Reversal strategies for vitamin K antagonists in acute intracerebral hemorrhage

Objective There is little evidence to guide treatment strategies for intracerebral hemorrhage on vitamin K antagonists (VKA‐ICH). Treatments utilized in clinical practice include fresh frozen plasma (FFP) and prothrombin complex concentrate (PCC). Our aim was to compare case fatality with different reversal strategies. Methods We pooled individual ICH patient data from 16 stroke registries in 9 countries (n = 10 282), of whom 1,797 (17%) were on VKA. After excluding 250 patients with international normalized ratio < 1.3 and/or missing data required for analysis, we compared all‐cause 30‐day case fatality using Cox regression. Results We included 1,547 patients treated with FFP (n = 377, 24%), PCC (n = 585, 38%), both (n = 131, 9%), or neither (n = 454, 29%). The crude case fatality and adjusted hazard ratio (HR) were highest with no reversal (61.7%, HR = 2.540, 95% confidence interval [CI] = 1.784–3.616, p < 0.001), followed by FFP alone (45.6%, HR = 1.344, 95% CI = 0.934–1.934, p = 0.112), then PCC alone (37.3%, HR = 1.445, 95% CI = 1.014–2.058, p = 0.041), compared to reversal with both FFP and PCC (27.8%, reference). Outcomes with PCC versus FFP were similar (HR = 1.075, 95% CI = 0.874–1.323, p = 0.492); 4‐factor PCC (n = 441) was associated with higher case fatality compared to 3‐factor PCC (n = 144, HR = 1.441, 95% CI = 1.041–1.995, p = 0.027). Interpretation The combination of FFP and PCC might be associated with the lowest case fatality in reversal of VKA‐ICH, and FFP may be equivalent to PCC. Randomized controlled trials with functional outcomes are needed to establish the most effective treatment. Ann Neurol 2015;78:54–62

HR 5 2.540, 95% confidence interval [CI] 5 1.784-3.616, p < 0.001), followed by FFP alone (45.6%, HR 5 1.344, 95% CI 5 0.934-1.934, p 5 0.112), then PCC alone (37.3%, HR 5 1.445, 95% CI 5 1.014-2.058, p 5 0.041), compared to reversal with both FFP and PCC (27.8%, reference). Outcomes with PCC versus FFP were similar (HR 5 1.075, 95% CI 5 0.874-1.323, p 5 0.492); 4-factor PCC (n 5 441) was associated with higher case fatality compared to 3-factor PCC (n 5 144, HR 5 1.441, 95% CI 5 1.041-1.995, p 5 0.027). Interpretation: The combination of FFP and PCC might be associated with the lowest case fatality in reversal of VKA-ICH, and FFP may be equivalent to PCC. Randomized controlled trials with functional outcomes are needed to establish the most effective treatment. ANN NEUROL 2015;78:54-62 A round 20% of all intracerebral hemorrhage (ICH) patients are on vitamin K antagonists (VKA), with the incidence of VKA-ICH increasing as the population grows older. 1 The 3-month case fatality of the condition is high at 50%. [2][3][4] One-third of ICH patients develop significant early hematoma expansion, 5 and this risk is doubled in VKA-ICH. 6 Vitamin K takes several hours to initiate sufficient endogenous clotting factor production, so urgent treatments to rapidly replace vitamin Kdependent clotting factors (II, VII, IX, X) are widely used, with the aim of limiting further bleeding. Prothrombin complex concentrate (PCC), fresh frozen plasma (FFP), recombinant factor VIIa, or combinations of these are used, with practice varying between different centers and countries. 7 Although there is a clear rationale for the use of these agents, none has been conclusively shown to improve outcome after VKA-ICH.
Evidence from patients with major VKA-associated bleeding (predominantly gastrointestinal hemorrhage) demonstrates that relative to FFP, PCC normalizes the international normalized ratio (INR) more quickly, reduces the need for red blood cell transfusion, and does not lead to an increase in adverse events. 8,9 Although PCC is more expensive, it has practical advantages including more rapid administration, smaller infusion volume, and no need for ABO blood type match. This has led to PCC being recommended as a reasonable alternative to FFP in the USA 10 and the first-line treatment in the United Kingdom. 11 The 2014 European consensus-based ICH guidelines do not provide a recommendation, citing lack of evidence. 12 Furthermore, different preparations of PCC have different concentrations of the vitamin K-dependent clotting factors, classified as 3factor or 4-factor depending on the concentration of factor VII (FVII). Three-factor PCCs are widely used in some countries, but may be less effective in correcting the INR than 4-factor PCC. 13 Although national and international guidelines recommend clotting factor replacement agents for the treatment of VKA-ICH, there is currently no definite evidence of benefit and no international consensus. Our aim was to utilize the existing international variation in practice to test for an association between the choice of VKA reversal strategy and survival, adjusted for key prognostic factors, in a large population of patients with VKA-ICH pooled from 16 registries in Europe, North and South America, and Australia.

