Transcatheter or surgical aortic valve replacement in patients with advanced kidney disease: A propensity score–matched analysis

Abstract

Background

Transcatheter aortic valve replacement (TAVR) is an alternative for surgically inoperable patients with severe aortic stenosis. Advanced kidney disease may significantly affect outcomes in patients treated with TAVR and surgical aortic valve replacement (SAVR).

Hypothesis

TAVR is associated with better in‐hospital outcomes compared with SAVR in patients with advanced kidney disease.

Methods

We identified our sample from the National Inpatient Sample between 2012 and 2014, using International Classification of Diseases, Ninth Revision, Clinical Modification codes. We included patients with chronic kidney disease stages IV and V and end‐stage renal disease as advanced kidney disease patients. We excluded patients with acute kidney injury on admission and patients on dialysis.

Results

After propensity matching, 2485 patients were included in each group. The primary outcome of in‐hospital mortality (12.9% vs 6.2%; P < 0.01) was higher with SAVR as compared with TAVR. Patients who underwent SAVR reported higher acute kidney injury (50.3% vs 33%; P < 0.01) and dialysis requirements (26.8% vs 20.1%; P < 0.01). Other secondary outcomes including blood transfusion, atrial fibrillation, iatrogenic cardiac complications, pericardial complications, perioperative stroke, perioperative infections, and postoperative shock were more common with SAVR. With SAVR, the length of hospitalization and hospitalization costs were significantly higher; however, permanent pacemaker placement was more common with TAVR compared with SAVR.

Conclusions

In patients with advanced kidney disease, SAVR was associated with higher mortality and higher periprocedural complications, as compared with TAVR. Thus, benefits of TAVR could be extended in patients with advanced kidney disease who cannot undergo surgery.

1. INTRODUCTION

Transcatheter aortic valve replacement (TAVR) is now a safe and feasible option for the treatment of severe aortic stenosis (AS) in high‐risk as well as intermediate‐risk patients.1, 2 There has been rapid growth in the use of TAVR for the treatment of severe AS after its initial approval from the US Food and Drug Administration.1 Patients with preprocedural advanced kidney disease showed an increase in operative risk for the TAVR and surgical aortic valve replacement (SAVR) procedures.3, 4, 5 Also, in patients with advanced kidney disease, aortic‐valve calcification occurs earlier and progresses more rapidly.6 More than 19 million patients in the United States are estimated to suffer from chronic kidney disease (CKD) or end‐stage renal disease (ESRD).7 The epidemic of CKD has been rising in recent years.8 The Placement of Aortic Transcatheter Valve Trial (PARTNER) 1 study showed renal dysfunction to be a significant predictor of mortality in patients undergoing TAVR.5 Previous studies showed impact of preoperative acute kidney injury (AKI), CKD, or ESRD on patients undergoing TAVR or SAVR.9, 10, 11 According to Iung et al., approximately a third of the patients were refused by the surgeons based on the severity of the kidney disease.11 Moreover, major clinical trials have excluded patients with severe renal insufficiency from the trials.1 However, previous small studies showed TAVR to be an effective procedure in patients with CKD or ESRD.12

There is a paucity of data comparing TAVR and SAVR in advanced kidney disease patients with severe AS and large number of “real‐world” cohorts. The main purpose of this study was to evaluate the in‐hospital outcomes in real‐world cohorts with advanced kidney disease patients undergoing TAVR or SAVR with severe AS.

2. METHODS

2.1. Study design and patient population

Our retrospective, observational cohort study's population is derived from the National Inpatient Sample (NIS) database.13 The NIS database has been described previously.14 Briefly, NIS is a part of the Healthcare Cost and Utilization Project (HCUP), which is sponsored by the Agency for Healthcare Research and Quality (AHRQ). NIS includes >1000 hospitals around the United States, which is nearly a 95% representation of the US population. The NIS database is assessed annually to maintain its internal validity.

