Association between intrarenal venous flow and postoperative acute kidney injury in cardiac surgery patients following cardiopulmonary bypass: a prospective cohort study

Abstract

Objective

Postoperative acute kidney injury (AKI) after cardiac surgery is a major complication that worsens patient outcomes and increases healthcare costs. This study aimed to investigate the association between intrarenal venous flow (IRVF) and the risk of postoperative AKI, duration of mechanical ventilation, and intensive care unit (ICU) length of stay.

Methods

This prospective observational cohort study enrolled patients undergoing cardiopulmonary bypass (CPB) cardiac surgery at a provincial tertiary hospital in Zhejiang Province from 1 April 2022 to 31 March 2023. Clinical, laboratory and ultrasonographic data were collected. The primary outcome was AKI within 7 days after surgery; secondary outcomes were duration of mechanical ventilation and ICU length of stay. Covariates were selected using LASSO and Boruta algorithms. Multivariable logistic regression assessed the association between IRVF grade and AKI. Receiver operating characteristic (ROC) and calibration curves evaluated predictive performance, and sensitivity analyses tested the robustness of findings. Violin plots and bar charts compared ventilation duration and ICU stay across IRVF grades.

Results

Among the 240 patients included, 59 (24.6%) developed AKI after surgery. In multivariable analysis, IRVF grade was associated with postoperative AKI. After full adjustment, patients with IRVF grade 3 had 6.35 times higher odds of AKI than those with grade 1 (OR 6.35, 95% CI 2.37–17.02; P < 0.001). Sensitivity analyses confirmed the robustness of this association. The IRVF-based model showed modest discrimination (AUC 0.688, 95% CI 0.614–0.761) and acceptable calibration. AKI predominantly occurred in patients with IRVF grades 2 and 3. Patients with IRVF grade 3 had a significantly longer mean duration of mechanical ventilation. Mean ICU length of stay was longer for grades 2 and 3, although only the difference between grades 1 and 2 reached statistical significance.

Conclusion

IRVF is associated with AKI following CPB cardiac surgery and correlates with prolonged mechanical ventilation and extended ICU length of stay. Bedside assessment of IRVF has the potential to serve as a useful tool for early risk stratification of postoperative AKI.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12872-026-06040-z.

Introduction

Postoperative AKI after cardiac surgery is common and portends worse clinical outcomes, prolonged hospitalisation and increased healthcare costs [1]. Reported incidence approaches 30%, and AKI is linked to persistent renal impairment and higher long-term mortality [2]. Although multiple risk factors have been identified, clinical management remains suboptimal, particularly with respect to early recognition of high-risk patients and timely intervention [3]. Accordingly, improving understanding of AKI pathophysiology and developing reliable methods for early risk stratification are urgent research priorities.

In practice, AKI diagnosis and monitoring depend largely on rises in serum creatinine (Scr) and reductions in urine output, both of which exhibit substantial latency. Serum creatinine typically increases hours to days after renal parenchymal injury, thereby often reflecting established, potentially irreversible damage; it is also influenced by non-renal factors and lacks specificity [4]. Novel molecular markers such as neutrophil gelatinase-associated lipocalin (NGAL) and Tissue Inhibitor of Metalloproteinases-2 (TIMP-2) have been proposed for earlier detection, but their predictive performance after cardiac surgery is inconsistent, limiting widespread adoption [5, 6]. Biomarkers of inflammation or endothelial injury (for example IL-6 or syndecan-1) are promising in concept but are constrained by assay cost and complexity, reducing their suitability for rapid bedside decision making [7]. These limitations justify exploration of complementary approaches, including haemodynamic imaging modalities that more directly reflect renal pathophysiology.

Venous congestion has emerged as a key mechanism in the complex pathogenesis of post‑cardiac surgery AKI. Assessment of IRVF by Doppler ultrasound permits real-time visualisation of renal venous return and converts venous pressure phenomena into quantifiable flow waveforms [8]. In heart failure cohorts, altered IRVF correlates with renal dysfunction and may outperform conventional congestion metrics [9], and in paediatric congenital heart disease abnormal intrarenal venous Doppler patterns predict postoperative AKI [10]. However, among adults undergoing cardiac surgery with CPB, the independent relationship between distinct IRVF patterns (for example continuous, intermittent and monophasic flow) and early postoperative AKI—as well as the association between IRVF and short-term postoperative outcomes—remains inadequately defined. This study therefore aimed to determine whether early postoperative IRVF grading after CPB cardiac surgery is associated with AKI within 7 days and to evaluate the predictive performance of IRVF for postoperative AKI.

