Early Clinical and Radiological Course, Management, and Outcome of Intracerebral Hemorrhage Related to New Oral Anticoagulants

Importance  Intracerebral hemorrhage (ICH) is the most devastating adverse event in patients receiving oral anticoagulation. There is only sparse evidence regarding ICH related to the use of non–vitamin K antagonist oral anticoagulant (NOAC) agents.

Objective  To evaluate the early clinical and radiological course, acute management, and outcome of ICH related to NOAC use.

Design, Setting, and Participants  Prospective investigator-initiated, multicenter observational study. All diagnostic and treatment decisions, including administration of hemostatic factors (eg, prothrombin complex concentrate), were left to the discretion of the treating physicians. The setting was 38 stroke units across Germany (February 1, 2012, to December 31, 2014). The study included 61 consecutive patients with nontraumatic NOAC-associated ICH, of whom 45 (74%) qualified for the hematoma expansion analysis.

Main Outcomes and Measures  Hematoma expansion, intraventricular hemorrhage, and reversal of anticoagulation during the acute phase. Recorded were the 3-month functional outcome, factors associated with an unfavorable outcome (modified Rankin Scale score, 3-6), any new intraventricular extension or an increase in the modified Graeb score by at least 2 points, and the frequency of substantial hematoma expansion (defined as relative [≥33%] or absolute [≥6-mL] volume increase).

Results  In total, 41% (25 of 61) of patients with NOAC-associated ICH were female, and the mean (SD) patient age was 76.1 (11.6) years. At admission, the median National Institutes of Health Stroke Scale score was 10 (interquartile range, 4-18). The mean (SD) baseline hematoma volume was 23.7 (31.3) mL. In patients with sequential imaging for the hematoma expansion analysis, substantial hematoma expansion occurred in 38% (17 of 45). New or increased intraventricular hemorrhage was observed in 18% (8 of 45). Overall mortality was 28% (17 of 60 [follow-up data were missing in 1 patient]) at 3 months, and 65% (28 of 43) of survivors had an unfavorable outcome (modified Rankin Scale score, 3-6). Overall, 57% (35 of 61) of the patients received prothrombin complex concentrate, with no statistically significant effect on the frequency of substantial hematoma expansion (43% [12 of 28] for prothrombin complex concentrate vs 29% [5 of 17] for no prothrombin complex concentrate, P = .53), or on the occurrence of an unfavorable outcome (modified Rankin Scale score, 3-6) (odds ratio, 1.20; 95% CI, 0.37-3.87; P = .76).

Conclusions and Relevance  Non–vitamin K antagonist oral anticoagulant–associated ICH has a high mortality and an unfavorable outcome, and hematoma expansion is frequent. Larger-scale prospective studies are needed to determine whether the early administration of specific antidotes can improve the poor prognosis of NOAC-associated ICH.

Intracerebral hemorrhage (ICH) is responsible for most deaths caused by bleeding complications during long-term anticoagulation.^1,2 Because intracerebral hematoma size and secondary hematoma expansion are important prognostic factors in ICH,³ prevention of hematoma expansion is a major therapeutic target of ICH management. Intracerebral hemorrhage during anticoagulation with vitamin K antagonists (VKAs) accounts for 10% to 25% of all occurrences ICH.^4,5 It is associated with a higher risk and prolonged period of hematoma growth,^6,7 as well as with a higher mortality compared with ICH in patients not receiving anticoagulation.^6,8 Compared with VKAs, all non-VKA oral anticoagulant agents (NOACs) carry a substantially lower risk of intracranial hemorrhage.^9,10 Nevertheless, given the rising prescription rates,¹¹ the occurrence of ICH during treatment with NOACs has become an important issue.

