Association between arterial to end-tidal carbon dioxide gradient and prehospital hemodynamic status in patients with moderate-to-severe traumatic brain injury

Abstract

Introduction

We aimed to assess the relationship between arterial partial pressure of carbon dioxide (PaCO₂) and end-tidal carbon dioxide (EtCO₂) according to three prehospital hemodynamic status and to evaluate whether hemodynamic-based EtCO₂ targets could improve agreement with PaCO₂.

Methods

Over a 4-year period, 100 adult patients with moderate-to-severe TBI admitted to a level 1 trauma center under mechanical ventilation were retrospectively included. Patients were classified into three clusters based on prehospital hemodynamic status (Cluster 1: Stable without vasopressors; Cluster 2: Stabilized with low-dose vasopressors; Cluster 3: Not stabilized with low-dose vasopressors). Correlation and agreement between paired PaCO₂ and EtCO₂ values were assessed overall and across clusters. Agreement was compared between non-adjusted and cluster-adjusted EtCO₂ thresholds.

Results

The mean PaCO₂ - EtCO₂ gradient increased according to the hemodynamic status (Cluster 1: 3.5 ± 5.5 mmHg, Cluster 2: 9.9 ± 7.3 mmHg, Cluster 3: 12.6 ± 9.5 mmHg). The agreement between paired PaCO₂ and non-adjusted EtCO₂ levels decreased between the three clusters (61% [95%CI: 44–77%] in Cluster 1, 42% [95%CI: 25–59%] in Cluster 2, 31% [95%CI: 16–46%] in Cluster 3). Adjusting EtCO₂ targets according to hemodynamic status did not improve agreement between pairs (53% vs. 44%, p = 0.204).

Conclusion

In moderate-to-severe TBI, there is a strong relationship between the PaCO₂–EtCO₂ gradient and prehospital hemodynamic status. However, our results did not support the relevance of a prehospital ventilation strategy guided by cluster-adjusted EtCO₂ targets.

Introduction

In patients with severe traumatic brain injury (TBI), a prehospital physician-staffed trauma system enables early interventions aimed at optimizing cerebral perfusion and limiting secondary brain injury. Current guidelines recommend the administration of vasopressors when a restrictive fluid resuscitation fails to achieve a target mean arterial pressure (MAP) ≥ 80 mmHg [1]. Moreover, the early use of mechanical ventilation is needed to secure the airway and ensure adequate oxygen delivery. However, careful adjustment of ventilatory settings is essential to prevent hypercapnia or hypocapnia, both associated with increased in-hospital mortality [2, 3].

In the prehospital settings, end-tidal carbon dioxide (EtCO₂) monitoring is commonly used as a surrogate for the arterial partial pressure of carbon dioxide (PaCO₂). An EtCO₂ between 30 and 35 mmHg is recommended prior to getting arterial gas samples when aiming for a normocapnic ventilation, assuming a 0 to 5 mmHg gradient between EtCO₂ and PaCO₂ [4, 5]. On the other hand, previous studies reported a wide inter- and intra-variability in the PaCO₂ - EtCO₂ gradient, especially in trauma patients with hemodynamic instability [6–8]. In this regard, relying on EtCO₂ to adjust prehospital ventilation can lead to iatrogenic hypercapnia and might thereby impact patient outcomes [9, 10].

Despite the physiological relationship between EtCO₂ and cardiac output, few studies have explored how the severity of hemodynamic failure might influence the PaCO₂ - EtCO₂ gradient [11]. In this context, we hypothesized that a hemodynamic cluster-based approach could reduce the previously observed variability in the PaCO₂ - EtCO₂ gradient. A strong agreement between paired PaCO₂ and EtCO₂ values according to hemodynamic status could allow the determination of cluster-adjusted EtCO₂ targets when aiming for normocapnic ventilation during the prehospital management of TBI patients.

The primary objective was to assess the association between prehospital hemodynamic status and an increased PaCO₂–EtCO₂ gradient at emergency department (ED) admission. The secondary objective was to determine the correlation and agreement between paired PaCO₂ and EtCO₂ values according to hemodynamic status.

