Effect of intraoperative hypotension depth and duration on acute kidney injury after Type A acute aortic dissection repair: An observational cohort study based on early risk stratification models Authors: Zhou Zhou, Fan Jiang, Anran Dai, Zhixiang Li, Yaoyi Guo, Xinlong Zhang, Yue Feng, Rui Ding, Chen Chen, Kaizong Huang, Ruipeng Jia, Hongwei Shi, Sheng Lou, Jianjun Zou, Yanna Si Published: 2025 ABSTRACT AbstractBackgroundThe impact of adjusting intraoperative mean arterial pressure (MAP) management strategies on early detection and prevention of acute kidney injury (AKI) after Type A acute aortic dissection (TA-AAD) repair has not been elucidated. This study sought to investigate the association between different degrees of hypotension exposure and stage 3 AKI, and examine how intraoperative time-series dynamic variables influence the performance of risk stratification models.MethodsWe analyzed intraoperative data and divided 336 adult patients into groups based on MAP below different thresholds (< 65, 60, 55, 50 mmHg). Logistic regression algorithms were used to identify and screen for predictors other than blood pressure (BP) features and develop initial model. Hypotensive exposure indicators, including cumulative time and area under the curve below a certain MAP threshold, were considered for confounding correction. Subsequently all independent predictors were incorporated to develop upgraded models. Predictive performance was assessed by area under the receiver operating characteristic curve (AUROC) and calibration curves.Results227 patients (67.6 %) developed postoperative AKI, including 114 (33.9 %) stage 1, 54 (16.1 %) stage 2, and 59 (17.6 %) stage 3. Multivariate logistic regression analysis identified preoperative serum creatinine (odds ratio [OR] = 1.007 [95 % CI 1.002–1.015], P = 0.047), operation duration (OR = 1.007 [95 % CI, 1.002–1.012], P = 0.008) and intraoperative urine output (OR = 0.576 [95 % CI, 0.417–0.768], P < 0.001) as independent predictors of stage 3 AKI. After confounding correction, hypotensive exposure indicators were significant at all four thresholds, and ORs increased with decreasing thresholds. Integrating BP features yielded eight upgraded models with AUROC ranging from 0.797 to 0.805.ConclusionsWith a worsening degree of intraoperative hypotension, i.e., lower absolute MAP thresholds and longer exposure times, the odds of stage 3 AKI risk after TA-AAD repair increased. Incorporating BP time-series variables into models could improve the accuracy of early prediction, while the two presentations of hypotension features yield scarcely any difference in predictive outcomes.Keywords: Acute kidney injury, Intraoperative hypotension, Prediction model, Cardiac surgery, Aortic dissection Introduction Acute kidney injury (AKI) as a serious complication is common after the repair of Type A acute aortic dissection (TA-AAD), with the incidence ranging from 23 % to 72 %.1, 2, 3, 4 Its threat to the long-term prognosis is linearly related to the severity of the renal injury itself.5, 6 Among them, severe AKI (the Kidney Disease: Improving Global Outcomes [KDIGO] stage 3) often reflects irreversible renal parenchymal injury and deserves more attention from all clinicians.7 In our past studies, the risk-adjusted relationship between intraoperative hypotension (IOH) and AKI in cardiac surgery has already been demonstrated.8 In general, it is recommended that clinicians preventatively maintain blood pressure (BP) in the appropriate “physiological” range rather than emergency management until the patient has sustained organ-damaging hypotension.9 However, there is still room for in-depth discussion about how the specific depth and duration of the hypotensive exposure affect the development of postoperative stage 3 AKI. Thus, integrating potential risk factors to achieve early and precise identification of high-risk individuals is extremely landmark. Currently, numerous studies have focused solely on the significant impact of perioperative hypotension on postoperative complications,10, 11, 12 or attempted to develop predictive models to assist in the early identification of those at high risk of AKI.13, 14, 15, 16, 17 However, they have not yet adequately integrated the two components mentioned above, i.e., they have not considered including dynamic continuous intraoperative BP measurements in the models. In addition, although the models performed well in terms of predictive accuracy, with their area under the receiver operating characteristic curve (AUROC) ranging from 0.73 to 0.78,13, 14, 15, 16, 17 the reliability of models constructed by relying only on preoperative conventional characteristic variables may be significantly compromised given the complex and diverse pathophysiological pathways that lead to cardiac surgery-associated AKI,18 and some of the laboratory physicochemical indicators are not readily accessible in emergency cases. Hence, the above-reported findings are of limited merit in guiding clinicians to optimize BP management. We hypothesized that lower mean arterial pressure (MAP) thresholds and prolonged exposure to hypotension would have a more severe impact on organ perfusion. This study attempted to clarify whether and to what extent intraoperative hypotension features improve model performance, thereby facilitating the development of perioperative prophylactic intervention strategies. Methods Study design and participants Adult patients with TA-AAD from January 2019 to May 2023 at Nanjing First Hospital were included in this study, and all underwent surgical treatment. According to computed tomography angiography (CTA) results, acute Stanford Type A aortic dissection was diagnosed when the primary tear is located in the ascending aorta, aortic arch, or proximal descending aorta. The dissection involves the ascending aorta, as well as the aortic arch, descending aorta, and abdominal aorta. The flowchart of the study design was shown in the Fig. 1. Exclusion criteria were: (1) patients lacking