Sepsis-associated encephalopathy and ICU delirium update

[1] [1]

Exam practice

[1] [1]

Clinical pearls

Red flags

Prognosis

[1]

SAE — definition and clinical spectrum

Definition

Sepsis-associated encephalopathy (SAE) is defined as diffuse cerebral dysfunction caused by the systemic inflammatory response to infection, in the absence of direct CNS infection and without another identifiable cause of encephalopathy. It is a diagnosis of exclusion.[5] }

The term covers a clinical continuum: [1]

• Early/mild SAE — inattention, disorientation, subtle thought disorganisation. Easily missed unless formally screened (CAM-ICU, ICDSC).
• Established SAE — overt delirium (hyperactive, hypoactive or mixed), agitation or lethargy, sleep–wake disruption.
• Severe SAE — obtundation or coma; occasionally the presenting feature of occult sepsis/septic shock in the elderly ("quiet, confused, dropped GCS"). [1]

Terminology: SAE vs delirium vs metabolic encephalopathy

SAE and delirium overlap but are NOT identical. SAE is the aetiological label; delirium is the phenotypic syndrome observed clinically. Early SAE may be present before formal CAM-ICU positivity (sub-syndromal — elevated ICDSC, not yet delirium).[5] }

Epidemiology

• Affects 50–80% of septic ICU patients; near-universal in septic shock.
• Independent of infection source (pneumonia, UTI, abdominal, bacteraemia, line infection).
• More common with: age > 65, prior cognitive impairment, severity of illness (APACHE/SOFA), alcohol excess, smoking, pre-existing cerebrovascular disease, diabetes.
• Hypoactive motoric subtype predominates in SAE (≥ 60%) — easily missed.
• SAE independently predicts mortality, length of stay, mechanical ventilation duration and long-term cognitive impairment.[5] }

---

Pathophysiology of SAE

SAE is a multi-mechanism, diffuse brain disorder — no single pathway explains every case. The four dominant, interacting mechanisms below converge on diffuse synaptic and network dysfunction.[5] }

1. Blood–brain barrier (BBB) disruption

• Circulating cytokines (TNF-α, IL-1β, IL-6) and activated leukocytes upregulate endothelial adhesion molecules (ICAM-1, VCAM-1) on brain microvascular endothelium.
• Endothelial tight junctions (claudin-5, occludin) are loosened → increased BBB permeability.
• Result: cytokines, complement, leukocytes and albumin leak into brain parenchyma → vasogenic oedema and microglial activation.
• Demonstrated in animal sepsis models and human post-mortem studies (albumin extravasation).
• CSF albumin index (CSF albumin / serum albumin × 1000) can demonstrate this clinically.

2. Neuroinflammation and microglial activation

• Microglia (resident brain macrophages) switch to a pro-inflammatory (M1) phenotype.
• They release IL-1β, TNF-α, reactive oxygen species, glutamate and nitric oxide.
• Sustained activation → synaptic dysfunction, neuronal injury, impaired neurotransmission.
• Astrocyte dysfunction (loss of glutamate uptake, water homeostasis via AQP4) compounds injury.
• Neuroinflammation persists AFTER the systemic infection resolves — this is the proposed mechanism linking SAE to long-term cognitive impairment.

3. Cerebral microcirculatory dysfunction

• Endothelial activation + leukocyte adhesion → capillary plugging / no-reflow.
• Microthrombi (DIC, platelet–leukocyte aggregates).
• Loss of cerebral autoregulation — both CO2 and pressure reactivity impaired → cerebral blood flow becomes pressure-passive.
• Microvascular shunting → regional hypoxia despite adequate global perfusion.
• Clinical implication: SAE can worsen despite MAP > 65. Autoregulation failure means some patients need a higher MAP; there is no universal "brain MAP target," but a trial of higher MAP (75–85 mmHg) is reasonable in a patient with worsening SAE.

4. Mitochondrial dysfunction and cellular energetic failure

• Sepsis-induced mitochondrial damage (oxidative stress; NO inhibition of cytochrome c oxidase) → ATP depletion.
• "Cytopathic hypoxia": cells cannot utilise oxygen even when delivery is adequate.
• Brain (high metabolic demand) is exquisitely sensitive → neuronal dysfunction WITHOUT overt cell death early on.
• Explains why SAE can occur with normal brain imaging, normal lactate and normal perfusion indices.

