Safety of Discontinuation of Anticoagulation in Patients With Intracranial Hemorrhage at High Thromboembolic Risk FREE

ANNUALLY, intracranial hemorrhage (ICH) causes complications in 1% to 2% of patients with a prosthetic heart valve or atrial fibrillation who are taking warfarin for thromboembolic prophylaxis.^{1 - 2} Rapid reversal of anticoagulation in the presence of ICH is needed to prevent enlargement of the hematoma, resulting in brain herniation. While not taking warfarin, the immediate concern in an unprotected patient is ischemic stroke and systemic embolization. In addition, the interval to resumption of anticoagulation and its associated potential risk of recurrence of cerebral hemorrhage is not known. Discontinuation of warfarin seems safe, but lack of strong scientific underpinnings continues to disconcert physicians confronted with these complex patients. Data have only come from small retrospective series.^{3 - 7} To assess the risk of discontinuation and resumption of warfarin treatment in patients with ICH, we conducted a hospital-based study in a large cohort of patients.

We retrospectively reviewed the medical records of 141 patients undergoing anticoagulation who had ICH between January 1, 1976, and December 31, 1999. This study was approved by the Institutional Review Board. Twenty-six patients with a metallic valve and ICH have been described previously.⁷ Patients with ICH who were taking anticoagulation treatment for deep venous thrombosis or pulmonary embolus were considered to be at low risk of embolic stroke and were excluded from the analysis. The hospital course and 30-day follow-up were reviewed until the first defining event, any embolic complication, ICH, or death. Abnormal transthoracic or transesophageal echocardiographic results were defined as a left atrial size greater than 40 mm, an ejection fraction of less than 40%, hypokinesis, a septal defect, and thrombus.^{8 - 9} If patent foramen ovale was the only abnormality on echocardiography, then it was coded as a normal echocardiographic study because of controversy over its significance as an independent risk factor for ischemic stroke.^{9 - 10} In addition, none of the patients in our series were taking warfarin because of patent foramen ovale. Either intravenous heparin or oral warfarin treatment was then restarted, depending on the treating physicians.

Intracranial hemorrhage was divided into intracerebral hemorrhage (including cerebellar hemorrhage, brainstem hemorrhage, lobar hemorrhage, and basal ganglia hemorrhage), subarachnoid hemorrhage (SAH), subdural hematoma (SDH), and primary intraventricular hemorrhage (IVH). Patients with a hemorrhagic infarct were excluded from the analysis. Patients were divided into 3 groups according to the indications for long-term anticoagulation. Group 1 consisted of 52 patients with prosthetic heart valves. Group 2 consisted of 53 patients with atrial fibrillation and cardioembolic stroke.¹¹ Group 3 consisted of 36 patients with recurrent transient ischemic attacks or ischemic stroke despite therapy with aspirin, ticlopidine hydrochloride, or clopidogrel.

Statistical analysis using the log-rank test was used to compare the likelihood of mortality among the various types of ICH. The 30-day risk of ischemic stroke while not taking anticoagulation treatment was determined using Kaplan-Meier survival estimates (censored for when heparin or warfarin therapy was restarted).

We studied 141 patients with 146 episodes of ICH while taking warfarin. The median patient age was 74 years (range, 23-98 years). Sixty percent of the patients were men. The demographic characteristics are shown in Table 1. The type of hemorrhage was intracerebral hemorrhage in 87 patients (61.7%), SDH in 43 (30.5%), SAH in 8 (5.7%), and IVH in 3 (2.1%). The median duration of anticoagulation for the cohort taken from the start of warfarin therapy to admission with ICH was 5 years (range, 2 months to 14.7 years). A prior transient ischemic attack or ischemic stroke occurred in group 1 in 27% and in group 2 in 43% of the patients. The median admission Glasgow Coma Scale score was 14. Anticoagulation was reversed with fresh frozen plasma and vitamin K on admission in all patients. The median and mean prothrombin times were 26.4 and 21.3 seconds (range, 9.7-240.0 seconds) on admission for the 3 groups. In group 1, 14 patients had mitral metallic valves, 31 had aortic metallic valves, and 7 had combined mitral and aortic metallic valves. In this group, 14 patients had Björk-Shiley (single-tilting disk) valves, 17 had Starr-Edwards (caged-ball) valves, and 11 had St Jude (bileaflet–tilting disk) valves. Seventeen patients had various types of valves, including 3 Medtronic-Hall (single–tilting disk) valves, 3 Braunwald-Cutter (caged-ball) valves, and 2 Carbomedic (bileaflet-tilting disk) valves; 9 had bioprosthetic valves.

