Carbapenem-resistant Klebsiella pneumonia (CRKp) blood stream infections (BSI) cause substantial mortality among hospitalized patients. Treatment options for CRKp infections are limited and increasing resistance rates to few available drugs, i.e., colistin, is a big concern. This prospective multicenter observational study is designed to describe clinical characteristics and outcomes of patients with CRKp bacteremia in an oxacillinase-48 (OXA-48) endemic country to define predictors of mortality with a focus on the impact of mono versus combination therapies on mortality. The study will also investigate risk factors associated with colistin-resistant CRKp BSI.

OBJECTIVES

MAIN OBJECTIVE

1. To compare the impact of combination therapy and monotherapy on mortality for patients with CRKp BSI

   SECONDARY OBJECTIVES

2. To identify predictors of mortality for patients with CRKp BSI

3. To compare the impact of appropriate and inappropriate empiric and targeted therapies on mortality for patients with CRKp BSI

4. To identify the risk factors for colistin resistance among patients with ColR CRKp BSI.

HYPOTHESES

MAIN HYPOTHESIS

1. Combination therapy is better than monotherapy to prevent 30-day mortality.

   i.Combination of one active drug with one inactive drug is not associated with lower probability of mortality as compared to monotherapy with one active drug if:

   • the inactive drug is carbapenem with a minimum inhibitory concentration (MIC) of ≥16 mg/L or
   • the inactive drug is any other antibiotic other than carbapenem or
   • the patient is in low mortality score

   ii.Combination of two or more active drugs is associated with lower probability of mortality as compared to monotherapy with one active drug.

   iii.For colistin resistant CRKp, monotherapy with one active drug is associated with lower probability of mortality as compared to combination therapies including carbapenem with a MIC of ≥16 mg/L or colistin.

   SECONDARY HYPOTHESES

2. Severe disease at BSI onset (admission to ICU, septic shock, INCREMENT-CPE (carbapenemase-producing Enterobacteriaceae) score 8-15, higher sequential organ failure assessment (SOFA) and Pitt bacteremia scores, Charlson Comorbidity Index ≥2, higher procalcitonin levels), carbapenemase type and carbapenem MIC of the isolate (CRKp isolates with a carbapenem MIC ≥16 mg/L, OXA-48 producing isolates), sequence type of the isolate (ST101 and ST258), colistin resistance delayed source control (>48 hours) and infection sources other than urinary tract are independent predictors of mortality.

3. Appropriate empirical therapy, early targeted therapy, are independent predictors of survival. For colistin susceptible CRKp, empirical use of colistin within 24 hours of BSI onset is associated with lower probability of mortality.

4. Previous colistin therapy, previous CRKp infection, previous intensive care unit (ICU) stay (last three months), length of hospital stay until index date is associated with colistin resistance. Mortality rates and length of hospital stay are higher among colistin resistant strains as compared to colistin susceptible strains.

5. Effect of treatment type on mortality is modified by disease severity.

STUDY DESIGN

Prospective cohort study.

Setting and study period

Fourteen tertiary care hospitals from the four most populous cities in Turkey (Istanbul, Ankara, Izmir, Bursa) are included in the study. Study period is scheduled to be 25 June 2018- 25 June 2019 or longer until pre-defined sample size is reached. Local investigators (infectious disease physicians and clinical microbiologists) will collect microbiological and clinical data. Site investigators will screen the microbiology laboratory logs daily to identify patients with CRKp bacteremia. All consecutive patients with CRKp bacteremia will be eligible to be evaluated for inclusion in the study. Patients who meet the inclusion criteria will be selected for the prospective cohort study and/or case-control study depending on colistin resistance pattern of the isolate. Site investigators will access patients' medical records and register relevant information using a web-based data entry system. Site investigators will also collect CRKp isolates and keep them in appropriate conditions until time of transfer for analysis. All CRKp isolates will be sent to the central microbiology laboratory for molecular analysis and colistin and ceftazidime-avibactam sensitivity testing.

Sample size

The expected mortality rate for CRKp BSI is approximately 40% based on previous studies (7, 12) To detect a 17% decrease in mortality rate with combination therapy as compared to monotherapy, we calculated an overall sample size of 178 (α=0.05, power=0.8) (STATA 15.0, power analysis for the cox proportional hazards model) (7). Considering withdrawals and incomplete data, we aimed to recruit 200 patients.

Given that we were able to recruit patients faster than expected within our study period (25 June 2018- 25 June 2019), we decided to increase our power to 0.9.

Follow up

Patients will be evaluated for microbiologic eradication on day 3 by control blood cultures and primary source cultures. If CRKp is not eradicated on day 3, consecutive cultures will be taken on day 7 and 14, as required. Patients will also be evaluated for clinical response on day 7 and day 14 using 7th and 14th day SOFA scores, C-reactive protein, procalcitonin and leucocyte levels and clinical signs of infection. All patients will be followed for 30 days after index date. If the patient is discharged before completion of the 30 days, he/she will be contacted by phone on day 30 to assess the outcome.

DEFINITIONS

CRKp: K. pneumonia non-susceptible to at least one carbapenem according to EUCAST 2018 breakpoints.

Colistin-resistant (ColR) CRKp: CRKp resistant to colistin according to EUCAST breakpoints

CRKp bacteremia: Presence of at least one set of blood culture positive for CRKp

CRKp BSI: Presence of at least one set of blood culture positive for CRKp with evidence of systemic inflammatory response.

Index date: The day of collection of the CRKp positive blood culture, infection onset, day 0.

