MNM on ECMO, embolus burden, aEEG continuity, and the silent stroke

1. Three patient vignettes#

Vignette A. Canonical school-age VA-ECMO post-arrest#

Daniyal, 6 years old, 22 kg. Out-of-hospital cardiac arrest from drowning; eCPR initiated in the emergency department after 35 minutes of CPR. VA-ECMO via the right femoral artery and vein. Day 3: stable circuit, ACT 180 seconds, aPTT 65, no bleeding. RASS −4, intermittent paralysis. Modalities: bilateral frontal NIRS pads, aEEG, intermittent TCD twice daily by a dedicated technologist, pupillometry every 4 hours. On day 3 morning the TCD technologist reports a cluster of 10 HITS on the right MCA over 1 hour (well above the usual background of 1 to 2 per hour on a stable circuit). aEEG continuity grade is unchanged. NIRS L 68 / R 60 (asymmetry approaching 8%, R lower). Pupils symmetric NPi 4.2 / 4.0. No clinical exam change (sedated). The question: clinically silent embolus burden or impending stroke? 1 2

Vignette B. Neonatal VV-ECMO for severe meconium aspiration#

Layla, term newborn, day of life 4, on VV-ECMO for severe persistent pulmonary hypertension after meconium aspiration. Right internal jugular dual-lumen cannula. Pulsatile native cardiac output preserved (PI on TCD ~ 0.8, lower than off-ECMO but present). Sedated and paralysed. Modalities: continuous aEEG (the neonatal mainstay), bilateral neonatal NIRS pads, daily head ultrasound (anterior fontanelle still open), intermittent TCD. Day 4 ultrasound shows a small right-sided intraventricular haemorrhage (grade II); aEEG remains continuous, normal voltage; NIRS symmetric. Anticoagulation target was aPTT 50 to 60; the team relaxes to 40 to 50 given the ICH and reassesses with a repeat ultrasound at 24 hours. The neonatal-specific point: ICH is the dominant neurological complication on neonatal ECMO and is often clinically silent; serial head ultrasound is the bedside primary surveillance tool, supplemented by aEEG continuity. 3 1

Vignette C. Atypical: silent left M1 stroke#

Yara, 4 years old, 18 kg. VA-ECMO day 5 for refractory cardiogenic shock from acute myocarditis. The circuit is stable; aPTT 70 seconds. Routine TCD shows a sudden absence of the left MCA signal (previously normal MFV 75 cm/s); the right MCA is unchanged at 70 cm/s. NIRS L 48% / R 64% (asymmetry 16%, well above any threshold). aEEG over the left hemisphere is suppressed compared with right. Pupillometry NPi 3.0 / 4.2 (subtle asymmetry, R higher). The team performs an urgent CT-angiography: left M1 occlusion. This is the silent stroke the modalities together caught despite a fully sedated, paralysed patient with no clinical exam to drift. Decision: discuss with neurointerventional; in this setting (anticoagulated, on circuit) thrombectomy is technically feasible but high-risk; the team chooses anticoagulation optimisation and supportive care given the timing (presumed > 6 hours from ictus). MRI day 7 confirms M1 territory infarct. The bedside lesson: the four-modality bundle (TCD + NIRS + aEEG + pupillometry) caught the stroke that the clinical exam could not see. 3

2. The clinical question#

In a child on ECMO with a clinically muted exam (sedation, paralysis, post-arrest), how do you detect evolving neurological injury before it becomes irreversible? The integration question is which combination of TCD, aEEG, NIRS, pupillometry, and serial exam catches the most neurological events at the bedside, and how the multimodal pattern guides the anticoagulation balance.

3. Pathophysiology refresher#

ECMO predisposes to neurological injury through several mechanisms. Embolic events arise from circuit-generated micro-emboli (gas bubbles, fibrin, platelet aggregates) and from the cannulation site (especially central cannulation of the carotid in neonates). Ischaemic stroke from low cerebral perfusion can develop during run-in or with circuit failure; in VA-ECMO with central or right common carotid cannulation, ipsilateral middle cerebral artery flow may be compromised. Haemorrhagic stroke results from systemic anticoagulation against a backdrop of consumptive coagulopathy and platelet dysfunction. Global hypoxic-ischaemic injury is common in eCPR survivors. Seizures occur in 10 to 20% of children on ECMO, often non-convulsive given paralysis. 1 3 4

VA-ECMO physiology and TCD. Fully-supported VA-ECMO produces a non-pulsatile arterial waveform; the native heart may eject little or nothing. The TCD spectral envelope flattens (low PI, often < 0.3), the systolic peak disappears, and the trace looks like a low-amplitude oscillating line at the support flow rate. Recovery of native pulsatility (rising PI, returning systolic peak) is a positive sign of cardiac recovery. The MFV remains computable and is a useful index of cerebral blood velocity. 2

