The consult reads: "78F, recurrent TIAs, please see." The story is oddly consistent — two minutes of tingling that marches up the left hand to the face, three episodes this month, each identical. The MRI, when someone finally orders the right sequence, answers the question the story was asking: a dozen lobar microbleeds and a ribbon of dark signal tracing the right central sulcus — cortical superficial siderosis.

Those are not TIAs. They are amyloid spells, and the difference is not academic: the treatment for TIA is an antiplatelet, and an antiplatelet is precisely what this cortex does not need.1

What CAA is

Cerebral amyloid angiopathy is deposition of amyloid-β in the walls of cortical and leptomeningeal arterioles, stiffening and weakening them with age. It travels with Alzheimer pathology but is its own disease: where hypertensive arteriopathy ruptures deep vessels (basal ganglia, thalamus, pons), CAA ruptures superficial ones, producing lobar hemorrhages — often recurrent, often in the same regions — plus the smaller stigmata that let us diagnose it in life: strictly lobar microbleeds and cortical superficial siderosis, the residue of bleeding over the convexity.23 It is also a contributor to vascular cognitive impairment in its own right, hemorrhage or no hemorrhage.

Boston 2.0, in plain terms

The Boston criteria formalize the in-life diagnosis; version 2.0 (2022) validated the additions.1 Stripped of nesting, the logic for probable CAA in a patient 50 or older is:

  • At least two strictly lobar hemorrhagic lesions on blood-sensitive MRI — any mix of lobar ICH, lobar microbleeds, and foci of cortical superficial siderosis — with none in deep territory; or
  • One lobar hemorrhagic lesion plus a qualifying white-matter feature — severe visible perivascular spaces in the centrum semiovale, or the multispot pattern of white matter hyperintensities;
  • and no other adequate cause — no deep bleeds arguing hypertensive disease, no lesion-specific explanation like the ones covered in the cavernoma and AVM chapters.

The practical prerequisite is the right sequence: SWI or GRE. A CAA brain can look nearly normal on CT and FLAIR; the diagnosis lives on the susceptibility images — the same detection point we make in the microbleeds chapter.

Amyloid spells: the TIA mimic with opposite treatment

Transient focal neurologic episodes occur in a meaningful minority of CAA patients: brief, recurrent and stereotyped, most characteristically spreading — paresthesias marching over minutes in a way that resembles migraine aura more than an embolic TIA — usually referable to a region carrying siderosis or a prior bleed. Their importance is double. First, they predict a genuinely elevated near-term risk of lobar hemorrhage at that site. Second, the reflexive TIA pathway — load an antiplatelet, discharge on it — adds bleeding risk to exactly the patient least able to afford it. The honest bedside rule: an older adult with recurrent stereotyped spreading spells earns a susceptibility MRI before anyone commits to the word "TIA" — the same discipline about that label we teach in what is a TIA, pointed in the other direction.

Stereotyped, spreading, and repeating is not the signature of embolism. Get the SWI before you get the aspirin.

The antithrombotic problem

CAA's clinical center of gravity is a risk calculation. Recurrent ICH risk after a CAA-related lobar hemorrhage is substantial — roughly 7–9% per year in cohort data, and higher still when cortical superficial siderosis is present, which is the single most useful imaging predictor.2 Against that backdrop:

  • Anticoagulation after CAA-related ICH is generally avoided; the 2022 ICH guideline frames resumption after any ICH as an individualized weighing, and probable CAA loads the bleeding side of that scale heavily.2 When atrial fibrillation forces the question, it deserves an explicit, documented risk conversation — the general timing framework is on the anticoagulation timing page, and bleeding-risk arithmetic on HAS-BLED — with left atrial appendage occlusion increasingly the escape hatch discussed for exactly this phenotype.
  • Antiplatelets are a smaller but real question: reasonable when a strong ischemic indication exists, avoided when the "indication" was an amyloid spell mislabeled as TIA.
  • Blood pressure control is the intervention with the clearest payoff — long-term control reduces recurrent ICH and is guideline-endorsed after any hemorrhage.2
  • Statins after lobar ICH are their own nuanced conversation — individualized, not reflexive.

Severity of the index bleed is scored like any ICH (ICH score); it is the etiology that rewrites the aftercare.

Two corners worth knowing

CAA-related inflammation is the treatable emergency inside this disease: subacute cognitive decline, headache, seizures, or focal deficits in an older adult, with asymmetric confluent white-matter edema on FLAIR overlying microbleed-rich cortex. It behaves like an autoimmune response to vascular amyloid and often responds to corticosteroids — a diagnosis worth making quickly, because it is the one CAA presentation that improves with treatment. And the anti-amyloid era has made CAA everyone's business: the microhemorrhage-and-edema changes (ARIA) seen with amyloid-lowering immunotherapies concentrate in patients with underlying CAA, which is why baseline microbleed counts now gate those treatments — one more reason the susceptibility sequence and an honest CAA read matter beyond the stroke service.

The bottom line

Think CAA whenever hemorrhage or hemosiderin sits at the surface of an older brain: lobar ICH, strictly lobar microbleeds, siderosis. Get the SWI, apply Boston 2.0, and let the diagnosis do its real work — reframing the "TIAs" that are spells, restraining the antithrombotic reflex, tightening blood pressure, and catching the steroid-responsive inflammatory variant. Fragile vessels forgive few second chances; the diagnosis exists so you stop spending them.