Acute Kidney Injury Following Transcatheter Aortic Valve Implantation: Association with Contrast Media Dosage and Contrast Media Based Risk Predication Models

The effect of contrast media (CM), delivered prior to- and during transcatheter aortic valve implantation (TAVI), on kidney function, following the procedure, is debatable. Consequently, the performance of CM-based, acute kidney injury (AKI) risk prediction models is also questionable. We retrospectively studied 210 patients that underwent TAVI. We recorded the dose of CM used prior and during TAVI, calculated the results of different AKI risk assessment models containing a CM module, and tested their association with AKI after the procedure. AKI was diagnosed in 38 patients (18.1%). The baseline estimated glomerular filtration rate (eGFR) was lower in the AKI+ group compared to AKI− group (51 ± 19.3 versus 64.5 ± 19 mL/min/1.73 mr2, respectively). While the dose of CM delivered prior to TAVI, during TAVI or the cumulative amount of both did not differ between the groups, the results of all tested risk models were higher in AKI+ patients. However, by multivariable analysis, only eGFR had a consistent independent association with AKI. We suggest that the dose of CM delivered prior or during TAVI is not associated with AKI and that the predictive power of CM based AKI risk models is, in all probability, limited to eGFR alone.


Introduction
Acute kidney injury (AKI) following transcatheter aortic valve implantation (TAVI) is frequent and strongly associated with adverse outcomes [1][2][3][4][5]. Therefore, it is imperative to recognize the parameters that inflict an increased risk for AKI following TAVI. Previous studies have linked, although inconsistently, the volume of radiographic contrast media (CM) administered during TAVI, with the development of AKI after the procedure [5][6][7][8][9][10]. It is possible, although not tested thus far, that preparatory exposure to CM, before TAVI, may impose additional deleterious effects on post-TAVI kidney function. Risk assessment tools, incorporating CM volume, have proved useful in predicting the development of AKI following percutaneous coronary procedures [11][12][13][14]. However, the utility of such risk models in predicting post-TAVI AKI is less established [6,15,16].
We investigated the association between the total volume of CM delivered during TAVI as well as during near-past CM utilizing procedures and the occurrence of post-TAVI AKI. We additionally tested the ability of CM based AKI risk assessment models, previously verified with percutaneous coronary interventions (PCI), to predict post-TAVI AKI. We also CR4EATME3AD3 model was calculated as follows: contrast volume > 200 mL, 2 points; eGFR < 60 mL/min/1.73 m 2 , 4 points; emergency procedure, 2 points; age > 70 years, 2 points; hypotension, 2 points; previous myocardial infarction, 2 points; left ventricular ejection fraction < 45%, 3 points; anemia, 2 points; diabetes mellitus, 3 points [14].
Statistical analysis: Categorical variables were reported as absolute numbers and percentages and were compared using the chi-square test. Continuous data with normal distribution were reported as the mean and standard deviation (SD) and were compared using an independent samples t-test. Continuous variables with a non-normal distribution were reported as medians and interquartile ranges (IQR), and were compared using the Mann-Whitney U test. Univariable and multivariable logistic regression analyses were performed to evaluate for predictors of AKI following TAVI with reported odds ratios (OR's) and 95% confidence intervals (CI). We excluded any variable with more than 10% missing values. First, univariable associations were determined for all potential covariates. Covariates with a p value of less than 0.1 on univariate analyses were entered into the multivariable logistic regression models. Risk variables that were determined to be significant in the univariable analysis were subsequently tested with the multivariable modeling. A 2-sided p-value of ≤0.05 was considered statistically significant.
Statistical analysis: Categorical variables were reported as absolute numbers and percentages and were compared using the chi-square test. Continuous data with normal distribution were reported as the mean and standard deviation (SD) and were compared using an independent samples t-test. Continuous variables with a non-normal distribution were reported as medians and interquartile ranges (IQR), and were compared using the Mann-Whitney U test. Univariable and multivariable logistic regression analyses were performed to evaluate for predictors of AKI following TAVI with reported odds ratios (OR's) and 95% confidence intervals (CI). We excluded any variable with more than 10% missing values. First, univariable associations were determined for all potential covariates. Covariates with a p value of less than 0.1 on univariate analyses were entered into the multivariable logistic regression models. Risk variables that were determined to be significant in the univariable analysis were subsequently tested with the multivariable modeling. A 2-sided p-value of ≤0.05 was considered statistically significant.

