Optimal Timing of Resumption of Warfarin After Intracranial Hemorrhage
Background and Purpose—The optimum timing of resumption of anticoagulation after warfarin-related intracranial hemorrhage in patients with indication for continued anticoagulation is uncertain. We performed a large retrospective cohort study to obtain more precise risk estimates.
Methods—We reviewed charts of 2869 consecutive patients with objectively verified intracranial hemorrhage over 6 years at 3 tertiary centers. We calculated the daily risk of intracranial hemorrhage or ischemic stroke with and without resumption of warfarin; we focused on patients who survived the first week and had cardiac indication for anticoagulation or previous stroke. Using a Cox model, we estimated rates for these 2 adverse events in relation to different time points of resumed anticoagulation. The combined risk of either a new intracranial hemorrhage or an ischemic stroke was calculated for a range of warfarin resumption times.
Results—We identified warfarin-associated intracranial hemorrhage in 234 patients (8.2%), of whom 177 patients (76%) survived the first week and had follow-up information available; the median follow-up time was 69 weeks (interquartile range [IQR] 19–144). Fifty-nine patients resumed warfarin after a median of 5.6 weeks (IQR 2.6–17). The hazard ratio for recurrent intracranial hemorrhage with resumption of warfarin was 5.6 (95% CI, 1.8–17.2), and for ischemic stroke it was 0.11 (95% CI, 0.014–0.89). The combined risk of recurrent intracranial hemorrhage or ischemic stroke reached a nadir if warfarin was resumed after approximately 10 to 30 weeks.
Conclusion—The optimal timing for resumption of warfarin therapy appears to be between 10 and 30 weeks after warfarin-related intracranial hemorrhage.
Intracranial hemorrhage is the most feared complication of treatment with vitamin K antagonists. The incidence of this hemorrhage is approximately 0.2% per year of treatment.1 Despite the underlying risk of a thromboembolic event in patients requiring long-term prophylaxis, an intracranial hemorrhage necessitates temporary discontinuation of a vitamin K antagonist. Depending on the indication for prophylaxis, the withdrawal of anticoagulation exposes the patient to a substantial risk of thromboembolic complications. Annual rates of 12% to 22% have been reported in patients with St. Jude’s Medical bileaflet heart valves,2 4% to 18% in patients with atrial fibrillation and additional risk factors for stroke,3 and 10% for patients with recurrent venous thromboembolism.4 The decision about when to resume anticoagulation requires the physician to balance the risk of an additional bleeding event (particularly of recurrent intracranial hemorrhage, the risk of which is likely high shortly after the index bleed, although poorly studied) with the risk of thromboembolism in the absence of anticoagulation.
A limited number of observational studies have addressed the question posed in this study.5–12 These studies generally included few patients who were followed at a single center and had limited follow-up. Several of the authors and the American Stroke Association13 suggest restarting anticoagulation after a period of 7 to 14 days following intracranial hemorrhage. A recent systematic review of studies in patients with mechanical heart valves reached the same conclusion based on only 2 recurrent hemorrhages and 4 strokes.14 Seven experts expressed the opinion that warfarin could be restarted in stable patients 3 to 10 days after intracranial hemorrhage.15
We performed this study to address the timing of resumption of warfarin after intracranial hemorrhage and assess the risks and benefits both in patients who restarted warfarin and those who did not. To our knowledge, this is the first multicenter study with analysis of the optimal timing of resumption of anticoagulation performed by calculating hazards of recurrent hemorrhage and of ischemic stroke, with or without warfarin.
The study was conducted at 3 tertiary care hospitals: Karolinska University Hospital in Solna-Stockholm, and Karolinska University Hospital in Huddinge-Stockholm, both in Sweden; and at Hamilton Health Sciences-General Hospital, Ontario, Canada. One investigator in each country (A.M. in Sweden and Y.-K.K. in Canada) reviewed the charts of all patients admitted between January 2002 and December 2007 in Canada, and between January 2004 and December 2008 in Sweden, who had a diagnosis of intracranial hemorrhage (ICD-10 diagnosis code I600-I629). Data from patients with warfarin-associated intracranial hemorrhage underwent a final review by a third investigator (S.S.).
