Abstract
Background:
Short courses of antibiotics (7–10 days) are effective for uncomplicated gram-negative bloodstream infections (GN-BSI). However, prior studies have been limited to small cohorts of critically ill patients.
Objective:
The objective of this study was to evaluate the safety and efficacy of short courses of therapy compared with longer courses in patients admitted to the intensive care unit (ICU) with GN-BSI.
Methods:
Propensity-matched, retrospective cohort study of critically ill patients with GN-BSI. The primary outcome was a composite of 30-day mortality or 60-day relapse. Secondary endpoints were components of the composite, 30-day relapse, cure with or without adverse drug events (ADE), and ADEs. Regression analysis was performed to identify factors predictive of the composite outcome.
Results:
225 patients were included in the propensity analysis, 145 in the long cohort and 80 in the short cohort. The primary outcome occurred in 3.8% of patients in the short group and 9.0% of patients in the long group (P = 0.24). There was no difference in 30-day mortality (3.8% vs 5.5%, P = 0.79), 60-day relapse (0% vs 3.4%, P = 0.23), or 30-day readmission (20% vs 22.8%, P = 0.76). ADEs were more common in the long group (47.2% vs 34.1%, OR 1.7, 95% CI 1.04–2.9), primarily attributable to diarrhea.
Conclusion and Relevance:
In critically ill patients with GN-BSI, there were no efficacy outcome differences in patients treated with a short course of antibiotics compared with longer. However, patients in the short group were less likely to experience ADE. These findings suggest that short courses of antibiotics are effective for GN-BSI in critically ill patients.
Keywords: antimicrobial stewardship, drug resistance, microbial, durations of therapy, intensive care unit, bacteremia
Introduction
Antibiotic overuse leading to antimicrobial resistance (AMR) is a global health emergency.1 In an effort to curb rising rates of AMR, antimicrobial stewardship (AMS) clinicians have sought out strategies to minimize inappropriate antibiotic use. One such method includes focusing on optimal durations of therapy and challenging dogmas surrounding the need for prolonged courses of antimicrobials for certain infections or high-risk patient populations.2 Clinicians must balance the necessity of sufficient antimicrobial courses with potential consequences of unnecessarily prolonged treatment. Longer exposure to antibiotics is correlated with increased rates of adverse drug events (ADE) and may lead to development of multidrug-resistant organisms (MDRO).3,4 However, critically ill patients with gram-negative bloodstream infections (GN-BSI) may be particularly vulnerable to poor outcomes due to comorbid conditions, high inoculum infections, and residual organ damage due to severity of initial presentation.5
In recent years, clinical trials have demonstrated that shorter durations of antibiotic therapy (7–10 days) are non-inferior to longer courses for uncomplicated GN-BSI.6–8 Notably, these trials excluded many patients with complicated bacteremia and, therefore, had low rates of recruitment of critically ill patients. Across all cohorts of these studies, reported SOFA scores were low with only 1 study describing rates of sepsis or septic shock on admission.6 Furthermore, critically ill cohorts include higher subpopulations with MDRO such as Pseudomonas aeruginosa9 and those with persistent or uncontrolled sources of infection, leading to a degree of nuance and individualized care required for each patient. As a consequence, the optimal duration of antimicrobial therapy in critically ill patients with GN-BSI has not been conclusively established. Due to the paucity of evidence for short courses, anecdotal practice often leads to prolonged courses of 14 days or more of antibiotics in this patient population.10,11
The objective of this study was to compare the outcomes of patients with GN-BSI treated with a short course (≤10 days) versus a prolonged course (>10 days) of antibiotics in patients admitted to the intensive care unit (ICU).
Patients and Methods
Study Cohort
This was a propensity-matched, retrospective cohort study of patients ≥18 years of age with GN-BSI admitted to the ICU from January 1, 2017, to February 28, 2023, at a large academic medical center in the Southeastern United States. Data were collected and entered by infectious diseases trained physicians and pharmacists or trainees closely supervised by infectious diseases specialists. This study was reviewed and approved by the IRB committee with a waiver of HIPAA authorization and informed consent due to the retrospective nature of the study.
Eligibility Criteria
Patients ≥18 years of age were included if they were admitted to any 1 of the 5 ICUs (neurological, cardiovascular, surgical, shock trauma, and medical) with a Gram-negative rod (GNR) isolated in at least 1 blood culture. Patients were excluded if they (1) had a concomitant Gram-positive cocci (GPC), yeast, or mold deemed to be pathogenic isolated in blood cultures during GN-BSI treatment; (2) did not receive active antibiotic treatment within 72 hours of index cultures; (3) had a metastatic site of infection or unresolved infectious source requiring prolonged therapy (osteomyelitis, undrained intra-abdominal abscesses, or infective endocarditis); (4) were discharged on antibiotics for an unknown duration; or (5) experienced a mortality outcome or withdrew care prior to completion of antibiotic therapy.