Results
Between February 2015 and March 2020, a total of 355 patients underwent TAVI at PMC. One hundred and forty-five patients were excluded from the analysis. The reasons for excluding these patients are depicted in Figure 1. The baseline data of the remaining 210 patients, which were included in the analysis, is summarized in Table 1. AKI was diagnosed in 38 patients (18.1%). Of them, 26 (68.4%) patients were classified as suffering from stage-1 AKI, eight (21.1%) patients were classified as stage-2 AKI, three (7.9%) patients had stage-3 AKI and one (2.6%) patient was diagnosed with stage-4 AKI. For the most part, the two groups had a similar baseline profile with the exception of a few significant differences. AKI+ patients, compared to AKI− patients, had a higher prevalence of prior neurological insult, higher surgical risk, as assessed by the STS score, lower hemoglobin levels and a higher rate of anemia (86.8% versus 69.2%, respectively, p = 0.028). They also had worse baseline kidney function with higher baseline creatinine and lower eGFR. Overall, 60.5% of AKI+ patients had eGFR ≤ The baseline data of the remaining 210 patients, which were included in the analysis, is summarized in Table 1. AKI was diagnosed in 38 patients (18.1%). Of them, 26 (68.4%) patients were classified as suffering from stage-1 AKI, eight (21.1%) patients were classified as stage-2 AKI, three (7.9%) patients had stage-3 AKI and one (2.6%) patient was diagnosed with stage-4 AKI. For the most part, the two groups had a similar baseline profile with the exception of a few significant differences. AKI+ patients, compared to AKI− patients, had a higher prevalence of prior neurological insult, higher surgical risk, as assessed by the STS score, lower hemoglobin levels and a higher rate of anemia (86.8% versus 69.2%, respectively, p = 0.028). They also had worse baseline kidney function with higher baseline creatinine and lower eGFR. Overall, 60.5% of AKI+ patients had eGFR ≤ 60 mL/min/1.73 m 2 compared to 37.2% of AKI− patients (p = 0.008). Baseline echocardiographic data was not found to be different between the two groups. After TAVI, AKI+ patients, compared to their counterparts, had higher SCr and lower eGFR during their post procedural hospitalization both at short-term (i.e., 30 days) and long-term (i.e., 12-month) follow-up (Table 2). CM data and risk models data, at the different tested time intervals, indexed according to the presence or absence of AKI after TAVI, are presented in Table 3. AKI+ patients and AKI− patients did not significantly differ in the volume of CM administered during TAVI. Nor did they differ in the amount of CM delivered, prior to TAVI, within a seven-day or a 30-day periods or in the cumulative amount of CM administered during all procedures. The scores for all tested risk models were significantly higher for AKI+ patients compared to AKI− patients either when they included the volume of CM administered during TAVI or the cumulative volume of CM as described earlier. Additionally, the modifications of Mehran risk model and CR4EATME3AD3 model, which were calculated without the CM element, were also higher in AKI+ patients.  Table 4 shows the results of univariable analysis. Baseline eGFR < 45 mL/min/1.73 m 2 , hemoglobin level and the STS score were significantly associated with AKI following TAVI. Additionally, all risk models, whether computed without the CM volume module or with it, as administered during TAVI or during the prespecified seven-day and 30-day intervals, were also predictive of AKI. In contrast, neither the administered volume of CM during TAVI nor the seven-day or 30-day cumulative dose of CM were associated with AKI development.
Each of the risk models was discretely entered into a multivariable logistic regression model together with the hemoglobin level, eGFR < 45 mL/min/1.73 m 2 and STS score. By multivariable analysis none of the risk models predicted post-TAVI AKI. Neither did the hemoglobin level nor the STS score. Only pre-TAVI eGFR, in most of the analyses, was independently associated with AKI after it (Table 5).

Discussion
The main findings of the study were: (1) The volume of CM administered during TAVI and the cumulative volume of CM administered within seven days or within 30 days from TAVI were not associated with AKI post-TAVI; and (2) None of the tested risk assessment models that included a CM module, were independently associated with post-TAVI AKI.
(3) eGFR was the only independent predictor for AKI.
