Major Cardiovascular Events After Spontaneous Intracerebral Hemorrhage by Hematoma Location

Key Points Question Does the risk of major adverse cardiovascular events (MACEs) after an intracerebral hemorrhage (ICH) differ by hematoma location? Findings In this cohort study of 2819 patients from 2009 to 2018, compared with patients with nonlobar ICH, those with lobar ICH had significantly higher rates (per 100 person-years) of MACEs (10.84 vs 7.91) and recurrent ICH (3.74 vs 1.24) but not ischemic stroke (1.45 vs 1.77) or myocardial infarction (0.42 vs 0.64). Meaning In ICH survivors, lobar hematoma location was associated with an increased risk of MACEs, mostly driven by higher risk of recurrent ICH.


eMethods. Supplement Methods.
eSetting Acute stroke is regarded as a medical emergency and suspected cases are offered rapid evaluation with easy access to brain imaging. In Denmark, patients who have a stroke are admitted to stroke units. The Region of Southern Denmark (RSD; population 1.2 million) includes four neurological departments with stroke units and one neurosurgery department.
During the study period, brain computed tomography (CT) was the first-line imaging investigation for suspected acute stroke in hospitals in RSD.

Identification and verification of first-ever ICH cohort
As previously described, 1 we identified a cohort of first-ever intracerebral hemorrhage (ICH) in Southern Denmark through two nationwide Danish registries, the Danish Stroke Registry (Stroke Registry) and the Danish National Patient Registry (Patient Registry). We retrieved data on patients residing in Southern Denmark who were recorded under International Classification of Diseases version 10 codes for ICH in the Stroke Registry (for the period January 1, 2003the year the registry became operativeto December 31, 2018), or the Patient Registry (primary or secondary code position, admission, emergency room, or outpatient contacts for the period January 1, 2007 to December 31, 2018). To minimize misclassification of prevalent ICH cases classified as first-ever ICH, we excluded patients recorded in either registry with ICH diagnosis codes predating the study period. For a total of 4,621 patients with hospital contacts with a first-ever ICH code recorded in either registry during the study period, we retrieved primarily discharge summaries and brain scan reports using a previously validated method to ascertain the spontaneous nature of ICH and hematoma location. 2 We could not trace medical records in 191 (4%) of cases, mainly due to paper-based medical records no longer being available in the archives of some hospitals. 1 Other reasons for exclusion of patients were: ICH diagnosis verified but not spontaneous (n=816), intracranial hemorrhage other than ICH (n=469), ischemic stroke or TIA (n=122), other coding errors (n=194), and spontaneous ICH but prevalent (i.e., first-ever ICH onset before study period) (n=10). 3 A total of 2,819 patients with verified first ever spontaneous ICH were eligible for this study. After exclusion of a further 500 patients for reasons outlined in eFigure 1, the final sample for the main analysis comprised 1,034 lobar and 1,255 nonlobar patients with first-ever ICH.
Although we cannot document the exact completeness of our data regarding first-ever ICH we feel confident that it is high. However, some caveats must be kept in mind. We were not able to identify patients who died with ICH before reaching a hospital. As the rate of autopsy in Denmark is low, 4 supplementing our data with information from the Causeof-Death Registry would most likely not improve our estimates to any measurable degree. 1 Also, we did not have access to information on patients suffering an ICH that were not referred to hospital for evaluation. However, given the structure of the Danish health system, including the availability of fast-track evaluation of potential stroke, we believe that the magnitude of this selection bias is relatively small. 1

Identification and verification of recurrent stroke and spontaneous intracranial hemorrhages other than ICH
To track recurrent ICH, ischemic stroke (IS) and intracranial hemorrhages other than ICH occurring during follow-up of the inception cohort, we interrogated the Patient Registry and the Stroke Registry. In the Patient Registry, we identified in-hospital contacts with International Classification of Disorders version 10 (ICD-10) discharge diagnosis codes (primary or non-primary coding position) corresponding to ICH (ICD-10 code: I61), IS (I63, I66), stroke unspecified (I64), some less specific codes that might have been used to identify stroke (I67.8, G46), transient ischemic attack (G45.9), subarachnoid hemorrhage (SAH) (I60), subdural hematoma (SDH) (I62.0, S06.5), non-traumatic epidural hematoma (I621) and intracranial hematoma unspecified (I62.9). From the Stroke Registry we retrieved admissions coded as ICH or IS. For all admissions concerned, whether in Patient Registry or Stroke Registry, we traced medical records (including reports of initial and control brain scans). We could not trace medical records for less than 2% of interrogated hospital contacts. Rather than accept the diagnosis code at face value, we chose to exclude these contacts from further consideration.
Based on information from retrieved medical records, 4 study physicians, supervised by a neurologist with special interest in stroke (DG), classified symptomatic spontaneous (non-traumatic) intracranial events after index ICH into re-ICH, IS, TIA, and intracranial extra axial hemorrhages (ICrExH) (i.e., SAH, SDH, or epidural). In cases of doubt regarding imaging issues, original scans were reviewed.
Information on episodes of TIA was not included in this study.

Potential confounders considered for inclusion in multivariable analyses
The following were considered for inclusion as potential confounders: hypertension, atrial fibrillation (AF), previous IS, myocardial infarction (MI), peripheral arterial disease, diabetes, chronic obstructive pulmonary disease (as a marker of smoking), diagnoses indicative of high alcohol use, current use of medications [separate covariates for each of the following drug classes: platelet antiaggregants (low-dose aspirin or clopidogrel), anticoagulants (direct oral anticoagulants (DOACs) or vitamin K antagonist (VKA)), statins, and antihypertensives]. These variables were chosen on the basis of subject matter knowledge. The above potential confounders were classified at baseline, i.e., based on information recorded in registries in a fixed time window of the 15 years leading up to and including the start of followup (start date). Use of antithrombotics, statins, and antihypertensive medications was also tracked during follow-up in a sensitivity analysis (see below).

