The central bedside principle is that cerebral edema is not a single therapeutic target. It is one possible cause of intracranial volume expansion, and the dangerous variable is often not the absolute ICP but the pressure gradient producing tissue displacement, vascular compression, or herniation. A patient with diffuse traumatic swelling may have globally elevated ICP, but a patient with a malignant hemispheric infarction can develop lethal subfalcine and transtentorial shift while a monitor in the contralateral hemisphere still reports an unimpressive number. A cerebellar infarction may compress the fourth ventricle and brainstem before supratentorial ICP becomes markedly elevated. Conversely, severe hypercapnia, jugular obstruction, coughing, seizures, or acute hydrocephalus can produce an ICP crisis with little new parenchymal edema. The first question is therefore not simply, “How do I lower the ICP?” It is, “Which intracranial compartment is expanding, and what definitive action does the temporary ICP reduction need to buy time for?”

The Monro–Kellie framework remains the cleanest starting point. Brain tissue, blood, and CSF occupy a nearly fixed cranial volume. Early compensation displaces venous blood and CSF, so a sizable volume change may initially produce little change in pressure. Once this reserve is exhausted, the pressure–volume curve becomes steep and even a few additional milliliters can cause a dramatic rise in ICP. This is why a patient can remain stable for hours and then deteriorate rapidly, and why a small intervention—draining several milliliters of CSF, correcting acute hypercapnia, or relieving jugular compression—may suddenly restore compliance. The core physiology summarized by Deranged Physiology’s ICP framework remains valid: every treatment either removes intracranial volume, increases the container’s effective volume, prevents further volume accumulation, or protects perfusion while the underlying lesion is corrected.

When deterioration is abrupt, assessment and treatment occur simultaneously. Verify that the ICP value is real by examining the waveform, transducer level, drain status, and temporal relationship to stimulation, coughing, suctioning, positioning, or ventilator dyssynchrony. Examine the pupils and motor response, review the trend rather than one number, and obtain immediate imaging when it will not delay treatment of clear herniation. Check PaCO₂, oxygenation, blood pressure, sodium, glucose, temperature, seizure activity, and whether an EVD is blocked or inadvertently clamped. A new unilateral pupil, progressive decline in consciousness, extensor posturing, or a rapidly worsening focal deficit should be treated as a pressure-gradient emergency even without documented ICP above 22 mm Hg.

---

The initial physiologic bundle begins with oxygenation, perfusion, and venous drainage. Place the head in a neutral position, usually elevated around 30 degrees, and remove avoidable jugular obstruction from rotation, tight ties, or an excessively constrictive cervical collar while maintaining appropriate spine precautions. Head elevation lowers cerebral venous pressure, but it can also lower arterial pressure at the brain. If CPP is being calculated with the head elevated, the arterial transducer should be referenced near the external auditory meatus rather than the right atrium; otherwise cerebral MAP and CPP are overestimated by the hydrostatic gradient. In a hypotensive patient, improving MAP may be more important than further elevation of the head.

Intubation is undertaken for inadequate airway protection, hypoxemia, ventilatory failure, or a rapidly progressive herniation syndrome, but the induction itself is dangerous. Hypoxemia, apnea-associated hypercapnia, and postinduction hypotension can each accelerate secondary injury. Preoxygenation, hemodynamically appropriate induction, immediate confirmation of ventilation, and vasopressor readiness matter more than loyalty to one induction drug. Ketamine is not contraindicated solely because ICP is elevated; contemporary evidence does not support the old assumption that it predictably raises ICP in a ventilated patient. Propofol is useful when blood pressure tolerates it, but a large bolus into marginal hemodynamics can convert a pressure problem into a perfusion catastrophe.

Hypertonic saline and mannitol both lower brain water and often reduce ICP, but neither has convincingly improved functional outcomes simply by being selected over the other. Hypertonic saline increases extracellular tonicity, draws water from brain regions where an effective osmotic barrier remains, expands intravascular volume, and may improve microcirculatory rheology. Mannitol also establishes an osmotic gradient but produces more subsequent diuresis. Typical crisis doses are 3% saline approximately 2–5 mL/kg, concentrated 23.4% saline 30 mL, or mannitol approximately 0.5–1 g/kg; these are protocol-dependent starting ranges rather than evidence-derived universal doses. The established NCS cerebral-edema guideline conditionally favors hypertonic saline in TBI and ICH and favors symptom-triggered boluses rather than sodium-target infusions in SAH, but the certainty is low. The current Brain Trauma Foundation guidance explicitly concludes that hyperosmolar therapy lowers ICP while comparative outcome evidence remains insufficient to endorse one agent.

