Prediction of acute kidney injury incidence following acute type A aortic dissection surgery with novel biomarkers: a prospective observational study

This prospective observational study unfolded across three successive cohorts. Perioperative plasma samples and pertinent clinical data were meticulously collated from a consecutive stream of ATAAD-diagnosed patients at Nanjing Drum Tower Hospital, spanning from June 1, 2021, to March 31, 2023. The study inclusion criteria encompassed patients diagnosed with ATAAD via computed tomography angiography (CTA), having undergone surgery within five days of ATAAD onset. Exclusions were extended to encompass patients with preoperative end-stage renal failure, preoperative renal malperfusion, or those succumbing within 48 h post-surgery. Instances of missing plasma samples were similarly omitted. From an initial screening pool of 371 patients, a final cohort of 328 was established and subdivided into three distinct groups: the plasma proteomic cohort (comprising 15 ASA-AKI and 15 non-ASA-AKI patients), the validation cohort (comprising 76 eligible patients), and the model construction cohort (comprising 222 eligible patients).

Concurrently, plasma S100A8/A9, pentraxin 3 (PTX3), and chitinase 3-like 1 (CHI3L1) levels were gauged and juxtaposed against those of 37 age-, gender-, and body mass index- (BMI-) matched healthy individuals. An additional independent set, consisting of 52 patients (equivalently age-, gender-, and BMI-matched), who had undergone cardiopulmonary (CPB) surgery, was enlisted to assess the specificity of plasma proteins in the context of ATAAD.

The incidence of AKI consistently served as the primary outcome throughout the study. The follow-up period remained consistent across all phases of the research, and the sample method was consistently upheld, ensuring uniformity in participant selection. Our emphasis was on maintaining a high level of methodological consistency to guarantee the reliability and validity of our study findings. The acquisition of biological samples adhered to a protocol endorsed by the Institutional Research Ethics Committee of Nanjing Drum Tower Hospital (Approval Number: 2021–084-01), accompanied by informed consent secured from all participating patients.

The diagnostic criteria for AKI were grounded in the Kidney Disease Improving Global Outcomes (KDIGO) guidelines, with assessments conducted within seven days following surgery [17]. The baseline sCr or initial sCr for patients with ATAAD is typically assessed just before the surgical procedure after admission to the ICU. Patient medical records were carefully assessed to validate the diagnosis of chronic kidney disease (CKD) and determine the respective CKD stages. Malperfusion in the context of ATAAD was defined as a decrement in blood flow or complete artery occlusion, substantiated through preoperative CTA and corroborated by clinical indicators such as coma, extremity paralysis, or abdominal pain, along with pertinent laboratory results including elevated myocardial enzymes. The analysis of all CTA images was undertaken independently by two experienced radiologists. Preoperative hypotension was defined by mean arterial pressure falling below 75 mmHg.

Routine laboratory samples, collected via an indwelling venous cannula pre-operatively in the ICU, were continued immediately after surgery and daily up to post-operative day 7 for sCr measurements. These samples were diligently sent for analysis to the hospital's clinical laboratory department. In cases where patients were diagnosed with AKI, plasma samples were extracted almost daily throughout their hospitalization period. This frequent sampling aimed to capture the dynamic fluctuations in sCr levels, enabling a comprehensive understanding of its behavior.

