Safety and Effectiveness of Contemporary P2Y12 Inhibitors in an East Asian Population With Acute Coronary Syndrome: A Nationwide Population-Based Cohort Study.

Background Prior reports indicate that the effect of P2Y12 inhibitors may be different in East Asian patients ("East Asian paradox"); therefore, understanding the outcomes associated with potent P2Y12 inhibitors in different populations is clinically important. Methods and Results In this observational cohort study using administrative healthcare data sets, we compared safety and effectiveness of contemporary P2Y12 inhibitors in patients with acute coronary syndrome. The primary safety outcomes were major and any bleeding, and the primary effectiveness outcomes were major cardiovascular events (a composite of cardiovascular death, myocardial infarction, or stroke) and all-cause mortality. Among 70 715 patients with acute coronary syndrome, 56 216 (79.5%) used clopidogrel, 11 402 (16.1%) used ticagrelor, and 3097 (4.4%) used prasugrel. The median follow-up period was 18.0 months (interquartile range: 9.6-26.4 months). In a propensity-matched cohort, compared with clopidogrel, ticagrelor was associated with a higher risk of any bleeding (hazard ratio: 1.23; 95% CI, 1.14-1.33) but a lower risk of mortality (hazard ratio: 0.76; 95% CI, 0.63-0.91). Prasugrel, compared with clopidogrel, was associated with higher risks of any bleeding (hazard ratio: 1.23; 95% CI, 1.06-1.43) and major bleeding (hazard ratio: 1.50; 95% CI, 1.01-2.21) but a similar risk of effectiveness outcomes. No significant difference was noted between ticagrelor and prasugrel with respect to key safety or effectiveness outcomes. Several sensitivity analyses showed similar results. Conclusions In East Asian patients with acute coronary syndrome, compared with clopidogrel, ticagrelor was associated with an increased risk of bleeding but a decreased risk of mortality. Prasugrel was associated with an increase of any bleeding without difference in effectiveness outcomes. The risks of bleeding and ischemic events were similar between ticagrelor and prasugrel.

Clinical Perspective

What Is New?

Some previous reports indicate that the effect of P2Y12 inhibitors might be different in East Asian patients (“East Asian paradox”); therefore, understanding the outcomes associated with diverse P2Y12 inhibitors in different populations is clinically important.

This population‐based study was the first with a specific focus on East Asian patients with acute coronary syndrome to investigate the comparative safety and effectiveness of different oral P2Y12 inhibitors (clopidogrel, ticagrelor, and prasugrel).

What Are the Clinical Implications?

Compared with clopidogrel, ticagrelor was associated with increased rates of bleeding, a significant reduction in mortality rate, and no decrease in the rate of major cardiovascular events.

Compared with clopidogrel, prasugrel was associated with an increase in bleeding events but no differences in effectiveness outcomes.

No significant differences were noted between ticagrelor and prasugrel with respect to the rate of any bleeding and ischemic events, and further randomized clinical trials are necessary to confirm the findings of this study regarding different levels of risk for bleeding and ischemic events among different P2Y12 inhibitors.

Introduction

Dual‐antiplatelet therapy involving aspirin and a P2Y12 antagonist is the standard antithrombotic therapy in patients with acute coronary syndrome (ACS) and in those undergoing percutaneous coronary intervention (PCI).1 Given greater and more consistent platelet inhibition and a documented clinical benefit of newer P2Y12 antagonists (ticagrelor or prasugrel) over clopidogrel,2, 3 current European and US guidelines recommend that use of ticagrelor or prasugrel in preference to clopidogrel is reasonable for ACS patients with or without PCI.4, 5 However, compared with a Western population, a differential propensity for thromboembolic and bleeding risks in response to P2Y12 inhibitors was reported in an East Asian population (“East Asian paradox”).6, 7 Although East Asian ethnic groups are among the most populous (>1.5 billion people), few East Asian patients were included in the large, phase III, randomized controlled trials (RCTs) of potent P2Y12 antagonists for atherosclerotic cardiovascular disease.2, 3, 8, 9, 10, 11 Consequently, concerns exist regarding whether potent P2Y12 inhibitors have acceptable safety and efficacy profiles in an East Asian population with differential ischemic and bleeding tendency.

Although an RCT setting with strict inclusion and exclusion criteria is required to obtain high‐quality scientific evidence on the effects of antithrombotic drugs, well‐conducted postapproval observational studies might complement the RCTs and provide additional clinical information in diverse groups of patients or in clinical circumstances encountered in daily practice. In this study, we sought to evaluate the relative safety and effectiveness of contemporary P2Y12 inhibitors using a nationwide population‐based cohort of Korean patients presenting with ACS.

Methods

Data Sources

Anonymized data and study materials have been made publicly available. The analytic methods have been made available within the article to other researchers for purposes of reproducing the results or replicating the procedure.

This study is based on data from nationwide administrative claims–based databases of the National Health Insurance Service (NHIS), which is the universal health coverage system in South Korea. All residents must be enrolled in the NHIS either as an National Health Insurance beneficiary or a Medical Aid recipient. Consequently, these data sets can enable unrestricted collection of large ACS cohorts with information about medical visits and prescriptions and no specific inclusion or exclusion criteria apart from the beneficiary status, minimizing selection bias. The NHIS databases maintain comprehensive healthcare data sets for diagnoses, treatments, procedures, surgeries, prescriptions, hospital admissions, and discharge records of all insured patients who are reimbursed by the government according to the National Health Insurance Act.12 The prescription claims data identify dispensed prescriptions, including medications, date filled, days supplied, number of pills, and dosage. Medical claims include diagnostic and procedure information coded in accordance with the International Classification of Diseases, Tenth Revision (ICD‐10) for inpatient and outpatient encounters. Based on these data sets, we collected information on demographics, clinical covariates, all diagnostic and procedure information, study drugs, and concomitant cardioactive medications (for details, see Table S1). We also collected the available self‐reported medical history, smoking status, and general laboratory variables from the general health examination data, which were provided periodically by NHIS to all insured persons.13 The NHIS databases were validated in prior antithrombotic studies.14, 15

Study Population

We constructed a study cohort of adult patients who presented with ACS (ie, unstable angina or acute myocardial infarction [MI]) who had newly initiated P2Y12 inhibitors between January 1, 2013, and November 30, 2015 (Figure 1). A new‐user cohort design was used to compare patients who were prescribed clopidogrel, ticagrelor, or prasugrel as the initial treatment for ACS. Exclusion criteria were as follows: (1) prior use of any P2Y12 inhibitor in the 12 months preceding the index date, (2) concomitant use of anticoagulants, (3) receipt of fibrinolytic therapy, (4) history of any cancer before the index date, (5) cardiogenic shock, (6) no hospital admission for a principal diagnosis of ACS, and (7) use of antiplatelet drugs <30 days. We also excluded users of dual P2Y12 inhibitors. This study was approved by the institutional review board of the National Evidence‐Based Healthcare Collaborating Agency (no. NECAIRB16‐009‐2), and informed consent was waived.