perioperative serum creatinine (Scr) value; (2) patients with preoperative AKI; (3) patients with severe preoperative renal dysfunction [preoperative renal replacement therapy (RRT) dependence, creatinine concentration > 300 µmol/L or urine output less than 400 mL/d]; (4) patients who died within 48 h after surgery; (5) patients with a history of aortic aneurysm, traumatic and iatrogenic aortic dissection; and (6) patients with connective tissue diseases such as Marfan syndrome, Ehlers-Danlos syndrome, Loeys-Dietz syndrome, and Turner syndrome. Flowchart of study design. We analyzed data from patient electronic health records, including demographics, past medical history, laboratory test results, and anesthesia information statistics. Due to the nature of the retrospective study, the Ethics Committee of Nanjing First Hospital approved the study protocol (permit number: KY20221122–01-KS-01) and waived the requirement of written informed consent for each patient, in accordance with the Declaration of Helsinki. Procedures Patients with TA-AAD were initially stabilized in the intensive care unit (ICU) with a focus on controlling BP and heart rate. In the operating room, the MAP was monitored via invasive cannulation of both the radial artery and the dorsalis pedis arteries. All patients received a combination of intravenous and inhaled general anesthesia to maintain the bispectral index between 40–60. Cardiopulmonary bypass (CPB) commenced post-systemic heparinization, ensuring an activated clotting time longer than 480 s. During CPB, the perfusion flow was 2.0–2.8 L·min−1·m−1, and MAP was maintained at 55–85 mmHg in most cases. As a routine part of the procedure, moderate hypothermic circulatory arrest (MHCA) was maintained at a temperature of 28 ± 0.3 °C to facilitate bloodless surgical conditions and reduce metabolic demands. Surgical monitoring included electrocardiograms, pulse oximetry, nasopharyngeal temperature, bladder temperature, invasive arterial blood pressure, central venous pressure, blood gas analysis results, the end-tidal carbon dioxide concentration and intraoperative fluid management. Post-surgery, patients were transferred back to the ICU with a tracheal tube for further support. Exposures Invasive BP was monitored by ultrasound-guided puncture placement and recorded every five minutes. BP Data from the transient phases of CPB initiation and weaning, i.e., the first 5 and last 5 min, were excluded to avoid artifacts arising from hemodynamic instability. MAP was calculated as 2/3 × diastolic blood pressure (DBP) + 1/3 × systolic blood pressure (SBP), and MAP was recorded directly during CPB. Two metrics were used to quantify IOH: cumulative exposure time and total area under the curve below the MAP threshold (MAP_AUT). We evaluated the risk of postoperative AKI for different hypotensive thresholds in terms of the duration below the threshold and the area under the curve. The area was calculated based on the trapezoidal rule, i.e., area = (d1 + d2) × h × 0.5, where d1 and d2 were the first and second BP deviations (mmHg) from the given threshold, respectively, and h was the duration (minutes) between the two deviations.19 MAP thresholds of 65, 60, 55, and 50 mmHg, respectively, were selected based on prior studies investigating the link between IOH and AKI following cardiac surgery.8 To avoid BP outliers due to machine measurement tools, we used the following criteria to remove intraoperative BP outliers: pressures out-of-range defined by SBP greater than or equal to 300 or SBP less than or equal to 20 mmHg, SBP less than or equal to DBP + 5 mmHg, or DBP less than or equal to 5 mmHg or DBP greater than or equal to 225 mmHg.20 We collected vasopressor and inotropic data during surgery. We calculated the Vaso-Inotropic Score (VIS) using the equation: VIS = dobutamine dose (mcg·kg−1·min−1) + dobutamine dose (mcg·kg−1·min−1) + 100 × epinephrine dose (mcg·kg−1·min−1) + 10 × milrinone dose (mcg·kg−1·min−1) + 10000 × vasopressin dose (U·kg−1·min−1) + 100 × norepinephrine dose (mcg·kg−1·min−1) + 50 × levosimendan dose (mcg·kg−1·min−1).21 Outcomes The primary outcome of this study was the occurrence of postoperative stage 3 AKI, which represents the most severe renal impairment condition. AKI was defined according to the KDIGO criteria.7 Specifically, stage 3 AKI, was defined as a Scr increase ≥ 3.0-fold from baseline, or an absolute Scr ≥ 4.0 mg/dL (353.6 μmol/L), or initiation of RRT within the first 7 postoperative days. The KDIGO urine output guideline was omitted, as it requires urine output of less than 0.5 mL·kg−1·h−1 for 6 h to diagnose AKI. The baseline Scr was determined as the latest Scr measurement taken within 48 h before surgery. Postoperative renal function monitoring followed the standardized protocol of our center to ensure consistency in patient assessment. Initially, the SCr was measured immediately upon admission to the ICU for all patients, and subsequently daily throughout their ICU stay. For patients with faster recovery who were transferred to the general ward, SCr was routinely measured every 48–72 h. Additional tests were performed if clinical symptoms indicative of renal insufficiency emerged or an upward trend in SCr levels was observed. This monitoring schedule adequately covered the early high-risk window for AKI, as the majority of cardiac surgery-related AKI developed within the first 72 h post-operatively.22, 23 Statistical analysis Descriptive statistics were performed on two cohorts in our patient population, with continuous variables expressed as mean (standard deviation) or median (quartile range) and categorical variables expressed as numbers (percentage). For continuous variables, the Student t-test or the Mann-Whitney U-test was applied as appropriate. Categorical variables were analyzed using Fisher’s exact test or the χ2 test. All tests were two-sided, and P < 0.05 was considered statistically significant. We performed statistical analysis using IBM SPSS software (version 25.0) and R version 