Clinical practice and effect of carbon dioxide on outcomes in mechanically ventilated acute brain-injured patients: a secondary analysis of the ENIO study

Abstract

Purpose: The use of arterial partial pressure of carbon dioxide (PaCO₂) as a target intervention to manage elevated intracranial pressure (ICP) and its effect on clinical outcomes remain unclear. We aimed to describe targets for PaCO₂ in acute brain injured (ABI) patients and assess the occurrence of abnormal PaCO₂ values during the first week in the intensive care unit (ICU). The secondary aim was to assess the association of PaCO₂ with in-hospital mortality.

Methods: We carried out a secondary analysis of a multicenter prospective observational study involving adult invasively ventilated patients with traumatic brain injury (TBI), subarachnoid hemorrhage (SAH), intracranial hemorrhage (ICH), or ischemic stroke (IS). PaCO₂ was collected on day 1, 3, and 7 from ICU admission. Normocapnia was defined as PaCO₂ > 35 and to 45 mmHg; mild hypocapnia as 32-35 mmHg; severe hypocapnia as 26-31 mmHg, forced hypocapnia as < 26 mmHg, and hypercapnia as > 45 mmHg.

Results: 1476 patients (65.9% male, mean age 52 $\pm$ 18 years) were included. On ICU admission, 804 (54.5%) patients were normocapnic (incidence 1.37 episodes per person/day during ICU stay), and 125 (8.5%) and 334 (22.6%) were mild or severe hypocapnic (0.52 and 0.25 episodes/day). Forced hypocapnia and hypercapnia were used in 40 (2.7%) and 173 (11.7%) patients. PaCO₂ had a U-shape relationship with in-hospital mortality with only severe hypocapnia and hypercapnia being associated with increased probability of in-hospital mortality (omnibus p value = 0.0009). Important differences were observed across different subgroups of ABI patients.

Conclusions: Normocapnia and mild hypocapnia are common in ABI patients and do not affect patients' outcome. Extreme derangements of PaCO₂ values were significantly associated with increased in-hospital mortality.

Keywords: Brain injury; Carbon dioxide; Critical care; Hyperventilation; ICH; Intensive care; Invasive ventilation; PaCO2; SAH; Stroke; TBI.

Fig. 1

Number of patients experiencing abnormally low and high partial pressure of carbon dioxide-PaCO₂-episodes during the ICU stay. The bars indicate the number of patients with 0, 1, 2, and 3 episodes of abnormal PaCO₂. Incidence rates (as number of episodes per 1 person-day) of hypocapnia and hypercapnia during ICU stay are provided. IR, incidence rate; CI, Confidence Interval

Fig. 2

Sankey plot providing a visual representation of the transition of patients with different categories of abnormal PaCO₂ at day 1, 3, and 7. The different colors show the number of patients in each PaCO₂ category and how they changed from day 1 to day 7. Most of the transitions occurring in the hypocapnia group were toward normocapnia. Most patients with forced hypocapnia remained in the same category, whereas most patients with hypercapnia evolved toward normocapnia. Overall, the pattern of flows between days 3 and 7 was similar to the pattern of flows between days 1 and 3

Fig. 3

A Association between PaCO₂ (modeled as a continuous variable) and in-hospital mortality using a longitudinal survival regression. PaCO_2c was modeled with 5-df restricted cubic splines (RCS). For esthetic purposes, PaCO₂ distribution was trimmed between 20 mmHg and 60 mmHg. The graph shows the adjusted hazard ratio (aHR) for in-hospital mortality on the y-axis across the full range of PaCO₂ values on the x-axis. The solid line represents the hazard ratio and the shaded area is the 95% confidence interval. B The association between categorical PaCO₂ values and in-hospital mortality risk using a longitudinal survival regression. The graph shows the hazard ratios with 95% confidence intervals using normocapnia a reference category

Fig. 4

a and b Relative distribution analysis aimed at exploring the relationship between the entire distribution of partial pressure of carbon dioxide, PaCO_2c with in-hospital mortality. The horizontal axis shows the cumulative proportion of the reference group (PaCO_2c values when in-hospital mortality = 0). The vertical axis shows the relative density, which is the ratio of densities between the two distributions at each percentile of the reference group. The 95% confidence intervals indicate whether the relative density at each percentile is significantly different from 1