Serial serum biomarkers at 72 h are associated with subsequent brain death/death by neurologic criteria after acute brain injury

Background Circulating brain injury biomarkers have been widely studied for predicting mortality and neurological outcome, but their prognostic value for subsequent brain death/death by neurologic criteria (BD/DNC) remains less well established. We evaluated the association and prognostic discrimination of serial serum neuron-specific enolase (NSE) and S100 calcium-binding protein B (S100B) measurements for subsequent BD/DNC in patients with acute brain injury requiring intensive care. Methods This prospective observational study included 107 adults, of whom 41 progressed to BD/DNC during hospitalization. NSE and S100B were measured early after injury and at the scheduled 72-h assessment. Twelve patients died before repeat sampling; 72-h analyses therefore used a landmark population of 95 patients. Biomarkers were compared with concurrent 72-h Glasgow Coma Scale (GCS), GCS motor component, and Sequential Organ Failure Assessment (SOFA), and parsimonious logistic models examined complementary information beyond concurrent GCS. Results In paired landmark analyses, NSE AUC increased from 0.769 initially to 0.941 at 72 h ( p = 0.0014), and S100B AUC from 0.820 to 0.976 ( p < 0.001). Neither 72-h biomarker differed significantly from concurrent total GCS. In the GCS-plus-S100B model, both S100B (adjusted OR 1.963 per 1 μg/L, 95% CI 1.215–3.171; p = 0.006) and GCS (adjusted OR 0.226 per point, 95% CI 0.058–0.882; p = 0.032) remained independently associated with subsequent BD/DNC; the optimism-corrected AUC was 0.986. Among the 33 patients who progressed after the 72-h landmark, the median interval from sampling to BD/DNC was 20.8 h, 18 (54.5%) progressed within 24 h, and 24 (72.7%) had bilaterally non-reactive pupils at the concurrent assessment. Conclusion Serial NSE and S100B showed stronger discrimination at 72 h than initially but did not outperform concurrent neurological examination. The short interval to BD/DNC and frequent pupillary non-reactivity indicate that 72-h values may often reflect already advanced injury; these findings support a complementary role within multimodal neuroprognostication rather than prediction of a temporally remote event.