Journal of the American Heart Association

ORIGINAL RESEARCH

Risk of Postdischarge Bleeding From Dual

Antiplatelet Therapy After Percutaneous
Coronary Intervention Among US Black and
White Adults
Brittain Heindl , MD; Stephen Clarkson , MD; Vibhu Parcha , MD; Chrisly Dillon, MD; Renuka Narayan , MPH;
Ebikere Usifo, MBBS, MPH; William Hillegass , MD, PhD; Marguerite R. Irvin, PhD; Pankaj Arora , MD;
Guihua Zhai, PhD; Mark Beasley, PhD; Nita Limdi , PharmD, PhD, MSPH

BACKGROUND: Dual antiplatelet therapy after percutaneous coronary intervention reduces myocardial infarctions but increases
bleeding. The risk of bleeding may be higher among Black patients for unknown reasons. Bleeding risk scores have not
been validated among Black patients. We assessed the difference in bleeding risk between Black and White patients along
with the performance of the Predicting Bleeding Complications in Patients Undergoing Stent Implantation and Subsequent
Dual Anti Platelet Therapy, Patterns of Nonadherence to Antiplatelet Regimens in Stented Patients, and Academic Research
Consortium for High Bleeding Risk scores among both groups.

METHODS AND RESULTS: This was a single-­center prospective study of patients who underwent percutaneous coronary
intervention (2014–­2019) and were followed for 1 year. The outcome was postdischarge Bleeding Academic Research
Consortium 2 to 5 bleeding. Incidence rates were reported. Cox proportional hazards models measured the effect of self-­
reported Black race on bleeding and determined the predictors of bleeding among 19 a priori variables. The 3 risk scores
Downloaded from http://ahajournals.org by on September 21, 2022

were assessed among Black and White patients separately using the Harrell concordance index. Of 1529 included patients,
342 (22.4%) self-­reported as being Black race. Unadjusted bleeding rates were 22.7 per 100 person-­years among Black
patients versus 16.3 among White patients (hazard ratio, 1.41 [95% CI, 1.00–­2.00], P=0.052). Predictors of bleeding were
age, glomerular filtration rate <30 mL/min per 1.73 m2, prior bleeding, ticagrelor or prasugrel use, and anticoagulant use.
Among Black and White patients, respectively, the C-­indexes were the following: 0.644 versus 0.600 for Predicting Bleeding
Complications in Patients Undergoing Stent Implantation and Subsequent Dual Anti Platelet Therapy (P<0.001 for both), 0.620
versus 0.612 for Patterns of Nonadherence to Antiplatelet Regimens in Stented Patients (P=0.003 and P<0.001, respectively),
and 0.600 versus 0.598 for Academic Research Consortium for High Bleeding Risk (P=0.006 and P<0.001, respectively).

CONCLUSIONS: The risk of dual antiplatelet therapy–­associated postdischarge Bleeding Academic Research Consortium 2 to
5 bleeding was not significantly different between self-­reported Black and White patients. Bleeding risk scores performed
similarly among both groups.

Key Words: incidence ■ percutaneous coronary intervention ■ platelet aggregation inhibitors ■ prasugrel hydrochloride ■ proportional
hazards models ■ prospective studies ■ ticagrelor

D
ual antiplatelet therapy (DAPT) is recommended reduces the incidence of subsequent myocardial infarc-
after an acute myocardial infarction or percutane- tion, it causes increased bleeding and, through unclear
ous coronary intervention (PCI).1 Although DAPT mechanisms, is associated with excess noncardiac

Correspondence to: Brittain Heindl, MD, University of Alabama at Birmingham, Boshell Diabetes Building, Room 201, 1808 7th Avenue South, Birmingham,
AL 35294. Email: bfheindl@uabmc.edu
Supplemental Material is available at https://www.ahajo​urnals.org/doi/suppl/​10.1161/JAHA.121.024412
For Sources of Funding and Disclosures, see page 9.
© 2022 The Authors. Published on behalf of the American Heart Association, Inc., by Wiley. This is an open access article under the terms of the Creative
Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use
is non-commercial and no modifications or adaptations are made.
JAHA is available at: www.ahajournals.org/journal/jaha

J Am Heart Assoc. 2022;11:e024412. DOI: 10.1161/JAHA.121.0244121

 Heindl et al Bleeding Risk Prediction Among Black Adults

CLINICAL PERSPECTIVE PRECISE-­DAPT Predicting Bleeding

Complications in Patients
Undergoing Stent Implantation
What Is New?
and Subsequent Dual Anti
• The risk of dual antiplatelet therapy–­associated
Platelet Therapy
postdischarge bleeding was not statistically
higher for Black patients compared with White
patients.
• A nonsignificant numerical difference was pre-
sent, and this difference was primarily explained mortality2 and a lower quality of life.3 Patients at high
by a higher proportion of Black patients having bleeding risk experience worse outcomes from DAPT,4
severe kidney disease, defined by a glomeru- and US guidelines recommend shorter durations of
lar filtration rate <30 mL/kg per 1.73 m2 or end-­ DAPT for patients at high bleeding risk.1
stage renal disease. Risk factors associated with bleeding while on
• The Predicting Bleeding Complications in DAPT have been combined into risk scores to de-
Patients Undergoing Stent Implantation termine high bleeding risk. Three scores (Predicting
and Subsequent Dual Anti Platelet Therapy, Bleeding Complications in Patients Undergoing Stent
Academic Research Consortium for High
Implantation and Subsequent Dual Anti Platelet
Bleeding Risk, and Patterns of Nonadherence
Therapy [PRECISE-­DAPT],5 Patterns of Nonadherence
to Antiplatelet Regimens in Stented Patients
scores had moderate predictive abilities among to Antiplatelet Regimens in Stented Patients [PARIS],6
both Black and White patients. and Academic Research Consortium for High Bleeding
Risk [ARC-­HBR7]) were designed to predict bleeding
What Are the Clinical Implications? after hospital discharge and have been validated in ex-
• Race should not be considered by clinicians ternal cohorts. They have been referenced in European
when assessing bleeding risk while on dual an- antiplatelet guidelines,8 but there is no consensus on
tiplatelet therapy. which score should be used in clinical practice. Current
• Clinicians should consider the following 5 pre- US guidelines have not referenced these risk scores.
dominant factors when assessing bleeding risk: Prior studies have demonstrated increased bleeding
severe kidney disease, age, prasugrel or ticagre- from DAPT among Black adults compared with White
lor use, anticoagulant use, and prior bleeding.
Downloaded from http://ahajournals.org by on September 21, 2022

adults.9,10 Factors contributing to this difference are not

• The Predicting Bleeding Complications in
known and may not be measured by the PRECISE-­
Patients Undergoing Stent Implantation
and Subsequent Dual Anti Platelet Therapy, DAPT, PARIS, or ARC-­HBR risk scores. In addition,
Academic Research Consortium for High the cohorts used to validate these 3 scores have been
Bleeding Risk, and Patterns of Nonadherence predominantly from European or Asian countries.11,12
to Antiplatelet Regimens in Stented Patients They have not been validated in self-­reported Black
scores can be confidently applied to both Black adults—­a subgroup underrepresented in PCI trials.13
and White patients in clinical practice, with the In the present study, we aim to (1) compare postdis-
Predicting Bleeding Complications in Patients charge bleeding between self-­ reported Black and
Undergoing Stent Implantation and Subsequent White patients; (2) identify clinical factors that contrib-
Dual Anti Platelet Therapy score better measur- ute to this difference; and (3) assess the ability of the
ing gradations in age and kidney disease.
PRECISE-­ DAPT, PARIS, and ARC-­ HBR risk scores
to predict postdischarge bleeding among Black and
White patients separately.

Nonstandard Abbreviations and Acronyms METHODS

ARC-­HBR Academic Research The data that support the findings of this study are
Consortium for High Bleeding available from the authors on reasonable request.
Risk The PRiME-­ GGAT (Pharmacogenomic Resource
BARC Bleeding Academic Research to Improve Medication Effectiveness-­ Genotype-­
Consortium Guided Antiplatelet Therapy) prospective cohort study
DAPT dual antiplatelet therapy enrolled patients aged ≥18 years who underwent
PARIS Patterns of Nonadherence to PCI at the University of Alabama at Birmingham
Antiplatelet Regimens in Hospital. The study was approved by the University of
Stented Patients Alabama at Birmingham Hospital institutional review
board. Consent was obtained at enrollment, which

