Acute Kidney Injury Following In-Patient Lower Extremity Vascular Intervention From the National Cardiovascular Data Registry

Abstract

OBJECTIVES

The authors analyzed data from the NCDR (National Cardiovascular Data Registry) PVI Registry and defined acute kidney injury (AKI) as increased creatinine of ≥0.3 mg/dl or 50%, or a new requirement for dialysis after PVI.

BACKGROUND

AKI is an important and potentially modifiable complication of peripheral vascular intervention (PVI). The incidence, predictors, and outcomes of AKI after PVI are incompletely characterized.

METHODS

A hierarchical logistic regression risk model using pre-procedural characteristics associated with AKI was developed, followed by bootstrap validation. The model was validated with data submitted after model creation. An integer scoring system was developed to predict AKI after PVI.

RESULTS

Among 10,006 procedures, the average age of patients was 69 years, 58% were male, and 52% had diabetes. AKI occurred in 737 (7.4%) and was associated with increased in-hospital mortality (7.1% vs. 0.7%). Reduced glomerular filtration rate, hypertension, diabetes, prior heart failure, critical or acute limb ischemia, and pre-procedural hemoglobin were independently associated with AKI. The model to predict AKI showed good discrimination (optimism corrected c-statistic = 0.68) and calibration (corrected slope = 0.97, intercept of −0.07). The integer point system could be incorporated into a useful clinical tool because it discriminates risk for AKI with scores ≤4 and ≥12 corresponding to the lower and upper 20% of risk, respectively.

CONCLUSIONS

AKI is not rare after PVI and is associated with in-hospital mortality. The NCDR PVI AKI risk model, including the integer scoring system, may prospectively estimate AKI risk and aid in deployment of strategies designed to reduce risk of AKI after PVI.

Acute kidney injury (AKI) is an important complication of medical procedures that use iodinated contrast, such as invasive angiography. The incidence of AKI following coronary revascularization is well described. Validated risk models have been developed and used as the foundation for quality improvement initiatives to reduce AKI from its current rate of 4% to 7% (1–3). Iodinated contrast is also used in peripheral vascular intervention (PVI), with AKI occurrence as high as 10%. Because of this, the incidence, predictors, and outcomes of AKI after PVI have become the focus of study (4–6). Prior analyses suffer from the absence of a comprehensive database for PVI procedures and outcomes from which to describe care and build valid prediction models. The American College of Cardiology’s NCDR (National Cardiovascular Data Registry) PVI Registry addresses this gap by providing multicenter data of patients undergoing PVI throughout the United States (7).

Leveraging the NCDR PVI Registry, we sought to identify the incidence and predictors of AKI after PVI using the Acute Kidney Injury Network (AKIN) criteria (8). Additionally, we assessed the outcomes of patients with and without AKI to underscore its clinical importance. Finally, we developed and validated an AKI risk prediction model and a derivative simple integer-based risk model. These efforts are intended to define the magnitude and importance of AKI in PVI and to lay the foundation for future quality improvement initiatives to improve procedural safety and outcomes.

METHODS

STUDY POPULATION.

The NCDR PVI Registry records data on patient, hospital, and procedural characteristics and outcomes at over 200 U.S. locations. Data are collected and transmitted to the NCDR through certified processes, then undergo quality assurance and auditing (9). All lower extremity PVI cases at participating sites are included. This is confirmed through the audit process. A waiver of written informed consent and authorization for this study was granted by the Advarra Institutional Review Board.

INCLUSION AND EXCLUSION CRITERIA.

Consecutive patients undergoing PVI were eligible for inclusion. Those on hemodialysis, without creatinine assessments, and those discharged on the day of PVI were excluded. Patients who underwent PVI between March 30, 2014, and June 30, 2017, were included in the derivation cohort. This analysis was completed on a “per procedure” level; therefore, individual patients undergoing multiple procedures during the study period may be included multiple times. Eligible patients who underwent PVI between July 1, 2017, and December 31, 2018, were included in the validation cohort.

OUTCOMES.

