Preoperative acute kidney injury predicts the need for post-LVAD renal replacement therapy

Abstract

Background

Postoperative renal failure following LVAD implantation is known to negatively impact outcomes. However, the role of preoperative predictors of renal failure remains unclear. Therefore, we sought to identify and assess preoperative factors potentially predictive of postoperative acute kidney injury (AKI) requiring renal replacement therapy (RRT) in destination therapy LVAD patients.

Methods

A retrospective review of all primary LVAD implantations at a single, destination therapy center from 2022 to 2024 was conducted. LVAD exchanges were excluded. The primary outcome was postoperative AKI requiring RRT. Independent predictors of postoperative RRT were assessed using multivariable logistic regression modeling. The impact of postoperative RRT on survival was assessed using the Kaplan-Meier method.

Results

In total, 103 patients underwent primary LVAD implantation. The mean preoperative creatinine was 1.37 ± 0.46 mg/dL with an eGFR_Cr of 66.60 ± 35.25 mL/min/1.73m2. Twenty-one patients (20.39%) required postoperative RRT. Preoperative creatinine (OR: 4.64 [1.24–17.32], p = 0.02), proteinuria (OR: 3.52 [1.15–10.82], p = 0.03), and preoperative AKI (OR: 3.57 [1.03–12.37], p = 0.04) were all independently associated with increased risk of postoperative RRT. The need for RRT was associated with increased operative mortality (19.05% vs. 1.22%, p < 0.001) and decreased 1-year survival (42.59% vs. 87.97%, p < 0.001).

Conclusions

The need for RRT after LVAD implantation is associated with a high operative and short-term mortality. In our series of high acuity and surgically complex patients, the preoperative variables identified as the best predictors of postoperative RRT were baseline creatinine, baseline proteinuria, and preoperative AKI.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12882-025-04674-3.

Introduction

Left Ventricular Assist Device (LVAD) implantation represents a potential life-saving procedure for patients with advanced heart failure, serving as both a bridge to transplant and as destination therapy. Acute kidney injury (AKI) is a common complication following LVAD implantation, and patients who require renal replacement therapy (RRT) have reduced short- and long-term survival rates [1–4]. While LVADs improve cardiac output, reduce venous pressure, and potentially improve right ventricular function, renal trajectories vary post implantation. Our ability to predict which patients are most at risk for post-operative RRT remains limited. Heterogeneity of patient populations, peri- and postoperative management, and varying definitions of AKI across existing literature present a challenge to drawing conclusions.

Multiple studies have shown that pre-LVAD renal dysfunction increases morbidity and mortality, partially through the increased risk of needing postoperative RRT [3, 5]. Current guidelines recommend the use of serum creatinine as the primary biomarker for evaluation of LVAD candidates [6]. However, its validity as an accurate estimate of glomerular filtration rate (eGFR) and as a marker of true tubular injury in advanced heart failure patients has been questioned [7]. At the very least, a single serum biomarker is likely not sufficient to stratify risk across this complex patient population.

The objectives of this study were to identify risk factors associated with RRT after LVAD implantation in our high acuity, surgically complex population and to assess the impact of RRT on mortality in a cohort that is inherently high-risk for kidney injury. By identifying novel variables, we aim to refine preoperative risk stratification, improve patient selection, and develop new clinical strategies to mitigate renal complications.

Methods

We conducted a retrospective review of all primary LVAD implantations from 2022 to 2024 at Baylor Scott and White, The Heart Hospital, Plano. All durable LVADs were included. In the study period, the only durable LVAD available at our institution was the Heartmate 3 (Abbott) device. All LVAD exchanges were excluded. The Baylor Research Institute Institutional Review Board approved the study and informed consent was waived.

Baseline demographics, co-morbidities, laboratory values, urine analysis, Intermacs profiles, operative data and outcomes were extracted from the electronic medical record. Our primary outcome was the need for postoperative RRT of any duration, whether volume removal, clearance, or both. To assess preoperative renal function, immediate preimplantation creatinine values were used. In cases where Cystatin C data were available, this was also included in our analysis. eGFR was estimated using the 2021 CKD-EPI equation. Preoperative AKI was defined according to KDIGO criteria and occurred during the index hospitalization but prior to LVAD implantation. Likewise, postoperative AKI was also defined by KDIGO criteria. All patients were considered eligible for development of postoperative AKI, regardless of preoperative kidney function. Proteinuria was assessed by urine dipstick.

