Background: Patients with sepsis are routinely treated with empiric broad-spectrum antibiotics while awaiting source identification, as recommended by the surviving sepsis campaign.2 Vancomycin, in combination with either piperacillin-tazobactam or cefepime, is commonly used for empiric treatment in these cases. Literature comparing the two regimens has largely focused on the adverse effects of each medication, rather than on overall efficacy. In 2021, Buckley et. al. found no difference in mortality in critical care patients treated with vancomycin and piperacillin-

tazobactam compared to vancomycin and cefepime or meropenem.3 The same year, Ross et. al. compared cefepime to piperacillin-tazobactam in a cohort of 250 patients with septic shock and found a signal toward higher mortality in those receiving cefepime.4 The 2023 ACORN Trial (covered here on REBEL EM) randomized 2,511 patients requiring antipseudomonal coverage to receive either cefepime or piperacillin-tazobactam. Investigators found no statistically significant difference in the incidence of acute kidney injury (AKI)  or death at 14 days, however, those who received cefepime had higher rates of neurologic dysfunction.5 Interestingly, while the effects of piperacillin-tazobactam and cefepime on rates of organ dysfunction have been explored, there are far fewer studies evaluating all-cause mortality as a primary endpoint, which Chanderraj et. al. address in the article discussed below.

Article: Chanderraj R, Admon AJ, He Y, et al. Mortality of Patients With Sepsis Administered Piperacillin-Tazobactam vs Cefepime. JAMA Intern Med. 2024;184(7):769-777. PMID: 38739397

Clinical Question: In adult patients with suspected sepsis treated empirically, does the administration of piperacillin-tazobactam compared to cefepime affect 90-day mortality?

What They Did:

• Investigators performed a retrospective cohort study in the Emergency Department at the University of Michigan between July 1, 2014 through December 31, 2018.
• They analyzed the electronic health records of patients presenting to the ED who met modified Center for Disease Control and Prevention (CDC) sepsis surveillance criteria and received empiric treatment with either vancomycin/piperacillin-tazobactam or vancomycin/cefepime.
• When determining inclusion and exclusion criteria, only the first 24 hours were considered to better emulate a randomized controlled trial.
• Evaluated data  90-day mortality and organ-failure free days, comparing unadjusted data to that adjusted for patient demographics and comorbidities.
• Ran an instrumental variable analysis, using the piperacillin-tazobactam shortage to simulate a randomized control trial in a retrospective manner
• Performed a subgroup analysis, comparing patients that received anaerobic coverage (i.e. Piperacillin-tazobactam or Cefepime and Metronidazole) and those that did not (i.e. Cefepime only)
• In response to the publication of the ACORN trial, they repeated their analysis, only looking at the first 14 days

Population:

Inclusion Criteria:

• Age ≥18, presenting to the ED
• Met the modified Centers for Disease Control and Prevention sepsis surveillance criteria
• Blood samples drawn for culture on arrival
• Evidence of acute organ dysfunction within the first 24 hours
• Antibiotics given for at least 1 day

Exclusion Criteria:

• Patients with an indication for anti-anaerobic therapy 
  • Necrotizing infections, intra-abdominal infections, head and neck infections
• CNS infections
• Transfers from outside hospitals
• Missing data around the source of infection or antibiotic administration

Intervention: Administration of cefepime in combination with vancomycin in the setting of a piperacillin-tazobactam shortage.

Comparator: Empiric therapy with piperacillin-tazobactam and vancomycin for treatment of suspected sepsis.

Outcomes:

Primary outcome: All-cause mortality within 90 days of ED presentation

Secondary outcomes: Organ-failure-free days at 28 days, that is how many days in the first 28 after arriving in the ED not requiring mechanical ventilation, vasopressors, or dialysis.

Results:

Patient Selection:

• 28,627 Screening admissions (July 2014 – December 2018)
  • 9,794 Excluded
    • 1,297 Outside hospital transfers
    • 8,497 Repeat admissions
• 18,833 First admission via emergency department
  • 8,958 Excluded
    • 6,491 Vancomycin not administered
    • 2,467 Neither piperacillin-tazobactam nor cefepime administered
• 9,875 Administration of vancomycin and either piperacillin-tazobactam or cefepime
  • 2,306 Excluded
    • 1,583 Intra-abdominal infection
    • 438 Necrotizing infection
    • 24 Head and neck infection
    • 22 Central nervous system infection
    • 239 Multiple excluded diagnoses
• 7,569 Clinical equipoise between piperacillin-tazobactam and cefepime

• Overall, the sensitivity and subgroup analyses supported the primary findings, suggesting that the association between piperacillin-tazobactam treatment and higher 90-day mortality is robust and not due to confounding factors or biases.