AKI was the primary outcome, assessed as the difference between pre-procedural serum creatinine and peak creatinine level 6 to 24 h after PVI. Any AKIN injury was considered AKI. Stages of AKIN injury were used to assess the relationship of AKI severity with in-hospital mortality. Briefly, AKIN stages of AKI are Stage 1: increase in serum creatinine ≥0.3 mg/dl or 50% from baseline; Stage 2: 200% to 300% increase in creatinine; and Stage 3: >300% increase in creatinine or increase to ≥4.0 mg/dl with an acute increase of ≥0.5 mg/dl or a new requirement for dialysis (8). Urine output is not recorded in the registry and was not included in the definition. In the reporting of mortality, patients were counted based only on the final procedure of any hospital admission.

VARIABLES.

The estimated glomerular filtration rate (eGFR) was calculated using the Modification of Diet in Renal Disease equation. Patients were further classified as having normal renal function (eGFR ≥60 ml/min/1.73 m²), mild (eGFR 45 to 59 ml/min/1.73 m²), moderate (eGFR 30 to 44 ml/min/1.73 m²), or severe chronic kidney disease (CKD, eGFR <30 ml/min/1.73 m²). This is a slight modification to the National Kidney Foundation classification scheme (10). Bleeding was defined as a hemoglobin drop of ≥3 g/dl, blood transfusion, and/or any bleeding requiring intervention/surgery. All variables were prospectively collected. Ankle-brachial index testing was not required and was performed in a minority of patients included in the registry and in this study (41.7%).

STATISTICAL ANALYSES.

Data are described as mean ± SD or proportion (%). Comparisons between patients who did and did not develop AKI was performed using chi-square or Student’s t-test, as appropriate. Hierarchical logistic regression used pre-procedural variables defined a priori, based on clinical judgment and existing published reports (2), with site as a random effect to account for clustering of patients. Potential predictor variables included age, sex, race (African American vs. not) and smoking history. Clinical factors included GFR, hypertension, diabetes mellitus, statin use, coronary artery disease (CAD), prior heart failure, prior percutaneous coronary intervention (PCI), prior PVI, and hemoglobin. Severity of disease was captured as peripheral artery disease (PAD) status (critical limb ischemia [CLI] vs. acute limb ischemia [ALI] vs. intermittent claudication). Contrast volume was intentionally excluded from risk model development because it is not known before the procedure and is a potential target for intervention. Linearity between AKI risk and continuous variables were tested with cubic splines. We then developed a reduced model that retained 90% of the full model’s R-square value (11). Internal validation was performed with 1,000 bootstrap replications. Model performance was assessed with the c-statistic for discrimination and by evaluating the slope and intercepts of the observed and predicted AKI risks to describe its calibration. Lastly, the beta coefficients of the regression model were converted to a simple integer scoring system that was used to categorize risk as low (0 to 4 points), moderate (5 to 11), or high (>12 points). A sensitivity analysis was then performed utilizing randomly selected procedures if more than 1 PVI was performed on any given hospital admission. The low- and high-risk categories roughly corresponded to the upper and lower 20% of risk and represent clinically reasonable thresholds for stratifying risk and developing prospective strategies to reduce AKI. All analyses were performed with SAS version 9.4 (SAS Institute, Cary, North Carolina).

RESULTS

A total of 27,119 PVI procedures between March 30, 2014, and June 30, 2017, were eligible for the derivation cohort. The following were excluded: 1,658 (6.1%) requiring hemodialysis before PVI, 3,655 (13.5%) missing creatinine assessments, and 11,800 (43.5%) discharges on the day of PVI. The remaining 10,006 procedures formed the derivation cohort (Figure 1). Baseline characteristics and procedural variables are presented in Table 1. A total of 7,880 distinct patients are analyzed, with 79.9% included in the registry a single time. Two procedures were included for 15.3% of patients and 3 or more for 4.8%. There are 81 sites included, each contributing a median of 85 cases (interquartile range [IQR]: 25 to 157) performed by 564 operators who contributed a median of 5 cases each (IQR: 2 to 19). The median rate of exclusion per site was 65% (IQR: 53% to 76%). Demographic data were over 99% complete.

