The Clamp and the Collateral

There is a moment in this operation when the surgeon deliberately stops the blood supply to one cerebral hemisphere. Everything else in the day is preparation for that moment, or recovery from it.

The uncomfortable part is how little of what happens around it is supported by evidence, in an operation that is itself among the best-proven in vascular surgery.

The lottery you can look up

Cerebral blood flow comes about 80% from the internal carotids and 20% from the vertebrals, meeting at the circle of Willis. That circle is the primary collateral channel between the hemispheres, and it is the reason cross-clamping one carotid is survivable.

A complete circle of Willis is the exception rather than the rule, and it is worth knowing how firmly that is established, because the numbers in circulation vary a great deal. There are 22 anatomical configurations in the Lazorthes system. Five vessels determine whether the collateral supply works: the anterior communicating artery, the two A1 segments of the anterior cerebral arteries, the two P1 segments of the posterior cerebral arteries, and the two posterior communicating arteries.

The best primary figure available comes from MR angiography in 150 healthy volunteers: an entirely complete circle — both the anterior and the posterior parts — in 42%. So a majority do not have one, and the proportion was slightly higher in younger people and in women.

On MR angiography, agenesis or obstruction of two or more of those five predicts cross-clamp intolerance with a sensitivity of 82% and a specificity of 90%. Cross-filling of contrast from the contralateral carotid into both A1 segments and the anterior communicating artery on conventional angiography is highly suggestive of adequate collateral flow. And the circle’s configuration correlates with the stump pressure the surgeon will measure once the clamp is on.

Which means the single most useful thing about this patient’s tolerance of the operation is available on the imaging before induction, and the source says so directly: the extent and patency of the collateral circulation should be discussed before anaesthesia begins.

Why this day does not quote a single completeness percentage. The textbook’s own chapters disagree, and neither matches the primary study. Chapter 16 gives a complete circle in 15 to 20%. Chapter 14 says the circle may be incomplete in as many as 21% — very nearly the reciprocal claim. The imaging series above says 42%.

Three numbers, one anatomical question, and the explanation is almost certainly that “complete” is not one thing. Whether you require every component merely to be present or also to be of normal calibre; whether you count the anterior and posterior parts separately or insist on both; and what your imaging can resolve — each of those moves the figure, and none of the three sources states its criterion in the same terms as the others.

So this day quotes the primary study, names the range, and declines to pick a winner between two textbook chapters that do not define their term. Nothing here depends on which is right. The predictive statistic above, the fact that collateral supply varies enormously between patients, and the instruction to look at this patient’s imaging all survive any of the three. Neuro 2 day 11 sets the disagreement out in full.

Two knobs, and only two

The brain has essentially no oxygen or glucose reserve and depends on continuous delivery. In severe carotid stenosis the usual protective coupling of flow to metabolic rate is often impaired. And of everything that influences cerebral blood flow, only two things are readily manipulable intraoperatively: arterial blood pressure and PaCO₂.

Carbon dioxide. Flow changes 1 to 2 mL/100 g/min per mm Hg of PaCO₂ between 20 and 80 mm Hg. Both directions have been argued for. Hypocapnia was proposed on a “Robin Hood” premise — vasoconstrict normal brain and divert flow to ischaemic regions — but if hypotension coexists it simply hypoperfuses the normal brain. Hypercapnia was proposed to dilate everything and raise global flow, but it dilates only tissue with preserved CO₂ responsiveness, so it can produce a reverse steal away from ischaemic zones in up to 23% of patients. Regional flow responses during cross-clamping are not predictable, and diabetes — common in this population — impairs the response unpredictably.

The conclusion is deliberately unexciting: maintain normocapnia, referenced to the preoperative arterial blood gas, or to a PaCO₂ giving a normal pH.

Blood pressure. The textbook autoregulatory range of 50 to 150 mm Hg is misleading here. The lower limit in humans averages at least 70 mm Hg, with large inter-individual variation, in patients who usually have coexisting hypertension. Routine deliberate hypertension is not recommended, and today’s pitfall is why. The working rule: patent collaterals, normal preoperative range, accept spontaneous rises up to 20% at clamping; poor collaterals, cautiously 20 to 30% above baseline.

And a rule about interpretation rather than treatment: excessive spontaneous hypertension may reflect cerebral ischaemia, and that should be considered before it is pharmacologically suppressed.

The operation, in the order it happens

Knowing the steps lets you anticipate the haemodynamics rather than react to them.