To simplify analysis, eGFR was calculated using creatinine, cystatin C, or the combination of both, and nonparametric receiver operator characteristic (ROC) curves were developed to preliminarily determine each method’s ability to predict the need for RRT. Areas under the curve (AUC) were calculated and then compared using generalized U-statistics to assess predictive value. After statistical evaluation, baseline creatinine alone had the highest univariate predictive value of the need for postoperative RRT. A multivariable logistic regression model was then constructed to determine which preoperative covariates were most predictive of post-implant RRT. To construct the model, variables were initially tested in univariate fashion. Independent covariates with significant association on preliminary univariate testing, those with biologic plausibility, and those with previous literature support were then incorporated into the model in nested stepwise fashion using the likelihood ratio test to build the most parsimonious model. Collinear variables were excluded.

Survival was assessed using Kaplan-Meier estimators. Survival comparisons were performed using the log-rank test. Continuous, normally distributed variables were compared using one-way analysis of variance (ANOVA) testing or the Student’s t-test as appropriate. Continuous, non-normally distributed variables were compared using the Wilcoxon rank-sum test. Categorical variables were compared using the Chi-squared or Fisher’s exact test as appropriate. For all statistical measures, a two-tailed p-value < 0.05 was considered statistically significant. Continuous parametrically distributed variables are presented as mean (± standard deviation), while continuous nonparametrically distributed variables are presented as median (interquartile range). Categorical variables are presented as whole numbers with percentages. Odds ratios are shown with 95% confidence intervals. Statistical analysis was performed using StataNow/BE 18.5 (StataCorp, College Station, Texas).

Results

From 2022 to 2024, 103 patients underwent primary LVAD implantation and were included in our analysis. The median age was 63.0 [54.0–70.0] years, 85 (82.52%) were male, and the majority (65, 63.11%) were White (Table 1). The cohort was of high acuity, with 46 (44.66%) patients requiring preoperative Impella support and 47 (45.63%) patients being Intermacs profile 1. The median preoperative creatinine was 1.31 [1.02–1.60] mg/dL with a mean eGFR_Cr of 58.26 [46.37–77.57] mL/min/1.73m2. Many patients (47, 45.63%) were diabetic and suffered from proteinuria (27, 26.21%). Only 41 (39.81%) patients did not have chronic kidney disease, and 55 (53.40%) patients experienced AKI during their hospitalization prior to LVAD implantation. A single patient was on RRT at the time of surgery. Baseline TAPSE was 1.6 cm [1.3–1.8]. Operative complexity was high with 23 (22.33%) patients requiring a reoperative sternotomy and 43 (41.75%) patients requiring a concomitant valvular procedure. The mean cardiopulmonary bypass time was 91.42 ± 34.90 min.

Table 1.