Figure 1

Flowchart of the study population. ACS indicates acute coronary syndrome; f/u, follow‐up.

In Korea, the recommended dose of P2Y12 inhibitors for the management of ACS was identical for standard‐dose labeling: clopidogrel at 300‐ to 600‐mg loading dose, 75‐mg daily maintenance dose; ticagrelor at 180‐mg loading dose, 90‐mg twice maintenance dose; and prasugrel at a 60‐mg loading dose, 10‐mg daily maintenance dose.

Outcomes and Definition

The primary safety outcomes were any bleeding and major bleeding. Bleeding events were also assessed according to the site of the bleeding source. The primary effectiveness outcomes were major cardiovascular events and all‐cause mortality. Detailed definitions of the safety and effectiveness outcomes on the basis of ICD‐10 codes are summarized in Table S2.

Major bleeding was defined as a fatal bleeding event, bleeding necessitating hospitalization, or bleeding that occurred in the critical sites (intracranial, intraspinal, intra‐articular, intraocular, pericardial, retroperitoneal, or intramuscular with compartment syndrome).16 Any bleeding included intracranial bleeding, gastrointestinal bleeding, urogenital bleeding, respiratory bleeding (hemoptysis), nasal bleeding, intraocular bleeding, intra‐articular or intramuscular bleeding, and other types of bleeding. Major cardiovascular events were defined as the composite of cardiovascular death, MI, or stroke. Death certificate linkage data were provided by the Korean National Statistical Office. According to the ICD‐10 codes for primary cause of death, mortality was categorized into cardiovascular disease (disease of the circulatory system: I00–I99; sudden death: R96) and other (non–cardiovascular disease) causes (all other ICD‐10 codes).

Statistical Analysis

Given the differences in the baseline characteristics among eligible participants in the treatment groups, propensity‐score matching was used to identify a cohort of patients with similar baseline characteristics.17 In each cohort for comparison, the propensity score was estimated using a nonparsimonious logistic regression model,18 with the treatment group of P2Y12 inhibitors as the dependent variable and all the baseline characteristics outlined in Table 1 as covariates. Propensity‐score matching was performed using bootstrapping with 1:1 nearest neighbor matching without replacement (caliper distance of 0.2 SD of the pooled propensity scores) to identify matched cohorts representing the 2 treatment groups. Covariate balance was evaluated using standardized differences of means, and standardized differences of <10.0% for a given covariate indicate a relatively small imbalance.19

Table 1

Baseline Characteristics Before and After Propensity‐Score Matching Among Patients With Ticagrelor and Clopidogrel Usea