The median time not taking warfarin was 10 days (range, 0-30 days). There were 3 patients with ischemic events within 30 days: 1 in the posterior cerebral artery region, 1 lacunar stroke, and 1 vertebrobasilar transient ischemic attack (Table 2). Using Kaplan-Meier survival estimates, the probability of having ischemic events at 7 days following warfarin treatment cessation in groups 1, 2, and 3 was 2.9% (95% confidence interval [CI], 0%-8.0%), 2.6% (95% CI, 0%-7.6%), and 4.8% (95% CI, 0%-13.6%), respectively. This remained unchanged at 14 and 30 days after ictus. By 7 day, 18 (35%) of the patients in group 1 died, 16 (30%) in group 2 died, and 18 (50%) in group 3 died. By day 14, 20 (38%) of the patients in group 1 died, 20 (38%) in group 2 died, and 18 (50%) in group 3 died. By day 30, 22 (42%) of the patients in group 1 died, 22 (42%) in group 2 died, and 18 (50%) in group 3 died. Anticoagulation treatment with either intravenous heparin or warfarin was restarted by day 7 in 7 (13%) of the patients in group 1, 4 (8%) in group 2, and 1 (3%) in group 3. By day 14, the frequency of anticoagulation was 26 patients (50%) in group 1, 7 (13%) in group 2, and 1 (3%) in group 3. By day 30, the frequency of anticoagulation was 28 patients (54%) in group 1, 12 (23%) in group 2, and 5 (14%) in group 3. An abnormal echocardiographic result was found in 30 (65%) of those in group 1, 20 (56%), in group 2, and 3 (21%) in group 3.

Of the 35 patients in whom anticoagulation therapy was restarted with intravenous heparin or oral warfarin, none had recurrence of ICH during hospitalization.

There were 45 patients who underwent neurosurgical procedures, ranging from external ventricular drain to craniectomy. This consisted of 14 patients in the ICH group, 28 in the SDH group, and 3 in the SAH and IVH group. In the SDH group, 6 patients did not undergo surgery and died. All except 1 patient had a large SDH with mass effect and a low Glasgow Coma Scale score (<5) at presentation. Warfarin treatment was restarted (5-72 days after ictus) without evacuation of the hematoma in 4 patients with small layers of SDH; none of these patients developed further clinical deterioration.

Death occurred in 43% of the patients and within 26 days (median, 2.8 days) of hospitalization. Causes of death were cerebral hemorrhage (n = 46), respiratory failure (n = 4), ischemic heart disease (n = 3), septicemia (n = 2), and uncertain (n = 6). The mortality according to the type of ICH was as follows: intracerebral hemorrhage, 48%; SDH, 20%; SAH, 38%; and IVH, 100%. The difference in mortality between the intracerebral hemorrhage group and the SDH group reached statistical significance (P =.005).

