Meta-analysis of intracranial hemorrhage in acute coronary syndromes: incidence, predictors, and clinical outcomes.

BACKGROUND: Little is known about the incidence, predictors, or outcomes of intracranial hemorrhage (ICH) in patients with non-ST-segment elevation acute coronary syndromes (NSTE ACS). We aimed to determine the incidence and timing of ICH, characterize the location of ICH, and identify independent baseline predictors of ICH in NSTE ACS patients. METHODS AND
RESULTS: We pooled patient-level data from 4 contemporary antithrombotic therapy trials. Multivariable modeling identified independent predictors of ICH. ICHs were adjudicated by a clinical events committee. Of 37 815 patients, 135 (0.4%) had an ICH. The median (25th, 75th percentiles) follow-up was 332 (184, 434) days but differed across trials. Locations of ICH were intracerebral (50%), subdural (31%), subarachnoid (18.5%), and intraventricular (11%). Independent predictors of ICH were older age (HR per 10 years, 1.61; 95% CI, 1.35 to 1.91); prior stroke/transient ischemic attack; HR, 1.95; 95% CI, 1.14 to 3.35), higher systolic blood pressure; HR per 10 mm Hg increase, 1.09; 95% CI, 1.01 to 1.18), and larger number of antithrombotic agents (HR per each additional agent, 2.06; 95% CI, 1.49 to 2.84). Of all ICHs, 45 (33%) were fatal.
CONCLUSIONS: In patients with NSTE ACS enrolled in recent clinical trials of antithrombotic therapies, ICH was uncommon. Patients with older age, prior transient ischemic attack/stroke, higher systolic blood pressure, or larger number of antithrombotic agents were at increased risk. One-third of patients with ICH died. These data may be useful to trialists and data and safety monitoring committees for trial conduct and monitoring. CLINICAL TRIAL REGISTRATION: URL: https://www.clinicaltrials.gov/. Unique identifiers: TRACER: NCT00527943, PLATO: NCT00391872, APPRAISE-2: NCT00831441, TRILOGY ACS: NCT00699998.

Little has been published about patients with non–ST-segment elevation acute coronary syndrome (NSTE ACS) who experience intracranial hemorrhage (ICH). Individual trials have reported ICH as part of bleeding or safety assessments. Proportions of patients with ICH have ranged from 0.2% to 0.5%.1–4 Heterogeneity of patient populations driven largely by inclusion and exclusion criteria, different management strategies, and different durations of follow-up have limited comparisons of ICH incidence across trials. Low event rates in individual trials also preclude statistical analyses.

We pooled patient-level data from 4 large global clinical trials coordinated at least in part at the Duke Clinical Research Institute in Durham, NC. The key objectives of these analyses are to determine the incidence and timing of ICH, to characterize the location of ICH, and to identify independent baseline predictors of ICH in NSTE ACS patients.

Methods

Study Population

The study population consisted of patients enrolled in 4 contemporary ACS trials: TRACER (Thrombin Receptor Antagonist for Clinical Event Reduction in Acute Coronary Syndrome), PLATO (Study of Platelet Inhibition and Patient Outcomes), APPRAISE-2 (Apixaban for Prevention of Acute Ischemic Events 2), and TRILOGY ACS (Targeted Platelet Inhibition to Clarify the Optimal Strategy to Medically Manage Acute Coronary Syndromes). The design and primary results of all 4 trials have been previously published.1–7 These trials were chosen because they were the 4 most recent, consecutive trials with individual patient-level data available at the DCRI that allowed comprehensive analyses to be performed. All studies were approved by the appropriate institutional review boards and ethics committees, and all participants provided informed consent.