HITS detection. TCD detects micro-emboli passing under the probe as high-intensity transient signals: brief (< 300 ms), high-amplitude (> 3 dB above background) Doppler artefacts. Background rate on a stable circuit is typically 1 to 3 per hour; a cluster of > 5 per hour, or > 10 over 1 hour, is suggestive of an embolic shower that warrants investigation (circuit thrombus, gas entrainment, recent line manipulation, or an active intracardiac thrombus). HITS counting requires a dedicated technologist or fixed-headframe monitoring; it is not a routine bedside measurement. 2 1

NIRS on ECMO. Bilateral frontal NIRS is the most-used bedside neurological monitor on ECMO. The interpretation is asymmetry-based: a persistent difference > 5 to 8% between hemispheres flags regional ischaemia. Common patterns: right > left (ipsilateral right common carotid cannulation in some neonates, lowering left hemisphere flow), or new asymmetry suggesting embolic stroke. A bilateral fall below 40% is concerning for global hypoperfusion; absolute values < 40 to 45% predict worse outcome in pediatric ECMO cohorts. 4 5

aEEG continuity. Amplitude-integrated EEG compresses the EEG trace into a continuity grade: continuous normal voltage, discontinuous, burst-suppression, low voltage, isoelectric. The trajectory matters: a previously continuous trace becoming discontinuous or burst-suppression is a marker of evolving injury. Sansevere 2023 demonstrated that continuous EEG on ECMO detects subclinical seizures in 10 to 15% of pediatric ECMO patients and that abnormal background predicts worse outcome. 6 7 4

The anticoagulation tension. ECMO requires systemic anticoagulation (usually unfractionated heparin titrated to ACT, aPTT, or anti-Xa). Too little anticoagulation: circuit clotting, embolic stroke. Too much: intracranial haemorrhage, the leading cause of death on neonatal ECMO. The MNM stack informs the conversation: a rising HITS rate or asymmetric NIRS argues for tighter anticoagulation; a new haemorrhage on head ultrasound argues for loosening it. 1

4. The multimodal picture table#

The most useful pairings on ECMO: TCD HITS + NIRS asymmetry (embolic-shower-to-stroke conversion), aEEG + cEEG (global function plus seizure detection), and NIRS + head ultrasound (neonatal-specific bundle).

5. Decision tree#

flowchart TD
  Start[Child on ECMO, sedated, paralysed] --> MNM[Continuous NIRS + aEEG; intermittent TCD; pupillometry q4h]
  MNM --> Surveillance{Any new MNM signal?}
  Surveillance -->|HITS cluster > 10/h| Embolic[Check circuit, recent manipulation, intracardiac thrombus]
  Surveillance -->|NIRS asymmetric drop > 5-8%| Regional[Look for regional stroke]
  Surveillance -->|aEEG voltage drop or burst-suppression| Global[Global injury suspected]
  Surveillance -->|Pupillometry asymmetric NPi drop| Brainstem[Brainstem or herniation concern]
  Surveillance -->|HU IVH in neonate| Bleed[Adjust anticoagulation; repeat HU]
  Embolic --> Anticoag1{Anticoagulation adequate?}
  Anticoag1 -->|Loose| Tighten[Tighten anticoagulation; circuit inspection]
  Anticoag1 -->|Adequate| Image[Brain imaging if exam feasible]
  Regional --> Image
  Global --> Image
  Brainstem --> Image
  Bleed --> Loosen[Loosen anticoagulation; repeat imaging]
  Image --> Strokey{Infarct or bleed?}
  Strokey -->|Ischaemic, recent| Discuss[Neurointerventional discussion; weigh thrombectomy vs supportive]
  Strokey -->|Haemorrhagic| Loosen
  Strokey -->|Neither| Recheck[Recheck MNM in 1-2 h]

