The Lay Medical Man

Author: Dr Unnikrishnan Prathapadas

  • 🧠 The Slope You’ve Been Ignoring: What the Aperiodic EEG Tells the Anesthesiologist

    🧠 The Slope You’ve Been Ignoring: What the Aperiodic EEG Tells the Anesthesiologist

    Every time we watch a processed EEG monitor, we stare at a single number and a busy little waveform and make a decision about depth. BIS says 45, so we leave the vaporizer alone. But underneath that number sits the raw EEG spectrum, and most of us were trained to look at only half of it: the oscillations, the familiar alpha and delta “humps” that rise and fall as anesthesia deepens.

    The other half, the part that looks like featureless background, turns out to carry some of the most interesting physiology on the screen. That background has a name: the aperiodic, or 1/f, component. This post is a plain-language tour of what it is and why a working anesthesia practitioner should care.

    📌 In one line: The downward tilt of the EEG power spectrum is not noise. It may be a readout of the brain’s excitation–inhibition balance, the very thing our hypnotics are designed to shift.

    📈 Two signals in one trace

    Take a stretch of frontal EEG and plot its power against frequency on a log-log scale, and two things jump out.

    First, there are peaks: oscillations at specific frequencies, like the frontal alpha that appears under propofol or sevoflurane. These are the “periodic” parts, the rhythms we already talk about.

    Second, and easy to overlook, the whole spectrum slopes downward from left to right. There is always more power at low frequencies than at high ones, and the line connecting them is roughly straight on a log-log plot. That tilt is the aperiodic component, and its steepness is the aperiodic slope (often reported as the “exponent”).

    The key insight of the last few years is that this slope is a physiologically meaningful quantity, not something to throw away. Modern tools, of which FOOOF (now called specparam) is the best known, deliberately separate the peaks from the slope so each can be measured on its own. Older band-power indices never fully solved this: a change in the slope can masquerade as a change in alpha or delta power, and vice versa.

    ⚖️ Why the slope tracks the brain’s balance

    Computational and animal work suggests the aperiodic slope is a window onto the brain’s excitation–inhibition (E/I) balance: the tug-of-war between glutamatergic excitation and GABAergic inhibition. When inhibition dominates, the spectrum steepens and power falls off faster toward high frequencies. When excitation dominates, it flattens.

    Now think about what our drugs do. Propofol and the volatiles are, broadly, GABA-A potentiators: they tilt the brain toward inhibition. Ketamine, an NMDA antagonist, tilts it toward disinhibition. The aperiodic slope moves in the direction this pharmacology predicts. It steepens as we deepen GABAergic anesthesia and behaves differently under dissociatives.

    In other words, the slope is arguably a more mechanistic readout of what the anesthetic is doing to cortical dynamics than a proprietary index that blends everything into one dimensionless score.

    🏥 What this could change at the head of the table

    It is not on your monitor yet, but there are three practical reasons to care.

    • It may explain individual sensitivity. Two patients at the same MAC, or the same effect-site propofol concentration, can sit in very different cortical states. The slope captures some of that variation in how strongly a given dose shifts the brain toward inhibition, and could flag the patient who is being driven deeper than the number suggests.
    • It sharpens the burst-suppression conversation. We increasingly worry that excessive suppression in older patients is linked to postoperative delirium. A single index can lag behind, or under-report, suppression. A slope read from the raw trace is one more way to quantify how far toward profound inhibition we have pushed a vulnerable brain.
    • It is monitor-agnostic and cheap. It needs only a raw frontal EEG and open-source analysis, with no new hardware. That makes it attractive for research and, eventually, for bedside decision support.

    ⚠️ The honest caveats

    This is an emerging tool, not a validated bedside target, and it would be irresponsible to pretend otherwise.

    • The E/I link is inferred, not directly measured in patients.
    • Dose and sensitivity are tangled. The anesthetic itself steepens the slope, so separating a patient’s intrinsic sensitivity from the drug effect needs careful, dose-normalised analysis.
    • Method moves the number. Electrode montage, filtering, the frequency range fitted and the fitting method all change the result, so values are not yet comparable across setups without a harmonised protocol.
    • No outcome data yet. No trial has shown that titrating to the slope improves a clinical outcome such as delirium.

    The right stance today is curiosity with discipline: understand the parameter, read the raw EEG with it in mind, and watch the literature.

    🧾 Take-home

    The number on your depth monitor is a summary, and like all summaries it discards information. The aperiodic slope is a reminder that the “boring” background of the EEG encodes something real: the balance between excitation and inhibition that our drugs exist to manipulate.

    You do not need to change your practice tomorrow. But the next time you glance at the raw trace beneath the index, remember that its overall tilt is telling a story about the brain’s state that a single number cannot.

    This post is intended for clinical education and discussion among anesthesia professionals. It describes an emerging research parameter, not a validated monitoring target or a standard of care.

  • 🧠 LAIS: A Stroke Neuroprotectant Finally Works? Read the mRS Table Before You Celebrate

    🧠 LAIS: A Stroke Neuroprotectant Finally Works? Read the mRS Table Before You Celebrate

    Visual abstract above: Li S, Feng B, He D, et al. JAMA. 2026;336(18):1612-1620. © American Medical Association. Open full-size image.

    Neuroprotection is the graveyard of stroke research. Over a thousand agents have protected neurons in the laboratory, and almost none have survived a large human trial. Even nerinetide, the PSD-95 inhibitor that looked so promising, stumbled in ESCAPE-NA1 (cancelled out by alteplase) and then failed outright in ESCAPE-NEXT.

