PaCO2 Association with Traumatic Brain Injury Patients Outcomes at High Altitude: A Prospective Single-Center Cohort Study

Background partial pressure of carbon dioxide (PaCO2) is generally known to influence outcome in patients with traumatic brain injury (TBI) at normal altitudes. Less is known about specific relationships of PaCO2 levels and clinical outcomes at high altitudes. Methods This is a prospective single-center cohort of consecutive TBI patients admitted to a trauma center located at 2600 meter above sea level. An unfavorable outcome was defined as the Glasgow Outcome Scale-Extended (GOSE) < 4 at 6-month follow-up. Results 81 patients with complete data, 80% (65/81) were men, and median (IQR) age was 36 (25–50) years). Median Glasgow Coma Scale (GCS) on admission was 9 (6–14), 49% (40/81) were severe (GCS: 3–8), 32% (26/81) moderate (GCS 12 − 9), and 18% (15/81) mild (GCS 13–15) TBI. The median (IQR) Abbreviated Injury Score of the Head (AISh) was 3 (2–4). Frequency of an unfavorable outcome (GOSE < 4) was 30% (25/81), median GOSE was 4 (2–5), and 6-month mortality was 24% (20/81). Comparison between patients with favorable and unfavorable outcomes revealed that those with unfavorable outcome were older, median [49 (30–72) vs. 29 (22–41), P < 0.01], had lower admission GCS [6 (4–8) vs. 13 (8–15), P < 0.01], higher AIS head [4 (4–4) vs. 3(2–4), p < 0.01], higher APACHE II score [17(15–23) vs 10 (6–14), < 0.01), higher Charlson score [0(0–2) vs. 0 (0–0), P < 0.01] and higher PaCO2 (mmHg), mean ± SD, 39 ± 9 vs. 32 ± 6, P < 0.01. In a multivariate analysis, age (OR 1.14 95% CI 1.1–1.30, P < 0.01), AISh (OR 4.7 95% CI 1.55–21.0, P < 0.05), and PaCO2 (OR 1.23 95% CI: 1.10–1.53, P < 0.05) were significantly associated with the unfavorable outcomes. When applying the same analysis to the subgroup on mechanical ventilation, AISh (OR 5.4 95% CI: 1.61–28.5, P = 0.017) and PaCO2 (OR 1.36 95% CI: 1.13–1.78, P = 0.015) remained significantly associated with the unfavorable outcome. Conclusion Higher PaCO2 levels are associated with an unfavorable outcome in ventilated TBI patients. These results underscore the importance of PaCO2 level in TBI patients and whether it should be adjusted for populations living at higher altitudes.