Methods

Study design

This single-centre prospective observational cohort study was conducted at a provincial tertiary general hospital in Zhejiang Province, China. The protocol complied with the Declaration of Helsinki and was approved by the Ethics Committee of the First Affiliated Hospital of Zhejiang University School of Medicine (No. 2022IIT905), which granted a waiver of informed consent. The study involved no alterations to routine clinical care and posed no additional risk to participants; consequently, no patient consent forms were signed.

Sample size calculation

Sample size was estimated using the conventional rule of thumb of ≥ 10 events per variable (EPV). Assuming a post-CPB AKI incidence of 20–25% at our centre and a multivariable model including 5–10 covariates, the required sample size was projected at approximately 200–500 patients [11].

Study participants

Consecutive patients admitted to the surgical intensive care unit (SICU) after cardiac surgery between 1 April 2022 and 31 March 2023 were screened. Inclusion criteria were: age ≥ 18 years; CPB surgery; ICU stay ≥ 24 h; continuous monitoring of central venous pressure (CVP) and invasive arterial blood pressure during the ICU stay; invasive mechanical ventilation; and completion of renal ultrasonography with IRVF assessment on postoperative day 1. Exclusion criteria included end-stage renal disease on maintenance haemodialysis or preoperative chronic kidney disease; solitary kidney, prior nephrectomy or renal transplantation; renal vein thrombosis; major congenital or structural renal abnormalities or recent major renal surgery; unreliable renal Doppler signals or inadequate acoustic windows precluding accurate IRVF assessment; preoperative mechanical ventilation or hemodynamic instability requiring vasoactive support; preoperative AKI or delirium [12]; ureteral obstruction; intra-abdominal tumours involving the kidney; pregnancy; and major postoperative haemorrhage necessitating reoperation or pericardiocentesis for tamponade.

Renal and cardiac ultrasonography

IRVF was assessed by renal Doppler ultrasound at the bedside within three hours of SICU admission. Scans were performed by an experienced operator with a second experienced operator present to supervise image quality and jointly confirm grading; both operators had completed supervised departmental competency training. Operators were blinded to clinical data. Image acquisition and grading followed a standardised protocol with uniform Doppler settings and predefined parameters (probe position, sample-volume placement, velocity scale, insonation angle and respiratory-phase control) [9]. Discrepancies were resolved by consensus or adjudicated by a senior physician with > 10 years’ experience.

A phased-array transducer (2.5–5 MHz) was used. The inferior vena cava diameter was measured 2 cm distal to the hepatic vein confluence. The right kidney was imaged by translating the probe caudally and posteriorly [13]. Patients were scanned supine with the head-of-bed elevated 0–30^°. ECG gating was used to identify cardiac cycle phases [12, 14]). During end-expiratory apnea, pulsed-wave Doppler waveforms were recorded at the corticomedullary junction of the renal segments for 2–3 consecutive cardiac cycles; the Doppler velocity scale was set at ~ 16 cm/s. Interlobar arterial and venous traces were recorded concurrently. IRVF patterns were classified according to Iida et al.: pattern 1—continuous venous flow throughout the cardiac cycle (or transient interruption during atrial contraction); pattern 2—discontinuous biphasic flow with systolic and diastolic components; pattern 3—discontinuous flow detectable only in diastole. When multiple patterns coexisted, the pattern with the greatest flow velocity and clearest spectral definition was recorded [9]. These corresponded to IRVF grades 1–3 [8]. Bedside echocardiography was performed by a specialised cardiac sonographer in the ICU and left ventricular ejection fraction (LVEF) was measured using Simpson’s method in the apical four‑chamber view.