Active drug: Antibiotic with in vitro activity against the CRKp isolate (antibiotic which the isolate is susceptible or intermediate to)

Inactive drug: Antibiotic without in vitro activity against the CRKp isolate (antibiotic which the isolate is resistant to)

Empiric therapy: Treatment administered following blood culture collection and prior to antimicrobial susceptibility test results

Targeted therapy: Treatment initiated or maintained after obtaining antimicrobial susceptibility test results

Appropriate therapy: Treatment regimen that includes at least one active drug initiated within 5 days of the index date and administered for at least 48 hours (24 hours for the patients who died within 2 days of the index date)

Early appropriate therapy: Treatment regimen that includes at least one active drug initiated within 2 days of the index date and administered for at least 48 hours (24 hours for the patients who died within 2 days of the index date)

Inappropriate therapy: Treatment regimen that does not include any active drugs and/or treatment administered for less than 48 h and/or treatment initiated after 5 days of the index date.

Monotherapy: Treatment with one active drug

Combination therapy: Treatment with more than one active drug

Primary bacteremia: Catheter-related bacteremia and bacteremia of unknown source

Secondary bacteremia: Identical CRKp growth from a specific body site within BSI window period (3 days before and 3 days after positive blood culture)

Polymicrobial bacteremia: Polymicrobial episodes (defined as ≥2 isolates of different microorganisms in samples taken ≤2 day apart from the same patient

Suspected source: The source of bacteremia determined by the attending physician according to the clinical presentation of the patient, rather than by the culture growth.

Source control: Central line removal, debridement, percutaneous or surgical drainage that are performed within 48 hours of BSI onset.

VARIABLES

Host variables: Age, sex, ward (ICU or non-ICU), hospital type and ward if transferred from another hospital, underlying diseases (Charlson Comorbidity Index), risk factors for acquisition of CRKp (previous infection: prior positive culture with CRKp, hospital stay for >48 hours in last three months, ICU stay for >48 hours in last three months, major surgery in previous month, antibiotic use in last three months: Carbapenem, colistin, other, invasive procedures performed within one week of BSI onset (insertion of central venous catheters, nasogastric tubes, foley catheters; endoscopy; endoscopic retrograde cholangiopancreatography; bronchoscopy; parenteral nutrition; mechanical ventilation), disease severity at index date: SOFA score, Pitt bacteremia score, INCREMENT-CPE score, presentation with septic shock, length of hospital stay until index date

Infection variables: Infection acquisition (Community-acquired vs. hospital-acquired), primary or secondary bacteremia, source control, procalcitonin & C-reactive protein & leucocyte at day 0, 7 and 14, creatinine and glomerular filtration rate (GFR).

Treatment variables: Antimicrobial therapy (duration, dosage (loading and maintenance), presence of prolonged infusion (only for carbapenems))

Microbiological variables: Automated system used for isolate identification and antimicrobial susceptibility testing in the local laboratory, susceptibility profile of the CRKp isolates, MICs for carbapenem, colistin, tigecycline and ceftazidime-avibactam, carbapenemase type, sequence type

MICROBIOLOGICAL STUDIES

Identification and antimicrobial susceptibility testing of CRKp isolates will be performed at local laboratories. MICs will be interpreted according to EUCAST (European Committee on Antimicrobial Susceptibility Testing) criteria. Colistin susceptibility test results will be confirmed and ceftazidime-avibactam susceptibility results will be determined at reference laboratory by broth microdilution. Molecular studies to identify carbapenem resistance genes, colistin resistance mechanisms and CRKp sequence types will be performed at the reference laboratory. Genetic relatedness of the isolates will also be determined at reference laboratory.

STATISTICAL ANALYSIS

Survivor and non-survivor patients with CRKp bacteremia will be compared to identify predictors of mortality. Continuous variables will be compared using t-tests or Mann-Whitney U tests. Categorical variables will be evaluated with the χ2 or Fisher's exact test. Two-tailed tests will be used to determine statistical significance and a p value of < 0.05 will be considered significant.

Effect of treatment type (mono vs. combination therapies, appropriate vs. inappropriate and early vs. late start empiric and targeted therapies) on clinical or microbiological response at day 7 (and 14 if applicable) and on mortality at day 30 will be assessed by logistic regression and cox proportional hazards model, respectively. Independent variables that are either identified from univariate analysis as statistically significant or reported in the literature as being associated with the outcome will be used as covariates. Interaction between treatment type and disease severity/mortality scores will also be assessed and stratified analyses according to disease severity will be performed. Survival curves will be obtained by Kaplan-Meier method and compared using log-rank test.

Propensity score for receiving combination therapy will be calculated using a logistic regression model. Age, sex, hospital, transfer from another hospital, history of hospitalization and antibiotic use, Charlson comorbidity index, ICU stay, SOFA score, Pitt bacteremia score, septic shock, procalcitonin level, infection acquisition type, source of infection will be included as independent variables in the model.

ColR CRKp cases will be compared with colistin-sensitive (ColS) controls in a nested case-control analysis. Competing risks of death for ColR CRKp BSI will be assessed.

Discriminating ability of the models will be assessed by estimating areas under the receiver operating characteristic (ROC) curves. Collinearity will be assessed by calculating variance inflation factors (VIFs) for each dependent variable in the analysis. Akaike information criteria (AIC) will be used for model selection.

Inclusion Criteria:

• All adult patients (≥18 years old) with clinically significant CRKp bacteremia

Exclusion Criteria:

• Polymicrobial infection (except coagulase-negative staphylococci or other skin contaminants isolated from one blood culture bottle, which is considered contamination)
• Subsequent CRKp bacteremia episodes from the same patient

All adult hospitalized patients (≥18 years old) with clinically significant CRKp bacteremia