Methods

Design, population, and setting

A retrospective review of the institution’s trauma registry was conducted to identify all adult patients with moderate-to-severe traumatic brain injury admitted in our level 1 Trauma Centre under mechanical ventilation over a 4-year period (2022–2025). Moderate-to-severe traumatic brain injury was defined by a prehospital Glasgow Coma Scale (GCS) ≤ 12 with abnormal CT scan at ED admission [12]. Secondary exclusion criteria were patients with prehospital cardiac arrest, patients with a first arterial blood gas analysis performed more than 15 min after ED admission or with missing data regarding the primary endpoint (EtCO₂, respiratory, and/or hemodynamic parameters).

The study protocol was declared to the Data Protection Officer in accordance with the French legislation. The observational character of the present study was confirmed by the Institutional Review Board (IRB number: CER-BDX 2025 − 110), exempting the need for specific written consent. According to national regulations, the patients and/or next of kin were informed about the potential inclusion of their anonymized data for retrospective studies, and none expressed opposition.

Data collection

The following variables were retrospectively collected from the prehospital medical records: patient characteristics (gender, age, height and ideal body weight [according to the Lorentz formula]), mechanism of injury (e.g., traffic accident, fall from height, assault or gunshot), prehospital functional assessment (initial Glasgow Coma Scale [GCS], pupillary abnormalities [unilateral or bilateral mydriasis], initial systolic blood pressure [SBP], heart rate [HR] and Shock index [HR/SBP ratio], minimum capillary oxygen saturation [SpO₂], and prehospital hemoglobin [Hb] levels), fluid collection on FAST ultrasound assessment, prehospital ventilation settings (latest recorded values of tidal volume [Vt], respiratory rate [RR], end-expiratory positive pressure [PEEP] and inspired fraction of oxygen [FiO₂]), other prehospital interventions (fluid intake, need for vasopressors, transfusion, exsufflation and/or thoracostomy, osmotherapy), total duration of prehospital management.

At ED admission, were recorded the functional assessment (first documented mean arterial pressure [MAP], doses of norepinephrine, SpO₂ and EtCO₂ values, body temperature), the timing and results of the first arterial blood gas analysis (PaO₂, PaCO₂, base deficit), the description of traumatic brain injuries (assessed through Marshall CT scoring system) and significant associated injuries (with an abbreviated injury scale [AIS] ≥ 3) on the initial CT scan [13].

For each patient, three clusters were predefined based on prehospital hemodynamic status:

Cluster 1: Stable without vasopressors

Patients who achieved hemodynamic targets (MAP ≥ 80 mmHg) without need for vasopressors.

Cluster 2: Stabilized with low-dose vasopressors

Patients who did not achieve hemodynamic targets (MAP < 80 mmHg) without vasopressors

Or patients who achieved hemodynamic targets (MAP ≥ 80 mmHg) under low-dose vasopressors (≤ 0.2 µg/kg/min)

Cluster 3: Not stabilized with low-dose vasopressors

Patients who did not achieve hemodynamic targets (MAP < 80 mmHg) under low-dose vasopressors (≤ 0.2 µg/kg/min)

Or patients who achieved hemodynamic targets (MAP ≥ 80 mmHg) under high-dose vasopressors (> 0.2 µg/kg/min)

Finally, other study endpoints included signs of cerebral herniation at ED admission (pupillary abnormalities, altered transcranial Doppler [TCD] values, midline or temporal shift on the initial CT scan), length of stay in the intensive care unit (ICU) and 30-day mortality.

Statistical analysis

Results are expressed as mean ± standard deviation or median (25%–75% interquartile range) for continuous variables, and as absolute or relative frequencies for categorical variables. The data distribution was assessed using the Kolmogorov-Smirnov test. Non-paired continuous variables were compared using the unpaired Student’s t-test or Mann–Whitney U test, as appropriate. Categorical variables were compared using the χ² test or Fisher’s exact test. Comparison between paired EtCO₂ and PaCO₂ values was assessed using Pearson correlation and Bland–Altman analysis in the overall population and compared across the three predefined clusters [14].

