Extended-Infusion versus Standard-Infusion Piperacillin-Tazobactam for Sepsis Syndromes at a Tertiary Medical Center

Abstract

Piperacillin-tazobactam (PTZ) is frequently used as empirical and targeted therapy for Gram-negative sepsis. Time-dependent killing properties of PTZ support the use of extended-infusion (EI) dosing; however, studies have shown inconsistent benefits of EI PTZ treatment on clinical outcomes. We performed a retrospective cohort study of adult patients who received EI PTZ treatment and historical controls who received standard-infusion (SI) PTZ treatment for presumed sepsis syndromes. Data on mortality rates, clinical outcomes, length of stay (LOS), and disease severity were obtained. A total of 843 patients (662 with EI treatment and 181 with SI treatment) were available for analysis. Baseline characteristics of the two groups were similar, except for fewer female patients receiving EI treatment. No significant differences between the EI and SI groups in inpatient mortality rates (10.9% versus 13.8%; P = 0.282), overall LOS (10 versus 12 days; P = 0.171), intensive care unit (ICU) LOS (7 versus 6 days; P = 0.061), or clinical failure rates (18.4% versus 19.9%; P = 0.756) were observed. However, the duration of PTZ therapy was shorter in the EI group (5 versus 6 days; P < 0.001). Among ICU patients, no significant differences in outcomes between the EI and SI groups were observed. Patients with urinary or intra-abdominal infections had lower mortality and clinical failure rates when receiving EI PTZ treatment. We did not observe significant differences in inpatient mortality rates, overall LOS, ICU LOS, or clinical failure rates between patients receiving EI PTZ treatment and patients receiving SI PTZ treatment. Patients receiving EI PTZ treatment had a shorter duration of PTZ therapy than did patients receiving SI treatment, and EI dosing may provide cost savings to hospitals.

INTRODUCTION

The current prevalence of multidrug-resistant (MDR) Gram-negative infections challenges clinicians to ensure appropriate and pharmacokinetically optimized antimicrobial therapy for patients with presumed Gram-negative infections (1–5). Piperacillin-tazobactam (PTZ) is frequently administered as empirical or targeted therapy for hospitalized patients who have risk factors for Pseudomonas aeruginosa or other MDR Gram-negative organisms.

β-Lactam antimicrobials, including PTZ, exhibit time-dependent killing properties, in which fractional time above the MIC is a critical factor determining pharmacological outcomes (6, 7). Monte Carlo simulations have been used to identify a higher probability of target attainment when PTZ is administered using extended infusion (EI) or continuous infusion (CI) for Gram-negative infections, including P. aeruginosa (8–12).

EI and CI PTZ protocols have been adopted in many hospitals in an effort to optimize the effectiveness of PTZ treatment. However, clinical studies have shown inconsistent benefits of EI versus standard-infusion (SI) PTZ protocols, in terms of mortality rates, length of stay (LOS), and clinical outcomes. Many of those studies were limited by small sample sizes, exclusion of patients who did not have microbiologically documented infections, and the use of PTZ doses that might be less effective against highly resistant organisms such as P. aeruginosa (13–17). A recent meta-analysis showed that EI and CI dosing regimens led to decreased mortality rates but no difference in clinical cure rates, compared to SI regimens (17). In order to help reconcile differences among previously published studies, we performed a large retrospective cohort study of patients who received either SI PTZ or EI PTZ therapy for presumed sepsis syndromes, using pharmacokinetically optimized PTZ dosing, and we analyzed clinical outcomes for the entire group and for various patient subsets.

MATERIALS AND METHODS

Study design and population.