Measurement of S100A8/A9, PTX3, CHI3L1, and NGAL levels was undertaken to employ commercial enzyme-linked immunosorbent assay (ELISA) kits (ABclonal, Wuhan, China) within a span of 3 months from sample collection, in accordance with the manufacturer's instructions. The laboratory analyses were meticulously conducted in a blinded manner, ensuring complete separation from both the associated clinical data and the KDIGO classification. Upon collection, plasma and urinary samples were promptly centrifuged (3000 rpm for 30 min) or refrigerated at 4 °C and subsequently centrifuged within 4 h. Serum and urine supernatants were meticulously collected into EP tubes and expeditiously stored at − 80 °C for subsequent utilization. The limitation of repeated freeze–thaw cycles was meticulously adhered to. Plasma S100A8/A9 was assessed with a 5.0% intra-assay coefficient of variation (CV) and a 7.2% inter-assay CV, establishing a lower detection limit of 3.0 μg/L. PTX3 measurements demonstrated a precision with a 4.1% intra-assay CV and a 4.3% inter-assay CV, highlighting a remarkable lower detection limit of 0.1 ng/mL. Serum concentrations of CHI3L1 were determined with a lower detection limit of 20 ng/mL, displaying an intra-assay CV of 5.0% and an inter-assay CV of 5.4%. In the case of urinary NGAL tests, the lower detection limit was established at 10 ng/mL, while both intra- and inter-assay CVs ranged between 5 and 10% for batched samples analyzed on the same day.

The plasma proteomic cohort consisted of 15 ASA-AKI and 15 non-ASA-AKI patients, matched for age, gender, and BMI, encompassing the period from June 1, 2021, to October 31, 2021. The quantification of protein expression, facilitated through isobaric tags for relative and absolute quantification labeling coupled with liquid chromatography-tandem mass spectrometry analysis, was executed by Applied Protein Technology (Shanghai, China), aligning with established protocols [18]. The schematic depiction of case inclusion and proteomic measurements is presented in Fig. 1a.

Validation cohort: the validation cohort, comprised of 76 eligible patients, was enrolled between November 1, 2021, and March 10, 2022. Blood samples were diligently collected at seven distinct time points: preoperative, postoperative 0 h, postoperative 6 h, postoperative 12 h, postoperative 24 h, postoperative 48 h, and postoperative 72 h. The comprehensive depiction of case inclusions is presented in Fig. 1b.

Healthy and disease control cohort: a total of 37 healthy volunteers, including 27 males and 10 females (with ages ranging from 40 to 72 years; mean age of 55.8 ± 8.2 years), who sought annual physical examinations at Nanjing Drum Tower Hospital between January 1, 2022, and January 1, 2023, were encompassed in this study. A meticulous assessment, inclusive of consultation, physical examination, blood pressure measurements, specialized examinations, and scrutiny of medical records, was conducted for each volunteer. Fasting blood samples were collected from all volunteers on a single occasion. In parallel, an additional group of 52 patients, who had undergone CPB surgery at our institution between January 1, 2022, and January 1, 2023, were enrolled as a disease control cohort. This subgroup consisted of 28 patients who underwent isolated valve replacement surgery and 24 patients who underwent isolated coronary artery bypass grafting surgery. Blood samples were obtained from these patients both preoperatively and immediately postoperative.

Model Construction Cohort: the model construction cohort, consisting of 222 eligible patients, was enrolled between March 11, 2022, and March 31, 2023. Blood samples were meticulously collected at 0 h post-surgery. The comprehensive depiction of case inclusions for this cohort is depicted in Fig. 1c.

Statistical analyses were meticulously conducted utilizing SPSS 26.0 (IBM, Chicago, IL, USA). Continuous variables underwent comparison through the Student's t-test or the Mann–Whitney U test, with representation as mean ± standard deviation for normally distributed data, and as median along with interquartile range for non-normally distributed data. The chi-square test or Fisher’s exact test was employed for the comparison of categorical data, presented as counts and percentages. Spearman's correlation analysis was executed to explore associations between serum levels of S100A8/A9, PTX3, and CHI3L1 at 0 h post-surgery and diverse variables. The diagnostic efficiency of the biomarkers was assessed through the construction of ROC curves and the calculation of the corresponding area under the curve (AUC) values.