Outcomes
The primary outcome was a composite of 30-day all-cause mortality or 60-day relapse (defined as isolation of same pathogen at any site in the 60 days after completion of antibiotics). Additional secondary endpoints included 30-day all-cause readmission, clinical cure (defined as any patient not meeting the composite outcome) with or without ADE, rates of individual ADEs, development of new antimicrobial resistance (defined as isolation of the same pathogen at any site with resistance to at least 1 new antibiotic compared with the index culture), hospital length of stay (LOS), and total duration of therapy. ADEs evaluated included acute kidney injury (AKI) (defined as an increase in serum creatinine by ≥0.3 mg/dL within 48 hours or 1.5 times baseline serum creatinine), new renal replacement therapy requirement, Clostridioidies difficile infection (CDI), diarrhea (defined as physician documented diarrhea, presence of test for diarrheal illness such as CDI, or use of anti-motility agent), liver function test (LFT) abnormalities (defined as bilirubin >1.5 times the upper limit of normal or alanine transaminase (ALT) or aspartate transaminase (AST) >2.5 time the upper limit of normal), or peripherally inserted central catheter (PICC)-associated readmission.
Sample Size
Sample size was calculated for the primary composite outcome of 30-day all-cause mortality and 60-day relapse. Previous epidemiological studies have displayed mortality and relapse rates between 5% and 45% for GN-BSI in various populations.5,6,12–14 Using a 1:2 ratio of short to long courses, 71 patients in the short group and 142 patients in the long group were required to provide an 80% power and two-sided α = 0.05 to detect a difference of 15% between the 2 groups.
Statistical Analysis
All statistical analyses were performed using R Statistical Software (v4.3.1, R Core Team 2023, Vienna, Austria) in R Studio. The proportion of patients experiencing the primary composite outcome of 30-day mortality or 60-day relapse was analyzed using the Chi-squared test. To determine the independent predictors of the composite outcome, a multivariable logistic regression was performed. Independent predictors of the composite on univariate regression were included into the multivariable model in a backwards and stepwise fashion with a P-value threshold of <0.1. Pearson’s Chi-squared test or Fischer’s Exact were used to analyze other categorical variables. Independent t-test or Mann-Whitney U were used to analyze continuous variables for normally and nonnormally distributed data, respectively. Results were considered statistically significant at an alpha level of less than 5%.
A propensity-matched analysis was conducted to control for potential confounders for receiving a shorter course of antimicrobials. The propensity score was calculated via multivariable logistic regression of independent predictors of receiving a shorter course of therapy or the composite outcome. The following covariates were included in the regression model to determine propensity scores: age, ICU admission on hospital day 1, cancer or HIV diagnosis, vasopressor use, urinary source of infection, isolation of an MDRO, Charlson Comorbidity Index (CCI), Pitt Bacteremia Score on day 1, and time to active antimicrobial therapy (in hours). Patients receiving antimicrobials for ≤10 days were matched to those receiving >10 days with a 1:2 ratio using the nearest neighbor method and a caliper of 0.2 times the standard deviation of the propensity score. Patients without a match were excluded from the propensity-matched analysis. Standard mean difference (SMD) was calculated for each variable in the matched cohort and an SMD < 0.1 was considered a good match. After propensity matching, the primary outcome, 30-day mortality, and 60-day relapse were reanalyzed using multivariable logistic regression with adjustment for variables with an SMD ≥ 0.1 for a “doubly robust” estimation.15 All other variables were analyzed using Chi-squared, Fischer’s Exact, independent t-test, and Mann-Whitney U as previously stated.
Results
Propensity Matching
A total of 340 patients were included in the unmatched cohort and 225 patients were included in the propensity-matched cohort (Figure 1). Of the 225 patients included in the matched cohort, 80 patients were in the short course group and 145 patients were in the long course treatment group. A majority of patients had a urinary source of infection and Enterobacterales species, particularly Escherichia coli, were the most frequently encountered pathogens. Analysis of baseline characteristics before and after matching demonstrates a well-matched cohort (SMD < 0.1) with the exception of CCI and Pitt Bacteremia Score on day 1 (Table 1). The distribution of propensity scores (Figure 2) in the prematched and postmatched cohort also demonstrates satisfactory matching.
Figure 1.
Study attrition table.
Table 1.