We identified our study cohorts undergoing TAVR and SAVR by analyzing the NIS data from 2012 to 2014, using International Classification of Diseases, Ninth Revision, Clinical Modification (ICD‐9‐CM) procedure codes. We included implantation of a bioprosthetic (code 35.21) or mechanical (code 35.22) replacement of the aortic valve in the SAVR group and a transfemoral replacement (code 35.05) or transapical replacement of the aortic valve (code 35.06) in the TAVR group. As TAVR was approved in late 2011, we decided to include patients from 2012 onward. Data after 2014 are not available. Afterward, patients age < 18 years were excluded. Using appropriate ICD‐9‐CM diagnosis codes, patients having primary AKI (code 584.xx) on admission were excluded. Those patients on hemodialysis or peritoneal dialysis (codes 54.98 and 39.95, respectively) prior to the admission were excluded. We then excluded those patients who were found to have TAVR and SAVR procedures in the same admission. From the remaining hospitalizations, we included the patients with CKD stages IV (585.4), V (585.5) and ESRD (585.6). (See Supporting Information, Figure 1, in the online version of this article.) Stage IV kidney disease patients have an estimated glomerular filtration rate between 29 and 15 mL/min/1.73 m², stage V, less than 15 mL/min/1.73 m² and ESRD, irreversible kidney damage. We used Elixhauser comorbidities in our study.16 Additional comorbidities were included using ICD‐9‐CM codes (see Supporting Information, Table 1, in the online version of this article). Deyo modification of the Charlson Comorbidity Index was used to define the severity of comorbid conditions (see Supporting Information, Table 2, in the online version of this article).17

Table 1.

Demographics, baseline characteristics and In‐hospital outcomes stratified by TAVR and SAVR (unmatched cohorts)

Variable	SAVR, n = 7855	TAVR, n = 3350	P Value
Age, y	66.3 ± 13.1	78.5 ± 9.2	<0.01
Sex
M	66.5	58.7	<0.01
F	33.5	41.3
Racea
White	60.2	72.8	<0.01
Black	19	9.1
Other	20.8	18.1
Elixhauser comorbidities
DM	23.5	29.7	<0.01
HTN	85	82.8	<0.01
CHF	6.9	14.6	<0.01
Chronic pulmonary disease	22.9	30.7	<0.01
Coagulation disorders	39.6	28.2	<0.01
PVD	21.7	33.1	<0.01
Obesity	19.9	14.2	<0.01
Pulmonary circulation disorders	3.1	6.1	<0.01
Smoking	22.9	23.6	0.44
Mitral stenosis	2.9	2.5	0.33
Mitral insufficiency	10.1	11	0.14
PCI (same admission)	0.8	4.9	<0.001
CABG (same admission)	38.7	0.3	<0.001
Percutaneous mitral valve repair (same admission)	0	0	NA
Surgical mitral valve repair/replacement (same admission)	16.6	0.3	<0.001
In‐hospital outcomes
In‐hospital mortality	11.5	6.9	<0.001
AKI	40.5	35.5	<0.001
Dialysis requirement	27.9	24	<0.001
Blood transfusion	50	37.8	<0.001
Vascular complications requiring surgery	4.3	4.6	0.39
PPM requirement	10.4	28.1	<0.001
AF	42.3	45.5	0.002
Iatrogenic cardiac complications	12.5	10.1	<0.001
Pericardial complications	3	1.2	<0.001
Perioperative stroke	1.3	1.2	0.54
Perioperative infections	2	0.6	<0.001
Postoperative shock	5.1	1.9	<0.001
LOS, d	14 (9–24)	8 (5–15)	<0.001
Mean cost, US$	90 966	69 983	<0.001
Disposition
Home	20.2	26	<0.001
Transfer to short‐term hospital	1.5	1
Transfer to other facilityb	42.7	34.8
Home healthcare	23.9	31.3

Abbreviations: AF, atrial fibrillation; AKI, acute kidney injury; CABG, coronary artery bypass grafting; CHF, congestive heart failure; DM, diabetes mellitus; F, female; HTN, hypertension; IQR, interquartile range; LOS, length of stay; M, male; NA, not applicable; PCI, percutaneous coronary intervention; PPM, permanent pacemaker; PVD, peripheral vascular disease; SAVR, surgical aortic valve replacement; SD, standard deviation; TAVR, transcatheter aortic valve replacement; US, United States.

Frequencies presented as %, mean ± SD, or median (IQR).

^aMissing observations: 720.

^bIncludes skilled‐nursing facility, intermediate‐care facility, and other types of facilities.

Table 2.