The primary endpoint was an increased PaCO₂–EtCO₂ gradient, defined as a difference between PaCO₂ and EtCO₂ values > 5 mmHg at ED admission, assessed in the overall population and across the three predefined clusters using Kruskal-Wallis analysis. Assuming a mean standard deviation of 8 mmHg, 84 patients (28 patients per group) were needed to observe a minimal clinically relevant difference of 6 mmHg between groups, with a two-sided α level of 0.05 and an 80% power [15]. A minimum of 100 patients was determined to account for the disparities of sample size across clusters.

The secondary study end-point was the agreement between paired EtCO₂ targets and PaCO₂ values for normocapnia (PaCO₂ 35–45 mmHg), hypercapnia (PaCO₂ > 45 mmHg) or hypocapnia (PaCO₂ < 35 mmHg). Non-adjusted EtCO₂ targets were classified as below threshold (EtCO₂ < 30mmHg), within threshold (EtCO₂ 30–40 mmHg) and over threshold (EtCO₂ > 40 mmHg). Moreover, theoretical cluster-adjusted EtCO₂ targets were derived based on the mean bias observed between EtCO₂ and PaCO₂ in each hemodynamic cluster. The percentage agreement was compared between non-adjusted and cluster-adjusted thresholds. The agreement between EtCO₂ and PaCO₂ was also tested using Cohen’s κ coefficient (interpreted as slight for values of less than 0.2, fair for values of 0.21 to 0.4, moderate between 0.41 and 0.6, substantial between 0.61 and 0.8 and very good between 0.81 and 1) [16].

Finally, association between initial PaCO₂ values, PaCO₂–EtCO₂ gradient, and 30-day mortality was assessed using Kaplan–Meier curves. Multivariate analysis was conducted to assess whether the increased PaCO₂–EtCO₂ gradient was an independent risk factor for 30-day mortality, adjusting with 4 predefined covariates determined a priori: age, initial GCS, hemodynamic severity (type of cluster) and associated chest trauma. The choice of the covariates was justified by the need to minimize sparse-data bias, limit multicollinearity and ensure clinical relevance (variables known to be associated with clinical outcome in TBI) [17].

A p value < 0.05 was considered statistically significant. Statistical analyses were performed XLSTAT 2026 for Windows (Addinsoft Paris, France).

Results

Characteristics of the population

Over the study period, 578 patients were admitted for a moderate-to-severe TBI under mechanical ventilation, but 478 did not meet the inclusion criteria (lack of data regarding etCO₂ values at ED admission, N = 401; first arterial blood gas analysis after 15 min of admission, N = 66; prehospital cardiac arrest, N = 11). Finally, 100 moderate-to-severe TBI patients who required mechanical ventilation during the prehospital settings were analyzed in the present study.

The mean timing for arterial arterial blood gas analysis was 8 [6–10] minutes. At ED admission, 46 patients (46%) had a first PaCO₂ measurement within the predefined target range (35-45mmHg), 21 (21%) presenting hypocapnia and 26 (26%) with hypercapnia. The main characteristics of the population are summarized in Table 1.

Table 1.

