African American-Caucasian American Differences in Aortic Valve Replacement in Patients with Severe Aortic Stenosis

Abstract

BACKGROUND

Among patients with severe aortic stenosis (AS), there are limited data on aortic valve replacement (AVR), reasons for non-receipt and mortality by race.

METHODS

Utilizing the Duke Echocardiography Laboratory Database, we analyzed data from 110,711 patients who underwent echocardiography at Duke University Medical Center between 1999–2013. We identified 1,111 patients with severe AS who met ≥1 of 3 criteria for AVR: ejection fraction ≤50%, diagnosis of heart failure, or need for coronary artery bypass surgery. Logistic regression models were used to assess the association between race, AVR and 1-year mortality. Chi-squared testing was used to assess potential racial differences in reasons for AVR non-receipt.

RESULTS

Among the 1,111 patients (143 AA and 968 CA) eligible for AVR, AA were more often women, had more diabetes, renal insufficiency, aortic regurgitation and left ventricular hypertrophy. CA were more often smokers, had more ischemic heart disease, hyperlipidemia and higher median income levels. There were no racial differences in surgical risk utilizing logistic euroSCORES. Relative to CA, AA had lower rates of AVR (aOR 0.46, 95% CI 0.3–0.71, p<0.001) yet similar 1-year mortality (aHR 0.81, 95% CI 0.57–1.17, p=0.262). There were no significant differences in reasons for AVR non-receipt.

CONCLUSION

Among patients with severe AS eligible for AVR, AA patients were less likely to undergo AVR. Despite racial differences in AVR, there were no significant differences in mortality within 1-year or reasons for AVR non-receipt.

CONDENSED ABSTRACT

Between 1999–2013 we identified 1,111 patients (143 AA and 968 CA) with severe AS who met prespecified criteria for AVR. AA relative to CA were more often women, had more diabetes, renal insufficiency, and left ventricular hypertrophy, however had less tobacco use, ischemic heart disease, hyperlipidemia and lower median income levels. Among patients with severe AS, AA relative to CA had lower rates of AVR (aOR 0.46, 95% CI 0.3–0.71, p<0.001) without significant differences in reasons for AVR non-receipt and similar 1-year mortality.

Introduction

Aortic stenosis (AS) is among the most common valvular heart diseases in developing nations, second only to mitral regurgitation.^1–2 Aortic stenosis has an estimated prevalence of 2–7% for persons 65 years of age and older.³ Studies suggest that severe AS disproportionately affects Caucasian-Americans (CA) relative to African Americans (AA) with a ratio of 3:1.⁴ While CA have a higher prevalence of severe AS compared with AA, AA have more traditional risk factors associated with the development of AS including hypertension, hyperlipidemia, diabetes mellitus (type II), metabolic syndrome, smoking, and renal failure.⁵ In addition, AA with severe AS may suffer greater morbidity and mortality relative to other racial and ethnic groups.^6–7 This paradoxical observation whereby AA cluster higher AS risk factors relative to CA, yet display a lower burden of disease and have worse outcomes has been observed in other cardiovascular diseases, most notably atrial fibrillation.⁸

Progressive severe AS has low survival rates (50–60%) in the subsequent 1–2 years after diagnosis.⁹ The mainstay of treatment has been surgical aortic valve replacement (SAVR)¹⁰ however, in the last decade, transcatheter aortic valve replacement (TAVR) has evolved as a treatment for moderate/high-risk surgical candidates and recently has been extended to low risk populations.^11–13

Few studies have analyzed rates of aortic valve replacement (AVR) in individuals with severe AS and associated outcomes as a function of race. We sought to address the following: 1) assess rates of AVR by AA and CA race among patients with severe AS; 2) determine reasons for AVR non-receipt by race among indicated patients; 3) evaluate 1-year all-cause mortality by AA and CA race among patients with severe AS; 4) determine the effect of AVR on 1-year mortality post-AVR in AA and CA individuals.

