Aneurysmal subarachnoid hemorrhage, the bleeding that occurs when a weakened vessel at the base of the brain bursts into the space surrounding it, carries a reported incidence of early seizures ranging from 8 to 24 percent. The widespread adoption of continuous electroencephalography in neurocritical care units has only complicated the picture, revealing a hidden burden of nonconvulsive seizures and even nonconvulsive status epilepticus that would otherwise go undetected. These silent electrical storms matter: patients with a high burden of electrographic seizures show measurably worse cognitive and functional outcomes. Known risk factors for early seizures include older age, rupture of an aneurysm in the anterior circulation, hydrocephalus requiring drainage of cerebrospinal fluid, seizures occurring before hospital arrival, higher clinical grade on admission, surgical clipping of the aneurysm, and the presence of a thick subarachnoid clot.

Against this backdrop, clinicians have long prescribed prophylactic antiseizure medications, but with little rigorous justification. Most of the supporting evidence comes from observational cohort studies, and professional societies have hedged accordingly. The earlier American Heart Association guidelines suggested considering prophylactic medication during the post-bleed period while warning against long-term use; the 2023 revision offered only a weak recommendation for patients with high-seizure-risk features; and the Neurocritical Care Society’s own comprehensive management guidelines from 2023 conspicuously made no statement on prophylaxis at all. The new guideline panel, formed in October 2019 and co-chaired by A. Shaun Rowe of the University of Tennessee Health Science Center and Jennifer A. Frontera of New York University Grossman School of Medicine, set out to close this gap using the Grading of Recommendations Assessment, Development, and Evaluation methodology, the international standard for translating evidence into clinical guidance.

The panel structured its work around three carefully framed questions, each following the population, intervention, comparator, and outcome format. First, in adult patients hospitalized with aneurysmal subarachnoid hemorrhage who have never had a clinical or electrographic seizure, should an antiseizure medication be given at all? Second, if treatment is chosen, should clinicians prefer levetiracetam or phenytoin and its water-soluble prodrug fosphenytoin? Third, if treatment is chosen, should it last a short period of three days or less, or extend beyond three days? The outcomes judged critical were early seizures, defined as occurring within fourteen days of hemorrhage or during hospitalization; late seizures at or beyond fourteen days; and adverse events attributable to the drugs themselves. Mortality and functional and cognitive outcomes, scored on instruments such as the modified Rankin Scale, were rated as important but not decisive.

The literature search, initially spanning databases including PubMed, Medline, Embase, Emcare, and the Cochrane Library from 1946 through July 2020 and updated through September 2024, retrieved nearly two thousand articles. Only ten survived the screening process, and just seven entered the meta-analyses. That winnowing alone tells the story: an intervention applied to tens of thousands of patients annually worldwide rests on a handful of small, nonrandomized investigations. Risk of bias was assessed with the Cochrane RoB-2 tool for randomized trials and the ROBINS-I instrument for nonrandomized studies, and all meta-analyses used random-effects models with heterogeneity flagged at an I-squared value of fifty percent or greater.

On the first question, the pooled evidence was sobering. Two nonrandomized studies encompassing 437 patients compared antiseizure medication with no treatment, and the meta-analysis found no significant reduction in early seizures, with a relative risk of 0.72 and a confidence interval spanning from 0.37 to 1.41. No study at all examined late seizures in this comparison. Adverse events, documented in a single study of 84 patients, were numerically more than four times as common in the treated group—nine of forty-four versus two of forty—though the difference fell just short of statistical significance with a p value of 0.06. Functional outcomes at discharge and six months showed no meaningful difference. The panel’s verdict: antiseizure medication may be used or withheld, a conditional recommendation grounded in low-quality evidence, with the observation that clinicians may reasonably favor treatment in high-risk patients such as those with high-grade hemorrhage, intraparenchymal clot, hydrocephalus, or surgical clipping.