Central Venous Pressure (CVP) and invasive Mean Arterial Pressure (MAP)

CVP was measured at end‑expiration with the patient supine; the transducer was zeroed at the mid-axillary line (approximately the fifth intercostal space) and values were averaged over three cardiac cycles. Invasive arterial pressure was obtained via a radial arterial catheter; transducers were zeroed and calibrated per standard practice, and systolic, diastolic and mean arterial pressures were recorded from the monitor.

Variable collection

In addition to IRVF pattern, CVP, MAP and inferior vena cava (IVC) diameter, we systematically recorded: Demographics: age, sex, body mass index (BMI), New York Heart Association (NYHA) class, EuroSCORE II, major comorbidities (e.g. hypertension, diabetes). Personal history: smoking status, alcohol consumption. Intraoperative variables: CPB time; intraoperative transfusion volume (Autoblood, red blood cells, plasma); intraoperative fluid balance (transfusions + crystalloids + colloids + albumin − estimated blood loss − urine output). Postoperative variables: percentage decrease in haemoglobin on SICU admission; the cardiovascular system score from SOFA (Table S1); left ventricular ejection fraction(LVEF); PaO₂/FiO₂; duration of mechanical ventilation; ICU length of stay; occurrence of AKI; and use of continuous renal replacement therapy (CRRT), among others.

Outcome and AKI definition

The primary outcome was AKI within 7 days of CPB surgery. AKI was defined by KDIGO criteria: increase in Scr by ≥ 0.3 mg/dL (26.5 µmol/L) within 48 h, or ≥ 1.5× baseline within 7 days, or urine output < 0.5 mL/kg/h for 6 h. Baseline Scr was the most recent value within 7 days before surgery. Secondary outcomes were duration of mechanical ventilation and ICU length of stay during the index hospitalisation.

Statistical analysis

Continuous variables with approximate normal distributions are presented as mean ± standard deviation and compared by one-way ANOVA. Non-normally distributed continuous variables are presented as median (interquartile range) and compared by the Kruskal–Wallis test. Categorical variables are reported as counts (percentages) and compared by Chi-square test or Fisher’s exact test as appropriate. Where the overall test was significant, post-hoc pairwise comparisons were performed with Bonferroni correction.

Potential prognostic variables were screened using LASSO regression in parallel with the Boruta algorithm; candidate sets were compared by the AUC of their respective models to identify the optimal variable combination. Four covariates were selected for adjustment: BMI, CPB time, EuroSCORE II and MAP. Multicollinearity in multivariable models was assessed using the generalized variance inflation factor (GVIF); GVIF, degrees of freedom and adjusted GVIF were reported. Associations between IRVF grade and outcomes were evaluated using multivariable logistic regression with three hierarchical models: Model I (crude); Model II (adjusted for BMI and EuroSCORE II); and Model III (fully adjusted for BMI, CPB time, EuroSCORE II and MAP). A test for trend across IRVF grades was performed. Results are presented as odds ratios with 95% confidence intervals and two-sided P values.

Model discrimination was assessed by ROC curves and AUC; calibration by calibration plots comparing predicted and observed probabilities. Sensitivity analyses included additional adjustment for intraoperative transfusion volume, intraoperative fluid balance, postoperative CVP and other haemodynamic indices. Prespecified subgroup variables were used for interaction testing and stratified analyses. E-values were calculated to assess the potential impact of unmeasured confounding. Differences in duration of mechanical ventilation and ICU length of stay across IRVF grades were compared and visualised using violin plots and bar charts. Data were independently collected and quality-controlled by three investigators. Missingness was minimal (< 5%) for intraoperative blood loss and urine output; missing values were addressed by multiple imputation. All analyses were performed using R Statistical Software (Version 4.2.2) and the Free Statistics analysis platform (version 2.2, Beijing, China, http://www.clinicalscientits.cn/freestatistics). For all analyses, a two-sided p-value < 0.05 was deemed statistically significant.