Characteristic	Before Matching			After Matching
	Ticagrelor (n=11 402)	Clopidogrel (n=56 216)	Standardized Difference (%)	Ticagrelor (n=11 402)	Clopidogrel (n=11 402)	Standardized Difference (%)
Age
Mean, y	60.9 (12.1)	65.4 (12.1)	37.6	60.9 (12.1)	60.8 (12.1)	0.5
≥75 y	1741 (15.3)	14 404 (25.6)	25.9	1741 (15.3)	1741 (15.3)	0.0
Sex
Male	8876 (77.9)	36 770 (65.4)	27.9	8876 (77.9)	8963 (78.6)	1.8
Female	2526 (22.1)	19 446 (34.6)	28.0	2526 (22.1)	2439 (21.4)	1.7
Socioeconomic status
Low tertile	3623 (31.8)	18 287 (32.5)	1.6	3623 (31.8)	3703 (32.5)	1.5
Middle tertile	3995 (35.0)	18 165 (32.3)	5.8	3995 (35.0)	3907 (34.3)	1.6
High tertile	3784 (33.2)	19 764 (35.2)	4.2	3784 (33.2)	3792 (33.3)	0.1
Body mass indexa
Mean	24.8 (2.7)	24.52 (2.7)	9.2	24.8 (2.7)	24.8 (2.7)	1.5
<20.0	392 (3.4)	2444 (4.4)	4.7	392 (3.4)	379 (3.3)	0.7
20.0 to <22.5	1458 (12.8)	7725 (13.7)	2.8	1458 (12.8)	1385 (12.2)	1.9
22.5 to <25.0	4275 (37.5)	23 719 (42.2)	9.6	4275 (37.5)	4211 (36.9)	1.2
25.0 to <27.5	3759 (33.0)	15 550 (27.7)	11.6	3759 (33.0)	3905 (34.3)	2.7
27.5 to <30.0	1053 (9.2)	4759 (8.5)	2.7	1053 (9.2)	1038 (9.1)	0.5
≥30.0	465 (4.1)	2019 (3.6)	2.6	465 (4.1)	484 (4.2)	0.8
Hypertension	5267 (46.2)	33 565 (59.7)	27.3	5267 (46.2)	5233 (45.9)	0.6
Dyslipidemia	1487 (13.0)	10 540 (18.8)	15.7	1487 (13.0)	1459 (12.8)	0.7
Current smoking	3323 (29.1)	11 425 (20.3)	20.6	3323 (29.1)	3311 (29.0)	0.2
Diabetes mellitus
Any	4214 (37.0)	26 515 (47.2)	20.8	4214 (37.0)	4203 (36.9)	0.2
Requiring insulin	66 (0.6)	582 (1.0)	5.1	66 (0.6)	78 (0.7)	1.3
Prior MI	379 (3.3)	2576 (4.6)	6.5	379 (3.3)	383 (3.4)	0.2
Prior PCI	45 (0.4)	481 (0.9)	6.0	45 (0.4)	51 (0.5)	0.9
Prior CABG	1 (0.0)	7 (0.0)	0.0	1 (0.0)	0 (0.0)	1.4
Prior CHF	58 (0.5)	812 (1.4)	9.5	58 (0.5)	79 (0.7)	2.3
Prior stroke	127 (1.1)	1285 (2.3)	9.1	127 (1.1)	113 (1.0)	1.2
PVD	1479 (13.0)	9972 (17.7)	13.3	1479 (13.0)	1373 (12.0)	2.8
Chronic renal failure	273 (2.4)	2728 (4.9)	13.2	273 (2.4)	270 (2.4)	0.1
Chronic lung disease	627 (5.5)	4892 (8.7)	12.5	627 (5.5)	675 (5.9)	1.8
Charlson comorbidity index
Mean (±SD)	2 (2.1)	2.8 (2.5)	33.9	2 (2.1)	2 (2.1)	1.4
0	3420 (30.0)	10 440 (18.6)	26.9	3420 (30.0)	3465 (30.4)	0.9
1–2	4273 (37.5)	19 678 (35.0)	5.2	4273 (37.5)	4324 (37.9)	0.9
≥3	3709 (32.5)	26 098 (46.4)	28.7	3709 (32.5)	3613 (31.7)	1.8
Clinical presentation
Unstable angina	2306 (20.2)	28 893 (51.4)	68.8	2306 (20.2)	2315 (20.3)	0.2
Acute MI	9096 (79.8)	27 323 (48.6)	68.8	9096 (79.8)	9087 (79.7)	0.2
Index treatment
PCI	10 938 (95.9)	48 291 (85.9)	35.4	10 938 (95.9)	10 941 (96.0)	0.2
CABG	128 (1.1)	1648 (2.9)	12.9	128 (1.1)	132 (1.2)	0.4
Medical therapy	336 (3.0)	6277 (11.2)	32.5	336 (3.0)	329 (2.9)	0.4
Concomitant mediations at index hospitalization
Aspirin	11 368 (99.7)	55 347 (98.5)	13.1	11 368 (99.7)	11 366 (99.7)	0.4
Statins	11 225 (98.5)	52 767 (93.9)	24.0	11 225 (98.5)	11 212 (98.3)	1.0
β‐Blockers	9544 (83.7)	41 440 (73.7)	24.6	9544 (83.7)	9559 (83.8)	0.4
Calcium‐channel blockers	4052 (35.5)	27 155 (48.3)	26.1	4052 (35.5)	4104 (36.0)	0.9
ACEIs or ARBs	8543 (74.9)	40 429 (71.9)	6.8	8543 (74.9)	8596 (75.4)	1.1
Diuretics	2215 (19.4)	13 725 (24.4)	12.1	2215 (19.4)	2237 (19.6)	0.5

Data are mean (SD) or number (percentage). The standardized differences are reported as percentages; a difference of <10.0% indicates a relatively small imbalance. ACEI indicates angiotensin‐converting enzyme inhibitor; ARB, angiotensin II receptor blocker; CABG, coronary artery bypass grafting; CHF, congestive heart failure; MI, myocardial infarction; PCI, percutaneous coronary intervention; PVD, peripheral vascular disease.

Weight in kilograms divided by the square of the height in meters.

In the matched cohort, paired comparisons were performed with the use of the McNemar test for binary variables and a paired Student t test or paired‐sample test for continuous variables. The comparative risks of safety and effectiveness outcomes were compared using Cox proportional hazards regression models with robust standard errors that accounted for the clustering of matched pairs. Kaplan–Meier survival curves were estimated in each matched cohort of P2Y12 inhibitors, and the survival curves were compared according to methods appropriate for matched data.20 All analyses for outcomes were truncated at 2 years of follow‐up, owing to the different follow‐up durations according to type of P2Y12 inhibitor and the small number of patients with data thereafter.

Several sensitivity analyses were performed considering that drug switching occurred over time. Adherence to P2Y12 inhibitors was shown at 3, 6, 9, 12, 18, and 24 months (Table S3). Drug exposure was considered as a time‐dependent variable. These same time points were used in the time‐dependent variable analysis of the Cox model. Several supplementary analyses were also performed to confirm the risk of safety and effectiveness outcomes in the various groups: (1) patients including a population with <30 days use of P2Y12 inhibitors; (2) ST‐segment–elevation MI patients; (3) patients according to initial presentation (acute MI versus unstable angina cohort); and (4) healthy PCI cohort (body weight ≥60 kg, <75 years old, and no history of stroke or transient ischemic attack). We conducted many sensitivity and subgroup analyses. The hazard ratios (HRs) were adjusted for propensity score in the each propensity‐score–matched cohort. In case of stratified analysis according to initial presentation, HRs used all data and were adjusted for covariates directly in the Cox model.

This observational data analysis used administrative claims–based data sets. To carefully define the population of interest and to minimize the data‐dredging processes, we prespecified study objectives, a hypothesis, and a statistical approach using a statistical analysis plan.21 All reported P values are 2‐sided, and those <0.05 were considered statistically significant. For all statistical analyses, SAS v9.3 (SAS Institute) was used.

Results

Study Population and Patient Characteristics

In the initial cohort of 324 937 patients with a diagnosis of ACS who were prescribed P2Y12 inhibitors, we identified 218 770 incident users of P2Y12 inhibitors. Among them, a total of 70 715 patients requiring hospitalization with a principal diagnosis of ACS met the study inclusion criteria and none of the exclusion criteria (Figure 1). Of these, 56 216 (79.5%) received clopidogrel, 11 402 (16.1%) received ticagrelor, and 3097 (4.4%) received prasugrel. In the study period, clopidogrel use steadily decreased, but ticagrelor use rapidly increased over time, and prasugrel use was consistently low at <5% (Figure S1). Before propensity‐score matching, there were between‐group differences regarding several of the baseline variables in each cohort for comparisons (Tables 1, 2 through 3). Prematched data showed that users of potent P2Y12 inhibitors (ticagrelor or prasugrel) were generally younger, were predominantly male, had higher body mass index, and had fewer comorbidities than users of clopidogrel. After propensity‐score matching was completed, there were 11 402 matched pairs for ticagrelor versus clopidogrel, 3097 matched pairs for prasugrel versus clopidogrel, and 3095 matched pairs for ticagrelor versus prasugrel. After matching, the standardized differences were <10.0% for most of variables, indicating only small differences between the 2 groups (Tables 1, 2 through 3).