Baseline Demographics.
| Variable | Unmatched cohort |
Matched cohort |
||||
|---|---|---|---|---|---|---|
| Short (n = 88) | Long (n = 252) | P-value | Short (n = 80) | Long (n = 145) | SMD | |
| Age (year), mean (SD) | 64.68 (16.1) | 58.84 (15.5) | 0.003 | 63.34 (16.2) | 62 (14.3) | 0.087 |
| Weight (kg), mean (SD) | 81.85 (26.8) | 83.64 (28.2) | 0.603 | 83.34 (26.7) | 85.9 (30.1) | 0.092 |
| Male, n (%) | 45 (51.1) | 136 (54.0) | 0.738 | 39 (48.8) | 75 (51.7) | 0.06 |
| CCI, mean (SD) | 5.42 (3.1) | 4.87 (3.3) | 0.170 | 5.28 (3.2) | 4.74 (2.9) | 0.174 |
| CHF, n (%) | 18 (20.5) | 45 (17.9) | 0.704 | 16 (20.0) | 29 (20.0) | <0.001 |
| Liver disease, n (%) | 6 (6.8) | 28 (11.1) | 0.342 | 6 (7.5) | 15 (10.3) | 0.100 |
| Diabetes, n (%) | 36 (40.9) | 96 (38.1) | 0.734 | 32 (40.0) | 58 (40.0) | <0.001 |
| CKD, n (%) | 20 (22.7) | 60 (23.8) | 0.952 | 15 (18.8) | 29 (20.0) | 0.032 |
| Cancer, n (%) | 12 (13.6) | 67 (26.6) | 0.020 | 12 (15) | 22 (15.2) | 0.005 |
| HIV/AIDS, n (%) | 2 (2.3) | 7 (2.8) | 1.000 | 2 (2.5) | 3 (2.1) | 0.029 |
| Pitt Bacteremia Day 1, mean (SD) | 2.9 (2.4) | 3.48 (2.7) | 0.078 | 2.88 (2.4) | 3.48 (2.8) | 0.235 |
| Pitt Bacteremia Day 5, mean (SD) | 1.53 (2.6) | 1.67 (2.44) | 0.648 | 1.46 (2.63) | 1.70 (2.39) | 0.093 |
| Delta Pitt Day 5, mean (SD) | –1.63 (2.64) | –1.80 (2.35) | 0.146 | –1.41 (2.57) | –1.79 (2.56) | 0.146 |
| ICU on Day 1, n (%) | 75 (85.2) | 229 (90.9) | 0.200 | 71 (88.8) | 129 (89) | 0.007 |
| Vasopressor use, n (%) | 51 (58) | 156 (61.9) | 0.598 | 48 (60) | 87 (60) | 0.001 |
| ID Consult, n(%) | 33 (37.5) | 99 (39.3) | 0.866 | 29 (36.2) | 57 (39.3) | 0.063 |
| Pathogen information | ||||||
| Escherichia coli, n(%) | 39 (44.3) | 108 (42.9) | 0.910 | 37 (46.2) | 66 (45.5) | 0.015 |
| Klebsiella pneumoniae, n (%) | 14 (15.9) | 53 (21.0) | 0.376 | 12 (15.0) | 24 (16.6) | 0.043 |
| Pseudomonas aeruginosa, n (%) | 6 (6.8) | 24 (9.5) | 0.581 | 4 (5.0) | 13 (9.0) | 0.156 |
| Proteus mirabilis, n (%) | 9 (10.2) | 21 (8.3) | 0.748 | 9 (11.2) | 11 (7.6) | 0.126 |
| Serratia marcescens, n (%) | 5 (5.7) | 11 (4.4) | 0.834 | 4 (5.0) | 7 (4.8) | 0.008 |
| Enterobacter cloacae, n (%) | 4 (4.5) | 9 (3.6) | 0.930 | 4 (5.0) | 5 (3.4) | 0.077 |
| Acinetobacter spp., n (%) | 2 (2.3) | 6 (2.4) | 1.000 | 1 (1.2) | 5 (3.4) | 0.146 |
| Other Enterobacterales, n (%) | 7 (8.0) | 19 (7.5) | 1.000 | 7 (8.8) | 13 (9.0) | 0.008 |
| Other NLF GNR, n (%) | 2 (2.3) | 1 (0.4) | 0.338 | 2 (2.5) | 1 (0.7) | 0.145 |
| MDR organism, n (%) | 12 (13.6) | 43 (17.1) | 0.560 | 12 (15) | 19 (13.1) | 0.055 |
| Source of infection | ||||||
| Urine, n (%) | 45 (51.1) | 110 (43.7) | 0.276 | 40 (50) | 72 (49.7) | 0.007 |
| Intra-abdominal, n (%) | 16 (18.2) | 46 (18.3) | 1.000 | 14 (17.5) | 24 (16.6) | 0.025 |
| Respiratory, n (%) | 8 (9.1) | 33 (13.1) | 0.422 | 7 (8.8) | 20 (13.8) | 0.160 |
| Primary BSI, n (%) | 5 (5.7) | 14 (5.6) | 1.000 | 5 (6.2) | 5 (3.4) | 0.131 |
| Central venous catheter, n (%) | 1 (1.1) | 12 (4.8) | 0.229 | 1 (1.2) | 7 (4.8) | 0.210 |
| SSTI, n (%) | 6 (6.8) | 13 (5.2) | 0.754 | 6 (7.5) | 6 (4.1) | 0.144 |