INTRODUCTION
Traumatic Brain Injury (TBI) accounts for a substantial global health burden, with approximately 27 million cases reported annually, particularly in low-and middle-income countries (LMICs) (1,2).As many as 50% of individuals with TBI do not regain their previous functionality (3), resulting in a reported age-standardized incidence rate of 111 (82-141) years lived with disability per 100,000 (4).The most frequently cited factors related to poor outcomes include age, trauma severity, and the Glasgow Coma Scale (GCS) at presentation.Other factors, such as imaging ndings, hypoxia, hypo or hypercapnia, and hypotension, have also been identi ed.(5,6,7).These ndings have allowed clinical teams and guidelines to establish goals in the acute setting to optimize care to limit secondary brain injury.These goals often include speci c hemodynamic and respiratory parameters to achieve a particular target, such as optimal levels of partial pressure of carbon dioxide (PaCO2) (8,9).
Carbon dioxide plays a central role in regulating cerebral blood ow, a notion supported by animal and human studies (10).Hypercapnia causes blood vessels to dilate due to cerebrospinal uid acidosis and the direct effect of extracellular H + on vascular smooth muscle (11), while hypocapnia constricts them via alkalosis, in uencing intracranial pressure and adjusting brain tissue perfusion in response to the environment (12).Maintaining optimal partial pressure of carbon dioxide (PaCO2) levels is crucial in cases of brain injury, as hypoperfusion and hypoxemia are closely linked to secondary brain injury and long-term consequences, impacting disability and survival rates (13,14).Guidelines recommend maintaining a target PCO2 range between 35-45 mm Hg to prevent cerebral ischemia in the case of low PaCO2 or hyperemia that could lead to elevated intracranial pressure if PaCO2 is high (15).Several studies have reinforced this concept of targeting a speci c range of PaCO2 as a goal of care for TBI patients in the neuro-intensive care unit (NICU) (16) and its potential systemic implications (17,18).There is also considerable variability in the management of PaCO2 in TBI patients within regions and centers (19).Furthermore, evidence indicates that normal PaCO2 levels can vary according to altitude and barometric pressure (20,21).Generally, the barometric pressure is 760 mmHg at sea level, with PaCO2 levels between 35-45 mmHg considered normal (21).At higher altitudes, the atmospheric pressure of O2 and CO2 is lower, reducing PaO2 and PaCO2 (alveolar pressure), which in turn stimulates alveolar ventilation (22,23).The implications of these differences on the physiology and management of patients with TBI are unclear.Further contributions in this area may help guide the management and care of this patient population (Fig. 1).
We hypothesize that the TBI population at higher altitudes may bene t from different PaCO2 level targets compared to sea-level populations.Additionally, we hypothesized that the initial management of respiratory care and support in the acute phase might in uence outcomes.This study evaluates the association between admission PaCO2 levels and outcomes at 6-month follow-ups in patients with TBI admitted to the NICU.

MATERIALS AND METHODS
The study was approved by the Institutional Review Board/Independent Ethics Committee (IRB/EC) under the local regulations and the Declaration of Helsinki for clinical practices, including obtaining informed consent from the patient representative.All clinical data were anonymized and collected using the Research Electronic Data Capture (REDCap), an electronic data collection form provided by the Universidad de La Sabana.

Study Population
This single-center prospective cohort study was conducted in a trauma center at the Universidad de La Sabana in Chía, Colombia.We consecutively recruited and collected data from TBI patients admitted to the NICU from December 2019 to June 2022.The diagnosis, inclusion, and exclusion criteria, as well as imaging studies, were obtained by chart review.
The study cohort included ≥ 18-year-old TBI patients admitted to the NICU within 24 hours post-injury and who stayed in the NICU for more than 48 hours.Patients with a previous history of disability or debilitating diseases measured by a modi ed Rankin Scale (mRS) > 2 and those admitted after 24 hours post-injury were excluded.

De nitions
To evaluate the severity of TBI, we utilized the Abbreviated Injury Scale of the head (AISh).We chose to use AISh because GCS was often obscured by sedation at the injury site or upon admission to the emergency department (ED).AISh incorporates both clinical and imaging ndings (24,25), enabling a more nuanced assessment of the severity of the lesion (Table S1) and provides a robust correlation with outcomes.The AIS ranks injury on an ordinal scale 0 to 6 (from no injury to fatal).AIS can be classi ed as 1 (minor injury), 2 (moderate), 3 (serious), 4 (severe), 5 (critical) or 6 (fatal) (26, 27).
To assess the severity of the traumatic injury overall, the injury severity score (ISS) was used.The ISS is a composite measure derived from the AIS score that includes a rating of the three most severely injured body regions and ranges from 0 to 75.An ISS of 15 or higher is usually considered major trauma, and the compromise of 2 or more body regions with an AIS ≥ 3 is considered multiple trauma (27).
To characterize the severity of the brain injury in a head CT scan, we used the Marshall classi cation.The Marshall scheme was rst published in 1992 and uses six categories (I to VI) of increasing severity based on non-contrast CT scan ndings, including midline shift, compression of cisterns, and mass lesions (28, 29) (Table S2).Its correlation with outcomes in TBI has been validated in several studies (30,31) The International Mission for Prognosis and Analysis of Clinical Trials in TBI (IMPACT) is a prognostic model that uses baseline characteristics and provides a probability of an unfavorable outcome and mortality at 6 months (Table S3).It de nes an unfavorable outcome as a Glasgow Outcome Scale of 1-3.The IMPACT model has accurately discriminated outcomes after TBI (32,33).We used the lab model that includes age, motor score of the GCS, pupillary reactivity, CT characteristics, and information on admission hemoglobin and glucose.(33).
To evaluate mortality and disability as outcomes, we selected the Glasgow Outcome Scale-Extended (GOSE), as outlined in Table S4, which is an ordinal scale of eight points ranging from death to good recovery (34).GOSE has been used widely to assess outcomes in TBI (35,36,37).A trained staff administered GOSE through a standardized phone interview with the patients or their caregivers 6 months post-injury.For the analysis, we dichotomized GOSE into favorable and unfavorable outcomes.A favorable outcome (GOSE ≥ 4) was considered for those with upper severe disability to upper good recovery, and an unfavorable outcome was de ned as a lower severe disability to death (GOSE < 4).