New York University Langone Medical Center-Tisch Hospital is a 705-bed tertiary referral hospital in New York City, with 63 adult intensive care unit (ICU)-level beds. An EI PTZ protocol was implemented on October 22, 2009, and was initially available only in adult medical and surgical ICUs. From October 22, 2009, to February 1, 2011, patients who initially received EI PTZ treatment in the ICU were changed to SI PTZ treatment upon transfer out of the ICU; on February 1, 2011, however, the EI protocol was expanded to include general medical and surgical wards. All patients who received EI PTZ treatment at the beginning of therapy were analyzed as members of the EI group even if they had transitioned to SI PTZ treatment outside the ICU. Patients in the EI group received PTZ administered over 4 h, while patients in the SI group received PTZ administered over 30 min. EI patients with sepsis of unknown source, health care-associated pneumonia (HCAP), ventilator-associated pneumonia (VAP), or suspected P. aeruginosa infection received PTZ at 4.5 g administered intravenously (i.v.) every 8 h with a creatinine clearance (CrCl) of >40 ml/min, 3.375 g administered i.v. every 8 h with a CrCl of 20 to 40 ml/min or while undergoing continuous renal replacement therapy (CRRT), or 3.375 g administered i.v. every 12 h with a CrCl of <20 ml/min or while undergoing hemodialysis (HD). EI patients with all other indications received PTZ at 3.375 g administered i.v. every 8 h with a CrCl of >20 ml/min or 3.375 g administered i.v. every 12 h with a CrCl of <20 ml/min. In the SI group, patients with sepsis of unknown source, HCAP, VAP, or suspected P. aeruginosa infection received PTZ at 4.5 g administered i.v. every 6 h with a CrCl of >40 ml/min, 3.375 g administered i.v. every 6 h with a CrCl of 20 to 40 ml/min, 2.25 g administered i.v. every 6 h with a CrCl of <20 ml/min, 2.25 g administered i.v. every 8 h while undergoing HD, or 3.375 g administered i.v. every 6 h while undergoing CRRT. SI patients with all other indications received PTZ at 3.375 g administered i.v. every 6 h with a CrCl of >40 ml/min, 2.25 g administered i.v. every 6 h with a CrCl of 20 to 40 ml/min, 2.25 g administered i.v. every 8 h with a CrCl of <20 ml/min, 2.25 g administered i.v. every 12 h while undergoing HD, or 2.25 g administered i.v. every 6 h while undergoing CRRT. Audit and recommendations for dosing were provided by clinical pharmacists through the hospital's antimicrobial stewardship program (ASP). This study received approval from the institutional review board of the New York University School of Medicine.

Inclusion and exclusion criteria.

All patients ≥18 years of age who received EI PTZ treatment for presumed sepsis syndromes, whether in an ICU or in a non-ICU ward, between October 22, 2009, and April 21, 2012, were eligible for inclusion in the EI group. Historical controls were selected at random from ICU and non-ICU patients who received SI PTZ treatment for presumed sepsis syndromes between April 21, 2009, and October 21, 2009, prior to implementation of an EI PTZ protocol at our institution. Patients who were eligible for inclusion in the EI group were excluded if they had received >24 h of SI PTZ treatment before starting EI PTZ treatment. Patients meeting two or more criteria for the systemic inflammatory response syndrome (SIRS) with a suspected source of infection were determined to have a presumed sepsis syndrome and were eligible for study inclusion. Patients who were eligible for inclusion in either the EI group or the SI group were excluded if they (i) had an absolute neutrophil count (ANC) of <1,000 cells/μl, (ii) had received PTZ treatment for <48 h, (iii) had been transferred from another facility with prior or unknown PTZ history at the referring facility, (iv) had no identifiable/presumed source of infection, or (v) had their goals of care changed to comfort care or withdrawal of care while receiving PTZ.

Data collection.