6. Step-by-step bedside actions#

1. Establish the MNM bundle on day 0 of ECMO: bilateral NIRS, aEEG (or cEEG if seizure risk high), pupillometry q4h, daily head ultrasound (if fontanelle open), intermittent TCD twice daily by a trained operator.
2. Document baseline NIRS, aEEG continuity, TCD MFV and PI, NPi in the first 6 hours on circuit. All subsequent values interpreted as deltas.
3. Daily sedation hold when safe to enable a brief clinical exam (motor response, brainstem reflexes, pupillary check). On ECMO this is rarely possible in unstable circuits but should be planned when stability allows.
4. Watch for the three canonical patterns:
   • HITS cluster > 5 per hour (or > 10 over 1 hour) on TCD
   • Asymmetric NIRS drop > 5 to 8% sustained > 30 minutes
   • aEEG voltage drop, new discontinuity, or burst-suppression
5. When any pattern triggers, walk the differential before imaging: sensor / probe drift, MAP change, sedation, circuit event. Document what you ruled out.
6. If two patterns converge (e.g., HITS + NIRS asymmetry, or aEEG + NIRS asymmetry), image (head CT or MRI). The transport risk on ECMO is real; weigh against the diagnostic gain.
7. Anticoagulation adjustment: convene the ECMO and neurology teams. New ICH = consider lowering anticoagulation target (ACT 160 to 180 rather than 180 to 200). New infarct with embolic shower = consider tightening (anti-Xa 0.3 to 0.7 IU/mL, ACT > 200), if no bleeding.
8. Seizure management on ECMO: levetiracetam 60 mg/kg load is the workhorse; phenobarbital is avoided when possible (hepatic sedation, drug interactions). Continuous EEG to confirm response.
9. Family conversation: neurological injury rate on ECMO is 15 to 30%; the team should be transparent early. The MNM stack provides honest evidence for prognostic conversations.
10. At decannulation, repeat MRI within 7 days to characterise any acquired injury for prognostic and rehabilitation planning.

7. Management ladder and endpoints#

Success looks like: no clinical neurological deterioration during ECMO, normal MRI at decannulation, age-appropriate developmental trajectory.

Failure looks like: large infarct or haemorrhage on imaging, refractory seizures, global hypoxic-ischaemic pattern incompatible with meaningful recovery, leading to WLST discussion.

8. Variant subsections#

8.1 VA-ECMO non-pulsatile considerations#

Full VA support flattens the TCD pulsatility (PI < 0.3); the spectral envelope is a low-amplitude oscillation around the mean velocity. Mx-based autoregulation (TCD MFV vs MAP correlation) becomes unreliable because MAP itself is non-pulsatile. NIRS and aEEG become disproportionately important. Carotid cannulation (rarely used now in older children) historically caused ipsilateral ischaemia; modern femoral or jugular approaches preserve cerebral inflow. Left-ventricular distension (the failing native heart cannot eject against the high arterial circuit pressure) raises pulmonary venous pressure and can cause cardiac thrombus, a source of embolic stroke. 2

8.2 VV-ECMO#

VV preserves native cardiac output and pulsatility; TCD looks closer to normal. Neurological injury rate is lower than VA but not zero (5 to 15%, mostly haemorrhagic). Anticoagulation requirements are similar (heparin target ACT 180 to 200 or anti-Xa 0.3 to 0.7). NIRS is still essential; cannulation site (right internal jugular) does not compromise cerebral inflow.

8.3 Embolic detection (HITS)#

Requires a dedicated TCD technologist or fixed-headframe continuous monitoring. Manual handheld twice-daily TCD does not adequately quantify HITS rate. The decision to invest in HITS surveillance is centre-specific; the literature supports it as a marker of stroke risk in adult VA-ECMO and is emerging in pediatric. Counting protocol: 30-minute monitoring of one MCA, with the technologist annotating each HITS by intensity and duration. Background rate < 3 per hour; clusters > 5 in 30 minutes warrant investigation. 2

8.4 Seizure monitoring#

10 to 20% of children on ECMO develop seizures; the majority are non-convulsive given paralysis and sedation. Continuous EEG is the gold standard; aEEG is a reasonable resource-limited substitute that detects 60 to 80% of seizures (misses focal posterior or basal seizures). Subclinical status epilepticus on ECMO predicts worse neurological outcome and warrants aggressive treatment. 6 8

8.5 Central cannulation vs peripheral#

Central cannulation (sternotomy, ascending aorta and right atrium) is used in post-cardiotomy ECMO; peripheral (femoral or right internal jugular) is the more common pediatric configuration. Central carries higher bleeding risk and higher stroke risk; peripheral risks limb ischaemia at the access site. The neurological consequences of either are similar in pooled data, but the bleeding risk profile differs. 1

8.6 Decannulation and post-ECMO surveillance#

Immediately after decannulation, the brain is re-exposed to the patient's native circulation (pulsatile, on the heart's recovered or recovering function). NIRS may show transient swings; aEEG should return to baseline if the patient was not neurologically injured during the run. MRI within 7 days of decannulation is recommended in many pediatric centres to characterise acquired injury; the literature shows previously undetected small infarcts in 20 to 40% of survivors. 4

9. Multimodal integration matrix#

10. Worked alternative scenarios#

10.1 What if the HITS cluster is gas, not solid emboli?#

A 12-year-old VA-ECMO patient. TCD shows 15 HITS in 30 minutes after a circuit oxygenator change. NIRS unchanged bilaterally. aEEG unchanged. The HITS are likely gas emboli from the circuit change, which the brain often clears without infarction if the cluster resolves. Action: monitor closely for 4 hours, repeat TCD, expect HITS rate to fall to baseline. If NIRS or aEEG deteriorates, the picture changes to embolic stroke and you image.

