Introduction

Bacteraemia due to carbapenem resistant Klebsiella pneumoniae (CR-KP) is associated with high morbidity and mortality [1]. Most studies on CR-KP bloodstream infection (BSI) have focused on the risk factors for all-cause mortality [2, 3].

In the clinical practice, recurrence after surviving the primary episode of CR-KP BSI is another important outcome that affects length of stay, need for readmission, and long-term mortality [4]. The knowledge of the risk factors for the recurrence of CR-KP BSI may be useful for treatment decisions.

The aim of our study was to determine how often patients cured of a first episode of CR-KP BSI developed a new episode, and to analyse the risk factors for recurrence.

Material and methods

Study design

We performed a prospective observational cohort study of all consecutive adult (≥18 years) patients cured of a first CR-KP BSI at our hospital over six-year period (June 2010–June 2016). The maximum follow-up period per patient was 180 days after the index blood cultures (BCs).

Setting

The setting was a 1,420-bed tertiary teaching hospital with approximately 72,000 yearly admissions. Performance of BCs was at the discretion of the attending physician and was not dictated by study protocol. Standard approach to the therapeutic management of CR-KP BSI consisted of source control when feasible and antibiotic therapy including high-dose (HD) meropenem (6 g/day) administered by extended infusion [5] plus at least one in vitro active drug depending on the infection source, as previously reported [6].

The institutional review board of the hospital approved the study.

Definitions

CR-KP BSI was defined as one or more BCs yielding CR-KP obtained from a patient with symptoms and/or signs of systemic inflammatory response syndrome [7]. Patients were considered only once at the time of first episode (index BCs). Subsequent positive BCs were defined as: (i) persistent CR-KP BSI, BCs persistently positive despite initiation of anti-CR-KP antibiotic treatment; (ii) breakthrough CR-KP BSI, new evidence of positive BCs after their clearance during an ongoing antibiotic course; or (iii) recurrent CR-KP BSI, new evidence of positive BCs in patients with documented clinical response after completing a course of anti-CR-KP therapy.

Polymicrobial BSI was defined as the growth of more than one microorganism, excluding potential contaminants (i.e. coagulase-negative staphylococci, Corynebacterium spp., Propionibacterium spp.) isolated in only one set of BCs.

BSI sources were established according to the Centres for Disease Control and Prevention criteria [8]. In the absence of a recognized source, BSI was considered as primary.

Clinical cure was defined as the resolution of all signs and symptoms of infection and was prospectively assessed according to vital signs, laboratory and imaging data at 14 days after index BCs by a trained clinical investigator.

Active empiric therapy was defined as at least one antibiotic with in vitro activity, based on susceptibility results, administered within 24 h from drawing the index BCs. Active directed therapy was defined as at least 1 antibiotic with in vitro activity administered after susceptibility results were available for at least 7 days. The length of antibiotic therapy was defined as the time period form the first to the last day of an active therapy.

Source control was defined as the removal of infection source, including the performance of nonsurgical or surgical procedures to treat an obstructive focus or abscess, and the removal of any device deemed as the source of BSI. In addition, removal of intravascular devices, foreign bodies and urinary catheters was recorded irrespectively of the BSI source.

Data

Variables were collected using a standardized case report form including demographic data (age and sex), admission and discharge dates, underlying diseases according to Charlson’s index score, solid organ transplantation, ward of admission at the time of index BCs, CR-KP rectal carriage at the time of index BCs, clinical severity at BSI onset according to the sepsis grading [9] and the APACHE II score, polymicrobial BSI, persistent and/or breakthrough BSI, the presumed source of BSI, source control, removal of central venous catheter (CVC), administered drugs with the relative start and end dates, all-cause mortality and date of death, recurrence and date of recurrence, as well as number of recurrences.

Microbiology

BCs were incubated using the BACTEC FX Automated Blood Culture System (Becton Dickinson, Franklin Lakes, NJ). All positive BCs were processed with MALDI-TOF for the rapid identification of the microorganism. Identification and susceptibility testing of strains were performed using the Vitek 2 automated system (bioMerieux, Marcy l’Etoile, France). Enterobacteriaceae MICs were interpreted using EUCAST clinical breakpoints for all tested antimicrobials. Carbapenemase production was confirmed by disc-diffusion synergy test (Rosco Diagnostica, Taastrup, Denmark).

Statistical analysis

Comparison between recurrent and non-recurrent cases was performed. Categorical variables were compared using Pearson chi-square or Fisher’s exact test when appropriate. Continuous variables were compared using Mann–Whitney U test.