Data were extracted from the electronic medical record (EMR) using a standardized chart review. Each member of the chart review team received formal instruction by the study coordinator regarding data collection and entry into a standardized electronic data form. Data elements included age, sex, level of care (ICU or non-ICU), inpatient death, mechanical ventilation, medical comorbidities (including structural lung disease, diabetes mellitus, and cancer), weight, body mass index (BMI), and creatinine level. Hospital LOS, ICU LOS, and duration of PTZ therapy were also calculated. Microbiological data, including source, organism, and MIC/Etest values, were extracted from the EMR when the samples were obtained at the start of PTZ administration. We applied 2012 Clinical and Laboratory Standards Institute (CLSI) guidelines to our MIC/Etest values, and we classified results as highly susceptible at <8 μg/ml, less susceptible at 8 to 16 μg/ml, or clinically resistant at >16 μg/ml. Because of concerns about unreliable automated PTZ susceptibility reporting from Vitek 2 systems (bioMérieux) that occurred during the study period, some Gram-negative rod isolates did not have PTZ susceptibility results reported; these isolates were omitted from further outcome analyses. Modified acute physiology and chronic health evaluation II (mAPACHE) scores were calculated for all patients who received ICU-level care at the time of PTZ administration, omitting Glasgow Coma Scale scores and alveolar-arterial oxygen gradient/arterial oxygen pressure data, as these data were not readily available for most patients. The mAPACHE scores included the most severe/abnormal physiological value for each variable recorded during the first 24 h of ICU admission. An ICU level of care was defined as care in an ICU or a step-down care unit, with both types of units being equipped to handle patients receiving mechanical ventilation or i.v. administration of drugs such as pressors and inotropes.

Progress notes, laboratory values, imaging, and microbiological data in the EMR were analyzed according to predefined criteria (http://tinyurl.com/AAC02759-14) in conjunction with CDC-National Healthcare Safety Network criteria to determine the presumptive source of infection (18). When no source of infection could be identified based on subjective and objective data, cases were classified as “no infection identified” and were excluded from further analysis. Clinical outcomes, defined as clinical improvement/cure or clinical failure, were also determined based on predefined criteria that included both subjective and objective data (http://tinyurl.com/AAC02759-14).

Outcomes.

Primary outcomes in this study included inpatient mortality rates, hospital LOS (from PTZ treatment start date to the date of discharge), and ICU LOS (from PTZ treatment start date to the date of discharge from the ICU). Secondary outcomes included rates of clinical failure and duration of PTZ therapy.

Data analysis.

Data were analyzed using either SPSS v.18 (IBM-SPSS, Chicago, IL) or the R statistical environment (see http://www.R-project.org). Categorical variables were compared using either Pearson's chi-square test or Fisher's exact test, while continuous variables were compared using the Wilcoxon test. P values of <0.05 were considered statistically significant.

RESULTS

Characteristics.

A total of 1,275 patients, including 1,026 in the EI group and 249 in the SI group, met the criteria for inclusion in this study. In the EI group, 364 patients (36%) were excluded, 195 for receiving >24 h of SI PTZ treatment, 12 for neutropenia, 111 for no identifiable infectious source, and 46 for the goals of care changing to comfort care or withdrawal of care. In the SI group, 64 patients (27%) were excluded, 6 for neutropenia, 47 for no identifiable infectious source, and 11 for the goals of care changing to comfort care or withdrawal of care. A total of 662 EI patients and 181 SI patients were subsequently available for analysis. SI patients were significantly more likely to be female (48.6% versus 38.7%; P = 0.016). There were no significant differences in age, weight, BMI, creatinine clearance, level of care, mAPACHE scores, use of mechanical ventilation, or the prevalence of diabetes mellitus, malignancy, or preexisting structural lung disease (Table 1).

TABLE 1.