FIGURE 1. Study Flow.

NCDR = National Cardiovascular Data Registry; PVI = peripheral vascular intervention.

TABLE 1.

Overall Procedural Characteristics and Comparison by AKI Groups

	Total (N = 10,006)	AKI (n = 737)	No AKI (n = 9,269)	p Value
Age, yrs	69.4 ± 11.8	69.9 ± 12.0	69.3 ± 11.8	0.226
Male	5,818 (58.1)	392 (53.2)	5,426 (58.5)	0.004
African American	1,413 (14.1)	134 (18.2)	1,279 (13.8)	0.001
Hispanic/Latino	461 (4.6)	45 (6.1)	416 (4.5)	0.044
Hypertension	9,004 (90.0)	694 (94.2)	8,310 (89.7)	<0.001
Dyslipidemia	8,096 (81.0)	601 (81.8)	7,495 (80.9)	0.573
Diabetes mellitus	5,235 (52.4)	480 (65.2)	4,755 (51.3)	<0.001
Coronary artery disease	5,172 (51.7)	409 (55.5)	4,763 (51.4)	0.032
Prior heart failure	1,877 (18.8)	221 (30.0)	1,656 (17.9)	<0.001
Current smoker	3,873 (38.7)	233 (31.6)	3,640 (39.3)	<0.001
Indication for PVI				<0.001
Claudication	3,456 (34.5)	156 (21.2)	3,300 (35.6)
Critical limb ischemia	5,034 (50.3)	422 (57.3)	4,612 (49.8)
Acute limb ischemia	1,201 (12.0)	125 (17.0)	1,076 (11.6)
Other*	315 (3.1)	34 (4.6)	281 (3.0)
Hemoglobin, g/dl	12.3 ± 2.3	11.4 ± 2.2	12.4 ± 2.2	<0.001
Glomerular filtration rate, ml/min/1.73 m2	76.5 ± 32.9	66.3 ± 36.6	77.4 ± 32.4	<0.001
Chronic kidney disease class				<0.001
Normal, GFR ≥60 ml/min/1.73 m2	6,688 (67)	368 (50)	6,320 (68)
Mild, GFR 45–59 ml/min/1.73 m2	1,822 (18)	157 (21)	1,665 (18)
Moderate, GFR 30–44 ml/min/1.73 m2	1,106 (11)	127 (17)	979 (11)
Severe, GFR <30 ml/min/1.73 m2	390 (3.9)	85 (12)	305 (3)
Baseline creatinine, mg/dl	1.12 ± 0.63	1.40 ± 0.96	1.10 ± 0.59	<0.001
Creatinine change, mg/dl	0.00 ± 0.54	0.84 ± 1.35	−0.07 ± 0.32	<0.001
Post-PVI bleeding	789 (7.9)	178 (24.2)	611 (6.6)	<0.001
Blood transfusion	679 (6.8)	171 (23.3)	508 (5.5)	<0.001
New hemodialysis	41 (0.4)	41 (5.6)	0 (0)	<0.001
Ankle-brachial index value	0.8 ± 0.3	0.7 ± 0.3	0.8 ± 0.3	0.001
Contrast volume, ml, mean	138.3 ± 90.3	131.9 ± 87.6	138.8 ± 90.5	0.048
Fluoroscopy time, min	23.0 ± 19.3	24.6 ± 20.5	22.9 ± 19.2	0.018
Chronic total occlusion	3,111 (31.1)	219 (29.7)	2,892 (31.2)	0.401
Lesion length, mm	152.7 ± 148.8	167.7 ± 166.0	151.4 ± 147.2	0.009

Values are mean ± SD or n (%).

^*Other indications for PVI included treatment or prevention of aneurysmal disease, maintenance of patency in asymptomatic patients and facilitation of other procedures.

AKI = acute kidney injury; GFR = glomerular filtration rate; PVI = peripheral vascular intervention.