The physiologic choice is therefore patient-specific. Hypertonic saline is usually more attractive when hypotension, hypovolemia, hyponatremia, or concern for mannitol-induced diuresis is present. Mannitol may be useful when substantial hypernatremia or hyperchloremia limits additional sodium administration, provided renal function and hemodynamics are adequate. Anuria or major renal dysfunction makes mannitol accumulation, intravascular expansion, and rebound more concerning. Hypertonic saline is not benign in renal failure either; sodium accumulation, hyperchloremic acidosis, pulmonary edema, and volume overload remain possible. In shock, the diuresis after mannitol can reduce MAP more than it reduces ICP, causing CPP to fall despite an apparently successful pressure response.

Acute treatment should be response-directed rather than driven by a ceremonial sodium goal. Reassess the pupils, examination, ICP burden, waveform, CPP, hemodynamics, and, when available, regional oxygenation after the bolus. A serum sodium of 150 mEq/L is not itself evidence that brain swelling is controlled, and a sodium of 142 is not an indication to administer hypertonic saline prophylactically. Routine scheduled mannitol and indiscriminate continuous hypertonic infusions have not demonstrated improved neurologic outcomes. The NCS safety discussion identifies sodium approximately 155–160 mEq/L and chloride approximately 110–115 mEq/L as reasonable upper monitoring boundaries associated with renal risk, but these are guardrails, not therapeutic targets. With mannitol, follow renal function, volume status, measured osmolality, and osmolar gap; the traditional osmolality cutoff of 320 mOsm/kg should not be treated as a validated binary stopping rule.

CSF drainage is the most direct treatment when the expanding compartment is CSF. An EVD can rapidly reverse hydrocephalus that would respond poorly to repeated osmotherapy. In severe TBI, continuous drainage may lower overall ICP burden more effectively than intermittent drainage, although the evidence is limited and continuous drainage sacrifices continuous pressure measurement unless a separate monitor is present. In aneurysmal SAH or IVH with symptomatic hydrocephalus, urgent CSF diversion addresses the mechanism rather than merely reducing brain water.

Posterior-fossa disease deserves special caution because supratentorial ICP can underestimate the local threat. The 2026 AHA/ASA ischemic-stroke guideline recommends ventriculostomy for obstructive hydrocephalus after cerebellar infarction, but drainage should be controlled because an infratentorial pressure gradient can theoretically promote upward displacement. The same guideline now recommends suboccipital decompression with dural expansion for neurological deterioration from brainstem compression or a cerebellar infarct volume of at least 35 mL. That volumetric threshold derives from observational evidence and should complement—not replace—examination, hydrocephalus, cisternal compression, and the trajectory of deterioration. These updated recommendations are summarized in the 2026 AHA/ASA brain-swelling guidance.

---

Ventilation manipulates cerebral blood volume rather than brain water. Lowering PaCO₂ constricts cerebral arterioles and can reduce ICP within minutes, making hyperventilation a valuable bridge during active herniation. The same vasoconstriction can worsen ischemia, especially when autoregulation is impaired or CBF is already reduced. A temporary PaCO₂ around 30–35 mm Hg is generally sufficient; more profound hypocapnia should be reserved for the briefest possible rescue interval. The effect also fades as CSF pH re-equilibrates. Prolonged prophylactic hyperventilation, particularly to PaCO₂ of 25 mm Hg or less, is not recommended, and the danger may be greatest during the first day after severe TBI when CBF is often already depressed. Once the definitive intervention is underway, return toward normocapnia deliberately rather than leaving the patient chronically vasoconstricted.

Ventilator settings should be judged by their net effect on brain and lung rather than by the myth that PEEP is automatically harmful. PEEP that recruits lung, improves oxygenation, and does not materially increase right atrial pressure or reduce MAP may have little adverse effect on ICP. PEEP that produces alveolar overdistention, raises intrathoracic and central venous pressure, or lowers cardiac output can obstruct cerebral venous drainage and reduce CPP. The bedside answer comes from observing ICP, CPP, oxygenation, compliance, and hemodynamics during the adjustment.