Patient characteristics

Characteristic	EI PTZ (n = 662)	SI PTZ (n = 181)	P
Median age (yr)	72	74	0.110
Female (%)	38.7	48.6	0.016a
Mean weight (kg)	75.3	73.8	0.433
Mean body mass index (kg/m2)	26.3	26.1	0.749
History of diabetes mellitus (no. [%])	228 (34.4)	61 (33.7)	0.853
History of malignancy (no. [%])	185 (27.9)	60 (33.1)	0.172
History of structural lung disease (no. [%])	149 (22.5)	47 (26)	0.329
Median creatinine clearance (ml/min)	57.1	54.7	0.359
Level of care (no. [% of all patients])			0.626
ICU	493 (74.5)	138 (76.2)
Non-ICU	169 (25.5)	43 (23.8)
mAPACHE score (median ± SD)	14 ± 5.5	14 ± 5.9	NSb
Ventilator use at start of PTZ treatment (no. [%])	187 (28.2)	49 (27.1)	0.755
Presumed source of infection (no. [%])			0.621
Pulmonary	371 (56.0)	95 (52.5)
Intra-abdominal	110 (16.6)	32 (17.7)
Urinary	80 (12.1)	30 (16.6)
Skin and soft tissue	62 (9.4)	16 (8.8)
Bloodstream	24 (3.6)	4 (2.2)
Other	15 (2.3)	4 (2.2)
Microbiologically confirmed infections (no. [%])	371 (56.0)	95 (52.5)	0.494
Pseudomonas isolates (no. [% of microbiologically confirmed infections])	42 (11.4)	12 (12.5)	0.426
Highly susceptible (no. [% of Pseudomonas isolates])	24 (57.1)	5 (41.7)
Less susceptible (no. [% of Pseudomonas isolates])	12 (28.6)	5 (41.7)
Clinically resistant (no. [% of Pseudomonas isolates])	3 (7.1)	2 (8.3)
No susceptibility reported (no. [% of Pseudomonas isolates])	3 (7.1)	0 (0)

^aReached statistical significance.

^bNS, not significant.

Primary outcomes.

There were no statistically significant differences among primary outcomes (Table 2). Overall inpatient mortality rates were 10.9% in the EI group and 13.8% in the SI group (P = 0.282). Among ICU patients, inpatient mortality rates were 14.2% for EI patients and 17.4% for SI patients (P = 0.352). Among patients with mAPACHE scores of ≥15, mortality rates were 19.0% for EI patients and 23.1% for SI patients (P = 0.833). The median hospital LOS was 10 days in the EI group and 12 days in the SI group (P = 0.171), and the ICU LOS was 7 days in the EI group and 6 days in the SI group (P = 0.061).

TABLE 2.

Clinical outcomes of patients receiving EI versus SI PTZ treatment

Patient group and outcome	EI PTZ	SI PTZ	P
All patients
Total no.	662	181
Death (no. [%])	72 (10.9)	25 (13.8)	0.282
Mean LOS (days [median])	14.9 (10)	15.0 (12)	0.171
Mean duration of PTZ treatment (days [median])	5.8 (5)	6.8 (6)	<0.001a
Clinical failure (no. [%])	122 (18.4)	36 (19.9)	0.756
ICU patients
Total no.	493	138
Death (no. [%])	70 (14.2)	24 (17.4)	0.352
Death according to mAPACHE score (%)
Score of <15	10.4	11.4	0.833
Score of ≥15	19.0	23.1	0.833
Mean LOS (days [median])	17.1 (13)	17.1 (14)	0.469
Mean LOS in ICU (days [median])	10.9 (7)	9.2 (6)	0.061
Mean duration of PTZ (days [median])	6.2 (6)	6.8 (6)	0.055
Clinical failure (%)	22.1	23.9	0.624
Non-ICU patients
Total no.	169	43
Death (%)	1.1	2.3	0.571
Mean LOS (days [median])	8.8 (6)	8.4 (8)	0.014a
Clinical failure (%)	7.7	4.7	0.487
Patient subsets (no./total no. [%])
Diabetes mellitus
Death	27/228 (11.8)	7/61 (11.5)	0.937
Malignancy
Death	31/185 (16.8)	8/60 (13.3)	0.529
Structural lung disease
Death	16/149 (10.7)	11/47 (23.4)	0.028a
Microbiologically confirmed infections
Death	46/371 (12.4)	13/95 (13.7)	0.737
Clinical failure	76/371 (20.5)	23/95 (24.2)	0.400
Pseudomonas aeruginosa infections
Death	5/42 (11.9)	3/12 (25.0)	0.260
Clinical failure	11/42 (26.2)	3/12 (25.0)	0.934
Gram-negative rod infections
Death	29/222 (13.1)	8/69 (11.6)	0.749
Clinical failure	50/222 (22.5)	18/69 (26.1)	0.501
Enterobacteriaceae infections
Death	18/162 (11.1)	5/53 (9.4)	0.732
Clinical failure	31/162 (19.1)	13/53 (24.5)	0.502