Methods

Data Capture and Study Population

Adult patients (age >18 years) were identified from the Duke Echocardiography Laboratory Database (DELD) (1999–2013). Details of the DELD have been previously described and will be briefly summarized.¹⁴ The DELD includes a digital archive of all echocardiograms performed at Duke University Hospital (DUH) and its satellite clinics. The database has been prospectively maintained from 1995 to 2015 and includes clinical information drawn from a variety of sources. Follow-up information for DELD patients was obtained from medical records, with events outside Duke ascertained by cross-linking with patients in the Duke Databank for Cardiovascular Disease (DDCD) and with a query of the National Death Index to establish vital status for patients in whom this was not known.^15–18

For inclusion in this analysis AS was defined by index echocardiographic indices, valve area <1cm² or mean gradient >40mmHg. Additionally, to be considered eligible for AVR, individuals had to have severe AS and ≥ 1 of 3 criteria, 1) left ventricular ejection fraction (LVEF) ≤50%, 2) diagnosis of heart failure (as documented by ICD-9 codes; 428.0–428.99), or 3) need for coronary artery bypass surgery (CABG).¹⁹ Patients indicated for CABG included those with left main coronary disease or three-vessel coronary disease by coronary angiography prior to or within 30 days of echo. These criteria for AVR were based on 2014 American Heart Association/American College of Cardiology (AHA/ACC) valvular heart disease guidelines.¹⁰ Race was self-reported with the designation Caucasian-American (includes European-American or White) and African-American (includes Black-American or Black). Median household income was obtained from the American Community Survey (2011) and linked to individual patients at the census block group level using current patient addresses. Patients with prior valve surgery, endocarditis, race not CA or AA, recent LVAD (prior to or within 30 days of echocardiogram), solid organ or bone marrow transplant (prior to or within 30 days), metastatic cancer and cirrhosis were excluded. For each patient, the baseline time point was defined as the date of the first echocardiogram procedure showing evidence of severe AS. Information on follow-up procedures and events was collected through December 31, 2014.

Medical records were manually reviewed for clinical presentation, comorbidities and justification for AVR non-receipt. Justification was determined via review of available clinical documentation in the Duke University Medical System electronic medical record.

Study Endpoints

The following outcomes were evaluated as a function of race: AVR receipt within 1 year, reasons for AVR non-receipt, all-cause mortality within 1 year of the diagnosis of AS from the index echocardiogram, and the impact of AVR on all-cause mortality through 1 year of AVR receipt. Aortic valve replacement was defined as surgical or transcatheter AVR at Duke University Health System within 1 year of the index echocardiogram date revealing severe AS. All-cause mortality was determined based on death within 1 year of index echocardiogram that revealed AS irrespective of AVR status. Next, to assess the impact of AVR on the mortality outcome, all-cause mortality was evaluated within 1 year of the patient’s first AVR procedure. Individuals not receiving AVR were followed from the median time-to-AVR among those who did receive AVR to ensure the assessment for mortality between AVR and non-AVR patients would be during the same time window to minimize the effect of immortal time bias. Patients in the AVR non-receipt group who died prior to this time point were excluded. Mortality follow-up was censored at 1 year or December 31, 2014, for deaths after 2014.

Statistical Methods

Patient Characteristics

To determine the distribution of severe AS in our population, the raw frequency and percentages of patients with severe AS were determined by race group. Patient baseline characteristics of those eligible for AVR were compiled for both racial groups with continuous variables summarized with median, 25^th, & 75^th percentiles whereas binary and categorical variables were summarized with frequencies and percentages. For continuous and ordinal variables, Wilcoxon’s Rank Sum test was conducted to test for differences in characteristics among AA and CA patients eligible for AVR. For categorical variables, percentages were compared using Chi-Square or Fisher’s Exact test, where appropriate. We used complete case analysis for all descriptive analyses.

Dependent variables for regression modeling were chosen based on clinical relevance in the decision for AVR and subsequent outcomes.^20–21 Adjustment variables included age, end-stage renal disease (defined by ICD-9 code 585.6 prior to or within 30 days of index echocardiogram), hypertension, sex, diabetes, eGFR, 3-vessel/left-main disease, ischemic heart disease, prior CABG, aortic regurgitation, mitral regurgitation, hyperlipidemia, smoking, logistic euroSCORE and median household income. The logistic euroSCORE is a risk prediction algorithm for cardiac surgery, which incorporates 17 risk factors.²²

The extent of missing adjustment covariates was low, except for eGFR, which had 15% missingness. To account for missingness, we used multiple imputation on all adjustment variables. We used 50 imputed datasets with fully conditional specification models, using logistic regression for categorical variables and regression for continuous variables.