Baseline demographics, comorbidities, laboratory values, and operative variables

	Overall (N = 103)	RRT (N = 21)	No RRT (N = 82)	P-Value
Demographics
Age (years)	63.0 [54.0–70.0]	59.0 [48.0–70.0]	63.5 [56.0–70.0]	0.296
Male sex	85 (82.52%)	20 (95.24%)	65 (79.27%)	0.086
White race	65 (63.11%)	11 (52.38%)	54 (65.85%)	0.254
Black race	25 (24.27%)	7 (33.33%)	18 (21.95%)	0.278
Hispanic race	13 (12.62%)	3 (14.29%)	10 (12.20%)	0.797
Co-morbidities
Diabetes mellitus	47 (45.63%)	11 (52.38%)	36 (43.90%)	0.486
Hemoglobin A1c > 7	37 (35.92%)	8 (38.10%)	29 (35.37%)	0.816
Body mass index (kg/m2)	26.9 [23.5–30.4]	30.4 [26.3–34.6]	26.7 [23.3–29.8]	0.003*
Intermacs profile 1	47 (45.63%)	15 (71.43%)	32 (39.02%)	0.008*
Intermacs profile 2	25 (24.27%)	2 (9.52%)	23 (28.05%)	0.077
Intermacs profile 3	27 (26.21%)	4 (19.05%)	23 (28.05%)	0.403
Intermacs profile 4	4 (3.88%)	0 (0.0%)	4 (4.88%)	0.579
Ischemic cardiomyopathy	44 (42.72%)	8 (38.10%)	36 (43.90%)	0.631
Atrial fibrillation	51 (49.51%)	14 (66.67%)	37 (45.12%)	0.078
Preoperative Impella support	46 (44.66%)	15 (71.43%)	31 (37.80%)	0.006*
Preoperative ECMO support	1 (0.97%)	1 (4.76%)	0 (0.0%)	-
Measures of Renal Function
Creatinine (mg/dL)	1.31 [1.02–1.60]	1.50 [1.25–2.03]	1.22 [0.99–1.52]	0.002*
Cystatin C (mg/L)	1.80 [0.49–1.72]	1.96 [1.56–2.38]	1.71 [1.45–2.09]	0.214
eGFRCr (mL/min/1.73m2)	58.26 [46.37–77.57]	55.23 [34.21–63.17]	60.86 [47.99–79.72]	0.032*
eGFRCysC (mL/min/1.73m2)	35.48 [26.94–46.37]	30.90 [24.10-46.29]	36.94 [27.84–46.21]	0.391
eGFR (CKD-EPI Cystatin C-Cr) (mL/min/1.73m2)	46.0 [34.0-55.5]	37.5 [29.5–50.5]	47.0 [37.5–58.0]	-
Blood urea nitrogen (mg/dL)	31.0 [19.5–39.0]	36.0 [33.0–48.0]	27.5 [18.3–36.0]	0.002*
CKD Stage 1	16 (15.53%)	4 (19.05%)	12 (14.63%)	0.618
CKD Stage 2	34 (33.01%)	6 (28.57%)	28 (34.15%)	0.628
CKD Stage 3	19 (18.45%)	7 (33.33%)	12 (14.63%)	0.064
CKD Stage 4	2 (1.94%)	1 (4.76%)	1 (1.22%)	0.368
Preoperative AKI	55 (53.40%)	17 (80.95%)	38 (46.34%)	0.005*
Protein dip 1+	19 (18.45%)	7 (33.33%)	12 (14.63%)	0.049*
Protein dip 2+	7 (6.80%)	2 (9.52%)	5 (6.10%)	0.578
Protein dip 3+	1 (0.97%)	1 (4.76%)	0 (0.0%)	-
Urine protein to creatinine ratio (mg/mg)	0.2 [0.0-0.5]	0.3 [0.0-0.5]	0.2 [0.0-0.5]	0.548
Laboratory values
Total bilirubin (mg/dL)	0.9 [0.7–1.6]	1.1 [0.8–1.6]	0.9 [0.7–1.4]	0.159
Albumin (g/dL)	3.10 [2.6–3.3]	2.90 [2.3–3.1]	3.10 [2.6–3.3]	0.085
Operative variables
Reoperative sternotomy	23 (22.33%)	6 (28.57%)	17 (20.73%)	0.442
Concomitant operation	43 (41.7%)	12 (57.14%)	31 (37.80%)	0.109
Cardiopulmonary bypass time (minutes)	91.42 ± 34.90	106.14 ± 52.96	87.65 ± 27.75	0.030*

Data represented as n (%), mean ± SD, or median [IQR]. Abbreviations: AKI, acute kidney injury; ECMO, extracorporeal membrane oxygenation; eGFR, estimated glomerular filtration rate; CKD, chronic kidney disease

A total of 79 (76.70%) patients experienced postoperative AKI: 37 (46.84%) KDIGO stage 1, 17 (21.12%) KDIGO stage 2, and 25 (31.65%) KDIGO stage 3. Twenty-nine (36.71%) of whom did not have preoperative AKI. The mean peak postoperative creatinine level among the cohort was 3.98 ± 3.51 mg/dL which represented a median creatinine increase of 44.00% [19.21%-91.43%]. A total of 21 (20.39%) patients required at least temporary RRT after surgery. Five patients recovered prior to discharge and 8 (38.10%) recovered during longer term follow up.

Patients with CKD were more likely to require RRT than those without CKD but this did not reach statistical significance (16, 25.81% vs. 5, 12.20%; p = 0.09). Similarly, patients with diabetes mellitus (p = 0.49) and elevated Hemoglobin A1c (p = 0.82) had similar rates of RRT as those who did not, and TAPSE was also not significantly different between groups (1.7 cm [1.3–1.8] vs. 1.6 cm [1.3–1.8]; p = 0.57). However, patients who experienced preoperative AKI (4, 8.33% vs. 17, 30.91%; p < 0.01) or who had preoperative proteinuria (11, 14.47% vs. 10, 37.04%; p = 0.01) were more likely to require RRT postoperatively. On multivariable logistic regression modeling, higher baseline creatinine (OR 4.64 [1.24–17.32]; p = 0.02), preoperative proteinuria (OR 3.52 [1.15–10.82]; p = 0.03), and preoperative AKI (OR 3.57 [1.03–12.37]; p = 0.04) were strongly associated with postoperative RRT (Table 2). This model was predictive of RRT with an AUC of 0.783 (Fig. 1).