J Am Heart Assoc. 2022;11:e024412. DOI: 10.1161/JAHA.121.0244122

 Heindl et al Bleeding Risk Prediction Among Black Adults

occurred during the index PCI hospitalization (June Black and White patients separately. Incidence rates
2014–­November 2019). of bleeding were calculated and reported as per 100
A structured form was used to record age, self-­ person-­years.
reported race, sex, and smoking status. Height, We assessed the influence of each baseline variable
weight, and laboratory values were recorded from on bleeding with a time-­to-­event analyses by using the
the medical record, measured on the day of PCI. fit proportional hazards function to develop Cox pro-
Laboratory values included serum creatinine, white cell portional hazards models. An unadjusted analysis was
count, platelet count, and hemoglobin. The Chronic first performed and then a multivariable analysis using
Kidney Disease Epidemiology Collaboration equation only those variables associated with bleeding with an
was used to derive an estimated glomerular filtration unadjusted P<0.20. Hazard ratios with 95% CIs were
rate (GFR; mL/min per 1.73 m2), which does not con- reported. A sensitivity analysis was performed using
sider self-­reported race as a variable.14 The following the same methodology but accounting for the compet-
variables were obtained from the medical record: his- ing risk of death using the PHREG procedure (PROC
tory of diabetes (or the use of glucose-­lowering med- PHREG, SAS software, version 9.4).
ications), hypertension (or the use of antihypertensive Cox proportional hazards models were then used
medications), stroke (or transient ischemic attack), to measure the mediating effect of each predictor vari-
prior bleeding, and liver cirrhosis (with portal hyperten- able (those with P<0.05 after multivariable adjustment)
sion). Home and discharge medications (antiplatelets, on the association between self-­reported Black race
anticoagulants, proton pump inhibitors, nonsteroidal and bleeding. Each model was adjusted for age and
anti-­inflammatory drugs, and corticosteroids) were re- sex and then for each predictor variable independently.
corded from the medical record. A sensitivity analysis was performed to determine the
Patients were followed for 1 year. All University of mediating effect of 3 forms of the GFR variable (GFR
Alabama at Birmingham medical records were re- <30 mL/min per 1.73 m2, GFR <45 mL/min per 1.73 m2,
viewed, and records from facilities outside of the and continuous) on the association between self-­
University of Alabama at Birmingham system were re- reported Black race and bleeding. The significance of
quested and reviewed. Postdischarge bleeding events each mediating effect was measured by the Sobel test.
were documented by study personnel and adjudicated We then assessed the performance of 3 commonly
by 2 physicians. Patients were right-­censored from the used risk scores among Black and White patients
analysis for the following 4 reasons: (1) they had any separately.
Downloaded from http://ahajournals.org by on September 21, 2022

Bleeding Academic Research Consortium (BARC) 2 to

5 bleeding event, (2) they were no longer taking DAPT 1. The PRECISE-­DAPT score5 is composed of 5 vari-
therapy, (3) they died, or (4) they were lost to follow-­up. ables: age, GFR, hemoglobin, white cell count, and
For patients lost to follow-­up, the last known clinic visit previous clinically significant bleeding. Each com-
or hospitalization was the point of censor. ponent is assigned point values as per Table S2.
Postdischarge bleeding events were categorized A score ≥25 denotes high bleeding risk, 18 to
based on the BARC statement (Table S1).15 The out- 24 denotes moderate risk, 11 to 17 denotes low
come for each of our analyses was postdischarge bleeding risk, and ≤10 denotes very low bleeding
BARC 2 to 5 bleeding. The following summarizes the risk. For this study, PRECISE-­DAPT was condensed
BARC schema: type 1 bleeding does not cause the into high (≥25), moderate (18–­ 24), and low (≤17)
patient to seek unscheduled care, type 2 bleeding categories.
prompts evaluation and care but does not meet types 2. The PARIS score6 is composed of 6 variables: age,
3 to 5 criteria, type 3 bleeding is major (hemoglobin current smoking, body mass index, GFR, hemo-
drop >3 g/dL, cardiac tamponade, intracranial, intra- globin, and oral anticoagulant use. Each is assigned
ocular, required transfusion, surgical intervention, or a point value as per Table S3. A score ≥8 denotes
vasoactive agents), type 4 bleeding is a coronary ar- high bleeding risk, 4 to 7 denotes moderate bleeding
tery bypass graft related, and type 5 bleeding is fatal. risk, and ≤3 denotes low bleeding risk.
We compared baseline variables between Black 3. The ARC-­HBR score7 is composed of 15 variables,
and White patients using t tests for continuous vari- classified as either major or minor criteria. For the
ables and χ2 tests for categorical variables. We chose present study, 7 variables were modified or excluded
these 19 variables because they were either included to fit our data set (Table S4). Having either 1 major or
in the PRECISE-­DAPT, PARIS, or ARC-­HBR risk scores 2 minor criteria denote high bleeding risk.
or they were determined a priori to be associated with
bleeding. We distributed patients into categories of risk for
We reported the number of bleeding events, each score. For the PRECISE-­DAPT and PARIS scores,
person-­years of follow-­up, and the anatomical location we distributed patients into quartiles of risk because
of each bleeding event among all patients and then these scores were intended to be continuous. For the

J Am Heart Assoc. 2022;11:e024412. DOI: 10.1161/JAHA.121.0244123

 Heindl et al Bleeding Risk Prediction Among Black Adults

ARC-­HBR score, we distributed patients into 2 catego- patients had a hemoglobin <12 g/dL. A greater pro-
ries of risk (low or moderate versus high risk) as this score portion of Black patients had diabetes and hyperten-
was intended to be binary. Incident rates of bleeding sion. Other baseline variables, including antithrombotic
were reported for each risk category, stratified by race. medication use, were not different between groups.
Proportional hazard models were used to compare each Overall, 159 patients (10.2%) experienced a post-
category with the lowest category of risk, also stratified discharge BARC 2 to 5 bleeding event, of which 44
by race. Adjustment was made for sex and the predictor were Black patients (12.9%) and 115 were White pa-
variables determined from the aforementioned analyses tients (9.7%). The number of events per BARC cate-
(those with P<0.05 after multivariable adjustment), un- gory, and the anatomical location of each event, are
less the predictor variable was included in any 1 of the 3 reported in Table S5. The largest proportion of BARC 2
risk scores. to 5 events were from gastrointestinal bleeding (37.1%),
We quantified the discriminative abilities of the 3 risk followed by nonprocedural hematomas (21.4%). The
scores by measuring the Harrell concordance index incidence of bleeding by each BARC category is pre-
(C-­index).16 To calculate this, we used the PHREG pro- sented in Table 2. Among all patients, the incidence of
cedure (PROC PHREG, SAS software, version 9.4) to postdischarge BARC 2 to 5 bleeding was 15.5 per 100
produce Cox proportional hazards models and se- person-­years (Table 2).
lected the option to compute a Harrell C-­index. The For the time-­to-­event analysis, the unadjusted pre-
scores were included as single continuous variables dictors of postdischarge BARC 2 to 5 bleeding were
for PRECISE-­DAPT and PARIS and as a nominal vari- age, GFR <30 mL/min per 1.73 m2, previous bleeding
able for ARC-­HBR. C-­indexes were calculated for all (requiring medical attention), ticagrelor or prasugrel
patients and then Black and White patients separately. use, anticoagulant use, hemoglobin, and prior isch-
All statistical tests were 2-­sided, with main effects emic stroke or transient ischemic attack (Table 3). Sex,
tested at an α level of 0.05 unless otherwise specified. liver cirrhosis, and proton pump inhibitor use each had
Incidence rates were calculated by using the OpenEpi a P value between 0.05 and 0.20 and were also in-
online software platform.17 The other analyses were cluded in the adjusted analyses. After adjustment for
performed by using JMP software, version 16.2, and these 11 variables, only age, GFR <30 mL/min per
SAS software, version 9.4. 1.73 m2, previous bleeding, ticagrelor or prasugrel use,
and anticoagulant use remained predictors.
Self-­reported Black race was not a significant pre-
Downloaded from http://ahajournals.org by on September 21, 2022

dictor of bleeding (unadjusted hazard ratio, 1.41 [95%

RESULTS CI, 1.00–­2.00]; P=0.052). Unadjusted and adjusted
Of the 1558 patients enrolled in the study, 29 were ex- time-­to-­ event curves for Black and White patients
cluded from the analysis because their self-­reported separately are displayed in Figures S1 and S2, respec-
race or ethnicity was other than Black race or White tively. The results of a sensitivity analysis incorporating
race, leaving 1529 patients included in the final analy- the competing risk of death were similar and are pre-
sis, of which 22.4% were Black patients. The analysis sented in Table S6.
included 1027.1 person-­years of follow-­up. The mean The mediating effects of each predictor variable
follow-­up was 0.62 years per person for Black patients individually on the relationship between self-­reported
and 0.69 years for White patients. Among all patients, Black race and BARC 2 to 5 bleeding are presented in
908 (59.3%) were censored for any reason (64.3% Table S7. Only GFR<30 mL/min per 1.73 m2 was a sig-
Black patients versus 57.9% White patients), and 39 nificant mediator (34.8% reduction in effect; P<0.001).
patients (2.5%) were censored because of death (1.5% Alternative forms of the GFR variable (<45 mL/min per
Black patients versus 2.8% White patients). Of the in- 1.73 m2 and continuous) did not have a mediating ef-
cluded patients, <1% had missing data elements. fect on the association between Black race and bleed-
Black patients were younger than White patients, ing (Table S8).
and a smaller proportion of Black patients were aged There were differences in the proportions of pa-
≥75 years (Table 1). A greater proportion of Black pa- tients classified as high risk, compared with low–­
tients were women and were current smokers. body moderate risk, between Black and White patients, for
mass index was higher among Black patients, and a the PRECISE-­DAPT and PARIS scores. The PRECISE-­
larger proportion of Black patients had a body mass DAPT score classified 30.4% of all patients as high
index ≥35 kg/m2. Mean GFR was lower among Black bleeding risk (35.4% of Black patients compared with
patients, and a greater proportion of Black patients 29.9% of White patients; P=0.023), the PARIS score
had a GFR <30 mL/min per 1.73 m2. Black patients classified 12.3% of all patients as high risk (15.8% of
had a lower mean white cell count and a higher mean Black patients compared with 11.3% of White patients;
platelet count. Black patients had a lower mean hemo- P=0.031). There were no differences in the proportions
globin concentration, and a greater proportion of Black classified as high risk, compared with low–­moderate