AKI is a serious complication, commonly encountered following TAVI with an incidence of any level of AKI, reported from large registries and metanalyses, at around 20% [1,3]. The development of AKI, following TAVI, significantly increases the risk of both short-term and long-term deleterious complications, such as myocardial infarction, life threatening bleeding, permanent renal dysfunction necessitating renal replacement therapy, and mortality [1,[3][4][5][20][21][22][23]. Various baseline characteristics, co-morbidities and peri-procedural factors have been linked with TAVI related AKI. Some of the most reported are the baseline decreasing level of hemoglobin, chronic kidney dysfunction, acute bleeding necessitating blood transfusions and hemodynamic instability during the procedure [2,5,22,24,25]. Still, there is much uncertainty with a multitude of proposed risk factors previously associated with AKI following other procedures in which CM is used, and the role they play in the development of AKI following TAVI with conflicting reported results. For instance, while some studies demonstrated a relationship between increasing age and AKI development following TAVI, other studies did not [3,6,7,10,25]. While in some studies LVEF was dis-concordantly associated with post-TAVI AKI, in other publications it was not [6,7,10,23,25,26]. Even decreasing baseline eGFR and chronic kidney dysfunction have an uncertain magnitude of effects on post-TAVI AKI [3,[6][7][8]10,[25][26][27]. The relationship between CM volume and TAVI-related AKI is also controversial. In several previous studies there was no significant difference in the volume of CM delivered during TAVI, between AKI+ patients and AKI− patients, nor between CM dose and post-procedural AKI [2,4,5,8,16]. Conversely, other studies described higher dosages of CM in AKI+ patients compared to AKI, and an association between the volume of CM delivered during TAVI and the subsequent development of AKI [6,9,10,25,26].
The uncertain association between administered CM volume and AKI has been described with other medical procedures utilizing radiographic CM. In patients with eGFR above 30 mL/min/1.73 m 2 who underwent CT testing, there was no significant difference in the incidence of AKI between those who received intravenous CM and those who did not, and the results for patients with eGFR less than 30 mL/min/1.73 m 2 were conflicting [28,29]. An analysis of observational studies concluded that radiocontrast use in CT scanning was not causally related to changes in kidney function [30]. These results were echoed in the most recent update of the American College of Radiology consensus statement in which it was declared that the risk of AKI developing following exposure to intravenous iodinated CM has been exaggerated and that the true risk of AKI, related to CM administration, remains uncertain even for patients with severe kidney disease [31]. It has been suggested that intravenous administration of CM, like with CT angiography, may impose a different risk of AKI than the one associated with arterial administration of CM, like with PCI and TAVI; however, the data is inconclusive [32][33][34][35][36][37].
CM related AKI has been extensively studied in the context of PCI. However, assessing the true impact of CM on the development of AKI, following PCI, has proven to be a difficult task. Studies that evaluated this coupling between CM and AKI widely differed in the type, chemical characteristics, pharmacokinetics and volume of CM, the route it was administered, the medical procedure it was given for and in the definition used to diagnose AKI [38]. Notably, a broad range of contrast volume (from below 100 mL to above 800 mL) has been associated with post-PCI nephropathy [12,13,[39][40][41]. There are several fundamental differences between patients treated with TAVI and patients treated with PCI which, potentially, position the former at higher risk to develop AKI than the later. Patients undergoing TAVI are usually elderly, fragile, frequently suffering from multiple co-morbidities and have reduced GFR [7,42]. Additionally, PCI-treated patients are usually exposed to CM once during their index procedure. TAVI treated patients, however, are frequently exposed to a large cumulative volume of CM during several diagnostic and interventional procedures and over a relatively short period of time. Interestingly, despite these differences, Venturi et al. found that AKI actually occurred less frequently in patients undergoing TAVI than in patients undergoing PCI, even after propensity score matching [7]. Following TAVI, unlike other procedures that utilize CM, there is an immediate and sustained improvement in hemodynamics resulting in an increase in cardiac output and systemic perfusion [43][44][45]. Still, even though this can possibly translate into the reduced incidence of AKI following TAVI, the rate of immediate kidney function improvement after TAVI was found to be small in a recent publication (5%) [46].