Baseline characteristics stratified by TAVR and SAVR procedures (propensity score matched analysis)

Variable	SAVR, n = 2485	TAVR, n = 2485	P Value	Standardized Difference, %
Age, y	75.9 ± 8.7	76.0 ± 9.2	0.95	1.1
Sex
M	62.8	64.8	0.14	2.9
F	37.2	35.2
Racea
White	71.2	70.8	0.72	4.0
Black	10.5	10.1
Other	18.3	19.1
Elixhauser comorbidities
DM	31.4	31.9	0.54	5.2
HTN	82.3	82.7	0.41	3.8
CHF	9.9	9.3	0.47	4.2
Chronic pulmonary disease	28.6	27	0.20	2.7
Coagulation disorders	33.2	33.2	1.00	0
PVD	27.8	28.8	0.43	1.3
Obesity	17.3	17.7	0.71	5.7
Pulmonary circulation disorders	4.4	3.4	0.07	7.4
Smoking	23.5	23.5	1.00	0
Mitral stenosis	3.6	2.8	0.11	6.6
Mitral insufficiency	12.1	10.9	0.18	5.6

Abbreviations: CHF, congestive heart failure; DM, diabetes mellitus; F, female; HTN, hypertension; M, male; PVD, peripheral vascular disease; SAVR, surgical aortic valve replacement; SD, standard deviation; TAVR, transcatheter aortic valve replacement.

Frequencies presented as % or mean ± SD.

^aMissing observations: 720.

The primary endpoint of our study was all‐cause in‐hospital mortality. Secondary in‐hospital outcomes included individual AKI, dialysis requirement, blood transfusion, vascular complications requiring surgery, permanent pacemaker (PPM) requirement, atrial fibrillation (AF), iatrogenic cardiac complications, pericardial complications (included hemopericardium, cardiac tamponade, and pericardiocentesis), perioperative stroke, perioperative infections, postoperative shock, and disposition (see Supporting Information, Table 1, in the online version of this article). We used secondary diagnosis or procedural codes to identify in‐hospital outcomes. A similar selection process for outcomes has been used previously.14 Additionally, we included length of hospitalization stay and mean cost of hospitalization. To estimate the cost, NIS data were merged with cost‐to‐charge ratio files available from HCUP. We estimated the final cost by multiplying total hospital charge with the cost‐to‐charge ratio.

2.2. Statistical analysis

We used SAS version 9.4 (SAS Institute, Inc., Cary, NC) for our analysis. Continuous variables were presented as mean ± SD, for which we performed the Student t test. Categorical variables were presented as frequencies in percentage, for which we performed a χ2 test. We constructed a propensity score–matched18 analysis to reduce selection bias and to reduce baseline differences between groups. First, a logistic regression model with forward selection was performed to calculate a propensity score for each individual patient. In this logistic regression model, we included age, sex, race, all Elixhauser comorbidities, and other comorbidities. Then, we matched all patients based on their propensity score using a 1:1 scheme without replacement using the nearest‐neighbor matching method. In the SAVR group, 31.6% patients were matched, and 74.2% in the TAVR group. Standardized difference < 10% is acceptable for clinical studies after propensity score matching.19 Those patients who did not match were excluded from the analysis. Comparisons of in‐hospital outcomes were performed in matched cohorts using the McNemar test or paired t test, as appropriate. This method has been used earlier.20

3. RESULTS

3.1. Comparison of baseline characteristics prior to propensity score matching

A total of 11 205 (weighted) patients with advanced CKD underwent aortic valve replacement between 2012 and 2014. Of these, 7855 patients were included in the SAVR group and 3350 patients were included in the TAVR group. Baseline differences existed between the groups. The patients presented at a higher age for TAVR than SAVR (78.5 vs 66.3 years; P < 0.01). Regarding the patient population, there was a higher proportion of males within the groups. Before matching, diabetes mellitus, heart failure, chronic pulmonary disease, peripheral vascular disease, and pulmonary circulation disorders were higher in the TAVR group. In unmatched population, in‐hospital mortality, AKI, dialysis requirement, blood transfusion, iatrogenic cardiac complications, pericardial complications, perioperative infection, and postoperative shock were higher in the SAVR group. Although PPM rates and AF rates were higher in the TAVR group, higher in‐hospital outcomes in the SAVR group translated into overall longer length of stay and mean cost of hospitalization (Table 1).

