Two patients, same lesion, opposite problems. The first: a 24-year-old with a first seizure, MRI showing a 3 cm right frontal flow-void tangle — an unruptured AVM found by accident of physiology. The second: a 31-year-old with a lobar hemorrhage and a nidus at its rim. The first patient's enemy is decades of small annual risk and the temptation to intervene. The second's enemy is the next twelve months. Most of the confusion in AVM care comes from letting evidence about one patient answer for the other.

What an AVM is, and how it announces itself

A brain AVM is a developmental error of vasculogenesis: arteries feeding directly into veins through a tangled nidus, no capillary bed to drop the pressure. The veins arterialize, the shunt recruits flow, and the lesion lives under hemodynamic stress for the patient's whole life. Roughly half declare themselves with hemorrhage — typically lobar, in a young person, the scenario that puts AVM on every young-ICH differential alongside the entities in the cavernoma and CAA chapters.1 Most of the rest surface as seizures, headaches, or incidental findings. Catheter angiography remains the definitive map — feeders, nidus, drainage, and the associated flow-related aneurysms that matter below.

The rupture math

The teaching number — 2–4% per year — is real but lumpy.1 Risk concentrates when the anatomy says pressure and the history says precedent: prior hemorrhage (the strongest factor, especially in the first year), exclusively deep venous drainage, deep or infratentorial location, and intranidal or feeder aneurysms. A superficial unruptured AVM with cortical drainage sits near the bottom of the range; a previously-ruptured deep lesion with deep drainage compounds toward the top. Lifetime framing matters for a 24-year-old: even 2% per year is a large cumulative number over 50 years — which is precisely the intuition ARUBA forces us to test rather than assume.

Spetzler-Martin: a table for the right question

Feature Points Note
Size<3 cm = 1 · 3–6 cm = 2 · >6 cm = 3Largest nidus dimension
Eloquent locationNo = 0 · Yes = 1Sensorimotor, language, visual cortex, thalamus, brainstem, deep nuclei
Deep venous drainageNo = 0 · Yes = 1Any deep draining vein counts
GradeSum 1–5: grades I–II generally favorable surgical candidates; grade III intermediate/heterogeneous; IV–V high surgical risk2

Keep the scale honest about what it measures: operability, not rupture risk. A grade V AVM is dangerous to resect, not necessarily likely to bleed; deep drainage is the lone feature that appears on both lists.

ARUBA, honestly read

ARUBA randomized adults with unruptured AVMs to medical management alone versus interventional therapy (any modality) and stopped early: death or symptomatic stroke occurred in about 10% of the medical arm versus roughly 30% of the intervention arm over a mean ~33 months.3 Extended follow-up out to five years told the same story.4 The criticisms are substantive and should be said in the same breath: intervention was a heterogeneous mix light on surgery for favorable grades, follow-up remains short against a lifetime of annual risk, enrollment selection was debated, and the trial cannot speak to the low-grade lesion a high-volume surgeon removes with minimal morbidity. What survives the criticism is the direction: for the average unruptured AVM over the horizon actually measured, intervention hurt more than the disease did — a burden of proof that now sits on treating, not watching. Guidelines accordingly treat unruptured-AVM intervention as a selective, team-based call.1

ARUBA did not say "never treat an unruptured AVM." It said the treatment must beat 2% a year — and over the trial's horizon, it did not.

Ruptured AVMs are the other disease. After hemorrhage, rebleeding risk runs high in the early years, ARUBA does not apply, and securing the nidus — surgery, radiosurgery, embolization, or combinations, sequenced by the team — is standard thinking. The acute bleed itself is managed as any ICH (severity framed with the ICH score, care per the hemorrhage pathway), with the lesion addressed once the patient and brain allow.

The rest of the visit

  • Seizures — treat AVM-related epilepsy like structural epilepsy: an antiseizure medication after a first seizure is generally warranted (a lesion is exactly the high-recurrence substrate), with drug choice logic borrowed from the post-stroke epilepsy chapter.
  • Blood pressure and the basics — no trial forbids ordinary life, but hypertension control is sensible physiology in a high-flow lesion.
  • The referral — every AVM deserves a multidisciplinary neurovascular conference (surgery, endovascular, radiosurgery) rather than a single-specialty verdict; modality mix and sequencing are exactly the decisions that go wrong in silos.
  • The counseling — a young patient with an unruptured AVM is carrying a probability for decades: give the annual number, the modifiers, the ARUBA result and its limits, and a surveillance plan — the same "number, plan, permission to live" formula as the aneurysm visit.

The bottom line

Split every AVM conversation at the first question: has it bled? Ruptured lesions get secured, by a team, with urgency proportional to front-loaded rebleeding risk. Unruptured lesions get arithmetic — annual risk with its multipliers, Spetzler-Martin for what repair would cost, ARUBA for humility about intervening — and a team decision that usually lands on watching, sometimes on treating, and never on reflex. The tangle is congenital; the outcomes are mostly decided by the decisions.