were excluded from the outcome analyses.
Statistical analysis (pending)
We used the χ2 test or Fisher exact test to compare categorical variables, as appropriate. To compare continuous variables, we used the Kruskal-Wallis test for 3-group comparisons and Wilcoxon rank sum test for 2-group comparisons. If a significant result (P < .05) was detected for a test that compared 3 groups, then pairwise comparisons were performed, with α levels adjusted using Holm’s sequential Bonferroni adjustment to control the type I error. A multivariate logistic regression model was used to identify factors that were independently associated with mortality. All tests were 2-sided, with a significance level of 0.05, except pairwise comparisons with the α adjustment. The statistical analyses were performed using SAS software version 9.3 (SAS Institute, Cary, NC).