Main characteristics of the population

	Overall population	PaCO2 - EtCO2 gradient	PaCO2 - EtCO2 gradient	p
	(N = 100)	≤ 5 mmHg (N = 41)	> 5 mmHg (N = 59)
Demographic data
-           Male gender	72 (72)	26 (63)	46 (78)	0.113
-           Age (years)	56 [27–68]	59 [27–68]	56 [28–69]	0.93
Mechanism of injury
-           Traffic accident (motor vehicle, bicycle, pedestrian…)	52 (52)	23 (56)	29 (49)	0.494
-           Fall from height	38 (38)	14 (34)	24 (41)	0.507
-           Other (brawl, gunshot injury…)	10 (10)	4 (10)	6 (10)	0.946
Prehospital initial assessment
-           Initial Glasgow Coma Scale	6 [3–8]	7 [4–9]	4 [3–7]	0.002
-           Pupillary abnormality	46 (46)	16 (39)	30 (51)	0.242
-           Hypoxemia (SpO2 < 90%)	35 (35)	8 (20)	27 (46)	0.006
-           Hypotension (SBP < 90mmHg)	33 (33)	4 (10)	29 (49)	< 0.001
-           Shock index > 1	32 (32)	5 (12)	25 (46)	< 0.001
-           Fluid collection on FAST ultrasound	5 (5)	0 (0)	5 (8)	0.019
-           Prehospital Hb levels	13 [10–14]	13 [12–14]	12 [10–14]	0.051
Prehospital management
-           Use of osmolar therapy	54 (54)	20 (49)	34 (58)	0.383
-           Need for vasopressors	46 (46)	8 (20)	38 (64)	< 0.001
-           Prehospital fluid resuscitation > 1 L	13 (13)	2 (5)	11 (19)	0.033
-           Need for prehospital transfusion	8 (8)	0 (0)	8 (14)	0.003
-           Need for exsufflation / thoracostomy	5 (5)	1 (2)	4 (7)	0.306
Duration of prehospital management before ED admission	90 [60–125]	90 [60–120]	85 [60–130]	0.753
Functional assessment at ED admission
-           Mean arterial pressure at ED admission (mmHg)	85 [75–99]	90 [77–105]	83 [73–96]	0.183
-           Mean dose of noradrenaline (µg/kg/min)	0 [0–0.2]	0 [0–0]	0.1 [0.0–0.3]	< 0.001
-           Tidal Volume (ml/kgIBW)	6.8 [6.3–7.2]	7.0 [6.5–7.3]	6.6 [6.2–7.0]	0.021
-           Respiratory rate	16 [15–18]	16 [15–18]	15 [15–18]	0.492
-           First SpO2 value at ED admission	99 [96–100]	100 [98–100]	98 [95–100]	< 0.001
-           First EtCO2 value at ED admission	32 [29–37]	35 [32–38]	30 [27–35]	< 0.001
-           Hypothermia (≤ 35 °C)	14 (14)	4 (10)	10 (17)	0.299
First arterial gas sample analysis
-           PaCO2 (mmHg)	41 [35–46]	36 [35–41]	43 [38–48]	< 0.001
-           PaO2/FiO2 ratio	257 [138–380]	371 [260–436]	163 [111–297]	< 0.001
-           Base Deficit (mmol/L)	4 [2–8]	2 [0–5]	6 [2–10]	0.001
Brain traumatic injuries and associated injuries
-           Marshall CT scoring system	2 [2–4]	2 [2–3]	2 [2–4]	0.838
-           Cerebral herniation (median or temporal shift)	44 (44)	18 (44)	26 (44)	0.987
-           Isolated intracranial injuries (vs. multiple trauma)	44 (44)	23 (56)	21 (36)	0.042
-           Associated chest trauma (AIS ≥ 3)	35 (35)	7 (17)	28 (47)	0.001
-           Associated spine injury (AIS ≥ 3)	13 (13)	1 (2)	12 (20)	0.004
-           Associated abdominal or pelvic trauma (AIS ≥ 3)	5 (5)	0 (0)	5 (8)	0.019
Clinical outcome
-           30 – days mortality	46 (46)	10 (24)	36 (61)	< 0.001
-           ICU Length of stay in survivors (N = 54)	17 [7–26]	15 [10–29]	18 [7–23]	0.93

Results expressed as numbers (percentages) and median [25–75 interquartile]

Association between prehospital hemodynamic status and increased PaCO₂–EtCO₂ gradient at ED admission

Overall, the PaCO₂ - EtCO₂ gradient at ED admission was 6 [3–13] mmHg, and 59 patients (59%) presented an increased PaCO₂ - EtCO₂ gradient at ED admission. The PaCO₂ levels at ED admission were statistically higher in patients with increased PaCO₂ - EtCO₂ gradient (44 ± 9 vs. 37 ± 6 mmHg, p < 0.0001). Other covariates associated with increased PaCO₂ - EtCO₂ gradient at ED admission are described in Table 1.

The mean PaCO₂ - EtCO₂ gradient increased according to the hemodynamic status at ED admission (Cluster 1: 3.5 ± 5.5 mmHg, Cluster 2: 9.9 ± 7.3 mmHg, Cluster 3: 12.6 ± 9.5 mmHg), with a statistical difference between groups (H = 21, df = 2, p < 0.0001) (Fig. 1). The characteristics of the three clusters according to the hemodynamic status are described in Table 2.