| Endovascular, n (%) | 0 (0.0) | 3 (1.2) | 0.714 | 0 (0.0) | 1 (0.7) | 0.118 |
| Unknown, n (%) | 7 (8.0) | 21 (8.3) | 1.000 | 7 (8.8) | 10 (6.9) | 0.069 |
| Source control, No, n (%) | 20 (22.7) | 60 (23.8) | 0.988 | 19 (23.8) | 31 (21.4) | 0.059 |
| Active therapy within 24 hours, n (%) | 71 (80.7) | 217 (86.1) | 0.295 | 66 (82.5) | 120 (82.8) | 0.007 |
| Time to antibiotics (hours), mean (SD) | 13.39 (25.6) | 6.77 (25.7) | 0.038 | 12.62 (26.3) | 10.28 (24.2) | 0.093 |
| Duration of therapy (days), mean (SD) | 7.81 (1.43) | 15.19 (3.87) | <0.001 | 7.80 (1.45) | 15.13 (3.69) | 2.613 |
| Inpatient DOT (days), mean (SD) | 6.62 (2.24) | 10.40 (5.14) | <0.001 | 6.49 (2.27) | 10.50 (5.08) | 1.020 |
| Outpatient DOT (days), mean (SD) | 1.21 (2.11) | 4.80 (5.25) | <0.001 | 1.33 (2.18) | 4.63 (4.81) | 0.883 |
| Discharge on PO antibiotics, n (%) | 20 (22.7) | 97 (38.5) | 0.011 | 20 (25.0) | 58 (40.0) | 0.324 |
| High bioavailability, n (%) | 0 (0.0) | 4 (1.6) | 0.036 | 0 (0.0) | 2 (1.4) | 0.367 |
| Moderate bioavailability, n (%) | 14 (15.9) | 54 (21.4) | 14 (17.5) | 34 (23.4) | ||
| Low bioavailability, n (%) | 6 (6.8) | 30 (15.5) | 6 (7.5) | 22 (15.2) | ||
Abbreviations: CCI, Charlson Comorbidity Index; NLF, non-lactose fermenting; CHF, congestive heart failure; CKD, chronic kidney disease; DOT, duration of therapy; HIV/AIDS, human immunodeficiency virus/acquired immunodeficiency syndrome; SD, standard deviation; SMD, standard mean difference; SSTI, skin and soft tissue infection.
Figure 2.
Distribution of propensity scores for matched and unmatched cohort.
Outcomes.
There was no difference in the primary composite outcome between the short course group and the long course group in the propensity matched model (3.8% vs 9%, P = 0.24). In addition, no differences were detected in 30-day mortality (3.8% vs 5.5%, P = 0.79), 30-day readmission (20% vs 22.8%, P = 0.76), 60-day relapse (0% vs 5%, P = 0.23), or hospital LOS (16.8 days vs 22 days, P = 0.2). The percentage of patients who were cured without experiencing any ADEs was significantly higher in the short course group (53.4%) compared with the long course group (41.3%), with an OR of 0.61 (95% CI: 0.38–0.99). Patients cured with ADEs and the 30-day readmission rate did not differ significantly between groups (Table 2).
Table 2.
Patient Outcomes.
| Outcome | Unmatched cohort |
Matched cohort |
||||
|---|---|---|---|---|---|---|
| Short (n = 88) | Long (n = 252) | P-value | Short (n = 80) | Long (n = 145) | P-value | |
| Composite, n (%) | 3 (3.4) | 25 (9.9) | 0.091 | 3 (3.8) | 13 (9.0) | 0.236 |
| Hospital LOS (day), mean (SD) | 17.76 (22.1) | 19.64 (26.2) | 0.547 | 16.8 (22.3) | 22.0 (31.6) | 0.195 |
| ICU LOS (day), mean (SD) | 8.77 (11.3) | 8.54 (13.1) | 0.879 | 8.2 (10.3) | 9.6 (14.7) | 0.464 |
| 30-day mortality, n (%) | 3 (3.4) | 15 (6.0) | 0.522 | 3 (3.8) | 8 (5.5) | 0.791 |
| 60-day relapse, n (%) | 0 (0.0) | 10 (4.0) | 0.126 | 0 (0.0) | 5 (3.4) | 0.227 |
| 30-day readmit, n (%) | 18 (20.5) | 54 (21.4) | 0.967 | 16 (20.0) | 33 (22.8) | 0.756 |
Abbreviations: ICU, Intensive care unit; LOS, length of stay; SD, standard deviation;.