    So when a phase 3 trial in JAMA reports that a neuroprotective drug improved functional outcome, and that it did so even when given more than a day after stroke onset, it deserves attention. It also deserves a careful look at the numbers. That trial is LAIS, testing a new molecule called loberamisal.

    📌 Bottom line: In moderate ischaemic stroke treated within 48 hours (median 25 hours), a 10-day course of intravenous loberamisal raised the rate of mRS 0–1 at 90 days from 56.3% to 69.7% (NNT about 8), with no safety signal. But mRS 0–2, NIHSS improvement and Barthel Index were the same in both groups. The benefit sits almost entirely at the boundary between “slight disability” and “no significant disability”. Promising, not yet practice-changing.

    💊 What is loberamisal?

    Loberamisal is a small molecule with two targets, chosen to hit excitotoxicity from both sides:

    • Turning down excitation: ischaemia over-activates NMDA receptors, which pulls neuronal nitric oxide synthase (nNOS) onto the scaffolding protein PSD-95. That coupling drives nitric oxide–mediated injury. Loberamisal uncouples PSD-95 from nNOS, the same target as nerinetide, but by a different molecule.
    • Turning up inhibition: it is a positive allosteric modulator of the α2 subunit-containing GABAA receptor, restoring the inhibitory tone that ischaemia erodes.
    • Beyond the acute hit: in preclinical work it also suppresses myeloperoxidase-driven neuroinflammation, and sustained PSD-95–nNOS uncoupling appears to favour neurogenesis and synaptic remodelling. The authors call it “neuroprotective and neuroreparative”, which matters for the late treatment window.

    👥 Who was studied

    • 998 patients randomised at 32 hospitals in China (July to December 2024); 997 analysed
    • Age 18 to 80 years, NIHSS 7 to 20, no prestroke disability (mRS 0–1), within 48 hours of onset
    • Median age 64; two thirds men; median NIHSS 8, so mostly moderate strokes
    • Median time from onset to first dose about 25 hours; more than half were treated after 24 hours
    • Thrombolysis allowed (about 17% received it); thrombectomy patients were excluded
    • Double-blind, placebo-controlled phase 3

    💉 What was compared

    • Loberamisal: 40 mg IV over 60 minutes, once daily for 10 days, plus standard stroke care
    • Placebo: matching infusion for 10 days, plus standard stroke care

    The primary outcome was a “full functional outcome”, mRS 0–1 at 90 days. Edaravone and edaravone-dexborneol were not permitted.

    📊 What they found

    OutcomeLoberamisal (n = 502)Placebo (n = 495)
    mRS 0–1 at 90 days (primary)350 (69.7%)279 (56.3%)
    Sliding dichotomy mRS at 90 days49.0%39.8%
    NIHSS improvement ≥4 at day 1044.7%49.0%
    NIHSS improvement ≥4 at day 3073.5%72.7%
    Barthel Index ≥95 at 90 days70.7%70.4%
    Serious adverse events8.6%10.7%
    Death1.2%2.0%
    Primary outcome: relative risk 1.24 (95% CI 1.12 to 1.36); risk difference 13.28% (95% CI 7.24% to 19.32%). No heterogeneity across prespecified subgroups, including time to treatment and thrombolysis.

    The safety profile was clean. Adverse event rates were nearly identical, there was no excess haemorrhagic transformation in thrombolysed patients, and no unexpected serious reactions were reported.

    🔎 Now look at where the benefit sits

    This is the part worth slowing down for. Here is the full 90-day mRS distribution:

    mRS at 90 daysLoberamisalPlacebo
    024.5%22.7%
    146.1%35.1%
    27.8%21.1%
    314.1%12.6%
    45.7%5.7%
    50.4%0.4%
    6 (death)1.3%2.5%
    mRS 0–2 (adds up)78.4%78.9%
    Percentages from Table 2 of the paper (imputed data). The mRS 0–2 row is my own addition of the rows above.

    Almost the entire treatment effect is a shift of about 13% of patients from mRS 2 to mRS 1. The proportion who were independent (mRS 0–2) is the same in both arms, and the proportion left with moderate-to-severe disability (mRS 3–5) is also the same. The usual ordinal shift analysis could not even be run, because the proportional odds assumption failed. That is exactly what you expect when an effect lives at one cut-point and nowhere else.

    Why does this matter? The difference between mRS 1 (“no significant disability, able to carry out usual activities”) and mRS 2 (“slight disability, cannot do everything as before”) is the softest boundary on the scale. It rests on how the patient describes their own life, and here it was assessed with a simplified questionnaire, often by telephone. Inter-rater agreement is weakest exactly at this boundary. Meanwhile, the more objective measures (NIHSS change at 10 and 30 days, Barthel Index ≥95) did not move at all.

    To be fair, the trial was blinded, so a measurement artefact should affect both arms equally. A real effect on fatigue, mood, cognition or confidence, which mRS 1 versus 2 can capture and NIHSS cannot, is possible. The exploratory depression and anxiety numbers leaned the right way (severe depression 4.9% vs 7.8%), though not significantly, and that fits an α2-GABAA mechanism. But a single-boundary shift with no support from other scales needs replication before it is believed.

    ⚠️ Other things to keep in mind

    • Late treatment window: at a median of 25 hours, there is little penumbra left to “protect”. If the drug works, it is probably working on inflammation and recovery rather than classical neuroprotection.
    • Narrow population: moderate strokes (median NIHSS 8), no thrombectomy, few thrombolysed patients, China only. Large-vessel occlusions treated with thrombectomy, the group that sees anaesthetists, were excluded.
    • Effect larger than planned: the trial was powered for a 10.8-point difference and found 13.4. Big early effects in single-country trials often shrink with replication.
    • Funding: supported by a national science grant and by the drug’s manufacturer, NeuroDawn Pharmaceutical. The academic team analysed the locked data independently, which is reassuring.
    • No imaging or biomarker data: nothing tells us how, or in whom, the drug worked.