The drug comparison proved equally murky. Three nonrandomized studies addressed early seizure prevention, but two reported zero events, leaving essentially one study to drive the finding. That investigation showed fewer early seizures with phenytoin or fosphenytoin—a relative risk of 2.46 favoring phenytoin against levetiracetam with a confidence interval of 1.09 to 5.55. Yet the safety picture ran the other way. In a study by Shah and Husain, patients on phenytoin or fosphenytoin experienced dramatically more elevated transaminases, at 30.1 percent versus zero percent, more thrombocytopenia at 11.5 percent versus zero, and more unexplained fever. A separate case-control study of 259 patients, excluded from the meta-analysis because of medication-switching confounding, linked phenytoin to worse functional outcomes and more delayed cerebral ischemia. There was also a mechanistic concern: phenytoin induces CYP3A4, the enzyme that metabolizes nimodipine, the drug given to all these patients to prevent delayed brain ischemia, potentially undermining its effectiveness. The panel suggested either agent could be used, conditional on very low quality evidence, but its expert commentary notes that most practitioners prefer levetiracetam for its favorable safety profile and fewer drug interactions.

The dosing problem received particular attention. Levetiracetam clearance is known to be enhanced in these patients because of augmented renal clearance, meaning standard doses of 500 milligrams twice daily may leave blood levels far below therapeutic targets. Population pharmacokinetic work cited in the guideline found that patients with augmented renal clearance may require as much as 1500 milligrams twice daily, and studies in broader neurocritical care populations suggest that doses of 750 to 1000 milligrams twice daily or total daily doses exceeding 1000 milligrams achieve better target levels and fewer seizures. The panel’s expert guidance explicitly raises the possibility of higher dosing or loading strategies and recommends serum level monitoring where feasible, acknowledging that apparent failures of levetiracetam prophylaxis may partly reflect systematic underdosing.

On duration, the evidence tilted cautiously toward brevity. A single small randomized trial of 84 patients comparing three days of levetiracetam with continued treatment until discharge found no difference in in-hospital seizures but more sedation-related discontinuation in the extended group; it was terminated early for slow enrollment and thus underpowered. A retrospective study of 449 patients comparing three-day phenytoin with multiweek courses found the short course reduced hypersensitivity reactions from 8.8 percent to 0.5 percent without any increase in hospital or long-term seizures. Conversely, another retrospective comparison of short-course levetiracetam against extended phenytoin showed more seizures with levetiracetam, though the levetiracetam dose used was low for this population. The pooled analysis favored extended treatment for seizure prevention with a relative risk of 2.45 for short courses, but the adverse event data favored short courses emphatically, with a relative risk of 0.06 for complications. Reconciling these competing signals, the panel suggested individualized decisions, and its practical commentary states plainly that prophylaxis should generally not continue beyond three days in patients without seizures or high-risk electroencephalographic findings, such as elevated scores on the validated 2HELPS2B risk instrument.

Perhaps the most striking feature of the guideline is its candor about what remains unknown. The authors note that the evidence base consists almost entirely of nonrandomized studies with inconsistent seizure definitions, variable drug regimens, minimal drug-level monitoring, and little systematic capture of adverse events or long-term cognitive outcomes. They call for large randomized controlled trials with standardized prophylaxis regimens, therapeutic drug titration, combined clinical and electrographic seizure surveillance, data safety monitoring boards, and subgroup analyses of low- versus high-grade hemorrhage. They also point to troubling signals that phenytoin exposure itself is associated with functional and cognitive disability after subarachnoid hemorrhage, independent of any seizure-prevention benefit. For a field that has prescribed prophylactic anticonvulsants reflexively for decades, the message is paradoxically both liberating and unsettling: doing nothing may be as defensible as treating, and when treatment is chosen, a short, carefully dosed course of levetiracetam—guided where possible by the electroencephalograph rather than tradition—represents the current best judgment of expert consensus in the face of genuinely sparse science.

The protocol was developed through the CLINICCA initiative, a global effort focused on the clinical use of continuous cerebral autoregulation information. A 2023 survey of 44 experts found that approximately half were already using an autoregulation index to influence clinical decisions, but only 39% of those clinicians had incorporated the information into a written local protocol. PRx was the most frequently used index among respondents. The new workgroup therefore invited clinicians who used PRx or PRx-derived cerebral perfusion pressure targets in adult or paediatric traumatic brain injury to edit a common template, share examples of local practice and, where available, submit written protocols. Eight invited experts responded from centres with established protocols or structured practice, while two additional clinicians contributed paediatric applications. After excluding practices unrelated to traumatic brain injury, the authors synthesised the feedback into a five-part framework.