To investigate risk factors for recurrence, univariate and multivariate cause-specific Cox proportional hazards analysis were performed. Patients were observed from the index BCs up to recurrence, death or 180 days. The variables with a p value ≤0.10 at univariate analysis were entered into the multivariate model. As the correlation test for source control and CVC removal was significant (0.673, p < 0.001), only source control was maintained into the final model. Statistical significance was set for p value <0.05. The analysis was performed with SPSS 21.00 software package.

Results

During the study period 249 patients were diagnosed with CR-KP BSI at our hospital. According to study criteria, 193 patients were deemed as cured within 14 days after index BCs and were analysed.

All patients received a combination therapy; the most frequent administered individual antibiotics and combination regimens are shown in Table 1.

Table 1 General characteristics of patients cured of a first CR-KP BSI episode, and comparison between recurrent and non-recurrent cases
Full size table

Recurrence occurred in 32 out of 193 patients (16.6%) within a median of 35 (IQR 25–45) days after index BCs. All but one of the recurrences occurred within 60 days after index BCs. Five patients developed more than one recurrent CR-KP BSI: three patients developed two episodes, one patient developed three episodes, and one patient developed four episodes. Susceptibility of recurrent strains to the antibiotics administered during the index CR-KP BSI was analysed. Increase in the MICs of colistin, tigecycline and gentamicin was observed in 37%, 36.8% and 50% of exposed cases, respectively. Meropenem MIC of the initial strain was ≥16 mcg/mL in all but three of the 31 recurrent cases treated with a meropenem-based combination regimen. In two of them the subsequent isolate showed an increase in MIC.

Comparison of recurrent and non-recurrent cases is shown in Table 1. Significant differences were found for colistin use (84.4% vs. 62.2%, p = 0.01), meropenem-colistin-tigecycline regimen (43.8% vs. 24.8%, p = 0.03) and length of therapy (median 18 vs. 14 days, p = 0.004).

All-cause 180-day mortality was significantly higher in recurrent cases (34.4% vs. 16.1%, p = 0.02).

The analysis of risk factors for recurrence is shown in Table 2. At the multivariate analysis the only variable that remained significantly associated with recurrent CR-KP BSI as a protective factor was source control.

Table 2 Univariate and multivariate analysis of risk factors for recurrence in patients cured of a first CR-KP BSI episode
Full size table

Discussion

We observed that recurrent CR-KP BSI among patients cured of the first episode is frequent and associated with higher long-term mortality. Source control was the only protective factor associated with recurrent CR-KP BSI at multivariate analysis.

Messina and colleagues recently investigated how often patients colonized or infected with CR-KP were readmitted with repeat positive cultures for CR-KP [4]. CR-KP readmission was observed in 20% of cases, a figure similar to that of our study. In addition, the authors examined the impact of treatment on the risk for CR-KP readmission [4]. Independent risk factors for 90-day CR-KP readmission were: malignant tumour, tigecycline regimen, and receipt >1 in vitro active antibiotic [4]. We observed in the univariate analysis that colistin use, meropenem-colistin-tigecycline regimen, and longer duration of antibiotic therapy during the index BSI were risk factors for recurrent CR-KP BSI. However, at the multivariate model the only variable associated with recurrence as a protective factor was source control. In our opinion, the impact of antibiotic treatment on CR-KP recurrence should be further investigated in larger multicentre studies.

It is unknown whether the introduction of the new active drugs against CR-KP, such as ceftazidime-avibactam, will reduce the rate of recurrence. Unfortunately, in a recent series on the use of ceftazidime-avibactam for treating 37 patients with CR-KP infection, the recurrence rate was 23% [10].

Our study has limitations. First, the single centre design with limited sample size hampers the generalizability of our results and the statistical power of our analysis. Second, genotyping of the strains was not performed thus we were not able to differentiate recurrences between relapses and re-infections [11]. The fact that all but one of the recurrences occurred within 60 days after the index episode may suggest that most episodes were relapses [12]. Third, follow-up blood cultures were drawn at the discretion of attending physicians, thus in several cases the microbiological clearance was presumed on the basis of the good clinical, laboratory and imaging response as normally performed with Gramme negative BSI. Finally, the impact of resistance to the administered antibiotics was analysed using the susceptibility results obtained with the Vitek 2 automated system. Although there are more accurate methods, this is one of the most common systems used.

To conclude, recurrence is frequent among patients cured of CR-KP BSI and it is associated with higher long-term mortality. When feasible, source control is mandatory to avoid recurrence. The impact of antibiotic therapy should be further investigated in large multicentre studies.