Three large randomized clinical trials consistently reported a mortality of NOAC-associated ICH ranging between 45% to 67%, and most survivors had permanent disability.^2,12,13 Despite this profound effect on the long-term outcome, the characteristics and natural history of NOAC-associated ICH in the acute phase are largely unknown. To our knowledge, there are no prospective data on hematoma expansion or on the effect of hemostatic management in patients receiving NOACs. The available evidence is limited to small retrospective studies^14-24 without detailed analysis of the clinical and radiological course of NOAC-associated ICH, and there is little information on the effectiveness of unspecific hemostatic factors. Clinical guidelines^25-27 and 2 recent large observational studies^7,28 regarding VKA-associated ICH support reversing the effect of VKAs using coagulation factors to reduce the risk of hematoma expansion. However, the extrapolation of treatment concepts from VKA-associated ICH to NOAC-associated ICH has limitations because NOACs and VKAs have dissimilar pharmacokinetics, and their effects on hemostasis in the brain may differ.^12,13,29-31 Moreover, although specific antidotes to reverse anticoagulation with NOACs are being tested clinically, none of them are available for clinical routine practice.^32,33 Unspecific hemostatic factors such as prothrombin complex concentrate are effective in preclinical models of NOAC-associated ICH and in healthy volunteers, but their effectiveness in acute bleeding is unknown.^34-37 Despite the absence of evidence in patients, current expert recommendations suggest antagonization of the effect of NOACs by using prothrombin complex concentrate.^29,30

Herein, we report the results from the ICH substudy of the Registry of Acute Stroke Under New Oral Anticoagulants (RASUNOA). The RASUNOA was a prospective multicenter observational study designed to describe the clinical and radiological course, management, and outcome of ICH during therapy with NOACs in routine clinical practice.

The RASUNOA is an investigator-initiated, multicenter, prospective, observational registry assessing the clinical and radiological course, management, and outcome after acute stroke related to NOAC use (clinicaltrials.gov identifier: NCT01850797). It involves 38 departments of neurology with certified stroke units in Germany. Study approval was obtained from the ethics committee of the Medical Faculty of Heidelberg, Heidelberg, Germany, as well as the ethics committees of each participating center.

Between February 1, 2012, and the per-protocol agreed end of patient inclusion on December 31, 2014, patients with acute nontraumatic ICH fulfilling the following eligibility criteria were included in the ICH substudy: age 18 years or older, NOAC therapy (ie, apixaban, dabigatran etexilate, or rivaroxaban) at the time of ICH, and receipt of written informed consent by the patient or a legal representative. Intracerebral hemorrhage had to be present on baseline neuroimaging (computed tomography or magnetic resonance imaging). There were no exclusion criteria regarding the modified Rankin Scale (mRS) score before the index ICH.

All diagnostic and treatment decisions were left to the discretion of the treating physicians. These orders included performance of follow-up imaging, selection of imaging modalities, and administration of hemostatic factors (eg, prothrombin complex concentrate).

Observational data were collected by staff members of local centers using a paper-based case report file to document baseline characteristics, including cardiovascular risk factors, clinical observations, and laboratory findings. Double data entry was performed by 2 independent staff members of the Institute of Medical Biometry and Informatics, University of Heidelberg, Heidelberg, Germany.

Neurological status was assessed using the National Institutes of Health Stroke Scale (NIHSS) score at admission, as well as 24, 48, and 72 hours later. Functional outcome was determined using the mRS score before stroke, at hospital admission, at the time of discharge, and during follow-up. The CHA₂DS₂VASc score (Cardiac Failure or Dysfunction, Hypertension, Age ≥75 Years [Doubled], Diabetes, Stroke [Doubled], Vascular Disease, Age 65-74 Years, and Sex Category [Female]) and the HAS-BLED score (Hypertension, Abnormal Renal/Liver Function, Stroke, Bleeding History or Predisposition, Labile INR [international normalized ratio], Elderly, Drugs/Alcohol Concomitantly) were calculated, excluding the index event.^38,39 The HAS-BLED score item “labile INR” was set to zero.