AVR Utilization

The cumulative incidence of AVR within 1-year of baseline echocardiogram was estimated by race. To examine the association between race and AVR through 1-year, unadjusted and adjusted logistic regression models were fit, and odds ratios and 95% confidence intervals were estimated comparing odds of AVR in AA versus CA patients. In this analysis of AVR, patients who died within 1-year without AVR were included in the analysis as a non-event. In these models, we assessed linearity of the continuous predictors and multicollinearity for all predictors. We examined both SAVR and TAVR in this analysis. Given TAVR was not available at Duke University Medical Center until 4/11/2011, there were only 418 patients out of the total 1,111 that were potentially eligible for TAVR based on the time period of assessment (40 AA, 378 CA).

Reasons for AVR non-receipt

Reasons for AVR non-receipt were identified, and grouped into 5 distinct categories (patient refusal, high operative risk, inadequate evaluation, severe AS reclassified and death). “Patient refusal” category was defined as any instance in which provider documentation cited the patient’s (or primary caregiver’s) refusal to undergo an AVR procedure. “High operative risk” category was defined as any instance in which provider documentation cited or referred to high operative risk as a concern if a patient were to undergo AVR or was otherwise noted by the provider to have a short life expectancy. “Inadequate evaluation” category was defined as any instance in which a patient had no encounter with a cardiovascular specialist (cardiologist or cardiothoracic surgeon) or an instance in which a patient was lost to follow-up. The “not symptomatic severe AS” category was defined as any instance in which provider notes suggested the patient had symptoms not clearly associated with AS, was asymptomatic, or had severe AS reclassified after further evaluation. Patients with severe AS reclassified after further evaluation were either reclassified based on heart catherization data or reevaluation of echo data by a cardiologist. The“death”category was defined to include any patient that died within 30 days of index echocardiogram or died prior to a planned AVR procedure. Chi-squared testing was performed to identify differences in the distribution of reasons for AVR non-receipt between the two racial groups.

1- Year Mortality

Unadjusted and adjusted Cox models were fit to assess the association between race and death through 1 year of baseline echocardiogram. Hazard ratios and 95% confidence intervals were reported for comparing the relative hazard of death in AA versus CA patients during 1-year of follow-up. Continuous covariates were checked for linearity and flexible transformations were used for nonlinearity. Proportional hazards assumption was checked and not violated. Lastly, a Cox model was used to assess the effect of AVR on 1-year mortality post-AVR in each race group. For patients who received AVR, time zero was defined as time of AVR. For patients who did not receive AVR, time zero was defined as the median AVR time (19.5 days). Patients who did not receive AVR but died before median AVR date were excluded from the analysis (N = 62). An interaction term was added to the Cox model to determine if the hazard of mortality in patients receiving AVR versus no AVR differed by race.

This study was approved by the Duke Institutional Review Board. All P values presented are two-sided. All statistical analyses were performed at the Duke Clinical Research Institute using SAS software (Version 9.4, SAS Institute, Cary, NC). The authors are solely responsible for the design and conduct of this study, all study analyses, the drafting and editing of the paper and its final contents. In part, support was provided by the Duke Center for Research to Advance Healthcare Equity (REACH Equity), which is funded by the National Institute on Minority Health and Health Disparities under award number U54MD012530.

Results

Baseline Characteristics

We identified 1,848/110,711 (1.6%) patients within the DELD database with severe AS of which 236/1,848 (12.8%) were AA and 1,612/1,848 (87.2%) were CA. Among patients with severe AS, 1,111 met pre-specified criteria for AVR (143 AA and 968 CA) (Figure 1). African-American individuals were more often female (66.4% vs 43.1%, p<0.001), had lower median household incomes ($38,529 vs $47,600, p<0.001), were more likely to have diabetes (45.5% vs 31.4%, p<0.001), hypertension (84.6% vs 75.7%, p=0.019), and end stage renal disease (8.4% vs 1.1%, p<0.001). Caucasian-American individuals had more ischemic heart disease (74% vs 61.5%, p=0.002), hyperlipidemia (58% vs 48.3%, p=0.029) and history of smoking (31.5% vs 21%, p=0.01). There were no significant racial differences in heart failure (HF), syncope or logistic euroSCOREs (Table I). Additional echocardiographic characteristics of the study population are available in Supplemental Table 1. The distribution of the 3 AVR eligibility criteria by AA vs CA race was: HF [123/236 (52%) vs 788/1612 (48.9%), p=0.181], LVEF<=50% [61/236 (25.8%) vs 453/1612 (28.1%), p=0.354], need for CABG [17/236 (7.2%) vs 252/1612 (15.6%), p<0.001] (Supplemental Table II). There were notably more CA compared with AA [420/968 (43.4%) vs 47/143 (32.9%)] who met >1 AVR eligibility criteria.