Table 2

Baseline Characteristics Before and After Propensity‐Score Matching Among Patients With Prasugrel and Clopidogrel Use

Characteristic	Before Matching			After Matching
	Prasugrel (n=3097)	Clopidogrel (n=56 216)	Standardized Difference (%)	Prasugrel (n=3097)	Clopidogrel (n=3097)	Standardized Difference (%)
Age
Mean, y	55.9 (9.5)	65.4 (12.1)	88.1	55.9 (9.5)	55.9 (9.4)	0.2
≥75 y	55 (1.8)	14 404 (25.6)	73.9	55 (1.8)	55 (1.8)	0.0
Sex
Male	2767 (89.3)	36 770 (65.4)	59.7	2767 (89.3)	2772 (89.5)	0.6
Female	330 (10.7)	19 446 (34.6)	59.7	330 (10.7)	325 (10.5)	0.6
Socioeconomic status
Low tertile	962 (31.1)	18 287 (32.5)	3.2	962 (31.1)	933 (30.1)	2.0
Middle tertile	1143 (36.9)	18 165 (32.3)	9.7	1143 (36.9)	1200 (38.8)	3.8
High tertile	992 (32.0)	19 764 (35.2)	6.6	992 (32.0)	964 (31.1)	1.9
Body mass indexa
Mean	25.3 (2.7)	24.52 (2.7)	30.1	25.3 (2.7)	25.3 (2.6)	3.0
<20.0	52 (1.7)	2444 (4.4)	15.7	52 (1.7)	52 (1.7)	0.0
20.0 to <22.5	282 (9.1)	7725 (13.7)	14.6	282 (9.1)	262 (8.5)	2.3
22.5 to <25.0	962 (31.1)	23 719 (42.2)	23.3	962 (31.1)	1008 (32.6)	3.2
25.0 to <27.5	1313 (42.4)	15 550 (27.7)	31.3	1313 (42.4)	1314 (42.4)	0.1
27.5 to <30.0	313 (10.1)	4759 (8.5)	5.7	313 (10.1)	288 (9.3)	2.7
≥30.0	175 (5.7)	2019 (3.6)	9.8	175 (5.7)	173 (5.6)	0.3
Hypertension	1185 (38.3)	33 565 (59.7)	43.9	1185 (38.3)	1157 (37.4)	1.9
Dyslipidemia	370 (12.0)	10 540 (18.8)	18.9	370 (12.0)	359 (11.6)	1.1
Current smoking	1027 (33.2)	11 425 (20.3)	29.3	1027 (33.2)	1031 (33.3)	0.3
Diabetes mellitus
Any	964 (31.1)	26 515 (47.2)	33.3	964 (31.1)	935 (30.2)	2.0
Requiring insulin	17 (0.6)	582 (1.0)	5.5	17 (0.6)	16 (0.5)	0.4
Prior MI	99 (3.2)	2576 (4.6)	7.1	99 (3.2)	96 (3.1)	0.6
Prior PCI	11 (0.4)	481 (0.9)	6.4	11 (0.4)	14 (0.5)	1.4
Prior CABG	1 (0.0)	7 (0.0)	1.4	1 (0.0)	1 (0.0)	0.0
Prior CHF	8 (0.3)	812 (1.4)	12.9	8 (0.3)	12 (0.4)	2.3
Prior stroke	20 (0.7)	1285 (2.3)	13.7	20 (0.7)	13 (0.4)	3.2
PVD	306 (9.9)	9972 (17.7)	22.9	306 (9.9)	282 (9.1)	2.6
Chronic renal failure	43 (1.4)	2728 (4.9)	20.0	43 (1.4)	39 (1.3)	1.1
Chronic lung disease	125 (4.0)	4892 (8.7)	19.2	125 (4.0)	133 (4.3)	1.3
Charlson comorbidity index
Mean (±SD)	1.7 (1.9)	2.8 (2.5)	50.4	1.7 (1.9)	1.6 (1.9)	4.8
0	1044 (33.7)	10 440 (18.6)	35.0	1044 (33.7)	1134 (36.6)	6.1
1–2	1225 (39.6)	19 678 (35.0)	9.4	1225 (39.6)	1206 (38.9)	1.2
≥3	828 (26.7)	26 098 (46.4)	41.7	828 (26.7)	757 (24.4)	5.3
Clinical presentation
Unstable angina	734 (23.7)	28 893 (51.4)	59.7	734 (23.7)	734 (23.7)	0.0
Acute MI	2363 (76.3)	27 323 (48.6)	59.7	2363 (76.3)	2363 (76.3)	0.0
Index treatment
PCI	3033 (97.9)	48 291 (85.9)	45.2	3033 (97.9)	3041 (98.2)	1.9
CABG	18 (0.6)	1648 (2.9)	18.0	18 (0.6)	16 (0.5)	0.8
Medical therapy	46 (1.5)	6277 (11.2)	40.6	46 (1.5)	40 (1.3)	1.7
Concomitant mediations at index hospitalization
Aspirin	3083 (99.6)	55 347 (98.5)	11.1	3083 (99.6)	3082 (99.5)	0.4
Statins	3043 (98.3)	52 767 (93.9)	22.8	3043 (98.3)	3049 (98.5)	1.5
β‐Blockers	2521 (81.4)	41 440 (73.7)	18.5	2521 (81.4)	2538 (82.0)	1.4
Calcium‐channel blockers	1034 (33.4)	27 155 (48.3)	30.7	1034 (33.4)	1015 (32.8)	1.3
ACEIs or ARBs	2334 (75.4)	40 429 (71.9)	7.8	2334 (75.4)	2347 (75.8)	1.0
Diuretics	495 (16.0)	13 725 (24.4)	21.1	495 (16.0)	490 (15.8)	0.4

Data are mean (SD) or number (percentage). The standardized differences are reported as percentages; a difference of <10.0% indicates a relatively small imbalance. ACEI indicates angiotensin‐converting enzyme inhibitor; ARB, angiotensin II receptor blocker; CABG, coronary artery bypass grafting; CHF, congestive heart failure; MI, myocardial infarction; PCI, percutaneous coronary intervention; PVD, peripheral vascular disease.