Regarding safety, the rate of ADEs was significantly higher in the long course group (47.2%) compared with the short course group (34.1%) with an OR of 1.7 (95% CI: 1.04–2.9). This difference was driven by rates of diarrhea, which was significantly lower in the short group (12.5% vs 24.6%, P = 0.033). There were no differences in rates of AKI, new renal replacement therapy requirement, CDI, LFT abnormality, or PICC-related readmission. Overall rates of new resistance were low (5.7% vs 7.9%, p = 0.46) with no significant difference between groups (Table 3).
Table 3.
Adverse Drug Events.
| Variable | Short (n = 88) | Long (n = 252) | P-value |
|---|---|---|---|
| Any adverse drug event, n (%) | 30 (34.1) | 119 (47.2) | 0.033 |
| Acute kidney injury, n (%) | 24 (27.3) | 68 (27.0) | 0.958 |
| New renal replacement therapy, n (%) | 4 (4.5) | 17 (6.7) | 0.460 |
| Clostridioidies difficile infection, n (%) | 4 (4.5) | 16 (6.3) | 0.536 |
| Diarrhea, n (%) | 11 (12.5) | 62 (24.6) | 0.017 |
| Liver function abnormalities, n (%) | 9 (10.2) | 35 (13.9) | 0.378 |
| PICC line complication, n (%) | 1 (1.1) | 3 (1.2) | 1.000 |
| Development of resistance, n (%) | 5 (5.7) | 20 (7.9) | 0.485 |
| Time to resistance, median days (IQR) | 95 (68–197) | 121 (23–229) | 0.921 |
Abbreviations: PICC, peripherally inserted central catheter.
Regression analysis, which was modeled to include short duration of therapy, along with CCI and Pitt Bacteremia Score on day 1 due to > 10% SMD after propensity matching did not indicate association for either the composite outcome or 30-day mortality with ≤ 10 days of antimicrobial therapy (Table 4).
Table 4.
Regression-Matched Cohort.
| Variable | Odds ratio (95% CI) | P-value |
|---|---|---|
| Composite outcome | ||
| Short course | 0.457 (0.10–1.51) | 0.239 |
| CCI | 0.898 (0.73–1.08) | 0.275 |
| Pitt day 1 | 1.14 (0.94–1.35) | 0.162 |
| Intercept | 0.10 (0.03–0.30) | <0.001 |
| 30-day mortality | ||
| Short course | 0.74 (0.16–2.73) | 0.671 |
| CCI | 1.00 (0.81–1.21) | 0.978 |
| Pitt day 1 | 1.16 (0.94–1.42) | 0.156 |
| Intercept | 0.03 (0.01–0.12) | <0.001 |
Abbreviations: CCI, Charlson Comorbidity Index; Pitt, Pitt bacteremia score.
Discussion
The results of this study suggest that critically ill patients with GN-BSI can be safely and efficaciously treated with ≤10 days of active antibiotics assuming the absence of metastatic sites of infection or uncontrolled infectious sources. Patients in the short course group received an average of 7.8 days of antibiotic therapy compared with 15.2 days in the long course group which is similar to prior studies evaluating short versus long courses of antibiotics.6–8 In addition, patients in the long course group were more likely to be discharged on oral antibiotics (22.7% vs 38.5%, P = 0.011). Patients were most frequently discharged on oral antibiotics considered to have moderate bioavailability (ciprofloxacin and trimethoprim/sulfamethoxazole) based on definitions by Kutob et al.16
In the treatment of GN-BSI, clinicians are tasked with the challenge of providing sufficient durations of therapy to eradicate organisms and prevent recurrence while also using the shortest efficacious treatment courses to prevent ADEs and AMR. In an effort to curb unnecessary antimicrobial use, there has been a recent influx of evidence evaluating shorter durations of therapy compared with historical, longer durations of therapy for a variety of infectious disease states such as intra-abdominal infections,17 pneumonia,18,19 and complicated urinary tract infections.20 Several additional retrospective studies have provided data to support durations of 7–10 days in treating uncomplicated GN-BSI, especially in patients with a urinary source of infection.21–24 To date, there have been 3 randomized trials addressing optimal durations of therapy for GN-BSI.6–8 While all 3 preexisting trials were well designed and provided valuable evidence for the treatment of GN-BSI, particularly those of the uncomplicated nature, many practitioners point to a lack of critically ill and immunocompromised patients and a small percentage of MDROs included as possible exceptions to shorter courses of antibiotics. A recent retrospective analysis by Soto et al utilized an inverse probability of treatment weighting model for patients with carbapenem-resistant Enterobacterales (CRE). Their evaluation of 183 patients demonstrated no difference in 30-day mortality or recurrent BSI in patients who received shorter durations of therapy (7–10 days) compared with prolonged courses (14–21 days).25 In addition, among a cohort of patients with hematological malignancies and febrile neutropenia diagnosed with Pseudomonas aeruginosa BSI, short courses (8 – 11 days) demonstrated similar outcomes compared with longer courses (12 – 21 days).26 Given results from these recent studies, critically ill patients remain a prominent unstudied group for shorter durations of therapy among the cohort of patients with GN-BSI.