Figure 1. Inclusion/Exclusion Criteria.

Inclusion and exclusion criteria applied to finalize study population.

Table I.

Characteristics of Individuals with Severe AS Eligible for AVR^*

Characteristic	All Patients (N=1111)	African-American (N=143)	Caucasian-American (N=968)	P-Value
Age, years	78 (70– 85)	78 (68– 86)	78 (70– 84)	0.775
Female	512 (46.1%)	95 (66.4%)	417 (43.1%)	<.001
BMI	28 (24– 32)	28 (25– 34)	27(24– 32)	0.353
Systolic blood pressure	129 (117– 145)	135 (122– 152)	129 (116– 144)	0.010
Hypertension	854 (76.9%)	121 (84.6%)	733 (75.7%)	0.019
Diabetes	369 (33.2%)	65 (45.5%)	304 (31.4%)	<.001
Hyperlipidemia	630 (56.7%)	69 (48.3%)	561 (58.0%)	0.029
CHF	911 (82.0%)	123 (86.0%)	788 (81.4%)	0.181
Cerebrovascular disease	241 (21.7%)	31 (21.7%)	210 (21.7%)	0.997
PVD	141 (12.7%)	21 (14.7%)	120 (12.4%)	0.443
COPD	72 (6.5%)	9 (6.3%)	63 (6.5%)	0.923
Smoking	335 (30.2%)	30 (21.0%)	305 (31.5%)	0.010
Liver disease	5 (0.5%)	1 (0.7%)	4 (0.4%)	0.499
Ischemic heart disease/CAD	804 (72.4%)	88 (61.5%)	716 (74.0%)	0.002
End-stage renal disease	23 (2.1%)	12 (8.4%)	11 (1.1%)	<.001
Prior MI	311 (28.0%)	42 (29.4%)	269 (27.8%)	0.694
Prior CABG	218 (19.6%)	17 (11.9%)	201 (20.8%)	0.013
Prior PCI	235 (21.2%)	26 (18.2%)	209 (21.6%)	0.351
Atrial fibrillation/flutter	420 (37.8%)	38 (26.6%)	382 (39.5%)	0.003
eGFR	55.2 (40.3– 71.3)	48.8 (29.1– 70.7)	55.6 (42.1– 71.3)	0.004
Syncope	74 (6.7%)	12 (8.4%)	62 (6.4%)	0.374
Logistic euroSCORE	12.6 (6.4– 22.9)	13.5 (6.0– 26.2)	12.5 (6.5– 22.6)	0.322
3-vessel/left main disease	269 (24.2%)	17 (11.9%)	252 (26.0%)	<.001
LVEF, median	55 (40– 55)	55 (40– 55)	55 (40– 55)	0.591
AVA < 1 cm2 -no./total no.	712/740 (96.2%)	77 /81 (95.1%)	635/659 (96.4%)	0.535
Mean aortic gradient				0.254
> 40 mmHg	738 (67.1%)	88 (62.9%)	650 (67.7%)
< 40 mmHg	362 (32.9%)	52 (37.1%)	310 (32.3)
Household income, $	46,154 (35,481– 62,383)	38,529 (30,556– 51,707)	47,600 (36,070– 63,516)	<.001

^*Categorical variables expressed as n (%), continuous variables expressed as median (interquartile range).

AVA: Aortic valve area, BMI: Body Mass Index, CABG: Coronary Artery Bypass Graft, CAD: Coronary Artery Disease, CHF: Congestive Heart Failure, COPD: Chronic Obstructive Pulmonary Disease, eGFR: estimated Glomerular Filtration Rate, IQR: Interquartile Range, LVEF: Left Ventricular Ejection Fraction, MI: Myocardial Infarction, PCI: Percutaneous Coronary Intervention, PVD: Peripheral Vascular Disease

Racial Differences in AVR

By 1 year from qualifying echocardiogram, 664/1,111 (59.8%) of eligible individuals had received either SAVR (594) or TAVR (70) (Supplemental Table III). There was a lower proportion of AA who received AVR compared with CA (40.6% vs 63%, adjusted odds ratio [aOR] 0.46, 95% confidence interval [CI]: 0.30–0.71, p<0.001) (Table II). In patients receiving AVR, the median time to AVR was 19.5 days and differed by race: 7.5 days (IQR=4–31) vs 21 days (IQR=6–59) for AAs vs CAs respectively (Figure 2).