4.2.2. Variable selection Variable selection was conducted using univariate and multivariate logistic regression analysis methods, incorporating only those variables that achieved statistical significance (P < 0.05) in both rounds of logistic regression analysis into the model. Firstly, univariate analysis was performed to identify potential risk variables for stage 3 AKI. Moreover, we employed collinear analysis to evaluate multicollinearity among the variables in the univariate analysis, using the Variance Inflation Factor (VIF) as the collinear diagnostic index. A factor was removed from the analysis if it had a VIF > 5.24 Subsequently, multivariate logistic regression analysis was conducted to adjust the p-values, odds ratios (OR), and two-sided 95 % confidence intervals (95 % CI) of each variable. Model development and evaluation A three-step sequential approach was employed to construct the predictive model. Firstly, we performed a multivariate logistic regression analysis, including all significant variables from the univariate analysis except for any intraoperative hypotension indicators. The resulting model, termed “the initial model”, was defined as the baseline multivariate logistic regression model adjusted for the independently significant non-hypotension covariates. Then, hypotensive exposure indicators (i.e., cumulative time and total area under the curve below the threshold) at each of the four thresholds (MAP < 65, 60, 55 and 50 mmHg) were integrated with the above independent risk factors, and logistic regression analysis was performed again to correct for confounding. Finally, we developed “the upgraded model” by sequentially incorporating significantly associated hypotensive exposure indicators as additional covariates into “the initial model”, thereby quantifying the incremental value of intraoperative BP management. Model performance was assessed through internal validation with 1000 resamples. First, discriminative ability was evaluated by calculating the AUROC. Then, the optimal threshold was determined using the Youden index (Youden index = sensitivity + specificity − 1). Furthermore, a bootstrap of 1000 resamples was generated to create calibration plots, assessing the gap between the predicted and true values. The performance of the model was further evaluated utilizing the Hosmer-Lemeshow test, where a P-value > 0.05 indicated a good fit. Decision curve analysis (DCA) and clinical impact curve (CIC) analysis were used to evaluate the clinical utility of models.25, 26 Results Baseline Characteristics Based on the exclusion criteria, a total of 336 out of 401 enrolled patients were eligible for this study (Fig. 2). Table 1 showed the baseline characteristics and clinical features of this population. It can be seen that the median age was 54.00 [45.00, 63.00] years. Males accounted for 74.1 % of the patient population. Additionally, 81.2 % of patients had a history of hypertension. Flowchart of participant inclusion and exclusion. Table 1The demographic and clinical characteristics in the whole cohort.Table 1VariablesAll Patients (N = 336)No Stage 3 AKI (N = 277)Stage 3 AKI (N = 59)P-valueDemographicsAge, years54.00 [45.00, 63.00]54.00 [45.00, 65.00]53.00 [43.00, 60.00]0.348Male, n (%)249 (74.1)206 (74.4)43 (72.9)0.942Weight, kg75.00 [65.00, 85.00]75.00 [65.00, 85.00]75.00 [65.00, 89.00]0.697BMI, kg/m²25.71 [23.51, 28.72]25.35 [23.39, 28.37]26.67 [24.30, 29.39]0.051Smoking, n (%)153 (45.5)124 (44.8)29 (49.2)0.638NYHA functional class, n (%)0.396 Ⅰ–Ⅱ196 (58.3)165 (59.6)31 (52.5) Ⅲ–Ⅳ140 (41.7)112 (40.4)28 (47.5)ASA, n (%)0.939 Ⅲ32 (9.5)27 (9.7)5 (8.5) Ⅳ268 (79.8)220 (79.4)48 (81.4) Ⅴ36 (10.7)30 (10.8)6 (10.2)Past medical historyHypertension, n (%)273 (81.2)222 (80.1)51 (86.4)0.347Diabetes mellitus, n (%)12 (3.6)10 (3.6)2 (3.4)1Stroke, n (%)22 (6.5)16 (5.8)6 (10.2)0.343Previous cardiac surgery, n (%)24 (7.1)20 (7.2)4 (6.8)1Preoperative laboratory examinationSerum creatinine, μmol/L79.35 [62.15, 102.05]76.10 [60.70, 96.40]93.30 [72.80, 138.60]< 0.001Urea nitrogen, mmol/L6.70 [5.33, 8.11]6.55 [5.17, 8.05]7.36 [6.28, 8.48]0.002Blood glucose, mmol/L7.47 [6.38, 8.57]7.23 [6.27, 8.44]7.87 [6.77, 8.86]0.031Hemoglobin, g/L133.00 [121.00, 144.25]133.00 [121.00, 143.00]132.00 [119.00, 147.00]0.839Platelet count, 109/L161.00 [135.00, 196.50]164.00 [138.00, 203.00]152.00 [120.00, 177.00]0.032AST, U/L29.00 [23.00, 47.00]28.00 [23.00, 44.00]31.00 [22.00, 63.50]0.299Neutrophil to lymphocyte ratio11.73 [7.41, 15.90]11.41 [6.99, 15.93]13.00 [8.59, 15.82]0.176Leukocyte, 109/L12.12 [9.73, 14.89]11.97 [9.50, 14.57]13.35 [10.58, 15.96]0.043Plasma D-dimer, μg/mL5.16 [2.59, 8.47]4.88 [2.55, 7.82]6.19 [3.29, 10.69]0.044Pulmonary artery pressure, mmHg24.50 [22.00, 29.25]24.00 [22.00, 28.00]26.00 [24.00, 30.00]0.022Left ventricular ejection fraction, %63.00 [61.00, 65.00]63.00 [61.00, 65.00]62.00 [60.00, 64.00]0.077Operative variablesSurgery time, min400.00 [350.00, 465.00]395.00 [345.00, 455.00]450.00 [395.00, 512.50]< 0.001CPB duration, min168.00 [146.00, 190.00]164.00 [144.00, 187.00]177.00 [154.00, 207.00]0.004AO time, min91.00 [79.00, 109.00]90.00 [77.00, 108.00]93.00 [81.50, 113.00]0.334Emergency surgery, n (%)308 (91.7)251 (90.6)57 (96.6)0.21Intraoperative medicationsTypeDopamine, n (%)9 (2.7)6 (2.2)3 (5.1)0.414Dobutamine, n (%)156 (46.4)126 (45.5)30 (50.8)0.545Epinephrine, n (%)25 (7.4)15 (5.4)10 (16.9)0.005Levosimendan, n (%)5 (1.5)3 (1.1)2 (3.4)0.461Milrinone, n (%)89 (26.5)67 (24.2)22 (37.3)0.056Vasopressin, n (%)4 (1.2)3 (1.1)1 (1.7)1Norepinephrine, n (%)209 (62.2)166 (59.9)43 (72.9)0.086Intraoperative diuretics use, n (%)59 (17.6)40 (14.4)19 (32.2)0.002DoseDopamine, μg/kg/min0.00 [0.00, 0.00]0.00 [0.00, 0.00]0.00 [0.00, 0.00]0.198Dobutamine, μg/kg/min0.00 [0.00, 4.00]0.00 [0.00, 4.00]3.00 [0.00, 5.00]0.132Epinephrine, μg/kg/min0.00 [0.00, 0.00]0.00 [0.00, 0.00]0.00 [0.00, 0.00]0.002Levosimendan, μg/kg/min0.00 [0.00, 0.00]0.00 [0.00, 0.00]0.00 [0.00, 0.00]0.19Milrinone, μg/kg/min0.00 [0.00, 0.11]0.00 [0.00, 0.00]0.00 [0.00, 