Patients discharged on the day of procedure were younger, more likely male and African American, with fewer comorbidities. They were less likely to have limb-threatening conditions; 31.4% had CLI or ALI, whereas 62.4% of those included in the derivation cohort had these indications (Supplemental Table 1). An additional 6,759 PVI patients between July 1, 2017, and December 31, 2018, represent an external validation cohort. Characteristics of the validation cohort are presented in Table 2.

Table 2.

Derivation and Validation Cohorts: Procedural Characteristics

	Derivation (n = 10,006)	Validation (n = 6,759)
Age, yrs	69.4 ± 11.8	69.5 ± 11.6
Male	5,818 (58.1)	3,946 (58.4)
African American	1,413 (14.1)	1,092 (16.2)
Hypertension	9,004 (90.0)	6,077 (90.0)
Dyslipidemia	8,096 (81.0)	5,344 (79.1)
Diabetes mellitus	5,235 (52.4)	3,765 (55.7)
Coronary artery disease	5,172 (51.7)	3,363 (49.8)
Prior heart failure	1,877 (18.8)	1,455 (21.5)
Current smoker	3,873 (38.7)	2,582 (38.2)
Critical limb ischemia	4,768 (47.7)	3,659 (54.2)
Acute limb ischemia	1,219 (12.2)	858 (12.7)
Hemoglobin, g/dl	12.3 ± 2.3	12.1 ± 2.2
GFR, ml/min/1.73 m2	76.5 ± 32.9	76.9 ± 34.1
Chronic kidney disease class
Normal GFR, ≥60 ml/min/1.73 m2	6,688 (67)	4,518 (67)
Mild, 45–59 ml/min/1.73 m2	1,822 (18)	1,192 (18)
Moderate, 30–44 ml/min/1.73 m2	1,106 (11)	786 (12)
Severe, <30 or dialysis	390 (3.9)	263 (3.9)
Baseline creatinine, mg/dl	1.12 ± 0.63	1.13 ± 0.73
Creatinine change, mg/dl	0.00 ± 0.54	0.00 ± 0.62
Post-PVI bleeding	789 (7.9)	534 (7.9)
Blood transfusion	679 (6.8)	427 (6.3)
Ankle-brachial index value	0.8 ± 0.3	0.7 ± 0.3
Contrast volume, mL	138.3 ± 90.3	119.2 ± 75.3
Fluoroscopy time, min	23.0 ± 19.3	21.5 ± 21.8
Chronic total occlusion	3,111 (31.1)	1,926 (28.5)
Lesion length, mm	152.7 ± 148.8	158.8 ± 152.9

Values are mean ± SD or n (%).

Abbreviations as in Table 1.

Overall, the average age was 69.4 ± 11.8 years, 58% were male, and over 90% had hypertension, and over 80% had hyperlipidemia. Most patients were Caucasian (83.3%) or African American (14.1%), and 4.6% were Hispanic or Latino. Diabetes and CAD were present in over one-half of patients. Prior heart failure was present in 19% and almost 40% were current smokers. Post-procedural laboratory testing was performed a median of 15.5 h after PVI (IQR: 12.3to 19.1). AKI occurred in 737 (7.4%), including a need for new hemodialysis in 41 patients (0.4%). Stage 1 AKI occurred in 605 patients (6.0%), Stage 2 in 50 (0.5%), and Stage 3 in 82 (0.8%).

Patients with AKI were more likely to be female, African American or Hispanic/Latino with hypertension, diabetes, and have CAD or prior heart failure (Table 1). Statin use was higher among patients with AKI (57.2% vs. 51.8%) but did not remain significant in multivariable analysis. Claudication was the indication for PVI in 36.4% of patients overall. CLI was the indication in 48%, and ALI was the indication in 12%—both of which were more common among patients with AKI. Among patients with AKI, the average baseline serum creatinine was 1.40 ± 0.96 mg/dl and increased by 0.84 ± 1.35 mg/dl after PVI, whereas for patients without AKI serum creatinine was 1.10 ± 0.59 mg/dl and decreased by 0.07 ± 0.32 mg/dl. GFR, pre-procedural hemoglobin, and ankle-brachial index were lower among patients with AKI. The prevalence of chronic total occlusions was 31% and did not differ between AKI groups though lesions were longer among those with AKI. The contrast load during PVI was lower among patients with AKI.

