Oral anticoagulant therapy is an effective treatment for stroke prevention in patients with atrial fibrillation and other forms of cardiac sources of embolism. However, vitamin K antagonists (VKAs) and novel direct oral anticoagulants (NOACs), such as direct thrombin inhibitors or direct factor Xa inhibitors, are associated with an increased bleeding risk, including intracerebral hemorrhage (ICH). In the absence of a previous history of ICH, reduction in the risk of thromboembolic events outweighs the increased risk of ICH. In subjects with previous ICH, the equilibrium changes and the risk/benefit ratio is unpredictable since those subjects were excluded from clinical trials on the prevention of thromboembolic events with oral anticoagulants. Even in the absence of reliable research data, it is widely accepted that anticoagulants should not be resumed in the acute stroke phase for the risk of hematoma expansion and early re-bleeding [1]. At variance, long-term prophylaxis is concerned and opinions are controversial; guidelines do not, in most cases, deal with this issue or in cases where they do, the level of evidence is generally low [1–3].

Some observational studies have evaluated the possible risks or benefits of resuming the use of anticoagulants in subjects with ICH [4–8]. Those studies are somewhat conflicting, and their results are of limited utility in clinical practice since in some instances studies only provide the risk of hemorrhagic but not of thrombotic events, only address the incidence of recurrences without taking into account other major outcomes such as mortality and disability, have a retrospective design, or lack statistical power. In addition, in all those studies, the subjects who resumed anticoagulants were not randomly selected. Consequently, results are not generalizable since only subjects considered at low bleeding risk had probably been selected to resume this treatment.

Further data, from randomized studies, are needed to establish the optimal treatment for subjects with a cardiac source of embolism who had an ICH. Nonetheless, in our opinion, at the present state of our knowledge, the use of anticoagulants should be avoided in most if not in all subjects with ICH in the presence of a cardiac source of embolism, since patients would be exposed to the deadliest and least treatable type of stroke event. Actually, even in subjects who do not use anticoagulants, ICH is the deadliest form of stroke with a 30-day case-fatality-rate close to 50 % [9, 10] while for ischemic stroke associated with atrial fibrillation, the 30-day case-fatality rate is equal to approximately 30 % [11]. In anticoagulant users, compared with non-users, hematoma volume is higher, and there is also an increased rate of early hematoma expansion [12]. This correlates with a worse clinical outcome in anticoagulant-related ICH than in spontaneous ICH [12]. Up to 76 % of patients with VKA-associated ICH may have severe disability or may die in the short term [13]. Mortality rates remain high and clinical outcomes generally remain poor despite reversal of anticoagulation [14]. Although NOACs have been associated with a reduced risk of ICH in clinical trials, we do not yet know if this reduced risk also applies to subjects with a previous ICH since they were excluded from those studies. Despite the reduced ICH occurrence associated with NOAC, those drugs have not been associated with reduced mortality rates compared to warfarin [15]. We should also consider that if the patient has a recurrent ICH we do not have any treatment that can significantly improve the prognosis, while if the patient has an ischemic event thrombolysis, within the appropriate time window, can significantly improve the prognosis. The probability of harming patients may be even higher in subjects with lobar ICH. The location of ICH may reflect the underlying microvascular pathology, which is amyloid angiopathy for most cases of lobar ICH occurring in subjects aged >55 years, while for deep ICH it is arterial hypertension. Lobar ICH is associated with a higher probability of re-bleeding compared to deep hemispheric ICH, with a 1-year risk of recurrence of 15 and 2.1 %, respectively [16]. Antihypertensive therapy likely reduces the risk of recurrence in patients with hypertensive hemorrhage. Unfortunately, there is currently no way to modify the risk of ICH associated with cerebral amyloid angiopathy. This concept is also supported by data from a decision analysis model based on retrospective data that suggest that in settings in which the risk of ICH recurrence is high, such as in lobar ICH, the increased mortality associated with warfarin ICH is itself sufficient to argue against anticoagulation, while risks and benefits of anticoagulation are more closely balanced when applied to patients with deep hemispheric ICH [17].

In patients with atrial fibrillation, the thromboembolic and hemorrhagic risk has been classically estimated using the CHADS2 or CHA2DS2-VASc and HAS-BLED score systems, respectively. A favorable balance for resuming anticoagulants has been suggested when CHADS2 score is ≥4 or CHA2DS2-VASc score is ≥5 with a risk of ischemic stroke >6.5 % per year and the HAS-BLED is ≤3 [18–20]. However, in our opinion this paradigm is not unconditionally true; secondary rather than primary ischemic stroke prevention is a stronger indication for resuming anticoagulants, and a CHADS2 score ≥4 is not necessarily indicative of a previous stroke, while the HAS-BLED scoring does not consider the risk associated with a previous ICH. Based on these data, if we try to define the characteristics of a patient for whom it is relatively safe to resume anticoagulation after an ICH, we should consider a patient with a previous ischemic stroke, a deep ICH and a good blood pressure control, a typical scenario of the minority of cases. In addition, the risk of new intracranial bleeding associated with resumption of anticoagulants may be magnified in the elderly because of the increased risk of falls, the presence of comorbidities (shared risk factors for ischemic and hemorrhagic stroke), the poor compliance to therapy or polytherapy, and the presence of white matter lesions and microbleeds. All these factors are more common with increasing age.

In conclusion, we want to underscore that the chance to have an ICH and to require anticoagulants for cardioembolism is an extremely difficult situation for the patient and for the physician who has to take a troublesome decision. At the moment, while hoping for new clinical evidence to emerge, we suggest to avoid the prescription of a treatment that may contribute to worsening of the natural history of our patients.