Resumption of oral anticoagulation after spontaneous intracerebral hemorrhage

After few and small-sized case series, the first larger retrospective observational analysis was published by Majeed et al. who investigated 234 patients with intracranial hemorrhage including 83 patients with ICH (55%, n = 83/234) documenting an overall increased risk for intracranial hemorrhage recurrence if OAC was resumed (11.5% versus 17.8%) [28]. In contrast, data from a Canadian registry (including 89% ICH patients, n = 252/284) reported annual recurrence rates less than 2.5% as well as a decreased mortality in patients resuming OAC [56]. Similar, a large Italian multicenter study documented an annual recurrence rate of 2.6% among 267 patients with intracranial hemorrhage resuming OAC (including 33% ICH patients, n = 88/267). All these studies remained inconclusive on whether or not risks for ICH recurrence were increased in a setting of OAC resumption [25].

From the year 2015 onwards there was growing evidence. The observational “geRman-widE mulTicenter Analysis of oRal Anticoagulation-associated intraCerebral hEmorrhage” (RETRACE) study included patients with OAC-associated ICH and investigated thromboembolic and hemorrhagic complication rates according to OAC exposure during one year of follow-up [24]. Among 719 survivors with AF, resumption of OAC significantly reduced thromboembolic events (OAC: 9/172 [5.2%] versus no-OAC: 82/547 [15.0%]; p < 0.001) without leading to increased rates of re-bleeding (OAC: 14/172 [8.1%] versus no-OAC: 36/547 [6.6%]; p = 0.48) [24]. Furthermore, OAC resumption was associated with a decreased long-term mortality risk among patients included in a propensity-matched survival analysis (HR: 0.258 (95% Confidence Interval [CI], 0.125–0.534; p < 0.001); i.e. 9 patients with OAC of 108 died (8.3%) compared to 47 patients without OAC of 153 (30.7%; p < 0.001) [24].

The same year, a large Danish registry including 1752 patients reported data strongly supporting these results [32]. The authors found a significantly decreased adjusted HR [0.55, 95% CI (0.39–0.78)] for all-cause mortality, stroke, and systemic embolism in patients on oral anticoagulant treatment in comparison with no treatment during 1-year follow-up [32]. The annual incidence rate of ischemic stroke and systemic embolism among patients using OAC was halved (5.3, 95% CI (3.3–8.5) per 100 patient years) compared with patients without antithrombotic treatment (10.4) or on antiplatelet therapy (10.3). For recurrent ICH, rates of 8.0 for OAC treated patients again did not significantly differ from 8.6 for patients with no antithrombotic treatment [adjusted HR, 0.91, 95% CI (0.56–1.49)], and 5.3 for patients using antiplatelet therapy [adjusted HR, 0.60, 95% CI (0.37–1.03)]. Another Danish population-based cohort study (n = 2978) confirmed these results, showing a significant lower risk of death [adjusted HR, 0.59, 95% CI (0.43–0.82)] and thromboembolic events [adjusted HR, 0.58 95% CI (0.35–0.97)] in ICH patients with post-discharge use of OAC, again without significantly increasing risks for major bleedings or recurrent ICH [adjusted HR 0.65, 95% CI (0.41–1.029] [34]. These results favoring resumption of OAC were further reproduced by several subsequent observational and registry studies [35, 36, 53].

To this day, the largest registry study was conducted in Taiwan and included 12,917 patients with intracranial hemorrhage from 1996 to 2011, reporting divergent results [4]. Chao and colleagues documented an increased risk for hemorrhage recurrence for both patients resuming OAC [HR 1.58, 95% CI (1.27–1.98)] as well as patients taking antiplatelet therapy [HR 1.36, 95% CI (1.19–1.57)] after propensity score matched analyses [4]. According to the authors only patients with a major thromboembolic risk (CHA₂DS₂-VASc-Score ≥ 6) would have a net-benefit resuming warfarin, shown by comparison of the number needed to treat of 27 (for preventing one ischemic stroke) versus the number needed to harm of 91 (for producing one ICH) [4]. However, these results might also be influenced by in general increased ICH risk in Asian patients [51].

