Some advances in technologies can help clinicians to better understand diseases and to modify their attitude towards management of their patients in terms of therapeutic strategies. Cerebral microbleeds (CMBs) are recently discovered lesions that have had a significant effect on neurologists’ conceptions about cerebral vasculopathies, and about which many questions still need to be answered for clinical practice.

CMBs were first described in the 1990s, after the development of magnetic resonance imaging (MRI) sequences called gradient echo T2*. CMBs appear as small round hypointense lesions with a black appearance and a diameter of < 5–10 mm; they are typically located either in the deep brain (basal ganglia) or in the corticosubcortical regions (lobar CMBs). In some patients, the CMB topography is mixed ( Fig. 1 ). To distinguish between a CMB and differential diagnoses, including calcification, iron deposition (especially when lesions are located in the basal ganglia), cavernoma and the end of veins, diagnostic criteria have been proposed by consensus . Histopathological analyses of these radiological constructs indicated that they corresponded to focal perivascular haemosiderin deposits, suggestive of an underlying vasculopathy . Since the initial description of CMBs, the literature on the subject has grown, and now describes the frequency of these lesions, their association with neurological diseases and their prognostic value.

Axial T2* gradient echo-weighted brain magnetic resonance imaging of two patients, showing: (A) multiple bilateral lobar cerebral microbleeds (some are shown with arrows); and (B) both lobar (arrows) and deep (arrow heads) cerebral microbleeds associated with bilateral basal haemorrhages (double arrow heads), suggestive of small-vessel disease.

The prevalence of CMBs was initially estimated at about 5% in the healthy population, with no difference between Asian and non-Asian people . However, this prevalence may have been largely underestimated. Actually, the prevalence of CMBs differs greatly depending on the setting of the studied populations and the MRI variables used for their detection . Hence, in the population-based Rotterdam scan study, 15.3% of the participants had at least one CMB, and the authors demonstrated that the prevalence doubled with each decade, from 6.5% in patients aged 45–50 years to 35.7% in those aged > 80 years . In addition, with age, hypertension was shown to be strongly associated with both the prevalence of CMBs and the occurrence of newly developed CMBs during follow-up . Other vascular risk factors associated with CMBs include smoking and diabetes .

The prevalence of CMBs was also shown to be greater in patients with a history of cerebrovascular diseases. Hence, 34% of patients with ischaemic stroke and 60% of those with spontaneous intracerebral haemorrhage (ICH) had at least one CMB, according to a large review of the literature . Furthermore, CMBs were more frequent among people with recurrent stroke than in those with first-ever strokes, both for ischaemic stroke (45% vs 23%) and ICH (83% vs 52%) . Several studies also pointed out a greater association between CMBs and lacunar infarcts than other ischaemic stroke subtypes; CMBs were identified in 53% of patients with lacunar strokes, in 36% of those with atherothrombotic stroke and in 19% of those with cardioembolic strokes . Moreover, lacunar strokes were associated with an increased risk of developing new CMBs over time . Patients with cerebral white matter hyperintensities, defining leukoaraiosis, also had a greater risk of CMBs and of developing new CMBs during follow-up . Finally, several studies have provided evidence that CMBs are associated with cognitive impairment, dementia and Alzheimer’s disease .

Based on these observations, it has been suggested that CMBs could reflect distinct vasculopathies, and that the anatomical distribution of the lesions may help to distinguish between them. On the one hand, CMBs may be a marker of small-vessel disease, as suggested by their association with both lacunes and leukoaraiosis. In these conditions, CMBs preferentially affect deep brain regions (basal ganglia, thalamus and brainstem), and occur more frequently in patients with vascular risk factors . On the other hand, CMBs may reflect cerebral amyloid angiopathy (CAA), a disease characterized by recurrent lobar ICH and a strong epidemiological association with Alzheimer’s disease . In this situation, CMBs are more frequently found in the corticosubcortical regions of the brain. Of note, a link between lobar CMB and the apolipoprotein E e4 allele was found in the Rotterdam scan study . Given the relationship between apolipoprotein E e4 and CAA, it is highly likely that CMBs could reflect CAA. Furthermore, an association between CMBs and either focal subarachnoid haemorrhage or superficial siderosis ( Fig. 2 ) is highly suggestive of CAA .

Axial T2* gradient echo-weighted brain magnetic resonance imaging of a patient with cerebral amyloid angiopathy, showing bilateral superficial siderosis appearing as hypointense linear lesions within the brain sulci (arrows).

