Service evaluation: Accuracy of using end-tidal CO 2 to estimate PaCO 2 values in suspected traumatic brain injury patients requiring mechanical ventilation

Introduction

Traumatic brain injury (TBI) frequently leads to disruption of cerebral autoregulation, rendering cerebral perfusion pressure (CPP) susceptible to the arterial partial pressure of carbon dioxide (PaCO₂). ¹ Dysregulated PaCO₂ can precipitate cerebral vasoconstriction or vasodilation, thereby altering oxygen delivery, intracranial pressure (ICP) and risking secondary brain injury. ² In the prehospital environment, direct measurement of PaCO₂ is possible using point-of-care testing (POCT), but it is rarely available outside of specialist prehospital critical care teams. End-tidal carbon dioxide (ETCO₂) monitoring is therefore used as a surrogate, with current guidance recommending a target ETCO₂ between 4.0 and 4.5 kPa to achieve normocapnia (i.e. a correction factor of approximately +0.5 kPa). ³ However, previous studies have shown poor correlation between ETCO₂ and PaCO₂, with wider gradients associated with increased mortality. ⁴ This service evaluation aimed to evaluate the accuracy of estimated PaCO₂ derived from ETCO₂ in patients with suspected TBI managed by the Hampshire and Isle of Wight Air Ambulance (HIOWAA).

Methods

A retrospective evaluation of HEMSBASE (MedicOne Systems) records was conducted from July 20, 2023, to July 20, 2024. Patients with suspected TBI and pre-hospital emergency anaesthesia performed by HIOWAA were included if they had a PaCO₂ recorded within 30 min of arrival at University Hospital Southampton. These were compared with the last recorded ETCO₂ at handover in the ED. Side-stream ETCO₂ sampling was used in all patients (Zoll X-Series Monitor – reported accuracy: 0–38 mmHg (0–5.1 kPa): ±2 mmHg (0.3 kPa); 39–150 mmHg (5.2–20.0 kPa): ±5% + 0.08% mmHg > 38).

Results were reported descriptively, and an inferential analysis was undertaken in R-Studio. Estimated PaCO₂ (EstPaCO₂) was calculated as ETCO₂ + 0.5 kPa. Data normality was assessed using the Shapiro–Wilk test. Group means were compared using a paired t-test. Agreement between measures was assessed using the Bland–Altman test. This service evaluation was reported in line with STROBE reporting guidelines.

Results

A total of 146 records were assessed for eligibility. Of these, 70 had suspected TBI, with a further 44 excluded because no PaCO₂ was recorded within 30 min of ED attendance. In total, 31 records were used in the final analysis. The median time between paired samples was 17 min (IQR 12–20 min).

An absolute difference between estimated PaCO₂ values and measured PaCO₂ values of 1.2 kPa was found (p < 0.001; Table 1). A paired boxplot was used to visually demonstrate the difference between individual values (Figure 1).

OPEN IN VIEWER

Table 1.

Descriptive statistics for estimated PaCO₂ (ETCO₂ + 0.5 kPa) and measured PaCO₂.

Variable	n	Mean	SD	Median	Min	Max	Range	Skewness	Kurtosis	SE
Estimated PaCO2 (kPa)	31	4.8	0.59	4.9	3.6	5.9	2.3	–0.26	–0.56	0.11
PaCO2 (kPa)	31	6.0	0.98	5.9	4.4	8.2	3.8	0.51	–0.33	0.18

OPEN IN VIEWER

Figure 1.

(A) Paired boxplot comparing estimated PaCO₂ (EstPaCO₂) with measured arterial PaCO₂. (B) Bland-Altman plot comparing agreement between EstPaCO₂ and PaCO₂.

The Bland–Altman bias between EstPaCO₂ and measured PaCO₂ was 1.17 (95% Limits of Agreement (LOA) 0.84–1.50; Figure 1). The differences are mostly above zero, so measured PaCO₂ is, on average, higher than the estimated PaCO₂ and shows moderate to wide scatter, with occasional large underestimations.

Discussion

In this study, prehospital ETCO₂ significantly underestimates PaCO₂ in suspected TBI patients requiring mechanical ventilation, highlighting the potential for unrecognised hypercapnia. Even mildly hypercapnic patients (arrival PaCO₂ > 6.53 kPa (49 mmHg)) have been shown to demonstrate worse outcomes in mild to moderate traumatic brain injury. ⁵ Nearly half (13/31) of patients demonstrated values that were outside the normocapnia range (4.5–6.0 kPa), despite their estimated PaCO₂ being within acceptable limits. Given that the measured PaCO₂ was on average 1.17 kPa higher than the estimated PaCO₂, one option may be to increase the correction factor from +0.5 kPa to a higher value, in the prehospital setting. However, discordance is not linear across all patients, so this may inadvertently cause hypocapnia in those with lower values, which may also be harmful. ⁶ Only two patients in our study were overestimated (by 0.31 and 0.34 kPa). Factors such as increased dead space ventilation, acidosis, chest trauma and hypotension may contribute to this variation.^7–10 Although our evaluation did not account for individual patient characteristics or accompanying injuries, it is possible that these factors contribute to some of the observed variation between Estimated PaCO₂ and measured PaCO₂. Our study did not examine haemodynamic data, temperature, ventilator settings or changes made during transitions of care, and is also limited by a small sample size (multiple exclusions due to delays in PaCO2 measurement). Nonetheless, the mean bias (>1 kPa) observed is clinically significant. Further work comparing prehospital ETCO₂ and PaCO₂ using point-of-care testing may better define their relationship in these patients.

Conclusion

ETCO₂-derived estimation of PaCO₂ may underestimate arterial values in suspected TBI patients in the pre-hospital setting. Lower ETCO₂ targets or point-of-care arterial analysis may be required to optimise ventilation prior to arrival at hospital and prevent secondary brain injury.