Fig. 1.

Distribution of the PaCO₂ - EtCO₂ gradient according to the hemodynamic status at ED admission

Table 2.

Characteristics of the three clusters according to the hemodynamic status at ED admission

	Cluster 1	Cluster 2	Cluster 3
	Stable without vasopressors	Stabilized with low-dose vasopressors	Not stabilized with low-dose vasopressors
	N = 33 (33%)	N = 31 (31%)	N = 36 (36%)
Patient’s characteristics and mechanism of injury
-           Male gender	23 (70)	25 (81)	24 (67)
-           Age (years)	56 [32–66]	56 [23–79]	55 [28–67]
-           Traffic accident	16 (48)	13 (42)	23 (64)
Prehospital initial assessment
-           Severe TBI (GCS ≤ 7)	21 (64)	24 (77)	25 (75)
-           Prehospital hypotension	2 (6)	7 (23)	24 (67)
-           Prehospital hypoxemia	8 (24)	10 (32)	17 (47)
-           Shock index > 1	3 (9)	6 (19)	23 (64)
-           Fluid collection on FAST ultrasound	1 (3)	2 (6)	2 (6)
-           Prehospital Hb levels	13 [12–14]	13 [11–15]	12 [10–13]
Prehospital interventions
-           Prehospital fluid intake (ml)	500 [500–500]	500 [500–1000]	1000 [950–1400]
-           Need for vasopressors	0 (0)	10 (32)	36 (100)
-           Need for prehospital transfusion	0 (0)	4 (13)	4 (11)
-           Need for exsufflation / thoracostomy	0 (0)	1 (3)	4 (11)
Hemodynamic status at ED admission
-           Mean arterial pressure (mmHg)	104 [90–115]	76 [73–89]	79 [73–88]
-           Doses of noradrenaline (mg/h)	0 [0–0]	0 [0–0.1]	0.3 [0.2–0.5]
Ventilation settings at ED admission
-           Tidal Volume (ml/kgIBW)	7.0 [6.5–7.4]	6.5 [6.1–7.2]	6.8 [6.3–7.1]
-           Respiratory rate	15 [15–16]	16 [15–20]	16 [15–18]
-           Inspiratory flow (L/min)	7 [7–8]	7 [7–9]	7 [6–8]
-           EtCO2 (mmHg)	35 [32–38]	32 [29–35]	30 [27–36]
First arterial gas sample analysis
-           PaCO2 (mmHg)	38 [35–41]	42 [35–46]	42 [36–48]
oHypocapnia (PaCO2< 35 mmHg)	9 (27)	6 (19)	6 (17)
oHypercapnia (PaCO2> 45 mmHg)	2 (6)	10 (32)	14 (39)
-           PaCO2 – EtCO2 gradient	3 [0–5]	9 [4–13]	11 [5–17]
-           Base Deficit (mmol/L)	2 [0–4]	4 [2–8]	7 [3–11]
-           PaO2/FiO2 ratio	335 [240–405]	248 [141–379]	169 [113–310]

Results expressed as numbers (percentages) and median [25–75 interquartile]

Correlation and agreement between paired PaCO₂ and EtCO₂ values

In the overall population, there was a poor correlation between PaCO₂ and etCO₂ levels at ED admission (r = 0.37; p < 0.001), and EtCO₂ was on average 9 mmHg (95% limits of agreement, -26 to + 8 mmHg) lower than its paired PaCO₂ value at ED admission. In this regard, only 44 patients (44%) had a paired CO₂ and EtCO₂ values in agreement at ED admission (κ = 0.13 for normocapnia, κ = 0.20 for hypocapnia and κ = 0.11 for hypercapnia).