Statistical Methods

Values are presented as N (percentage) and median with interquartile range (IQR) where appropriate. Time to ICH was explored using Kaplan–Meier methods. Rates of ICH are reported per 100 patient-years. Adjusted Cox multivariable regression modeling with both stepwise and backward elimination techniques were used to determine independent predictors of ICH. Candidate variables included age, sex, prior congestive heart failure, prior diabetes mellitus, prior hypercholesterolemia, prior hypertension, prior transient ischemic attack (TIA)/stroke, race, smoking status, baseline creatinine, baseline hemoglobin, systolic blood pressure (SBP), and weight. They were chosen after review of the baseline demographic and treatment factors to determine common variables across the 4 trials. They were also known risks factors for stroke or ICH for the ST-segment elevation myocardial infarction patient population.8 We created a candidate variable called “number of antithrombotic agents.” We included any oral antiplatelet agent or oral anticoagulant, including the randomized treatment assignment. This was done to assess the potential risk of ICH based on the number of agents but not to include comparisons across agents or trials. Missing values in candidate variables were not imputed, and multivariable models were based on complete case analysis.

ICH was defined similarly across the trials and included primary hemorrhagic strokes, ischemic strokes with hemorrhagic conversion, and ICH events that were not strokes but had bleeding within the cranium. All suspected ICH events were adjudicated by a clinical events committee using similar procedures across all 4 trials.

Results

Enrollment, number of ICH events, median follow-up, and median time to ICH for each of the trials and overall are shown in Table1. In total, 48 286 patients were enrolled in the 4 trials and 10 471 (21.7%) were excluded because they were enrolled in the ST-segment elevation cohorts of the trials that included this patient population in addition to NSTE ACS (PLATO [N=7544] and APPRAISE-2 [N=2927]). Overall, 135 patients (0.4%) had ICH events. The median duration of follow-up varied across the trials because of protocol differences in planned duration, and 1 trial3 was stopped early by the data and safety monitoring board. Figure 1 shows the cumulative probability of ICH in each of the trials, with similar curves for all 4 studies. The ICH rate was 0.239 per 100 patient-years for patients receiving single antiplatelet therapy, 0.255 for those receiving dual antiplatelet therapy, and 0.602 for those receiving 3 antithrombotic agents. In patients with and without prior TIA/stroke, the ICH rates were 0.741 and 0.299 per 100 patient-years, respectively. The median (IQR) days from randomization to ICH was 205 days (85 to 367) in the entire dataset. Median times from randomization to ICH differed by trial (Table1).

Table 1

Trials Included

	APPRAISE-23	PLATO2	TRACER1	TRILOGY ACS4	Total
No. of patients included in analysis	N=4456	N=11 080	N=12 944	N=9326	N=37 815
ICH, n/N (%)	7/4465 (0.2)	24/11 080 (0.2)	68/12 897 (0.5)	36/9326 (0.4)	135/37 768 (0.4)
ICH rate per 100 patient-y follow-up	0.264	0.263	0.393	0.272	0.319
Median follow-up time (d), median (25th, 75th)	180 (69, 299)	282 (179, 371)	481 (328, 651)	513 (312, 731)	—
Median time to ICH (d), median (25th, 75th)	56 (14, 214)	118 (33, 180)	224 (85, 369)	278 (166, 445)	—
Investigative therapies	Apixaban vs placebo on SOC	Ticagrelor vs clopidogrel on ASA	Vorapaxar vs placebo on SOC	Prasugrel vs clopidogrel on ASA	—

ASA indicates aspirin; ICH, intracranial hemorrhage; SOC, standard of care.

Figure 1

Product-limit failure curves.1–4 ICH indicates intracranial hemorrhage.

Table S1 shows the baseline demographics and patient characteristics by trial. Some differences are observed. The lower proportion of patients with prior stroke or TIA in TRILOGY ACS is due to planned exclusion of these patients.

Patients with ICH were older and more likely to have a history of hypertension, diabetes mellitus, smoking, coronary artery bypass graft surgery, prior stroke or TIA, and higher SBP (Table2). Patients with ICH were more often treated with antiplatelet and anticoagulant therapies.