Results

A total of 240 patients were included; 59 (24.6%) developed postoperative AKI (Fig. 1). The cohort was 58.3% male, with mean age 57.6 ± 12.6 years and mean BMI 23.4 ± 3.2 kg/m². Baseline characteristics by IRVF grade showed that higher IRVF grades were associated with older age, higher CVP, larger IVC diameter and higher EuroSCORE II, and with lower BMI, LVEF, MAP and PaO₂/ FiO₂; rates of AKI and use of CRRT increased with IRVF grade (Table 1). There were no significant differences between groups in the cardiac surgery type, volumes of crystalloids, colloids, autoblood, albumin, urine output, or bleeding (P > 0.05). Patients with higher IRVF grades received greater intraoperative plasma, blood cell and total transfusion volumes (P < 0.05) (Table S2). Compared with patients without AKI, those who developed AKI were more frequently male and had higher age, BMI, CVP, IVC diameter, IRVF grade and EuroSCORE II, and experienced longer CPB time, mechanical ventilation and ICU stay; they also had lower LVEF, MAP and PaO₂/ FiO₂ (Table S3).

Fig. 1.

Flowchart of the inclusion and exclusion process of the study population

Table 1.

The baseline data and clinical characteristics of patients in different IRVF groups.

Variables	Total (n = 240)	IRVF			P value
		1 (n = 112)	2 (n = 88)	3 (n = 40)
Gender, n (%)					0.711
Male	140 (58.3)	67 (59.8)	52 (59.1)	21 (52.5)
Female	100 (41.7)	45 (40.2)	36 (40.9)	19 (47.5)
Age, year	57.6 ± 12.6	56.6 ± 13.2	57.9 ± 12.5	60.1 ± 10.8	0.307
BMI, kg/m²	23.4 ± 3.2	23.7 ± 3.0	23.5 ± 3.5	22.3 ± 2.8	0.048
NYHA, n (%)					0.656
1	47 (19.6)	24 (21.4)	18 (20.5)	5 (12.5)
2	94 (39.2)	39 (34.8)	39 (44.3)	16 (40)
3	81 (33.8)	40 (35.7)	26 (29.5)	15 (37.5)
4	18 ( 7.5)	9 (8)	5 (5.7)	4 (10)
LVEF, %	60.8 ± 10.0	62.2 ± 8.6	61.1 ± 10.2	56.5 ± 12.1	0.008
CVP, mmHg	11.1 ± 2.8	10.4 ± 2.3	11.5 ± 3.0	12.2 ± 3.2	< 0.001
IVC, mm	18.1 ± 4.3	17.1 ± 4.6	18.4 ± 3.8	19.9 ± 3.7	< 0.001
CPB time, minute	129.2 ± 50.1	118.8 ± 45.6	134.4 ± 50.8	146.8 ± 55.2	0.005
EuroScore II	1.7 (1.0, 3.2)	1.4 (0.8, 2.5)	1.8 (1.0, 3.2)	2.6 (1.4, 4.4)	< 0.001
Hb.decreasedpercent, %	0.4 ± 0.2	0.4 ± 0.2	0.4 ± 0.2	0.4 ± 0.1	0.04
MAP, mmHg	75.6 ± 9.2	77.1 ± 8.7	74.4 ± 9.4	74.0 ± 9.6	0.055
PaO₂/FiO₂, mmHg	285.5 ± 80.6	290.5 ± 76.6	283.9 ± 89.1	275.3 ± 72.4	0.579
Hypertension, n (%)					0.903
No	173 (72.1)	80 (71.4)	63 (71.6)	30 (75)
Yes	67 (27.9)	32 (28.6)	25 (28.4)	10 (25)
Diabetes, n (%)					0.079
No	216 (90.0)	105 (93.8)	74 (84.1)	37 (92.5)
Yes	24 (10.0)	7 (6.2)	14 (15.9)	3 (7.5)
Smoke, n (%)					0.835
No	198 (82.5)	94 (83.9)	71 (80.7)	33 (82.5)
Yes	42 (17.5)	18 (16.1)	17 (19.3)	7 (17.5)
Alcohol, n (%)					0.382
No	219 (91.2)	101 (90.2)	83 (94.3)	35 (87.5)
Yes	21 ( 8.8)	11 (9.8)	5 (5.7)	5 (12.5)
AKI, n (%)					< 0.001
No	181 (75.4)	98 (87.5)	63 (71.6)	20 (50)
Yes	59 (24.6)	14 (12.5)	25 (28.4)	20 (50)
AKI stage, n (%)					0.673
1	45 (76.3)	12 (85.7)	20 (80)	13 (65)
2	3 ( 5.1)	0 (0)	1 (4)	2 (10)
3	11 (18.6)	2 (14.3)	4 (16)	5 (25)
CRRT, n (%)					0.025
No	229 (95.4)	110 (98.2)	84 (95.5)	35 (87.5)
Yes	11 ( 4.6)	2 (1.8)	4 (4.5)	5 (12.5)
Ventilation time, day	1.0 (1.0, 2.0)	1.0 (1.0, 1.0)	1.0 (1.0, 2.0)	1.0 (1.0, 3.2)	< 0.001
ICU stay, day	3.0 (2.0, 4.0)	3.0 (2.0, 4.0)	3.0 (2.0, 5.0)	3.0 (2.0, 5.2)	0.126