J Am Heart Assoc. 2022;11:e024412. DOI: 10.1161/JAHA.121.0244124

 Heindl et al Bleeding Risk Prediction Among Black Adults

Table 1. Baseline Characteristics of the Study Population

All patients, N=1529 Black patients, n=342 White patients, n=1187 P value

Age, y 62.2±11.9 58.9±11.4 63.1±11.8 <0.001

Age ≥75 y 224 (14.7) 22 (6.4) 202 (17.0) <0.001
Female sex 463 (30.3) 146 (42.7) 317 (26.7) <0.001
Smoking status
Current smoker 384 (25.8) 108 (32.6) 276 (23.9) 0.005
Former smoker 552 (37.1) 107 (32.3) 436 (37.7)
Never smoker 552 (37.1) 116 (35.1) 445 (38.5)
Body mass index, kg/m2 30.3±6.1 31.1±7.1 30.0±5.8 0.010
<25 kg/m2 282 (18.5) 66 (19.4) 216 (18.2) 0.007
25–­34.9 kg/m2 953 (62.5) 191 (56.0) 762 (64.4)
≥35 kg/m2 290 (19.0) 84 (24.6) 206 (17.4)
GFR, mL/min per 1.73 m2 74.8±25.7 67.8±30.4 76.8±23.8 <0.001
GFR ≥60 mL/min per 1.73 m2 1130 (74.2) 227 (66.8) 903 (76.3) <0.001
GFR 45–­59 mL/min per 1.73 m2 202 (13.2) 38 (11.2) 164 (13.9)
GFR 30–­44 mL/min per 1.73 m2 90 (5.9) 23 (6.8) 67 (5.7)
GFR <30 mL/min per 1.73 m2 or requiring dialysis 101 (6.6) 52 (15.3) 49 (4.1)
White cell count, ×103/μL 8.4 (6.5–­10.8) 7.6 (3.7–­10.3) 8.5 (6.7–­10.8) <0.001
Platelet count, ×109 per L 222.9±70.2 238.5±73.4 218.5±68.7 <0.001
9
Platelet count, <100×10 per L 27 (1.8) 3 (0.9) 24 (2.0) 0.240
Hemoglobin, g/dL 13.6±1.9 13.0±1.9 13.8±1.9 <0.001
Hemoglobin, <12 g/dL 291 (19.1) 98 (28.9) 193 (16.3) <0.001
Diabetes 653 (42.7) 165 (48.3) 488 (41.1) 0.019
Hypertension 1308 (85.5) 308 (90.1) 1000 (84.2) 0.007
Prior ischemic stroke or TIA 210 (13.7) 54 (15.8) 156 (13.1) 0.217
Downloaded from http://ahajournals.org by on September 21, 2022

Prior hemorrhage 21 (1.4) 2 (0.6) 19 (1.6) 0.194

Previous bleeding requiring medical attention* 105 (6.9) 27 (7.9) 78 (6.6) 0.397
†
ARC-­HBR major bleeding history 27 (1.8) 8 (2.3) 19 (1.6) 0.366
ARC-­HBR minor bleeding history† 8 (0.5) 3 (0.9) 5 (0.4) 0.388
Liver cirrhosis with portal hypertension 24 (1.6) 5 (1.5) 19 (1.6) 1.000
P2Y12 inhibitor use (in combination with aspirin)
Clopidogrel 1001 (65.8) 228 (66.9) 773 (65.5) 0.182
Prasugrel 36 (2.4) 4 (1.2) 32 (2.7)
Ticagrelor 475 (31.2) 104 (30.5) 371 (31.4)
Anticoagulant use 218 (14.3) 44 (12.9) 174 (14.8) 0.430
Proton pump inhibitor use 538 (35.4) 115 (33.7) 423 (35.9) 0.480
Long-­term NSAID use‡ 63 (4.1) 14 (4.1) 49 (4.1) 1.000
Long-­term corticosteroid use‡ 68 (4.4) 13 (3.8) 55 (4.6) 0.655

Continuous variables are displayed as mean±SD or median (interquartile range) if the distribution was skewed, whereas categorical variables are displayed as
number (percentage). ARC-­HBR indicates Academic Research Consortium for High Bleeding Risk; GFR, glomerular filtration rate; PRECISE-­DAPT, Predicting
Bleeding Complications in Patients Undergoing Stent Implantation and Subsequent Dual Anti Platelet Therapy; and TIA, transient ischemic attack.
*Definition used for the PRECISE-­DAPT score.
†
Defined in Table S4.
‡
Both a home medication at the time of index percutaneous coronary intervention and continued at discharge.

risk, between Black and White patients, for the ARC-­ hazard ratios comparing risk categories. The incidence
HBR criteria: 46.4% overall, with 48.3% of Black pa- rates of BARC 2 to 5 bleeding among risk categories
tients classified as high risk compared with 45.8% and adjusted hazard ratios comparing categories are
of White patients (P=0.46). The number of patients, presented in Figure.
events, and person-­years of follow-­up in each category For the PRECISE-­DAPT score, the Harrell C-­index
of risk are provided in Table S9, along with unadjusted was 0.614 for all patients (P<0.001), 0.644 for Black

J Am Heart Assoc. 2022;11:e024412. DOI: 10.1161/JAHA.121.0244125

 Heindl et al Bleeding Risk Prediction Among Black Adults

Table 2. Incidence Rates for Postdischarge Bleeding Events by BARC Category

All patients, N=1529 Black patients, n=342 White patients, n=1187

Person-­years of follow-­up 1027.1 213.2 813.9

Category of bleeding Incidence rate per 100 person-­years (95% CIs)
BARC 2 11.1 (9.3–­13.4) 12.7 (8.5–­18.8) 10.8 (8.7–­13.3)
BARC 3 4.0 (2.9–­5.4) 7.5 (4.4–­11.9) 3.1 (2.0– ­4.5)
BARC 4 0 0 0
BARC 5 0.3 (0.1–­0.8) 0.5 (0.02–­2.3) 0.2 (0.04–­0.8)
Combined BARC 2 to 5 15.5 (13.2–­18.0) 20.6 (15.2–­27.5) 14.1 (11.7–­16.9)

BARC indicates Bleeding Academic Research Consortium.

patients (P<0.001), and 0.600 for White patients Black patients compared with White patients. The only
(P<0.001). For the PARIS score, the C-­index was 0.617 predictor of bleeding that contributed to a numerical
for all patients (P<0.001), 0.620 for Black patients difference was severe chronic kidney disease (CKD),
(P=0.003), and 0.612 for White patients (P<0.001). For defined as a GFR <30 mL/min per 1.73 m2 or end-­stage
the ARC-­HBR score, the C-­index was 0.600 for all pa- renal disease. The proportions deemed high bleeding
tients (P<0.001), 0.600 for Black patients (P=0.006), risk by the PRECISE-­DAPT and PARIS scores were
and 0.598 for White patients (P<0.001). higher among Black patients, with the PRECISE-­DAPT
score classifying more patients as high risk than the
PARIS score (30.4% versus 12.3%, respectively). Each
score had a moderate predictive ability among both
DISCUSSION groups. Overall, our study suggests that differences in
In the present study of 1529 patients who underwent severe renal failure are the primary contributor to any
PCI and were placed on DAPT, the risk of postdis- differences in bleeding risk among self-­reported Black
charge bleeding was not significantly higher among individuals and race should not be considered when

Table 3. Unadjusted and Adjusted Associations of Bleeding Risk Factors With Bleeding Academic Research Consortium 2
to 5 Bleeding
Downloaded from http://ahajournals.org by on September 21, 2022