Table 2

Baseline Demographic and Clinical Characteristics by ICH

	No ICH (N=37 680)	ICH (N=135)
Trial
APPRAISE-23	4458/37 680 (11.8%)	7/135 (5.2%)
PLATO2	11 056/37 680 (29.3%)	24/135 (17.8%)
TRACER1	12 876/37 680 (34.2%)	68/135 (50.4%)
TRILOGY ACS4	9290/37 680 (24.7%)	36/135 (26.7%)
Age, y	65 (57, 72)	69 (62, 77)
Race or ethnic group
White	30 715/37 680 (81.5%)	110/135 (81.5%)
Black	770/37 680 (2.0%)	3/135 (2.2%)
Other	4861/37 680 (12.9%)	15/135 (11.1%)
Male	25 426/37 680 (67.5%)	91/135 (67.4%)
Female	12 254/37 680 (32.5%)	44/135 (32.6%)
Body mass index, kg/m2	27 (25, 31)	27 (24, 30)
Prior hypertension	28 003/37 680 (74.3%)	106/135 (78.5%)
Prior diabetes mellitus	12 858/37 680 (34.1%)	49/135 (36.3%)
Smoking status
Never smoked	17 079/37 582 (45.4%)	63/135 (46.7%)
Former smoker	10 976/37 582 (29.2%)	48/135 (35.6%)
Current smoker	9527/37 582 (25.3%)	24/135 (17.8%)
Prior dyslipidemia	21 651/37 680 (57.5%)	84/135 (62.2%)
Prior myocardial infarction	12 481/37 680 (33.1%)	49/135 (36.3%)
Prior congestive heart failure	5007/37 680 (13.3%)	22/135 (16.3%)
Prior transient ischemic attack/stroke	2224/37 616 (5.9%)	15/135 (11.1%)
Prior angina	12 477/23 932 (52.1%)	46/92 (50.0%)
Prior percutaneous coronary intervention	8512/37 680 (22.6%)	27/135 (20.0%)
Prior coronary artery bypass graft	4399/37 680 (11.7%)	23/135 (17.0%)
Prior chronic obstructive pulmonary disease	1646/23 932 (6.9%)	7/92 (7.6%)
Coronary artery disease	10 385/33 222 (31.3%)	44/128 (34.4%)
Baseline Killip score
I	30 512/33 214 (91.9%)	111/128 (86.7%)
II	2258/33 214 (6.8%)	13/128 (10.2%)
III	411/33 214 (1.2%)	4/128 (3.1%)
IV	33/33 214 (0.1%)	0/128 (0.0%)
Baseline heart rate, beats/min	70 (62, 79)	70 (62, 78)
Systolic blood pressure, mm Hg	130 (120, 141)	135 (120, 146)
Diastolic blood pressure, mm Hg	77 (70, 82)	79 (70, 85)
Glucose, mg/dL	113 (96, 149)	115 (93, 161)
Creatinine clearance, mL/min per 1.73 m²	77 (60, 96)	69 (52, 87)
Baseline hemoglobin, g/L	138 (127, 148)	136 (124, 147)
Baseline ST depression	17 170/35 142 (48.9%)	56/126 (44.4%)
Baseline transient ST elevation	3192/35 176 (9.1%)	9/126 (7.1%)
Baseline T-wave inversion	10 338/35 142 (29.4%)	31/126 (24.6%)
Selected medication administered between hospital admission and randomization by intracranial hemorrhage
Aspirin	35 664/37 680 (94.6%)	131/135 (97.0%)
Thienopyridines	26 750/37 680 (71.0%)	103/135 (76.3%)
Heparin (all forms)	23 539/37 680 (62.5%)	86/135 (63.7%)
Glycoprotein IIb/IIIa	4671/37 680 (12.4%)	19/135 (14.1%)
Direct thrombin inhibitors	1298/37 680 (3.4%)	7/135 (5.2%)
Statin	31 202/37 680 (82.8%)	107/135 (79.3%)
β-blockers	28 720/37 680 (76.2%)	112/135 (83.0%)
No. of antithrombotics*
0	89/37 302 (0.2%)	0/135 (0.0%)
1	2511/37 302 (6.7%)	7/135 (5.2%)
2	27 467/37 302 (73.6%)	78/135 (57.8%)
3	7230/37 302 (19.4%)	50/135 (37.0%)
4	5/37 302 (0.0%)	0/135 (0.0%)