The contrast-to-GFR ratio was 2.62 in those with AKI and 2.12 in those without (p < 0.001). In addition, patients with AKI were more likely to present with CLI or ALI and less likely to be current smokers. The in-hospital mortality rate was 7.1% when AKI occurred and 0.6% when it did not (p < 0.001). The median time to death after PVI was 5 days (IQR: 2 to 9). Additionally, higher stages of AKI were associated with higher risk of mortality (Figure 2) and AKI was associated with increased mortality regardless of presenting symptoms (Supplemental Table 2). Cardiovascular disease was the most common cause of in-hospital death (37.5%), followed by pulmonary (20.9%), infectious (9.2%), gastrointestinal (6.7%), renal failure (4.2%), bleeding (3.3%), neurological disease (3.3%), and peri-operative death from noncardiovascular procedures (1.7%). Other causes of death each represent <1%. The proportion of mortality from each cause separated by AKI groups is presented in Supplemental Table 3. Bleeding (24.2% vs. 6.6%; p < 0.001) and blood transfusion (23.3% vs. 5.5%; p < 0.001) were more likely among patients with AKI.

FIGURE 2. All-Cause Mortality.

All-cause mortality is shown with versus without acute kidney injury (AKI) and by stage of AKI or new requirement for hemodialysis after peripheral vascular intervention.

Variables independently associated with AKI included sex, race, GFR, diabetes, heart failure, hypertension, ALI or CLI, and hemoglobin (Figure 3). After bootstrap validation, model discrimination (optimism corrected c-statistic = 0.678) and calibration (corrected slope = 0.97, intercept of −0.07) were good in the derivation cohort and varied minimally when considered by presenting symptoms (claudi-cants c-statistic = 0.666, CLI = 0.665, and ALI = 0.631). In the validation cohort, the discrimination (c-statistic = 0.651) and calibration (corrected slope = 0.93, intercept of −0.187) remained reasonable. The simple integer point system showed good discrimination of risk for AKI (Central Illustration) in the validation cohort. Scores ≤4 and ≥12 correspond to the lowest and highest 20% of AKI risk. A plot of predicted risk among those with and without AKI is presented as Supplemental Figure 1. Sensitivity analysis revealed similar performance of the integer score for all procedures versus randomly selected procedures within a single admission (Supplemental Table 4). Variables with the strongest prediction included baseline GFR, ALI, history of hypertension or heart failure, and anemia.

FIGURE 3. Predictors of AKI.

Predictors of acute kidney injury (AKI) are shown with their associated odds ratios and 95% confidence intervals. ALI = acute limb ischemia; CLI = critical limb ischemia; GFR = glomerular filtration rate; HGb = hemoglobin; PAD = peripheral artery disease.

CENTRAL ILLUSTRATION. Simple Integer Point System.

Distribution of scores and prevalence of each score in the study population (A) with discrimination of PVI patients’ risk for AKI (B). AKI = acute kidney injury; eGFR = estimated glomerular filtration rate; PVI = peripheral vascular intervention.

DISCUSSION

AKI is potentially avoidable, but there is limited insight into its prevalence and predictors following PVI. We assessed pre-procedural risk factors and found that baseline renal dysfunction, hypertension, diabetes mellitus, congestive heart failure, lower hemoglobin levels, and the presence of CLI or ALI were independently associated with AKI. We found a 10-fold increase in in-hospital mortality and a tripling of PVI-related bleeding associated with AKI. After developing and validating a risk model for predicting AKI, we created an integer scoring system that could serve as the basis for a practical clinical tool to improve the safety of PVI. The positive predictive value of a low score is low (8.9%), whereas the negative predictive value is 95.7% when it is compared with the “non-low” scores. Also as expected, a high score has a relatively low positive predictive value of 14.5% because AKI remains a rare event. This also leads to the high negative predictive value of 93.5%.