Table II.

Association Between Race, AVR Treatment and Mortality within 1 year

	AA Patients N=143*	CA Patients N=968*	Unadjusted Effect Estimate (95% CI)	Unadjusted P-Value	Adjusted Effect Estimate (95% CI)	Adjusted P-Value
AVR	58 (40.6%)	606 (63.0%)	0.42 (0.29–0.60)	<0.001	0.46 (0.30–0.71)	<0.001
1-Year Mortality†	41 (28.9%)	265 (27.9%)	1.04 (0.75–1.45)	0.800	0.81 (0.57–1.17)	0.262

^*Effect estimates are odds ratios for 1 Year AVR-free mortality and hazard ratios for all-cause mortality through 1 year

^†Represents risk of death within 1 year of baseline echocardiogram for both AVR and non-AVR patients.

AA: African-Americans, AVR: Aortic Valve Replacement, CA: Caucasian-Americans

Figure 2. Cumulative Incidence of AVR rates within 1 Year, Stratified by Race.

Rates of AVR within 1 year of index echocardiogram showing severe aortic stenosis.

AVR Non-receipt

We identified 432 patients (84 AA, 348 CA) who met AVR criteria but did not receive AVR. Caucasian Americans relative to AAs were more likely to have high operative risk (34.8% vs 28.6%), death (19.8% vs 11.9%) and inadequate evaluation (12.6% vs 11.9%) as reasons for AVR non-receipt. African Americans were more likely to have the severity of AS reclassified (27.4% vs 19%) and patient refusal (20.2% vs 13.8%) as reasons for AVR non-receipt. Chi-squared testing did not show any significant racial differences in reasons for AVR non-receipt (p=0.12) (Figure 3). Among individuals initially deemed eligible for AVR who did not receive AVR, the logistic euroSCORES did not differ by race, (median: 18.4 (IQR: 9.6–31.5) vs 17.9 (9.7–31.1), p=0.836) for AA and CA respectively.

Figure 3. Association of Race with AVR Non-Receipt Within 1 Year.

Event distribution of AVR non-receipt categories separated by race.

AVR associated mortality and all-cause mortality by race

Among patients eligible for AVR, overall 1-year mortality was 306/1,111 (27.5%). Kaplan-Meier estimates of all-cause mortality through 1-year post-baseline echocardiogram, stratified by race, are shown in Figure 4. Through 1-year of follow-up from baseline echocardiogram, there was no significant difference in all-cause mortality in AA vs. CA independent of AVR status (28.9% vs 27.9%, aHR 0.81, 95% CI 0.57–1.17, p=0.262) (Table II).

Figure 4. Cumulative Incidence of All-Cause Mortality within 1 Year, Stratified by Race.

Estimated event probability of all-cause death within the first 1 year of index echocardiogram showing severe aortic stenosis. Numbers below time axis indicate number of patients at risk in each group

Figure 5 shows Kaplan-Meier estimates of mortality, through 1 year post-AVR, stratified by race and AVR status. Among patients that underwent AVR, independent of race, there was an overall decrease in mortality. Table III demonstrates an association of lower mortality with AVR 1-year post-AVR irrespective of race (AA [aHR 0.53, 95% CI 0.23–1.21] vs CA [aHR 0.25, 95% CI 0.18–0.33]) (Table III). Analysis of the interaction between race and AVR status on 1-year mortality demonstrated no significant difference (p=0.086).

Figure 5. Cumulative Incidence of Mortality within 1 Year post-AVR, Stratified by AVR Status & Race.

Estimated event probability of death within the first 2 years post-AVR in both Caucasian-Americans and African-Americans with & without AVR within 1 Year of index echocardiogram. For patients who did not receive AVR, follow-up for mortality started at 19.5 days post-echocardiogram. This was the median time to AVR, calculated among those who received AVR. Numbers below time axis indicate number of patient at risk in each group.

Table III.