0.40]0.045Vasopressin, units/kg/min0.00 [0.00, 0.00]0.00 [0.00, 0.00]0.00 [0.00, 0.00]0.694Norepinephrine, μg/kg/min0.10 [0.00, 0.20]0.10 [0.00, 0.20]0.10 [0.00, 0.20]0.03Vaso inotropic score10.00 [5.00, 24.00]10.00 [4.00, 20.00]19.00 [9.50, 32.00]0.001Intraoperative fluid managementCrystalloid infusion, mL/kg/h2.90 [2.18, 3.71]2.92 [2.17, 3.68]2.88 [2.23, 3.90]0.641Colloid infusion, mL/kg/h2.60 [2.00, 3.23]2.58 [2.02, 3.21]2.68 [1.92, 3.37]0.922Red blood cell transfusion, unit0.00 [0.00, 2.56]0.00 [0.00, 2.50]0.00 [0.00, 3.00]0.581Cryoprecipitate transfusion, unit17.00 [12.38, 20.00]17.00 [12.00, 20.00]17.25 [12.50, 20.00]0.945Blood platelet transfusion, unit2.00 [1.20, 2.40]2.00 [1.20, 2.40]2.00 [1.60, 2.40]0.611Autologous blood transfusion, mL1000.00 [750.00, 1500.00]1000.00 [750.00, 1500.00]1250.00 [750.00, 2000.00]0.038Intraoperative urine output, mL/kg/h2.69 [1.82, 4.04]2.91 [2.09, 4.41]1.51 [0.91, 2.59]< 0.001Intraoperative blood loss, mL1875.00 [1500.00, 2500.00]1800.00 [1450.00, 2500.00]2000.00 [1675.00, 3000.00]0.003Intraoperative hemodynamic dataMAP_50 time, min100.00 [75.00, 140.00]95.00 [75.00, 135.00]125.00 [75.00, 182.50]0.01MAP_55 time, min135.00 [100.00, 190.00]130.00 [95.00, 180.00]170.00 [112.50, 242.50]0.001MAP_60 time, min180.00 [125.00, 245.00]175.00 [125.00, 230.00]240.00 [157.50, 310.00]< 0.001MAP_65 time, min230.00 [165.00, 300.00]220.00 [165.00, 285.00]290.00 [217.50, 372.50]< 0.001MAP_AUT_50, mmHg*min1467.50 [946.88, 2135.62]1435.00 [922.50, 2022.50]1735.00 [1085.00, 2720.00]0.012MAP_AUT_55, mmHg*min2155.00 [1428.12, 2890.62]2112.50 [1397.50, 2780.00]2457.50 [1583.75, 3891.25]0.009MAP_AUT_60, mmHg*min3005.00 [2029.38, 3968.75]2955.00 [1977.50, 3802.50]3535.00 [2296.25, 5118.75]0.003MAP_AUT_65, mmHg*min4073.75 [2837.50, 5319.38]3930.00 [2810.00, 5055.00]4810.00 [3382.50, 6606.25]0.001Postoperative laboratory examinationPostoperative LDH, U/L611.00 [334.00, 865.50]586.00 [330.00, 848.00]753.00 [447.00, 940.50]0.007Postoperative hemoglobin, g/L99.00 [90.75, 108.25]100.00 [91.00, 109.00]93.00 [85.50, 104.50]0.013Postoperative serum potassium, mmol/L4.03 [3.73, 4.51]4.03 [3.77, 4.50]4.03 [3.56, 4.60]0.743BMI, body mass index; NYHA, New York Heart Association; ASA, American Society of Aneshesiologists; AST, aspartate transaminase; CPB, cardiopulmonary bypass; AO, aortic occlusion; MAP_50/55/60/65 time, cumulative time of mean arterial pressure lower than 50/55/60/65 mmHg; MAP_AUT_50/55/60/65, total area under curve below threshold mean artery blood pressure 50/55/60/65 mmHg; LDH, lactic dehydrogenase. The demographic and clinical characteristics in the whole cohort. BMI, body mass index; NYHA, New York Heart Association; ASA, American Society of Aneshesiologists; AST, aspartate transaminase; CPB, cardiopulmonary bypass; AO, aortic occlusion; MAP_50/55/60/65 time, cumulative time of mean arterial pressure lower than 50/55/60/65 mmHg; MAP_AUT_50/55/60/65, total area under curve below threshold mean artery blood pressure 50/55/60/65 mmHg; LDH, lactic dehydrogenase. The median times at the specified thresholds during surgery were 230.00 [165.00, 300.00] minutes (MAP < 65 mmHg), 180.00 [125.00, 245.00] minutes (MAP < 60 mmHg), 135.00 [100.00, 190.00] minutes (MAP < 55 mmHg), and 100.00 [75.00, 140.00] minutes (MAP < 50 mmHg), respectively. Fig. 3 illustrated the cumulative percentage of patients who spent the amount of duration displayed on the x-axis below a certain threshold. Cumulative percentage of patients with time below a certain threshold shown on the x-axis. MAP, mean arterial pressure. The proportion of patients with a diagnosis of postoperative AKI was 227/336 (67.6 %), with 33.9 %, 16.1 %, and 17.6 % having stage 1, 2, and 3, respectively. Compared with non-stage 3 AKI, patients with severe AKI had significantly higher pre-operative Scr (93.30 [72.80, 138.60] umol·L−1 vs. 76.10 [60.70, 96.40] umol·L−1, P < 0.001) and urea nitrogen (7.36 [6.28, 8.48] mmol·L−1 vs. 6.55 [5.17, 8.05] mmol·L−1, P = 0.002), as well as longer IOH (MAP < 65 mmHg) duration (290.00 [217.50, 372.50] min vs. 220.00 [165.00, 285.00] min, P < 0.001). Independent risk factors selection All variables were included in univariate logistic regression, as shown in Table S1. Several factors significantly influenced stage 3 AKI: ody mass index (BMI), pre-operative Scr, pre-operative urea nitrogen, pre-operative platelet count, operation duration, CPB duration, autologous blood transfusion, intraoperative urine output, intraoperative blood loss, and postoperative LDH. There was no collinearity among the ten factors (VIF < 5) (Table S2). After multivariate logistic regression, the results were shown in Table 2, pre-operative Scr ([umol·L−1], OR = 1.007, 95 % CI, 1.002–1.015, P = 0.047), operation duration ([min], OR = 1.007, 95 % CI, 1.002–1.012, P = 0.008) and intraoperative urine output ([mL·kg−1·h−1], OR = 0.576, 95 % CI, 0.417–0.768, P = 0.000) were independent risk factors. Table 2Univariable and multivariable analyses of associations between various characteristics (except intraoperative BP) and stage 3 AKI.Table 2VariablesUnivariable analysisMultivariable analysisOR95 % CIP valueOR95 % CIP-valueBMI1.0791.009–1.1540.0271.0170.936–1.1040.689Preoperative serum creatinine1.0101.005–1.0160.0001.0071.002–1.0150.047Preoperative urea nitrogen1.1421.047–1.2460.0020.9240.778–1.0730.332Preoperative platelet count0.9940.988–0.9990.0200.9940.986–1.0000.069Operation duration1.0081.004–1.0110.0001.0071.002–1.0120.008CPB duration1.0081.002–1.0150.0140.9970.987–1.0060.522Autologous blood transfusion1.0001.000–1.0010.0021.0000.999–1.0000.875Intraoperative urine output0.4790.360–0.6180.0000.5760.417–0.7680.000Intraoperative blood loss1.0001.000–1.0000.0021.0001.000–1.0000.712Postoperative LDH1.0011.000–1.0010.0011.0001.000–1.0010.295BP, blood pressure; BMI, body mass index; CPB, cardiopulmonary bypass; LDH, lactic dehydrogenase; OR, odds ratio; CI, confidence interval. Univariable and multivariable analyses of associations between various characteristics (except intraoperative BP) and stage 3 AKI. BP, blood pressure; BMI, body mass index; CPB, cardiopulmonary bypass; LDH, lactic dehydrogenase; OR, odds ratio; CI, confidence interval. IOH (MAP below different thresholds) and AKI The BP features under each threshold were added to the independent factors obtained above for the multivariate logistic regression analysis. As shown in Table 3, both the cumulative time and the total area under the curve were significant at all four thresholds, with OR values following an increasing trend as the thresholds decreased. The risk of AKI increased by 1.059 for every 10 min (OR = 1.059, 95 % CI, 1.005–1.118, P = 0.031), and 1.039 for every 100 min × mmHg (OR = 1.039, 95 % CI, 1.009–1.071, P = 0.010) for the most severe persistent hypotension in this study, that is, the MAP below 50 mmHg. Table 3Indicators of hypotension exposure under each threshold were incorporated into the initial model to correct for confounding.Table 3VariablesOR95 % CIP valueMAP_65 time, 10 min1.0401.000–1.0820.049MAP_60 time, 10 min1.0401.001–1.0810.046MAP_55 time, 10 min1.0441.001–1.0890.046MAP_50 time, 10 min1.0591.005–1.1180.031MAP_AUT_65, 100 mmHg·min1.0191.003–1.0350.021MAP_AUT_60, 100 mmHg·min1.0231.004–1.0430.018MAP_AUT_55, 100 mmHg·min1.0301.006–1.0540.013MAP_AUT_50, 100 mmHg·min1.0391.009–1.0710.010MAP_50/55/60/65 time, cumulative time of mean arterial pressure lower than 50/55/60/65 mmHg; MAP_AUT_50/55/60/65, total area under curve below threshold mean artery blood pressure 50/55/60/65 mmHg; OR, odds ratio; CI, confidence interval. Indicators of hypotension exposure under each threshold were incorporated into the initial model to correct for confounding. MAP_50/55/60/65 time, cumulative time of mean arterial pressure lower than 50/55/60/65 mmHg; MAP_AUT_50/55/60/65, total area under curve below threshold mean artery blood pressure 50/55/60/65 mmHg; OR, odds ratio; CI, confidence interval. Nonlinear association between duration of hypotension and AKI As shown in Fig. 4, we plotted the restricted cubic spline (RCS) based on multivariate logistic regression, which visually depicted the relationship between duration of hypotension and stage 3 AKI before and after matching. However, the results of the RCS analysis showed no nonlinear association (P-values for non-linearity were 0.801 and 0.661, respectively, before and after matching). Nonlinear associations between MAP_65 time and the risk of AKI. (a) RCS before matching based on initial model. (b) RCS after matching based on initial model. MAP_65 time, cumulative time of mean arterial pressure lower than 65 mmHg; AKI, acute kidney injury; RCS, restricted cubic spline; OR, odds ratio; CI, confidence interval. Model performance As shown in Fig. 5a, the AUROC of the initial model combining the above three predictors (pre-operative Scr, operative time, and intraoperative urine output) was 0.792 (0.790–0.794), with an optimal threshold of 0.157, at which point the model’s specificity and sensitivity were 0.671 and 0.780, respectively. The model was also well-calibrated (Fig. 5b), and the P-value of the Hosmer- Lemeshow goodness-of-fit test was 0.285. The DCA and the CIC demonstrate that the initial model yields superior overall net benefits based on the actual range of threshold probabilities (Fig. S1 and Fig. S2). By adding cumulative time or total area under the curve below different BP thresholds to the initial model, respectively, we constructed eight upgraded models. After internal validation, all models showed enhanced performance, with the AUROC ranging from 0.797 to 0.805 (Table 4). The P-values derived from the goodness-of-fit tests were uniformly greater than 0.05, indicating a commendable fit for the models. Performance of the initial models. (a) The ROC curves plot. (b) The calibration curves plot. ROC, the receiver operating characteristic curve; AUROC, area under the receiver operating characteristic curve. Table 4Performance evaluation of upgraded models.Table 4Upgraded modelsAUROC (95 % CI)Goodness-of-Fit-Test1Initial model + MAP_65 time0.801 (0.799–0.803)0.1672Initial model + MAP_60 time0.801 (0.799–0.803)0.2023Initial model + MAP_55 time0.799 (0.797–0.801)0.0684Initial model + MAP_50 time0.797 (0.795–0.799)0.2115Initial model + MAP_AUT_650.800 (0.798, 0.802)0.2856Initial model + MAP_AUT_600.800 (0.798, 0.802)0.2507Initial model + MAP_AUT_550.801 (0.799, 0.803)0.2318Initial model + MAP_AUT_500.805 (0.803, 0.807)0.111MAP_50/55/60/65 time, cumulative time of mean arterial pressure lower than 50/55/60/65 mmHg; MAP_AUT_50/55/60/65, total area under curve below threshold mean artery blood pressure 50/55/60/65 mmHg; AUROC, the area under the receiver operating characteristic curve. Performance evaluation of upgraded models. MAP_50/55/60/65 time, cumulative time of mean arterial pressure lower than 50/55/60/65 mmHg; MAP_AUT_50/55/60/65, total area under curve below threshold mean artery blood pressure 50/55/60/65 mmHg; AUROC, the area under the receiver operating characteristic curve. Discussion Based on previous findings from our team and other investigators, we conducted a robust multidimensional analysis of IOH events and attempted to integrate invasive BP measurements, a modifiable risk indicator, into the initial predictive model. Ultimately, this study successfully confirmed the association between multiple hypotensive thresholds and AKI, and reported that upgraded models comprising BP time-series variables could achieve early and accurate screening of high-risk groups for stage 3 AKI. Among the eight upgraded models, the one with the best performance had an AUROC of 0.805, surpassing the 0.792 of the initial 3-variable (pre-operative Scr level, operation duration, and intraoperative urine output) model. It is necessary to acknowledge that perioperative BP control is an interesting and important procedure, trying to identify an absolute threshold below which patients are more likely to develop serious complications. However, most previous studies have focused on