PRx is designed to provide a continuous estimate of how well cerebral blood vessels react to changes in perfusion pressure. Under normal conditions, small arteries and arterioles in the brain constrict or dilate to help maintain relatively stable blood flow despite fluctuations in arterial blood pressure. This buffering process is known as cerebral autoregulation. When autoregulation is impaired after traumatic brain injury, changes in arterial pressure may be transmitted more directly to the cerebral circulation, increasing the risk of inadequate perfusion or pressure-driven swelling. PRx is calculated from slow waves in arterial blood pressure and intracranial pressure, using a moving Pearson correlation over roughly five minutes. A positive correlation generally suggests impaired pressure reactivity, whereas a lower or negative value is more consistent with preserved vascular responsiveness. Because the calculation is global and depends on high-quality, continuously streamed physiological signals, PRx should not be interpreted as a direct measurement of regional blood flow or tissue oxygenation.

The first component of the harmonised protocol uses an elevated or persistently abnormal PRx value as a trigger for clinical review. In the participating centres, a threshold near 0.3 was commonly used, although published studies have examined values ranging from approximately 0.2 to 0.5. When PRx indicates impaired reactivity, clinicians may check the patient’s head and neck position, verify arterial and intracranial pressure transducers, inspect signal quality and review intracranial pressure, cerebral perfusion pressure, carbon dioxide and brain-tissue oxygenation. Other possible contributors include fever, sedation changes, ventilator alterations, seizures, systemic hypotension and evolving intracranial pathology. The workgroup stresses that an abnormal PRx is a warning signal rather than an automatic instruction to intervene. If conventional physiological variables remain acceptable and no corroborating evidence of cerebral hypoperfusion or dangerous intracranial hypertension is present, some centres would continue observation rather than launch a treatment escalation.

The second component concerns intracranial hypertension management and reflects a counterintuitive feature of autoregulation-guided care. When PRx is impaired, raising arterial pressure may not produce the expected reduction in intracranial pressure and can potentially increase cerebral blood volume or worsen pressure transmission. Several contributors therefore consider reducing arterial pressure, cautiously and under defined safety conditions, when impaired reactivity accompanies intracranial hypertension. By contrast, preserved PRx may support a carefully monitored increase in arterial pressure or cerebral perfusion pressure as part of a “MAP or CPP challenge.” If the cerebrovascular bed can constrict appropriately, increasing perfusion pressure may improve vascular tone and help lower intracranial pressure. The protocol does not present either strategy as universal. Each decision must be considered alongside the patient’s neurological examination, imaging, systemic circulation, oxygen delivery and other neuromonitoring data.

The third and fourth components address cerebral perfusion pressure targets derived from PRx. Cerebral perfusion pressure is commonly approximated as mean arterial pressure minus intracranial pressure, and it is a key determinant of the pressure gradient driving blood through the brain. A fixed target may be useful, but it cannot account for the changing autoregulatory state that often follows severe brain injury. One derived target is CPPopt, or optimal cerebral perfusion pressure, the pressure associated with the lowest PRx over a range of observed perfusion pressures. Other approaches estimate the lower limit of reactivity, where autoregulatory capacity begins to fail, or the upper limit of reactivity, where pressure may become excessive. More than half of the contributing centres reported using one or more derived targets to optimise cerebrovascular physiology, although the specific algorithms, thresholds, timing and degree of clinical reliance varied substantially.

The protocol also recognises that perfusion targets may be used to reduce treatment burden rather than simply maximise physiological measurements. If a patient maintains stable cerebral autoregulation and other monitoring signals at a lower pressure, clinicians may be able to avoid unnecessary fluid loading, high doses of vasopressors or prolonged exposure to interventions that can damage the heart, lungs or kidneys. This approach seeks a balance between supporting the injured brain and limiting harm elsewhere in the body. However, the authors caution that PRx is a global summary measure and may behave differently in diffuse injury compared with focal contusions, haematomas or areas of infarction. A pressure that appears favourable globally may not guarantee adequate perfusion in every vulnerable region. For that reason, derived targets should be treated as dynamic decision-support information, not as rigid numbers that override the clinical picture.