Weight in kilograms divided by the square of the height in meters.

Table 3

Baseline Characteristics Before and After Propensity‐Score Matching Among Patients With Ticagrelor and Prasugrel Usea

Characteristic	Before Matching			After Matching
	Ticagrelor (n=11 402)	Prasugrel (n=3097)	Standardized Difference (%)	Ticagrelor (n=3095)	Prasugrel (n=3095)	Standardized Difference (%)
Age
Mean, y	60.9 (12.1)	55.9 (9.5)	46.2	55.9 (9.4)	55.9 (9.5)	0.1
≥75 y	1741 (15.3)	55 (1.8)	49.8	55 (1.8)	55 (1.8)	0.0
Sex
Male	8876 (77.9)	2767 (89.3)	31.4	2766 (89.4)	2765 (89.3)	0.1
Female	2526 (22.2)	330 (10.7)	31.4	329 (10.6)	330 (10.7)	0.1
Socioeconomic status
Low tertile	3623 (31.8)	962 (31.1)	1.6	919 (29.7)	961 (31.1)	3.0
Middle tertile	3995 (35.0)	1143 (36.9)	3.9	1191 (38.5)	1143 (36.9)	3.2
High tertile	3784 (33.2)	992 (32.0)	2.5	985 (31.8)	991 (32.0)	0.4
Body mass indexa
Mean	24.8 (2.7)	25.3 (2.7)	20.9	25.3 (2.6)	25.3 (2.7)	0.8
<20.0	392 (3.4)	52 (1.7)	11.2	41 (1.3)	52 (1.7)	3.0
20.0 to <22.5	1458 (12.8)	282 (9.1)	11.8	285 (9.2)	282 (9.1)	0.3
22.5 to <25.0	4275 (37.5)	962 (31.1)	13.6	946 (30.6)	962 (31.1)	1.1
25.0 to <27.5	3759 (33.0)	1313 (42.4)	19.6	1334 (43.1)	1311 (42.4)	1.5
27.5 to <30.0	1053 (9.2)	313 (10.1)	2.9	324 (10.5)	313 (10.1)	1.2
≥30.0	465 (4.1)	175 (5.7)	7.3	165 (5.3)	175 (5.7)	1.4
Hypertension	5267 (46.2)	1185 (38.3)	16.1	1181 (38.2)	1184 (38.3)	0.2
Dyslipidemia	1487 (13.0)	370 (12.0)	3.3	367 (11.9)	370 (12.0)	0.3
Current smoking	3323 (29.1)	1027 (33.2)	8.7	1025 (33.1)	1027 (33.2)	0.1
Diabetes mellitus
Any	4214 (37.0)	964 (31.1)	12.3	963 (31.1)	964 (31.2)	0.1
Requiring insulin	66 (0.6)	17 (0.6)	0.4	5 (0.2)	17 (0.6)	6.6
Prior MI	379 (3.3)	99 (3.2)	0.7	117 (3.8)	99 (3.2)	3.2
Prior PCI	45 (0.4)	11 (0.4)	0.5	9 (0.3)	11 (0.4)	1.2
Prior CABG	1 (0.0)	1 (0.0)	1.4	0 (0.0)	1 (0.0)	2.4
Prior CHF	58 (0.5)	8 (0.3)	4.0	6 (0.2)	8 (0.3)	1.5
Prior stroke	127 (1.1)	20 (0.7)	4.9	18 (0.6)	20 (0.7)	0.9
PVD	1479 (13.0)	306 (9.9)	9.7	322 (10.4)	306 (9.9)	1.7
Chronic renal failure	273 (2.4)	43 (1.4)	7.3	31 (1.0)	43 (1.4)	3.6
Chronic lung disease	627 (5.5)	125 (4.0)	6.9	124 (4.0)	125 (4.0)	0.2
Charlson comorbidity index
Mean (±SD)	2 (2.1)	1.7 (1.9)	16.0	1.7 (1.9)	1.7 (1.9)	0.0
0	3420 (30.0)	1044 (33.7)	8.0	1080 (34.9)	1044 (33.7)	2.4
1–2	4273 (37.5)	1225 (39.6)	4.3	1170 (37.8)	1223 (39.5)	3.5
≥3	3709 (32.5)	828 (26.7)	12.7	845 (27.3)	828 (26.8)	1.2
Clinical presentation
Unstable angina	2306 (20.2)	734 (23.7)	8.4	709 (22.9)	734 (23.7)	1.9
Acute MI	9096 (79.8)	2363 (76.3)	8.4	2386 (77.1)	2361 (76.3)	1.9
Index treatment
PCI	10 938 (95.9)	3033 (97.9)	11.6	3033 (98.0)	3031 (97.9)	0.5
CABG	128 (1.1)	18 (0.6)	5.9	18 (0.6)	18 (0.6)	0.0
Medical therapy	336 (3.0)	46 (1.5)	9.9	44 (1.4)	46 (1.5)	0.6
Concomitant mediations at index hospitalization
Aspirin	11 368 (99.7)	3083 (99.6)	2.5	3086 (99.7)	3082 (99.6)	2.2
Statins	11 225 (98.5)	3043 (98.3)	1.5	3053 (98.6)	3042 (98.3)	2.8
β‐Blockers	9544 (83.7)	2521 (81.4)	6.1	2522 (81.5)	2520 (81.4)	0.2
Calcium‐channel blockers	4052 (35.5)	1034 (33.4)	4.5	1044 (33.7)	1032 (33.3)	0.8
ACEIs or ARBs	8543 (74.9)	2334 (75.4)	1.0	2311 (74.7)	2332 (75.4)	1.6
Diuretics	2215 (19.4)	495 (16.0)	9.0	515 (16.6)	495 (16.0)	1.8

Data are mean (SD) or number (percentage). The standardized differences are reported as percentages; a difference of <10.0% indicates a relatively small imbalance. ACEI indicates angiotensin‐converting enzyme inhibitor; ARB, angiotensin II receptor blocker; CABG, coronary artery bypass grafting; CHF, congestive heart failure; MI, myocardial infarction; PCI, percutaneous coronary intervention; PVD, peripheral vascular disease.

