LAUSR

To the Editor: The authors assessed the effects of ERAS protocol related fluid restriction on the occurrence of postoperative acute kidney injury (AKI) and 3-month kidney function. On multivariate analysis they found that use of the ERAS protocol was associated with an increased risk of postoperative AKI, and patients with postoperative AKI had a significantly higher odds ratio of stage 3 chronic kidney disease (CKD) at 3 months following surgery. In addition to limitations described in the discussion section, we noted other issues affecting the interpretation of the findings, and we invite the authors to comment. The occurrence and severity of AKI were defined by the KDIGO (Kidney Disease: Improving Global Outcomes) classification system. We noted that compared to patients on an ERAS protocol, those on a nonERAS protocol not only received significantly larger total fluid volumes during surgery, but also had increased fluid volumes during postoperative days 1 to 4. It was unclear whether the serum creatinine (SCr) levels used for diagnosis and staging of postoperative AKI had been corrected based on perioperative fluid balance. It has been shown that a positive fluid balance may dilute SCr and can significantly affect the definition and staging of postoperative AKI. Especially a small increase in early postoperative SCr levels adjusted for fluid balance can significantly improve diagnosis and severity classifications of postoperative AKI. We were concerned that this unknown factor would result in a false reduction of postoperative AKI due to SCr dilution by increased perioperative fluid volume in patients on a nonERAS protocol. Additionally when multivariate logistic regression analysis was performed to determine independent predictors of postoperative AKI, only age, gender, Charlson comorbidity index, preoperative estimated glomerular filtration rate (eGFR), operative time, ERAS, estimated blood loss, total operative time and type of diversion were included in the model. As a general principle, it is required that all variables identified as significant (p <0.2) in the univariate model be subsequently included in the multivariate model for statistical adjustment. The authors stated that the results of univariate analysis were provided in table 4 in the article but the contents of this table actually were the results of initial comparisons of demographics and intraoperative variables of patients with and without postoperative AKI. Because the readers were not provided details regarding multicollinearity among candidate independent variables by the univariate analysis, we argue that some known risk factors for AKI after noncardiac surgery, such as congestive heart failure, diabetes mellitus, intraoperative fluid volume and blood transfusion, would have been missed. This study also showed that patients with postoperative AKI had a significantly increased risk of stage 3 CKD at 3 months postoperatively, even when controlling for baseline renal function. In fact, this study only considered the effects of AKI occurrence on postoperative long-term adverse renal outcomes. The available evidence indicates that severity and duration of postoperative AKI also are important in the development of long-term adverse renal outcomes including CKD after noncardiac surgery, regardless of preexisting renal function. Moreover, persistent AKI is a main determinant of progression of postoperative CKD. Thus, we believe the findings of this study would be more informative and clinically valuable if the design included the contributions of AKI severity and duration to postoperative long-term adverse renal outcomes. 1352 LETTERS TO THE EDITOR/ERRATA