patients undergoing non-cardiac surgery,27, 28 and specific detailed intraoperative MAP management strategies for cardiac surgery remain elusive. The novelty of our study is to investigate how adjusting various BP thresholds during TA-AAD repair affects the odds of postoperative stage 3 AKI, based on the manifestation of hypotensive exposure across various dimensions, with horizontal and vertical comparisons. Also, different dimensions of hypotension exposure indicators at different BP thresholds were taken as risk factors, along with traditional predictors, to develop upgraded models and observe whether the accuracy of predicting outcomes could change. In terms of accurately obtaining real-world data, our study has achieved technological improvements. Acute Stanford Type A aortic dissection repair differs from other procedures in that intraoperative BP monitoring is particularly specific, which can be easily affected by the extent and location of the dissection. False lumen formation or branch involvement can cause discrepancies between measured and actual MAP.29 We therefore used preoperative bilateral BP measurements and imaging to guide artery selection for MAP monitoring. Specifically, if the dissection involves the left subclavian artery or brachiocephalic trunk, the higher BP side is monitored. For iliac artery involvement, upper limb BP is monitored. Based on the above method, the true hemodynamic state is not only recorded continuously at 5-minute intervals (including on-CPB and off-CPB periods), but is also presented within an acceptable margin of error, enabling continuous assessment of intraoperative renal hypoperfusion. The optimal presentation of intraoperative hypotensive features remains unclear. Addressing this gap represents a primary objective of the present study. We considered the cumulative time below MAP thresholds of 65, 60, 55, and 50 mmHg as one-dimensional indicators to describe hypotensive events; besides that, in line with our previous study, the impact of altered intraoperative hemodynamics on the development of AKI was investigated from two dimensions, the “depth” of the patient’s real-time BP below a specified absolute threshold and the “length” of the exposure time, whereby MAP_AUT was calculated to reflect the severity of hypotension dynamically. Consistent with de la Hoz et al.,10 we quantified comprehensive BP measurements for each patient throughout the procedure (except for the transient phases of CPB initiation and weaning) to adequately capture the cumulative effect of hypotensive events. A further uniqueness of this study lay in its focus on more targeted endpoints (stage 3 AKI) rather than including all stages. While this approach may sacrifice the additional clinical insights from exploring the earlier stages of AKI, it could offer greater benefits for the high-risk populations, where clinical decision-making is more urgent. Stage 3 AKI, as a severe and irreversible type of renal injury, has been demonstrated in a study with a median follow-up of 58 months following TA-AAD repair to be significantly associated with lower long-term survival.5 Conversely, the hazard ratios for patients with stage 1 and stage 2 AKI were not significantly higher than those for non-AKI patients. And, its diagnosis was less susceptible to confounding factors such as fluid balance and measurement bias, which can sometimes interfere with the identification of stage 1 AKI. Thus, employing stage 3 AKI as the endpoint enhanced the robustness of the model. More importantly, the pathophysiology and risk factors underlying mild and severe AKI were different.30, 31, 32 Tailored risk outcomes were less likely to compromise the ability of developed models or proposed interventional strategies to be deployed accurately at the bedside. Moreover, numerous publications concerning AKI stages 1–3 have been released, including by our team.8, 14, 16, 17 It is reasonable to believe that this exploratory study, based on perioperative hemodynamic optimization, may fill a gap in reducing the incidence of severe AKI after TA-AAD repair. Clinical relevance and practice recommendations To begin with, through multivariate logistic regression analyses, it was readily apparent that both the duration of hypotension and MAP_AUT were independently associated with AKI, irrespective of the definition of hypotension during cardiac surgery. These findings were similar to those of Gregory and colleagues.33 However, our study was the first known to be designed for the population undergoing complex on-pump cardiac surgery, which would be expected to fill the gap in the precise perioperative management of such patients. More importantly, we found that a lower BP threshold was more strongly associated with AKI, regardless of whether it was presented in one- or two-dimensional form. For instance, there was a 5.9 % increase in the risk of stage 3 AKI for every additional 10 min of exposure to MAP below 50 mmHg, whereas when MAP ranged from 60 to 64 mmHg, the increase in risk was only 4.0 %. This association may be attributed to the fact that excessively low BP leads to hypoperfusion, impairing tubule-glomerular functions and disrupting the intrarenal oxygen supply-demand balance34, 35—along with the higher doses of vasoactive agents required at lower BP levels., which may induce renal vasospasm, further increasing the risk of postoperative AKI.36 Nevertheless, given that most TA-AAD patients have a history of chronic hypertension, resulting in rightward shifting of the autoregulatory curves, higher MAP levels may be required.37, 38 It is noteworthy that preoperative comorbid hypertension does not independently increase postoperative AKI risk. Thus, an excessive concern about baseline BP compromising endpoint prediction reliability may be unwarranted. We suggest that, unlike the MAP target of 55 mmHg proposed in non-cardiac surgery,39, 40 MAP ≥ 65 mmHg should be maintained as much as possible to ensure adequate renal perfusion during aortic dissection