Sample size determination was carried out, indicating a need for approximately 194 patients to achieve 95% statistical power in detecting a clinically meaningful ASA-AKI rate of 45.0% at a two-sided α of 0.05. Proteomics were plotted and analyzed using the R program (version 4.1.3), with Volcano plots generated utilizing the ggplot2 package. Statistically significant differences were attributed to a protein fold change exceeding 2 or falling below 0.5 (p < 0.05). The predictive model was constructed and visualized employing the “rms” package under the R language platform. In summary, the patient cohort was randomly partitioned into training and validation groups at a 7:3 ratio. The associations between each preoperative variable and the occurrence of ASA-AKI were evaluated by Spearman’s correlation analysis. Univariate analysis and multivariate analysis were performed to identify risk factors for ASA-AKI. The selection of the final subset of variables was carried out through the forward, backward, or both elimination model selection procedures, utilizing the Wald test as a basis for assessment. To determine the most appropriate model fit, metrics such as deviance, Akaike information criterion (AIC), Bayesian information criterion (BIC), and pseudo-R² were employed, indicating the robustness and suitability of the model statistics. In establishing a risk scoring model, a range of identified factors was considered. These factors included demographic characteristics, comorbidities, preoperative laboratory values, preoperative radiologic characteristics, operative data, and postoperative variables. Their comprehensive inclusion allowed for a holistic approach, capturing various aspects that could contribute to the assessment and prediction of risk in the given context.

Plasma samples collected before and immediately after completion of ATAAD surgery were subjected to proteomic analysis. As indicated in Fig. 1a, 53 patients underwent ATAAD surgical repair between June 1, 2021, and October 31, 2021. Following the exclusion of four patients (1 due to renal malperfusion, 2 with end-stage renal failure, and 1 deceased within 48 h after surgery), 49 patients were included in the subsequent analysis. Among these, 24 patients (49.0%) developed AKI post-surgery, comprising 8 patients in KDIGO stage 1, 9 patients in KDIGO stage 2, and 7 patients in KDIGO stage 3. After carefully matching for age, gender, and BMI, samples from 15 patients who developed ASA-AKI and 15 patients who did not develop ASA-AKI were sent for further proteomic analysis. Characteristics of these 30 patients were outlined in Table 1, revealing no significant differences in preoperative variables between the groups. However, notable differences were observed in intraoperative parameters, with prolonged CPB duration (206.3 ± 49.2 min vs. 166.3 ± 35.5 min; p = 0.016) and cross-clamp duration (155.8 ± 47.3 min vs. 125.1 ± 33.0 min; p = 0.048) in the ASA-AKI group.

The list of proteins exhibiting differential expression was visualized in Fig. 2a (p < 0.05). A Venn diagram highlighted 77 proteins displaying postoperative differential expression between ASA-AKI and non-ASA-AKI patients. Additionally, 83 differentially expressed proteins were identified in ASA-AKI patients compared to preoperative measurements. Among these, 28 proteins were up-regulated and 4 were down-regulated postoperatively in ASA-AKI patients compared to both the non-ASA-AKI and preoperative groups. Figure 2b–c illustrate that S100A9 exhibited the highest fold change (3.13-fold) postoperatively in ASA-AKI patients compared to non-ASA-AKI patients, followed by S100A8 (2.92-fold), CHI3L1 (2.66-fold), and PTX3 (2.62-fold). These four proteins also displayed up-regulation postoperatively compared to preoperative measurements in ASA-AKI patients (Fig. 2d). Notably, S100A8 and S100A9 primarily exist as biologically functional S100A8/A9 heterodimers [19], rendering them subjects of further analysis.

To validate proteomics findings, S100A8/A9, PTX3, and CHI3L1 levels were measured using ELISA in the validation cohort, and compared with healthy and disease-control cohorts. Figure 1b outlines the enrollment of 76 ATAAD patients, 37 age-, gender-, and BMI-matched healthy individuals, and 52 age-, gender-, and BMI-matched patients who underwent CPB surgery for comparison. Among ATAAD patients, a collective total of 37 individuals, representing 48.7% of the cases, experienced postoperative AKI. This group comprised 13 patients classified under KDIGO stage 1, 12 patients under KDIGO stage 2, and 12 patients under KDIGO stage 3. Clinical and laboratory characteristics of enrolled patients are presented in Additional file 1: Table S1 and S2. The data revealed prolonged CPB duration and cross-clamp duration in the ASA-AKI group compared to the non-ASA-AKI group within the ATAAD cohort (Additional file 1: Table S1) and between the ATAAD and disease control groups (Additional file 1: Table S2). As illustrated in Fig. 3a–c, preoperative plasma levels of S100A8/A9, PTX3, and CHI3L1 were significantly elevated in ATAAD patients compared to preoperative levels of patients undergoing CPB surgery, as well as healthy controls. Postoperative levels of S100A8/A9, PTX3, and CHI3L1 were markedly higher in ATAAD patients compared to the disease control group.