This work confirms and extends prior studies on AKI after PVI, though the 7.4% rate of AKI is lower than rates of 10% to 13% reported in patients with CLI (5,12). Rates <1% have been reported in all-comers analyses (13). The low AKI rates are likely related to inclusion of patients with stable claudication as well as exclusion of patients with bleeding complications. Additionally, the majority of patients included the NCDR PVI database were discharged home on the day of PVI without assessment of renal function before discharge and were ineligible for the present analysis. Their inclusion would have likely lowered the AKI rate.

Recent work by Weisbord et al. (14) defines “clinically significant” AKI following contrast exposure as that associated with death, need for dialysis, or persistent renal dysfunction 90 days later. The rate of “clinically significant” AKI was low (1.2%), despite an overall incidence of AKI of 9.7%. They also noted that pre-procedural GFR mediated the outcome rather than AKI. Although this does raise the question of how to best define clinically relevant AKI, it is widely accepted that AKI is associated with adverse outcomes. The current study adds to the published reports in a substantial way and may help advance the discussion of this important outcome and how to best define it.

Previously, Bodewes et al. (15) demonstrated that AKI predicts readmission after PVI, and Hess et al. (16) demonstrated that AKI was associated with increased major adverse limb events including cardiovascular and all-cause rehospitalization. Even a modest increase in serum creatinine levels is associated with increased mortality rates, length of stay, and cost after adjusting for comorbidities (17). The results of the present study confirm worse outcomes in patients with AKI in the NCDR PVI Registry, most notably a 10-fold increase in in-hospital mortality. Also of note, pre-existing CKD among CLI patients leads to more AKI and decreased survival (18). This report confirms the findings of Grossman et al. (6) who found that diabetes, heart failure, anemia, baseline renal dysfunction, and CLI predict contrast-induced nephropathy. Prasad et al. (5) reported an association of AKI with age that was not evident in our data, and also found diabetes, CKD, and heart failure to predict AKI as demonstrated in the current report. Others found that baseline proteinuria predicted AKI, something that could not be tested in our dataset (19). In the current study, baseline hypertension independently predicted AKI, a finding that was not present in prior reports.

One approach to mitigating AKI is to consider other therapies for symptomatic PAD. For example, PVI causes more AKI than open surgical treatment of PAD in comparable patients (20) and is associated with more renal dysfunction after 1 year than supervised exercise therapy (21). It is imperative to consider the relative risks and benefits of all treatment options before making recommendations, particularly in patients at higher risk for AKI. Although alternative therapies for patients presenting with claudication may have a more favorable risk/benefit ratio, those presenting with ALI or CLI have a different goal of therapy (limb salvage) that must be considered when treatment decisions are made.

The potential clinical impact of the current report lies in avoidance of AKI following PVI. In coronary intervention, the authors (22) suggest that using “as low as reasonably achievable” contrast dosing neutralized the impact of diabetes mellitus and CKD on AKI development after PCI. This suggests that reducing the contrast dose may improve outcomes. Prior analyses suggest that a contrast dose >3 times the calculated creatinine clearance is associated with a significantly increased risk of contrast-induced nephropathy (6). Other possible interventions include hydration protocols (23,24), use of low-osmolar contrast (25), contrast-sparing strategies such as dose modulation and staging of additional procedures (26,27) or use of digital subtraction angiography at low frame rates with contrast dilution, and preprocedural treatment with statins (28–30). Unfortunately, information on these potential interventions is limited in the registry. Another possibility is identification of novel, noniodinated contrast agents. For example, carbon dioxide angiography decreases the incidence of AKI (31). Unfortunately, it also increases nonrenal events such as limb pain, nausea, and vomiting, limiting its usefulness. Gadolinium is also under investigation as an alternative endovascular contrast agent (32), though its utility is limited by concern for systemic sclerosis in patients with renal insufficiency (33).