BMI Body Mass Index, LVEF Left Ventricular Ejection Fraction, NHYA New York Heart Association, CVP Central venous pressure, IVC Inferior vena cava diameter, IRVF Intra-renal venous flow pattern, CPB Cardiopulmonary bypass, EuroScore II European System for Cardiac Operative Risk Evaluation II, Hb Hemoglobin, MAP Mean Arterial Pressure, PaO₂/FiO₂ Oxygenation index, CRRT Continuous Renal Replacement Therapy, AKI Acute Kidney Injury, ICU Intensive Care Unit

Variable selection by LASSO identified five predictors (IRVF, BMI, CPB time, EuroSCORE II, MAP) (Fig. 2A, B), while the Boruta algorithm identified six (CVP plus the five above) (Fig. 2C). ROC comparison of the two candidate models produced AUCs of 0.845 (95% CI 0.785–0.905) for the five‑variable model and 0.849 (95% CI 0.789–0.909) for the six‑variable model (P = 0.555) (Fig. 2D). GVIF values were < 2 for all predictors, indicating no problematic multicollinearity (Table S4). IRVF was defined as the exposure and four covariates (BMI, CPB time, EuroSCORE II, MAP) were chosen for adjustment.

Fig. 2.

Variable selection using LASSO and Boruta algorithms. A LASSO coefficient profiles of all candidate variables, with coefficients plotted against log(λ). B Tenfold cross-validation for the optimal λ selection, based on binomial deviance. C Boruta algorithm-based variable importance plot, ranking predictors by importance (blue: confirmed rejection, red: tentative, green: confirmed acceptance). D ROC curves comparing two models after variable selection

In multivariable logistic regression with stepwise adjustment, IRVF grade was associated with AKI risk. In the crude model, IRVF grade 2 (OR 2.78; 95% CI 1.34–5.75; P = 0.006) and grade 3 (OR 7.00; 95% CI 3.04–16.14; P < 0.001) were significantly associated with AKI, with a significant linear trend across grades (P for trend < 0.001). The association persisted after adjustment for BMI and EuroSCORE II. In the fully adjusted model (BMI, CPB time, EuroSCORE II, MAP), IRVF grade 3 remained an independent predictor of AKI (OR 6.35; 95% CI 2.37–17.02; P < 0.001), whereas the effect of grade 2 was attenuated and no longer statistically significant (OR 1.95; 95% CI 0.84–4.50; P = 0.118). A dose-response relationship across IRVF grades was retained in all models (P for trend < 0.001). Each one-grade increase in IRVF was associated with a 2.47-fold increase in AKI odds in the fully adjusted model (OR 2.47; 95% CI 1.51–4.04; P < 0.001) (Table 2).

Table 2.