A structured telephone follow-up was performed by trained mRS score raters of local centers 90 days after ICH and included the mRS score and current antithrombotic medication. If the patient was unable to be contacted in person, the interview was performed with a close relative, legal representative, or family physician familiar with the current functional and medical status of the patient.

Volumetric measurements of intracerebral hematoma volume were performed on 2 identical sets of computed tomographic and magnetic resonance imaging data by 2 independent, experienced readers (J.C.P. and M.W.) masked to patient characteristics. We used an open-source database and viewer (OsiriX and DICOM; Pixmeo). Regions of interest around intraparenchymal hemorrhage, excluding intraventricular hemorrhage (IVH), were drawn manually on each section. Hematoma volume was calculated using the region-of-interest volume calculator. In case of volume differences exceeding 30% or technical problems, images were reassessed by both readers to seek consensus. We used the arithmetic mean of the estimates obtained by both readers for further analysis. Based on a semiquantitative scale for IVH volume measurement, the modified Graeb score was calculated according to the method by Hinson et al.⁴⁰

The rate of hematoma expansion was prespecified as a primary aim of the study. Hematoma expansion was determined if sequential brain imaging was available. Substantial hematoma expansion was defined as a relative increase in hematoma volume by at least 33% or an absolute increase by at least 6 mL compared with the initial imaging. Substantial intraventricular expansion was defined as the occurrence of any new intraventricular hematoma expansion or an increase in the modified Graeb score by at least 2 points. To qualify for the hematoma expansion analysis, follow-up imaging had to be performed within 3 to 72 hours after the first imaging. If more than 1 follow-up imaging session was performed within the time frame, the one closest to the 24-hour time point was chosen. Patients with hematoma evacuation before any follow-up imaging within the time frame were excluded from the hematoma expansion analysis.

Continuous variables were described by the mean (SD) or the median (interquartile range [IQR]). For categorical variables, absolute and relative frequencies were reported. The Shapiro-Wilk test was used to ascertain the distribution of data. The χ² test or Fisher exact test, as appropriate, was used to compare the proportions of baseline and hematoma characteristics between patients with vs without follow-up images, with vs without hematoma expansion, and with vs without prothrombin complex concentrate administration. To compare continuous variables, the nonparametric Mann-Whitney test was used because of the skewness of the data. Bivariate correlations by the Kendall τ rank correlation were used to assess the association between hematoma volume at baseline and patient characteristics. Univariate logistic regression analyses were conducted to analyze the association of demographic and clinical characteristics with an unfavorable outcome (mRS score, 3-6) at the 3-month follow-up. In case of incomplete separation, Firth logistic regression was used if there was no outcome observation for a given category. In addition, a macro (SAS %fl; SAS Institute Inc) was applied to estimate odds ratios (95% CIs) using penalized likelihood estimation methods and associated penalized likelihood ratio tests.^41-43 Because of the limited number of patients, no multivariable analyses were performed. All statistical analyses were conducted using statistical software (IBM SPSS Statistics, version 22; IBM SPSS and SAS, version 9.3; SAS Institute Inc).

In total, 61 patients were enrolled in the study. Of the 38 sites participating in the RASUNOA, 21 reported at least 1 patient with a NOAC-associated ICH. Four of 21 centers reporting patients with NOAC-associated ICH enrolled 39 of the patients (64%). Within these 4 centers, 71% (39 of 55) of all eligible patients were included. With regard to these 4 centers, no statistically significant differences were found at an aggregated level between all patients with NOAC-associated ICH treated in the study period (including deceased ones) and the patients included in the study.

Table 1 summarizes baseline characteristics of the study cohort. Patients had a mean (SD) age of 76.1 (11.6) years (range, 46-97 years) and had a moderate to severe neurological deficit at admission (median NIHSS score, 10; IQR, 4-18). The median time since the last NOAC intake to the first brain imaging was 14.3 hours (IQR, 6.0-22.8) (Table 2).

