Warfarin-related intracerebral haemorrhage: better outcomes when reversal includes prothrombin complex concentrates


  • Funding: None.
  • Conflict of interest: None.


H. Carl Hanger, Canterbury District Health Board, PO Box 800, Christchurch, New Zealand, 8022 Email: carl.hanger@cdhb.govt.nz



Warfarin-related intracerebral haemorrhage (WRICH) has high mortality. Haematoma expansion is prolonged in WRICH and independently predicts worse outcomes. Guidelines recommend prompt reversal of the warfarin coagulopathy, but evidence of benefit is lacking.


To determine whether the introduction of a WRICH reversal protocol (late 2008), which includes prothrombin complex concentrates (PCC), improves outcomes


All patients presenting with WRICH between January 2004 and July 2010 were included. Retrospective case note and radiology review was performed, collecting data on intracerebral haemorrhage (ICH) severity, degree and timeliness of reversal, and patient outcomes. Cox's proportional hazards analysis was used to compare outcomes associated with and without PCC after controlling for ICH severity.


Eighty-eight patients were included (27 treated palliatively). Mean international normalised ratio was 2.9. Vitamin K, PCC and fresh frozen plasma were given alone or in combination to 68, 23 and 44 patients, and mean time from computed tomography scanning to administration was 2.2, 3.3 and 3.1 h respectively. Four patients received PCC pre-protocol (none before 2007), two during development and seventeen patients post-protocol. Those who received PCC had improved survival (P < 0.001). After controlling for ICH score, hazard ratio for death was 0.27 (P < 0.01) for use of PCC. Survival tended to be greater with earlier administration of PCC (P = 0.053). Despite improved survival, discharge domicile and function were not significantly worse.


PCC reversal was associated with improved survival without worsened disability. Delays in administration may have reduced the potential benefits.


One of the most feared complications of warfarin use is warfarin-related intracerebral haemorrhage (WRICH).[1] This is due both to its high mortality (44–68% at 30 days)[1-4] and morbidity. Evidence for effective options to ameliorate these poor outcomes is lacking, and there is only modest consensus on the best acute management strategies.[1, 5]

Estimates of WRICH frequency vary between 6% and 24% of all intracerebral haemorrhages (ICH).[3, 5-7] Higher international normalised ratios (INR) have a higher risk of WRICH, but most WRICH occur at an INR within the target therapeutic range.[3, 4, 7] The volume of ICH is a major predictor of outcomes.[8] Haematoma expansion (HE) can continue in the first 2–3 h after primary ICH but may be much longer in WRICH. Serial computed tomography (CT) scanning has shown that 30–50% of primary ICH expand between 6 h and 12 h, with an associated clinical deterioration,[9] whereas WRICH expansion may continue for up to 72 h.[10] HE may worsen the already high mortality after WRICH.[9, 11]

Therefore, treatments to limit final haematoma volume are potentially beneficial, and guidelines recommend prompt reversal of coagulopathy to achieve this.[5, 12] All WRICH, even those with relatively small volume and better prognosis bleeds, should have urgent reversal of the coagulopathy to avoid clinical deterioration from HE.[7] There are haematology guidelines for the reversal of warfarin-related coagulopathy,[13] but a consistent approach is lacking in the stroke literature.[5] Treatments with vitamin K and other products (fresh frozen plasma (FFP), prothrombin complex concentrate (PCC) and factor VIIa) have not been subjected to a randomised controlled trial (RCT) in this patient group, and furthermore there are concerns regarding thrombotic complications with rapid reversal of anticoagulation[7, 14]

A previous local audit (for the period 1996–2006) showed inadequate reversal of the warfarin-associated coagulopathy. Vitamin K and FFP were used variably, and no patient received PCC. There was also no apparent consideration given either to urgency of reversal or to targeting patients who were most likely to survive. To address these deficits, we developed a protocol for urgent reversal of the coagulopathy (see Appendix I)[13] This protocol was a major change in practice in that it included PCC (a three-factor PCC was used). It was widely disseminated within our institution, with education of medical and nursing staff, and the protocol was incorporated into existing guidelines[15, 16]

Our hypothesis was that improved reversal strategies would reduce HE and final ICH volume, and therefore improve both survival and functional outcomes. The study aimed to examine whether the introduction of a PCC-based reversal protocol improved patient mortality, function and discharge domicile outcomes.


