We read with great interest the comments made by Cheng et al1 on our work.2
In our study, we classified acute kidney injury (AKI) outcomes using the Kidney Disease Improving Global Outcomes (KDIGO) Clinical Pratice Guidelines for AKI.3 We did not correct 48-h or 7-d creatinine for fluid balance as it is not recommended to do so in KDIGO AKI guidelines.3 We used either the change in creatinine concentration or urine output that classified AKI into the highest grade (Table 1). Although body weight–adjusted creatinine may be interesting to increase the sensitivity of creatinine-based AKI classification and better reflect AKI-associated mortality, it is not clear that such a correction has any added value when both creatinine and urine output criterion are used.4 As sensitivity analyses, we used only the change in creatinine criteria to classify AKI, which gave similar results. We discussed many limitations of our observational cohort study. However, we do not believe that AKI classification bias had an effect on our findings.
TABLE 1. -
KDIGO-AKI Guidelines for adult patients3
||1.5–1.9 times baseline OR
||<0.5 mL/kg/h for 6–12 h
|≥26.5 mmol/L increase
||2.0–2.9 times baseline
||<0.5 mL/kg/h for ≥12 h
||3.0 times baseline OR
||<0.3 mL/kg/h for ≥24 h OR
|Increase in serum creatinine to ≥353.6 mmol/L OR
||Anuria for ≥12 h
|Initiation of renal replacement therapy
KDIGO, Kidney Disease Improving Global Outcome; AKI, acute kidney injury.
Although we excluded patients under renal replacement therapy before the surgery, we did not include in our models other preoperative organ support (use of vasoactive medications or requirement for mechanical ventilation) that may be needed in acute-on-chronic liver failure (ACLF) or acute liver failure before liver transplantation. We agree that it would have been interesting to better define preoperative illness severity by including these variables. However, such data were not collected. Our objective was to measure the effects of fluid balance on AKI outcomes, while adjusting such association for potential confounders. We adjusted our models for illness severity using preoperative hemoglobin and creatinine concentrations, model for end-stage liver disease, and acute liver failure as a transplantation indication. Because ACLF patients will have lower hemoglobin, higher creatinine, and higher model for end-stage liver disease, we may already have captured the effects that ACLF might have on the association between fluid balance and AKI. Nonetheless, we did not measure the specific effects of preoperative organ support on our outcomes.
Finally, our objective was to assess the epidemiological association between fluid balance and AKI after adjusting for potential confounders. We did not build models using all available data to best predict the incidence of AKI at different time points. Because 7-d AKI is probably influenced by many postoperative events and interventions, we chose 48-h AKI as our primary outcome. Including any variable that may be caused by fluid balance and caused or be caused by 7-d AKI in our model (such as early AKI, graft failure, calcineurin inhibitors doses, or any other postoperative event or intervention) may have falsely reduced the observed effects of fluid balance.5,6 Such variables are mediators in the causal pathway between fluid balance and 7-d AKI, and adjusting for mediators is not part of standard association models.7 Including postoperative events in a postoperative AKI predictive model may improve discriminative abilities of such model but may bias epidemiological association models.
We thank Cheng et al1 for their interest in our work, and we hope we addressed their concerns.
1. Cheng Y, Xue F-S, Wan L. Assessing association between intraoperative fluid balance and the risk of acute kidney injury after liver transplantation: methodological issues. Transplantation. 2020; 104:e303
2. Carrier FM, Chassé M, Sylvestre M-P, et al. Effects of intraoperative fluid balance during liver transplantation on postoperative acute kidney injury: an observational cohort study. Transplantation. 2020; 104:1419–1428doi:10.1097/TP.0000000000002998
3. Kidney Disease: Improving Global Outcomes (KDIGO) Acute Kidney Injury Work Group. KDIGO clinical practice guideline for acute kidney injury. Kidney Inter Suppl. 2012; 2:1–138doi:10.1038/kisup.2012.2
4. Liu KD, Thompson BT, Ancukiewicz M, et al.; National Institutes of Health National Heart, Lung, and Blood Institute Acute Respiratory Distress Syndrome Network. Acute kidney injury in patients with acute lung injury: impact of fluid accumulation on classification of acute kidney injury and associated outcomes. Crit Care Med. 2011; 39:2665–2671doi:10.1097/CCM.0b013e318228234b
5. Greenland S. Quantifying biases in causal models: classical confounding vs collider-stratification bias. Epidemiology. 2003; 14:300–306
6. VanderWeele TJ. Principles of confounder selection. Eur J Epidemiol. 2019; 34:211–219doi:10.1007/s10654-019-00494-6
7. Rothman KJ, Lash TL, Greenland S. Modern Epidemiology. 2008;3rd ed. Philadelphia, PA: Lippincott Williams & Wilkins