The management of anticoagulation in these patients with ICH has yet to be established, and guidelines cannot be obtained from the existing literature.^{3 - 6} Most of the patients described in prior studies had mechanical heart valves, and the same observation cannot be extrapolated to patients taking warfarin for atrial fibrillation or recurrent stroke, despite antiplatelet therapy. Babikian et al³ did not observe any ischemic events in 6 patients over a median of 19 days (range, 8-42 days) while they were not taking warfarin for ICH. Butler and Tait⁴ observed cerebral ischemic events in 3 of 13 patients. Kawamata et al⁶ did not find any ischemic strokes when warfarin treatment was discontinued for 3 days in 27 patients with ICH. An analogy to the risk of warfarin discontinuation in the setting of ICH in high-risk patients for ischemic stroke is the risk in patients who have mechanical heart valves undergoing elective surgery. Tinker and Tarhan¹² retrospectively reviewed the risk of discontinuing warfarin therapy in 159 patients with mechanical heart valves undergoing elective surgery, and none of the patients had in-hospital thromboembolic complications. Further indirect evidence to support discontinuation of warfarin therapy in these patients with ICH comes from the recent International Stroke Trial. Early recurrence of ischemic stroke in patients with atrial fibrillation was low even while taking aspirin.¹³

Our data help put into context the potential hazards of management of anticoagulation in patients with ICH. Warfarin was given for the prevention of stroke in patients selected for this study on the premise of a high risk of embolization. We found that the estimated risk of having an ischemic stroke after discontinuation of warfarin therapy within 30 days was relatively low. The cumulative risk for ischemic stroke at 30 days in patients with a metallic valve was 3% (95% CI, 0%-8%); in patients with atrial fibrillation, 3% (95% CI, 0%-7%); and in those with a recurrent transient ischemic attack or minor stroke, 5% (95% CI, 0%-13%). This is an uncontrolled observational study, but we believe bias from selection of patients at low risk is highly unlikely. Prior ischemic stroke before ICH occurred in 27% of the group 1 patients and in 43% of the group 2 patients. In addition, the median time not taking warfarin was 10 days, not only allowing for ICH to stabilize or to treat definitively with surgical intervention but to also observe the risks of a treatment-free interval. In the group 1 patients, 14 had a mitral valve and 20 had caged-ball (Starr-Edwards and Braunwald-Cutter) valves, conditions associated with a high embolic risk.¹⁴ We acknowledge that the long period involved in this study may introduce bias because of changing clinical practice, but despite the 23-year period we were able to find only 141 patients who had ICH while taking warfarin and were at high risk of thromboembolic events. This is reflected in the scant data in the literature on how to manage such patients.

Recurrence of ICH with subsequent deterioration after starting anticoagulation with intravenous heparin or oral warfarin was low. In fact, no ICH recurred during the same hospitalization. Kawamata et al⁶ did not observe any early recurrent cerebral hemorrhage in their neurosurgical series when anticoagulation was started within 3 days. The alternative to discontinuing warfarin therapy in the presence of ICH is to maintain anticoagulation with heparin. Only 1 group¹⁵ has reported data on the safety of maintaining anticoagulation with heparin in the presence of ICH in 4 patients with a mechanical heart valve. Leker and Abramsky¹⁵ did not find enlargement of the hematoma when they continued anticoagulation therapy with heparin. It is possible that subtle enlargement occurred, not causing any clinically detectable deterioration in patients with SDH or intracerebral hematoma. However, to answer this question, a serial computed tomographic scan study is needed. In view of the limited data on the safety of continuation of warfarin in this clinical dilemma, our data on the relative safety of discontinuation of warfarin therapy warrant consideration. In addition, there are no safety data on the continuation of anticoagulation in the presence of SDH, IVH, and SAH.

Our study showed that brief (1- to 2-week) discontinuation of anticoagulation therapy in patients with a high embolic risk is relatively safe. These results compare with the long-term risk of embolic events of 4% per year in patients with a metallic valve without anticoagulation but are lower than the 10% to 12% per year found in patients with nonvalvular atrial fibrillation.^{1 - 2} Our study provides observations that should be of value to neurologists and cardiologists who manage this clinical dilemma.

Accepted for publication May 15, 2000.

This study was supported by fund IRB 1264-98 from the Neurologic Research Committee, Mayo Clinic, Rochester, Minn.

Reprints: Eelco F. M. Wijdicks, MD, Department of Neurology, Mayo Clinic, W8B, 200 SW First St, Rochester, MN 55905 (e-mail: wijde@mayo.edu).