The Institutional Ethics Committees in both countries approved this retrospective study and waived the need for consent from patients. Patients who were being treated with warfarin at the time of diagnosis of intracranial hemorrhage, and who had an international normalized ratio of >1.5 at the time the bleeding occurred, were included in the study. The diagnosis had to be radiologically confirmed. Any combination of intracerebral hemorrhage with subdural or subarachnoid hemorrhage was classified as intracerebral. We excluded from our analysis patients with hemorrhage caused by severe accidents, with multi-trauma, or those with hemorrhagic transformation of ischemic stroke.
Data regarding patient characteristics, indication for warfarin treatment, risk score for stroke in patients with atrial fibrillation,3 additional antithrombotic treatment, international normalized ratio at presentation, cause of bleeding (spontaneous or traumatic), location of bleeding according to radiological investigations, and neurosurgical intervention were collected. From the follow-up period after intracranial hemorrhage, we collected data on resumption of warfarin, recurrent intracranial bleeding, ischemic stroke, systemic embolism, transitory ischemic attacks, heart valve thrombosis, venous thromboembolism, death, and the timing of these events in relation to the index intracranial hemorrhage. The data were retrieved from electronic hospital charts and by contacting each patient’s family physician.
Our aim was to describe the occurrences of recurrent bleeding, thromboembolic complications, and death in patients surviving the first week after the index intracranial hemorrhage. The reason for this focus was that resumption of anticoagulation with vitamin K antagonists during the first week has rarely been recommended. Furthermore, patients with such short survival are usually very ill and are not considered for anticoagulant therapy during this limited period. In addition, we focused the analysis on patients who were at a moderate-to-high risk of ischemic stroke, including those with atrial fibrillation, mechanical heart valves, left ventricle thrombus, or previous ischemic stroke.
Binomial data were analyzed with the χ2 test, and continuous variables were analyzed with Student t-test or Mann–Whitney for skewed distributions. Median value and interquartile range (IQR) were used to describe these skewed distributions. Cumulative risks were estimated with Kaplan–Meier analysis. A Cox model was fitted for recurrent intracranial hemorrhage and separately for ischemic cerebral events, with a time-dependent variable indicating the time at which warfarin was resumed (if at all). The assumption in the Cox model was that from the time of resumption, warfarin exerts a constant proportional effect on the risk of recurrent intracranial hemorrhage, and exerts a constant proportional reduction of the risk of ischemic events. The software for calculations was SAS7STAT 9 (SAS Institute Inc.). The results of the Cox models yielded estimates of the risks per day (hazards) of intracranial hemorrhage or of thromboembolic event, with and without warfarin treatment, over time since the qualifying event. Based on these estimates, we computed the cumulative risks over treatment horizons between 3 and 6 years, and with warfarin resumption commencing between 1 and 50 weeks after the index intracranial hemorrhage.
We identified and reviewed retrospectively the records of all 3287 admissions for 2869 consecutive patients diagnosed with intracranial hemorrhage at the 3 centers. Two hundred thirty-four patients (8.2%) fulfilled the criteria for inclusion in the study. The baseline characteristics of the patients are described in Table 1. Few patients were identified with intracranial pathology as secondary cause of intracerebral hemorrhage: 4 patients with aneurysms, 4 patients with amyloid angiopathy, 2 patients with tumors, and 1 patient with a ruptured colloid cyst. The number of patients who were reviewed, included in the study, and had resumption of anticoagulation at the 3 sites of investigation is shown in Table 2.