The median baseline hematoma volume at presentation was 10.8 mL (IQR, 4.0-30.0) (Table 2). A lobar location (41% [25 of 61]) was slightly more frequent than a deep location (38% [23 of 61]) of hematoma. Neurological status (NIHSS score) at admission was correlated with baseline hematoma volume (Kendall τ = 0.347, P < .001) (eTable 1 in the Supplement). The elapsed time between symptom onset and initial brain imaging was not significantly correlated with baseline hematoma volume. Six patients (10%) were receiving concomitant treatment with at least 1 platelet inhibitor (Table 1). These patients were older and had significantly higher baseline hematoma volumes compared with patients without concomitant platelet inhibition (median, 30 [IQR, 26-39] vs 9 [IQR, 3-19] mL; P = .03).

Hematoma expansion could be analyzed in 45 patients with sequential cranial imaging within 3 to 72 hours (median time between imaging, 21.1 hours; IQR, 13.2-27.5). Baseline characteristics of the hematoma expansion analysis group did not differ significantly from those of the entire study cohort (Table 1). Substantial hematoma expansion occurred in 38% (17 of 45) of patients (Table 2). In 7% (3 of 45), intraventricular extension had developed since the initial imaging, and a relevant increase in intraventricular bleeding was observed in an additional 11% (5 of 45).

Thirty-five of the 61 patients (57%) received 4-factor prothrombin complex concentrate (mean [SD] dose, 2390 [980] IU). Patients receiving prothrombin complex concentrate had a worse clinical status and tended to more frequently have a deep hemorrhage (eTable 2 and eTable 3 in the Supplement). Larger baseline hematoma volumes were found in the subgroup of patients administered prothrombin complex concentrate who received follow-up imaging (P = .04). Moreover, the interval between the last NOAC intake and the initial brain imaging tended to be shorter in patients receiving prothrombin complex concentrate. Administration of prothrombin complex concentrate had no statistically significant effect on the early hematoma expansion (43% [12 of 28] for prothrombin complex concentrate vs 29% [5 of 17] for no prothrombin complex concentrate, P = .53) and the functional outcome at 3 months (Table 3 and Table 4).

Overall mortality rates were 16% (10 of 61) during the acute inpatient stay and 28% (17 of 60) (data missing for 1 patient) at 3 months (Figure). Sixty-five percent (28 of 43) of the survivors had an unfavorable outcome (mRS, 3-5). There was a strong association between clinical deficit at admission and death and dependency at 3 months (Table 4). Larger baseline hematoma volume (OR, 2.37; 95% CI, 1.02-5.53, P = .046) and intraventricular extension at baseline (OR, 8.13; 95% CI, 1.64-40.27; P = .01) were associated with an unfavorable outcome (mRS score, 3-6), including death at 3 months. In contrast, no statistically significant association with an unfavorable outcome was found for substantial hematoma expansion. The 43 survivors at 90 days after the hemorrhagic event had a median mRS score of 4 (IQR, 2-5). Five of the 61 enrolled patients (8%) underwent surgical hematoma evacuation, of whom 4 were still alive at 3 months. A sensitivity analysis restricted to the most frequently used NOAC among the study population showed no major differences in demographic and clinical characteristics, as well as main outcomes, compared with the entire group with NOAC-associated ICH (eTable 4 and eTable 5 in the Supplement).

In terms of subsequent stroke prevention, no patient received anticoagulation at the time of discharge from the acute hospital. In 10 of the 43 survivors (23%), oral anticoagulation had been resumed by day 90 after the hemorrhagic event.

Our prospective observational study provides major new insights into the clinical and radiological course, management, and outcome of NOAC-associated ICH. The characteristics of NOAC-associated ICH at baseline, including hematoma volume and location, are similar to those previously reported for VKA-associated ICH.^6,7 Subsequent substantial hematoma expansion occurred in more than one-third (38% [17 of 45]) of patients with NOAC-associated ICH herein. Although recommended by current expert guidance,^29,30 little more than half (57% [35 of 61]) of the patients received prothrombin complex concentrate for anticoagulation reversal, but no statistical significant association with outcome was observed.