Dynamic changes of these biomarkers in ASA-AKI patients were depicted in Fig. 4a–c. S100A8/A9 and CHI3L1 levels peaked at 0 h after surgery, whereas PTX3 levels peaked at 6 h post-surgery, followed by a subsequent decline in ASA-AKI patients. In contrast, patients without ASA-AKI experienced slight postoperative increases in these proteins compared to preoperative levels. Importantly, plasma levels of these biomarkers at various time points post-surgery were significantly higher in the ASA-AKI group compared to the non-ASA-AKI group. Evaluation of diagnostic value indicated that AUCs of S100A8/A9 at 0 h, 6 h, 12 h, and 24 h post-surgery were 0.823, 0.781, 0.744, and 0.774, respectively (Fig. 4d). The AUCs for PTX3 at corresponding time points were 0.786, 0.774, 0.745, and 0.742 (Fig. 4e), while those for CHI3L1 were 0.803, 0.750, 0.715, and 0.741 (Fig. 4f). The urinary levels of NGAL exhibited a gradual increase post-surgery, with a notable difference observed in the early postoperative period between ASA-AKI patients and non-ASA-AKI patients (Additional file 1: Fig. S1a). The diagnostic assessment revealed that the AUCs of urinary NGAL at 0 h and 24 h post-surgery were 0.633 and 0.605, respectively (Additional file 1: Fig. S1b). These findings suggest the potential utility of these plasma proteins as diagnostic markers for ASA-AKI.

Correlations between plasma levels of S100A8/A9, PTX3, CHI3L1, and their urinary expressions were explored. Positive correlations were observed between plasma levels of S100A8/A9 (r = 0.634, p < 0.001; Additional file 1: Fig. S2a) and PTX3 (r = 0.380, p < 0.001; Additional file 1: Fig. S2b) with their corresponding urinary levels at 0 h post-surgery, whereas no significant correlation was noted between CHI3L1 plasma level and its urinary counterpart (r = 0.163, p = 0.160; Additional file 1: Fig. S2c). Urinary levels of these proteins peaked at 24 h post-surgery and showed a significant difference in the ASA-AKI group compared to the non-ASA-AKI group at 24 h and 48 h post-surgery (Additional file 1: Fig. S2d–f). Urinary levels of S100A8/A9, PTX3, and CHI3L1 at 24 h post-surgery exhibited superior predictive value (AUCs: 0.770, 0.742, and 0.624, respectively) compared to other time points (Additional file 1: Fig. S2g–i). Notably, their predictive value in diagnosing ASA-AKI appeared less optimal than their plasma levels.

Associations were investigated between plasma levels of S100A8/A9, PTX3, CHI3L1, and urinary NGAL levels at 0 h post-surgery, as well as Cleveland Clinic scores. As listed in Additional file 1: Fig. S3a and Fig. S3d, plasma levels of S100A8/A9 correlated positively with urinary NGAL levels (r = 0.286, p = 0.012) and Cleveland Clinic scores (r = 0.394, p < 0.001). Similarly, plasma levels of CHI3L1 exhibited positive correlations with urinary NGAL levels (r = 0.325, p = 0.004; Additional file 1: Fig. S3c) and Cleveland Clinic scores (r = 0.245, p = 0.033; Additional file 1: Fig. S3f). Conversely, plasma levels of PTX3 correlated solely with urinary NGAL levels (r = 0.311, p = 0.006; Additional file 1: Fig. S3b), lacking correlation with Cleveland Clinic scores (r = 0.016, p = 0.891; Additional file 1: Fig. S3e). ROC analyses indicated better predictive values of S100A8/A9, PTX3, and CHI3L1 plasma levels at 0 h post-surgery compared to urinary NGAL levels and Cleveland Clinic scores (Additional file 1: Fig. S3g).