We had access to follow-up information on the 234 patients regarding major clinical events for a median of 34 weeks (IQR 1–115). For the 177 patients who survived the first week, the median follow-up duration was 69 weeks (IQR 19–144). Overall, 113 patients (48%) died during the follow-up, corresponding to a median survival of 4.5 years. Fifty-seven patients (24%) died within the first week: 54 patients because of progression of bleeding and/or herniation (1 patient also had massive pulmonary embolism), 2 patients from ischemic stroke, and 1 patient from myocardial infarction. Fatal outcome within the first week occurred in 47 of all patients (36%) with intracerebral hemorrhage, 3 patients (14%) with subarachnoid hemorrhage, and 7 patients (8%) with subdural hematoma. During the entire follow-up, mortality among those with intracerebral hemorrhage was 59%, compared with 32% after subdural hematoma (Kaplan–Meier analysis, P<0.001). Recurrent intracranial hemorrhage was fatal in 4 of 18 patients (22%). In contrast, with the 21 patients (12%) who experienced a subsequent arterial thromboembolic event, none of those events were fatal.
Resumption of Warfarin
Fifty-nine patients (33%) resumed warfarin. The patients for whom warfarin was restarted were younger and had a longer follow-up than did the remainder (Table 1). Warfarin was restarted in 22 of the patients with atrial fibrillation (22%), 8 of the patients with venous thromboembolism (27%), 15 of the patients with mechanical aortic valve (79%), and 7 of the patients with mechanical mitral valve (77%). There was a highly significant difference between the proportions of restarted patients among those with mechanical valves compared with those of the remainder (P<0.001). Warfarin was resumed in 23 of all first-week survivors after intracerebral hemorrhage (28%), in 7 patients with subarachnoid hemorrhage (39%), and in 29 patients with subdural hematoma (38%). Resumption occurred after a median of 4.4 weeks (IQR 2.3–14) in patients with intracerebral hemorrhage, compared with 6.4 weeks (IQR 3.6–26) after subdural hematoma.
Finally, warfarin was restarted in 10 of those with previous stroke (31%) versus 49 of those without (34%). There were no statistically significant differences in any of those proportions.
Resumption of warfarin occurred after a median duration of 5.6 weeks (IQR 2.6–17) for all first-week survivors, after 9.2 weeks (IQR 5.6–34) for patients with atrial fibrillation, after 4.6 weeks (IQR 2.3–15) for patients with mechanical aortic valve, after 3.1 weeks (IQR 2.2–14) for patients with mechanical mitral valve, and after 2.3 weeks (IQR 1.7–19) for patients with venous thromboembolism.
Recurrent Intracranial Hemorrhage
Recurrent intracranial hemorrhage was observed in 18 patients (10%); 8 of those patients restarted warfarin and 10 of those patients did not, although 2 of the latter patients subsequently resumed warfarin treatment. Twelve patients with subdural hematoma (16%; 4 had resumed warfarin and 8 patients had not) and 6 patients with intracerebral hemorrhage (7%; 4 had resumed warfarin and 2 patients had not) developed recurrent intracranial hemorrhage, versus none of those with subarachnoid hemorrhage (P=0.1). Kaplan–Meier analysis showed a trend toward higher risk of recurrence in patients with subdural hematoma (16%) compared with those with intracerebral hemorrhage (8.4%); (P=0.07). Among the patients with intracerebral hemorrhage, it recurred in 4 patients with lobar localization and in 1 patient each with deep hemispheric or brain stem localization. The type of recurrence was generally the same as it was in the index bleeding, except for 1 patient with initial intracerebral hemorrhage in which it recurred as subdural, and another patient with the opposite sequence. The cumulative risk of recurrent intracranial hemorrhage without restarted anticoagulation or from the time point of resumption of anticoagulation is shown in Figure 1A and 1B, respectively.
Among the 21 patients who suffered a subsequent arterial thromboembolic event, 2 patients had systemic embolism (brachial artery and femoral artery), 1 patient had a transient ischemic attack, and 18 patients suffered an ischemic stroke. Strokes occurred in 12 patients (12%) with atrial fibrillation, in 4 patients (14%) with mechanical heart valves, in 1 patient (3%) with venous thromboembolism, and in 1 patient (6%) with another indication for warfarin. Kaplan–Meier analysis showed a significantly higher risk of arterial thromboembolism in patients with previous stroke (28%) compared with those without (8.6%); (P=0.004). The cumulative risk of ischemic events without restarted anticoagulation or from the time point of resumption of anticoagulation is shown in Figure 1A and 1B, respectively.