Multivariable logistic regression model to evaluate the relationship between IRVF and the AKI

Variable	Model I		Model II		Model III
	OR (95%CI)	P value	OR (95%CI)	P value	OR (95%CI)	P value
IRVF continuous variable	2.65 (1.75 ~ 4.02)	< 0.001	2.85 (1.78 ~ 4.57)	< 0.001	2.47 (1.51 ~ 4.04)	< 0.001
IRVF grade
1	1(Ref)		1(Ref)		1(Ref)
2	2.78 (1.34 ~ 5.75)	0.006	2.59 (1.18 ~ 5.68)	0.018	1.95 (0.84 ~ 4.5)	0.118
3	7.00 (3.04 ~ 16.14)	< 0.001	8.3 (3.23 ~ 21.32)	< 0.001	6.35 (2.37 ~ 17.02)	< 0.001
Trend.test		< 0.001		< 0.001		< 0.001

Model I: Crude model

Model II: Adjusted for BMI, EuroScore II

Model III: Adjusted for BMI, CPB time, EuroScore II, MAP

The predictive performance of IRVF alone resulted in an AUC of 0.688 (95% CI 0.614–0.761) (Fig. 3A), and calibration plots demonstrated a match between the predicted and observed AKI incidence (Fig. 3B). The full model incorporating IRVF (BMI, CPB time, EuroSCORE II, MAP) demonstrated strong discrimination (AUC 0.845; 95% CI 0.785–0.905) (P < 0.001) (Fig. 4A). By comparison, EuroSCORE II alone produced an AUC of 0.679 (95% CI 0.598–0.760) (P = 0.851) (Fig. 4B).

Fig. 3.

Predictive performance and calibration of IRVF. A ROC curve for IRVF. B Calibration plot for IRVF, showing the ideal line, apparent calibration, and bias-corrected calibration

Fig. 4.

Comparison of ROC curves for Model 1 and Model 2. A Model 1 represents the base model (IRVF), and Model 2 represents the full model (IRVF, BMI, CPB time, EuroScore II, MAP). B Model 1 represents IRVF, and Model 2 represents EuroScore II. The areas under the curves (AUCs) with 95% CIs are shown, along with the P value for the comparison of the two curves

Sensitivity analyses that additionally adjusted for IVC diameter, cardiac surgery type, intraoperative fluid balance and transfusion volume, and postoperative circulatory status produced results consistent with the primary model (Table 3). No multicollinearity was detected in these models (Table S5).Prespecified subgroup analyses (sex; age < 60 vs. ≥ 60 years; BMI < 24 vs. ≥ 24 kg/m2; CPB duration < 120 vs. ≥ 120 min; hypertension history) showed that IRVF grade 3 was associated with increased AKI risk relative to grade 1 across all strata; interaction tests were nonsignificant (all P for interaction > 0.05), indicating consistency of the IRVF–AKI association (Figure S1). The E-value for the primary association was 4.48, suggesting robustness to unmeasured confounding (Figure S2).

Table 3.

Sensitivity analysis of the association between IRVF and postoperative AKI

Variable	Model I		Model II		Model III
	OR (95%CI)	P value	OR (95%CI)	P value	OR (95%CI)	P value
IRVF continuous variable	2.47 (1.51 ~ 4.04)	< 0.001	2.45 (1.47 ~ 4.07)	0.001	2.57 (1.52 ~ 4.36)	< 0.001
IRVF grade
1	1(Ref)		1(Ref)		1(Ref)
2	1.95 (0.84 ~ 4.5)	0.118	1.94 (0.83 ~ 4.51)	0.126	1.81 (0.74 ~ 4.44)	0.195
3	6.35 (2.37 ~ 17.02)	< 0.001	6.27 (2.27 ~ 17.32)	< 0.001	7.10 (2.47 ~ 20.38)	< 0.001
Trend.test		< 0.001		0.001		< 0.001

Model I: Adjusted for BMI, CPB time, EuroScore II, MAP

Model II: Adjusted for BMI, CPB time, EuroScore II, MAP, IVC

Model III: Adjusted for BMI, CPB time, EuroScore II, MAP, Intraoperative transfusion volume, Intraoperative fluid balance, Cardiac surgery type, Circulatory status

AKI patients were concentrated in IRVF grades 2–3, whereas non-AKI cases predominated in grades 1–2 (Fig. 5A). Mean durations of mechanical ventilation were 1.4, 2.3 and 2.7 days for IRVF grades 1, 2 and 3, respectively; the difference between grades 1 and 3 was statistically significant (P = 0.009) (Fig. 5B). Mean ICU length of stay was 3.3, 4.5 and 4.8 days for grades 1, 2 and 3, respectively, with a significant difference between grades 1 and 2 (P = 0.014) (Fig. 5C).