The predictive model was developed for AKI using pre-procedural variables. This allows more accurate review of procedural risks with patients, a more informed consent, and potential tailoring of treatment to individual patients’ risk. Similar preprocedural risk scores were developed for PCI with varying degrees of success and utilization (2,3,34). The ability to incorporate individual risk factors into a personalized risk score allows a global risk assessment that may better inform decisions regarding interventions designed to decrease AKI risk. To develop the risk model variables are assigned weights according to their beta coefficient, or the strength of their association with the outcome. Similar methods were used in the development of the post-PCI scores mentioned above. Those models include a wider variety of clinical presentations such as acute coronary syndromes, ST-segment elevation myocardial infarction, cardiogenic shock, and cardiac arrest. Acute events such as these are uncommon in PVI and did not merit inclusion here.

Knowing the risks of an individual patient is 1 step in a systematic approach to reducing AKI. Developing protocols and monitoring physician adherence to these protocols is the next step in implementing an effective quality improvement strategy. It is reasonable initially to focus on high-risk patients (e.g., AKI risk >10%), for whom the number needed to treat with preventative measures would be lowest. It may be beneficial to explicitly define recommended preventative strategies and provide education to gain physician buy-in to the approach (35). For example, previously published recommendations for renal function–based dosing of contrast during PCI (36) could be extended to PVI, or further research with this model to define safe contrast limits could be pursued. Similar recommendations deserve further exploration for intravenous volume expansion (23,24), statin dosing (28–30,37), and contrast-sparing interventional equipment (26,38). Prevention of post-PVI bleeding may also mitigate risk, because bleeding is significantly more common with AKI and is associated with substantial mortality risk (39). It may be the cumulative total of several, if not all, of these strategies that leads to decreasing AKI and improving outcomes after PVI.

STUDY LIMITATIONS.

First, we are subject to the limitations of the registry itself—notably, that post-PVI creatinine measurements 6 to 24 h after the procedure are reported. A longer window up to 72 h may have altered the results. If renal dysfunction developed after discharge, it would be unrecognized and therefore not included. Second, there was a higher than expected rate of patients with CLI included in our cohort, likely due to exclusion of stable patients who were discharged on the day of PVI. This potentially limits applicability of our results to stable outpatient procedures, although over 5,000 procedures performed without ALI or CLI were included. Pre-procedural creatinine levels may vary, and values in this study may not be reflective of true baseline levels, particularly because peri-procedural hydration is recommended in many patients. Potential other nephrotoxic agents may have been used but not recorded in the database (such as antibiotics for CLI). Additionally, as with any observational, nonrandomized study, unrecognized biases may affect study results.

CONCLUSIONS

This analysis of over 10,000 procedures included in the NCDR PVI Registry confirms that AKI is relatively common after PVI, occurring in over 7% of procedures. Mortality is increased 10-fold in association with AKI. We identified several variables associated with AKI and used them to develop the NCDR PVI AKI risk model—a practical integer scoring system to predict patient risk of AKI. This clinical tool can be used to identify patients for preventative strategies to reduce the risk of AKI and should also foster further study focused on decreasing the incidence of AKI following PVI.

PERSPECTIVES.

WHAT IS KNOWN?

Acute kidney injury following iodinated contrast exposure is known to be associated with adverse outcomes.

WHAT IS NEW?

This report details the assessment of acute kidney injury after in-patient peripheral vascular intervention in the NCDR PVI Registry and proposes an integer scoring system to predict risk.

WHAT IS NEXT?

Future studies of the application of the risk score and the effects of renal protective efforts may help improve outcomes following inpatient peripheral vascular intervention.

ABBREVIATIONS AND ACRONYMS

AKI

acute kidney injury

AKIN

Acute Kidney Injury Network

ALI

acute limb ischemia

CAD

coronary artery disease

CKD

chronic kidney disease

CLI

critical limb ischemia

eGFR

estimated glomerular filtration rate

IQR

interquartile range

PAD

peripheral artery disease

PCI

percutaneous coronary intervention

PVI

peripheral vascular intervention