Prior to this review, we are aware of one other retrospective cohort review by Havey et al evaluating short versus long courses of antibiotics in critically ill patients with BSI. That study was confounded by the inclusion of GPC BSI (~56% of the patient population), foci requiring prolonged therapy, persistent bacteremia, high mortality rates, and otherwise dissimilar patient populations with no propensity matching. As a result, the investigators were able to comment on the heterogeneity of prescribing patterns, but unable to draw meaningful conclusions related to outcomes based on duration of therapy.27 We deduce that many clinicians choose to employ prolonged courses of therapy in critically ill patients with GN-BSI, despite the fact that prolonged durations of therapy are based on anecdote rather than high-quality, reproducible evidence.
Similar to the finding of the RCT by Molina et al, the present study demonstrates that in addition to providing similar efficacy outcomes, using shorter durations of therapy is associated with lower rates of ADEs. The aforementioned study completed a DOOR/RADAR (desirability of outcome ranking and response adjusted for duration of antibiotic risk) analysis that demonstrated patients receiving a short course of antibiotics had a 77.7% probability of achieving better outcomes than those treated with a prolonged course.6 While we did not complete a DOOR/RADAR analysis, similar outcomes were evaluated as secondary endpoints (cure, cure with ADE, 30-day readmission, 60-day relapse, and 30-day mortality) and demonstrated supportive findings.
Strengths of this study include well-balanced demographics, comorbidities, and severity of illness on presentation after propensity matching. This reduced the risk of dissimilar patient populations and confounding of outcomes or altered durations of therapy caused by covariates or presenting disease severity. Propensity analysis does not completely eliminate the risk of selection bias, but this risk was further minimized by including CCI and Pitt Bacteremia Score on day 1 in the regression analysis due to a > 10% SMD between groups. In addition, many studies comparing outcomes stratified by treatment duration are likely to encounter immortal time bias whereby events occurring prior to study inclusion time frames are unaccounted for and thus not reflected in the findings. We attempted to minimize this confounder by excluding patients who had a mortality outcome prior to completion of antibiotic therapy. While this prevented a potential skew toward worse outcomes in the short group due to truncated courses, we were unable to completely eliminate immortal time bias as patients had to survive to the end of therapy to be evaluated. This may explain why the mortality rate in the present study is markedly lower than previous characterization and epidemiology studies evaluating GN-BSI.12–14 While this may be viewed as a limitation in some sense, this study design characteristic delineates findings and removes additional confounders and barriers to interpreting our results for applicable patient populations, namely those who survive and display clinical improvement prior to day 5 of therapy.
This study was not without limitations, which are primarily due to the retrospective nature. As mentioned previously, findings may have been subject to confounding from biases such as selection bias and immortal time bias; though efforts were taken to minimize these risks. Moreover, limitations in documentation in the medical record may have led to incomplete data in medical history, source of infection, and insight into clinical decision-making. We attempted to minimize this risk by evaluating ICU length of stay, Pitt Bacteremia score on day 5, and change in Pitt score through day 5 to assess clinical trajectory and included relevant baseline variables in the propensity model. Patients in the long course group saw a similar reduction in Pitt score after 5 days of therapy compared with the short course group, demonstrating improvement in objective measures of clinical status reducing the risk that patients were treated for longer due to lack of response to initial therapy. Finally, the sample size is relatively small, and the study was conducted at a single center. Both of these may limit the generalizability of the results.