    🏥 What it means for the neuro-OR and neuro-ICU team

    For now, very little changes at the bedside. Loberamisal is not available outside trials, and the patients we meet most, thrombectomy cases under GA or sedation, were not studied. The questions for us are the next ones: will it help after successful recanalisation, as a true “reperfusion plus protection” strategy? And does a GABAA modulator given daily for 10 days interact with our sedatives in the neuro-ICU? The trial does not tell us.

    The bigger lesson is conceptual. LAIS adds to a growing idea that the window for helping the injured brain may be days, not hours, if the target is inflammation and repair rather than the dying penumbra. That is a field neurocritical care should watch closely.

    🧾 Take-home

    LAIS is the most positive phase 3 neuroprotection result in years, with a clean safety profile and a striking headline number. But the benefit is concentrated in moving patients from mRS 2 to mRS 1, while independence, neurological recovery and daily-living scores were unchanged. Treat it as a strong signal that needs a second trial, ideally outside China and in thrombectomy patients, before it changes how we manage stroke.

    Reference: Li S, Feng B, He D, et al; LAIS Investigators. Loberamisal for acute ischemic stroke: the LAIS randomized clinical trial. JAMA. 2026;336(18):1612-1620. doi:10.1001/jama.2026.16557. ClinicalTrials.gov NCT06517173. See also the accompanying editorial by Jeffrey L. Saver, “Neuroprotective Therapy for Acute Ischemic Stroke”, in the same issue.

    For education only. Follow your local stroke and anaesthesia protocols.

  • 🧠 When Raising CPP Won’t Help: The Type 2 Picture on Cerebral Microdialysis

    🧠 When Raising CPP Won’t Help: The Type 2 Picture on Cerebral Microdialysis

    Your ICP is 14. Your CPP is sitting where the autoregulation index says it should. PbtO2 is comfortable. And the lactate/pyruvate ratio from the microdialysis catheter has been climbing all night.

    This is the moment microdialysis earns its place, because it is telling you something no pressure-based monitor can. The question is what to do next — and the instinct to push the pressure up is often wrong.

    📌 The hook: High LPR with low pyruvate means no oxygen. High LPR with normal or high pyruvate means sick mitochondria. The first responds to flow and oxygen. The second does not.

    🔬 What the catheter actually measures

    A 20 kDa membrane perfused at 0.3 µL/min, giving roughly 70% recovery, with vials collected hourly. Three properties follow from that, and all three matter at the bedside: the readout is focal, it is delayed, and it is labour-intensive. You are sampling one small volume of brain, an hour behind events.

    📊 Thresholds worth knowing

    From the 2014 international consensus statement:

    AnalyteThreshold
    Lactate/pyruvate ratio> 25 abnormal; > 40 metabolic crisis
    Brain glucose< 0.8 mmol/L low; < 0.2 mmol/L critical
    GlutamateAdjunct — excitotoxicity, seizures
    GlycerolAdjunct — membrane breakdown

    🎯 Where the catheter sits decides what it means

    The tip belongs in at-risk tissue: pericontusional brain, or territory prone to vasospasm. Not in the dead core, where the numbers are meaningless, and not in distant normal brain, where they are reassuring and irrelevant. Confirm the gold tip on CT before you believe anything the analyser tells you.

    🩺 Type 1 or Type 2?

    Type 1, ischaemic. High LPR with low pyruvate. Oxygen or flow is not arriving. This one you treat with delivery.

    Type 2, non-ischaemic. High LPR with normal or high pyruvate. Substrate is arriving; the cell cannot use it. Raising CPP here buys you nothing, and the oedema and cardiac cost of chasing pressure are real. The rest of this post is about Type 2.

    1️⃣ Exclude hidden ischaemia first

    Mixed pictures exist, and a focal catheter can miss a regional supply problem. Before you accept a mitochondrial label, go back through the delivery chain: PbtO2, CPP measured against the autoregulation-guided optimum rather than a protocol number, ICP, haemoglobin, PaO2 and PaCO2.

    If you find a supply deficit, fix it regardless. It is cheap, and it is the safe default when you are not sure which picture you are looking at.

    2️⃣ Hunt for the cause of the metabolic stress

    • Seizures or non-convulsive status. Get continuous EEG. A high LPR with high glutamate points here, and non-convulsive status is easy to miss in a sedated patient.
    • Fever or hyperthermia. Treat to normothermia to cut demand.
    • Glucose. Read brain glucose alongside the LPR, never alone. Low brain glucose with tight systemic control is neuroglycopenia, and the answer is to liberalise.
    • Sepsis, or a mitochondrial toxin. This includes propofol above 4 mg/kg/h — the propofol infusion syndrome link. Reduce the dose and change agent.
    • Catecholamine- and glutamate-driven hyperglycolysis. This raises lactate without any true mitochondrial failure, and it is worth recognising before you escalate.

    3️⃣ Cut demand, support the substrate

    • Deepen sedation, or use a barbiturate if the ICP situation allows it.
    • Avoid tight glycaemic control. Aim for roughly 140–180 mg/dL, following the NICE-SUGAR logic. The injured brain tolerates tight control badly.
    • Start nutrition early.

    4️⃣ Experimental options — name them, and say so

    In a viva, knowing these exist is worth marks. Claiming they are standard is not.