We undertook a retrospective audit of all patients presenting with a WRICH between January 2004 and July 2010. The acute stroke unit (ASU) was opened in 2004, and our stroke register also began then. Apart from introduction of WRICH protocol in 2008, there were no other changes in the delivery of service for ICH patients during this study (Fig. 1). This study period included 56 months pre-protocol, 5 months during protocol development and dissemination, and 18 months after protocol implementation.

Figure 1.

Timeline of study. ASU, acute stroke unit; PCC, prothrombin complex concentrates; SRU, stroke rehabilitation unit; WRICH, warfarin-related intracerebral haemorrhage.

In our region (total catchment population 460 000), there is one acute hospital and two rehabilitation hospitals to which all patients with ICH are admitted. Discharge coding data from these hospitals were searched for patients with an International Classification of Diseases (ICD)-10 discharge coding of ICH (I61 or I62.9). The admission stroke register in the ASU was checked for additional cases. The electronic records and laboratory results of all these patients were reviewed to identify any patient with an elevated INR and taking warfarin at the time of ICH. Included patients had (i) ICH confirmed by imaging or post-mortem, (ii) taking warfarin and (iii) had an INR >1.2. The clinical notes and radiology of each WRICH patient were reviewed in addition to the electronic screening above. Data extracted included basic demographic data; ICH location (lobar, deep, cerebellum or brainstem), volume (using ABC/2 estimation[17]) and intraventricular extension; indications for anticoagulation; reversal agents used, including time of reversal; and patient outcomes of mortality, discharge function (Functional Independence Measure (FIM)) and discharge domicile.

Exclusion criteria were any of the following: INR ≤1.2 at presentation, ICH secondary to trauma, thrombolysis or heparin-related ICH, thrombocytopenia, haemorrhagic transformation of an infarct, subdural and subarachnoid bleeding, and asymptomatic microbleeds. While intraventricular haemorrhage (IVH) alone without parenchymal bleeding is important, has a high mortality and should have urgent treatment,[18] we could not calculate ICH volumes (and hence severity). Therefore, these IVH patients were excluded from this outcomes analysis. WRICH patients presenting late (>24 h of onset) were also excluded, as reversal is unlikely to be effective.

Time of onset of ICH was not reliably recorded, but all patients had the time of presentation to emergency department (ED). The time of CT scan (or magnetic resonance imaging) was taken as time recorded on the scan images. For each reversal agent given, the administration time, as recorded by the administering nurse, was used, as opposed to when it was prescribed.

Some patients were deemed palliative from the outset by the treating clinicians and unlikely to benefit from any reversal. Therefore, where the clinical notes clearly stated the palliative intent at the first review, these patients were excluded from any further analysis for outcomes.

The primary outcome of this study was survival after the ICH, with secondary outcomes being function and domicile on discharge among survivors.

Outcomes after ICH are predominantly determined by age, location and size of bleed, and Glasgow Coma Scale (GCS) at presentation. These are all included in the ICH score, which has been shown to predict mortality.[19] ICH scores are not used in routine clinical practice in our hospital, so were calculated for each patient, retrospectively. Comparisons between cohorts were adjusted for confounding factors, such as severity (using ICH score) and palliation.

Survival rates according to the type of reversal agent were compared using Cox's proportional hazard analyses after controlling for the confounding variables. Kaplan–Meier survival curves were plotted according to the nature of reversal agents administered. Differences in continuous variables, according to the type of reversal agent given, were compared using univariate analysis of variance after controlling for ICH severity and excluding those deemed to be palliative.

The local ethics committee considered this study as audit and did not require formal ethics committee review (URB/10/EXP/016).


During the study period, there were 827 patients with a discharge coding of ICH, of whom 104 (12.6%) were taking warfarin. Of these people on warfarin, 88 patients were included in the study. The remaining 16 were excluded, as shown in Figure 2. All patients except one person (an overseas visitor who returned home) were followed up until death or survival as of 24 July 2011. Mean follow up was 524 days (range 0–2700).