Unadjusted HR (95% CI) P value Adjusted HR (95% CI) P value

Self-­reported Black race 1.41 (1.00–­2.00) 0.052 1.37 (0.94–­1.99) 0.098

Age, per y 1.01 (1.00–­1.03) 0.039 1.01 (1.00–­1.03) 0.044
GFR <30 mL/min per 1.73 m2 or 2.65 (1.72–­4.09) <0.001 1.90 (1.15–­3.11) 0.011
end-­stage renal disease
Previous bleeding requiring 2.45 (1.57–­3.81) <0.001 1.80 (1.13–­2.89) 0.014
medical attention
Ticagrelor or prasugrel use vs 1.80 (1.32–­2.46) <0.001 2.15 (1.56–­2.97) <0.001
clopidogrel use
Anticoagulant use 2.41 (1.67–­3.46) <0.001 2.38 (1.64–­3.46) <0.001
Hemoglobin, g/dL 0.86 (0.80–­0.93) <0.001 0.94 (0.86–­1.03) 0.189
Female sex 1.26 (0.91–­1.75) 0.170 1.07 (0.77–­1.52) 0.693
Prior ischemic stroke or TIA 1.53 (1.03–­2.28) 0.034 1.24 (0.81–­1.88) 0.319
Liver cirrhosis with portal 2.10 (0.86–­5.11) 0.144 1.70 (0.67–­4.35) 0.265
hypertension
Proton pump inhibitor use 1.25 (0.91–­1.71) 0.168 1.02 (0.74–­1.42) 0.905
Current smoking, yes 0.81 (0.55–­1.19) 0.269 … …
2
Body mass index ≥35 kg/m 1.05 (0.71–­1.55) 0.783 … …
White cell count per 103/μL 1.02 (0.97–­1.05) 0.974 … …
Platelet count per 109/L 1.39 (0.39– ­4.68) 0.597 … …
Diabetes 1.11 (0.81–­1.51) 0.525 … …
Hypertension 1.03 (0.40–­2.65) 0.95 … …
Long-­term NSAID use 0.54 (0.20–­1.45) 0.219 … …
Long-­term corticosteroid use 1.52 (0.80–­2.88) 0.200 … …

GFR indicates glomerular filtration rate; HR, hazard ratio; and TIA, transient ischemic attack.
The variables included in the adjusted model were those with an unadjusted P value <0.20.

J Am Heart Assoc. 2022;11:e024412. DOI: 10.1161/JAHA.121.0244126

 Heindl et al Bleeding Risk Prediction Among Black Adults

Black patients per category

60 PRECISE-DAPT Risk Score 400
White patients per category
Incidence rate, Black patients
350

Incidence Rate for BARC 2-5 Bleeding

50 Incidence rate, White patients
300

Patients Within Each Risk Category (n)

(Per 100 Person-years)
40 37.5
250

30 200
24.5
24.0 150
20
12.3 100
10.8 14.8
10
10.1 10.4 50

0 0
Point category 11 12-19 20-27 >28
Adjusted* hazard ratio (95% confidence intervals)
0.74 1.60 2.24
Black Patients referent
(0.25-2.15, p=0.579) (0.60-4.25, p=0.348) (0.95-5.29, p=0.066)
1.05 1.47 2.46
White Patients referent
(0.60-1.85, p=0.856) (0.83-2.59, p=0.185) (1.46-4.14, p<0.001)

60 400
PARIS Risk Score
350
Incidence Rate for BARC 2-5 Bleeding

50
300

Patients Within Each Risk Category (n)

(Per 100 Person-years)

40
36.0 250

30 28.1 200

24.9 150
20
15.6
17.1 100
9.3
10 11.3
8.4 50
Downloaded from http://ahajournals.org by on September 21, 2022

0 0
Point category 3-4 5-6
Adjusted* hazard ratio (95% confidence intervals, P-value)
1.46 2.80 2.28
Black Patients referent
(0.50-4.30, p=1.46) (0.99-7.88, p=0.052) (0.76-6.84, p=0.143)
1.31 2.07 3.11
White Patients referent
(0.75-2.31, p=0.347) (1.19-3.62, p=0.010) (1.78-5.44, p<0.001)

60 700
ARC-HBR Risk Score
Incidence Rate for BARC 2-5 Bleeding

600
50
Patients Within Each Risk Category (n)
(Per 100 Person-years)

500
40

400
30.9
30
300

20 20.9
200
12.1
10
9.2 100

0 0
Category Low or moderate risk High risk
Adjusted* hazard ratio (95% confidence intervals), P-Value
1.95
Black Patients referent
(0.99-3.85, p=0.053)
2.39
White Patients referent
(1.62-3.53, p<0.001)

Figure. Incidence of BARC 2 to 5 bleeding among categories of risk, stratified by race.

Models were adjusted for sex, ticagrelor or prasugrel use, and proton pump inhibitor use.
ARC-­HBR indicates Academic Research Consortium Criteria for High Bleeding Risk; BARC,
Bleeding Academic Research Consortium; PARIS, Patterns of Nonadherence to Antiplatelet
Regimens in Stented Patients; and PRECISE-­DAPT, Predicting Bleeding Complications in
Patients Undergoing Stent Implantation and Subsequent Dual Anti Platelet Therapy.

J Am Heart Assoc. 2022;11:e024412. DOI: 10.1161/JAHA.121.0244127

 Heindl et al Bleeding Risk Prediction Among Black Adults

deciding DAPT duration. The PRECISE-­DAPT score variables were not included in widely cited risk scores.
categorized GFR with greater granularity and better Because we felt that we could not adequately mea-
stratified patients at higher risk. sure such differences, we chose to focus on clinical
The unadjusted risk of bleeding was 41% higher variables.
among Black patients, but this difference did not meet Among all patients, our analysis demonstrated a C-­
statistical significance (P=0.052). A larger proportion of index of 0.614 for the PRECISE-­DAPT score, 0.617 for
Black patients were censored from the analysis, for rea- the PARIS score, and 0.600 for the ARC-­HBR score.
sons other than death, and some events that occurred C-­indexes were higher among Black patients for all 3
may not have been observed. Other studies have re- scores compared with White patients. However, the
ported a higher unadjusted risk of DAPT-­associated C-­index values we reported for these scores are lower
bleeding among self-­reported Black adults compared than those described in other studies. For example, the
with other racial or ethnic groups.9,10,18–­20 However, PRECISE-­DAPT derivation study reported a C-­index in
these studies were different than ours in multiple ways. the derivation cohort of 0.73 and 0.70 and 0.66 in the
The most prominent difference was that prior studies 2 validation cohorts.5 Studies of the PRECISE-­DAPT
used definitions of bleeding other than the BARC cri- score by Choi et al reported C-­statistics between 0.75
teria, such as transfusion requirements, International and 0.81.28–­30 The reasons for the different statistics re-
Classification of Diseases (ICD) codes, and non-­BARC ported by these studies, compared with our own, are
definitions of major bleeding. To our knowledge, our likely methodological. Costa et al5 used the end point of
study is the only to apply the BARC criteria to examine TIMI (Thrombolysis in Myocardial Infarction) major and
bleeding among self-­reported Black patients. minor bleeding as the outcome rather than BARC 2 to 5
The only variable that reduced the effect of self-­ bleeding. Choi et al stratified the PRECISE-­DAPT score
reported Black race on postdischarge bleeding was the into 2 to 3 categories of risk rather than evaluating it as
presence of severe CKD (GFR <30 mL/min per 1.73 m2, a continuous score. In each referenced study, patients
including end-­stage renal disease). It has been well requiring oral anticoagulation were excluded, whereas
demonstrated that CKD increases the risk of bleeding in our study we included these patients. Systematic re-
while taking antiplatelet medications.21 Multiple mech- views of commonly used risk scores have reported sim-
anisms have been reported by which uremic toxins ilar wide variation in risk score performances because
and increased fibrinogen levels reduce platelet adhe- of methodological and study population differences.31
sion and aggregation.22 Also well documented is that Although the C-­indexes were similar between the
Downloaded from http://ahajournals.org by on September 21, 2022

self-­reported US Black adults have a higher prevalence PRECISE-­DAPT, PARIS, and ARC-­HBR scores, there
of severe CKD,23 compared with White adults, partially are differences between these scores that should be
because of a higher prevalence of diabetes,24 lower considered. Apart from ticagrelor, prasugrel, and anti-
blood pressure control,25 and more frequent homozy- coagulant use, age and GFR <30 mL/min per 1.73 m2
gosity for variants of the Apolipoprotein L1 gene.26 We were the strongest contributors to bleeding risk, and
also observed a higher prevalence of diabetes and hy- the PRECISE-­DAPT score contains 25 categories for
pertension among Black patients in our cohort. GFR and 19 for age, compared with between 2 and 5,
In the present study, severe CKD alone did not en- respectively, for PARIS and 3 and 2, respectively, for
tirely explain the numerical difference in postdischarge ARC-­HBR. This contributed to large differences in the
bleeding between Black and White adults, and other proportions of patients classified as high risk between
unmeasured variables must have contributed. We only scores: 35.4% for the PRECISE-­DAPT score, 15.8% for
included clinical variables that have been consistently the PARIS score, and 46.7% for the ARC-­HBR score.
and repeatedly associated with an increased risk of Because so few patients were deemed high risk by
bleeding. Socioeconomic and structural differences the PARIS score, the incidence of bleeding among pa-
between these groups almost certainly contribute to tients in the quartile below the high-­risk quartile was
higher rates of bleeding as well as the development 20.6 per 100 person-­years compared with 10.4 per
of diabetes, hypertension, and subsequently CKD 100 person-­years for the PRECISE-­DAPT score. The
among Black patients. An analysis of the National ARC-­HBR score simply classified so many patients
Cardiovascular Data Acute Coronary Treatment and as being high risk that it is not clinically useful. Finally,
Intervention Outcomes Network Registry found zip the PARIS and ARC-­ HBR scores both include oral
code, as a surrogate for socioeconomic status, to anticoagulation, whereas the PRECISE-­ DAPT score
be associated with major bleeding events after mul- does not. Oral anticoagulation is already included in
tivariable adjustment (odds ratio, 1.10 [95% CI, 1.05–­ guideline-­recommended algorithms for choosing the
1.16]).27 We did not report differences in variables that duration of DAPT, and therefore its inclusion in a risk
demonstrate this structural bias because no single so- score is not beneficial. For these reasons, we believe
cioeconomic variable has been repeatedly associated that the PRECISE-­DAPT score should be used for risk-­
with DAPT-­associated bleeding, and socioeconomic stratifying patients taking DAPT.