Patients
We performed a retrospective pooled analysis of 16 stroke registries from Argentina, Australia, Finland, France, Germany, Italy, the Netherlands, the United Kingdom, and the USA. Patient registration methods and registration periods varied. Three registries were population-based, 1 from an international observational study, and 12 from single centers representing both large tertiary teaching hospitals and smaller regional hospitals. Registration of cases was prospective in 11 registries and retrospective in 5. Patient consent was required in 3 registries, there was opt-out in 3 registries, and the remaining 10 registries were approved as quality registries with consecutive registration of all cases. Patient registration years ranged from 1993 to 2014, with 90% of patients from 2004 to 2013. Registry methods are summarized in Supplementary Table I, with further details previously published. [2][3][4][14][15][16][17][18][19][20] According to a prespecified protocol, we included patients aged 18 years taking any VKA at the time of their ICH. We excluded patients with ICH secondary to trauma or tumor, primary subarachnoid hemorrhage, or hemorrhagic transformation of ischemic stroke, and those with baseline INR < 1.3. We also excluded those with missing data on reversal therapy received or variables used for adjustment: age, gender, INR, Glasgow Coma Scale (GCS), and imaging parameters (infratentorial location, intraventricular extension, baseline ICH volume). We prespecified exclusion of patients treated >24 hours from ICH onset, but later added these cases following a request from a manuscript reviewer.

Procedures
The registries provided data for the pooled analysis using a standardized form. Baseline volume was defined as the volume of intraparenchymal hemorrhage on the first scan, excluding any intraventricular hemorrhage (IVH). ICH volume was estimated with the ABC/2 method 21 in 12 and with planimetric methods in 4 registries. IVH volume was estimated with planimetric methods (2 registries), Hallevi score 22 (3 registries), or an estimate of the Hallevi score based on the Graeb score multiplied by 2 (6 registries). Immediate palliation was defined as the decision to withhold active treatment and to provide palliative care only immediately after ICH diagnosis. Therefore, patients who received either FFP or PCC were, by definition, not immediately palliated. Patients were classified into treatment arms of "FFP alone" (FFP but no PCC), "PCC alone" (PCC but no FFP), "combination" (both FFP and PCC), and "no reversal" (neither FFP nor PCC). PCC preparations were further classified as 4-factor PCC or 3-factor PCC according to the presence or absence of FVII. Patients receiving both 3factor PCC and FVIIa were considered to have received 4factor PCC. We also recorded whether patients were given vitamin K. Our primary outcome measure was all-cause case fatality by 30 days or end of follow-up if earlier.