Figure 2.

Diagram of patients included in study. CT, computed tomography; ICH, intracerebral haemorrhage; INR, international normalised ratio; WRICH, warfarin-related intracerebral haemorrhage.

The characteristics of the cohort are shown in Table 1. Approximately half the cohort was taking concurrent antiplatelet therapies, predominantly aspirin (42/88), with a smaller number on dipyridamole (four) or clopidogrel (one). Thirteen were also taking other medications that can affect bleeding (selective serotonin re-uptake inhibitor for eleven patients, and non-steroidal anti-inflammatory drugs for two patients). Atrial fibrillation was the main indication for anticoagulation. Prosthetic heart valves were the indication for nine (10%) patients – of these, four were treated palliatively, and of the remaining five, three received PCC.

Table 1. Descriptive data for whole warfarin-related ICH cohort (n = 88), and for those who did and did not receive PCC (after exclusion of palliated patients)
 Total cohortAfter exclusion of palliated patients
PCC cohort (n = 22)No PCC cohort (n = 39)
  1. †Not measured in two patients (died before measured, or not appropriate), but had proven ICH and were taking warfarin. ‡Some patients had more than one indication for anticoagulation. ICH, intracerebral haemorrhage; INR, international normalised ratio; IQ, interquartile; PCC, prothrombin complex concentrate.
Age (years)   
Mean (median)77 (79)76.2 (76)77 (79)
(IQ range)(73–82)(71–81)(76–82)
M : F (%)53:4759:4154:46
INR on presentation   
Mean (median)2.9 (2.5)2.9 (2.6)2.7 (2.5)
(IQ range)(2.2–3.2)(2.2–3.2)(2.2–3.2)
Location of ICH   
Lobar3611 (50%)16 (41%)
Deep389 (41%)16 (41%)
Cerebellum132 (9%)7 (18%)
Intraventricular extension – n (%) of total cohort45 (51%)9 (41%)15 (38%)
Volume of ICH (mL)   
Mean (median)30 (11)13.9 (6.7)17.6 (9.2)
(IQ range)4.6–401.6–12.44.2–17.6
Volume of ICH – n (%) of total cohort   
Small (<5 mL)22 (25%)8 (36%)11 (28%)
Medium (5–30 mL)38 (43%)11 (50%)22 (56%)
Large (>30 mL)28 (32%)3 (14%)6 (15%)
Indication for anticoagulation – N (%) of total cohort   
Atrial fibrillation70 (77%)20 (91%)28 (72%)
Prosthetic valve9 (10%)3 (14%)2 (5%)
Venous thromboembolism7 (8%)1 (5%)6 (15%)
Left ventricular thrombus1 (1%)02 (5%)
Stroke1 (1%)02 (5%)
Transient ischaemic attack2 (2%)02 (5%)
Other or unclear from notes1 (1%)01 (3%)
Palliation from outset – n (%)27 (31%)N/AN/A
No reversal agents given at all – n (%)20 (23%)N/A3 (7.7%)
ICH score at presentation – N (%)   
015 (17%)5 (22%)10 (26%)
123 (26%)9 (41%)11 (28%)
222 (25%)7 (32%)11 (28%)
315 (17%)1 (5%)6 (15%)
47 (8%)01 (3%)
56 (7%)00

The range of GCS on admission was 3–15 but was skewed to higher values (median 14), with 36 having normal GCS of 15. A further 15 patients scored 14.

The average time (h) from ED presentation to CT scanning was 3.1 h (median 1.9; interquartile (IQ) range 1.1–3.4), and there were four outliers beyond 10-h post-presentation (11, 11, 22, 28 h).

Vitamin K was given according to protocol in 61 (90%) of the 68 patients receiving this agent. The remaining patients were given less than 5 mg intravenously (five patients), or oral vitamin K (one), and in one the dose was not stated.

FFP was given in 44 patients. In the 15 of these patients who also received PCC, the dose of FFP was 1–2 units, as specified by the protocol (when given with PCC). However, in the remaining patients (FFP and vitamin K alone), doses of FFP were also small (median 2 units, range 1–6).