3.2. Comparison of in‐hospital outcomes after propensity score matching

After performing propensity score matching, 2485 patients were matched in each group. Characteristics of propensity matched variables are shown in Table 2. We demonstrated standardized difference between matched variables in the same table (Table 2). In‐hospital mortality was higher with SAVR compared with TAVR (12.9% vs 6.2%, respectively; P < 0.01). AKI (50.3% vs 33%; P < 0.01) and dialysis requirements (26.8% vs 20.1%; P < 0.01) were higher with the SAVR group as well. Other secondary outcomes were higher with SAVR compared with TAVR, including transfusion (49.7% vs 38.2%, respectively; P < 0.01), AF (50.3% vs 41.8%; P < 0.01), iatrogenic cardiac complications (14.5% vs 10.5%; P < 0.01), pericardial complications (2.2% vs 1%; P < 0.01), perioperative stroke (1.6% vs 0.8%; P < 0.01), perioperative infection (2.2% vs 0.2%; P < 0.01), and postoperative shock (4.6% vs 2%; P < 0.01). Interestingly, a PPM placement with TAVR compared with SAVR (9.3% vs 27.8%; P < 0.01) showed higher rates of implantation. The median length of stay was higher with SAVR (14 vs 8 days; P < 0.01). Additionally, the mean cost of hospitalization was higher with SAVR ($62 295 vs $58 927; P < 0.01). More patients were discharged home after TAVR compared with SAVR (28.4% vs 10.1%; P < 0.01; Table 3).

Table 3.

In‐hospital outcomes stratified by TAVR and SAVR procedure (propensity score matched cohorts)

Variable	SAVR, n = 2485	TAVR, n = 2485	P Value
In‐hospital mortality	12.9	6.2	<0.01
AKI	50.3	33	<0.01
Dialysis requirement	26.8	20.1	<0.01
Blood transfusion	49.7	38.2	<0.01
Vascular complications requiring surgery	3.4	4.4	0.07
PPM requirement	9.3	27.8	<0.01
AF	50.3	41.8	<0.01
Iatrogenic cardiac complications	14.5	10.5	<0.01
Pericardial complications	2.2	1	<0.01
Perioperative stroke	1.6	0.8	<0.01
Perioperative infections	2.2	0.2	<0.01
Postoperative shock	4.6	2	<0.01
LOS, da	14 (0–28)	8 (0–17)	<0.01
Mean cost, US$	62 295	58 927	<0.01
Disposition
Home	10.1	28.4	<0.01
Transfer to short‐term hospital	1.8	1.2
Transfer to other facilityb	51.1	32.8
Home healthcare	23.7	31.4

Abbreviations: AF, atrial fibrillation; AKI, acute kidney injury; IQR, interquartile range; LOS, length of stay; PPM, permanent pacemaker; SAVR, surgical aortic valve replacement; TAVR, transcatheter aortic valve replacement; US, United States.

Frequencies presented as % or median (IQR).

^aDay 0 represents discharge on the same day or within 24 h.

^bIncludes skilled‐nursing facility, intermediate‐care facility, and other types of facilities.

4. DISCUSSION

In this nationally representative observational study, we compared TAVR and SAVR in patients with advanced kidney disease and severe AS. First, we demonstrated that TAVR has less in‐hospital mortality. Additionally, secondary outcomes such as AKI, dialysis requirement, transfusion, AF, iatrogenic cardiac complications, pericardial complications, perioperative stroke and infection, and postoperative shock were lower with TAVR when compared with SAVR. Fewer in‐hospital outcomes with TAVR translated into reduced length of hospitalization stay and mean cost of hospitalization compared with SAVR in advanced kidney disease patients. However, we found higher PPM placement rates with TAVR when compared to SAVR.

TAVR is shown to be safe compared with SAVR in high‐risk patients21; however, there is a paucity of data comparing TAVR and SAVR in advanced kidney disease patients. Our result showed that benefits of TAVR could be extended in patients with advanced kidney disease who cannot undergo surgery. Overall mortality in patients with advanced CKD is higher with TAVR as well as SAVR compared with previous studies with normal to mildly impaired kidney functions.12, 22, 23, 24 This is observed in our study as well. A few recent studies showed that severity of CKD is associated with poorer outcomes,25, 26 and hence risk stratification prior to the procedure is important.27 However, one of the previous studies showed that worsening renal dysfunction is not associated with poor outcomes with TAVR when compared with SAVR.12 It is still unclear how renal dysfunction affects TAVR and SAVR outcomes.