Association between AVR, Race and Outcomes over 1 Year of Follow-up in patients surviving to AVR^*

Clinical Endpoint	AA Unadjusted HR (95% CI)	CA Unadjusted HR (95% CI)	Interaction P-Value	AA Adjusted HR (95% CI)	CA Adjusted HR (95% CI)	Interaction P-Value
1- Year Mortality	0.43 (0.19–0.97)	0.20 (0.15–0.27)	0.079	0.53 (0.23–1.21)	0.25 (0.18–0.33)	0.086

^*Time zero for this analysis was time of AVR for patients undergoing AVR. For those not undergoing AVR, time zero was the median time to AVR (19.5 days) in those who received it. Adjusted model is stratified by year of echo group and adjusts for end-stage renal disease, hypertension, diabetes, eGFR, 3-vessel/left-main disease, ischemic heart disease, prior CABG, aortic regurgitation, mitral regurgitation, median household income, and smoking.

^†Hazard ratio represents the increase in hazard of event in patients in indicated racial group undergoing AVR versus those who do not.

AA: African-Americans, CA: Caucasian-Americans

Discussion

In contrast with prior studies that have assessed AVR utilization and mortality by race we additionally determined reasons for AVR non-receipt. Moreover, outside of large registries such as the National Cardiovascular Data Registry (NCDR) or Transcatheter Valve Therapy (TVT), our paper has one of the largest patient populations to assess racial differences in the treatment, and outcomes of patients with severe AS.

Our analysis was notable for several findings. Despite a higher proportion of risk factors associated with development of AS, AA patients undergoing echocardiography had a lower prevalence of severe AS relative to CA. Among individuals with severe AS and indications for AVR, a lower proportion of AA received AVR compared with CA that did not appear to be explained by early mortality after diagnosis, co-morbidities, or surgical risk. Among patients receiving AVR, median time to AVR was shorter for AA relative to CA. There were no AA/CA differences in mortality rates 1-year post-echocardiogram for all patients eligible for AVR. Finally, AVR was associated with lower rates of subsequent 1-year mortality in both AA and CA patients with no evidence of differential benefit between races.

Our study found that AA had a significantly lower odds of AVR compared with CA adjusting for co-morbidities, euroSCOREs and median household incomes. This is consistent with other studies that have reported lower rates of AVR in AA compared with CA. Yeung et. al showed AA received less AVR than their CA counterparts [(39% vs 53%, p-value 0.02) (aOR 0.55, 95% CI 0.35–0.87, p-value 0.01)].²³ Alqahtani et al examined 96,728 total patients over a 12 year span (2003–2014) and showed a lower ratio of AVR to AS-related admissions in AA vs CA (6.7% vs 11.3%, p<0.001).²⁴

We observed that AAs were more likely to refuse AVR or have their severe AS reclassified whereas CAs were more likely to die prior to intervention or be deemed non-operative candidates. Dharmajian et al’s analysis of heart centers participating ACC’s Championing Care for the Patient with Aortic Stenosis noted patient refusal as the most common reason for AVR non-receipt.²⁵ Freed et al noted patient refusal as one of the reasons patients did not receive guideline-directed AVRs.²⁶ Patient refusal as a proposed reason for AVR underutilization has many associated layers. Cultural differences resulting in poor communication between patient/provider, lack of understanding of the procedure, or patient/provider mistrust may all be contributing factors. AAs relative to CAs may be more risk averse.^27–28 African American patients may be more amenable to TAVR given the less invasive approach. A recent analysis of the Optum database suggests that racial differences in TAVR/SAVR use have decreased as TAVR use has increased.²⁹ As we continue to expand TAVR implementation among lower risk populations, patient education and shared decision making may represent a strategy to mitigate racial disparities in AVR utilization.³⁰ Patient–related factors such as disease prevalence and differences in life expectancy would be expectant factors for differential AVR rates between AA and CA. Conversely differences in other patient-related factors that vary by race such as symptom awareness, access to care, socioeconomic factors, cultural beliefs, and trust/mistrust of health are potentially modifiable factors that if targeted may close disparities in AVR.³¹

In our analysis, AAs received AVR earlier than their CA counterparts. The reasons for this finding warrant further investigation. One potential explanation may be that AAs present with more advanced disease requiring urgent/emergent intervention relative to CA. This magnifies the importance of prompt referral and follow-up as AA patients that are ready to move forward with their procedure may want it performed soon or not at all. Additionally, healthcare/system factors, including referral bias, cultural/language barriers, and regional access to subspecialists have been noted to be factors contributing to underutilization of AVR.^29,32 Implementing performance quality metrics used to assess referral patterns/practices within a health system may address subspecialty access and ensure patient follow-up.