Weight in kilograms divided by the square of the height in meters.

Comparative Safety and Effectiveness Outcomes

The median follow‐up period was 17.5 months (interquartile range: 9.0–26.2 months). During the follow‐up period, adherence to the index P2Y12 regimen was shown in Table S3. Absolute event rates at 2 years were shown in Table S4. In a propensity‐matched cohort, compared with clopidogrel, ticagrelor use was associated with a higher risk of any bleeding (HR: 1.23; 95% CI, 1.14–1.33; P<0.001; Table 4 and Figure 2). With respect to effectiveness outcomes, ticagrelor was associated with a similar risk of major cardiovascular events (HR: 1.00; 95% CI, 0.92–1.09; P=0.96) but a lower risk of all‐cause mortality (HR: 0.76; 95% CI, 0.63–0.91; P=0.002). With regard to each component of major cardiovascular events, compared with clopidogrel, ticagrelor was significantly associated with a lower risk of cardiovascular death or stroke, but the risk of MI was similar. In a matched cohort of prasugrel versus clopidogrel, prasugrel was associated with a higher risk of any bleeding (HR: 1.23; 95% CI, 1.06–1.43; P=0.01) and major bleeding (HR: 1.50; 95% CI, 1.01–2.21; P=0.04), but there was no statistically significant difference in effectiveness outcomes (Table 5 and Figure 3). In a matched cohort of ticagrelor versus prasugrel, there was no statistically significant between‐group difference with respect to safety or effectiveness outcomes except nasal bleeding (Table 6 and Figure 4).

Table 4

Risk of Safety and Effectiveness Outcomes in the Propensity‐Score–Matched Cohort of Ticagrelor and Clopidogrela

Outcomes	Outcome Rate at 2 Years (%)b		HR (95% CI)c	P Value
	Ticagrelor (n=11 402)	Clopidogrel (n=11 402)
Safety outcomes
Any bleeding	18.1	15.1	1.23 (1.14–1.33)	<0.001
Major bleeding	3.1	2.5	1.18 (0.98–1.43)	0.07
Site of bleeding events
Intracranial bleeding	0.8	1.0	0.85 (0.61–1.18)	0.33
Gastrointestinal bleeding	6.1	5.3	1.10 (0.96–1.26)	0.15
Urogenital bleeding	2.3	2.1	1.12 (0.89–1.39)	0.33
Respiratory bleeding	1.0	0.8	1.29 (0.93–1.78)	0.13
Nasal bleeding	4.4	2.8	1.73 (1.47–2.04)	<0.001
Intraocular bleeding	5.0	4.4	1.18 (1.01–1.36)	0.03
Other bleeding	0.5	0.5	1.21 (0.78–1.86)	0.40
Transfusion	1.8	1.5	1.22 (0.96–1.56)	0.10
Effectiveness outcomes
Major cardiovascular eventsd	13.1	13.0	1.00 (0.92–1.09)	0.96
Death from cardiovascular causes	1.0	1.7	0.62 (0.47–0.82)	0.001
MI	10.6	10.0	1.07 (0.97–1.18)	0.20
Stroke	2.1	2.5	0.82 (0.66–1.00)	0.05
All‐cause mortality	3.1	3.9	0.76 (0.63–0.91)	0.002

HR indicates hazard ratio; MI, myocardial infarction.

The propensity‐score–matched cohort included 11 402 patients in the ticagrelor user group and 11 402 patients in the clopidogrel user group.

Outcome rates were derived from paired Kaplan–Meier curves.

HRs are for ticagrelor compared with clopidogrel.

Major cardiovascular events were defined as a composite of death from cardiovascular causes, MI, or stroke.

Figure 2

Cumulative risks of the study outcomes in the matched cohort of ticagrelor and clopidogrel. Cumulative incidence curves are shown for any bleeding (A), major bleeding (B), major cardiovascular events (C), and all‐cause mortality (D).

Table 5

Risk of Safety and Effectiveness Outcomes in the Propensity‐Score–Matched Cohort of Prasugrel and Clopidogrela

Outcomes	Outcome Rate at 2 Years (%)b		HR (95% CI)c	P Value
	Prasugrel (n=3097)	Clopidogrel (n=3097)
Safety outcomes
Any bleeding	14.8	12.5	1.23 (1.06–1.43)	0.01
Major bleeding	2.6	1.8	1.50 (1.01–2.21)	0.04
Site of bleeding events
Intracranial bleeding	0.8	0.5	1.21 (0.59–2.49)	0.60
Gastrointestinal bleeding	5.2	3.9	1.33 (1.02–1.73)	0.03
Urogenital bleeding	1.8	1.6	1.13 (0.73–1.75)	0.58
Respiratory bleeding	0.6	0.5	1.48 (0.71–3.10)	0.30
Nasal bleeding	4.0	2.3	1.88 (1.36–2.60)	<0.001
Intraocular bleeding	4.0	4.1	0.95 (0.71–1.26)	0.72
Other bleeding	0.3	0.2	1.88 (0.63–5.61)	0.26
Transfusion	1.5	1.0	1.60 (0.96–2.64)	0.07
Effectiveness outcomes
Major cardiovascular eventsd	10.3	11.4	0.88 (0.74–1.05)	0.17
Death from cardiovascular causes	0.6	0.9	0.66 (0.35–1.26)	0.21
Myocardial infarction	9.0	9.8	0.91 (0.75–1.10)	0.32
Stroke	1.3	1.3	0.95 (0.58–1.57)	0.85
All‐cause mortality	1.6	1.9	0.78 (0.50–1.22)	0.28

HR indicates hazard ratio; MI, myocardial infarction.