Eleven patients (5%) developed symptomatic venous thromboembolism following warfarin cessation, after a median of 2 weeks (IQR, 1.5–5) from the intracranial hemorrhage; in 4 of these patients, the initial indication for warfarin was venous thromboembolism.
The modeling of risk for recurrent intracranial hemorrhage versus ischemic stroke in patients with or without resumption of warfarin therapy is based on the population with cardiac indication for anticoagulation and/or with previous ischemic stroke and who had survived the first week without a recurrent event (n=132; Table 1).
As might be expected, fitting separate Cox models with a time-dependent variable to indicate the point at which warfarin was restarted revealed strong effects of anticoagulation resumption. Restarting warfarin increased the risk of recurrent intracranial hemorrhage by more than a factor of 5 (HR 5.57; 95% CI, 1.80–17.25; P=0.0029), whereas the risk of a thromboembolic event was reduced by almost 90% (HR 0.11; 95% CI, 0.14–0.87; P=0.036). These model-based estimates are apparent in the raw data displayed in the top panel of Table 3. There, time is divided into periods with reasonably constant risk, which differ somewhat among recurrent hemorrhage and ischemic stroke, as these have independent natural courses. Within each time period, we show the observed number of events and the number of days at risk, separately for days when patients are receiving versus not receiving warfarin. For example, in the first 35 days there were 7 recurrent intracranial hemorrhages in 3829 patient-days while not on warfarin, equaling a daily rate of 0.18%; this is compared with 2 events in 265 patient-days while on warfarin, equaling a daily rate of 0.75%. The ratio of these two rates is 4.13, which is the estimated hazard ratio associated with starting warfarin on or before day 35. The corresponding hazard ratio in the 36- to 93-day period is 4.46, and then becomes infinite in later periods because there are no subsequent intracranial bleeds in patients not on warfarin. The Cox model hazard ratio estimate of 5.57 is essentially a weighted average of these individual estimates from the separate time intervals. The data in Table 3 clearly suggest a pattern of reducing risk of recurrent intracranial hemorrhage with time since the index bleed, both in the presence and absence of warfarin. Despite the lack of observed events beyond 63 days in the absence of warfarin, the data are quite consistent with a constant hazard ratio over time; thus, whatever the time of warfarin resumption, it would be anticipated that retreatment would multiply the background risk of an intracranial bleed by a factor of 5.57.
The risk of a subsequent thromboembolic event is also much higher immediately following the index intracranial bleed and then declines with time (Table 3). Only 1 thromboembolic event was observed in a patient receiving warfarin, in keeping with its antithrombotic efficacy. The assumption of a constant proportional risk reduction of approximately 90% is hard to verify in this data set, but the results of warfarin trials, for example in atrial fibrillation, tend to show diverging cumulative risk curves consistent with an ongoing constant proportional effect.16
Resumption of warfarin at any given time point will increase the subsequent risk of recurrent intracranial hemorrhage and reduce the risk of a thromboembolic event. The optimal restart time must balance these 2 competing cumulative risks over the entire warfarin “treatment horizon.” Because the underlying need for prophylaxis will continue indefinitely and the median survival rate in our data was 4.5 years, we evaluated a range of treatment horizons from 3 to 6 years in our modeling. To determine the optimal restart time, we varied warfarin resumption between 1 and 50 weeks after the index intracranial bleed and calculated the “total” risk (ie, before plus after the selected time point of resumption) of recurrent intracranial hemorrhage, and of a thromboembolic event through to the end of treatment. The calculation of cumulative risks used the rates displayed in the bottom panel of Table 3. Given that it is implausible that the actual risk of intracranial bleed or thromboembolic event is 0, we have not used the observed daily risks directly, but instead we have blended the observed rates and Cox model hazard ratios as indicated in the footnote to Table 3.