Data Collection
Upon admission to the NICU, demographic data and trauma severity and prognostication scales that include the GCS, ISS, IMPACT model, and Marshall CT scan classi cation were recorded consecutively and prospectively.We collected admission vital signs and lab tests, which were reviewed and con rmed directly from the electronic medical record.Medical interventions during NICU stay, including mechanical ventilation, blood components transfusion, and use of vasopressors within 72 h of admission were reported.Finally, infections in the NICU, total hospital and NICU length-of-stay (LOS), and hospital mortality were also recorded.At 6-month follow-up, patients or their legal representatives were contacted via phone by a trained research team member to administer GOSE.

Statistical Analysis
Continuous variables were summarized based on clinical relevance and distribution using minimum and maximum values, mean ± standard deviation (SD), or median and interquartile range (IQR).Dichotomous variables were presented as frequencies and percentages.Differences between intervention groups were assessed by applying the chi-square and Fisher's exact tests for categorical variables.In contrast, continuous variables were evaluated using the student's t-test or Mann-Whitney U test, depending on their distribution.
A multivariate logistic regression model was constructed for the general cohort to investigate the risk factors associated with unfavorable outcomes at 6-month follow-up.The model was adjusted for admission demographic data, vital signs, and lab tests.The logistic regression used the best subset method for the variable selection and included variables with a P value of less than 0.10 in the univariate analysis.Odds ratios (OR) with a 95% con dence interval (95% CI) were calculated based on the exponential values of the coe cients obtained from the nal model D. We used R Studio (Version 2023.09.1 + 494) for the analysis.
Comparison between patients with a 6-month favorable outcome and unfavorable outcome (Table 1) revealed that those with an unfavorable outcome were older [49 (30-72) 1).

PaCO2 and Outcome for Patients on Mechanical Ventilation
When evaluating the group on mechanical ventilation (n = 49), the PaCO2 mean ± SD was 39.0 ± 7.7 mmHg, which was signi cantly higher for those with a 6-month unfavorable outcome compared to the group with a favorable outcome (42.0 ± 7.8 vs. 35.3± 4.4, P < 0.01, Fig. 3).In the group without ventilatory support, the PaCO2 mean ± SD was 28.1 ± 5.8 mmHg, and it was signi cantly lower for the group with an unfavorable outcome (21.6 ± 2.5 vs 28.9 ± 5.6, P < 0.01) compared to those with favorable outcome at 6 months.Mean PaCO2 was lower in the group without ventilator support than those on mechanical ventilation (28.1 ± 5.8 vs. 39.0 ± 7.7, P < 0.001).Finally, NICU LOS was longer for the unfavorable outcome group, 14 days (6-23) vs. 5 days (4-8), P < 0.01.