J Am Heart Assoc. 2022;11:e024412. DOI: 10.1161/JAHA.121.0244128

 Heindl et al Bleeding Risk Prediction Among Black Adults

Limitations were present in our study. First, self-­ REFERENCES

reported race is a social construct, and we did not report 1. Levine GN, Bates ER, Bittl JA, Brindis RG, Fihn SD, Fleisher LA,
the social determinants of health that contribute to dispa- Granger CB, Lange RA, Mack MJ, Mauri L, et al. 2016 ACC/AHA
guideline focused update on duration of dual antiplatelet therapy in pa-
rate outcomes. The definitions for socioeconomic vari- tients with coronary artery disease: a report of the American College
ables remain heterogenous and without clear standards, of Cardiology/American Heart Association task force on clinical prac-
and we felt that including these variables would not pro- tice guidelines: an update of the 2011 ACCF/AHA/SCAI guideline for
percutaneous coronary intervention, 2011 ACCF/AHA guideline for
duce clinical value. Second, past medical history items coronary artery bypass graft surgery, 2012 ACC/AHA/ACP/AATS/
were obtained by screening the patient’s medical record, PCNA/SCAI/STS guideline for the diagnosis and management of pa-
yet some data elements may have not been recorded tients with stable ischemic heart disease, 2013 ACCF/AHA guideline
for the management of ST-­elevation myocardial infarction, 2014 AHA/
in the medical record. Third, 4 of the ARC-­HBR crite- ACC guideline for the management of patients with non-­ST-­elevation
ria were excluded because the data required for these acute coronary syndromes, and 2014 ACC/AHA guideline on perioper-
variables were incomplete. However, the prevalence of ative cardiovascular evaluation and management of patients undergo-
ing noncardiac surgery. Circulation. 2016;134:e123–­e155. doi: 10.1161/
these variables in the general population is low, and their CIR.0000000000000404
inclusion would likely not have affected the performance 2. Navarese EP, Andreotti F, Schulze V, Kołodziejczak M, Buffon A,
of the score. Finally, this was primarily a descriptive and Brouwer M, Costa F, Kowalewski M, Parati G, Lip GY. Optimal duration
of dual antiplatelet therapy after percutaneous coronary intervention
exploratory analysis with substantial probability of type with drug eluting stents: meta-­analysis of randomised controlled trials.
I error greater than nominal (α=0.05 for some of the hy- BMJ. 2015;350:h1618. doi: 10.1136/bmj.h1618
pothesis tests) because of multiple comparisons. 3. Amin AP, Bachuwar A, Reid KJ, Chhatriwalla AK, Salisbury AC, Yeh RW,
Kosiborod M, Wang TY, Alexander KP, Gosch K. Nuisance bleeding
In conclusion, self-­reported Black patients did not with prolonged dual antiplatelet therapy after acute myocardial infarc-
have a statistically higher risk of postdischarge bleed- tion and its impact on health status. J Am Coll Cardiol. 2013;61:2130–­
ing while on DAPT compared with White patients. The 2138. doi: 10.1016/j.jacc.2013.02.044
4. Généreux P, Giustino G, Witzenbichler B, Weisz G, Stuckey TD, Rinaldi
PRECISE-­ DAPT, PARIS, and ARC-­ HBR risk scores MJ, Neumann F-­J, Metzger DC, Henry TD, Cox DA. Incidence, predic-
performed similarly among both Black and White pa- tors, and impact of post-­discharge bleeding after percutaneous coro-
tients, although more Black patients were classified as nary intervention. J Am Coll Cardiol. 2015;66:1036–­1045. doi: 10.1016/j.
jacc.2015.06.1323
high risk by the PRECISE-­DAPT and PARIS scores. 5. Costa F, van Klaveren D, James S, Heg D, Räber L, Feres F, Pilgrim T,
Clinicians should not consider self-­reported Black race Hong M-­K, Kim H-­S, Colombo A. Derivation and validation of the pre-
in their assessment of bleeding risk, although they dicting bleeding complications in patients undergoing stent implanta-
tion and subsequent dual antiplatelet therapy (PRECISE-­DAPT) score: a
should consider age, GFR <30 mL/min per 1.73 m2, pooled analysis of individual-­patient datasets from clinical trials. Lancet.
anticoagulant use, ticagrelor or prasugrel use, and
Downloaded from http://ahajournals.org by on September 21, 2022

2017;389:1025–­1034. doi: 10.1016/S0140-­6736(17)30397-­5

prior bleeding, as these were the strongest predictors. 6. Baber U, Mehran R, Giustino G, Cohen DJ, Henry TD, Sartori S, Ariti C,
Litherland C, Dangas G, Gibson CM. Coronary thrombosis and major
The PRECISE-­DAPT score better characterized differ- bleeding after PCI with drug-­eluting stents: risk scores from PARIS. J
ences in GFR and thus may be the most appropriate Am Coll Cardiol. 2016;67:2224–­2234. doi: 10.1016/j.jacc.2016.02.064
score for this indication among the US population. 7. Urban P, Mehran R, Colleran R, Angiolillo DJ, Byrne RA, Capodanno D,
Cuisset T, Cutlip D, Eerdmans P, Eikelboom J, et al. Defining high bleed-
ing risk in patients undergoing percutaneous coronary intervention: a
consensus document from the academic research consortium for high
ARTICLE INFORMATION bleeding risk. Eur Heart J. 2019;40:2632–­2653. doi: 10.1093/eurheartj/
ehz372
Received November 25, 2021; accepted July 7, 2022. 8. Collet J-­P, Thiele H, Barbato E, Barthélémy O, Bauersachs J, Bhatt DL,
Dendale P, Dorobantu M, Edvardsen T, Folliguet T. 2020 ESC guidelines
Affiliations for the management of acute coronary syndromes in patients present-
Division of Cardiovascular Diseases, Department of Medicine (B.H., S.C., ing without persistent ST-­segment elevation: the task force for the man-
V.P., P.A.); and Department of Neurology (C.D., R.N., E.U., N.L.), University agement of acute coronary syndromes in patients presenting without
of Alabama at Birmingham, Birmingham, AL; Department of Data Science, persistent ST-­segment elevation of the European Society of Cardiology
School of Public Health, University of Mississippi Medical Center, Jackson, (ESC). Eur Heart J. 2021;42:1289–­1367. doi: 10.1093/eurheartj/ehaa575
MS (W.H.); Department of Epidemiology (M.R.I.); Department of Biostatistics 9. Mak K-­H, Bhatt DL, Shao M, Hankey GJ, Easton JD, Fox KA, Topol EJ.
(G.Z.) and School of Public Health, University of Alabama at Birmingham, Ethnic variation in adverse cardiovascular outcomes and bleeding com-
Birmingham, AL (M.B.). plications in the clopidogrel for high atherothrombotic risk and ischemic
stabilization, management, and avoidance (CHARISMA) study. Am
Sources of Funding Heart J. 2009;157:658–­665. doi: 10.1016/j.ahj.2008.08.031
This work was supported by the National Institutes of Health (T32HL129948, 10. Kobayashi T, Glorioso TJ, Armstrong EJ, Maddox TM, Plomondon ME,
RO1HL092173, and K24HL133373), Clinical and Translational Science Award Grunwald GK, Bradley SM, Tsai TT, Waldo SW, Rao SV. Comparative
UL1TR000165, University of Alabama Birmingham’s Health Service Foundations’ outcomes after percutaneous coronary intervention among black and
General Endowment Fund, and the Hugh Kaul Personalized Medicine Institute. white patients treated at US veterans affairs hospitals. JAMA Cardiol.
Hillegass is supported by Institutional Development Award (IDeA) Networks for 2017;2:967–­975. doi: 10.1001/jamacardio.2017.2180
Clinical and Translational Research (IDeA-­CTR) (U54GM115428). 11. Ueki Y, Bär S, Losdat S, Otsuka T, Zanchin C, Zanchin T, Gragnano F,
Gargiulo G, Siontis G, Praz F. Validation of bleeding risk criteria (ARC-­
Disclosures HBR) in patients undergoing percutaneous coronary intervention and
None. comparison with contemporary bleeding risk scores. Eurointervention.
2020;16:371–­379. doi: 10.4244/EIJ-­D-­20-­0 0052
Supplemental Material 12. Abu-­Assi E, Raposeiras-­Roubin S, Cobas-­Paz R, Caneiro-­Queija B,
Tables S1–­S9 Martínez-­Reglero C, Rodríguez-­Rodríguez JM, Baz A, Iniguez-­Romo
Figures S1–­S2 A. Assessing the performance of the PRECISE-­DAPT and PARIS risk