Data presented as n/N (%) or median (25th, 75th); 1341 patients (3.8%) had missing race. ICH indicates intracranial hemorrhage.

Definition of number of antithrombotics: PLATO—1 for ticagrelor or clopidogrel (randomized treatment)+1 if the patient is taking aspirin at randomization+1 if the patient is taking vitamin K antagonists at randomization; APPRAISE2—1 if the patient is taking aspirin at randomization+1 if the patient is taking clopidogrel/ticlopidine/prasugrel at randomization+1 if the patient was randomized to apixaban; TRACER—1 if the patient is taking aspirin at randomization+1 if the patient is taking thienopyridines at randomization+1 if the patient is taking oral anticoagulants at randomization+1 if patient randomized to vorapaxar; TRILOGY ACS—1 for clopidogrel or prasugrel (randomized treatment)+1 if the patient is taking aspirin at randomization.

The anatomic locations of the ICH events are shown in Table3. One-half of all ICHs were intracerebral (50.4%), and nearly one-third (31.1%) were subdural hematomas. In total, 45 of the 135 (33.3%) ICH events were fatal.

Table 3

ICH, ICH Type, and Significant Mortality Following ICH by Clinical Trial

	APPRAISE-23 (N=7)	PLATO2 (N=24)	TRACER1 (N=68)	TRILOGY ACS4 (N=36)	Total (N=135)
Intracerebral hemorrhage	3/7 (42.9%)	14/24 (58.3%)	40/68 (58.8%)	11/36 (30.6%)	68/135 (50.4%)
Subdural hematoma	1/7 (14.3%)	1/24 (4.2%)	23/68 (33.8%)	17/36 (47.2%)	42/135 (31.1%)
Subarachnoid hemorrhage	0/7 (0.0%)	3/24 (12.5%)	15/68 (22.1%)	7/36 (19.4%)	25/135 (18.5%)
Intraventricular hemorrhage	1/7 (14.3%)	0/24 (0.0%)	14/68 (20.6%)	0/36 (0.0%)	15/135 (11.1%)
Fatal intracranial hemorrhage	4/7 (57.1%)	9/24 (37.5%)	23/68 (33.8%)	9/36 (25.0%)	45/135 (33.3%)

Data are presented as n/N (%). ICH indicates intracranial hemorrhage.

The results of the modeling to determine the independent predictors of ICH are shown in Tables S2 and4. Table S2 shows the full model, and Table4 shows the model after backward selection. The same model was obtained by using stepwise model selection. Four independent predictors of ICH were identified: older age, prior stroke or TIA, higher SBP, and treatment with a larger number of antithrombotic agents. Because patients with prior stroke or TIA were supposed to be excluded from TRILOGY ACS but 23 patients with prior stroke or TIA were enrolled and randomized, treatment was stopped as soon as this information became known, the modeling was performed without patients from the TRILOGY ACS trial, and the results were similar (data not shown). None of these 23 patients had an ICH.

Table 4

Multivariable Analyses of ICH Model After Backward Selection

Parameter	χ2		HR	95% CI
	Value	P Value	Value	Lower	Upper
Older age (HR in units of 10 years)	28.63	<0.001	1.61	1.35	1.91
Number of antithrombotics (HR per 1-unit increase)	13.48	0.002	2.10	1.41	3.13
Prior stroke/transient ischemic attack (yes vs no)	6.94	0.008	2.10	1.21	3.66
Systolic blood pressure (HR in units of 10 mm/Hg)	4.12	0.042	1.09	1.01	1.18

Analysis based on 37 241 complete cases. HR indicates hazard ratio; ICH, intracranial hemorrhage.