Several meta-analyses of reported data have been conducted until now, all of them showing a significant reduction of thromboembolic complications without leading to increased risk of ICH recurrence [3, 22, 30, 57]. Furthermore, antiplatelet agents – sometimes considered as a safer alternative approach – were not beneficial neither for thromboembolism prophylaxis nor prevention of ICH recurrence [22]. One recent meta-analysis of individual patient data (n = 1012) did also address the association of OAC resumption with functional outcome, documenting that OAC resumption increases chances for a favorable outcome after 12 months (modified Rankin Scale 0–3) by 4-fold in both non-lobar and lobar ICH patients [2]. Even in the absence of recurrent clinically apparent stroke, patients resuming OAC seem to benefit with respect to better functional recovery, hypothetically due to prevention from micro- embolism cumulating to significant central nervous system damage influencing post-ICH recovery associated with cardioembolic stroke risk [29].

Of note, all of these observational studies harbor important limitations due to confounding by indication and selection bias [25]. Physicians individually weigh patient’s risk for ischemic versus hemorrhagic complications which results in selected patients with favorable risk-benefit-profiles restarting OAC. This might also be reflected by their younger age, less severe ICH, and better functional outcome in observational studies [24, 46]. In general, withholding therapy in severely affected patients is frequent in ICH care possibly further affecting post-discharge drug prescription [44, 46]. As statistical adjustment is to a large extent possible for quantifiable parameters, additional unmeasured variables likely introduce residual bias influencing the reported associations [22, 46]. Further, many investigations included heterogeneous patient cohorts combining different intracranial pathologies – mostly ICH, but also patients with subarachnoid hemorrhage, epidural or subdural hematomas – as well as OAC indications – atrial fibrillation, mechanical heart valves, deep vein thrombosis – each strongly influencing patients individual risk for recurrent hemorrhage or thromboembolism [25].

Although data from observational studies in the vast majority solely cover resumption of OAC using vitamin-K antagonists (VKA), it seems apparent that Non–vitamin K antagonist oral anticoagulants (NOAC) should be preferred [25]. Compared to VKA large randomized trials have demonstrated a halved ICH incidence in patients using NOAC [10, 20]. Although the mechanism behind risk reduction is not completely understood, it seems that the more selective mode of action, i.e. targeting only one clotting factor, together with limited crossing of the blood–brain barrier (dabigatran) or effluxing out of the brain by p-glycoprotein efflux pumps (rivaroxaban and apixaban) lead to a more beneficial safety profile [55]. Recently, a Bayesian network meta-analysis including 17 randomized controlled trials enrolling 116,618 patients evidenced that all NOACs were safer than warfarin for risk of ICH [55]. Relative ranking probability based on surface under the cumulative ranking curve (SUCRA) suggested the safest profile among NOACs for dabigatran (SUCRA, 0.86) followed by edoxaban (SUCRA, 0.81), apixaban (SUCRA, 0.61), and rivaroxaban (SUCRA, 0.32). VKA was ranked as the least safe drug among all anticoagulants (SUCRA, 0.06). In general, NOACs were associated with a significant 54% relative risk reduction compared with warfarin (Odds Ratio [OR] 0.46, 95% CI (0.35–0.59); p < 0.001). However, this analysis referred to drug safety only and did not focus on inter-class effects regarding prevention of thromboembolic events [55]. Available data for the comparison of severity of vitamin-K-antagonist related ICH with NOAC-ICH provide only little differences, however in-hospital mortality might potentially be reduced in ICH under NOAC [13, 19, 38, 49]. So far, studies specifically analyzing NOAC resumption after ICH do currently not exist [25].