J Am Heart Assoc. 2022;11:e024412. DOI: 10.1161/JAHA.121.0244129

 Heindl et al Bleeding Risk Prediction Among Black Adults

scores for predicting one-­year out-­of-­hospital bleeding in acute cor- 23. Saran R, Robinson B, Abbott KC, Agodoa LY, Albertus P, Ayanian
onary syndrome patients. EuroIntervention. 2018;13:1914–­1922. doi: J, Balkrishnan R, Bragg-­Gresham J, Cao J, et al. US renal data sys-
10.4244/EIJ-­D-­17-­0 0550 tem 2016 annual data report: epidemiology of kidney disease in the
13. Sardar MR, Badri M, Prince CT, Seltzer J, Kowey PR. Underrepresentation United States. Am J Kidney Dis. 2017;69:A7–­A8. doi: 10.1053/j.
of women, elderly patients, and racial minorities in the randomized trials ajkd.2016.12.004
used for cardiovascular guidelines. JAMA Inter Med. 2014;174:1868–­ 24. Wang L, Li X, Wang Z, Bancks MP, Carnethon MR, Greenland P, Feng
1870. doi: 10.1001/jamainternmed.2014.4758 YQ, Wang H, Zhong VW. Trends in prevalence of diabetes and con-
14. Levey AS, Stevens LA, Schmid CH, Zhang YL, Castro AF III, Feldman trol of risk factors in diabetes among US adults, 1999-­2018. JAMA.
HI, Kusek JW, Eggers P, Van Lente F, Greene T, et al. A new equation to 2021;326:1–­13. doi: 10.1001/jama.2021.9883
estimate glomerular filtration rate. Ann Intern Med. 2009;150:604– ­612. 25. Muntner P, Hardy ST, Fine LJ, Jaeger BC, Wozniak G, Levitan EB,
doi: 10.7326/0003-­4819-­150-­9 -­200905050-­0 0006 Colantonio LD. Trends in blood pressure control among US adults with
15. Mehran R, Rao SV, Bhatt DL, Gibson CM, Caixeta A, Eikelboom J, Kaul hypertension, 1999-­2000 to 2017-­2018. JAMA. 2020;324:1190–­1200.
S, Wiviott SD, Menon V, Nikolsky E. Standardized bleeding definitions doi: 10.1001/jama.2020.14545
for cardiovascular clinical trials: a consensus report from the bleeding 26. Genovese G, Friedman DJ, Ross MD, Lecordier L, Uzureau P, Freedman
academic research consortium. Circulation. 2011;123:2736–­2747. doi: BI, Bowden DW, Langefeld CD, Oleksyk TK, Knob ALU. Association of
10.1161/CIRCULATIONAHA.110.009449 trypanolytic ApoL1 variants with kidney disease in African Americans.
16. Harrell FE Jr, Lee KL, Mark DB. Multivariable prognostic models: issues Science. 2010;329:841–­845. doi: 10.1126/science.1193032
in developing models, evaluating assumptions and adequacy, and mea- 27. Udell JA, Desai NR, Li S, Thomas L, de Lemos JA, Wright-­Slaughter
suring and reducing errors. Stat Med. 1996;15:361–­387. doi: 10.1002/ P, Zhang W, Roe MT, Bhatt DL. Neighborhood socioeconomic
(SICI)1097-­0258(19960229)15:4<361::AID-­SIM168>3.0.CO;2-­4 disadvantage and care after myocardial infarction in the National
17. Dean AG, Sullivan KM, Soe MM. OpenEpi: Open source epidemiologic Cardiovascular Data Registry. Circulation. 2018;11:e004054. doi:
statistics for public health, Version. 2021. Available at: https://www. 10.1161/CIRCOUTCOMES.117.004054
OpenE​pi.com. Accessed August 30, 2022. 28. Choi KH, Song YB, Lee JM, Park TK, Yang JH, Choi JH, Choi
18. Sullivan LT II, Mulder H, Chiswell K, Shaw LK, Wang TY, Jackson LR SH, Oh JH, Cho DK, Lee JB, et al. Clinical usefulness of
II, Thomas KL. Racial differences in long-­term outcomes among black PRECISE-­DAPT score for predicting bleeding events in patients
and white patients with drug-­eluting stents. Am Heart J. 2019;214:46–­ with acute coronary syndrome undergoing percutaneous coro-
53. doi: 10.1016/j.ahj.2019.04.005 nary intervention: an analysis from the SMART-­ DATE random-
19. Faggioni M, Baber U, Chandrasekhar J, Sartori S, Weintraub W, Rao ized trial. Circ Cardiovasc Interv. 2020;13:e008530. doi: 10.1161/
SV, Vogel B, Claessen B, Kini A, Effron M. Use of prasugrel and clinical CIRCINTERVENTIONS.119.008530
outcomes in African-­American patients treated with percutaneous cor- 29. Choi SY, Kim MH, Cho YR, Sung Park J, Min Lee K, Park
onary intervention for acute coronary syndromes. Catheter Cardiovasc TH, Yun SC. Performance of PRECISE-­ D APT score for pre-
Interv. 2019;94:53–­60. doi: 10.1002/ccd.28033 dicting bleeding complication during dual antiplatelet ther-
20. Cai A, Dillon C, Hillegass WB, Beasley M, Brott BC, Bittner VA, Perry apy. Circ Cardiovasc Interv. 2018;11:e006837. doi: 10.1161/
GJ, Halade GV, Prabhu SD, Limdi NA. Risk of major adverse cardiovas- CIRCINTERVENTIONS.118.006837
cular events and major hemorrhage among white and black patients 30. Choi SY, Kim MH, Lee KM, Ko YG, Yoon CH, Jo MK, Yun SC.
undergoing percutaneous coronary intervention. J Am Heart Assoc. Comparison of performance between ARC-­HBR criteria and PRECISE-­
2019;8:e012874. doi: 10.1161/JAHA.119.012874 DAPT score in patients undergoing percutaneous coronary interven-
21. Wallentin L, Becker RC, Budaj A, Cannon CP, Emanuelsson H, Held tion. J Clin Med. 2021;10:2566. doi: 10.3390/jcm10122566
Downloaded from http://ahajournals.org by on September 21, 2022

C, Horrow J, Husted S, James S, Katus H, et al. Ticagrelor versus 31. Chen JY, Zhang AD, Lu HY, Guo J, Wang FF, Li ZC. CHADS2 versus
clopidogrel in patients with acute coronary syndromes. N Engl J Med. CHA2DS2-­VASc score in assessing the stroke and thromboembolism
2009;361:1045–­1057. doi: 10.1056/NEJMoa0904327 risk stratification in patients with atrial fibrillation: a systematic review
22. Boccardo P, Remuzzi G, Galbusera M. Platelet dysfunction in renal failure. and meta-­analysis. J Geriatr Cardiol. 2013;10:258–­266. doi: 10.3969/j.
Semin Thromb Hemost. 2004;30:579–­589. doi: 10.1055/s-­2004-­835678 issn.1671-­5411.2013.03.004

J Am Heart Assoc. 2022;11:e024412. DOI: 10.1161/JAHA.121.02441210

SUPPLEMENTAL MATERIAL
Downloaded from http://ahajournals.org by on September 21, 2022
 Table S1. Bleeding Academic Research Consortium (BARC) definition for bleeding events.
Type 0 No bleeding.
Bleeding that is not actionable and does not cause the patient to seek unscheduled
performance of studies, hospitalization, or treatment by a healthcare professional; may
Type 1
include episodes leading to self-discontinuation of medical therapy by the patient without
consulting a healthcare professional.
Any clinically overt sign of hemorrhage (more bleeding than would be expected for a
clinical circumstance) that is actionable but does not meet criteria for type 3, 4, or 5 BARC
bleeding.
Type 2 The bleeding must require diagnostic studies, hospitalization, or treatment by a healthcare
professional.
Examples include: Hematocrit testing, hemoccult testing, endoscopy, colonoscopy,
computed tomography scanning, or Urinalysis
Overt bleeding plus hemoglobin drop of 3 to <5 g/dL(provided hemoglobin drop is related
Type 3a to bleed)
Any transfusion with overt bleeding.
Overt bleeding plus hemoglobin drop ≥5 g/dL (provided hemoglobin drop is related to
bleed)
Cardiac tamponade
Type 3b
Bleeding requiring surgical intervention for control (excluding
dental/nasal/skin/hemorrhoid)
Bleeding requiring intravenous vasoactive agents.
Intracranial hemorrhage (does not include microbleeds or hemorrhagic transformation,
Downloaded from http://ahajournals.org by on September 21, 2022

does include intraspinal)

