RESPIRATORY RESPONSE AND NEUROLOGICAL TOLERANCE OF PRONE POSITIONING IN BRAIN-INJURED PATIENTS WITH ACUTE RESPIRATORY DISTRESS SYNDROME

Background/Objective: Prone positioning is established therapy for moderate-to-severe acute respiratory distress syndrome (ARDS), but its use after acute brain injury remains controversial because it may increase intracranial pressure (ICP). Where invasive ICP monitoring is unavailable, transcranial Doppler (TCD) may provide a non-invasive assessment of cerebral hemodynamic tolerance. We evaluated respiratory and neurological tolerance of proning in brain-injured patients with ARDS.
Methods: We retrospectively studied patients aged ≥16 years who developed ARDS after brain injury in our ICU from January 2020 to March 2023. Patients were grouped by proning exposure. Proning required neurological stability and absence of a pathological TCD pattern, defined as pulsatility index >1.4 with middle cerebral artery diastolic velocity <20 cm/s. TCD and pupillary assessments were repeated twice daily. A pathological pattern suggested intracranial hypertension only after optimization of PaCO₂ and invasive arterial blood pressure. The primary outcome was change in PaO₂/FiO₂.
Results: Of 613 brain-injured patients, 29 developed ARDS; 14 were proned and 15 were not. In proned patients, median PaO₂/FiO₂ increased from 133 (62–182) to 222 (185–257), improving in all 14 (p<0.001). TCD documentation sufficient to assess cerebral hemodynamic tolerance was available in eight patients: seven had no new pathological pattern and one developed findings suggestive of intracranial hypertension, prompting osmotherapy then external ventricular drainage. No new pupillary abnormality occurred. ICU mortality was 42.9% versus 60.0% (risk ratio 0.71, 95% CI 0.34–1.49; p=0.47).
Conclusions: In this selected cohort, prone positioning improved oxygenation without an evident signal of excess neurological intolerance. Serial TCD may be a pragmatic non-invasive adjunct where invasive ICP monitoring is unavailable. The small sample, retrospective single-center design, incomplete TCD data and selection bias preclude causal conclusions; these findings are hypothesis-generating.