Logistic regression analysis
Univariate analysis of in-hospital variables and their association with a 6-month unfavorable outcome were performed through a univariate logistic regression (P < 0.1) ( Afterward, the same analysis was applied to the subgroups of patients with and without ventilator support.A multivariate analysis was performed on the mechanical ventilation group using the same variables: age, AISh, APACHE II, and PaCO2.In this case, again, AIS head (OR 5.4, 95% CI: 1.61-28.5,P = 0.017) and PaCO2 (OR 1.36, 95% CI: 1.13-1.78,P = 0,015) remained signi cantly associated with the 6month unfavorable outcome (Table 3).The same analysis for the group without mechanical ventilation did not yield a signi cant result for any of the variables (P = 1).Hosmer-Lemeshow test for binary logistic regression models demonstrated the goodness-of-t test (P = 0.97).

DISCUSSION
This study initially characterizes a prospective cohort of patients with TBI admitted to the NICU in an academic center in the Andean region in Colombia.The group with an unfavorable outcome was older, lower GCS on admission, higher AISh, higher probability of an unfavorable outcome by the IMPACT-TBI model, higher APACHE II, and higher Charlson score.Among vital signs and laboratory data, the only documented difference was a higher PaCO2 on admission for those with an unfavorable outcome.In terms of in-hospital procedures, the group with an unfavorable outcome required more ventilatory and hemodynamic support, underwent neurosurgical interventions and tracheostomy more often, and had a longer LOS in the NICU.After adjusting for age, severity of TBI, and APACHE II, PaCO2 remained directly correlated with an unfavorable outcome at 6 months.A higher PaCO2 was associated with an unfavorable 6-month outcome for all the study groups and the group on ventilatory support.In the subgroup, without ventilatory support, this correlation was not maintained.The mean PaCO2 in the subgroup without ventilatory support was lower than those on mechanical ventilation.The lower PaCO2 levels observed in the non-ventilated group may be associated with the inherently lower baseline levels of PaCO2 in populations residing at higher altitudes.Consequently, this suggests a potential difference in the way regulatory mechanisms are established (10,13).
The demographic characteristics of the studied cohort are similar to what others have found in terms of age and cause of trauma (42,43).TBI affects predominantly the adult male population in their fourth or fth decade of life, and the leading causes of injury are road accidents and falls.This has been consistent in several prospective studies, including the European and Chinese cohorts of CENTER-TBI and the TRACK-TBI for the US (5,42,44).Regarding mortality and functional outcomes, the ICU stratum of the European Center-TBI found 43.1% and 21.3% rates of an unfavorable outcome (GOSE < 5) and mortality, respectively.The results in our study are similar in both mortality (24%) and unfavorable outcome (30%), bearing in mind that the de nition we used for unfavorable outcome was GOSE < 4 (44).There is no standardized manner to dichotomize GOSE, and de nitions vary across studies (45,46).TBI patients might show functional and cognitive improvement even 1 year after the trauma (47,48), depending on their recovery trajectory.GOSE equal to 4 refers to a person who requires partial supervision and assistance but can be on their own at home for at least 8 hours a day.Therefore, we considered it reasonable to de ne GOSE ≥ 4 as the favorable outcome, considering that those patients are already partially independent at home and still have the potential for further progress.
Several studies have pointed out that older and more severely injured TBI patients have more frequent severe disability and functional dependence after TBI (49,50).Moderate and severe TBI cases are usually admitted to the NICU, where interventions are guided by targets that aim to protect the brain from a secondary injury (51).Henceforth, it is also the more severely traumatized patient who needs more assistance in terms of respiratory, hemodynamic, and metabolic support as well as surgical interventions (52,53).In our cohort, the group with unfavorable outcomes was older and had a more severe TBI on admission.Therefore, it could be expected that it is, in turn, the group that received a higher burden of care, including mechanical ventilation, vasopressors, neurosurgical and tracheostomy procedures, and was more exposed to complications like in-hospital infections and longer ICU stays.This re ects the complexity of treatment and prognosis when many factors are involved, leaving aside the variability of management across centers and regions (53).Despite this challenge, some prognostic models have been developed and validated, for instance, The Corticosteroid Randomization After Signi cant Head Injury (CRASH) model and the International Mission for Prognosis and Analysis of Clinical Trials (IMPACT) in TBI model (54,55,56).These models estimate the probability of disability and mortality and consider factors such as age, Glasgow motor score, pupillary reactivity, and imaging ndings on head CT scans.
We did not intend to develop a model, but we did identify some factors on admission associated with outcomes, including age, severity of TBI, APACHE II, and need for hemodynamic and ventilatory support.However, when assessing vital signs and laboratory tests, higher levels of PaCO2 on admission were associated with the unfavorable outcome, even after controlling for the age and severity of the injury.The role of PaCO2 in this context relies on its effect on the cerebral vasculature or vasoreactivity (57,58).The brain has high metabolic demand, requiring a constant supply of oxygen and glucose (59).This supply is ensured through a tightly regulated cerebral blood ow that matches each brain region's temporal and spatial metabolic requirements (60).One of those mechanisms is the vasomotor response to carbon dioxide, where cerebral arterioles dilate or contract according to changes in PaCO2.This response has a sigmoidal shape and functions within the 20-60 mmHg of PaCO2.Every 1 mmHg increase in PaCO2 corresponds to roughly a 4% increase in cerebral blood ow (61, 62), which in turn increases the cerebral blood volume resulting in an intracranial pressure elevation and nally affecting the cerebral perfusion pressure.Several cohorts have demonstrated the effect of PaCO2 management on outcomes, including mortality (21).However, variability in management exists across centers (63).Guidelines recommend a normal range ventilation, PaCO2 35-45 mmHg, and avoidance of hyperventilation and severe (< 25 mmHg) or moderate (< 30 mmHg) hypocapnia (8, 9) given the risk of brain ischemia.
In our cohort, we found a higher PaCO2 for those cases with an unfavorable outcome, and the multivariate analysis revealed a direct relation between admission PaCO2 levels and the probability of death and disability.The association remained for the subgroup on mechanical ventilation but not for those patients without ventilatory support.This could be expected given that PaCO2 in a ventilated patient depends mostly on the ventilator settings and can be adjusted to a speci c goal.However, we would like to point out that most of our patients had PaCO2 levels within the recommended range of 35-45 mmHg and even below for those with a favorable outcome, 32 ± 6 mmHg.In addition, non-ventilated patients had even lower PaCO2 levels.These results underscore the importance and impact of PaCO2 as a crucial target in the management of ventilated TBI patients and raise the question of whether, for populations at higher altitudes, different PaCO2 goals should be pursued.Further investigation would be needed to answer this question, which will bene t a substantial proportion of the global TBI population living at higher altitudes.
Limitations of our study include a single-center study that requires further validation to make the results more generalizable.In addition, we only recorded the admission PaCO2 values rather than serial values.