The mean (SD) hematoma volume in NOAC-associated ICH herein (24 [31] mL) was at the upper end of the range previously reported for ICH in patients not receiving anticoagulation (13-26 mL)^3,4,44,45 but was within the range reported for VKA-associated ICH (14-48 mL) (eTable 6 in the Supplement).^4,45,46 Moreover, our observed intraventricular hematoma extension had a frequency similar to that reported for VKA-associated ICH (international normalized ratio, <3.0).^{3,4,6,7,28,44,47} Therefore, our prospective data do not support a previous retrospective study⁴⁸ showing smaller ICH volumes in NOAC-associated ICH compared with VKA-associated ICH. Hematomas were frequently found in a lobar location, possibly reflecting the increased contribution of cerebral amyloid angiopathy to the risk of intracerebral bleeding in the elderly.^5,49 Compared with investigations⁷ on VKA-associated ICH, the proportion of patients with previous stroke, including prior ICH, was higher herein. This finding might reflect the fact that, given the reduced risk of ICH, NOACs are considered first-line treatment for patients at high risk of ICH.¹⁰

Our definition of substantial hematoma expansion included thresholds for relative (≥33%) or absolute (≥6-mL) volume increase. As already noted, substantial hematoma expansion was found in 38% (17 of 45) of our patients with NOAC-associated ICH. This proportion is within the range reported for VKA-associated ICH (36%-56%)^6,7,50 and is higher compared with that related to ICH in patients not receiving anticoagulation (12%-26%).^4,51,52 When limiting the analysis to the relative increase in ICH volume only, the rate of hematoma expansion was at the upper end of a range reported for ICH in patients not receiving anticoagulation,^2,3,28,29 almost identical to that in a large recent study⁷ on patients with VKA oral anticoagulation, and resembled that of other VKA-associated ICH investigations.³ We found no association of substantial hematoma expansion with the dichotomized 3-month functional outcome because the initial hematoma size largely determined an unfavorable outcome. Initial IVH was associated with an unfavorable 3-month outcome. Secondary IVH extension and IVH expansion occurred in 18% (8 of 45) of our patients, in accord with data reported for VKA-associated ICH (6%-18%).^53,54

No specific NOAC antidote has been approved for the reversal of anticoagulation in severe hemorrhage in routine clinical practice. Studies in experimental ICH^34,35 and healthy volunteers^36,37 suggested the efficacy of prothrombin complex concentrate for the reversal of NOACs. Consequently, current expert guidance suggests administering 30 to 50 IU per kilogram of body weight of prothrombin complex concentrate in severe acute hemorrhage.^18,19 However, we failed to observe any association of the use of prothrombin complex concentrate with hematoma expansion and outcome in our observational study. This finding might be because patients receiving prothrombin complex concentrate had different baseline characteristics, including a more frequent deep hematoma location and a more severe initial neurological deficit being associated with a poor outcome. In addition, in contrast to preclinical NOAC-associated ICH investigations and the optimal efficacy in VKA-associated ICH,^7,28,34,35 prothrombin complex concentrate was often not administered in the first 5 hours after symptom onset. Our study design, the limited sample size, and the potential for confounding by indication do not allow any conclusions regarding a potential association between prothrombin complex concentrate treatment and outcome. Although idarucizumab, an antibody fragment binding to dabigatran, showed effective reversal of anticoagulation with dabigatran³³ and other antidotes are under development,³² their effectiveness in NOAC-associated ICH remains to be shown.