From a prognostic point of view, several studies have demonstrated that individuals with CMBs are at a greater risk of incident stroke . Of note, a recent study showed that CMBs considered unrelated to CAA were associated with an increased risk of both ischaemic stroke and ICH, whereas patients with CAA-related CMBs (i.e. lobar CMBs) had an increased risk of ICH only . In addition, CMBs were also associated with a greater risk of recurrent stroke. A meta-analysis of 10 studies showed that the risk of recurrent stroke was 2.2 times higher for patients with CMBs (odds ratio: 2.25; 95% confidence interval [CI]: 1.70–2.98) . Interestingly, the increased risk was more pronounced for spontaneous ICH (odds ratio: 8.52; 95% CI: 4.23–17.18) than for ischaemic stroke (odds ratio: 1.55; 95% CI: 1.12–2.13).

The excess of haemorrhagic stroke conferred by the presence of CMBs raises important questions regarding the management of antithrombotic agents in patients with atrial fibrillation. Indeed, in the general population, warfarin users seem to have a higher prevalence of deep or infratentorial CMBs, and a higher incidence of any CMB . Furthermore, a pooled analysis, which included 1460 patients with ICH and 3817 patients with ischaemic stroke or transient ischaemic attack, concluded that CMBs were more prevalent in patients with ICH, whatever the premorbid use of antithrombotic agents . The authors demonstrated that the excess of CMB prevalence (odds ratio) in patients with ICH increased from 2.8 (95% CI: 2.3–3.5) in non-antithrombotic users to 5.7 (95% CI: 3.4–9.7) in antiplatelet users and 8.0 (95% CI: 3.5–17.8) in warfarin users, and that the presence of CMBs at baseline was associated with an increased risk of subsequent ICH . A study that included patients with ischaemic stroke, among whom 92.5% were treated with a single antiplatelet agent, 4.3% were treated with warfarin and 3.2% were switched from an antiplatelet agent to warfarin or vice versa during follow-up, suggested that the risk of subsequent ICH increased with the number of CMBs. Hence, during a mean follow-up of 26.6 ± 15.4 months, the incidence of ICH was 0.6% in patients with no CMBs, 1.9% in those with one CMB, 4.6% in those with two to four CMBs and 7.6% in those with at least five CMBs . The risk of future ICH did not differ according to the antithrombotic regimen used. In contrast, a longitudinal study indicated that in stroke patients with at least three deep CMBs, warfarin use was associated with an increased incidence of deep ICH (14% versus 4%) . In a study that enrolled more than 500 patients who had had an ischaemic stroke related to non-valvular atrial fibrillation, the prevalence of CMBs was 31%. About 97% of patients were treated with warfarin at discharge. After 2.5 years of follow-up, the mortality rate was 34.9% ; this rate was higher in patients with CMBs than in those without CMBs (41.9% vs 31.8%), and the authors found that having at least five CMBs was independently associated with all-cause mortality (hazard ratio: 1.99; 95% CI: 1.03–3.85) and ischaemic stroke mortality (hazard ratio: 3.39; 95% CI: 1.39–8.28). In addition, patients with strictly lobar CMBs had a greater risk of haemorrhagic stroke mortality (hazard ratio: 5.91; 95% CI: 1.58–22.11).

As non-vitamin K antagonist oral anticoagulants (NOACs) have been shown to be associated with a lower risk of ICH compared with warfarin , one could question the relationship between NOACs and CMBs. To date, few data on this topic are available. A recent study enrolled 69 patients with atrial fibrillation who underwent MRI at inclusion and at 1 year to evaluate the occurrence of new CMBs with regard to the antithrombotic regimen prescribed (23 on NOACs, 21 on warfarin, nine on warfarin plus an antiplatelet agent, nine on an antiplatelet agent and seven without antithrombotic treatment) . The authors reported that new-onset CMBs occurred in 13.0% of patients overall. No new lesions were observed in the NOAC or antiplatelet groups, whereas new-onset CMBs were noted in five patients on warfarin, three patients on warfarin plus an antiplatelet agent and one patient with no antithrombotic treatment.

Based on our current knowledge, there is no clear evidence to justify different treatment decisions according to the presence of CMBs, especially with regard to antithrombotic use in patients with atrial fibrillation. However, further long-term observational studies are needed to assess whether the systematic evaluation of CMBs by MRI, in association with other markers, would be useful in such patients to better estimate their risk of subsequent ICH, and whether therapeutic strategies regarding the use of warfarin or NOACs and left atrial appendage closure could be adjusted according to this risk.