surgery. Secondly, what is noteworthy to clinicians would be that intraoperative MAP < 65 mmHg lasting more than 230 min substantially increases the risk of stage 3 AKI. As we all know, although anesthetists take active countermeasures, such as vasopressors, inotropic drugs, modulating intravascular volume or pump flow to stabilize intraoperative hemodynamics, sometimes this does not go well enough to avoid hypotensive events. Fortunately, blood flow to the kidneys can be automatically regulated to overcome external shocks.41, 42 We believe that a certain range of BP variations might not exacerbate the development of AKI.43, 44 However, previous research teams, including ours, have also revealed that the prognosis of patients undergoing cardiac surgery gradually worsens with prolonged exposure to hypotension.8, 10, 45, 46 To date, there are unfortunately no guidelines clarifying how long the hypotension lasts before a patient’s AKI risk will skyrocket. Therefore, we sought to plot RCS curves to determine the critical duration. The salient advantage of RCS analysis is the ability to visualize the relationship between continuous variables, like intraoperative BP time-series variables, and the dependent variable without converting the former into classified variables.47 The adjusted curve visually demonstrated that 230 min was the optimal threshold. Until this point was reached, the risk of patients suffering stage 3 AKI remained relatively stable. Moreover, there was no nonlinear association between the duration of IOH (MAP < 65 mmHg) and postoperative AKI. This threshold could serve as a risk stratification tool to alert clinicians that intraoperative persistent hypotension should be avoided whenever possible, and more attention should be directed to patients with prolonged hypotension exposure (> 230 min) during postoperative follow-up care. It is also a distinctive finding that differs from our previous studies. Furthermore, the two-dimensional indicator MAP_AUT provides a more comprehensive picture of hypotensive exposure compared to the cumulative duration under a single dimension. However, taking the accuracy of early assessment of AKI as a measure, we surprisingly found that the performance of the upgraded model combining the duration of hypotension was scarcely inferior to the upgraded model incorporating MAP_AUT. Both models effectively informed clinicians about the likelihood of stage 3 AKI, which is challenging to reverse. As an example, when the MAP threshold was set to 65 mmHg, both achieved an AUROC of 0.800 and 0.801, respectively, outperforming the initial model that failed to consider the BP time-series variables. The development of individualized perioperative medicine in primary healthcare centers often faces greater challenges at present. Given the novel insights of the present study which could not be realized in previous studies, we recommend that the duration of hypotension may still be incorporated into predictive models as reliable dynamic features to assist in clinical decision-making and adjust timely intervention strategies to maximize functional outcomes, especially in resource-limited settings. Lastly, this study simplified the 9-variable machine learning model we have previously developed.8 Apart from intraoperative BP characteristics, only three easily accessible variables, preoperative Scr level, operation duration, and intraoperative urine output, were included. Specifically, elevated preoperative Scr levels imply that the kidneys have already been subjected to a larger metabolic burden.48 Renal injury is more likely to be exacerbated following intraoperative hemodynamic disturbance.49 Extended surgical durations expose patients to increased hazards due to factors such as intraoperative hypothermia, oxidative stress, systemic inflammatory activation and hemolysis.35, 50, 51 Particularly in the face of more complex cardiac surgeries like aortic dissection repair, longer procedures will have more pronounced physiological effects on patients. Intraoperative oliguria, as a response to inadequate renal perfusion, had been proven to be remarkably associated with AKI and was likewise taken into account for the development of predictive models.52, 53 By using the above predictors collected right after surgery, an instant postoperative risk assessment is feasible. This timeliness is crucial, as it enables clinicians to identify high-risk patients for stage 3 AKI at the earliest possible moment, facilitating prompt and targeted interventions upon ICU admission. It is essential to note several limitations of this study. Firstly, as a retrospective cohort study, it was not possible to determine whether there were confounders that had not yet been adjusted for. Despite the observation that IOH was strongly associated with stage 3 AKI after TA-AAD repair, a clear causal link between the two cannot be guaranteed. Secondly, the collection window regarding BP data was limited to the operation period. Hemodynamics during the postoperative care phase, including ICU monitoring, were not analyzed, potentially overlooking the benefit of postoperative management in reducing the incidence of AKI. However, on the one hand, Coulson et al. showed that the inclusion of postoperative variables did not significantly improve model performance.54 This is especially true after one day in the ICU, where it may be too late to predict and intervene in the outcome; on the other hand, to provide a more realistic and precise representation of intraoperative BP fluctuations and compensate for the lack of postoperative data, we ensured that the dynamic continuous BP variables were recorded in detail for each patient at 5-minute intervals, which was not possible in the majority of studies. Thirdly, setting relative MAP thresholds (changes relative to baseline MAP measurements) was not considered in this study. However, this does not detract from the fact that anesthesiologists implement strategic control by observing the absolute changes in BP