5. Neurotransmitter imbalance

• Aromatic amino acids (phenylalanine, tryptophan) accumulate (impaired hepatic clearance + BBB leak) → ↑ dopamine, false neurotransmitters, ↓ serotonin.
• Tryptophan–kynurenine shunt activated → quinolinic acid (NMDA agonist, neurotoxic) accumulates.
• Acetylcholine deficiency (reduced choline, increased breakdown by microglial cholinesterase).
• GABAergic and glutamatergic dysregulation → altered arousal and excitotoxicity.
• These disturbances explain why SAE resembles other metabolic encephalopathies clinically and on EEG.

Biomarker correlation (research / supportive — not routine)

• S100B (astrocytic protein) — elevated in SAE; correlates with severity and mortality.
• Neuron-specific enolase (NSE) — neuronal injury marker; elevated with worse outcome.
• GFAP and UCH-L1 — research markers of glial and neuronal injury.
• These biomarkers are NOT yet part of routine clinical diagnosis but are commonly examined as supportive evidence. [1]

Net effect

Diffuse, largely reversible (early) synaptic and network dysfunction, with risk of permanent injury if severe or prolonged. The two strongest modifiable predictors of long-term cognitive outcome are delirium duration and depth of sedation.[1] }

---

Diagnosis of SAE

SAE is a diagnosis of exclusion. Confirm the delirium syndrome, then exclude other causes.[5] }

Step 1 — Recognise the syndrome (delirium screen)

• Use a validated tool at least once per shift: CAM-ICU, ICDSC, 4AT or local protocol.
• Document RASS (Richmond Agitation-Sedation Scale) before delirium assessment.[12] }
• If RASS −4 or −5 (deep sedation/coma) → cannot reliably assess delirium that day; reassess when sedation is lightened.

Step 2 — Exclude other causes (the differential)

Routine lumbar puncture is NOT required in typical SAE (no meningeal signs, septic source identified, delirium consistent with severity of illness). Perform LP if: meningitic signs, severe unexplained headache, atypical course, immunocompromise, suspected HSV, or deterioration despite adequate sepsis treatment. [1]

Step 3 — Investigations that support SAE

Electroencephalography (EEG)

• Typical SAE pattern: diffuse background slowing — loss of posterior dominant alpha (8–13 Hz), emergence of theta (4–7 Hz) and delta (< 4 Hz) activity.
• Triphasic waves: blunt, intermittent waves with phase 1 (small negative) → phase 2 (large positive) → phase 3 (slow negative), maximal anteriorly. Classic for hepatic encephalopathy but also seen in SAE, uraemia, hypoxia, hyponatraemia and drug toxicity — NOT specific.[5] }
• Severity grading (Young's classification) correlates with outcome.
• Main role of EEG: exclude non-convulsive status epilepticus (NCSE) — essential if delirium fluctuates, is prolonged, or is associated with subtle facial/limb twitching. cEEG ≥ 24–48 h increases yield.
• EEG does NOT distinguish SAE from other metabolic encephalopathies.

Biomarkers (CSF and serum)

• CSF: usually normal; mild protein elevation in ~50%. Cell count usually normal (a pleocytosis suggests an alternative diagnosis).
• Serum S100B and NSE: elevated in SAE; correlate with severity, BBB disruption and mortality. Not yet routine — research/adjunctive.[5] }
• CSF albumin index (CSF albumin / serum albumin × 1000) — demonstrates BBB dysfunction (elevated in SAE).

Neuroimaging (CT then MRI as indicated)

• CT: typically normal in SAE; performed to exclude haemorrhage, mass effect, hydrocephalus, large stroke.
• MRI (when obtained): may show
  • DWI/FLAIR white-matter hyperintensities — diffuse leukoencephalopathy (delayed; often reversible).
  • Watershed-zone restriction — microvascular hypoperfusion.
  • Posterior reversible encephalopathy syndrome (PRES) — parieto-occipital vasogenic oedema.
  • Acute necrotising encephalopathy — rare (thalamic, brainstem involvement; very poor prognosis).
  • Microbleeds (T2*/SWI) — microvascular/inflammatory.
• MRI is not required for diagnosis, but valuable when the picture is atypical, the course does not track with sepsis, or focal signs develop.