The results, as shown in Figure 2, demonstrate how the total risk of intracranial hemorrhage and an ischemic event for the whole treatment period varies according to when warfarin is restarted. Based on this combined risk, the optimal period of resumption of warfarin seems to be between 10 and 30 weeks from the index intracranial hemorrhage over a survival- and treatment-horizon of 3 years. This total risk did not change when the treatment horizon was expanded to 4, 5, or 6 years (data not shown). Resumption of anticoagulation within the first month results in a much higher “total” risk, which is driven by the high risk of recurrent intracranial hemorrhage. Inclusion of patients with a noncardiac indication for warfarin or without previous stroke in this analysis did not change the outcome (data not shown).
We are reporting results from a large cohort of patients with warfarin-induced intracranial hemorrhage. Our analysis of patients with a high risk for arterial thromboembolic complications and the extended follow-up enabled us to estimate the optimal time interval for resumption of warfarin, taking into account the risk for recurrent intracranial hemorrhage as well as for ischemic stroke. In contradiction to previous studies, our data suggest that warfarin resumption should be delayed at least a month from the index event. With 100 patients resuming anticoagulation 2 weeks, 5 weeks, or 10 weeks after warfarin-associated hemorrhage, there would, according to our modeling, be during a 3-year treatment horizon 36, 24, or 19 recurrent intracranial hemorrhages versus 4, 5, or 7 ischemic events, respectively.
The ideal design would have been a randomization between early and late resumption of anticoagulation, but this poses substantial logistical challenges. We needed approximately 5 years of consecutive cases from 3 tertiary referral centers to collect sufficient data for our analysis. We chose to analyze only recurrent major bleeds that were intracranial because of the clinical burden and the high risk of chronic sequel and death. The only thromboembolic complication of similar severity is ischemic stroke, and we therefore excluded transitory ischemic attacks and systemic embolism from our analysis. Our hazards model was focused on patients with a cardiac indication for anticoagulation or with previous stroke because when untreated, these conditions confer a moderate-to-high risk of ischemic stroke. Clinicians are usually less hesitant about permanent discontinuation of anticoagulation if the initial indication was venous thromboembolism.
In a systematic review of 6 retrospective studies, which included 120 patients with mechanical heart valves and intracranial hemorrhage, the authors concluded that it was safe to resume warfarin within 2 weeks.14 All 6 studies were either small or had a very limited follow-up, and only 2 recurrent intracranial hemorrhages and 4 ischemic strokes were reported. These studies did not include patients with atrial fibrillation and additional risk factors for stroke, or patients with myocardial infarction and intramural thrombus; both groups are also of interest due to high risk of stroke. A recently published systematic review with 492 patients concluded that anticoagulation may be resumed after 72 hours.17 This review included a large number of single-case reports with a high risk for bias toward successful cases, and the authors considered all anticoagulants of any dose together. The hitherto largest cohort included 141 patients, of whom 35 patients resumed warfarin; however, the follow-up was only 30 days, which may explain why no recurrent intracranial hemorrhage was observed.10 The second-largest cohort had 52 patients, of whom 23 patients resumed warfarin, reported 3 recurrent intracranial hemorrhages and 3 ischemic strokes during a mean follow-up of 43 months.9 These small numbers of events do not allow statistical analysis of the optimal time of resumption of anticoagulation. Eckman et al addressed this question by using a Markov model, and concluded that patients with deep hemispheric location of the hemorrhage (which had a reported low risk of recurrence) would qualify for resumption of anticoagulation, provided that the risk of ischemic stroke was high.18 Conversely, patients with lobar hemorrhage, which appears to carry a higher risk of recurrence, should never be restarted on anticoagulation. This analysis focused on patients who had atrial fibrillation and an average risk of stroke of 4.5% per year. We had too-small numbers of each of the subtypes of intracerebral hemorrhage with recurrence to confirm those assumptions. Furthermore, the risk for ischemic stroke varies widely, and for a patient with a previous stroke, the risk of recurrence (28% in our material) may justify resumption of anticoagulation for a patient with lobar hemorrhage as well.
