Related Experiment Video
Updated: Sep 17, 2025

Multiplex Therapeutic Drug Monitoring by Isotope-dilution HPLC-MS/MS of Antibiotics in Critical Illnesses
Published on: August 30, 2018
Does Piperacillin-Tazobactam Increase Mortality Risk Compared With Cefepime? Collider Bias and the Importance of
Fergus Hamilton1,2, Todd C Lee3, George Davey Smith1
1MRC Integrative Epidemiology Unit, University of Bristol, Bristol, United Kingdom.
Background:
Instrumental variable (IV) analysis is a statistical method allowing causal inference under certain assumptions. A recent high-profile IV analysis suggested cefepime was superior to piperacillin-tazobactam in treating sepsis. This study used a worldwide piperacillin-tazobactam shortage as an IV to infer mortality effects. However, this result starkly contrasts with the well-powered ACORN trial, which showed no effect. We discuss important limitations of the IV study, potentially explaining this discrepancy.
Methods:
We used causal diagrams and the potential outcomes framework to describe potential biases. We identified 2 sources: (1) statistical adjustment for metronidazole treatment, leading to collider bias, and (2) operationalization of the treatment variable (exposure coarsening). We performed simulations demonstrating collider bias can explain the results. Finally, we used summary data from the original paper to obtain alternative causal estimates robust to these biases.
Results:
Adjusting for metronidazole, a choice influenced by both the IV (via initial antibiotic) and underlying factors such as disease severity, induces collider bias. Analyses not adjusting for metronidazole show no strong evidence for a mortality difference. However, bias risk from exposure coarsening remains even without adjustment. Re-analyzing summary data provides no compelling evidence for a benefit of cefepime over piperacillin-tazobactam.
Conclusions:
The recent IV analysis does not support a mortality benefit for cefepime; results appear dependent on incorrect analytical choices introducing bias. Clinicians should be aware of IV analysis complexities and assumptions when making causal inferences from observational data, especially when results contradict high-quality trials and antibiotic choice is the exposure.
More Related Videos
09:26Antibiotic Efficacy Testing in an Ex vivo Model of Pseudomonas aeruginosa and Staphylococcus aureus Biofilms in the Cystic Fibrosis Lung
Published on: January 22, 2021
06:55Inverse Probability of Treatment Weighting Propensity Score using the Military Health System Data Repository and National Death Index
Published on: January 8, 2020
Related Concept Videos
Bias in Epidemiological Studies
Types of Biopharmaceutical Studies: Controlled and Non-Controlled Approaches
Non-controlled studies, commonly employed for initial exploration, lack a control group, rendering them susceptible to biases and external influences. In contrast,...
Comparing the Survival Analysis of Two or More Groups
Confounding in Epidemiological Studies