    • Hypertonic sodium lactate as an alternative fuel (Ichai, Bouzat).
    • Normobaric hyperoxia. Small studies improved the LPR, but PET showed no gain in CMRO2. The evidence is genuinely mixed.
    • Cyclosporin, as a blocker of the mitochondrial permeability transition pore.

    🗣️ Viva essentials

    Q. What do the lactate/pyruvate ratio and glucose tell you that ICP cannot, and why must the catheter not sit in the contusion core?

    Say: LPR > 25 and > 40 · type 1 versus type 2 · penumbral placement · 0.3 µL/min. ICP and CPP describe the plumbing; microdialysis describes whether the cell is actually making ATP. In dead core tissue there is no metabolism left to measure, so the numbers mean nothing.

    Probe: how does tight insulin therapy show up? As falling brain glucose with a rising LPR — neuroglycopenia in a patient whose blood sugar looks respectable.

    🧾 Take-home

    Pyruvate is the discriminator. Low pyruvate sends you to the delivery chain; normal pyruvate sends you looking for seizures, fever, sepsis, propofol and tight glycaemic control. The temptation with a rising LPR is always to push the pressure higher. In the Type 2 picture that is effort spent on the wrong organ system.

    For education only. Follow your local neurocritical care protocols.

  • 🧠 The Sleep Switch: Why a Patient on Dexmedetomidine Wakes When You Call Them

    A patient on a dexmedetomidine infusion looks asleep. You say their name. They open their eyes, squeeze your hand, answer a question, and drift off again. Nothing else in the sedation cupboard behaves like that.

    The reason is anatomical. Dexmedetomidine does not blanket the cortex the way propofol does. It flips a switch the brain already owns.

    📌 In one line: Dexmedetomidine silences the locus coeruleus, which releases the VLPO from inhibition and tips the sleep–wake flip-flop towards sleep. Everything clinically distinctive about the drug follows from that one step.

    🔀 The sleep switch

    The VLPO — ventrolateral preoptic nucleus — is a small cluster of GABA and galanin neurons in the anterior hypothalamus. It is the brain’s off switch for wakefulness.

    It sends inhibitory projections to the arousal nuclei: the locus coeruleus (noradrenaline), the tuberomammillary nucleus (histamine), the raphe nuclei (serotonin), and the orexinergic neurons of the lateral hypothalamus. Those nuclei inhibit the VLPO straight back.

    Mutual inhibition of that kind makes a flip-flop circuit: two stable states, awake or asleep, with fast transitions between them and no comfortable middle ground. That is the design intent. Nobody benefits from being half asleep while crossing a road.

    💉 Where dexmedetomidine gets in

    Dexmedetomidine is an alpha-2A agonist, and the locus coeruleus is dense with those receptors. Agonism there cuts noradrenergic output. The LC falls quiet, the VLPO is released from inhibition, and the switch tips towards the sleep state.

    In other words, dexmedetomidine borrows an endogenous pathway. Propofol and the volatile agents do something quite different: they act directly on GABA-A receptors across the cortex and thalamus — a broader and far less physiological suppression.

    That is why dexmedetomidine sedation resembles natural stage-2 NREM sleep rather than GABAergic hypnosis, and why a sufficient stimulus can flip the switch back. The rousable patient is not a quirk of dosing. It is the mechanism showing itself.

    📈 What it means on the EEG

    A stage-2 NREM-like state has stage-2 features: sleep spindles and K-complexes, not the deep delta slowing or burst suppression of heavy GABAergic hypnosis.

    The practical consequence is that interictal epileptiform discharges stay visible. This is what makes dexmedetomidine suited to MRI-EEG in epilepsy surgery planning, where propofol risks suppressing the very activity you came to record. The same property lets you rouse the patient for a neurological check and then allow them to settle again — useful in awake craniotomy and in the neuro-ICU.

    🎯 The drug needs an intact target

    If the mechanism depends on switching off the locus coeruleus, then the LC has to be there to switch off. It sits in the dorsal pontine tegmentum, close to the floor of the fourth ventricle — a crowded neighbourhood, and one we operate in.

    • Pontine infarct or haemorrhage
    • Surgery on the floor of the fourth ventricle
    • Brainstem traumatic injury
    • Infiltrative lesions, such as diffuse midline glioma, H3 K27-altered (formerly DIPG)

    The two you will actually meet are the fourth-ventricle tumour resection and the pontine stroke. The glioma is an instructive example rather than a common anaesthetic encounter. Where the target is damaged, the sedation you predicted may simply not arrive — and the answer is a different agent, not a larger dose.

    🌙 REM rebound and the sleep apnoea patient

    REM is the respiratorily unsafe stage of sleep: atonia of the pharyngeal dilators and of the accessory muscles. Opioids and volatile agents suppress REM during and immediately after surgery, and that debt is repaid later as REM rebound, classically on postoperative nights 2 to 4.

    That is precisely when the patient with obstructive sleep apnoea desaturates — well after the recovery unit, usually on a general ward, often unmonitored, and frequently on the night everyone has stopped worrying.

    Dexmedetomidine suppresses REM far less than the GABAergic sedatives do. Together with its minimal respiratory depression, that is a substantial part of the argument for it in this group.

    🗣️ A viva probe

    Q. Your patient on dexmedetomidine looks asleep but obeys commands. Which sleep stage does that resemble, and what does it tell you about the EEG?

    A. Stage-2 NREM, with spindles on the trace. The practical point is that epileptiform activity stays readable, which is why the drug suits MRI-EEG in epilepsy work-up.