Statistical Analysis
Statistical analyses were performed according to a prespecified protocol. For all descriptive analyses, we compared patient characteristics by treatment arm using the Kruskal-Wallis and chisquare tests, as appropriate. For the outcome of case fatality by 30 days, Kaplan-Meier estimates and the log-rank test were used for univariate analysis and Cox proportional hazard models for multivariate analysis, with the multivariate comparison predefined as the primary outcome measure. We adjusted for prespecified baseline characteristics known to be associated with case fatality in ICH: age, log-transformed ICH volume, INR, and GCS as continuous variables, and sex, infratentorial location, and intraventricular extension as binary variables. 10 We confirmed the assumption of proportional hazards by visual examination of the log (minus log) curves. Hazard ratios (HRs) with 95% confidence intervals (CIs) are reported, together with a survival plot at covariate means.
To examine for potential heterogeneity in the association between treatment and all-cause case fatality, we estimated this in several subgroups with prespecified cutoffs using the same Cox model, including a treatment by subgroup interaction term. As sensitivity analysis, we performed a prespecified propensity score-matched analysis. The propensity score for receiving the combination treatment was estimated using the same variables as in the primary analysis. We then matched patients in all 4 treatments arms using the nearest neighbor method with caliper set at 0.2 standard deviations of the logit of the propensity score. The primary analysis was duplicated in the propensity score-matched population.
Our study was powered for the primary outcome only. Assuming 50% case fatality by 30 days in the no reversal arm, and 25% HR reduction with any of the active treatments, we would have 80% power to detect this difference with 1,087 patients. The 2-sided threshold for statistical significance was set at p 5 0.05, with no correction for multiple comparisons justified by the hypothesis-generating exploratory nature of this analysis.

Results
The pooled registries contained 10,282 ICH patients over the study period, and after exclusions 1,547 were included in the analysis (Fig 1, Supplementary Table II).
Finally, we performed sensitivity analyses of the primary outcome in a propensity score-matched population of 131 patients from each treatment arm well balanced on baseline characteristics (Supplementary Table III