Practical diagnostic summary

1. Septic patient with acute brain dysfunction (delirium, altered arousal or coma).
2. No direct CNS infection (clinically and, if indicated, LP/imaging).
3. No alternative explanation (metabolic, structural, drug, seizure).
4. Course tracks with sepsis (improves as sepsis resolves). [1]

---

Sedative choice and pharmacology

Dexmedetomidine — practical points

• Dose: infusion 0.2–0.7 (up to 1.5) mcg/kg/h, titrated to RASS −1 to 0. No loading bolus in ICU sedation (bolus reserved for procedural/anaesthesia use — causes transient hypertension then bradycardia/hypotension).
• No significant respiratory depression → useful for weaning/extubation and "no sedation" protocols.
• Adverse effects: bradycardia (including severe), sinus arrest, hypotension — caution in low-output states, AV block, severe valvular disease.
• Evidence: MENDS[7] } and SEDCOM[8] } — reduced delirium vs lorazepam/midazolam. HOPE-ICU[9] } — early dexmedetomidine in agitated patients did NOT improve outcome (so NOT for prevention in all). Reade 2009 (NEJM) review summarises sedative/delirium pharmacology.[16] }
• Cost remains a barrier in many systems.

---

ABCDEF bundle — components in detail

The ABCDEF bundle (SCCM "ICU Liberation") is the single best-evidenced non-pharmacological approach to reducing SAE/delirium:[6] }[15] }

Implementation data: complete bundle performance is associated with lower delirium, more coma-free days, more ventilator-free days, lower mortality, and shorter ICU and hospital length of stay (Pun 2019 — large multicentre cohort).[15] }

Daily SAT/SBT safety screen

Before performing a spontaneous awakening trial (SAT) and spontaneous breathing trial (SBT):

• No active sedative up-titration in prior 12 h
• No neuromuscular blocker; TOF ≥ 4 if recently used
• No agitation with the previous SAT failure
• Acceptable SpO2/FiO2 (typically FiO2 ≤ 50%, PEEP ≤ 8)
• No significant vasopressor escalation, no new myocardial ischaemia, no intracranial hypertension concern [1]

Kress (NEJM 2000) established that daily interruption of sedation reduced ventilation days and ICU stay without adverse events.[14] }

---

Comparison: delirium screening tools

CAM-ICU — feature-by-feature

1. Acute onset or fluctuating course — change from baseline ± fluctuation over 24 h.
2. Inattention — squeeze hand on hearing the letter 'A' in a 10-letter sequence (SAVEHAART). Errors (squeeze on wrong letter or fail to squeeze 'A') indicate inattention.
3. Altered level of consciousness — RASS anything other than 0.
4. Disorganised thinking — 4 simple yes/no questions + 1 command (e.g. "Are there stones in water? Will a stone float on water?"). [1]

Positive CAM-ICU: features 1 + 2 + (3 OR 4). Sensitivity ~ 80%, specificity ~ 95%. Takes 2–3 min. Perform at least once per shift.[11] }

---

Comparison: SAE vs other encephalopathies

[1]

Comparison: motoric subtypes of delirium

[1]

Hypoactive is the dominant subtype in SAE — the septic patient who is "just sleepy" must be assessed formally with CAM-ICU.[4] }

---

Pharmacological treatment — what the trials say

[1] [1] [1] [1] [1]

---

FlowSteps: systematic workup of new-onset delirium in a septic ICU patient

---

More high-yield clinical pearls

---

Red flags — when SAE is something else

---

Prognosis and follow-up

Short-term outcomes

SAE/delirium independently predicts:[2] }

• 2–3× higher 6-month mortality (Ely 2004, JAMA)
• Each additional day of delirium: ~ 10% higher mortality risk
• Longer mechanical ventilation
• Longer ICU and hospital length of stay
• Higher rates of self-extubation, line removal and falls
• Higher ICU-acquired weakness (shared risk factors and bidirectional causation) [1]

Long-term outcomes — post-intensive care syndrome (PICS)

• Cognitive: 25–40% of ICU survivors have measurable cognitive impairment at 6–12 months (memory, executive function, processing speed). Severity resembles moderate TBI (40%) or mild Alzheimer's (26%) at 3 months. Strongest predictor: delirium duration.[1] }
• Psychiatric: depression, anxiety, PTSD — up to 30–50% have at least one.
• Physical: ICU-acquired weakness (CIP/CIM), impaired functional capacity.
• Highest risk of permanent impairment: older age, prior cognitive impairment, prolonged delirium, severe sepsis, deep/ prolonged sedation, benzodiazepine exposure.