Strengths:

• Patient-Centered Outcomes: The study reports meaningful outcomes directly relevant to patient care, reflecting the real impact of treatments on patient survival and quality of life.
• Real-World Relevance: The findings apply to real-world clinical settings where the choice of antibiotics is critical, providing meaningful insights into antibiotic stewardship and potential harm in empirical therapy decisions.
• Focused Population with Equipoise Between Treatment Groups: The study carefully selects patients with conditions of presumed equipoise between piperacillin-tazobactam and cefepime, which helps reduce bias due to underlying differences in the indication for antibiotic choice.
• Followed STROBE Guidelines: The study adhered to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines, which increase transparency, reproducibility, and methodological rigor, enhancing the credibility of the findings.
• Large Sample Size and Diverse Cohort: The study includes a large cohort of 7,569 patients with sepsis, increasing the statistical power and generalizability of the findings within the context of a single healthcare system.
• Consistency Across Multiple Sensitivity Analyses: Various sensitivity analyses consistently support the primary findings, enhancing the reliability and robustness of the results.

Limitations:

• Retrospective Analysis Design: The study is a retrospective cohort analysis. While using an instrumental variable aims to mimic the conditions of an RCT, it remains susceptible to unmeasured confounding factors that could bias the results.
• Single-Center Study: Conducted at a single healthcare center which limits external validity. The findings may not be generalizable to other settings or populations, especially given the potential variability in local bacterial susceptibility and treatment practices.
• Potential Bias from Unmeasured Confounders: Although the study adjusted for several confounders, there is still a possibility that unmeasured confounders could have influenced the outcomes.
• Use of Antibiotics During Shortage Periods as an Instrument: The natural experiment design based on the antibiotic shortage period may not perfectly emulate random assignment, and there might be other concurrent factors influencing antibiotic choice and patient outcomes that were not accounted for.
• Reliance on Electronic Health Records Data: As with many retrospective studies, reliance on EHR data for identifying eligible patients and outcomes is prone to inaccuracies, missing data, and variability in clinical documentation practices.
• Shortage Period Impact: The use of the piperacillin-tazobactam shortage as a natural experiment may have introduced confounding factors related to antibiotic stewardship changes, healthcare system strain, or other concurrent shifts in clinical practice.
• Instrumental Variable Assumptions Cannot Be Proven: The validity of the instrumental variable analysis relies on several assumptions (e.g., independence of the instrument from unmeasured confounders), which cannot be definitively proven in an observational study.
• Imbalances Between Treatment Groups:
  • ICU Admissions: Higher in piperacillin-tazobactam (33% vs. 30%).
  • Comorbidity Index: Higher in the cefepime group (median 2 vs. 1).
  • Sex: More males in piperacillin-tazobactam (57.1% vs. 52.2%).
  • Racial Minorities: Higher in piperacillin-tazobactam (18.6% vs. 15.3%).
  • Chronic Pulmonary Disease: More in cefepime (25.7% vs. 22.3%).
  • Cancer: Higher in cefepime (19.8% vs. 17.8%).
  • Pneumonia: More in cefepime (18.6% vs. 15.8%).
• No data on antibiotic therapy: The study did not capture information on antibiotic adherence or the duration of antibiotic treatment. Variations in compliance could have impacted patient outcomes.
• Limited information on other interventions: There is minimal data on other critical interventions beyond antibiotics, such as intravenous fluids (IVF), vasopressors, and mechanical ventilation, which could also influence mortality outcomes.