Discussion
Our international, multicenter, observational study of 1,547 VKA-ICH patients demonstrates considerable variation in the practice of anticoagulation reversal after acute ICH. Most centers used PCC alone and some centers used solely FFP, whereas Australian centers  5 1,176), published after the submission of our study, suggested that rapid reversal of INR is associated with less ICH expansion and lower in-hospital mortality, but was not powered to compare reversal therapies due to a uniform national practice of using PCC. 24 Previous observational studies comparing treatments have been small (17-181 patients), with variable results with regard to case fatality, 14,25,26 functional outcome, 27,28 and hematoma expansion. 29 These studies also combined patients receiving FFP with either PCC 26,29 or no reversal, 14,23,27 precluding direct comparison of specific reversal strategies.
Our data show that receiving no clotting factor replacement is associated with more than twice the risk of death by 30 days, when compared to all 3 treated groups. This might be expected, as patients without reversal are a select group, usually palliated early due to a perceived dismal prognosis. However, it is possible that reversal treatment does reduce mortality, given that we found similar results in our propensity score analysis. Our study had sufficient power to compare survival in those receiving FFP and PCC. The finding that patients with either therapy have an equivalent adjusted risk of death in our large cohort is thus an important and novel finding.  The observed trend toward a reduced adjusted risk of death in patients receiving a combination of FFP and PCC must be interpreted in the context of our observational study design. This group was relatively small (131 patients vs >350 patients in each of the other 3 groups), and it is of note that the combination group had slightly higher rates of surgery (15% vs 10-13%). Surgery was not adjusted for in our prespecified Cox regression model, given the existing uncertainty around the impact of surgery on outcome. 10,12,30 However, adding surgery into the model post hoc did not markedly change our findings. The combination group consisted largely of patients from 2 Australian centers; thus, there may be unmeasured confounding factors related to the characteristics of the local population or aspects of clinical care that could account for the observed difference. It is also possible that factors present in FFP such as FXIII, fibrinogen, antithrombin, and von Willebrand factor may confer additional benefit when combined with PCC. For example, FXIII is present in FFP (but not PCC) and creates fibrin cross links, making clots more stable. A drop in FXIII activity has previously been associated with reduced hematoma growth in patients not on VKA, suggesting that consumption of FXIII stabilizes clots. 31 FFP also contains endogenous inhibitors of fibrinolysis (eg, plasminogen activator inhibitor 1 and 2). Our finding that combination treatment was associated with a lower risk of death remained robust in all the prespecified subgroups (see Fig 4).
It is more difficult to interpret the finding that among patients receiving PCC only, 3-factor PCC was associated with lower case fatality compared to 4-factor PCC. That patients who received the combination therapy also had 3-factor PCC as their PCC product raises the possibility that there is a true difference between different PCC products. In patients taking warfarin, suppression of prothrombin (FII) activity levels seems to be more important for sustained antithrombotic efficacy than suppression of other vitamin K-dependent clotting factors. 32 Thus, administration of 4-factor PCC with significant FVII content may not be essential to reducing further bleeding and could increase thrombotic complications. The compositions of the multiple PCC products on the market also vary by heparin, antithrombin, and protein C and S concentrations, which all may be relevant. We did not have sufficient numbers to compare individual products, or collect data on thrombotic complications or causes of death, so further studies are required to test this hypothesis.
Our study benefits from a large sample size of 1,547 patients. The patients included in the study are representative of routine clinical care in varied health care systems across a range of countries. The majority of patients presented between 2004 and 2013 and thus are more closely representative of current practice than cohorts described in previous studies. All patients included in our analysis had key prognostic factors available describing age, level of consciousness on presentation, imaging findings, and baseline INR, allowing us to adjust for these in our multivariate models.
There are also some potential weaknesses inherent in our observational study design. Our results might be limited by bias related to unmeasured or hidden covariates. Allocation to treatment group was dictated by the preferences of the attending physician in consultation with the patient and/or their family, guided by local and national policies. One of the main determinants of outcome in ICH is the decision to palliate early, which often forms a so-called self-fulfilling prophecy. 10 Half of the patients without any reversal fell into this category. The decision not to treat patients may also be influenced by unmeasured factors, such as poor premorbid health or frailty, and these may contribute to the observed association between poor outcome and no treatment. It is possible that other factors specific to each center (such as the use of other interventions and level of supportive care, or ethnic composition) could also have influenced survival.
As each center followed a certain treatment strategy and thus the center and treatment effects were strongly correlated, we performed a post hoc shared frailty analysis, which within its limitations suggested that the differences in outcome were not explained by center-specific differences. Although our data set is large, the numbers in each treatment arm are still relatively small, and we thus risk both overmodeling and type II errors. As our observational study can only be considered hypothesis generating, this is not a major risk. The methodology of ICH volume estimation differed across the centers but is unlikely to be a major source of error, as the ABC/2 and planimetric methods used produce similar results in regular and irregular bleeds alike. 21 Estimates of IVH volume varied and were missing in many patients, as were data on baseline blood glucose. Also, we did not adjust for other factors that have been variably associated with outcome in ICH, such as blood pressure or treatment delays, as they were not in our statistical analysis plan and not available. Finally, we do not have data describing functional outcomes, and thus do not know whether the association of improved survival with treatment is at the expense of a greater proportion of patients living with severe disability. Our findings support the hypothesis that the use of clotting factor replacement therapy in VKA-ICH is associated with improved survival. Although some guidelines recommend PCC over FFP based on limited evidence and expert opinion, 11 we found no evidence to support this policy. The clinical significance of our preliminary finding of improved survival in those treated with combination therapy, or with 3-factor PCC compared to 4-factor PCC, requires further investigation. Based on our findings, we would recommend prompt clotting factor replacement for VKA-ICH, with the choice of preparation guided by local policy. A definitive randomized controlled trial testing the effect of PCC, FFP, and combination therapy on functional outcomes is urgently needed to resolve the existing uncertainty regarding the best reversal strategy for VKA-ICH. It has been previously estimated that a population base of around 67 million would be required to conduct a 5-year trial of this nature, 33 so such a trial would be a major undertaking, although the effect sizes noted in our observational study and adaptive or 3-arm trial designs may make such a trial more feasible.