CONCLUSION
We evaluated the relationship between PaCO2 levels and functional outcomes in patients with TBI admitted to the ICU.Interestingly, in our center, situated at a higher altitude above sea level, we observed that in the sample of patients on mechanical ventilation, a PaCO2 below the recommended target was associated with improved outcomes.Although this is a single-center prospective cohort study, it raises the question of whether the target PaCO2 levels need adjustment in populations at higher altitudes.

Figure 1 Effect
Figure 1

Table 1
Baseline characteristics of patients admitted to NICU for TBI and comparison between groups with favorable and unfavorable outcomes.
Abbreviations: AIS head: Abbreviated Injury Score of the head, DI I: Diffuse Injury I, DI II: Diffuse Injury II, DI III: Diffuse Injury III, DI IV: Diffuse Injury IV, EML V: Evacuated Mass Lesion V, GCS: Glasgow Coma Scale, IMPACT TBI: International Mission for Prognosis and Analysis of Clinical Trials in Traumatic Brain Injury, ISS: Injury Severity Score, NEML VI: Non.-evacuated Mass Lesion VI, NICU: Neuro-Intensive Care Unit, SBP: Systolic Blood Pressure, WBC: White Blood Cell Count.

Table 2
Univariate and multivariate analysis of variables on admission and a 6-month unfavorable outcome in patients with TBI admitted to NICU.

Table 3
Multivariate analysis of clinical variables and a 6month unfavorable outcome in TBI patients admitted to the NICU on mechanical ventilation.