Incision along the anterior border of sternocleidomastoid, platysma divided, dissection to expose the common, external and internal carotid arteries. Manipulation of the carotid sinus during this exposure causes bradycardia, hypotension, or hypertension — the baroreflex misreading traction as a rise in pressure. Local anaesthetic blockade of the sinus nerve usually attenuates it.

Heparin 70 to 100 units/kg to double the activated clotting time.

Cross-clamping in the order distal internal carotid, proximal common, distal external — the mnemonic is ICE — to minimise embolic stroke. Clamp application often raises the blood pressure through loss of stretch on the vessel wall; rises up to about 20% are acceptable.

Cross-clamp intolerance is detected by neurological examination in an awake patient or by neuromonitoring in an anaesthetised one. Shunt insertion is at the surgeon’s discretion.

Arteriotomy, plaque dissected from the wall, intimal flaps resected, closure direct or with a patch.

Declamping in the reverse order — external, common, then internal — so that luminal debris goes into the external circulation rather than the brain.

Heparin is often deliberately not reversed, because of the risk of acute internal carotid thrombosis. If haemostasis is inadequate, protamine 0.5 mg/kg.

Awake or asleep

This is the argument the room will have, and GALA is the answer: no difference. 3526 patients, 95 centres, 24 countries, and 4.8% versus 4.5% for stroke, myocardial infarction or death at 30 days. No difference in the components taken separately, and none in cranial nerve injury.

Local anaesthesia was cheaper and produced better early neurocognitive performance. A trend favoured it in patients with contralateral carotid occlusion but did not reach significance. A Cochrane review pooling 16 trials and 4839 patients — of which GALA supplied 3526 — found no difference in stroke or in stroke and death at 30 days.

The source’s own conclusion is the one worth carrying: outcome after endarterectomy depends more on patient selection, preoperative optimisation and an experienced, committed surgical team than on the choice of anaesthetic technique.

For a general anaesthetic, the practical points are unremarkable: neuromuscular blockade often maintained throughout, ventilation to normocapnia, hypertension from carotid traction treated first with local anaesthetic infiltration of the sheath or added depth and only then with esmolol or labetalol, hypotension with reduced depth, fluid, and phenylephrine 0.5 to 1.0 µg/kg or ephedrine 5 to 10 mg — which a randomized study during endarterectomy found equally effective at maintaining cerebral oxygenation.

Monitors, measured against a person

The reason the technique argument never resolves is visible in the accuracy data, because every monitor is validated against an awake, cooperative patient, and that patient is described in the source as a highly sensitive and specific means of monitoring neurological status.

Compared against awake testing:

Monitor	Sensitivity	Specificity	Threshold
Transcranial Doppler	0.88	0.92	↓48–70%
Evoked potentials	0.85	0.84	↓0–50% of N20/P25 amplitude
Near-infrared spectroscopy	0.84	0.89	↓15–20%
Jugular venous saturation	0.82	0.86	≤55%
Carotid stump pressure	0.76	0.88	25–50 mm Hg
EEG	0.70	0.96	—

The pattern is worth reading rather than memorising. EEG misses events but rarely invents them — low sensitivity, very high specificity, so a change is usually real. Doppler is the best single monitor and also detects emboli, which lets you distinguish a haemodynamic from an embolic cause of deterioration at clamping. Stump pressure — the mean pressure in the internal carotid above the clamp, reflecting retrograde collateral flow — is intuitive and under-validated: values below 25 mm Hg correlate with intolerance, but the threshold is disputed and the reading is influenced by the anaesthetic agent. NIRS is the most popular among anaesthetists at 28%, ahead of EEG at 24% and awake testing at 23%, and it is better at reassurance than at detection: a fall over 20% has a false positive rate of 67% but a false negative rate of only 2.6%.

The best reported performance is stump pressure combined with either Doppler or EEG. And the honest caveat is that a Cochrane review found the data too limited to support or refute routine or selective shunting, and no monitoring method has been shown to produce better outcomes.

Protecting the brain, or not

Barbiturates, lidocaine, benzodiazepines, etomidate, propofol, ketamine and the volatiles have all shown modest protection in animal models. Clinical efficacy is undetermined. A systematic review of 25 randomized trials in 3274 patients — 88% of them cardiac surgery — found only atorvastatin and magnesium possibly associated with fewer new deficits, and none reduced mortality. Preconditioning and postconditioning have shown a paucity of clinical benefit during endarterectomy specifically.