Discussion:

Gut Microbiome: The gut microbiome plays a crucial role in maintaining immune homeostasis and can be disrupted by antibiotics. Pettigrew et. al. evaluated the microbiome of patients in the ICU and found that those who received piperacillin-tazobactam had decreased amounts of protective bacteria, such as Lactobacillus and Clostridiales spp, and increased rates of Enterococcus in the gut microbiota, increasing the risk for serious infections.8 This disruption, likely due to piperacillin-tazobactam’s activity against anaerobic organisms, may explain the higher 90-day mortality rates. Similarly, in this paper, the subgroup analysis of patients who received cefepime with anaerobic coverage had comparable 90-day mortality rates to the piperacillin-tazobactam group, supporting this potential mechanism. Additionally, piperacillin-tazobactam infusions contain a significant sodium load, which may contribute to volume overload in patients with congestive heart failure and worsen outcomes.9 

Instrumental Variable Analysis: An instrumental variable analysis is a tool used in economics and social sciences to evaluate causality in retrospective data, helping to account for bias and distinguish causation from correlation.10 It simulates randomization by using a factor correlated with the intervention. In this case, a piperacillin-tazobactam shortage influenced antibiotic usage based on when patients presented to the ED, creating a “randomization” effect. However, the quality of the results is dependent on the strength of the instrument, and three key criteria must be considered:10, 11

1. Is there a strong correlation between the variable and the outcome?  
2. Is the variable associated with the outcome outside of the intervention? 
3. Is there an association between the instrumental variable and a measured or unmeasured confounder? 

Given the increased availability of pooled data and the increased availability of retrospective data available through the electronic health record, instrumental variable analysis will likely appear in an increasing number of studies. Learning to recognize a strong instrumental variable, as well as the inherent limitations, will likely become increasingly high yield. 

90-day Mortality: The investigators assessed 90-day mortality and conducted a 14-day mortality subgroup analysis for comparison with the ACORN trial. The initial results showed a 5% increase in 90-day mortality with piperacillin-tazobactam. However, the 14-day analysis revealed no significant difference in mortality between the groups. The choice of the 90-day mortality primary outcome is puzzling, antibiotics likely have the biggest impact early in the course of sepsis, within the initial days of treatment. 

Confounders: While instrumental variable analysis aims to mimic a randomized controlled trial, it is an imperfect tool, facing challenges similar to those faced by other retrospective studies. For example, in this analysis, 33% of patients receiving piperacillin-tazobactam were admitted to the ICU, compared to 30% of those receiving cefepime. This discrepancy may be due to limited supplies of piperacillin-tazobactam, with the sickest patients prioritized for the available doses. This confounding factor could contribute to higher mortality by shifting more critically ill patients into the “control” group.

Beyond demographic imbalances, clinical details are notably sparse. The study provides limited information about the patients’ critical care interventions, such as whether they required intubation, mechanical ventilation, vasopressors, or their duration of antibiotic treatment. We also lack adherence data, essential for understanding how consistently patients received their intended therapy. Without these details, it’s challenging to account for variations in patient management that could independently affect mortality rates, leaving room for significant confounding in the analysis.

Additionally, the study’s single-center design limits its external validity. For instance, local resistance patterns at the University of Michigan may have favored cefepime, while other regions might have higher resistance rates, altering outcomes. Similarly, the prevalence of anaerobic infections could make piperacillin-tazobactam more beneficial elsewhere. Differences in practice, like greater familiarity with cefepime or quicker recognition of neurotoxicity, can also influence results in ways that may not apply to other institutions.

Author’s Conclusions: “Among patients with suspected sepsis and no clear indication for anti-anaerobic coverage, administration of piperacillin-tazobactam was associated with higher mortality and increased duration of organ dysfunction compared with cefepime.  These findings suggest the widespread use of empirical anti-anaerobic antibiotics in sepsis may be harmful.”

Clinical Bottom Line:

In this study, cefepime was associated with a 5% reduction in 90-day mortality compared to piperacillin-tazobactam, among patients with undifferentiated sepsis without a clear indication for anaerobic coverage. However, the ACORN Study, a large, prospective, multicenter RCT demonstrated no mortality difference between the two antimicrobials. While the findings are interesting, the data require further validation due to the limitations of this single-center, retrospective study.

