This study was approved by the Ethics Committee of Chongqing University Cancer Hospital (No. CZLS2024020-A), as it was a retrospective study and the data was anonymous, the informed consents from patients were waived.

The participants were those who underwent non-cardiac surgery under general anesthesia in Chongqing University Cancer Hospital from January 2018 to May 2023. The inclusion criteria were age > 18 years, elective non-cardiac surgery, surgical duration > 1 h, general anesthesia, and hospitalization for at least 2 days after surgery. The exclusion criteria were end-stage kidney disease, not enough data to determine postoperative AKI and data on preoperative glucose was missing. For those who underwent multiple surgeries during the study period, the data were obtained from the initial surgery. The patients consist of two cohorts, the cohort with DM (known DM, referring to the population with a history of DM or long-term use of hypoglycemic drugs or insulin) and the cohort without DM (no known DM).

The data of surgical patients who met the inclusion criteria were collected through the electronic medical record system of Chongqing University Cancer Hospital. The characteristics of patients included age, sex, ASA physical status (ASA PS), body mass index (BMI), smoking, comorbidities (hypertension, coronary artery disease, heart failure, COPD, cancer, stroke, arrythmia, renal disease, liver disease, thyroid disease), long-term medication (NSAIDS, steroid, antihypotensive agents, hypoglycemic agents, insulin), preoperative laboratory tests (hemoglobin, albumin, fasting blood glucose, creatinine), intraoperative data (body temperature, blood pressure, fluid infusion rate, bleeding, transfusion, hydroxyethyl starch, noradrenaline, any episode of hypotension, NSAIDs), and postoperative creatinin levels within 7 days. The blood sample was withdrawn from a peripheral vein on the morning before oral intake within 72 h before surgery, and the preoperative serum glucose level was measured at the central laboratory of Chongqing University Cancer Hospital. If there were multiple fasting blood glucose values before surgery, the latest one to the surgery was used for data analysis.

The primary outcome was postoperative AKI. AKI was defined as an increase in serum creatinine of ≥ 26.5 μmol/L within 48 h or ≥ 1.5 times the baseline value within 7 days after surgery according to KDIGO(Kidney Disease Improving Global Outcomes) criteria (Kellum et al. 2012).

The categorized variables were presented as n (%), and the continuous variables were presented as median (interquartile range) in all patients, DM cohort, and non-DM cohort. The distribution of preoperative fasting glucose values in 1.0 mmol/L increments was plotted in frequency histograms for patients with and without DM.

When the data of preoperative glucose or intraoperative blood pressure or for postoperative AKI diagnosis were missing, the data of the patient were excluded from the final analysis. Other variables (e.g., the patients’ characteristics or laboratory results) had missing values of < 5%, and we imputed the missing data with the median of each cohort and incorporated all the data for the subsequent multivariable analysis.

The restrictive cubed spline model was used to determine the potential associations between preoperative fasting glucose level and postoperative AKI in cohorts with DM and without DM, and logistic regression analysis was used to calculate the adjusted odds ratio (OR) for all the covariables listed in Table 1. The adjusted OR and 95% confidence intervals (95% CI) per 1 mmol/L increment in glucose level for AKI were calculated by logistic regression. The receiver operating characteristic analysis was conducted and Youden’s index which maximizes the sum of sensitivity and specificity was used to identify the cut-off value for preoperative glucose to predict AKI. Wald χ2 test was used to evaluate the difference between the two cohorts, The adjusted OR and 95% CI comparing both cohorts with DM and without DM on AKI over the range of preoperative glucose values from 3 to 14 mmol/L were estimated.

The data were analyzed using Stata16 (Stata Corp.) and R software 4.12(R foundation for statistical computing). The significance of all tests was two-sided, and P < 0.05 was considered statistically significant.