Comparison between EtCO₂ and PaCO₂ showed a mean bias of -4 mmHg (95% limits of agreement, -15 to + 7 mmHg) in Cluster 1, -10 mmHg (95% limits of agreement, -25 to + 5 mmHg) in Cluster 2 and − 13 mmHg (95% limits of agreement, -32 to + 6 mmHg) in Cluster 3 (Fig. 2). The agreement between paired PaCO₂ and non-adjusted EtCO₂ levels decreased between the three clusters (61% [95%CI: 44–77%] in Cluster 1, 42% [95%CI: 25–59%] in Cluster 2, 31% [95%CI: 16–46%] in Cluster 3), with a statistical difference between groups (p = 0.041).

Fig. 2.

Scatter plots, histograms and bland-altman analysis comparing EtCO₂ and PaCO₂ in the overall population and according to the hemodynamic status

Based on these results, theoretical cluster-adjusted EtCO₂ targets were defined as follows: 30–40 mmHg for Cluster 1, 25–35 mmHg for Cluster 2 and 20–30 mmHg for Cluster 3. After cluster-adjustment of EtCO₂ targets, the agreement between paired PaCO₂ and EtCO₂ values at ED admission was 53% (κ = 0.17 for normocapnia, κ = 0.35 for hypocapnia, and κ = 0.36 for hypercapnia). However, cluster adjustment did not significantly improve the overall percentage agreement between paired PaCO₂ and EtCO₂ values (53% vs. 44%, p = 0.204).

Association between PaCO₂ levels, PaCO₂ - EtCO₂ gradient and patient outcomes

In multivariate analysis, there was a statistical association between 30-day mortality and PaCO₂ - EtCO₂ gradient at admission (adjusted OR = 5.1 [95%CI: 1.6–16.1], p < 0.0001). However, there was no statistical association between PaCO₂ levels (hypercapnia or hypocapnia) at ED admission and clinical symptoms of cerebral herniation at ED admission, ICU length of stay and 30-day mortality. The Kaplan–Meier curves are depicted Fig. 3.

Fig. 3.

Association between 30-day mortality and (A) hypercapnia, (B) hypocapnia and (C) increased PaCO₂ - EtCO₂ gradient (Kaplan – Meier curves)

Discussion

To our knowledge, this is the first study assessing the PaCO₂–EtCO₂ gradient according to a physiological approach and to test whether hemodynamic status–guided EtCO₂ targets can improve the correlation and agreement between paired PaCO₂ and EtCO₂ values when aiming for a normocapnic ventilation in the prehospital settings.

First, our results are in accordance with numerous studies demonstrating a wide variation in the gradient between PaCO₂ and EtCO₂, depending on patient conditions and over time [6–10]. In contrast, our results do not support more recent data suggesting that an EtCO₂ target range of 35–45 mmHg may represent a safe zone during prehospital care, particularly in unstable trauma patients [5]. A prehospital ventilation targeting an EtCO₂ threshold as a surrogate for PaCO₂ is associated with a high prevalence of hypoventilation and acidosis at hospital admission, especially in patients with hemodynamic failure [18].

Beyond this known disagreement, the PaCO₂–EtCO₂ gradient is known to be influenced by the hemodynamic status (increased alveolar dead space or reduced cardiac output) [18, 19]. We therefore hypothesized that a structured, cluster-based hemodynamic stratification could account for this variability and potentially improve the accuracy of EtCO₂-guided ventilation. Our results reflect this physiological relationship, suggesting a constant increase in PaCO₂ – EtCO₂ gradient according to the severity of hemodynamic failure (Cluster 1: 3.5 ± 5.5 mmHg, Cluster 2: 9.9 ± 7.3 mmHg, Cluster 3: 12.6 ± 9.5 mmHg). This effect is further amplified by an impaired ventilation–perfusion mismatch, particularly in the presence of associated thoracic injuries, which were more frequent in patients with higher gradients.

Based on this rationale, we evaluated whether the cluster-based hemodynamic stratification could improve the poor overall correlation between paired PaCO₂ and EtCO₂ values. In this context, our results suggested that the EtCO₂ may be relatively reliable in hemodynamically stable patients (Cluster 1: r = 0.56; p < 0.001), in accordance with a recently published study in cooperative and stable patients with respiratory distress [20]. However, there remained a poor correlation between PaCO₂ and EtCO₂ in hemodynamically unstable patients (Cluster 2: r = 0.48; p < 0.001; Cluster 3: r = 0.44; p < 0.001), with wide limits of agreement that limit the generalizability of these findings. Finally, cluster adjustment did not significantly improve the overall percentage agreement between paired PaCO₂ and EtCO₂ values, with only slight to fair agreement across capnia categories. In this regard, our results did not support the relevance of a ventilation strategy guided by cluster-adjusted EtCO₂ targets.

