C24-01 Individualized Treatment Effects of Cefepime vs Piperacillin-Tazobactam in Adults Hospitalized With Acute Infection

Abstract Rationale Among adults hospitalized with acute infection a recent randomized trial, Antibiotic Choice on ReNal Outcomes or ACORN, observed no significant average treatment effect of cefepime versus piperacillin-tazobactam with respect to acute kidney injury or death, but did observe fewer days alive and free of delirium and coma with cefepime. Whether the effects of cefepime versus piperacillin-tazobactam on acute kidney injury or on neurologic dysfunction differ for patients based on their individual characteristics is unknown. Methods In a secondary analysis of the ACORN trial, a machine learning model was derived (N = 1255) and externally validated (N = 1256) to predict the individualized effect of treatment with cefepime versus piperacillin-tazobactam on outcomes. The primary outcome was survival without new or worsening acute kidney injury and the secondary outcome was survival without delirium or coma. Results Among the 1256 patients in the validation cohort, the individualized treatment effect predicted by the model for each patient significantly modified the effect of cefepime versus piperacillin-tazobactam on survival without acute kidney injury (Qini coefficient, 1.68; 95% confidence interval, 0.25 to 3.18; P value = 0.01). Patients predicted to experience benefit from cefepime were younger, were more likely to have a skin and soft tissue infection, and had less organ dysfunction. Patients predicted to benefit from piperacillin-tazobactam had higher severity of illness and more severe kidney injury at enrollment. Among the 626 patients predicted to experience a ≥ 5% benefit from cefepime, 81.2% of patients randomized to cefepime survived without acute kidney injury, compared to 74.2% of patients randomized to piperacillin-tazobactam (risk difference, 7.1%; 95% confidence interval, 0.25% to 13.9%). Among the 203 patients predicted to experience a ≥ 5% benefit from piperacillin-tazobactam, 56.6% of patients randomized to cefepime survived without acute kidney injury, compared to 63.0% of patients randomized to piperacillin-tazobactam (risk difference, -6.3%; 95% confidence interval, -20.7% to 8.2%). Survival without delirium or coma was more common in the piperacillin-tazobactam group than the cefepime group, but there was no individual patient level treatment effect identified by the model. Conclusion In this secondary analysis of a randomized trial, the predicted effect of treatment for individual patients modified the effect of cefepime versus piperacillin-tazobactam on survival without acute kidney injury but did not modify the effect on survival without delirium or coma. Future research should evaluate whether use of evidence-based personalized antibiotic selection improves outcomes for adults hospitalized with acute infection. This abstract is funded by: None