Advanced kidney disease may affect platelet functions, which may eventually translate to higher incidence of hemorrhagic events.28 This could be the reason for overall higher rates of blood transfusion with TAVR as well as SAVR. However, transfusion rates were lower with TAVR, as it is a less invasive procedure. This result was supported by prior study that showed lower overall transfusion rates with TAVR as compared with SAVR in patients with advanced kidney disease.12 A study by D'Errigo et al29 showed higher rates of AKI with SAVR compared with TAVR. In contrast, one of the prior studies showed no difference with postprocedural AKI when comparing TAVR with SAVR (29.7% vs 24.1%; P = 0.21).30 Their study was limited, as there were only 195 patients in each arm after performing propensity score matching; our study has 2485 patients in each arm after performing propensity score matching. We showed higher AKI with SAVR when compared with TAVR. Measures to reduce the contrast agent and the duration of cardiopulmonary bypass may decrease the risk of AKI.31, 32 Additionally, appropriate strategy should be developed to prevent further AKI in such patients with collaboration from a nephrologist when performing TAVR or SAVR in patients with advanced kidney disease.33 In support of one previous study, perioperative stroke rates were higher with SAVR in our study.29 Prior studies showed shorter hospitalization stay with TAVR compared with SAVR.22 We noticed significantly shorter length of hospitalization with TAVR when compared with SAVR, even with advanced kidney disease. Additionally, the overall higher in‐hospital outcomes seen with SAVR may translate into longer length of stay and higher mean cost of hospitalization24 in our study. In our study, PPM placement rates were higher with TAVR compared with SAVR. This could be due to a higher risk of atrioventricular blockage seen earlier with TAVR.29

4.1. Study limitations

We recognize several limitations in our study. First, we may have biases inherent with any retrospective study, such as a coding error or missing data. Second, we do not have information on confounders such as valve type, defect size, radiation‐dose exposure, contrast volume, preexisting anatomic abnormalities, medication use, and concomitant procedures performed. Inclusion of such variables may show a different outcome. We do not have variables for any scoring, such as Valve Academic Research Consortium (VARC 2) for bleeding or the Society of Thoracic Surgeons (STS) score or European System for Cardiac Operative Risk Evaluation (EuroSCORE) for severity, as we do not control variables in our study. Additionally, we do not have information on whether the patient had AF34 or carotid stenosis prior to the procedure, which can significantly affect our outcomes with TAVR. Third, we do not have information on long‐term follow‐ups because of the nature of our database. As shown in our study, surgery may have more short‐term harmful effects; however, surgery may have comparable or better long‐term outcomes when compared with TAVR. Fourth, our study contains data collected up to 2014 only. Recent advancements in technique and technology (eg, new valve, direct aortic approach) may have different outcomes and may show further benefits with TAVR. Fifth, a major concern with TAVR is paravalvular leak; even the smallest leak could be associated with poorer outcomes.35, 36 We do not have information on paravalvular leak because of the nature of the database. Finally, we do not have information regarding whether these procedures were performed primarily or as redo procedures, as we do not control collected information. Despite these limitations, ICD‐9 codes are too specific for the given condition. Additionally, access to a large database such as HCUP‐NIS may aid in giving important information about a plethora of “real‐world” advanced kidney disease patients in relatively new procedure such as TAVR.

5. CONCLUSION

We noticed lower mortality and morbidity with the TAVR procedure in advanced kidney disease patients when compared with SAVR. TAVR has been shown to benefit a wider group of patients, now including those with advanced kidney disease. Hence, TAVR can safely be considered as a safe and effective alternative for high‐risk patients with advanced kidney disease who cannot undergo surgery. This is hypothesis‐generative only, and further research is required for more information on the comparison of TAVR and SAVR in low‐ to intermediate‐risk patients and an optimal approach strategy for TAVR in patients with advanced kidney disease.

Conflicts of interest

The authors declare no potential conflicts of interest.

Supporting information

Appendix S1 Supplementary Material

Supplementary Table 1: International Classification of Diseases, Ninth Revision, Clinical Modification (ICD 9 CM) codes used in our analysis

Supplementary Table 2: Deyo's Modification of Charlson's Co‐morbidity Index (CCI)

Supplementary Figure 1 Flow Chart for Patient Selection

Doshi R, Shah J, Patel V, Jauhar V, Meraj P. Transcatheter or surgical aortic valve replacement in patients with advanced kidney disease: A propensity score–matched analysis. Clin Cardiol. 2017;40:1156–1162. 10.1002/clc.22806