values. In the real world, patients are prone to induce elevated BP due to emotional stress before surgery. Consequently, baseline MAP values are usually not equivalent to measurements obtained prior to the induction of anesthesia, especially for non-elective procedures such as TA-AAD repair, which in turn biases the calculation of relative differences.55, 56, 57 This is one of the possible reasons why the PeriOperative Quality Initiative (POQI) International Consensus Statement only made relevant recommendations regarding absolute BP thresholds.58 Additionally, several studies have also demonstrated that the use of relative hypotensive events to predict AKI is not superior to absolute hypotension under a specific threshold.20, 59 Also, absolute thresholds are more adapted to the decision-making system, thereby promoting user convenience. Nevertheless, it remains worthwhile to incorporate broader indicators of BP variability in the future, crucially in conjunction with baseline BP levels for exploratory work, to determine personalized intervention thresholds. This would contribute a more comprehensive perspective on BP management. Fourthly, due to the absence of complete TEM classification data for some participants, we regretfully cannot use TEM classification to further evaluate preoperative organ perfusion. Moreover, our study lacked a dynamic assessment of renal artery obstruction. Follow-up studies will be deployed prospectively in a multicenter setting, incorporating TEM classification, to investigate the impact of perioperative organ perfusion on prognosis. Finally, given the limited sample size and ethnic homogeneity, it is undeniable that there may be resistance to generalizing the results of this single-center study, such as the proposed BP management strategies, to all cardiac surgery patients. Multicenter external validations and prospective studies for all types of procedures are urgently needed in the future. Conclusions This study confirmed a statistically significant association between intraoperative hypotensive exposure and increased risk of stage 3 AKI in patients undergoing TA-AAD repair. Renal function progressively deteriorated with decreasing absolute MAP thresholds and in a time-dependent manner. We recommend that intraoperative BP features be incorporated into early risk stratification models to improve predictive accuracy as assisting tools to optimize individualized perioperative management. Meanwhile, it was found that the cumulative time of hypotension may be able to replace the two-dimensional indicator MAP_AUT for precise screening of high-risk populations with stage 3 AKI. In the future, a large-scale multicenter prospective study will be designed to cover richer imaging parameters as well as perioperative dynamic hemodynamic and Scr features. In addition to accurately assisting in the prevention of severe AKI occurrence, attention will be paid to AKI progression time and functional prognosis in high-risk populations, as well as continually refining the thresholds for BP interventions to comprehensively implement individualized short- and long-term high-quality care. CRediT authorship contribution statement Xinlong Zhang: Data curation. Yaoyi Guo: Data curation. Zhixiang Li: Methodology. Yanna Si: Supervision, Conceptualization. Anran Dai: Writing – review & editing, Writing – original draft, Methodology, Formal analysis. Jianjun Zou: Supervision, Conceptualization. Fan Jiang: Writing – review & editing, Writing – original draft, Data curation. Sheng Lou: Supervision, Conceptualization. Hongwei Shi: Writing – review & editing, Conceptualization. Zhou Zhou: Writing – review & editing, Writing – original draft, Formal analysis, Conceptualization. Jia RuiPeng: Writing – review & editing. Kaizong Huang: Data curation. Chen Chen: Writing – review & editing. Rui Ding: Data curation. Yue Feng: Data curation. All authors have read and agreed to the published version of the manuscript. Consent for publication Due to the retrospective nature of the study, the committee waived the requirement for informed consent. Ethical statement The study was conducted in accordance with the Declaration of Helsinki and was registered at the China Clinical Trial Registration Center (trial registry number: ChiCTR2400089726). The study involving human participants was reviewed and approved by the Ethics Committee of Nanjing First Hospital, Nanjing Medical University, permit number KY20221122–01-KS-01, on 24 November 2022. Funding This study was funded by National Natural Science Foundation of China (grant numbers 81873954, 82173899); Six Talent Peaks Project of Jiangsu (grant number WSW-106); Nanjing Medical Science and Technical Development Foundation (grant number ZKX22030); Jiangsu Pharmaceutical Association (grant number H202108, A2021024, Q202202, JY202207, Z04JKM2023E040). Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Acknowledgements We are thankful to all anesthesiologists who aided in the operating theatre, as well as to Bongani Mbambara, Salama Habibu Saad, Chernor Sulaiman Bah, and Iddi Asha Khatib for their contributions to refining the manuscript. Footnotes Supplementary data associated with this article can be found in the online version at doi:10.1016/j.jatmed.2025.12.003. Contributor Information Sheng Lou, Email: lousheng66@126.com. Jianjun Zou, Email: zoujianjun100@126.com. Yanna Si, Email: siyanna@163.com. Supplementary material Supplementary material Data availability The original contributions presented in the study are included in the article and supplementary material. Further data that support the findings of this study are available from the corresponding author on reasonable request. References Associated Data Supplementary Materials Supplementary material Data Availability Statement The original contributions presented in the study are included in the article and supplementary material. Further data that support the findings of this study are available from the corresponding author on reasonable request.