Fig. 5.

The relationship between IRVF grades and AKI, mechanical ventilation, and ICU hospitalization time. A The distribution of IRVF grades in patients with and without AKI. B The duration of mechanical ventilation in patients with different IRVF grades. C The length of ICU stay in patients with different IRVF grades

Discussion

In this prospective cohort study, early bedside Doppler assessment of IRVF after CPB cardiac surgery was associated with postoperative AKI. Compared with IRVF grade 1, grade 3 conferred a 6.35-fold higher adjusted odds of AKI (OR 6.35; 95% CI 2.37–17.02; P < 0.001). IRVF alone showed modest discrimination (AUC 0.688; 95% CI 0.614–0.761), while a model combining IRVF with MAP, CPB time, EuroSCORE II and BMI achieved substantially improved discrimination (AUC 0.845; 95% CI 0.785–0.905). Higher IRVF grades were also associated with longer mechanical ventilation and prolonged ICU stay, suggesting that abnormal renal venous flow may reflect broader postoperative multisystem dysfunction rather than isolated renal impairment.

Earlier attempts to forecast AKI following cardiac surgery have predominantly depended on preoperative biomarkers, overall intraoperative perfusion indices, or conventional venous congestion [15–17]. These biomarkers fail to accurately reflect immediate, organ-specific hemodynamic disruptions, and the predictive power of individual indicators is still limited. The Classic scoring system, EuroSCORE II, demonstrated only modest predictive performance for AKI in our cohort (AUC 0.679; 95% CI: 0.598–0.760), which is consistent with prior reports (AUC 0.67; 95% CI: 0.62–0.73) [18]. Point-of-care ultrasound (POCUS) enables real-time haemodynamic phenotyping [19]; in particular, the IRVF grading proposed by Iida et al. correlates with right atrial pressure and provides a direct visual marker of renal venous congestion [9]. Prior POCUS studies linking abdominal and renal venous congestion to postoperative AKI were often single‑centre, small and heterogeneous [12, 20]. Our prior work established IRVF as a useful predictor after cardiopulmonary bypass, but a graded dose-response had not been demonstrated [21]; here we show a clear graded association (P for trend < 0.001). Conventional systemic indices have limitations: CVP cannot resolve organ-specific haemodynamic heterogeneity [22] and IVC diameter, while intuitive, adds limited independent predictive information [23]. By contrast, IRVF directly captures renal venous waveform abnormalities that likely reflect retrograde transmission of elevated right-sided pressures into the renal microcirculation, raising interstitial hydrostatic pressure, compressing peritubular capillaries and impairing tubular perfusion [24, 25]; animal data indicate venous occlusion can cause microvascular rupture and interstitial haemorrhage that exacerbate ischaemia–reperfusion injury, which may explain the high AKI risk observed with IRVF grade 3 even when MAP is preserved [26].

Renal injury after cardiac surgery is driven by the balance between arterial inflow and venous outflow pressures: postoperative hypotension diminishes forward perfusion, while elevated renal venous pressure increases backward resistance, together lowering net glomerular filtration pressure [27]. Consistent with this physiology, combining MAP and IRVF improved prediction (AUC 0.845; 95% CI 0.785–0.905), supporting integrated haemodynamic phenotyping [28]. At a molecular level, venous hypertension can induce tubular hypoxia, activate HIF-1αsignalling and trigger proinflammatory mediator release, perpetuating endothelial dysfunction and microvascular leak; superimposed arterial hypoperfusion accelerates these pathways toward irreversible cell injury [29, 30]. In the CPB setting, haemodynamic instability and inflammatory activation may precipitate these events, making early IRVF abnormalities a salient marker of pathogenic processes.

Additionally, higher IRVF grades in our cohort were associated with lower LVEF, worse oxygenation, longer mechanical ventilation and extended ICU stay, suggesting renal venous congestion may herald evolving multisystem organ dysfunction [31]. The combined use of LASSO and Boruta reduced overfitting and helped isolate IRVF’s independent prognostic contribution, and reported intraoperative variability of renal venous flow by transesophageal Doppler implies that early postoperative IRVF may reflect intraoperative insults that were incompletely corrected [32]. Thus, abnormal IRVF appears to integrate multiple pathophysiological signals and may identify patients likely to require intensified organ support.