Table 2.

Multivariable logistic regression model of need for postoperative renal replacement therapy

Variable	Odds Ratio	95% CI	p-value
Baseline creatinine (per mg/dL)	4.64	[1.24–17.32]	0.02
Proteinuria	3.52	[1.15–10.82]	0.03
Preoperative AKI	3.57	[1.03–12.37]	0.04
Diabetes mellitus	0.75	[0.24–2.33]	0.62

Abbreviations: CI, confidence interval; AKI, acute kidney injury

Fig. 1.

Sensitivity and specificity of logistic regression model

Operative mortality was 5 (4.85%) with a 1-year survival of 80.28%. Following LVAD implantation, mild AKI did not impact survival; however, patients experiencing KDIGO stage 3 AKI had significantly lower 30-day (82.94%; p < 0.01) and 1-year survival (46.21%; p < 0.01; Supplemental Fig. 1) compared to patients experiencing milder kidney injury. Patients requiring RRT had increased operative mortality (19.05% vs. 1.22%, p < 0.01) and decreased 1-year survival (87.97% vs. 42.59%, p < 0.01; Fig. 2) compared to those who did not require RRT.

Fig. 2.

1-year Kaplan-Meier survival curve stratified by the need for renal replacement therapy

Despite an early increase in creatinine levels, and therefore decrease in eGFR, creatinine and eGFR tended to return to baseline by discharge and were stable at 3, 6, and 12 months (Supplemental Figs. 2 & 3).

Discussion

In this retrospective, single-center analysis of high acuity and surgically complex patients, we identified 3 risk factors predicting postoperative RRT following LVAD implantation: elevated baseline serum creatinine, proteinuria, and preoperative AKI. The predictive value of our multivariable logistic regression model was reasonable, with an AUC of 0.783. These findings corroborate prior reports that baseline renal function and proteinuria predict renal outcomes after LVAD [8]. Novel findings of this study are the identification of preoperative AKI as an independent risk factor for RRT in LVAD patients and the superiority of serum creatinine over cystatin C in the predictive models, although the latter was only present in approximately one third of the population. We also confirmed the mortality risk associated with RRT post LVAD implantation. Combined with baseline Cr and proteinuria, assessment of preoperative AKI during the index admission may allow for improved risk stratification for patients undergoing LVAD implantation.

In our cohort, incidence of AKI was 76% after LVAD implantation. The severity of postoperative AKI was closely tied to clinical outcomes. While mild or moderate AKI (KDIGO stage 1 or 2) was not associated with increased mortality, patients with severe AKI (KDIGO stage 3) had an increased risk of mortality. These findings suggest that lesser degrees of renal dysfunction may be reversible and in fact, a majority of our patients show improving renal function at time of discharge which remains stable at 1-year follow up. Patients who required postoperative RRT had higher rates of in-hospital (19% vs. 1%) and 1-year mortality (57% vs. 12%), which is consistent with prior reports [1, 2, 9]. Of note, 20% of our LVAD patients required postoperative RRT compared to other studies reporting lower rates [2, 8–10]. This discrepancy may be partially explained by the severity of illness in our patient population, as 70% of patients were classified as INTERMACS profile 1 or 2, and 45% of patients required preoperative Impella support. Additionally, there may be inconsistent thresholds for initiation of RRT between institutions. Despite the high rate of postoperative RRT, 61% of patients recovered kidney function either prior to or within 3 months of discharge from the hospital, a higher percentage than would be expected [11].

Proteinuria is a known predictor of AKI in other patient populations and is a common complication in diabetic patients [12]. Interestingly, uncontrolled diabetes was not predictive of postoperative RRT in our study. The link between proteinuria and CSA-AKI (cardiac surgery associated AKI) is well established [13]. Proteinuria is not only a marker of glomerular damage but also likely a reflection of overall cardiovascular risk and global endothelial dysfunction [14]. The mean UPCR in our cohort was 0.2 mg/mg which suggests that even low-level proteinuria can increase postoperative RRT risk. However, it is important to note that dipstick protein and spot urine creatinine ratios are subject to limitations and may not be as precise as other measurements.