The fifth component places PRx within multimodality monitoring, combining it with measurements such as brain-tissue oxygen tension, transcranial Doppler, near-infrared spectroscopy, electroencephalography, arterial carbon dioxide and end-tidal carbon dioxide. In one possible decision pathway, a low brain-tissue oxygen signal accompanied by impaired PRx could prompt assessment of perfusion pressure, oxygen delivery and intracranial pressure together rather than isolated treatment of any single value. Preserved PRx may support a cautious perfusion-pressure challenge when oxygen delivery appears pressure-dependent. Paediatric centres described particularly intensive use of multimodal information, including PRx, transcranial Doppler, near-infrared spectroscopy and brain-tissue oxygenation, to balance sufficient cerebral perfusion against control of intracranial volume. Yet multimodal strategies were among the least consistently adopted elements of the survey, reflecting differences in equipment, staffing, expertise and confidence in how the signals should be combined.

A major message from the workgroup is that technical reliability is inseparable from clinical interpretation. PRx requires continuous acquisition of arterial blood pressure and intracranial pressure at adequate sampling rates, correct signal calibration, reliable time synchronisation and effective removal or recognition of artefacts. Disconnections, damping, flushing, waveform distortion, patient movement and abrupt therapeutic changes can create misleading correlations. The calculation also depends on slow spontaneous fluctuations; a flat or highly unstable signal may generate a number without providing meaningful physiological information. Derived CPP targets can be similarly unreliable when the data contain insufficient variation, when autoregulation is changing rapidly or when the algorithm is applied without quality-control indicators. The supplementary technical addendum accompanying the article provides practical advice for data acquisition, preprocessing and bedside interpretation, but the authors note that standardisation across devices and software remains incomplete.

The workgroup’s findings are likely to attract attention because they capture a field already moving into clinical practice before definitive outcome evidence has arrived. Observational studies have linked cerebrovascular reactivity with prognosis after traumatic brain injury, and a phase-two randomised feasibility trial found that targeting CPPopt was safe and feasible in selected adults. Nevertheless, no strong evidence yet demonstrates that treating patients according to PRx improves long-term neurological outcomes compared with conventional management. The new protocol therefore serves two purposes: it offers clinicians a transparent starting point for local implementation, and it exposes the areas where practice is most uncertain. Agreement was strongest around using PRx to prompt reassessment and to inform intracranial-hypertension management. The greatest variation involved CPP-derived targets and multimodal algorithms. By making these differences visible, the CLINICCA workgroup hopes to encourage collaboration, improve reproducibility and support future prospective trials capable of testing whether autoregulation-guided care can change the course of traumatic brain injury.

Subject of Research: Use of continuous cerebral autoregulation monitoring and the pressure reactivity index (PRx) to guide management of adult and paediatric traumatic brain injury patients with intracranial pressure monitoring.

Article Title: A Harmonised Protocol for the Use of PRx in Adult and Paediatric TBI Patients with ICP Monitoring: CLINICCA Workgroup Output

Article References: Beqiri E, Lavinio A, Dias C, et al. “A Harmonised Protocol for the Use of PRx in Adult and Paediatric TBI Patients with ICP Monitoring: CLINICCA Workgroup Output.” Neurocritical Care (2026). Related references include Czosnyka et al., “Continuous assessment of the cerebral vasomotor reactivity in head injury,” Neurosurgery (1997); Tas et al., the COGiTATE feasibility randomized controlled trial, Journal of Neurotrauma (2021); and Beqiri et al., studies of CPPopt and the lower limit of reactivity.

Image Credits: AI Generated

DOI: 10.1007/s12028-026-02595-z

Keywords: PRx, CPPopt, cerebral perfusion pressure, cerebral autoregulation, traumatic brain injury, intracranial pressure, paediatric neurocritical care, multimodal monitoring, lower limit of reactivity, upper limit of reactivity.