PCC was given to 23 patients all in appropriate weight-adjusted per-protocol doses and all with vitamin K. Eight of these patients did not receive concurrent FFP. The average dose of PCC was 2600 IU (range 500–4500) based on estimated weights of patient. Repeat INR were performed in 21/23 patients receiving PCC, with three patients still having a raised INR when next measured (median of 7 h). One of these three received only 500 IU PCC, whereas another had additional PCC to achieve complete reversal. Twenty patients had normal INR within 24 h (the remaining one had normal when first checked at 27 h).

Reversal in those given FFP without PCC was incomplete in 13 patients who had a persisting high INR on repeat testing (median of 9 h), and 9/21 (43%) had a raised INR (>1.2) beyond 24 h.

Mean times from CT scan to when each of the reversal agent was given are shown in Table 2. Negative values indicate reversal given before CT scan was performed. The first reversal agent given was vitamin K for the majority (91%) of patients.

Table 2. Time between computed tomography (CT) scan and administration of each reversal agent (h)
 CT scan to vitamin KCT scan to fresh frozen plasma (FFP)CT scan to prothrombin complex concentrates (PCC)§CT scan to first reversal givenEmergency department admission to first reversal given
  1. †Seven patients given vitamin K before scanned. Eleven patients were given vitamin K, even though they were receiving palliative care. ‡Timing of FFP administration in two patients not recorded. Eight patients who were palliated received FFP. §PCC only given since 2007, and protocol was fully implemented 1 February 2009. One patient deemed for palliative cares received PCC (as well as FFP and vitamin K). ¶Time of when first reversal agent was given, regardless of which one (was vitamin K in 90% of patients).
Number who were given each agent6844236866
IQ range0.6–2.61.6–3.71.6–4.20.6–2.21.9–6.6
Range−18 to 23.7−0.4 to 170.5 to 13.4−18 to 23.70.4 to 24

Few patients received mannitol or dexamethasone (one and five patients respectively). No patients in this cohort received either craniotomy or a drainage procedure. Additional antihypertensive therapy was used in 16 patients according to our institutions' guidelines at the time.[15]

As expected, ICH score was highly predictive of survival (Fig. 3), highlighting the importance of controlling for ICH score when examining outcomes.

Figure 3.

Kaplan–Meier survival curve grouped by ICH score[19] at presentation (P < 0.001). ICH score: (image) 0, (image) 1, (image) 2, (image) 3, (image) 4, (image) 5.

In a Kaplan–Meier survival analysis on unadjusted data, non-palliated patients were more likely to survive if they were given PCC (P = 0.007) but not vitamin K or FFP (Fig. 4). Cox regression analysis was used to control for ICH severity, and better survival with PCC remained statistically significant. The hazard ratio (HR) for death with each increase in ICH score was 2.33 (95% confidence interval (CI) 1.59–3.41) (P < 0.001), and the use of PCC was protective with an HR = 0.27 (95% CI 0.10–0.72) (P < 0.01). Most deaths (82%) were early (within 7 days), reflecting neurological death from ICH itself rather than comorbidities.[3, 9]

Figure 4.

Kaplan–Meier survival curve for patients given each reversal agent versus those not given that reversal agent (vitamin K (not significant), FFP (not significant) and PCC (P = 0.007) after exclusion of patients who were palliated from outset (raw data, unadjusted for ICH severity). FFP, fresh frozen plasma; ICH, intracerebral haemorrhage; PCC, prothrombin complex concentrate. Vitamin K: (image) no vitamin K, (image) vitamin K given, (image) no vitamin K-censored, (image) vitamin K given. FFP: (image) no FFP given, (image) given FFP, (image) no FFP given-censored, (image) given FFP-censored. PCC: (image) no PCC, (image) given PCC, (image) no PCC-censored, (image) given PCC-censored.

Survival of WRICH was not at the expense of increased disability. There were no differences in either discharge FIM or domicile for any of the reversal agents. However, surviving patients who received PCC had greater FIM gains than those who did not receive PCC (28.3 vs 12.3, P = 0.049).