The described observational studies documented a median starting point in between 4 to 6 weeks after ICH. Studies more specifically addressing this question reported a broad range of supposed optimal time points ranging from 72 h to 10–30 weeks [15, 28]. One large Swedish registry study (n = 2619) suggested an optimal time window within 7–8 weeks for resuming OAC after COX regression-based balancing between observed risk of ischemic and hemorrhagic complications [36]. Although having used sound statistical approaches, limitations of that study comprise censoring the first 4 weeks after ICH, narrow information on patient and ICH characteristics as well as treatment allocation gathered by outpatient dispensed drug registry [36]. A meta-analysis from the Kings College in London, UK, assessed the associations of resuming VKA six weeks after ICH with occurrence of both thromboembolic and hemorrhagic complications over a one-year follow-up time frame [22]. In essence, VKA-resumption was verified to be safe without increasing hemorrhagic complications over comparator treatments with platelet inhibitors [risk ratio: 1.34, 95% CI (0.79–2.30), p = 0.28], or no antithrombotic treatment respectively [risk ratio: 0.93, 95% CI (0.45–1.90), p = 0.84] [22]. As yet, it remains unclear when to optimally resume OAC after ICH but ongoing randomized trials might provide further evidence. Current expert opinion would suggest a timeframe between 4 to 8 weeks after index ICH depending on patient’s individual risk profile [25]. Application of a shorter time period to resumption may only be considered in life-threatening situations and compelling indications, such as symptomatic intracardiac thrombus formation or acute pulmonary embolism, and only after confirmation of hematoma stability by control imaging and strict blood pressure control.

As mentioned above, NOAC should be preferred today for OAC resumption in indications like AF or venous thromboembolism. However, this does not apply for patients with mechanical heart valves (MHV) in situ as NOACs were shown to be inferior to VKA in these patients [11]. Neither cardiologic nor neurologic international guidelines provide specific recommendations how to treat MHV patients after ICH [16, 33, 48, 50]. Compared to AF, patients with MHV are at risk of increased thromboembolic complications why a recent consensus paper from the European Society of Cardiology Working Group in Thrombosis recommended that systemic anticoagulation using heparins may be safe to start as early as 3 days after ICH and oral anticoagulation using VKA after 7 days, based on limited data from small observational studies and case series [14].

The best available and most recent data came from a sub-group analysis of the German-wide multicenter RETRACE program which included among 2504 OAC-associated ICH patients 166 patients with MHV in situ [26]. Resumption of therapeutic anticoagulation, using heparins or VKA, was related to a 10-fold increase in incidence of major extra- or intracranial hemorrhagic complications during hospital stay in MHV-patients [rate ratio: 10.3, 95% CI (3.7–35.7)] [26]. Adjusted COX regression modeling for timing of anticoagulation revealed that OAC resumption may be safe after two weeks regarding bleeding complications, whereas optimal balancing of hemorrhagic with thromboembolic complications resulted in an earliest starting point of one week after ICH, to be considered especially in patients with highest risk for thromboembolism, i.e. concomitant AF, mitral position, or older prosthesis types [26]. Main results of this study are summarized in Fig. 1.

Patients with ICH in lobar location compared to hypertensive non-lobar located ICH need special considerations due to strong relation to CAA, characterized by deposition of amyloid-β, micro-hemorrhages, and vascular fragility resulting in a greater risk for ICH recurrence [39, 45]. Genetically CAA shows association with apolipoprotein E alleles (subtypes epsilon 2 & 4) likely leading to amyloid deposition triggering recurrent ICH [16, 41]. A prediction model for CAA-associated lobar ICH integrating genetic characteristics – APOE ɛ4 carrier – and radiological CT parameters – subarachnoid hemorrhage, finger-like ICH projections – showed excellent discrimination [Area under the curve: 0.92, 95% CI (0.86–0.98)] in a recent prospective study [41]. Using MRI and clinical history CAA can be validly identified using the modified Boston criteria [27]. Observational studies documented recurrence rates for patients classified as having definite or probable sporadic CAA of 8.9 per 100 patient-years [95% CI (7.1–11)] [52].

MRI-detected presence and amount of cerebral microbleeds (CMB) seen on iron-sensitive sequences (T2*-weighted gradient-echo or susceptibility-weighted imaging) are associated with both first-ever ICH in ischemic stroke patients using OAC [OR 2.68, 95% CI (1.19–6.01), p = 0.017] as well as with recurrence of ICH [> 10 CMB versus none: OR 5.6, 95% CI (2.1–15), p = 0.001] [8, 9]. Pathologically, CMB correspond to hemosiderin deposits remaining in macrophages following a self-limiting microhemorrhage [47]. Therefore, they are seen as a neuroimaging marker for small vessel disease contributing to most lobar ICH [54]. Further, cortical superficial siderosis (cSS) and cortical or convexity subarachnoid hemorrhage (cSAH) were identified as independent marker for increased hemorrhagic risk [HR 3.92, 95% CI (1.38–11.17), p = 0.011, and HR 3.48, 95% CI (1.13–10.73), p = 0.030] [42]. If both probable CAA and cSS were present one investigation documented an ICH rate as high as 19% (95% CI, 11–32) compared to 6% (95% CI, 3–12) in patients without cSS during 5 years of follow-up [7].