The propensity‐score–matched cohort included 3097 patients in the prasugrel user group and 3097 patients in the clopidogrel user group.

Outcome rates were derived from paired Kaplan–Meier curves.

HRs are for prasugrel compared with clopidogrel.

Major cardiovascular events were defined as a composite of death from cardiovascular causes, MI, or stroke.

Figure 3

Cumulative risks of the study outcomes in the matched cohort of prasugrel and clopidogrel. Cumulative incidence curves are shown for any bleeding (A), major bleeding (B), major cardiovascular events (C), and all‐cause mortality (D).

Table 6

Risk of Safety and Effectiveness Outcomes in the Propensity‐Score–Matched Cohort of Ticagrelor and Prasugrela

Outcomes	Outcome Rate at 2 Years (%)b		HR (95% CI)c	P Value
	Ticagrelor (n=3095)	Prasugrel (n=3095)
Safety outcomes
Any bleeding	18.0	14.8	1.16 (1.00–1.35)	0.05
Major bleeding	2.6	2.6	0.99 (0.67–1.44)	0.94
Site of bleeding events
Intracranial bleeding	0.6	0.8	1.10 (0.52–2.30)	0.80
Gastrointestinal bleeding	5.6	5.2	1.00 (0.77–1.31)	0.98
Urogenital bleeding	2.2	1.8	1.08 (0.68–1.69)	0.75
Respiratory bleeding	0.6	0.6	1.01 (0.50–2.02)	0.99
Nasal bleeding	6.1	4.0	1.38 (1.05–1.80)	0.02
Intraocular bleeding	4.8	4.0	1.13 (0.83–1.53)	0.44
Other bleeding	0.4	0.3	1.25 (0.51–3.09)	0.63
Transfusion	1.6	1.5	0.95 (0.59–1.54)	0.84
Effectiveness outcomes
Major cardiovascular eventsd	11.1	10.3	1.14 (0.94–1.37)	0.18
Death from cardiovascular causes	0.5	0.6	0.76 (0.33–1.75)	0.52
MI	9.6	9.1	1.11 (0.90–1.36)	0.33
Stroke	1.3	1.3	1.18 (0.70–2.01)	0.54
All‐cause mortality	1.4	1.6	0.92 (0.53–1.59)	0.77

HR indicates hazard ratio; MI, myocardial infarction.

The propensity‐score–matched cohort included 3095 patients in the ticagrelor user group and 3095 patients in the prasugrel user group.

Outcome rates were derived from paired Kaplan–Meier curves.

HRs are for ticagrelor compared with prasugrel.

Major cardiovascular events were defined as a composite of death from cardiovascular causes, MI, or stroke.

Figure 4

Cumulative risks of the study outcomes in the matched cohort of ticagrelor and prasugrel. Cumulative incidence curves are shown for any bleeding (A), major bleeding (B), major cardiovascular events (C), and all‐cause mortality (D).

Sensitivity and Subgroup Analyses

Results of the sensitivity analyses with P2Y12 inhibitors exposures as a time‐varying covariate were similar to those of the overall analysis (Table S5). We performed additional analyses including a population with <30 days’ use of P2Y12 inhibitors (Tables S6 and S7). As a result, risks in safety and effectiveness outcomes were similar for the main results. We conducted focused analysis of ST‐segment–elevation MI patients, which ensured a more homogeneous patient group for comparison (Table S8). In the ST‐segment–elevation MI cohort, compared with clopidogrel, ticagrelor and prasugrel were associated with higher risk of bleeding. Compared with prasugrel, ticagrelor was associated with lower risk of major bleeding. With respect to effectiveness outcomes, no significant between‐group difference was noted in all matched subcohorts. The outcomes of stratification analyses according to the patient's initial presentation (acute MI versus unstable angina cohort) are shown in Table S9. The results for both groups of acute MI and the unstable angina cohort were similar. Results for another sensitivity analysis in the healthy PCI cohort are shown in Table S10. The risk of any bleeding was significantly higher in the ticagrelor group than in the clopidogrel group. No significant between‐group difference was noted in any matched subcohorts with respect to any safety and effectiveness outcomes.

Discussion

This nationwide population‐based cohort study had several major findings. First, potent P2Y12 inhibitors were prescribed substantially less often for Asian patients than for Western patients.22, 23 Second, compared with clopidogrel, ticagrelor was associated with an increased risk of bleeding but with lower risks of mortality for any cause and for cardiovascular causes and stroke. Third, compared with clopidogrel, prasugrel was associated with an increased risk of bleeding but with similar risks for effectiveness outcomes. Fourth, no significant differences were noted in the risk of bleeding and ischemic events between ticagrelor and prasugrel.

The key findings of our study conflicted with those of the pivotal RCTs of ticagrelor and prasugrel.2, 3 In our study, compared with clopidogrel, ticagrelor use significantly increased the rate of bleeding events without reducing major cardiovascular events. Some prior data suggested that the advantages of ticagrelor over clopidogrel and its net clinical benefit varied according to geography and ethnicity.24, 25 Similar to our findings, the PHILO trial showed that the 1‐year rates of major bleeding events (10.3% versus 6.8%) and minor bleeding events (15.2% versus 9.2%) were higher in the ticagrelor group than in the clopidogrel group without a clear benefit regarding ischemic events.26 Nevertheless, although our study used relatively weak criteria for major bleeding, the risk of major bleeding seems to be lower compared with the PHILO trial. This disparity might be explained by the differences in study design, population, definition and coding of events, and adjudication process. Similar to the PLATO (Platelet Inhibition and Patient Outcomes) trial,2 our study also showed that ticagrelor was associated with mortality reduction. In PLATO, the improved survival rate with ticagrelor might be due to a decrease in ischemic events without a concomitant increase in major bleeding. However, in our study, the plausible reasons for the mortality benefit of ticagrelor use without a significant reduction of major cardiovascular events are still unclear. With regard to each component of the composite major cardiovascular event, compared with clopidogrel, ticagrelor was significantly associated with lower risk of death from cardiovascular causes and stroke but not risk of MI. Because the proportion of MI was largest in the composite outcome, the benefit of ticagrelor on reduction of major cardiovascular events seems to be not significant. A differential effect of ticagrelor on mortality or MI needs to be addressed in future investigations (ie, the pleiotropic effects of ticagrelor associated with inhibition of adenosine reuptake).