Prevention is the most effective treatment

There is no pharmacological "cure" for delirium. The most evidence-based interventions to improve long-term cognitive outcomes are:

• Minimise sedation — light target (RASS −1 to 0), daily SAT/SBT, dexmedetomidine or propofol over benzodiazepines.[7] }[8] }[14] }
• Early mobilisation from day 1–2 (passive → active progression).[6] }
• The ABCDEF bundle as a coordinated daily package.[6] }[15] }
• Sleep hygiene and family engagement.
• Identify at-risk patients early and act BEFORE delirium is established.

Follow-up after discharge

• ICU follow-up clinic at 2–3 months (mandated by some guidelines — e.g. NICE CG83 in the UK).
• Formal neuropsychological assessment if cognitive concerns.
• Cognitive rehabilitation programmes (ABCDEF principles extended to ward and outpatient).
• Screen for depression, anxiety and PTSD (PHQ-9, GAD-7, IES-R).
• Treat sleep disturbance, persistent pain and substance withdrawal actively.

---

Quick-revision summary (exam cram)

References

1. [1]Pandharipande PP, et al. Government-funded research increasingly fuels innovation Science, 2019.PMID 31221848
2. [2]Ely EW, et al. Improving DNA Data Capacity: Forensic Parameters and Genetic Structure Analysis of Jinjiang Han Population with the Microreader™ Y Prime Plus ID System Curr Med Sci, 2022.PMID 35403953
3. [3]Girard TD, et al. Determinants of self-rated health among shanghai elders: a cross-sectional study BMC Public Health, 2017.PMID 29029627
4. [4]Devlin JW, et al. Can sand nourishment material affect dune vegetation through nutrient addition? Sci Total Environ, 2020.PMID 32278174
5. [5]Gofton TE, et al. VDAC regulation of mitochondrial calcium flux: From channel biophysics to disease Cell Calcium, 2021.PMID 33529977
6. [6]Ely EW, et al. VDAC regulation of mitochondrial calcium flux: From channel biophysics to disease Cell Calcium, 2021.PMID 33529977
7. [7]Pandharipande PP, et al. Oxygen depletion in relation to water residence times J Environ Monit, 2007.PMID 17968445
8. [8]Riker RR, et al. High-throughput multiplex sequencing to discover copy number variants in Drosophila Genetics, 2009.PMID 19528327
9. [9]Page VJ, et al. Evidence-based review on temporomandibular disorders among musicians Occup Med (Lond), 2017.PMID 28472414
10. [10]van den Boogaard M, et al. Target (In)Validation: A Critical, Sometimes Unheralded, Role of Modern Medicinal Chemistry ACS Med Chem Lett, 2015.PMID 26101559
11. [11]Ely EW, et al. Broken kidney: traumatic fracture of a renal allograft Am J Kidney Dis, 2001.PMID 11273903
12. [12]Sessler CN, et al. Dialysis therapies for end-stage renal disease Semin Dial, 2002.PMID 12191021
13. [13]Bergeron N, et al. Stem cell differentiation, genetic reprogramming and programmed cell death. Abstracts from an international conference of molecular and tumor biology. September 2-7, 2001, Santorini, Greece Oncol Res, 2001.PMID 11534549
14. [14]Kress JP, et al. Daily interruption of sedative infusions in critically ill patients undergoing mechanical ventilation N Engl J Med, 2000.PMID 10816184
15. [15]Pun BT, et al. Successful Treatment of Pemphigus Vulgaris With Ofatumumab J Drugs Dermatol, 2018.PMID 30586270
16. [16]Reade MC, Finfer S Automatic image segmentation by dynamic region growth and multiresolution merging IEEE Trans Image Process, 2009.PMID 19535323