    🧾 Take-home

    Propofol imposes unconsciousness on the cortex. Dexmedetomidine persuades the hypothalamus to do it for you. The rousable patient, the readable EEG, the preserved REM and the occasional disappointing response after a pontine lesion are all the same fact, seen from four directions.

    For education only. Follow your local sedation and anaesthesia protocols.

  • 🧠 ATTENTION-LATE: Does Tenecteplase Before Thrombectomy Help in Late Basilar Artery Occlusion?

    🧠 ATTENTION-LATE: Does Tenecteplase Before Thrombectomy Help in Late Basilar Artery Occlusion?

    Visual abstract above: Li R, Yao X, Nogueira RG, et al. JAMA. 2026;336(17):1529-1538. © American Medical Association. Open full-size image.

    The basilar artery is the most unforgiving territory in stroke medicine. Left untreated, a basilar artery occlusion kills or severely disables most patients. Thrombectomy changed that story: the ATTENTION trial showed that pulling the clot out nearly doubled the chance of a good outcome (46% vs 23%). Tenecteplase, meanwhile, has been pushing the clot-busting window towards 24 hours in some anterior circulation trials.

    So the obvious next question: when a patient with a basilar occlusion reaches a thrombectomy centre late, should we give tenecteplase on the way to the angio suite? ATTENTION-LATE, published in JAMA on 23 September 2026, gives a clear answer.

    📌 Bottom line: In basilar artery occlusion presenting 4.5 to 24 hours after onset directly to a thrombectomy-capable centre, adding tenecteplase before thrombectomy did not improve 90-day functional independence. About 30% of patients were independent at 90 days in both groups.

    🔬 The question

    Does intravenous tenecteplase before endovascular thrombectomy (EVT) improve outcomes in basilar artery occlusion when patients arrive between 4.5 and 24 hours after stroke onset, compared with thrombectomy alone?

    👥 Who was studied

    • 332 patients randomised at 40 thrombectomy-capable stroke centres in China (March 2023 to February 2025); 330 included and 329 completed follow-up
    • Adults with moderate to severe basilar artery occlusion (NIHSS 10 or more) arriving 4.5 to 24 hours after onset
    • Mean age 66 years; about three quarters were men
    • Open-label design with blinded assessment of outcomes

    💉 What was compared

    • Tenecteplase + EVT: a single intravenous bolus of 0.25 mg/kg (maximum 25 mg), then thrombectomy
    • EVT alone: thrombectomy within 24 hours of onset

    The primary outcome was functional independence at 90 days, defined as a modified Rankin Scale score of 0 to 2.

    📊 What they found

    OutcomeTenecteplase + EVTEVT alone
    Functional independence (mRS 0–2) at 90 days50/165 (30.3%)50/164 (30.5%)
    Symptomatic intracranial haemorrhage5%4%
    Death at 90 days40%43%
    Adjusted rate ratio for functional independence 0.92 (95% CI 0.67 to 1.25); risk difference −0.18% (95% CI −10.13% to 9.76%).

    The drug was not clearly harmful: bleeding and death were similar in both groups. It simply added nothing.

    🤔 Why didn’t the clot-buster help?

    The trial does not answer this directly, but several explanations are plausible:

    • Thrombectomy does the heavy lifting. When the catheter reaches the clot soon after arrival, the drug has little time to work first.
    • Old clots resist lysis. Many hours after onset, a thrombus tends to be more organised and less responsive to a fibrinolytic.
    • Atherosclerosis is common in basilar occlusion, particularly in Asian populations. A fibrinolytic cannot open a fixed stenosis.
    • It fits earlier evidence. In TIMELESS, late-window tenecteplase also showed no benefit when most patients went on to thrombectomy.

    ⚠️ Read with caution

    • Wide confidence interval: the result cannot rule out a modest benefit or harm of up to about 10 percentage points either way.
    • Open-label: clinicians knew who received the drug, though outcomes were assessed blind.
    • China only: the mix of stroke causes, especially intracranial atherosclerosis, may differ elsewhere.
    • Direct arrivals only: patients came straight to thrombectomy centres. The trial does not tell us about “drip and ship” patients who face long transfer delays from a peripheral hospital.
    • Severe strokes: 90-day mortality of about 40% reflects how sick these patients were.

    🏥 What it means for the neuro-OR and neuro-ICU team

    For a late-presenting basilar occlusion arriving at a thrombectomy centre, the priority is getting to the groin quickly, not waiting for lysis. Skipping the drug also removes a layer of complexity for the anaesthesia team, such as post-lytic caution around arterial and central line placement. The decisions where we can still make a difference remain the same: general anaesthesia versus conscious sedation, airway protection in a patient with brainstem signs, and blood pressure management before and after recanalisation.

    🧾 Take-home

    In late basilar artery occlusion at a thrombectomy-capable centre, go straight for the clot. Tenecteplase on the way in adds cost and complexity without measurable benefit.

    Reference: Li R, Yao X, Nogueira RG, et al; ATTENTION LATE Investigators. Tenecteplase before thrombectomy at 4.5 to 24 hours for basilar artery occlusion: the ATTENTION LATE randomized clinical trial. JAMA. 2026;336(17):1529-1538. doi:10.1001/jama.2026.15496. ClinicalTrials.gov NCT05701956.

    For education only. Follow your local stroke and anaesthesia protocols.

  • 🧠 The Skull Is Not a Sealed Box: How AI Is Learning to Read Brain Pressure Without a Drill

    🧠 The Skull Is Not a Sealed Box: How AI Is Learning to Read Brain Pressure Without a Drill

    For decades, measuring intracranial pressure (ICP) meant one thing: putting a hole in the skull. A bolt, a catheter, a strain gauge pushed into brain tissue — the gold standard, but one that carries real cost. A new frontier is quietly rewriting that rule, and it rests on a surprising fact of physiology.