Our study had some limitations. Vitamin K antagonist–associated ICH and ICH in patients not receiving anticoagulation were compared by referring to previously published, mostly retrospective observational studies. Therefore, validation of the results in future prospective studies using matched control groups is necessary, and current comparisons should be interpreted with caution. Recruitment was incomplete in most participating centers. However, we performed a sensitivity analysis, including aggregated baseline characteristics of eligible patients at the 4 top recruiting centers, which revealed no statistically significant differences compared with the patients actually included in our study. Furthermore, some patients may have died before informed consent could be obtained, and these data could not be included, as mandated by the ethics committee. However, although this limitation may have led to underestimation of mortality, the observed case fatality in our data set is consistent with previous data.⁵⁵ In addition, primary neurosurgical patients were not enrolled at all centers; therefore patients with a worse clinical status might be underrepresented, although the short-term mortality in our cohort is consistent with data from a recent retrospective neurosurgical case series.²³ Our study was not intended to detect potential differences in ICH-related features among different agents, classes of agents (ie, direct thrombin inhibitors vs factor Xa inhibitors), or various doses of a specific NOAC. In addition, the small sample size hampers more detailed statistical analyses. However, we are planning a larger prospective cohort (RASUNOA-Prime [clinicaltrials.gov identifier: NCT02533960]). Finally, because only a limited number of patients underwent very early imaging and sequential imaging for the hematoma expansion analysis was only available in 74% (45 of 61) of our patients, the actual rate of hematoma expansion may be inadequately estimated. It remains to be shown that successful reversal of anticoagulation translates into improved clinical outcome via prevention of hematoma expansion. Despite these limitations, the present study, to our knowledge, represents the largest, most detailed, and only prospective analysis of NOAC-associated ICH to date.

Intracerebral hemorrhage related to NOAC use is associated with a high mortality and an unfavorable outcome, and hematoma expansion is frequent. Larger-scale prospective studies are needed to determine whether the early administration of specific antidotes can improve the poor prognosis of NOAC-associated ICH.

Accepted for Publication: October 1, 2015.

Corresponding Author: Roland Veltkamp, MD, Department of Stroke Medicine, Imperial College London, Charing Cross Campus, 3 East 6 Fulham Palace Rd, London W6 8RF, England (r.veltkamp@imperial.ac.uk).

Published Online: December 14, 2015. doi:10.1001/jamaneurol.2015.3682.

Author Contributions: Drs Purrucker and Veltkamp had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Purrucker, Rizos, Veltkamp.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Purrucker, Haas, Heuschmann, Veltkamp.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Haas, Heuschmann.

Study supervision: Veltkamp.

Conflict of Interest Disclosures: Dr Purrucker reported receiving travel and congress participation support from Pfizer, outside of the present work. Dr Rizos reported receiving consulting honoraria, speaker honoraria, travel support, or research support from Boehringer Ingelheim, Bayer, BMS Pfizer, and Portola. Dr Poli reported receiving personal fees from Boehringer Ingelheim and Bayer. Dr Steiner reported receiving speaker fees and consultant honoraria from Boehringer Ingelheim, BMS Pfizer, Bayer, and Daiichy Sanyo. Dr Heuschmann reported receiving grants from the German Federal Ministry of Education and Research, European Union, Charité, Berlin Chamber of Physicians, German Parkinson Society, University Hospital Würzburg, The Robert Koch Institute, Charité–Universitätsmedizin Berlin (within MonDAFIS, supported by an unrestricted research grant to the Charité from Bayer), University Göttingen (within FIND-AF [randomized], supported by an unrestricted research grant to the University Göttingen from Boehringer-Ingelheim), and University Hospital Heidelberg (within RASUNOA-Prime, supported by an unrestricted research grant to the University Hospital Heidelberg from Bayer, BMS, Boehringer-Ingelheim), outside of the present work. Dr Veltkamp reported receiving speaker fees, consulting honoraria, and research support from Bayer, Boehringer Ingelheim, BMS Pfizer, Daiichi Sanyo, and CSL Behring. No other disclosures were reported.

Additional Contributions: We thank all the RASUNOA investigators, including participating centers, local physicians, radiologists, and study nurses, for their support during the study. The RASUNOA investigators are listed in the eAppendix in the Supplement.