Only one sub-analysis of an individual patient data meta-analysis to date aimed to analyze the impact of OAC on ICH recurrence in patients with probable/possible CAA (n = 190). Although OAC resumption was consistent with overall lobar ICH associated with improved functional outcome and decreased mortality, numbers of patients and events were insufficient for investigating the influence of OAC on complication rates [2]. Summing up, decision making regarding OAC resumption in lobar ICH patients should include extended diagnostic work-up including magnetic resonance imaging to evaluate characteristics such as a microbleed burden, cSS or cSAH [25]. Their presence should lead to even more critical weighing of the potential benefit of OAC resumption versus the increased bleeding risk, considering also alternative interventional strategies for thromboembolism prevention [25]. For a suggested flow chart on OAC resumption in patients with deep compared to lobar ICH location please see Fig. 2.

The role of the left atrial appendage as the most important source of cardiac thromboembolism related to AF has led to introduction of left atrial appendage occlusion (LAAO) into clinical practice as a potential alternative treatment to long-term OAC. Percutaneous LAAO using the to date only approved device (WATCHMAN) showed non-inferiority compared to warfarin for prophylaxis of a primary composite endpoint of stroke, cardiovascular death, and systemic embolism in the multicenter, randomized PROTECT AF trial including 707 patients with non-valvular atrial fibrillation [18]. As several concerns were raised by the U.S. Food and Drug Administration (FDA) especially regarding early acute safety events, a second trial – PREVAIL – was performed, documenting an improved safety over time (7-day procedure-related complications: PREVAIL 4.2% versus PROTECT AF 8.7%, p = 0.004) [17, 18]. However, compared to warfarin this study did only reach non-inferiority regarding stroke or systemic embolism > 7 days [17]. Regarding the coprimary endpoint of the overall composite of stroke, systemic embolism and cardiovascular/unexplained death (18-month rate ratio 1.07 [0.57–1.89]) the upper bound of 1.89 extended the pre-specified noninferiority margin of 1.75, triggered by a higher number of (early) strokes in the intervention group (ischemic and hemorrhagic stroke, intervention: 6 [2.2%] versus control: 1 [0.7%]) [17]. Next to questionable efficacy and peri-interventional complications, the indicated antithrombotic therapy subsequent to device implantation needs consideration. The recommended antithrombotic therapy is related to the patient’s individual bleeding risk. Based on the PROTECT-AF trial protocol, patients of low bleeding risk should receive VKA for 45 days, and then switch to dual antiplatelet therapy until 6 months, followed by life-long low-dose aspirin monotherapy [18]. Patients of high bleeding risk, as among others patients after ICH, should be treated with dual antiplatelet therapy for one to six months, followed by life-long low-dose aspirin monotherapy. The assumed safety of this approach is based on results from the prospective multicenter nonrandomized ASAP feasibility study investigating LAAO in patients who were ineligible for OAC, however, safety in ICH patients remains unclear [40]. One observational study compared 151 ICH patients with AF who underwent LAAO with a propensity score-matched group of 151 patients receiving standard medical therapy [31]. Analyses showed a decreased risk for the composite endpoint consisting of all-cause mortality, ischemic stroke and major bleeding (HR 0.16, 95% CI (0.07–0.37)) as well as for recurrent ICH [HR 0.10, 95% CI (0.01–0.81)] in patients having LAAO [31]. Therefore, LAAO might potentially represent an alternative strategy to chronic OAC therapy in high-risk ICH patients, provided its successful evaluation in ongoing randomized-controlled trials – especially compared with NOAC as a safer and potentially more effective comparator than VKA [18, 21]. Today, according to FDA approval, interventional LAAO is formally contraindicated in patients with high-bleeding risk such as ICH patients and its off-label use should be preceded by a critical and interdisciplinary decision making process [43].