Follow Me: 

References:

1. Chanderraj R, Admon AJ, He Y, Nuppnau M, Albin OR, Prescott HC, Dickson RP, Sjoding MW. Mortality of Patients With Sepsis Administered Piperacillin-Tazobactam vs Cefepime. JAMA Intern Med. 2024 Jul 1;184(7):769-777. PMID: 38739397
2. Evans L, Rhodes A, Alhazzani W, et al. Surviving sepsis campaign: international guidelines for management of sepsis and septic shock 2021. Intensive Care Med. 2021;47(11):1181-1247. PMID: 34599691
3. Buckley MS, Komerdelj IA, D’Alessio PA, Rangan P, Agarwal SK, Tinta NC, Martinez BK, Ziadat DS, Yerondopoulos MJ, Kobic E, Kane-Gill SL. Vancomycin with concomitant piperacillin/tazobactam vs. cefepime or meropenem associated acute kidney injury in the critically ill: A multicenter propensity score-matched study. J Crit Care. 2022 Feb;67:134-140. PMID: 34768175
4. Ross RC, Rosen AN, Tran KK, Smith KL, Franck AJ. A Comparison Between Cefepime and Piperacillin-Tazobactam in the Management of Septic Shock. Cureus. 2021 Oct 13;13(10):e18742. PMID: 34790489
5. Qian ET, Casey JD, Wright A, et al. Cefepime vs Piperacillin-Tazobactam in Adults Hospitalized With Acute Infection: The ACORN Randomized Clinical Trial. JAMA. 2023;330(16):1557–1567. PMID: 37837651
6. Jalili-Firoozinezhad S, Gazzaniga FS, Calamari EL, Camacho DM, Fadel CW, Bein A, Swenor B, Nestor B, Cronce MJ, Tovaglieri A, Levy O, Gregory KE, Breault DT, Cabral JMS, Kasper DL, Novak R, Ingber DE. A complex human gut microbiome cultured in an anaerobic intestine-on-a-chip. Nat Biomed Eng. 2019 Jul;3(7):520-531. PMID: 31086325
7. Pickard JM, Zeng MY, Caruso R, Núñez G. Gut microbiota: Role in pathogen colonization, immune responses, and inflammatory disease. Immunol Rev. 2017 Sep;279(1):70-89. PMID: 28856738
8. Pettigrew MM, Gent JF, Kong Y, Halpin AL, Pineles L, Harris AD, Johnson JK. Gastrointestinal Microbiota Disruption and Risk of Colonization With Carbapenem-resistant Pseudomonas aeruginosa in Intensive Care Unit Patients. Clin Infect Dis. 2019 Aug 1;69(4):604-613. PMID: 30383203
9. Scheerenberger HR, Kullab S, Elgazzar A, Lewis N, Shams WE. Piperacillin/Tazobactam Use vs Cefepime May Be Associated With Acute Decompensated Heart Failure. Fed Pract. 2024 Feb;41(2):44-47. PMID: 38835923
10. Fu R, Kim SJ. Inferring causality from observational studies: the role of instrumental variable analysis. Kidney Int. 2021 Jun;99(6):1303-1308. PMID: 33811980
11. Zhang Z, Uddin MJ, Cheng J, Huang T. Instrumental variable analysis in the presence of unmeasured confounding. Ann Transl Med. 2018 May;6(10):182. PMID: 29951504
12. Bellomo R, Warrillow SJ, Reade MC. Why we should be wary of single-center trials. Crit Care Med. 2009 Dec;37(12):3114-9. PMID: 19789447
13. van den Berghe G, Wouters P, Weekers F, Verwaest C, Bruyninckx F, Schetz M, Vlasselaers D, Ferdinande P, Lauwers P, Bouillon R. Intensive insulin therapy in critically ill patients. N Engl J Med. 2001 Nov 8;345(19):1359-67. PMID: 11794168.
14. NICE-SUGAR Study Investigators; Finfer S, Chittock DR, Su SY, Blair D, Foster D, Dhingra V, Bellomo R, Cook D, Dodek P, Henderson WR, Hébert PC, Heritier S, Heyland DK, McArthur C, McDonald E, Mitchell I, Myburgh JA, Norton R, Potter J, Robinson BG, Ronco JJ. Intensive versus conventional glucose control in critically ill patients. N Engl J Med. 2009 Mar 26;360(13):1283-97. PMID: 19318384.

Post Peer Reviewed By: Salim R. Rezaie, MD (Twitter/X: @srrezaie)