Classification of exposure to medications
We retrieved data on prescriptions dispensed to cohort members from the Danish National Prescription Registry. 5 Baseline medication exposure was classified into current use (latest prescription supply before start date lasted until start date, or ended within the 30 previous days), past use (supply ended 31-365 days before index date), and non-use (supply ended >365 days before index date or no prescription for the drug recorded in 15-year time-window). For the purposes of a sensitivity analysis, we also created time-dependent variables to capture varying exposure to platelet antiaggregants, anticoagulants, statins, or antihypertensive drugs (the latter as part of confounder control) during follow-up. Patients entered the follow-up with status of 'current use', 'past use', or 'non-use' classified as described above. A patient's status with regard to use of e.g., platelet antiaggregant drugs, could change during follow-up in a time dependent manner as further prescriptions were presented. For example, if no further prescriptions were presented 31-365 days after platelet antiaggregant drug supply run out, current use status changed to past use (or non-use, if supply had run out for >365 days). Conversely, past (or non-use) status was changed to current use if a prescription for the drug in question was presented during follow-up.

Validation of study method used to classify hematoma location
We previously investigated the validity of classifying location of the index ICH based exclusively on retrieved brain scan reports and discharge summaries (henceforth referred to as simple method) in a small sample of patients. 2 To further explore the validity of the simple method, we compared it to results achieved when using an internationally recognized rating instrument, CHARTS 6 in a large subsample of patients (all patients aged 55+ years with an index ICH in 2015-2018). For the purposes of this analysis, we included all index ICH patients fulfilling the age and time-period criteria with a single hematoma, irrespective of location. For each of these patients the first brain CT performed after the onset of the index ICH was evaluated blinded to all clinical data (including previous radiological reports) by two radiology trainees (approximately 1 year of experience), supervised by an experienced radiologist with a special interest in neuroradiology.
The simple method allows classification into 'lobar', 'non-lobar', 'isolated intraventricular hemorrhage', 'large unclassifiable ICH', and 'unclassifiable due to missing information'. The CHARTS classifications we employed were 'lobar', 'probable lobar', 'deep', 'probable deep', and 'holohemispheric'. We cross tabulated the results of the two methods and calculated kappa statistics for agreement; we also calculated positive predictive values and corresponding 95% CIs for lobar and non-lobar location according to the simple method using the CHARTS result as gold standard.

Subanalyses
Absolute event rates per 100 person-years for recurrent ICH, IS, and MI were higher in the day 0-30 period than in the day ≥31 period of follow-up (eTable 5). The type of event in the first 30-days after ICH varied by index hematoma location (i.e., the lobar cohort had a higher risk of early recurrent ICH whereas the non-lobar cohort was at higher risk of early ischemic events (i.e., IS or MI) (eTable 5). Annual IRs for the main outcomes were higher in the first year for MACE and recurrent ICH in the lobar cohort compared with the non-lobar cohort (eFigure 5).
The case fatality rate was higher and percentage of patients able to walk unaided was lower after recurrent ICH than after IS for patients with non-lobar ICH; this was also true for the ability to walk unaided in the lobar cohort (eTable 6).
Compared with the lobar cohort, patients in the non-lobar cohort with recurrent ICH had a higher case fatality on day 1 and day 7.
Absolute rates independent of other outcomes produced results similar to main analyses (eTable 7). We used this approach for calculating rates independent of other outcomes in analyses of second recurrent strokes. The IR per 100 person-years was similar between lobar and non-lobar cohorts for second ICH (lobar: 3 [thiazides and other non-loop diuretics, loop-diuretics, beta-blockers, calcium channel blockers, ACE-inhibitor and angiotensin II receptor blockers (plain or in combinations] C03A, C03D, C03E, C03C, C07, CO8, C09A, C09B, C09C, C09D eTable 2. ICH location established according to the study method (i.e, based on brain scan reports and discharge summaries) compared with re-evaluation of brain CT scans according to CHARTS by evaluators blinded to clinical data. Numbers where evaluation methods are in agreement in bold.  1 Location of hematoma on brain scan of first-ever intracerebral hemorrhage. 2 Classified based on information at baseline. 3 Medical history of ischemic stroke, myocardial infarction, or peripheral arterial disease at baseline. 4 Unadjusted hazard ratio. 5 Hazard ratio using Fine-Gray sub-distribution with all-cause death as competing event and adjusted for sex, age (<75 (ref); 75-84; 85+ years), hypertension, atrial fibrillation, previous ischemic stroke, myocardial infarction, peripheral arterial disease, diabetes, chronic obstructive pulmonary disease (as a marker of smoking), diagnoses indicative of high alcohol use, use of medications (separate covariates for each of the following drug classes: platelet antiaggregants (low-dose aspirin or clopidogrel), anticoagulants (direct oral anticoagulants or vitamin K antagonist), antihypertensives, and statins). 6 Major adverse cardiovascular event defined as stroke (ICH, or ischemic stroke), myocardial infarction, systemic embolism, or vascular death. 7 Not estimated due to sparse events. eTable 8. Absolute stroke event rates for second recurrent stroke by hematoma location of index ICH. .56) a Follow-up was performed separately for intracerebral hemorrhage and ischemic stroke, respectively. b Number of patients available for follow-up of the event being studied, e.g., 122 patients had a first recurrence of ICH and were available for follow-up for a second recurrence of ICH. c Deviates slightly from results of main analysis, as in this analysis follow-up is performed separately for ICH vs IS, whilst in main analysis a common follow-up is used. d Not reported to preserve anonymity in view of small cell counts.