Type 3c
Subcategories confirmed by autopsy or imaging or lumbar puncture
Intraocular bleed compromising vision
CABG-related bleeding
Perioperative intracranial bleeding within 48 hours
Type 4 Reoperation after closure of sternotomy for the purpose of controlling bleeding
Transfusion of ≥5 units whole blood or packed red blood cells within a 48 hour period
Chest tube output ≥2 L within a 24 hour period
Type 5a Probable fatal bleeding; no autopsy or imaging confirmation but clinically suspicious
Type 5b Definite fatal bleeding; overt bleeding or autopsy or imaging confirmation
 Table S2. Variables included in the Patterns of Nonadherence to Antiplatelet Regimens in Stented
Patients (PRECISE-DAPT) score, with associated point assignments.
Estimated glomerular White Blood Cell Previous bleed
Age Hemoglobin
filtration rate Count requiring medical
(years) (g/dL)
(mL/min/1.73m2) (103 cells per uL) attention
≤51.49 (0) >98 (0) ≥ 12.0 (0) ≤ 5.485 (0) No (0)
51.50-53.49 (1) 94.1-98.0 (1) 11.9 (1) 5.486-6.455 (1) Yes (25)
53.50-55.49 (2) 90.2-94.0 (2) 11.8 (2) 6.456-7.425 (2)
55.50-57.49 (3) 86.3-90.1 (3) 11.7 (2) 7.426-8.395 (3)
57.50-59.49 (4) 82.3-86.2 (4) 11.6 (3) 8.396-9.366 (4)
59.50-61.49 (5) 78.4-82.1 (5) 11.5 (4) 9.367-10.355 (5)
61.50-63.49 (6) 74.5-78.3 (6) 11.4 (5) 10.356-11.305 (6)
63.50-65.49 (7) 70.5-74.4 (7) 11.3(5) 11.306-12.275 (7)
65.50-68.49 (8) 66.6-70.4 (8) 11.2 (6) 12.276-13.245 (8)
68.50-70.49 (9) 62.7-66.5 (9) 11.1 (7) 13.246-14.215 (9)
70.50-72.49 (10) 58.7-62.6 (10) 11.0 (8) 13.216-15.285 (10)
72.50-74.49 (11) 54.8-58.6 (11) 10.9 (8) 15.186-16.155 (11)
74.50-75.49 (12) 50.8-54.7 (12) 10.8 (9) 16.156-17.125 (12)
76.50-78.49 (13) 46.9-50.7 (13) 10.7 (10) 17.126-18.095 (13)
Downloaded from http://ahajournals.org by on September 21, 2022

78.50-80.49 (14) 43.0-46.8 (14) 10.6 (11) 18.096-19.065 (14)

80.50-83.49 (15) 39.0-42.9 (15) 10.5 (11) ≥19.065 (15)
83.50-86.49 (16) 35.1-38.9 (16) 10.4 (12)
86.50-87.49 (17) 31.2-35.0 (17) 10.3 (13)
87.50-89.49 (18) 27.2-31.1 (18) 10.2 (13)
≥89.50 (19) 23.3-27.1 (19) 10.1 (14)
19.4-23.2 (20) ≤10.0 (15)
15.4-19.3 (21)
11.5-15.3 (22)
7.6-11.4 (23)
3.6-7.5 (24)
<3.6 (25)
 Table S3. Variables included in the Patterns of Nonadherence to Antiplatelet Regimens in Stented
Patients (PARIS) risk score, with associated point assignments.
Variable Increment
Variable
(Assigned Points)
<50 50-59 60-69 70-79 ≥ 80
Age (years)
(0) (1) (2) (3) (4)
No Yes
Current smoking
(0) (2)
<25 25-34.9 ≥35
Body mass index (kg/m2)
(2) (0) (2)
Estimated glomerular
Absent Present
filtration rate <60
(0) (2)
mL/min/1.73m2
No Yes
Hemoglobin <12 g/dL
(0) (3)
Triple therapy (aspirin,
No Yes
P2Y12 inhibitor, and
(0) (2)
anticoagulant) on discharge
Downloaded from http://ahajournals.org by on September 21, 2022
 Table S4. Academic Research Consortium Criteria for High Bleeding Risk, with modifications made for the present study.
Major Criteria Minor Criteria Modifications for the Present Study*
Age ≥ 75 years ~
Anticipated use of long-term oral anticoagulation ~
Severe or end-stage chronic kidney disease (eGFR <30 ml/min) Moderate chronic kidney disease (eGFR 30-59 mL/min) ~
Hemoglobin <11 g/dL Hemoglobin 11-12.9 g/dL for men and 11-11.9 g/dL for women ~
Spontaneous bleeding requiring hospitalization or transfusion in Spontaneous bleeding requiring hospitalization or transfusion
~
the past 6 months or at any time, if recurrent within the past 12 months not meeting the major criteria
Moderate or severe baseline thrombocytopenia (platelet count
~
<100x109/L)†
Chronic bleeding diathesis Data unavailable‡
Liver cirrhosis with portal hypertension ~
Defined as having one of these medications
Long-term use of oral NSAIDs or steroids
prescribed on admission and at discharge.
Downloaded from http://ahajournals.org by on September 21, 2022

Active malignancy (excluding nonmelanoma skin cancer)

Data unavailable‡
within the past 12 months§
Previous spontaneous ICH (at any time) Both traumatic ICH and spontaneous ICH were
Previous traumatic ICH within the past 12 months included as one major criteria.
Presence of a brain arteriovenous malformation ~
Stroke was not distinguished, and any past
Moderate or severe ischemic stroke within the past 6 months Any ischemic stroke at any time not meeting the major criterion
ischemic stroke was labeled as a minor criteria.
Nodeferrable major surgery on dual antiplatelet therapy Data unavailable‡
Recent major surgery or major trauma within 30 days before
Data unavailable‡
percutaneous coronary intervention
Definitions: eGFR, estimated glomerular filtration rate; ICH, intracranial hemorrhage; NSAID, nonsteroidal anti-inflammatory drug;
*
Fields with ~ represents variables for which no modification was necessary. †Baseline thrombocytopenia was defined by ARC-HBR as
thrombocytopenia before PCI. ‡Not included as an ARC-HBR variable for the present study. §Active malignancy was defined by ARC-HBR as
diagnosis within 12 months and/or ongoing requirement for treatment (including surgery, chemotherapy, or radiotherapy).
 Table S5. Number and anatomic location of events by Bleeding Academic Research Consortium (BARC)
category.
All patients Black patients White patients
Location of bleeding event
Number of events (proportion within category)

Total BARC 2 events 115 27 88

Hematoma 31 (27.0%) 8 (29.6%) 23 (26.1%)

Gastrointestinal 28 (24.3%) 8 (29.6%) 20 (22.7%)

Genitourinary 13 (11.3%) 1 (3.7%) 12 (13.6%)

Epistaxis 13 (11.3%) 2 (7.4%) 11 (12.5%)

Bruise 10 (8.7%) 2 (7.4%) 8 (9.1%)

Hemoptysis 3 (2.6%) 0 3 (3.4%)

Gingival 1 (0.9%) 0 1 (1.1%)

Not recorded 16 (13.9%) 6 (22.2%) 10 (11.4%)

Total BARC 3 events 41 16 25

Downloaded from http://ahajournals.org by on September 21, 2022

Gastrointestinal 28 (24.3%) 10 (37.0%) 18 (20.5%)

Genitourinary 3 (2.6%) 0 3 (12.0%)

Hematoma 1 (0.9%) 0 1 (4.0%)

Epistaxis 1 (0.9%) 1 (6.3%) 0

Gingival 1 (0.9%) 1 (6.3%) 0

Intracranial hemorrhage 1 (0.9%) 0 1 (4.0%)

Retroperitoneal Hemorrhage 1 (0.9%) 0 1 (4.0%)

Not recorded 5 (4.3%) 4 (25%) 1 (4.0%)

Total BARC 4 events 0 0 0

Total BARC 5 events 3 1 2

Intracranial hemorrhage 1 (33%) 0 1 (50.0%)

Retroperitoneal hematoma 2 (66.6%) 1 (100.0%) 1 (50.0%)

 Table S6. Unadjusted and adjusted association of bleeding risk factors on Bleeding Academic Research
Consortium 2-5 bleeding, accounting for the competing risk of death.
Unadjusted HR P- Adjusted HR
P-value
(95% CI) value (95% CI)

1.42 1.37
Self-reported Black race 0.047 0.106
(1.01-2.01) (0.94-2.01)

1.01 1.02
Age (per year) 0.061 0.047
(1.00-1.03) (1.00-1.03)

GFR <30 mL/min/1.73m2 or 0.39 0.52

<0.001 0.013
end-stage renal disease (0.26-0.59) (0.31-0.87)

Previous bleeding requiring 2.42 1.92

<0.001 0.011
medical attention (1.55-3.76) (1.16-3.17)

1.81 2.07
Ticagrelor or prasugrel use <0.001 <.0.001
(1.33-2.48) (1.49-2.87)