^aReached statistical significance.

Secondary outcomes and subset analyses.

The duration of PTZ therapy was significantly shorter among EI patients than among SI patients (mean, 5.8 versus 6.8 days [median, 5 versus 6 days]; P < 0.001). Similar trends were seen when the analysis was limited to ICU patients (mean, 6.2 versus 6.8 days [median, 6 versus 6 days]; P = 0.055). Clinical failure rates were almost identical between EI and SI patients, i.e., 18.4% versus 19.9% for all patients (P = 0.756) and 22.1% versus 23.9% for ICU patients (P = 0.624). There were no differences in mortality or clinical failure rates between the EI and SI groups among patients with diabetes mellitus or malignancy (Table 2).

Infection sources and microbiologically confirmed infections.

When results were stratified on the basis of the presumed source of infection, mortality rates were lower among EI patients when the presumed source of infection was either urinary (2.5% versus 16.7%; P = 0.016) or intra-abdominal (7.3% versus 18.8%; P = 0.086), and clinical failure rates were lower among EI patients when the presumed source of infection was urinary (6.3% versus 26.7%; P = 0.006) or intra-abdominal (14.5% versus 25%; P = 0.184) (Table 3). Similar differences in mortality and clinical failure rates were observed among the ICU subset of patients with presumed urinary or intra-abdominal infections (Table 3). While we observed decreased mortality rates among patients with preexisting structural lung disease who received EI PTZ treatment, nonsignificant increases in mortality and clinical failure rates were seen among patients with a presumed pulmonary source of infection who received EI PTZ treatment (Table 3). Subset analysis of patients with microbiologically confirmed infections or P. aeruginosa infections revealed similar mortality and clinical failure rates for the EI and SI groups (Table 2). Among patients with P. aeruginosa infections, there was no difference in mortality rates when data were stratified according to MIC/Etest values, although numbers were small in both the EI and SI groups (data not shown).

TABLE 3.

Clinical outcomes based on presumed source of infection

Patient group, source of infection, and outcome	No. of indicated outcome/total no. (%)		P
	EI PTZ	SI PTZ
All patients
Pulmonary
Death	53/371 (14.3)	10/95 (10.5)	0.403
Clinical failure	90/371 (24.3)	16/95 (16.8)	0.168
Urinary
Death	2/80 (2.5)	5/30 (16.7)	0.016a
Clinical failure	5/80 (6.3)	8/30 (26.7)	0.006a
Intra-abdominal
Death	8/110 (7.3)	6/32 (18.8)	0.086
Clinical failure	16/110 (14.5)	8/32 (25.0)	0.184
Skin and soft tissue
Death	6/62 (9.7)	3/16 (18.8)	0.380
Clinical failure	9/62 (14.5)	2/16 (12.5)	1
Bloodstream
Death	3/24 (12.5)	0/4 (0)	1
Clinical failure	1/24 (4.2)	0/4 (0)	1
Other
Death	0/15 (0)	1/4 (25.0)	1
Clinical failure	1/15 (6.7)	1/4 (25.0)	0.386
ICU patients
Pulmonary
Death	52/306 (17.0)	9/78 (11.5)	0.299
Clinical failure	81/306 (26.5)	15/78 (19.2)	0.240
Urinary
Death	2/47 (4.3)	5/22 (22.7)	0.030a
Clinical failure	3/47 (6.4)	8/22 (36.4)	0.003a
Intra-abdominal
Death	8/80 (10.0)	6/22 (27.3)	0.073
Clinical failure	14/80 (17.5)	7/22 (31.8)	0.149
Skin and soft tissue
Death	5/34 (14.7)	3/13 (23.1)	0.666
Clinical failure	9/34 (26.5)	2/13 (15.4)	0.702
Bloodstream
Death	3/18 (16.7)	0/1 (0)	1
Clinical failure	1/18 (5.6)	0/1 (0)	1
Other
Death	0/8 (0)	0/2 (0)	1
Clinical failure	1/8 (12.5)	1/2 (50.0)	0.378