The observed association between preoperative AKI and postoperative RRT underscores the cumulative risk of repeated renal insults, where incomplete recovery may make kidneys more susceptible to perioperative hemodynamic fluctuations inherent to LVAD surgery. Experimental animal models support that recurrent tubular injury predisposes to maladaptive repair, progressive nephron loss and increased expression of fibrotic pathways [15]. Our results raise the question of whether all increases in serum creatinine before surgery carry equivalent clinical significance. Creatinine is a non-specific biomarker and does not differentiate between true tubular injury or hemodynamic strain. Creatinine may also be influenced by other clinical factors such as muscle wasting and sarcopenia, which are common in patients requiring LVAD implantation. The discovery of novel renal biomarkers including neutrophil gelatinase-associated lipocalin (NGAL) and kidney injury molecule-1 (KIM-1) may provide new tools for phenotyping AKI [15]. Increased urine levels of these markers are reflective of tubular injury and can help predict and distinguish between intrinsic and hemodynamic rises in serum creatinine.

It remains unknown whether RRT following AKI from an intrinsic etiology (acute tubular necrosis, interstitial nephritis, etc.) will lead to similar mortality as RRT following AKI from hemodynamic changes (right ventricular failure, hypovolemia, vasoconstriction, etc.). In the setting of CSA-AKI, renal hypoperfusion due to hemodynamic changes is the primary driver of renal dysfunction [16, 17]. And indeed, proxies for hemodynamic changes including labs such as troponin and NTproBNP have been shown to correlate with AKI risk [18]. Therefore, we believe it reasonable to assume the majority of AKI in this patient population is primarily due to hypoperfusion rather than intrinsic pathology. Further study is needed to elucidate the impact of AKI from intrinsic causes from the impact of hemodynamic changes, as these causes may reflect different patient populations and lead to different outcomes.

From a management perspective, our results raise the question of whether delaying LVAD surgery in clinically stable patients with recent AKI might reduce the risk of postoperative RRT. This decision is complex as delay may increase the risk of worsening heart failure. Prospective studies are needed to determine whether allowing time for kidney recovery translates to better outcomes. Including biomarker-based risk stratification could help identify patients at greatest risk for RRT.

In summary, our findings highlight the prognostic significance of baseline serum creatinine, proteinuria, and preoperative AKI in predicting the need for postoperative RRT in high acuity and surgically complex LVAD patients. While others have shown various factors that may impact prognosis following LVAD implantation, the identification of preoperative AKI as a potentially modifiable risk factor suggests that assessment and optimization of renal function before LVAD implantation should be an integral part of preoperative planning and individualized timing strategies may be important to improving outcomes.

Limitations

As a retrospective analysis, certain biases are implicit including selection bias. Importantly, we should also note that practice patterns for initiation of RRT vary by provider. Certain providers may be more liberal or conservative regarding RRT initiation, and the timing and indication of RRT may significantly impact overall outcomes following its initiation. Furthermore, analysis of high-volume, single-center experiences, such as the present work, may generalize poorly to other centers. Residual confounding may persist due to variables unaccounted for in our analysis, (e.g., hemodynamic status, use of nephrotoxic medications, and indications for RRT initiation). As our outcomes are limited to 1-year mortality, extended follow-up is needed to determine whether patients who recover kidney function following RRT suffer from increased long-term mortality. It is also important to note that our cohort size is limited, and larger studies are necessary to more fully elucidate the impact of our findings.

Conclusions

Acute kidney injury is common following LVAD implantation. The need for renal replacement therapy following LVAD remains a harbinger of poor outcomes and increased mortality. Use of preoperative creatinine, proteinuria, and presence of preoperative AKI predict likelihood of renal replacement therapy in LVAD implantation, and may allow for improved decision-making between patients and providers.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

CWD data analysis, manuscript writing, manuscript review/editing. TJG data collection, analysis, manuscript writing, review. AR manuscript review/editing, data collection. AA data contribution, manuscript review/editing. NK data contribution, manuscript review/editing. GM data contribution, manuscript review/editing. DAR data contribution, manuscript review/editing. HN data contribution, manuscript review/editing. PVB manuscript review/editing. HL data collection, data analysis, project administration, manuscript writing, manuscript review/editing.

Data availability

All data generated or analyzed during this study are included in this article and its supplementary material files. Further enquiries can be directed to the corresponding author.

Declarations

Competing interests

The authors declare no competing interests.