Letter to “Associations of Intraoperative Hypotension and Vasopressor Administration With Postoperative Acute Kidney Injury in Children Undergoing Liver Transplantation: A Retrospective Cohort Study”

I read with great interest the recent study by Wingert et al.'s study published in Pediatric Anesthesia (2025) [1]. This retrospective cohort study addresses a critical gap in pediatric perioperative care by examining intraoperative hemodynamics and AKI outcomes in a high-risk population. The authors' rigorous methodology, including exploratory cluster analysis, provides valuable insights into hemodynamic phenotypes and their associations with adverse events. However, I wish to raise a methodological concern regarding the selection of variables for the multivariable logistic regression models. The study reports an AKI incidence of 39% in a cohort of 122 children, and the final multivariate model included variables such as emergent status, preoperative GFR, and total bilirubin, as shown in table 2. Notably, variables like Preoperative creatinine, Anesthesia time, and Preoperative ammonia—which seem biologically plausible predictors of AKI—were not included in the final analysis. From table 1, these variables demonstrated no significant associations with AKI in univariate comparisons (Preoperative creatinine: p = 0.736; Anesthesia time: p = 0.288; Preoperative ammonia: p = 0.135). This likely explains their exclusion from the multivariable model, as the authors prioritized variables with stronger univariate signals or clinical justification, such as Preoperative GFR and total bilirubin, which were significant. Conversely, the inclusion of emergent status (p = 0.222), primary surgeon time (p = 0.059), and weight (p = 0.275) in the multivariate model—despite their non-significance in univariate analyses—these inconsistent variable selection criteria may cause our reader confusion. Furthermore, the exclusion of these variables may overlook potential confounders. Preoperative creatinine directly reflects renal baseline, anesthesia time could indicate surgical complexity and exposure to insults, and elevated ammonia might signal hepatic dysfunction impacting renal perfusion. Future studies with larger sample sizes could benefit from incorporating such variables to enhance model comprehensiveness. The exploratory cluster analysis further highlights the complexity of hemodynamic patterns, as seen in figures 1 and 2, but underscores the need for robust variable selection in predictive models. In conclusion, while Wingert et al.'s study contributes importantly to the field, the variable selection process necessitates cautious interpretation of the multivariate results. I encourage further research with larger cohorts to validate these findings and explore omitted variables. Bingshen Han wrote the manuscript. The author has nothing to report. The author has nothing to report. The author has nothing to report. The author declares no conflicts of interest. The author has nothing to report.