The main limitation of our study is the retrospective design with resumption of anticoagulation occurring at different time points. The choice of time point might have been biased by the fear of the clinician of recurrent bleeding or of ischemic stroke; this seems to have been the case for patients with mechanical valves who were restarted on warfarin more frequently than for other subsets. However, those with previous stroke were treated in a similar manner to those without, despite higher risk for new ischemic events. Among patients with subdural hematoma and a high risk for recurrent hemorrhage, the rate of resumption was as high as it was for other patients. Although warfarin was resumed more frequently in younger patients, we did not observe a difference in the risk for recurrent intracranial hemorrhage or ischemic events depending on age (data not shown). Another limitation of our study is the significantly shorter follow-up of patients who were not restarted on warfarin. However, with a median of 39 weeks of observation in this subset, the period with the highest event rate should have been covered.
We acknowledge that our conclusions are based on relatively few clinical events. However, our study includes more events than does any previous study or systematic review. We have shown that the initial risk for recurrent intracranial hemorrhage increases 5-fold when warfarin is resumed, and that the risk of recurrent bleeding with resumed warfarin is higher than is the risk of ischemic events without anticoagulation. Resumption of warfarin can therefore be delayed by more than 1 month after the index event. In fact, the optimal period for resumption of anticoagulation seems to occur between week 10 and week 30 after warfarin-associated intracranial hemorrhage, when the risk for recurrent intracranial hemorrhage and the risk for ischemic stroke are taken into account. Our findings contradict the recommendations made in previous studies and expert opinion. The larger number of patients included in this study, the longer follow-up, and the modeling we performed have allowed for a more precise description of the total risk for serious events in relation to the timing of resumed anticoagulation. The span of this interval of 20 weeks may be helpful to select when patients with a perceived high risk for bleeding versus high risk for stroke should be restarted.
Sources of Funding
No external funding was obtained for this study. Y.-K.K. received salary support from Soonchunhyang University while performing the study at McMaster University. A.M. received PhD-student support from the Research and Development center, Eskilstuna, Sweden.
S.S. has received consulting fees from Astra-Zeneca, Bayer HealthCare, Boehringer Ingelheim, GlaxoSmith-Kline, Octapharma, and Sanofi-Aventis; lecture fees from LEO Pharma, Sanofi-Aventis, and Boehringer Ingelheim; and grant support from Bayer HealthCare. M.H. has received lecture fees from Baxter Medical, CSL Behring, Leo-Pharma, and Nycomed, and travel support from Leo Pharma and Swedish Orphan. A.M., R.R., and Y.-K.K. have no potential conflicts of interest to declare.
S.S. planned and supervised the study. M.H. contributed to the implementation of the study at the Swedish sites. A.M. and Y.-K.K. performed the chart review. R.S.R. and S.S. performed the statistical analyses. A.M. and S.S. wrote the manuscript, and all the other authors critically reviewed the manuscript.
- Received June 9, 2010.
- Accepted September 13, 2010.
Babikian VL, Kase CS, Pessin MS, Caplan LR, Gorelick PB. Resumption of anticoagulation after intracranial bleeding in patients with prosthetic heart valves. Stroke. 1988; 19: 407–408.
Broderick J, Connolly S, Feldmann E, Hanley D, Kase C, Krieger D, Mayberg M, Morgenstern L, Ogilvy CS, Vespa P, Zuccarello M. Guidelines for the management of spontaneous intracerebral hemorrhage in adults: 2007 update: A guideline from the american heart association/american stroke association stroke council, high blood pressure research council, and the quality of care and outcomes in research interdisciplinary working group. Circ. 2007; 116: e391–e413.
Aguilar MI, Hart RG, Kase CS, Freeman WD, Hoeben BJ, Garcia RC, Ansell JE, Mayer SA, Norrving B, Rosand J, Steiner T, Wijdicks EF, Yamaguchi T, Yasaka M. Treatment of warfarin-associated intracerebral hemorrhage: Literature review and expert opinion. Mayo Clin Proc. 2007; 82: 82–92.
Eckman MH, Rosand J, Knudsen KA, Singer DE, Greenberg SM. Can patients be anticoagulated after intracerebral hemorrhage? A decision analysis. Stroke. 2003; 34: 1710–1716.