The diagnosis of AKI tends to lag after the index exposure to CM, with gradual worsening of kidney function over the ensuing days. Accordingly, the assessment for CM associated nephrotoxicity was traditionally made within 48 to 72 h of exposure and according to VARC-3 consensus statements this time frame has been further extended to seven days [13,17,39]. Thus far, the effect of CM on kidney function was tested for CM delivered during TAVI alone, while the possible effect inflicted by near past exposures to CM was overlooked. This led us to test for an additive effect of preceding exposures to CM on the incidence of AKI following TAVI. In this study, the volume of CM administered during TAVI alone, during a 7-day period or during a 30-day period, did not significantly differ between AKI+ patients and AKI− patients, nor did it independently predict the development of AKI. Similarly, in a recent study, assessing the effect of multiple exposures to CM during recurrent diagnostic and interventional coronary procedures, although over a period of several years, worsening of renal function was associated with known risk factors for the progression of kidney disease but not with cumulative CM volume [47].
The difficulties in predicting AKI following TAVI led to the development of risk assessment models. Incorporating several parameters to form a risk model, can potentially improve the predictive power beyond that of the individual modules included in it. AKI risk prediction models have been developed, tested and validated for coronary procedures [11][12][13][14]48,49]. Sadly, only a few of these models have been specifically designed to assess for the risk of post-TAVI AKI. For instance, Zivcovic et al. introduced an AKI risk calculator that was meant to be utilized prior to TAVI and therefore did not include a CM component [27]. As in this study, several previous studies tested the effectiveness of CMbased, non-TAVI dedicated risk models in predicting AKI following TAVI [6,15,16,24,26,50]. In our study, none of the CM-based risk models we tested independently predicted post-TAVI AKI. Mach et al. reported that none of the six tested CM-based risk models (including the Mehran risk model) were significantly different between AKI+ and AKI− patients or, similarly to our results, independently predicted AKI [16]. Rosa et al. describes that all of their tested risk models, of which four included a CM module and two did not, had poor accuracy in terms of predicting the occurrence of any AKI. However, there was an improvement in AKI risk prediction for more advanced stages of AKI. By univariable logistic regression analysis, only risk models with a CM module were associated with AKI. The results of multivariable logistic regression coefficients of risk for any AKI or the different stages of AKI were not reported [15]. In contrast, a small study of 93 patients, of which AKI was diagnosed in 24 of them, found in a univariable analysis that the Mehran risk model, as well as CM volume, predicted post-TAVI AKI. The authors also reported that in multivariable analysis, CM volume, the Mehran score, SCr and eGFR were all independently associated with AKI [6]. However, the study was of limited sample size and, given the low number of endpoints achieved, was underpowered to accurately explore the aforementioned endpoints.
In our study, decreasing eGFR, as previously described by others, was significantly and independently associated with AKI across most analyses [5,8]. In fact, it was the only independent predictor for AKI. Therefore, with caution, we suggest that the AKI predictive power of risk score models containing a CM volume module, is almost entirely limited to pre-procedural eGFR.
Limitations: This is an individual center report and has the limitations associated with retrospective analysis. Given the relatively small number of AKI events, our study was underpowered to evaluate for the most severe stages of AKI. The assessment of baseline sCr was made prior to TAVI and not prior to the first exposure to CM within the assessed time interval. Additionally, most patients were discharged before seven days had elapsed from their procedure and therefore, even though it is our institutional practice to discharge the patient after kidney function has stabilized, it is possible that further deterioration in kidney function ensued after discharge without being recorded. The use of drugs with nephrotoxic properties close to contrast media exposure was also not recorded.

Conclusions
Neither the volume of CM delivered during TAVI nor the cumulative amount of CM delivered during a time period that starts either seven-days or 30 days before TAVI and ends with TAVI are associated with AKI. The power of non-TAVI dedicated CM based risk models, in predicting AKI, is limited to pre-procedural kidney dysfunction. Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.

Data Availability Statement:
The data presented in this study was obtained from PMC's local TAVI prospective registry and is available on request from the corresponding author.
Conflicts of Interest: D.S. received a speakers' fee from Abbott Medical Laboratories, Sanofi, Novartis and Novo Nordisk; E.Y.B. received a speakers' honoraria from CTS and Novo Nordisk received research support paid to the University of Pennsylvania by Medtronic Inc. and Impulse Dynamics Ltd.