10.2 What if aEEG suddenly drops without HITS or NIRS change?#

A neonate on VV-ECMO day 5 develops sudden aEEG burst-suppression. NIRS unchanged. No HITS. The differential: subclinical seizure (paradoxically aEEG narrows during ictal-postictal cycling), sedation change, or evolving global injury. Action: switch to cEEG to characterise; check sedation, check temperature (hypothermia depresses aEEG), check ABG. If subclinical status, load levetiracetam. If sustained suppression with no reversible cause, head ultrasound (in neonates) or CT (in older children) to look for diffuse injury.

10.3 What if NIRS bilaterally falls slowly over 12 hours?#

A 9-year-old VA-ECMO day 7. NIRS L 65 / R 67 at hour 0; L 50 / R 52 at hour 12. Symmetric bilateral fall. The differential: anaemia (transfusion threshold reached), inadequate flow (circuit recirculation), worsening native cardiac function with circuit demand exceeding supply, or fever raising CMRO2. Action: check haemoglobin (transfuse if < 8 g/dL), circuit flow audit, ABG, temperature. The aEEG and TCD will lag; the NIRS trend is informative early.

11. Outcome data#

• Cho 2024: large pediatric ECMO registry analysis. Neurological injury (stroke, ICH, seizure, HIE) in 22% of pediatric ECMO runs; brain injury independently associated with mortality (OR 3 to 5). 3
• Lorusso 2017 ELSO neurological complications: foundational analysis of the ELSO registry; details rates of stroke, ICH, seizures across configurations. 1 9
• La Rovere 2017 pediatric ECMO TCD: feasibility and patterns; PI flattens on full VA; HITS detection valuable; reduced TCD signal predicts cerebral injury. 2
• Sansevere 2023 neonatal cEEG: continuous EEG on ECMO detects subclinical seizures in 10 to 15% of neonates; abnormal background predicts outcome. 6 7
• Naim 2023 PCCM neurological monitoring review: places MNM as standard of care on pediatric ECMO; recommends NIRS plus aEEG as minimum bundle. 4 10
• Davies 2017 pediatric NIRS: validates NIRS asymmetry > 5 to 8% as the actionable threshold for regional injury. 5

12. Pitfalls#

• Relying on clinical exam alone. Sedation and paralysis mute the exam; MNM is essential.
• Failing to invest in HITS counting. Without dedicated technologist or fixed-headframe TCD, HITS rate is uncountable; an embolic shower can be missed.
• Ignoring small asymmetries. A 6% NIRS asymmetry sustained over 30 minutes is more informative than a brief 10% transient.
• Over-tightening anticoagulation. Tighter heparin reduces embolic risk but raises bleed risk; the MNM stack must inform the balance.
• Forgetting daily head ultrasound in neonates. ICH is the dominant neurological complication on neonatal ECMO; HU is the cheap, accessible primary surveillance tool.
• Hyperventilating to "protect the brain". Hypocapnia constricts cerebral vessels and worsens ischaemia; maintain normocapnia.
• Sedation with phenobarbital. Phenobarbital depresses aEEG (confounds monitoring) and prolongs the muting of the clinical exam; use levetiracetam first.
• Missing the silent stroke. A clinically silent patient can still be having an evolving stroke; the four-modality bundle (TCD + NIRS + aEEG + pupillometry) is what catches it.

13. Pediatric considerations#

14. Combine with#

• NIRS modality page: the workhorse on ECMO.
• TCD / TCCD modality page: HITS detection and vessel-level patency.
• aEEG and continuous EEG: aEEG continuity and seizure detection.
• Pupillometry: brainstem sentinel when exam is muted.
• Brain death MNM integration: when MNM informs the end-of-care conversation.
• HIE in the newborn integration: related neonatal MNM bundle.
• Discordance triage: when modalities disagree.
• Family communication MNM: the conversation framework.

15. Evidence summary and recent literature (2022 to 2025)#

Foundational#

Recent literature (2022 to 2025)#

• Cho 2024: large pediatric ECMO registry analysis on brain injury and outcome; the most current numbers on rates and mortality association. 3
• Naim 2023 PCCM: pediatric brain injury during ECMO; MNM bundle recommendations. 4 10
• Sansevere 2023 neonatal cEEG: subclinical seizure detection rates on neonatal ECMO. 6 7
• Helbok 2024 pediatric MMM update: ECMO chapter places NIRS + aEEG as minimum bundle; cEEG and TCD as tier 2. 14
• Figaji 2025 pediatric MMM consensus: same framework, with explicit MNM recommendations for ECMO surveillance. 15
• Tasker 2023 PCCM review: integrative pediatric MMM review including ECMO. 16 17

16. Self-check#