The page arrives at 4 p.m.: "MRI shows innumerable microbleeds. Patient has AFib, CHA₂DS₂-VASc of 5. Cards wants to know if we're still okay to anticoagulate." Somewhere behind that page is a radiology report that has frightened three services at once with dots almost nobody was looking for a decade ago.

Here is the framework that turns the dots back into a decision.

What a microbleed is — and why the sequence decides the count

A cerebral microbleed is a small (usually 2–10 mm) round hypointensity on susceptibility-weighted or gradient-echo MRI: hemosiderin left behind by a microscopic hemorrhage, marking a diseased small vessel that has already leaked once.1 Three properties matter clinically. They are sequence-dependent — invisible on CT, FLAIR, and T1/T2, increasingly numerous the more susceptibility-sensitive the acquisition, which is why "no microbleeds" on a scan without SWI/GRE means nothing. They are common in the elderly — a routine finding, not an automatic alarm. And they are a biomarker of the underlying arteriopathy, which is where distribution comes in.

Pattern Where the dots sit What it implies
Strictly lobarCortex and gray–white junction, sparing deep structuresCerebral amyloid angiopathy physiology — see the CAA chapter; Boston 2.0 may apply
Deep / infratentorialBasal ganglia, thalamus, brainstem, cerebellumHypertensive small vessel disease — the same substrate as lacunes and deep ICH
MixedBoth compartmentsUsually both arteriopathies; hypertensive physiology tends to dominate management

Mimics are worth one sentence each: calcifications (check the phase images or CT), cavernomas with their popcorn architecture (their own chapter: cavernous malformations), and axonal-injury hemorrhages in the right trauma context.

The pooled data: both risks rise — unevenly

The question that matters is not whether microbleeds predict hemorrhage (they do) but whether they predict it enough to outweigh ischemia. The individual-patient pooled analysis of 38 cohorts — 20,322 patients with recent ischemic stroke or TIA — answered with useful clarity: microbleed presence raised the risk of both recurrent ischemic stroke and ICH; the relative risk increase was steeper for ICH and climbed with burden; and yet, in absolute terms, recurrent ischemic strokes outnumbered hemorrhages in every microbleed stratum — including patients with many microbleeds.2

Even the microbleed-heavy brain, statistically, has more ischemic strokes ahead of it than hemorrhages.

That single fact reorganizes the ward conversation. Withholding a clearly indicated antithrombotic because of dots protects against the rarer event at the cost of the commoner one.

The decision, case by case

  • Atrial fibrillation with a stroke or TIA indication: anticoagulate in most cases — microbleeds alone are not a listed contraindication in the secondary-prevention guideline, and the ischemic math above backs that stance.3 Choose a DOAC over warfarin where eligible, score the modifiable bleeding factors on HAS-BLED (then fix them rather than using the score as a veto), and time the start per the anticoagulation timing page.
  • The CAA phenotype: strictly lobar burden — and above all cortical superficial siderosis or a prior lobar ICH — moves the patient out of "dots, proceed" territory into the individualized conversation of the CAA chapter, where appendage occlusion enters the differential of therapies.
  • Antiplatelets for non-cardioembolic stroke: proceed for real indications; the general framework lives on the antithrombotic selection page. The place to hesitate is the soft indication — primary prevention, or the "TIA" that was actually an amyloid spell.
  • Thrombolysis: a known microbleed burden is not a standard exclusion at the code-stroke decision point — eligibility runs off the lytic checklist, and in practice the burden is rarely known in the window.
  • Everyone with a microbleed-rich scan: long-term blood pressure control — the one intervention that serves both the ischemic and hemorrhagic sides of this ledger, and the same lever emphasized in vascular cognitive impairment, because microbleed-marked small vessel disease is also a cognitive substrate.

The bottom line

Read microbleeds in three steps: confirm the sequence (SWI/GRE), read the distribution (lobar CAA versus deep hypertensive), and only then join the antithrombotic question — where the pooled data say the ischemic threat usually still outweighs the hemorrhagic one. Reserve real hesitation for the CAA phenotype with siderosis or prior lobar bleeding, tighten everyone's blood pressure, and do not let dots — however innumerable the report makes them sound — quietly cost a patient with atrial fibrillation the anticoagulant that prevents the larger disaster.