This study has several limitations. First, the single-centre design may limit generalisability to other settings. Second, although scans were performed by trained operators following a standardised protocol and images were adjudicated by consensus, IRVF measurements were not independently double‑blinded at acquisition and formal interobserver reproducibility was not quantified. Third, a single IRVF assessment within the first three postoperative hours does not characterise temporal dynamics; one time point may under or overestimate cumulative venous exposure and cannot distinguish transient from sustained congestion. Fourth, despite calculating an E-value to assess robustness, we did not include intra-abdominal pressure in the adjustment models, and detailed information regarding vasopressor types, dosing and the duration and severity of intraoperative hypotension was not fully captured; these omissions leave the possibility of residual confounding. Finally, as an observational study, causal inference is limited. IRVF may primarily reflect the severity of postoperative systemic venous congestion or underlying cardiac dysfunction, rather than being a unique renal prognostic marker. Therefore, it remains unproven whether interventions targeting abnormal intrarenal venous flow can reduce the incidence of AKI. Large‑scale prospective interventional trials are required to test this hypothesis.

Conclusion

IRVF is associated with AKI after CPB cardiac surgery and correlates with longer duration of mechanical ventilation and extended ICU length of stay. Bedside assessment of IRVF, by detecting organ-specific venous congestion, has the potential to serve as a useful tool for early risk stratification of postoperative AKI. Multicentre prospective interventional studies incorporating serial IRVF monitoring are needed to define its temporal trajectories and relationship to AKI staging, and to determine whether IRVF-guided interventions can reduce AKI incidence and improve clinical outcomes.

Abbreviations

AKI

Acute kidney injury

ANOVA

Analysis of variance

AUC

Area under the receiver operating characteristic curve

BMI

Body mass index

Boruta

Boruta feature‑selection algorithm

CI

Confidence interval

CPB

Cardiopulmonary bypass

CRRT

Continuous renal replacement therapy

CVP

Central venous pressure

Df

Degrees of freedom

ECG

Electrocardiogram

LVEF

Left ventricular ejection fraction

EuroSCORE II

European System for Cardiac Operative Risk Evaluation II

GVIF

Generalized variance inflation factor

GVIF_adj

Adjusted generalized variance inflation factor

HIF‑1α

Hypoxia‑inducible factor‑1 alpha

IQR

Interquartile range

IVC

Inferior vena cava

IRVF

Intrarenal venous flow

KDIGO

Kidney Disease: Improving Global Outcomes

LASSO

Least absolute shrinkage and selection operator

MODS

Multisystem organ dysfunction (syndrome)

NGAL

Neutrophil gelatinase‑associated lipocalin

NYHA

New York Heart Association

OR

Odds ratio

POCUS

Point‑of‑care ultrasound

ROC

Receiver operating characteristic

Scr

Serum creatinine

SD

Standard deviation

SICU

Surgical intensive care unit

SOFA

Sequential Organ Failure Assessment

TIMP‑2

Tissue inhibitor of metalloproteinases‑2

Authors’ contributions

ZJF and LZT jointly conceived and designed this study. LZT, WW, and FS were responsible for data collection and measurement of ultrasound indicators. LZT collaborated with FQH to complete data cleaning, verification, and statistical analysis. ZJF performed statistical analysis, chart preparation, and initial manuscript drafting. LZT, WW, FS, FQH, and FSQ participated in critical revisions. FSQ oversaw the research supervision, and all authors contributed to the review and finalization of the manuscript.

Funding

No funding.

Data availability

The dataset used in this study is involved in another unpublished research and is currently not available to the public. If readers have a need for it, they may contact the corresponding author.

Declarations

Ethics approval and consent to participate

The study was conducted in accordance with the principles of the Declaration of Helsinki, approved by the Ethics Committee of the First Affiliated Hospital of Zhejiang University School of Medicine (No. 2022IIT905), and was granted an exemption from informed consent. The study did not involve any changes in medication or treatment protocols, and did not cause any harm to the patients’ interests. Therefore, no patient consent forms were signed.

Competing interests

The authors declare no competing interests.