Acute kidney injury following liver transplantation: Pitfalls in the interpretation of risk factors

To the editor, In the article by Berkowitz et al.,[1] the authors studied adult liver transplantation (LT) recipients between 2008 and 2018 to identify intraoperative risk factors contributing to postoperative acute kidney injury (AKI). Using the Kidney Disease Improving Global Outcomes (KDIGO) criteria based on the trends of creatinine or requirement of dialysis within 7 days after LT, the authors show that younger age, cardiac arrhythmias, lower serum creatinine/blood urea nitrogen, and lower Model for End‐Stage Liver Disease (MELD) score are risk factors for AKI. We commend the authors for this exhaustive analysis but are left with some important questions regarding interpretation and generalization of these results. These results are predicated on multiple independent variables that are highly associated with each other. Some examples are MELD score, which already includes serum creatinine, and electrolyte–acid balance abnormalities, which can be caused by either renal insufficiency or liver dysfunction or various medications (e.g., diuretics, fluids, transfusions). In a multivariate analysis, inclusion of such highly associated factors introduces a risk for multicollinearity, making the results less robust and minimizing its generalization across various clinical scenarios. As authors suggested, AKI after LT is a continuum of many pre‐, intra‐, and post‐operative factors. Few relevant immediate and remote pre‐LT factors that are not included in analysis are prior episodes of AKI (frequency, recovery, and remote transient hemodialysis), portal venous thrombosis, use of anticoagulation for arrhythmias/thrombosis, prior episodes of systemic infection impacting end‐organ perfusion, pulmonary hypertension, recent intravenous contrast use, types of antihypertensive medications, nutritional status, and albumin supplementation. All these syndromes can have a distinct, variable, and direct association with independent risk factors in this study. Similarly, the postoperative surgical complications, bleeding, sepsis, transfusion requirement, need for contrast studies, and nephrotoxic medications were not included. When evaluating AKI for up to 7 days after LT, the impact of intraoperative factors alone becomes very difficult to analyze, especially when these defined risk factors are not chronologically one‐off events, are often over‐lapping, and are cumulative. As timing of AKI is variable, it would be interesting if the authors can outline the time of onset of AKI to identify potentially modifiable risk factors that can have significant short‐ and long‐term impact on the LT recipient. We also suggest that patients with high MELD score (potentially high from kidney dysfunction) be excluded in this study. If possible, the authors can account for confounding where patients with lower MELD scores (often with hepatocellular carcinoma) can have a higher probability of receiving LT from a donor after cardiac death. The authors showed significant survival differences based on the dialysis requirement. As KDIGO stage 4 requires dialysis, it would be interesting if the authors could also show specific risk factors associated with AKI stage 4. Since the “safety net” policy was implemented for LT recipients in 2017, it would be of further interest to have outcome data regarding recovery of kidney function or need for subsequent kidney transplantation in this subset. We hope our suggestions will add to the strengths of this study. CONFLICT OF INTEREST Nothing to report.