In the current study, although the limited number of prasugrel users might provide less robust findings, prasugrel was associated with an increased risk of bleeding events and was not associated with a benefit for major cardiovascular events and mortality compared with clopidogrel. Given the lower body mass index and greater bleeding tendency of Asian patients, physicians were less likely to prescribe the usual dose of prasugrel. The PRASFIT‐ACS (Prasugrel Compared With Clopidogrel for Japanese Patients With ACS Undergoing PCI) trial, involving Japanese patients with ACS, showed that a reduced dose of prasugrel (a 20‐mg loading dose and a 3.75‐mg daily maintenance dose) was associated with a lower risk of ischemic and bleeding events compared with clopidogrel.27 After this trial, a low dose of prasugrel was approved as the recommended dosing for Japanese population. Further studies are required to define the optimal dosing of prasugrel targeting an East Asian population.

A head‐to‐head comparison of newer P2Y12 inhibitors remains a significant challenge. The PRAGUE‐18 trial showed that the 30‐day and 1‐year rates of ischemic, bleeding, and net clinical end points were similar for ticagrelor and prasugrel.28, 29 Similarly, our postapproval observational study showed no significant differences in bleeding or ischemic outcomes for ticagrelor and prasugrel. These observations might highlight the practical challenges faced by treating physicians considering head‐to‐head evaluations of active therapies for ACS care. However, because previous trials were underpowered and observational studies have inherent limitations, a definitive answer regarding the comparative effectiveness of ticagrelor and prasugrel warrants further investigation and should be confirmed or refuted through large RCTs.

Although East Asian data have come from several registries and cohorts, the results are conflicting. KAMIR‐NIH and this cohort's result favored for the concept of the East Asian paradox,30 whereas the Taiwan National Database and the international multicenter BleeMACS registry favored potent P2Y12 inhibitors for ACS patients.31, 32 Although exact reasons for the different results across these registries are still unknown, they might be explained in part by differences in patient characteristics, clinical practice or pattern, and end point definitions, as well as by confounding factors. The underlying mechanism of East Asian paradox with response to antiplatelet drugs has not been fully determined.6, 33 This phenomenon may be partly explained by interethnic differences in intrinsic thrombogenicity, pharmacokinetic and pharmacodynamic profiles, and propensity for bleeding complications.34 In addition, differences in genetic polymorphisms (ie, factor V Leiden [G1691A] and prothrombin [G20210A] gene mutations), plasma hemostatic factors (ie, fibrinogen, D‐dimer, and factor VIII), and endothelial activation markers (ie, VWF [von Willebrand factor], ICAM1 [intercellular adhesion molecule 1], and E‐selectin) may at least contribute to this disparity.35, 36

Our study has some potential limitations. First, our results rely on the completeness and accuracy of data from electronic and administrative databases. There is a possibility of coding errors, missing data, lack of clinically relevant data due to unmeasured variables, or concomitant over‐the‐counter drug use that usually cannot be captured in such data sources. However, the definition and coding of clinically relevant outcomes in our study were validated in recent clinical studies using the NHIS database.14, 15 Second, this study was observational and may have selection or ascertainment bias. Although all measured baseline differences were accounted for using robust propensity‐score matching, unmeasured confounder might influence observed results. Unfortunately, we did not have data on coronary lesion characteristics that affect clinical outcomes; therefore, this factor could not be included in the propensity scores. Third, the primary end points were not adjudicated, leaving substantial risk of bias and misclassification of the end points. Finally, we cannot accurately quantify the effects of treatment retention and adherence. Over time, P2Y12 de‐escalation (switching from ticagrelor/prasugrel to clopidogrel) was common (Table S3). However, even after additional adjustment of the status of P2Y12 inhibitors as a time‐varying covariate, the overall findings were similar.

Conclusions

Among East Asian patients who presented with ACS, compared with clopidogrel, ticagrelor was associated with an increased rate of bleeding but with a significant reduction in death from all causes and from cardiovascular causes and stroke. Compared with clopidogrel, prasugrel was associated with an increase in bleeding events without differences in effectiveness outcomes. No significant differences were noted between ticagrelor and prasugrel with respect to bleeding and ischemic events.

Sources of Funding

This study was supported by the National Evidence‐based Healthcare Collaborating Agency (NECA) under project number NECA‐A‐16‐003. There was no industry involvement in the design, conduct, or analysis of the study.

Disclosures

None.

Table S1. Definitions of Clinical Risk Factors or Comorbid Conditions and Concomitant Cardioactive Medications on the Basis of Codes and Prescriptions in the 365 Days Before Exposure

Table S2. Definitions of Safety and Efficacy Outcomes

Table S3. Adherence to the Index Drug During the Follow‐Up Period

Table S4. Observed Rates of 2‐Year Clinical Outcomes in the Overall Population

Table S5. Time‐Dependent Covariate Analysis

Table S6. Baseline Characteristics Between Included Study Population and Excluded Population With <30 Days Use of P2Y12 Inhibitors

Table S7. Risk of Safety and Effectiveness Outcomes in the Propensity‐Score–Matched Cohort of Each P2Y12 Inhibitor in Sensitivity Analyses Including Population with <30‐Day Use of P2Y12 Inhibitors

Table S8. Risk of Safety and Effectiveness Outcomes in the Propensity‐Score–Matched Cohort of Each P2Y12 Inhibitor in a Subgroup of ST‐Segment–Elevation Myocardial Infarction Patients

Table S9. Stratified Analysis According to Initial Presentation (Acute Myocardial Infarction vs Unstable Angina Cohort)

Table S10. Risk of Safety and Effectiveness Outcomes in the Propensity‐Score–Matched Cohort of Each P2Y12 Inhibitors in a Healthy Percutaneous Coronary Intervention Cohort (Body Weight ≥60 kg, Age <75 Years, and No History of Stroke or Transient Ischemic Accident)

Figure S1. Proportion of new antiplatelet drug use compared with clopidogrel use over time.

Click here for additional data file.