Discussion

We aggregated patient-level data from 4 large, contemporary trials of patients with ACS. In nearly 38 000 patients with NSTE ACS, ICH was uncommon and occurred in 0.4% of patients. However, 33% of ICH events were fatal. Four independent predictors of ICH were identified: older age, history of stroke or TIA, higher SBP, and larger number of antithrombotics. Investigators designing clinical trials of antithrombotic therapies in NSTE ACS populations or experts involved in data and safety monitoring committees may find these data useful, even though the overall event rate is low.

Prior large trials with similar NSTE ACS populations have individually reported ICH occurrences of 0.1% to 0.3% with different follow-up periods and patient populations.9–12 The proportions of patients with ICH from these other trials are consistent with the current analysis, particularly the TRITON–TIMI 38 trial, which was conducted in a contemporary era and (unlike SYNERGY and PURSUIT) had a longer follow-up period, similar to the 4 trials included in these analyses. ICH has been an important safety signal in several recent trials, including TRITON–TIMI 38,9 which supported approval of prasugrel for NSTE ACS but not in patients with prior stroke or TIA, as well as the TRA-2P TIMI 50 trial,13 in which the data and safety monitoring board stopped enrollment of patients with prior stroke due to excess ICH with vorapaxar compared with placebo. The follow-up phase of TRACER was also stopped, due to excess bleeding including ICH.1

In the design of trials, careful consideration of inclusion and exclusion criteria may be important to minimize risk to some patients while still being broad enough in patient recruitment to optimize generalizability of the trial results. Point estimates of ICH occurrence provided from these data of nearly 38 000 patients may also aid data and safety monitoring committees that are charged with monitoring human subject safety, as benchmarks of expected rates of ICH may be informative. From a clinical perspective, more work is needed before these data can be used to definitively guide treatment decisions for specific ACS patients. While we believe these data are robust, making patient treatment decisions based on 135 events is potentially hazardous.

Patient demographics and clinical characteristics, such as those used in our modeling, have been shown to be predictive of ICH events in other patient populations, including patients with ST-segment elevation myocardial infarction9 and for general populations.14 However, other important predictors of ICH are known that we could not include in these analyses, such as amyloid angiopathy.15 More work is needed to potentially integrate biomarker, proteomic, and genomic information that is available for some of the 4 trials included in these analyses.

Limitations

These analyses have limitations. Baseline information common to trials was included, but we did not include important postrandomization factors such as medication therapy and dosing or procedures. In addition, only blood pressure at time of randomization was systematically collected in the trials. Comparisons across the trials were not performed given the differences in patient populations and follow-up durations that were driven in part by protocol specifications. These trials included experiment arms with antithrombotic therapies with different mechanisms of action and different background therapies defined by the standard of care. The approach used to combine these different agents in a composite number of antithrombotic therapies, therefore, has limitations. Despite the large patient population, the number of ICH events remained modest and limits the confidence of the inferences that can be made from the data. We did not systematically assess stroke morbidity in all the trials and, thus, have limited information on patient outcomes after nonfatal ICH.

Conclusions

ICH was an uncommon event in patients with NSTE ACS and occurred in 0.4% of patients. Many events occurred early after enrollment, but there was continued risk for ICH over time. In total, 33% of patients with ICH died within 30 days of the event. Patients with older age, prior stroke or TIA, higher SBP, or treatment with larger number of antithrombotic agents were at an increased risk of ICH. Investigators designing clinical trials of antithrombotic therapies in NSTE ACS populations or experts involved in data and safety monitoring committees may find these data useful. Strategies to better identify patients at risk of ICH after NSTE ACS are needed and are part of ongoing clinical investigations, including incorporation of biomarker and genetic information.