Earlier treatment with PCC was associated with a trend to better survival after controlling for ICH score and compared with no treatment (P = 0.053) (Fig. 5).

Figure 5.

Survival curve grouped by timing of administration of prothrombin complex concentrate (PCC) after controlling for intracerebral haemorrhage severity and excluding those deemed to be palliative (P = 0.053). SCantoPCCcat: (image) no PCC, (image) <2 h, (image) 2–4 h, (image) >4 h.


This study has shown that reversal of the warfarin coagulopathy with PCC is associated with improved survival, after adjustment for confounding variables, such as ICH severity. This improved survival was not at the cost of survival of very disabled patients.[20, 21] Not only was disability in those surviving found to be no worse, but there were greater functional gains with PCC. There was also a trend suggesting that earlier treatment with PCC is better than late, but late still seemed better than none at all.

‘Time is brain’, yet as has been shown in other studies,[7, 21-25] we found significant delays in administering any of the reversal agents. This study did not examine reasons for delays, but all possible reasons, including procedural and attitudinal, need further exploration. Reversal of WRICH needs to be regarded as a medical emergency, similar to thrombolysis for ischaemic stroke.[26]

Strengths of this study are the large sample size (n = 88) for this uncommon condition. This number represents all eligible WRICH patients presenting in our region over a 6.5-year period, thus avoiding any sampling bias. The mean age of our cohort (77 years) is higher than our non-anticoagulant ICH population (72 years),[3, 27] reflecting the ‘real-world’ older patients taking warfarin for atrial fibrillation (the main indication for anticoagulation). There was complete follow up of all cases beyond hospital discharge, and the results were adjusted for ICH severity and palliative treatment. Both of these confounding factors are likely to bias whether PCC is given or not, as well as affect outcomes. Further selection bias was avoided by choosing controls from a time period when PCC was not used for WRICH at this hospital (see Fig. 1) but when all other selection criteria were the same.

We acknowledge the weaknesses of a non-randomised trial in a single centre (albeit large catchment area of 460 000). There was a high use of concurrent antiplatelet medications (mostly aspirin),[28] although there is debate whether this alters outcomes in WRICH.[3, 12] There was no change in the proportion of WRICH patients taking concurrent antiplatelet agents throughout the study period. The reviewers could not be blinded (to PCC use or not) when assessing the notes, but use of objective outcomes limits this potential bias. We were unable to ascertain the exact time of onset of WRICH, and instead used times from ED presentation and confirmed diagnosis (radiological scan) to measure our responsiveness. The three treatment groups are overlapping, and most patients received more than one treatment, as the guideline indicates they should.[13] This makes determining whether one component of the treatment is effective more difficult. However, as vitamin K and FFP were used throughout the study period, and the sole change was the introduction of PCC, we believe that the improvements seen reflect the added benefit of PCC over and above these existing treatments. While our CT scan to reversal times were longer than we desired, they are similar to other studies.[21-25] Had these delays been shortened, the effect of PCC may have been greater. High mortality rates with WRICH and some incomplete data collection contributed to smaller numbers of patients available for functional outcomes analysis. As a result, our study may have been underpowered for these latter functional outcomes.

To date, the best strategy for acute reversal remains unclear with differing combinations of vitamin K, FFP or PCC proposed.[1, 5, 7, 11, 12] There are no RCT data to guide us, although one is currently recruiting.[29] Reversal with PCC appears the most promising, with rapid reversal, and hints of improved outcomes.[11, 22, 25, 30] Fredriksson et al.[30] in 1992 showed a trend to lower modified Rankin scale outcomes in those receiving PCC, but this was confounded by selection bias as to who was given PCC versus FFP. In contrast, a larger study by Sjoblom et al. could not show any improvement in outcomes with the different agents given.[11] Similarly, Boulis et al.,[31] Huttner et al.[32] and Cartmill et al.[33] showed PCC produced a rapid reversal in coagulopathy but without improvements in clinical outcomes. Recently, Japanese investigators[22] reported better outcomes in those treated with PCC compared with a control group. This effect persisted after controlling for severity and other confounding variables, but as the authors themselves acknowledge, the groups were likely to be affected by selection bias. Selection bias, adjustment for severity and exclusion of those unlikely to survive are all important. We excluded those patients who were treated palliatively, adjusted for severity and tried to avoid bias by using historical controls, during a period when PCC was not used at all. Therefore, the improved survival found in our study appears real.

