The Association Between Arterial Oxygen Level and Outcome in Neurocritically Ill Patients is not Affected by Blood Pressure


By Jaana Humaloja, Markus B. Skrifvars, Rahul Raj, Erika Wilkman, Pirkka T. Pekkarinen, Stepani Bendel, Matti Reinikainen & Erik Litonius

First Online: 05 January 2021


In neurocritically ill patients, one early mechanism behind secondary brain injury is low systemic blood pressure resulting in inadequate cerebral perfusion and consequent hypoxia. Intuitively, higher partial pressures of arterial oxygen (PaO2) could be protective in case of inadequate cerebral circulation related to hemodynamic instability.

Study purpose

We examined whether the association between PaO2 and mortality is different in patients with low compared to normal and high mean arterial pressure (MAP) in patients after various types of brain injury.


We screened the Finnish Intensive Care Consortium database for mechanically ventilated adult (≥ 18) brain injury patients treated in several tertiary intensive care units (ICUs) between 2003 and 2013. Admission diagnoses included traumatic brain injury, cardiac arrest, subarachnoid and intracranial hemorrhage, and acute ischemic stroke. The primary exposures of interest were PaO2 (recorded in connection with the lowest measured PaO2/fraction of inspired oxygen ratio) and the lowest MAP, recorded during the first 24 h in the ICU. PaO2 was grouped as follows: hypoxemia (< 8.2 kPa, the lowest 10th percentile), normoxemia (8.2–18.3 kPa), and hyperoxemia (> 18.3 kPa, the highest 10th percentile), and MAP was divided into equally sized tertiles (< 60, 60–68, and > 68 mmHg). The primary outcome was 1-year mortality. We tested the association between hyperoxemia, MAP, and mortality with a multivariable logistic regression model, including the PaO2, MAP, and interaction of PaO2*MAP, adjusting for age, admission diagnosis, premorbid physical performance, vasoactive use, intracranial pressure monitoring use, and disease severity. The relationship between predicted 1-year mortality and PaO2 was visualized with locally weighted scatterplot smoothing curves (Loess) for different MAP levels.


From a total of 8290 patients, 3912 (47%) were dead at 1 year. PaO2 was not an independent predictor of mortality: the odds ratio (OR) for hyperoxemia was 1.16 (95% CI 0.85–1.59) and for hypoxemia 1.24 (95% CI 0.96–1.61) compared to normoxemia. Higher MAP predicted lower mortality: OR for MAP 60–68 mmHg was 0.73 (95% CI 0.64–0.84) and for MAP > 68 mmHg 0.80 (95% CI 0.69–0.92) compared to MAP < 60 mmHg. The interaction term PaO2*MAP was nonsignificant. In Loess visualization, the relationship between PaO2 and predicted mortality appeared similar in all MAP tertiles.


During the first 24 h of ICU treatment in mechanically ventilated brain injured patients, the association between PaO2 and mortality was not different in patients with low compared to normal MAP.

Read full article here.