Conclusion and Relevance
In conclusion, this trial suggests that ≤10 days of antibiotics should be considered for critically ill patients with GN-BSI who demonstrate clinical improvement (as seen in improvement in Pitt score by day 5). However, optimal durations of therapy should be individualized for each patient. Thus, patients with persistent sources of infection or delayed response to therapy may require longer treatment durations. While additional studies are needed to substantiate these findings, alternatives to prolonged oral or parenteral antimicrobial therapy are needed due to increased risk of ADEs such as CDI, PICC line-associated complications, AMR, and lengthier hospitalizations.
Acknowledgement
All authors affirm that this manuscript is an honest, accurate, and transparent account of the study being reported.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
Footnotes
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
References
- 1.CDC.COVID-19: U.S. Impact on Antimicrobial Resistance, Special Report 2022. U.S. Department of Health and Human Services, CDC; 2022. doi: 10.15620/cdc:117915 [DOI] [Google Scholar]
- 2.Davar K, Clark D, Centor RM, et al. Can the future of ID escape the inertial dogma of its past? The exemplars of shorter is better and oral is the new IV. Open Forum Infect Dis. 2022;10(1):706. doi: 10.1093/ofid/ofac706 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Tamma PD, Avdic E, Li DX, Dzintars K, Cosgrove E. Association of adverse events with antibiotic use in hospitalized patients. JAMA Intern Med. 2017;117(9):1938. doi: 10.1001/jamainternmed.2017.1938 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Kazakova S, Baggs J, McDonald LC, et al. Association between antibiotic use and hospital-onset clostridioides difficile infection in US Acute Care Hospitals, 2006–201: an ecologic analysis. Clin Infect Dis. 2020;70(1):11–18. doi: 10.1093/cid/ciz169 [DOI] [PubMed] [Google Scholar]
- 5.Chelazzi C, Pettini E, Villa G, De Gaudio AR. Epidemiology, associated factors and outcomes of ICU-acquired infections caused by Gram-negative bacteria in critically ill patients: an observational, retrospective study. BMC Anesthesiol. 2015;15:125. doi: 10.1186/s12871-015-0106-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Molina J, Montero-Mateos E, Praena-Segovia J, et al. Seven-versus 14-day course of antibiotics for the treatment of bloodstream infections by Enterobacterales: a randomized, controlled trial. Clin Microbiol Infect. 2022;28(4):550–557. doi: 10.1016/j.cmi.2021.09.001 [DOI] [PubMed] [Google Scholar]
- 7.von Dach E, Albrich WC, Brunel AS, et al. Effect of C-reactive protein-guided antibiotic treatment duration, 7-day treatment, or 14-day treatment on 30-day clinical failure rate in patients with uncomplicated gram-negative bacteremia: a randomized clinical trial. JAMA. 2020;323(21):2160–2169. doi: 10.1001/jama.2020.6348 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Yahav D, Franceschini E, Koppel F, et al. Seven versus 14 days of antibiotic therapy for uncomplicated gram-negative bacteremia: a noninferiority randomized controlled trial. Clin Infect Dis. 2019;69(7):1091–1098. doi: 10.1093/cid/ciy1054 [DOI] [PubMed] [Google Scholar]
- 9.Thaden JT, Park LP, Maskarinec SA, Ruffin F, Fowler VG Jr, van Duin D. Results from a 13-year prospective cohort study show increased mortality associated with bloodstream infections caused by pseudomonas aeruginosa compared to other bacteria. Antimicrob Agents Chemother. 2017;61:6. doi: 10.1128/AAC.02671-16 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Daneman N, Shore K, Pinto R, Fowler R. Antibiotic treatment duration for bloodstream infections in critically ill patients: a national survey of Canadian infectious diseases and critical care specialists. Int J Antimicrob Agents. 2011;38(6):480–485. doi: 10.1016/j.ijantimicag.2011.07.016 [DOI] [PubMed] [Google Scholar]
- 11.Fernandez-Lazaro CI, Brown KA, Langford BJ, Daneman N, Garber G, Schwartz KL. Late-career physicians prescribe longer courses of antibiotics. Clin Infect Dis. 2019;69(9):1467–1475. doi: 10.1093/cid/ciy1130 [DOI] [PubMed] [Google Scholar]