Analgesia and sedation reduce cerebral metabolic demand, coughing, sympathetic surges, and ventilator dyssynchrony. They are especially effective when the ICP elevation is stimulus-driven, but sedation will not resolve hydrocephalus or a growing hematoma. Propofol provides rapid titratability and ICP control but requires surveillance for hypotension and propofol-infusion syndrome at high cumulative exposure. A short course of neuromuscular blockade can control severe coughing or shivering after adequate sedation, but it eliminates the motor examination and masks convulsive activity; continuous EEG becomes important when seizures remain plausible.

Barbiturates belong in a later rescue tier and must be separated by etiology. For severe TBI with ICP refractory to maximum standard medical and surgical therapy, high-dose barbiturate therapy can be used to suppress cerebral metabolism and control ICP, provided hemodynamic stability is maintained. Prophylactic burst suppression is not recommended. In malignant ischemic swelling, the 2026 AHA/ASA guideline specifically recommends against barbiturates, hypothermia, and corticosteroids because efficacy is unproven and adverse effects are substantial. Hypothermia can lower metabolism and ICP, but that physiologic success has repeatedly failed to translate into better outcomes and can add hypotension, infection, coagulopathy, arrhythmia, electrolyte shifts, and difficult rebound during rewarming.

Corticosteroids demonstrate why edema phenotype matters. They are appropriate for vasogenic edema surrounding many intracranial tumors because they reduce abnormal capillary permeability. They are also indicated in selected CNS infections, most notably dexamethasone given before or with the first antibiotic dose in appropriate bacterial meningitis. They do not treat cytotoxic ischemic edema, TBI edema, or perihematomal edema; high-dose methylprednisolone increased mortality in severe TBI. A dramatic response to dexamethasone in tumor edema should never be generalized into “steroids treat brain swelling.”

---

The most important recent stroke update is that osmotherapy is explicitly framed as a bridge, not definitive treatment. The 2026 AHA/ASA guideline recommends osmotic therapy for neurological decline from swelling while surgical intervention is being arranged. In patients 60 years or younger with unilateral MCA infarction who deteriorate from swelling within 48 hours despite medical treatment, decompressive hemicraniectomy with dural expansion improves survival and functional outcome. In patients older than 60, surgery may reduce mortality, but survival is frequently accompanied by major disability, making early values-based discussion essential. Thrombolysis does not itself preclude decompression when malignant edema develops.

The same guideline now recommends against intravenous glibenclamide for large hemispheric infarction. The mechanism was compelling: SUR1–TRPM4 blockade might limit ionic edema and microvascular failure. In CHARM, however, glibenclamide did not produce a favorable 90-day modified Rankin shift, with a common odds ratio of 1.17 and a 95% confidence interval of 0.80–1.71; mortality was not reduced, and hypoglycemia occurred more often. The trial stopped early for operational reasons and was therefore underpowered for definitive subgroup conclusions, but the overall result does not support clinical use. This is a clean example of physiology surviving a negative trial: ionic edema and SUR1–TRPM4 remain biologically real, but modifying one pathway did not improve patient-centered outcomes in the studied population. The CHARM primary report supports continued investigation, not routine treatment.

Surgery for ICH provides another lesson in mechanistic precision. Removing clot reduces mass effect, potentially limits exposure to blood-breakdown products, and may improve compliance, but traversing eloquent tissue can create competing injury. ENRICH demonstrated benefit primarily in carefully selected lobar hemorrhages treated early, whereas the 2025 MIND trial enrolled a population dominated by deep hemorrhages and allowed minimally invasive evacuation within 72 hours. MIND achieved substantial clot reduction but did not improve 180-day disability: the odds ratio for superiority was 1.03, with a 96% confidence interval of 0.62–1.72, and 30-day mortality was 7.2% versus 9.8%. Enrollment stopped early, so location-specific effects remain uncertain. The defensible inference is not that evacuation never works, but that timing, hematoma location, surgical trajectory, technique, and the amount of irreversible primary injury determine whether volume removal translates into recovery. The contrast between MIND and ENRICH is exactly why an anatomically targeted intervention cannot be generalized to every swollen hemorrhagic brain.

Decompressive craniectomy in TBI makes the same point. Increasing cranial volume reliably lowers ICP. It does not restore destroyed tissue, and depending on the clinical threshold and timing, survival may shift toward severe disability. Early decompression for relatively modest refractory ICP and later rescue decompression for extreme refractory ICP are not interchangeable procedures. The decision must incorporate the underlying lesion, duration and intensity of the ICP burden, pupillary trajectory, systemic insults, multimodal evidence of viable but threatened brain, and the patient’s acceptable outcome states. A technically successful reduction in ICP is an intermediate physiologic endpoint, not proof of neurological benefit.