Overall, AA not undergoing AVR shared similar surgical risk profiles in this study. Current data however suggests surgical complications to be worse in AA relative to CA. Several studies have noted higher rates of peri-surgical complications in AA, including acute kidney injury, stroke, prolonged ventilation, reoperation for bleeding, increased cost of hospitalization and longer length of stay.^6,24,30 Our analysis showed no racial differences in logistic euroSCOREs, however current risk stratification tools may not adequately represent surgical risk in racially diverse patients. The STS and euroSCOREs were derived from populations that were 94.41% white.^19,20

We found no differences in 1-year all-cause mortality despite lower rates of AVR in AA relative to CA. Other studies have shown similar findings. Yeung et al showed no black-white differences in outcomes after AVR with similar 3-year survival. Additionally, patients undergoing AVR had better survival at 1 & 3 years without racial differences.²³ Moreover, Minha et al did not show black-white differences in 1-year survival after TAVR in AA vs CA.⁷ Additional studies utilizing the STS database and Medicare for patient sampling showed similar findings.^33–34 Conversely, other database studies suggest AAs have worse outcomes including higher rates of 1-year hospitalization, increased hospital stay/costs, and higher post-operative mortality.^30,35–37 The small number of AA patients included in our analysis may have limited our ability to determine AA/CA differences in mortality. The inability to subcategorize mortality did not allow for investigation of competing risks that may have contributed to a greater mortality among CAs.

Study Limitations

The findings of this study must be interpreted in the context of its limitations. This analysis involved a single center retrospective cohort with specific practice patterns, thus limiting generalizability. Procedures performed at outside hospitals were not captured, potentially leading to underreporting. Data was unavailable to report the rates of patients receiving CABG and AVR within the same surgery. However, we noted that Caucasian Americans were more likely than AA to have the need for CABG as an AVR indication and within 1-year of index echocardiogram, more CA received CABG than AA [415/968 (42.9%) vs 42/143 (29.4%) (p-value=0.002)] (Supplemental Table IV). We were unable to delineate whether there were contraindications to AVR based on the severity of coronary artery disease and inability to revascularize diseased vessels. In our analysis to determine if there was a differential impact of AVR as a treatment on mortality for AA relative to CA individuals, we utilized the median time to AVR for the study population (19.5 days) rather than the median time to AVR for each racial group. Given that the median time to AVR was longer for CA (21 days) than AA (7.5 days), this likely introduces immortal time bias towards a differential treatment benefit of AVR in AA relative to CA patient as AAs would have more deaths excluded and CAs more deaths added. Given the finding of no significant interaction of AVR benefit by race (p=0.086) the directionality of the bias would produce a less significant interaction, further supporting our conclusion (Table 3).

There may be residual confounding in the relationship between race, AVR, and the outcomes due to unmeasured variables, such as comorbidities, medications, and socioeconomic variables not captured in DELD. We did not account for a constellation of symptoms that may have impacted decisions for AVR. We did however determine AVR eligibility using guideline recommendations and thus our criteria are valid while not exhaustive. Lastly, each reason for AVR non-use was limited to information documented by providers in patient charts and thus subject to any bias associated with providers.

Conclusion

Despite a higher proportion of traditional risk factors, AA relative to CA had lower prevalence of severe AS. Among patients with severe AS eligible for AVR, AA patients were less likely to undergo AVR within 1-year. AS reclassification and patient refusal were the biggest drivers of AVR non-receipt for AAs, whereas high operative risk and death prior to procedure were the biggest drivers for CAs. Despite these differences, there was no significant racial difference in 1-year all-cause mortality. Based on these findings, clinicians should be mindful of AVR under-receipt in AA populations. Early identification of AA patients with severe AS is critical and measures of emphasis include adequate examination, follow-up visits and patient education/awareness of symptoms. After detection of severe AS, prompt referral is key.

ABBREVIATIONS

AA

African American

ACC

American College of Cardiology

AHA

American Heart Association

AS

Aortic Stenosis

AVR

Aortic Valve Replacement

CA

Caucasian American

DDCD

Duke Databank for Cardiovascular Disease

DELD

Duke Echocardiogram Laboratory Database

DUH

Duke University Hospital

ICD-9

International Classification of Diseases, 9^th Revision

NCDR

National Cardiovascular Data Registry

SAVR

Surgical Aortic Valve Replacement

STS

Society of Thoracic Surgeons

TVT

Transcatheter Valve Therapy

TAVR

Transcatheter Aortic Valve Replacement