This retrospective analysis used ICU patient records from two large critical-care databases: the Medical Information Mart for Intensive Care IV (MIMIC‑IV) and the eICU Collaborative Research Database (eICU‑CRD). The study population comprised adults with non‑traumatic subarachnoid hemorrhage (NSAH). Depending on the MBPV indicator and time window, the final analytic sample size ranged from 1,345 to 1,820 patients. The primary endpoint was all‑cause in‑hospital mortality.
Mean arterial pressure (MAP) measurements were extracted for three early admission windows: 24, 48, and 72 hours after ICU admission. Four MBPV metrics were computed from those MAP recordings:
These MBPV metrics provided complementary perspectives on blood pressure fluctuation: SD and range capture overall dispersion, ARV summarizes absolute sequential changes, and SV emphasizes immediate successive differences.
The investigators evaluated nonlinear associations between each MBPV metric and in‑hospital mortality using restricted cubic splines (RCS) incorporated into Cox proportional hazards models. Piecewise regression was applied to identify risk inflection points along nonlinear curves. Where the proportional hazards assumption was violated, time‑dependent Cox regression models were fitted to characterize how the hazard associated with an MBPV indicator changed over time.
Across all MBPV metrics and time windows, the association between early MBPV and in‑hospital mortality was U‑shaped (all P < 0.05). This pattern indicates that both relatively low and relatively high early MAP variability were associated with increased risk of in‑hospital death compared with intermediate levels of variability. Among the evaluated windows, MBPV measured within the first 24 hours after admission demonstrated the strongest and most stable association with mortality.
For MBP SD, piecewise regression identified an inflection point at 8.9. Values of SD below or above this threshold were associated with higher mortality; in other words, very low early MAP variability and very high early MAP variability both predicted worse in‑hospital survival relative to moderate variability around that inflection. The authors interpret this U‑shaped relationship as suggesting that both excessive stabilization and excessive instability of arterial pressure early after NSAH may be harmful, whereas moderate fluctuation may be associated with better outcomes.
Time‑dependent Cox analyses revealed differing temporal patterns across MBPV metrics. The adverse associations for ARV and SV became more prominent after 4–7 days post‑admission, indicating that sequential variability measures may exert increasing hazard later in the hospital course. In contrast, the risk effects associated with SD and range were concentrated mainly within the first 14 days after admission. These temporal distinctions suggest that different MBPV metrics may carry prognostic information at different phases after NSAH.
Two prespecified sensitivity analyses were performed to test the robustness of the main findings. One analysis excluded patients who underwent surgery within 72 hours of admission; the other excluded patients with missing medication data. Results from both sensitivity checks were consistent with the primary analysis, supporting the stability of the observed U‑shaped associations and the relative strength of the 24‑hour MBPV signals.
Early mean arterial pressure variability (MBPV) after non‑traumatic subarachnoid hemorrhage demonstrates a reproducible U‑shaped association with in‑hospital mortality in two large ICU datasets. Moderate early MBPV was associated with the best in‑hospital survival, whereas both low and high extremes of variability increased mortality risk. Among evaluated metrics, MBP SD within 24 hours had the strongest and most consistent relationship with mortality and is proposed by the authors as a preferred indicator for early blood pressure monitoring after NSAH.
Note: The article provides these findings based on MIMIC‑IV and eICU‑CRD analyses and reports the described statistical results; no additional clinical trial data, intervention effects, or external validation details beyond those in the original report were included in the source abstract.