- 12.Tang Y, Xu C, Xiao H, Wang L, Cheng Q, Li X. Gram—negative bacteria bloodstream infections in patients with hematological malignancies—the impact of pathogen type and patterns of antibiotic resistance: a retrospective cohort study. Infect Drug Resist. 2021;14:3115–3124. doi: 10.2147/IDR.S322812 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Kang CI, Kim SH, Park WB, et al. Bloodstream infections caused by antibiotic-resistant gram-negative bacilli: rack factors for mortality and impact on inappropriate initial antimicrobial therapy on outcome. Antimicrob Agents Chemother. 2005;49(2):760–766. doi: 10.1128/AAC.49.2.760-766.2005 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Blot S, Vandewoude K, De Bacquer D, Colardyn F. Nosocomial bacteremia caused by antibiotic-resistant gram-negative bacteria in critically ill patients: clinical outcome and length of hospitalization. Clin Infect Dis. 2002;34(12):1600–1606. doi: 10.1086/340616 [DOI] [PubMed] [Google Scholar]
- 15.Sawyer RG, Claridge JA, Nathens AB, et al. Trial of short-course antimicrobial therapy for intraabdominal infection. N Engl J Med. 2015;372(21):1996–2005. doi: 10.1056/NEJMoa1411162 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Kutob LF, Justo JA, Bookstaver PB, Kohn J, Albrecht H, Al-Hasan MN. Effectiveness of oral antibiotics for definitive therapy of Gram-negative bloodstream infections. Int J Antimicrob Agents. 2016;48(5):498–503. doi: 10.1016/j.ijan-timicag.2016.07.013 [DOI] [PubMed] [Google Scholar]
- 17.Nguyen TL, Collins GS, Spence J, et al. Comparison of the ability of double-robust estimators to correct bias in propensity score matching analysis. Pharmacoepidemiol Drug Saf. 2017;26(12):1513–1519. doi: 10.1002/pds.4325 [DOI] [PubMed] [Google Scholar]
- 18.Chastre J, Wolff M, Fagon JY, et al. Comparison of 8 vs 15 days of antibiotic therapy for ventilator-associated pneumonia in adults: a randomized trial. JAMA. 2003;290(19):2588. doi: 10.1001/jama.290.19.2588 [DOI] [PubMed] [Google Scholar]
- 19.Uranga A, España PP, Bilbao A, et al. Duration of antibiotic treatment in community-acquired pneumonia: a multicenter randomized clinical trial. JAMAIntern Med. 2016;176(9):1257–1265. doi: 10.1001/jamainternmed.2016.3633 [DOI] [PubMed] [Google Scholar]
- 20.Drekonja DM, Trautner B, Amundson C, Kuskowski M, Johnson JR. Effect of 7 vs 14 days of antibiotic therapy on resolution of symptoms among afebrile med with urinary tract infection: a randomized clinical trial. JAMA. 2021;326(4):324–331. doi: 10.1001/jama.2021.9899 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Heil EL, Bork JT, Abbo LM, et al. Optimizing the management of uncomplicated gram-negative bloodstream infections: consensus guidance using a modified delphi process. Open Forum Infect Dis. 2021;8(10):434. doi: 10.1093/ofid/ofab434 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.McAteer J, Lee JH, Cosgrove SE, et al. Defining the optimal duration of therapy for hospitalized patients with complicated urinary tract infections and associated bacteremia. Clin Infect Dis. 2023;76:9. doi: 10.1093/cid/ciad009 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Babich T, Naucler P, Valik JK, et al. Duration of treatment for pseudomonas aeruginosa bacteremia: a retrospective study. Infect Dis Ther. 2022;11(4):1505–1519. doi: 10.1007/s40121-022-00657-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Giannella M, Pascale R, Toschi A, et al. Treatment duration for Escherichia coli bloodstream infection and outcomes: retrospective single-centre study. Clin Microbiol Infect. 2018;24(10):1077–1083. doi: 10.1016/j.cmi.2018.01.013 [DOI] [PubMed] [Google Scholar]
- 25.Soto CL, Hsu AJ, Lee JH, et al. Identifying effective durations of antibiotic therapy for the treatment of carbapenem-resistant enterobacterales bloodstream infections: a multicenter observational study. Clin Infect Dis. 2023;78:27–30. doi: 10.1093/cid/ciad476 [DOI] [PubMed] [Google Scholar]
- 26.Feng X, Qian C, Fan Y, et al. Is short-course antibiotic therapy suitable for Pseudomonas aeruginosa bloodstream infections in onco-hematology patients with febrile neutropenia? Results of a multi-institutional analysis. Clin Infect Dis. 2023;5:605. doi: 10.1093/cid/ciad605 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Havey TC, Fowler RA, Pinto R, Elligsen M, Daneman N. Duration of antibiotic therapy for critically ill patients with bloodstream infections: a retrospective cohort study. Can J Infect Dis Med Microbiol. 2013;24(3):129–137. doi: 10.1155/2013/141989 [DOI] [PMC free article] [PubMed] [Google Scholar]