    🔬 The Principle: A Cranium That Breathes

    The adult skull is not perfectly rigid. With every heartbeat and breath, the bone expands by fractions of a micron — movement mechanically coupled to the pressure–volume state inside. A cranial extensometer (such as the brain4care / B4C sensor) captures these nanometre-scale displacements and turns them into a pulse waveform. The shape of that wave — its P2/P1 amplitude ratio, its time-to-peak, its harmonics — encodes intracranial compliance.

    The leap is conceptual: instead of measuring pressure, machine-learning models now infer it from waveform morphology. Pressure from shape.

    📊 The Evidence So Far

    A 2025 proof-of-concept study in npj Digital Medicine trained and tested this approach on 112 neurocritical patients with acute brain injury (92 for training/testing, 20 for independent validation).

    The results:

    ✅ Mean absolute error of 3.00 mmHg
    ✅ 95% confidence interval within ±7.5 mmHg
    ✅ Roughly 72% of validation estimates within 0–4 mmHg of invasive readings

    A separate multicentre evaluation confirmed the system can detect intracranial hypertension, not just estimate a number.

    💡 Why It Matters

    A non-penetrating sensor removes the great fears of invasive monitoring — haemorrhage, infection, ventriculitis, malposition. It opens the door to monitoring the coagulopathic patient, the pre-hospital and resource-limited setting, and all the patients for whom a bolt could never be justified: mild TBI, hepatic encephalopathy, the quietly deteriorating stroke.

    ⚠️ The Honest Caveats

    A rigorous examiner will press here, and rightly so. A ±7.5 mmHg band is wide when the treatment threshold sits at 20–22 mmHg — error of that size straddles the treat-or-watch decision. The models lean on small, single-technology cohorts, vulnerable to shifts in skull thickness, age, craniectomy status, and sensor placement. And the output is still a relative, trend signal anchored to population priors — not a patient-specific, zeroed absolute pressure, and certainly not a drain you can open.

    So its true place today is as a triage and compliance-trending adjunct — not a replacement for the invasive monitor in a patient already committed to ICP-directed therapy.

    🔭 The Frontier Ahead

    The real story is convergence. Waveform-based non-invasive ICP is merging with quantitative pupillometry (the Neurological Pupil index, now in the multicentre ORANGE trial) and optic nerve sheath diameter into a single, sensor-fused read of the intracranial state.

    The defensible claim is not that machine learning will retire the strain gauge. It is that it will expand the monitored population to patients currently invisible to us — and that regulatory-grade adoption now hinges on large, multi-device, externally validated trials reporting Bland–Altman limits of agreement and threshold-level sensitivity, not correlation coefficients alone.

    The skull was never truly silent. We are only now building the instruments to listen. 🩺

    📚 Sources
    brain4care ML validation — npj Digital Medicine (2025): https://www.nature.com/articles/s41746-025-01463-y
    Brain4care clinical performance — PubMed: https://pubmed.ncbi.nlm.nih.gov/40293697/
    Pupillometry / ORANGE — PMC: https://pmc.ncbi.nlm.nih.gov/articles/PMC10791071/

  • 🧠 The Stroke Surgery Trial That’s Forcing Neurosurgeons To Rethink The Rulebook

    🧠 The Stroke Surgery Trial That’s Forcing Neurosurgeons To Rethink The Rulebook

    For decades, surgery for brain hemorrhage carried a quiet stigma. Two landmark trials had already tried to prove it works and came up empty. In 2024, a new trial finally cracked the code, but only for a specific type of patient. Here is what changed, and why it matters.

    🩸 THE PROBLEM

    Spontaneous intracerebral hemorrhage, a stroke caused by bleeding directly into the brain rather than a clot blocking a vessel, kills roughly 40 percent of patients within a month. Only 10 to 25 percent ever regain functional independence. Removing the clot surgically seems obvious. Getting the evidence to support it has been anything but.

    📉 WHY SURGEONS HAD GIVEN UP HOPE

    Two major trials tried and failed to prove surgery helps.

    STICH I, published in 2005, randomized patients with spontaneous ICH to early surgical evacuation versus initial medical management. Result: no significant difference in outcomes. Surgery did not clearly help.

    STICH II, published in 2013, narrowed the focus to patients with more favorable, superficial lobar hemorrhages, betting that a more selective population would reveal a benefit. It still came up short on its primary outcome.

    Together, these trials cemented over a decade of caution. The 2022 AHA and ASA guidelines reflected that uncertainty, and blood pressure control rather than surgery became the dominant lever for improving ICH outcomes, guided by trials like INTERACT2 and ATACH-2, which tested how aggressively to lower blood pressure after a bleed. INTERACT2 suggested intensive blood pressure lowering was safe and possibly beneficial. ATACH-2 later found that pushing blood pressure even lower than standard targets added risk without added benefit. Between them, these trials shaped the medical management arm that every surgical trial since has had to compete against.

    ⚡ ENTER ENRICH

    Published in the New England Journal of Medicine in 2024, ENRICH is the trial that finally moved the needle. It tested a newer, less invasive surgical technique called minimally invasive trans-sulcal parafascicular surgery, MIPS for short, which uses a tubular retractor to reach and remove the clot while minimizing damage to surrounding brain tissue, a meaningfully gentler approach than the open craniotomies used in earlier trials.