^aReached statistical significance.

DISCUSSION

Patients receiving EI PTZ treatment had a significantly shorter duration of PTZ therapy than did patients receiving SI PTZ treatment, but we did not find significant differences in mortality rates, overall LOS, or ICU LOS between the treatment groups, even for patients with higher mAPACHE scores. Our results are consistent with those of Lodise et al., which showed no significant differences in mortality rates or LOS among all patients, and those of Patel et al. and Lee et al., which showed no differences in LOS between the two groups (13–15). Our findings are also consistent with those of Lee et al. demonstrating shorter duration of therapy for patients receiving EI PTZ treatment (13). The distribution of the presumed sources of infection varied considerably among all studies, including ours, and this variability limits interstudy comparisons, due to differing levels of antimicrobial tissue penetration at various sites of infection. Lee et al. reported that 86 to 88% of patients had pulmonary infections, compared to 52.5 to 56% of patients in our study (13). While urinary infections in prior studies ranged from 13 to 40%, only 12 to 17% of patients in our study had urinary infections (14–16). Our results also may differ from other studies due to differences in patient selection; our study included patients with presumed sepsis syndromes, while most other studies included only patients with microbiologically confirmed infections. Because accurate microbiological diagnoses can sometimes be elusive in sepsis syndromes, we thought that including all patients with sepsis syndromes would lead to increased generalizability of the results.

When results were stratified by mAPACHE scores, we found no difference in mortality rates for patients receiving EI versus SI PTZ treatment. Our results are consistent with those of Yost et al. but are in contrast to those of Lodise et al., who found that patients with higher APACHE II scores had lower mortality rates if they received EI versus SI PTZ treatment (14, 16). Presumably, septic patients with greater acuity of illness rely more heavily on the bactericidal properties of antimicrobials than do patients with lower acuity of illness (19). These higher-acuity patients would derive greater benefits from pharmacokinetically optimized EI dosing; indeed, this is suggested by the study by Lodise et al. (14). However, the study by Lodise et al. (14) calculated APACHE II scores for all patients and not just ICU patients, the latter of whom represented 65% of the entire study group. It is conceivable that the location of the patient (ICU versus non-ICU) might be more representative of the true severity of illness than the APACHE II score and that this score should be used only to stratify patients in ICU settings, as in our current study.

Subset analyses suggested that patients with preexisting structural lung disease and patients with presumed urinary or intra-abdominal sources of infection receiving EI PTZ treatment had improved clinical outcomes versus patients receiving SI PTZ treatment, while patients with a presumed pulmonary source of infection might have had worse clinical outcomes when receiving EI PTZ treatment. Improved outcomes may have resulted from the optimized pharmacodynamics associated with EI PTZ treatment. Pulmonary infections are often challenging to treat, with frequent involvement of mechanical ventilation and decreased tissue penetration of antimicrobials into lung tissue (19–22). These infections also can be polymicrobial in nature, including P. aeruginosa and other difficult-to-treat Gram-negative rods, which may explain poorer outcomes overall, regardless of the dosing regimen. Because of reduced lung penetration, higher PTZ doses as well as prolonged infusion may be needed to optimize pharmacokinetic/pharmacodynamic parameters when treating presumed pulmonary infections.