From an outcome perspective, there was no association between PaCO₂ levels at ED admission and neurological outcome and in-hospital mortality. This result is in accordance with a recent meta-analysis that did not find any statistical association between a prehospital hypercapnia and increased mortality [21]. In contrast, the PaCO₂–EtCO₂ remained associated with 30-day mortality after adjustment for key confounders. As such, the EtCO₂ monitoring should be interpreted more as a physiologic marker for risk stratification than a surrogate for PaCO₂ in the prehospital settings [22]. Our results reinforce the concept that early hemodynamic optimization should be prioritized in the prehospital management of TBI patients [11, 19]. Attempting to fine-tune ventilation based solely on EtCO₂ values, even when adjusted for hemodynamic status, may be misleading and ultimately of limited impact on patient outcomes.

Several limitations should be acknowledged. First, the retrospective single-center design exposes the study to selection bias, particularly given the large number of excluded patients due to missing data. Second, the timing between EtCO₂ recording and arterial blood gas sampling may not have been strictly simultaneous, potentially affecting the accuracy of paired comparisons. Third, the predefined hemodynamic clusters, although clinically relevant, may not fully capture the dynamic nature of circulatory failure. In addition, unmeasured confounders such as thoracic injuries, ventilation-perfusion mismatch, or variations in ventilatory management may have influenced the observed gradients. Finally, the relatively small sample size, especially within subgroups, may limit the generalizability of our findings.

Conclusion

Overall, our results suggest that although EtCO₂ monitoring remains indispensable in prehospital care, it should not be considered a reliable tool to accurately estimate PaCO₂ or to guide ventilatory settings in patients requiring strict control of normocapnia. Even when adapting EtCO₂ targets to hemodynamic status, the risk of inappropriate ventilation remains substantial, without clear impact on outcomes. These findings highlight the importance of prioritizing hemodynamic resuscitation as a key determinant of survival.

Abbeviations

AIS

Abbreviated injury scale

CT

Computed Tomography

ED

Emergency department

EtCO2

End-tidal carbon dioxide

FiO2

Inspired fraction of oxygen

GCS

Glasgow Coma Scale

Hb

Hemoglobin

HR

Heart rate

MAP

Mean arterial pressure

OR

Odd Ratio

PaCO2

Arterial partial pressure of carbon dioxide

PEEP

End-expiratory positive pressure

RR

Respiratory rate

SBP

Systolic blood pressure

SpO2

Capillary oxygen saturation

TBI

Traumatic brain injury

TCD

Transcranial Doppler

Vt

Tidal volume

95%CI

95% Confidence Interval

Author contributions

CC designed the study and wrote the manuscript. LPN and CC were responsible for the integrity of the data collection and the accuracy of the data analysis. XC, MG and MB helped to provide statistical advice on the design of the study, analyze the data, and write the manuscript. EA and LP confirmed an understanding of the statistical methods used for the efficacy analysis, and confirmed that the objectives and procedures of the study were honestly disclosed. All authors read and approved the final manuscript.

Funding

Only departmental funds were used for this study. No external funds were obtained.

Data availability

The data that support the findings of this study are not openly available due to reasons of sensitivity and are available from the corresponding author upon reasonable request. Data are located in controlled access data storage at Pellegrin Bordeaux University Hospital.

Declarations

Ethics approval and consent to participate

The study protocol was declared to the Data Protection Officer in accordance with the French legislation. The observational character of the present study was confirmed by the Institutional Review Board (Research Ethics Committee of the Health and Research Ethics Centre of Bordeaux, IRB number: CER-BDX 2025 − 110), exempting the need for specific written consent. According to national regulations, the patients and/or next of kin were informed about the potential inclusion of their anonymized data for retrospective studies, and none expressed opposition.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.