There have been concerns that rapid reversal may cause thrombotic complications in a population already at high risk of these events.[14] This risk appears to be low,[25, 34] possibly between 0% and 5%.[35] The risks of thrombosis particularly associated with the original indication for anticoagulation, such as atrial fibrillation or prosthetic heart valves, need to be balanced against the extremely high mortality associated with WRICH of up to 68%.[3-5, 12] While our study did not specifically address this issue, the above studies and our survival benefit suggest that any thrombotic risks are outweighed by the gains in rapid reversal.

If the effect of acute reversal is by reducing HE[5] than earlier, more aggressive reversal should be associated with better outcomes.[12] While we did not perform repeat CT scans to detect any expansion, our data support this, as when PCC was given earlier there was better survival (Fig. 5).

The ‘time is brain’ message has been widely accepted for ischaemic strokes with the urgency to restore blood flow with thrombolysis preferably within minutes but up to 4.5 h.[26] We were disappointed to see the treatment delays in our study, although other studies report similar delays.[21-25] Earlier treatment appears better. Vitamin K is readily available, yet was not given immediately following CT confirmation. Both FFP and PCC are blood products, which in our hospital are only available through the blood transfusion service. Accessing FFP requires compatibility testing and thawing before transfusion. PCC approval required the INR result and approval by the on-call haematologist. These institutional barriers run counter to the urgency required. We have subsequently introduced a revised ‘WRICH urgent reversal protocol’ for use in ED, which facilitates urgent reversal commencing as soon as possible after CT scan, and without need for pre-treatment INR result to be known. Greater emphasis on promptly assessing the adequacy of reversal is also needed.

PCC and vitamin K, without concurrent FFP, can rapidly reverse the coagulopathy,[5, 25, 31-33] and some authors suggest that the PCC/vitamin K combination may be sufficient to reverse life-threatening bleeding.[25, 36] However, the three-factor PCC used in our institution (Prothrombinex-HT[37]) may not contain sufficient factor VII, and given the dire consequences of inadequate reversal, this is covered by using small volumes of FFP.[7, 13]


The high mortality associated with WRICH is not immutable. Treatment with PCC is practical, reverses the coagulopathy and may make a positive difference for patients. We just need to reverse faster.

Appendix: Appendix I

Reversal guidelines for warfarin-related intracerebral haemorrhage[13]

All patients with a warfarin-related intracerebral haemorrhage and an elevated international normalised ratio (INR) (>1.2) should have rapid reversal of the coagulopathy. Do ALL of the following:

  • Cease warfarin.
  • Give 5–10 mg vitamin K intravenously.
  • Prothrombin complex concentrate (Prothrombinex-HT is available in New Zealand) 25–50 IU/kg intravenously.
  • Fresh frozen plasma (150–300 mL) intravenously.


  • Vitamin K takes 6–24 h to be effective.
  • Fresh frozen plasma contains all the relevant clotting factors but requires large volumes (2L or more) to adequately replace clotting factors.
  • Prothrombin complex concentrate acts rapidly (within 15 min) and is accessed through contacting New Zealand Blood Service doctor on-call.
  • Prothrombinex-HT used in New Zealand may not contain sufficient factor VII, hence concurrent use of a small amount of fresh frozen plasma as well.
  • Monitoring:
  • INR alone is not useful for monitoring the effectiveness of clotting factor replacement. It is only useful for monitoring warfarin use in steady state situations.
  • Monitoring should be done immediately after treatment using a coagulation screen (INR, activated partial thromboplastin time, thrombin time and fibrinogen). If still abnormal, more coagulation factors should be given immediately.
  • If normal recheck in 4–6 h (reflecting shortest half-life of factor VII and vitamin K onset of action).
  • If normal again, then recheck at 24 h, or sooner if patient is clinically unstable.
  • The risk of thrombotic events during this short-term reversal appears very low, even in patients with prosthetic heart valves.