    300 patients across 37 US centers Hemorrhage volume 30 to 80 mL Surgery within 24 hours of onset Primary outcome: a quality of life weighted disability score at 180 days

    📊 THE NUMBERS

    Surgery group score: 0.458 Medical management group score: 0.374 Difference: 0.084, with 98.1 percent Bayesian probability that surgery was better

    The benefit was not evenly spread. In lobar hemorrhages, the more superficial bleeds, the difference jumped to 0.127. In deep basal ganglia hemorrhages, there was no benefit at all, and enrollment for that subgroup was stopped early because it clearly was not working.

    Thirty day mortality told a similarly stark story. 9.3 percent died in the surgery group versus 18.0 percent in the medical management group, a difference researchers were 98.7 percent confident was real. Rebleeding after surgery, the complication everyone worries about, occurred in just 3.3 percent of surgical patients.

    🔬 WHY THIS ACTUALLY MAKES BIOLOGICAL SENSE

    The logic is straightforward once you see it. A blood clot sitting in brain tissue is not inert. It swells the surrounding area, raises pressure inside the skull, and releases breakdown products that are toxic to nearby neurons. The longer it sits, the more damage accumulates. Removing it early, through a corridor that avoids tearing through critical white matter tracts, interrupts that injury cascade before it becomes irreversible.

    🏥 WHAT THIS MEANS IN PRACTICE

    ENRICH does not say surgery works for everyone with a brain bleed. It says surgery works for a specific, well defined group, lobar location, right size window, right technique, right speed. For hospital teams, especially anesthesia and critical care staff, that translates into a much tighter operational demand: fast coagulopathy reversal, disciplined blood pressure control going into the OR, and readiness to move within hours rather than days.

    The open questions are exactly what you would expect. Does this benefit hold up in younger patients, in patients with bleeding disorders, or in deep hemorrhages using a different technique. Those answers are still pending, and they are precisely where the next round of trials is headed.

  • Rewriting the Rules for Brain Bleeds: What INTERACT3 Taught Us

    Spontaneous intracerebral haemorrhage has long carried an air of clinical resignation, the diagnosis that quietly lowered everyone’s expectations before treatment had even begun. INTERACT3, published in The Lancet in May 2023, unsettled that fatalism. It changed not a drug but a way of thinking, and it has since become impossible to discuss acute haemorrhagic stroke without reference to it.

    The Trial in Brief

    INTERACT3 was a pragmatic, international stepped-wedge cluster-randomised trial enrolling 7,036 patients across 121 hospitals, most of them in low- and middle-income countries. Instead of testing a single molecule, it asked whether a coordinated care bundle, begun within hours of the bleed, could alter the course of the illness. The bundle drew together four physiologically targeted actions: intensive systolic blood pressure lowering to a target below 140 mmHg; strict glucose control (6.1–7.8 mmol/L in patients without diabetes, 7.8–10.0 mmol/L in those with it); fever control holding temperature at or below 37.5°C; and rapid reversal of anticoagulation to an INR below 1.5.

    What the Numbers Showed

    The primary outcome was functional recovery, measured by a modified Rankin Scale shift analysis at six months. The bundle came out ahead, with a common odds ratio of 0.86 (95% CI 0.76–0.97; p=0.015). The advantage extended beyond function. Mortality fell from 17.0% to 14.1%, and serious adverse events dropped from 20.1% to 16.0%. In a disease so often treated with a shrug, these are shifts of real consequence.

    The Story of Secondary Injury

    The physiology explains why. Outcome after an intracerebral haemorrhage is shaped less by the original bleed than by the secondary injury that follows, and each element of the bundle disarms a different, time-sensitive part of that cascade.

    The first threat is early haematoma expansion, which occurs in roughly a third of patients within the opening hours and responds to blood pressure. That is the rationale for rapid, controlled lowering of systolic pressure. The trial’s smooth, protocol-guided titration also mattered: it likely spared patients the falls in renal and cerebral perfusion that may have blunted the effect in earlier blood-pressure trials such as ATACH-2.

    Alongside pressure sit hyperglycaemia and fever, each an independent driver of perihaematomal oedema, oxidative stress, and a faltering penumbral metabolism. Anticoagulation-associated bleeds carry the highest expansion risk of all, which places rapid reversal among the most valuable components of the bundle.

    The deeper lesson runs beneath the individual parts. None of these interventions had previously shown a convincing functional benefit on its own. Delivered together, early and in concert, they moved patients measurably towards independence. The whole outperformed the sum of its parts.

    From Nihilism to a Protocol Worth Running

    This reframes the disease. Intracerebral haemorrhage moves from a condition met with therapeutic nihilism to an actionable, protocol-driven emergency, one that sits naturally beside the bundled approaches already familiar from sepsis and STEMI care. For the neuroanaesthesiologist and neurointensivist, the implication is concrete: systems-level implementation — standing order sets, rapid-reversal pathways, and pre-specified physiological targets switched on in the emergency department and carried through into the operating room — achieves what isolated pharmacological effort has not.

    Reading the Evidence Honestly

    Two caveats temper the enthusiasm. The stepped-wedge design is vulnerable to temporal confounding and cannot, by its structure, isolate which single element carried the benefit, which means the bundle should be adopted as a complete package rather than dismantled into favoured parts. And because most patients were recruited in lower-resource settings, questions of external validity and generalisability deserve open acknowledgement rather than quiet assumption.

    These limits notwithstanding, INTERACT3 is well positioned to shape forthcoming AHA/ASA and European Stroke Organisation guidance. It establishes early, multimodal physiological control as the emerging standard of care in acute intracerebral haemorrhage.