The model construction cohort encompassed 222 patients (Fig. 1c). Detailed demographic and clinical characteristics were outlined in Table 2. A notable 45.9% of patients developed postoperative AKI, with 8.1% requiring renal replacement therapy. A total of 37 patients were categorized in KDIGO stage 1, 35 patients in KDIGO stage 2, and 30 patients in KDIGO stage 3. Hospital mortality reached 8.6%, climbing to 15.7% in patients with postoperative AKI. The cohort was randomly divided into training (155 patients) and nomogram validation (67 patients) cohorts. No significant differences were observed in perioperative variables between the two groups, except for the prevalence of diabetes mellitus (Additional file 1: Table S3).

The univariate logistic regression analysis identified potential risk factors for ASA-AKI, incorporating various factors such as age, BMI, pre-admission CKD status, preoperative estimated glomerular filtration rate, preoperative cystatin C, preoperative white blood cell count (WBC), preoperative sCr, preoperative D-dimer, duration of CPB exceeding 180 min, postoperative drainage volume 24 h after surgery, Cleveland Clinic score, and biomarkers levels (0 h post-surgery), encompassing NGAL, S100A8/A9, PTX3, and CHI3L1. Subsequently, the forward elimination model revealed specific influential factors, including age (odds ratio [OR] = 1.07, 95% confidence interval [CI]: 1.02–1.13, p = 0.005), WBC (OR = 1.31, 95% CI: 1.11–1.56, p = 0.002), CPB duration > 180 min (OR = 3.10, 95% CI: 1.13–8.96, p = 0.031), Cleveland Clinic score (OR = 1.26, 95% CI: 1.02–1.58, p = 0.037), NGAL (OR = 1.02, 95% CI: 1.01–1.03, p < 0.001), S100A8/A9 (OR = 1.33, 95% CI: 1.18–1.54, p < 0.001), PTX3 (OR = 1.25, 95% CI: 1.13–1.41, p < 0.001), and CHI3L1 (OR = 1.01, 95% CI: 1.00–1.01, p < 0.001). These factors displayed the minimal deviance, AIC, BIC, and the maximum pseudo-R², signifying the best goodness of fit (Table 3 and Additional file 1: Table S4). The coefficients identified were used to construct a nomogram for assessing the risk of ASA-AKI (Fig. 5), offering a practical tool for evaluating risk based on these influential factors.

In the model construction cohort, AUCs for S100A8/A9, PTX3, and CHI3L1 at 0 h after surgery were 0.821, 0.782, and 0.798, respectively (Fig. 6a), consistent with validation cohort results. When comparing various combinations of the clinical model with urinary NGAL, the Cleveland Clinic score, or novel biomarkers (S100A8/A9, PTX3, and CHI3L1), we noted that the highest AUC (0.963 [95% CI: 0.941–0.986]) was achieved with the integration of the clinical model, NGAL, Cleveland Clinic score, and novel biomarkers (Fig. 6b). Furthermore, the predominant influence on this effect was attributed to the scores derived from the novel biomarkers. The corresponding AUC in the precision-recall curve was also 0.963 (Additional file 1: Fig. S4). Calibration analysis demonstrated congruence between predicted and actual ASA-AKI occurrences (Additional file 1: Fig. S5). Decision curve analysis further validated the nomogram’s net benefit (Additional file 1: Fig. S6), while the clinical impact curve displayed effective stratification of high-risk probability (Additional file 1: Fig. S7). Notably, the prognostic model's net benefit surpassed the treat-all-patients and treat-none strategies.