Which makes the survey data honest reading. At cross-clamping, United States anaesthetists reported raising blood pressure (60.9%), giving 100% oxygen (34%), giving nothing (32.6%), and adding anaesthetic drug — usually propofol — in under 10%. The commonest intervention is the one today’s pitfall is about.

Endarterectomy or stent

CREST is the largest randomized comparison: 2502 patients, 108 centres, operator credentialing, mandatory embolic protection. The composite of stroke, myocardial infarction or death at 30 days plus ipsilateral stroke to 4 years was 7.2% for stenting versus 6.8% for surgery — not significantly different, and still not different at 10 years.

The components are what matter:

• Periprocedural stroke was higher with stenting: 4.1% versus 2.3%.
• Myocardial infarction was higher with surgery: 1.1% versus 2.3%.
• And the two are not equivalent to the patient — stroke cost 15.8 quality-of-life points, myocardial infarction 4.5.

A 2020 meta-analysis of 22 trials and 9753 patients found stenting associated with a higher periprocedural risk of death or stroke in symptomatic disease, with no significant difference under age 70 and a clear disadvantage over 70. Stenting had lower rates of myocardial infarction, cranial nerve palsy and access-site haematoma.

So the choice is a trade between an early neurological risk and an early cardiac one, weighted by age and by anatomy. Stenting has a defined set of favourable indications — previous endarterectomy, contralateral occlusion, previous neck dissection or radiotherapy, lesions above C2, inability to extend the neck, contralateral laryngeal nerve palsy, and patients at high risk of wound infection.

After

Two complications belong to the recovery period and both are yours.

Hyperperfusion syndrome. Flow rises after revascularisation in most patients, usually by less than 35%; in severe cases it can exceed 200% of preoperative levels. There is no agreed absolute definition, but many investigators use a doubling of baseline flow. The clinical picture is a triad of ipsilateral headache, seizures and focal neurological symptoms, with face and eye pain, cerebral oedema and intracerebral haemorrhage. Timing is the diagnostic clue: 3 to 7 days after endarterectomy, within 36 hours after stenting, with reported presentations up to a month after stenting. Pooled incidence 3.4% after endarterectomy and 2.2% after stenting; associated intracranial haemorrhage 0.1% and 0.8% respectively. Transcranial Doppler is the usual diagnostic modality, CT perfusion where the acoustic window is poor. And a sudden rise in regional oxygen saturation of 5% immediately after declamping has been reported to predict it with 100% sensitivity and 86% specificity — one of the few places NIRS earns a strong number.

This is why postoperative blood pressure is held at the preoperative normotensive level or 20% below it.

Postoperative hypertension itself is commoner in patients with pre-existing, poorly controlled hypertension and after procedures in which the carotid sinus is denervated. The differential for postoperative instability is broad — myocardial ischaemia, atrial fibrillation, hypoxia, hypercarbia, pneumothorax, pain, confusion, stroke, and a distended bladder.

Neck haematoma. In NASCET the incidence was 7%, about 45% of them moderate to severe, and its occurrence was itself a significant risk factor for perioperative stroke and death. It can progress rapidly to airway obstruction and death, and an expanding haematoma is an emergency requiring reintubation and surgical exploration. In a reported series, 64% were reintubated before induction and 36% after; awake fibreoptic intubation after topicalisation succeeded in 75% of attempts; awake direct laryngoscopy succeeded in 71% and post-induction direct laryngoscopy in 87%; one patient had an awake tracheostomy; and in three patients intubation became possible only once the haematoma had been decompressed.

That last detail is the one to remember. When the neck is too distorted to intubate, opening the wound is not a surgical step that follows airway management — it is part of it.

Cranial nerve palsy completes the list at 5 to 12%: hypoglossal, vagus and facial branches, usually traction injuries, presenting as dysphagia, with operations over two hours carrying more risk. GALA found no difference between anaesthetic techniques.

The shape of the day

Be confident about the indication. The operation prevents strokes and the thresholds are written in terms of the local complication rate — under 6% for symptomatic stenosis of 50 to 99%, under 3% for asymptomatic stenosis of 60 to 99% with more than five years of life expectancy.

Be humble about the conduct. Awake or asleep does not change outcome. No monitor has been shown to improve outcome. Shunting has neither been supported nor refuted. Pharmacologic neuroprotection does not work.

And spend your attention where the harm actually is: on the collateral anatomy before induction, on the embolic moments during manipulation and declamping, on the heart in a population where two in five have coronary disease, and on the neck afterwards.