    Sources: INTERACT3, The Lancet (2023) · INTERACT3 full text, PMC

  • Reading the Pupil Anew: Automated Pupillometry in Neurological Monitoring

    For over a century, the pupillary examination has rested on a simple ritual: a clinician, a penlight, and a judgement rendered in an instant — is the pupil brisk, sluggish, or unresponsive? It is among the most venerable of bedside assessments, and also among the least dependable. When clinicians examine the same patient, their agreement on whether the pupils are unequal falls below fifty percent, and the unaided eye is wholly incapable of detecting the subtle, incremental changes that so often precede deterioration.

    Automated pupillometry offers a more rigorous alternative. A calibrated flash of light is delivered, the pupil’s response recorded in fine detail, and the result distilled into a single value — the Neurological Pupil index (NPi), scaled from 0 to 5, with anything below 3 flagged as abnormal. In shining light and measuring the constriction that follows, the device tests the full pupillary light reflex — the afferent limb carried by the optic nerve (cranial nerve II) that senses the light, and the efferent limb carried by the parasympathetic fibres of the oculomotor nerve (cranial nerve III) that drives constriction. Its warning power rests chiefly on the efferent (CN III) pathway, which is exquisitely sensitive to rising intracranial pressure and is compressed early as the brain begins to herniate. A gradually falling NPi can therefore signal catastrophe hours before the dreaded “blown pupil” appears. Measured every four to six hours, it now helps anticipate dangerous swelling after stroke, head injury, and surgery, and — via the 2025 NPi-Connect system — feeds straight into the medical record as a trackable vital sign independent of who holds the light.

    The supporting evidence continues to accumulate. Automated pupillometry is now employed to anticipate dangerous brain swelling in stroke, with measurements taken every four to six hours to follow the trajectory rather than a single isolated reading. The same principle has proven valuable following traumatic brain injury and neurosurgery, where a falling NPi identifies patients on the verge of decline. In early 2025, a system known as NPi-Connect began transmitting these readings directly into the electronic medical record, establishing the NPi as a properly documented, trackable vital sign that no longer depends upon the individual performing the examination.

    The limitations, however, deserve candid acknowledgement. Commonly used sedatives and analgesics — opioids and dexmedetomidine among them — blunt the pupillary response, and recent work demonstrates that the depth of sedation can meaningfully distort the reading. Iris pigmentation, prior ocular surgery, direct orbital trauma, and certain neuromuscular conditions may all degrade its accuracy, and the measurement interrogates only a single nerve pathway, offering no insight into the wider brain. The four-to-six-hour intervals between readings likewise leave unmonitored windows.

    The most compelling horizon lies in estimating the pressure within the skull without breaching it at all. Emerging systems infer that pressure from the faint pulsatile expansion of the skull, from gentle microwave or near-infrared signals, or from ultrasound of the sheath surrounding the optic nerve. One such approach now falls within roughly 3 mmHg of the true value, achieving accuracy to within a few points in approximately 72 percent of cases. Combining these signals — the pupillary index, a non-invasive pressure estimate, and the optic nerve measurement — offers a credible route toward sparing many patients an invasive monitor placed through the skull. For the present, the invasive monitor retains its primacy, and the prudent posture is to regard pupillometry as a valuable early-warning companion rather than the sole arbiter of a swelling brain.

  • How Much Blood Does an Injured Brain Need? What do HEMOTION and TRAIN tell?

    The messages

    For years, ICUs treated anemia in brain-injured patients like anemia in anyone else: transfuse only when hemoglobin falls to about 70 g/L. Two large 2024 randomized trials HEMOTION (NEJM) and TRAIN (JAMA) have overturned that habit, marking the biggest shift in neurocritical-care transfusion practice in a decade.

    Why the injured brain is different

    Injured brain tissue has almost no reserve to pull extra oxygen from blood or raise blood flow in the vulnerable “penumbra” around the injury. So anemia isn’t just a lab number here it becomes a direct driver of secondary ischemic injury at hemoglobin levels the rest of the body tolerates. Both trials tested whether keeping hemoglobin higher protects the brain.

    HEMOTION: a strong signal, just short of significance

    Adults with moderate-to-severe traumatic brain injury and anemia were randomized to a liberal strategy (transfuse at Hb <100 g/L) or restrictive (<70 g/L), with unfavorable Glasgow Outcome Scale “Extended (GOS-E) at six months as the primary outcome. Unfavorable outcomes: 68.4% liberal vs 73.5% restrictive  a ~5-point improvement favoring liberal, but not statistically significant. A “negative” trial that still leans one way.

    TRAIN: larger, broader, clearly positive

    Bigger and wider  850 patients, 72 ICUs, 22 countries, including traumatic brain injury plus subarachnoid and intracerebral hemorrhage. Liberal trigger <9 g/dL vs restrictive <7 g/dL, GOS-E at 180 days. Liberal was clearly better: 62.6% vs 72.6% unfavorable, an absolute risk difference of 10.0%, with no excess clotting or lung injury.

    Do they really disagree? No.

    The trials are concordant in direction, differing only in power and breadth. HEMOTION’s confidence interval includes TRAIN’s result, and both favor transfusing sooner. The likely mechanism is better cerebral oxygen delivery at the microvascular level supported by benefit appearing in functional neurologic recovery rather than survival.

    What this means at the bedside

    Shift toward a liberal threshold of ~90-100 g/L, individualized by evidence of brain-tissue compromise (brain-tissue oxygen tension, an evolving infarct, vasospasm risk) rather than a rigid number. A 70 g/L default is now hard to defend in structural brain injury. Two caveats: neither trial showed a mortality benefit (the case rests on function), and both were underpowered for disease subgroups. The open frontier is titrating transfusion to real-time cerebral oximetry or brain-tissue oxygen monitoring rather than a single systemic haemoglobin value.