2.37 2.28
Anticoagulant use <0.001 <.0.001
(1.65-3.40) (1.54-3.37)

0.87 0.95
Hemoglobin (g/dL) 0.001 0.388
(0.79-0.95) (0.86-1.06)
Downloaded from http://ahajournals.org by on September 21, 2022

1.25 1.07
Female sex 0.175 0.725
(0.90-1.74) (0.74-1.54)

1.52 1.19
Prior ischemic stroke or TIA 0.036 0.409
(1.03-2.25) (0.78-1.82)

Liver cirrhosis with portal 1.96 1.66

0.137 0.345
hypertension (0.81-4.75) (0.58-4.72)

1.24 1.02
Proton pump inhibitor use 0.181 0.908
(0.91-1.70) (0.73-1.44)

0.81 0.96
Current smoking (yes) 0.290 0.866
(0.55-1.20) (0.62-1.50)

1.06 1.06
Body Mass Index ≥35 kg/m2 0.761 0.771
(0.72-1.56) (0.71-1.60)

1.01 1.01
White cell count (per 103/uL) 0.417 0.448
(0.98-1.05) (0.98-1.05)

1.00 1.00
Platelet count (per 109/L) 0.640 0.507
(1.00-1.00) (1.00-1.00)

Diabetes history 1.09 0.580 0.93 0.681

 (0.80-1.49) (0.67-1.30)

1.03 0.86
Hypertension 0.898 0.561
(0.66-1.62) (0.52-1.42)

0.55 0.47
Long-term NSAID use 0.228 0.184
(0.20-1.46) (0.15-1.44)
Abbreviations: CI, confidence intervals; GFR, glomerular filtration rate. HR, hazard ratio. *The variables
included in the adjusted model were those with an unadjusted p-value <0.2.
Downloaded from http://ahajournals.org by on September 21, 2022
 Table S7. The effect of self-reported Black race on Bleeding Academic Research Consortium 2-5 bleeding, with individual adjustment for
predictors of bleeding.*
Hazard ratio for self- Adjusted % change p-value for the
Variables in the model, other
reported Black race p-value in effect size from presence of
than self-reported Black race
(95% CI) the baseline model ‡ mediation§

1.46
Age and sex (baseline model)† 0.037 referent referent
(1.02-2.09)

1.30
GFR <30, age, and sex 0.158 -34.8% <0.001
(0.90-1.88)

Ticagrelor or prasugrel use, 1.53

0.020 15.2% 0.186
age, and sex (1.07-2.18)

1.52
Anticoagulant use, age, and sex 0.022 13.0% 0.817
(1.06-2.18)

Previous bleed requiring 1.44

Downloaded from http://ahajournals.org by on September 21, 2022

0.047 -4.3% 0.210

medical attention, age, and sex (1.00-2.05)
Abbreviations: CI, confidence intervals; GFR, glomerular filtration rate; HR, hazard ratio. *Variables found to be significant after multivariable
adjustment (p<0.05) †Age and sex were chosen as variables for the baseline model as they are commonly associated with exposure and outcome
differences. ‡The equation used for this column was: (1 – Adjusted HRBlack race) - (1 – Baseline HRBlack race) / (1 – Baseline HRBlack race) x 100.
§
Derived from Sobel’s test.
 Table S8. Unadjusted and adjusted effects of three forms of the glomerular filtration rate (GFR) variable on Bleeding Academic Research
Consortium 2-5 bleeding, and the mediating effect of each form on the association between self-reported Black Race and bleeding.

Adjusted* effect of GFR on

Mediation effect of GFR on the association between self-reported Black race and bleeding
BARC 2-5 Bleeding

Hazard ratio for Adjusted† hazard ratio for Change (%) in effect
GFR variable p-value for the presence
GFR P-value self-reported Black race P-value size compared with the
form of mediation§
(95% CI) (95% CI) baseline model ‡

1.46
~ ~ ~ 0.037 referent referent
(1.02-2.09)

GFR <30 2.41 1.30

<0.001 0.158 -33.8% 0.003
mL/min/1.73m2 (1.54-3.76) (0.90-1.88)

GFR <45 1.65 1.36

0.012 0.096 -20.0% 0.043
mL/min/1.73m2 (1.12-2.45) (0.95-1.96)
Downloaded from http://ahajournals.org by on September 21, 2022

GFR (per 0.99 1.36

0.057 0.098 -20.0% 0.159
mL/min/1.73m2) (0.99-1.00) (0.94-1.97)
*
Adjusted for age, sex, prior bleeding, ticagrelor or prasugrel use, and anticoagulant use. †Adjusted for age, sex, and the GFR form described in the
first column. The equation used for this column was: (1 – Adjusted HRBlack race) - (1 – Baseline HRBlack race) / (1 – Baseline HRBlack race) x 100.
§
Derived from Sobel’s test.
 Table S9. Number of patients, person-years, and events within categories of PRECISE-DAPT, PARIS,
and ARC-HBR risk, as well as unadjusted comparisons between categories, among Black and White
patients separately.
PRECISE-DAPT Risk Score
Points
composing risk ≤11 12-19 20-27 >28
category
Number of patients per risk group

Black Patients 98 86 56 102

White Patients 317 333 261 276

Person-years of follow-up per risk group

Black Patients 65.0 55.6 36.7 56.0

White Patients 228.0 240.0 175.3 170.7

Number of events per risk group

Black Patients 8 6 9 21

White Patients 23 25 26 41
Unadjusted hazard ratio
(95% confidence intervals), p-value
0.85 1.96 2.87
Black patients referent
Downloaded from http://ahajournals.org by on September 21, 2022

(0.30-2.46, p=0.769) (0.76-5.08, p=0.166) (1.27-6.47, p=0.011)

1.03 1.45 2.28
White patients referent
(0.59-1.82, p=0.907) (0.83-2.55, p=0.192) (1.37-3.80, p=0.002)
PARIS Risk Score
Points
composing risk ≤2 3-4 5-6 ≥7
category
Number of patients per risk group

Black patients 77 102 80 83

White patients 321 350 299 217

Person-years of follow-up per risk group

Black patients 54.0 68.5 44.5 46.2

White patients 241.2 247.4 193.0 132.3

Number of events per risk group

Black patients 5 10 16 13

White patients 21 28 33 33
Unadjusted hazard ratio
(95% confidence intervals), p-value
 1.56 3.51 2.89
Black patients Referent
(0.53-4.57, p=0.416) (1.28-9.58, p=0.014) (1.28-9.58, p=0.044)
1.28 1.89 2.68
White patients Referent
(0.73-2.25, p=0.397) (1.09-3.24, p=0.024) (1.55-4.64, p<0.001)
ARC-HBR Risk Score

Risk Categories Not high risk High risk

Number of patients per risk group

Black patients 177 165

White patients 643 544

Person-years of follow-up per risk group

Black patients 116.2 97.0

White patients 468.7 345.2

Number of events per risk group

Black patients 14 30

White patients 43 72
Unadjusted hazard ratio
(95% confidence intervals), p-value
2.45
Black patients referent
(1.30-4.62, p=0.006)
Downloaded from http://ahajournals.org by on September 21, 2022

2.18
White patients referent
(1.50-3.19, p<0.001)
Abbreviations: BARC, Bleeding Academic Research Consortium; PARIS, Patterns of Nonadherence to
Antiplatelet Regimens in Stented Patients; PRECISE-DAPT, PREdicting bleeding Complications In
patients undergoing Stent implantation and subsEquent Dual Anti Platelet Therapy. The PRECISE-DAPT
and PARIS risk scores have three categories of risk based on a point system, but the ARC-HBR schema
has a binary design, with only high and not-high risk categories.
 Figure S1. Unadjusted time-to-event curves for the endpoint of BARC 2-5 bleeding for all patients, and
between Black and White patients individually.

1.00
Unadjusted Survival for BARC 2-5 Bleeding

0.95

0.90

0.85
Black patients
White patients
0.80
Hazard ratio 1.41
95% Confidence intervals: 1.00-2.00; p=0.052

0.75
0 50 100 150 200 250 300 350
Downloaded from http://ahajournals.org by on September 21, 2022

Days

Abbreviations: BARC, Bleeding Academic Research Consortium. Survival estimates were generated by
applying the PHREG procedure to produce an unadjusted cox proportional hazard model (PROC
PHREG, SAS software, version 9.4).
 Figure S2. Adjusteda time-to-event curves for BARC 2-5 bleeding for all patients, and between Black and
White patients individually.

1.00
Adjusted Survival for BARC 2-5 Bleeding

0.95

0.90

0.85

Black patients
White patients
0.80
Hazard ratio 1.37
95% Confidence intervals: 0.94-1.99; p=0.098
Downloaded from http://ahajournals.org by on September 21, 2022

0.75
0 50 100 150 200 250 300 350
Days
Abbreviations: BARC, Bleeding Academic Research Consortium. Survival estimates were generated by
applying the PHREG procedure to produce a cox proportional hazard model. aAdustment was made for
age (per year), GFR <30 mL/min/1.73m2, previous bleeding requiring medical attention, ticagrelor or
prasugrel use, and anticoagulant use.