To our knowledge, this is the largest study to date comparing EI PTZ and SI PTZ dosing protocols. Our results may be generalizable to acutely hospitalized patients in medical/surgical wards, as well as ICU patients, and can be helpful for ASPs that aim to implement treatment and dosing guidelines. We specifically included patients with presumed sepsis syndromes instead of including only patients with microbiologically confirmed infections, which we thought would be more relevant and clinically applicable, given the frequent empirical utilization of PTZ and the often-unreliable collection of certain types of microbiological samples, including sputum and urine (14–16). In addition, with the increasing use of syndrome-based clinical pathways to guide treatment decisions, clinical data on the impact of empirical antimicrobial dosing strategies on clinical outcomes are greatly needed (23–25). Additional strengths of this study include the collection of data on the presumed source of infection and microbiological data, when available, and the determination of overall clinical outcomes.

Our study had several limitations. We cannot account for variations in antimicrobial prescribing patterns between the two time periods in this study, as well as potential variations in organisms and their susceptibilities over time. In addition, changes in hospital policies that began in 2009 might have introduced systematic confounding. Prior to implementation of our antimicrobial stewardship program in 2009, all PTZ orders required prior authorization from the infectious diseases service. After 2009, any clinician could order PTZ, the use of which was automatically followed by audit and feedback from clinical pharmacists after 72 h of treatment. The original requirement for prior authorization might have deterred some clinicians from ordering PTZ, and this might have excluded patients who might have benefitted most from PTZ treatment.

The retrospective nature of this study required review of both subjective and objective data to determine the presumed sources of infection and clinical outcomes. While our chart reviewers used a well-defined and standardized protocol to minimize personal biases, synthesis of subjective aspects of the chart review was highly reliant on clinical documentation. Many cases were categorized as “no infection identified” because clinical documentation did not support a rationale for PTZ usage. It is possible that patients' presumed sources of infection or clinical outcomes were misclassified because of poor or ambiguous clinical documentation; however, we considered the size of our study to minimize the potential impact of misclassification related to clinical documentation. Except for patients with confirmed P. aeruginosa infections, our study did not assess the concomitant use of other antimicrobials; therefore, some clinical outcomes observed in our study could potentially be attributed to antimicrobials other than PTZ. However, even in the subset of patients with confirmed Gram-negative infections, for whom empirical Gram-positive coverage (e.g., vancomycin) would likely have been stopped, we saw no difference in clinical outcomes between the two groups.

Our enrollment period for EI patients was longer than that for our historical SI control group; therefore, more EI patients were enrolled in our study. Given the similar baseline characteristics of the two groups, we did not think that our numerical imbalance between the EI and SI groups would reduce the power to detect a difference between the two groups. We excluded patients whose goals of care were changed to comfort care or withdrawal of care while they were receiving PTZ, because we thought that it would be impossible to attribute clinical failure to the antimicrobial itself or the dosing regimen in such cases and exclusion of these patients would allow for more-reliable assessment of PTZ efficacy.

Ideally, a randomized controlled trial comparing EI and SI regimens is needed to reconcile differences in outcomes observed in our study and other retrospective studies. However, all studies to date, including our own, suggest that EI PTZ administration is no worse than SI PTZ administration and that EI treatment may provide additional clinical benefits for certain subsets of patients. While we did not perform a formal cost analysis, patients receiving EI PTZ treatment had fewer days of PTZ administration and fewer doses administered per day, which could potentially translate into cost savings for hospitals. By reducing the number of defined daily doses, EI protocols can help optimize antimicrobial resource allocation, reduce antibiotic treatment duration, and minimize adverse events associated with prolonged antimicrobial therapy.