Effects of systemic mild hypercapnia on brain function in a refractory cardiac arrest model treated with extracorporeal cardiopulmonary resuscitation

Abstract

Introduction

Cardiac arrest (CA) is a major cause of mortality and neurologic disability. Extracorporeal cardiopulmonary resuscitation (ECPR) may be used in refractory CA to potentially reduce hypoxic-ischemic brain injuries. Currently, no recommendation exists regarding optimal carbon dioxide levels in ECPR patients.

Methods

Twelve pigs were submitted to 5 min of untreated ventricular fibrillation followed by 25 min of CPR; 30 min after CA induction, ECPR was initiated and defibrillations delivered until ROSC. Pigs were randomly assigned to normocapnia (NC; PaCO₂ 35–45 mmHg, n = 6) or hypercapnia (HC; PaCO₂ 50–55 mmHg, n = 6). Animals underwent continuous multimodal neurological and cardiovascular monitoring. The primary outcome of the study was the change of brain tissue oxygen tension (PbtO₂) over time. Other parameters were secondary outcomes. At 12 h post-ROSC, animals were euthanized and brain tissues were harvested for molecular analyses.

Results

Mean PbtO₂ tended to be higher in the HC group (49.9 vs 36.4 mmHg in the NC group; p = 0.13), while mean intracranial pressure was similar between the two groups (19.9 vs 22.6 mmHg, respectively – p = 0.42). Cerebral glucose concentrations were significantly higher in the HC group (p = 0.01), without differences in lactate/pyruvate ratio, glycerol or glutamate. Caspase-8 gene expression in the parietal lobe was significantly lower in the HC group, with no differences in EEG cortical activity or serum biomarkers of brain injury.

Conclusions

In this model of ECPR, HC was associated with increased cerebral glucose concentrations and reduced caspase-8 expression, without changes in PbtO₂ levels, intracranial pressure, cortical activity, or neurobiomarkers.

Background

Cardiac arrest (CA) represents a major public health challenge owing to its high incidence and substantial global mortality rate.¹ Among patients who achieve return of spontaneous circulation (ROSC), cardiovascular failure and hypoxic–ischemic brain injury (HIBI) are the main causes of poor neurological outcomes and death.² In recent years, significant progress has been made in improving resuscitation strategies and post–cardiac arrest care. In refractory cases, extracorporeal cardiopulmonary resuscitation (ECPR) has emerged as a potential rescue therapy. Although its use remains debated, ECPR may improve survival and neurological outcomes in carefully selected patient populations.3, 4, 5

In comatose patients after CA, international guidelines⁶ currently recommend several therapeutic interventions, including temperature control, optimization of oxygenation and careful control of ventilation. These measures aim to improve neurological outcomes by limiting HIBI and preventing secondary cerebral damage. However, most of the available evidence has been derived from patients treated with conventional cardiopulmonary resuscitation (CPR). The optimal management of patients undergoing ECPR remains less clearly defined, as prolonged low-flow duration and rapid reperfusion may exacerbate cerebral ischemia–reperfusion injury in this specific population.⁷

In a previous experimental study,⁸ we investigated the impact of therapeutic hypothermia on brain function in a refractory CA model treated with ECPR. Despite the theoretical neuroprotective effects of temperature control,9, 10 our results did not demonstrate a clear reduction in brain injury in this setting, prompting further investigation into other modifiable factors influencing cerebral injury after ECPR. Among these factors, arterial carbon dioxide tension (PaCO₂) has emerged as a key determinant of cerebral physiology and has been extensively investigated for its potential role in neuroprotection.¹¹ Indeed, PaCO₂ is a major regulator of cerebral blood flow (CBF); hypercapnia (HC) induces vasodilatation and increases cerebral perfusion, whereas hypocapnia promotes vasoconstriction and reduces CBF.¹² Beyond its vascular effect, HC has been associated with reduced oxidative stress¹³ and attenuation of excitotoxicity.¹⁴ Systemic mild HC (i.e. PaCO₂ between 50 and 55 mmHg) has been widely studied in subjects undergoing post-resuscitation care through experimental,15, 16, 17 observational,18, 19 and randomised controlled trials.20, 21, 22 The TAME trial,²² the largest randomised controlled trial on this topic, did not demonstrate improved neurological outcomes with mild HC compared with normocapnia (NC) in comatose CA survivors. However, the trial enrolled an unselected population, with a relatively favourable prognosis, and its findings may not apply to patients undergoing ECPR, a population characterised by prolonged ischemia, more severe brain injury, and higher mortality. The effects of mild HC in this specific context remain poorly studied.

In this study, we postulated that HC could reduce HIBI after refractory CA treated with ECPR. We therefore assessed the effects of HC on brain perfusion, extracellular metabolites, cortical activity and biomarkers of brain injury in a refractory CA animal model treated with ECPR.

Methods

Experimental setting

The Institutional Review Board for Animal Care of the Université libre de Bruxelles (ULB, Belgium) approved all experimental procedures (Ethical Committee approval number: 731N), which were also in compliance with ARRIVE 2.0 (Animal Research: Reporting in Vivo Experiments) guidelines.²³ The recommended checklist is provided in the Supplemental Appendix (Table S1). Care and handling of the animals were in accord with National Institutes of Health guidelines (Institute of Laboratory Animal Resources). For all experiments, both sex swine of six months old and weighing 45–55 kg (Sus Scrofa Domesticus) were used.

Animal preparation

A complete description of our experimental model developed in our laboratory was published previously.8, 24 The animal preparation, the neurosurgical procedure, the cannulation procedure, the ECMO preparation and the cardiac arrest procedure are described in the Supplemental Appendix.

Group allocation and additional treatment

On the day of the experiment, pigs were randomly assigned in a 1:1 ratio, using simple randomization, to either the HC group (target systemic PaCO₂ 50–55 mmHg) or the normocapnia group (NC; target systemic PaCO₂ 35–45 mmHg). Sweep gas flow and minute ventilation were then adjusted accordingly to maintain the PaCO₂ within the predefined range of the allocated group. According to our previous results,⁸ normothermia (body temperature between 37 and 38°C) was maintained in all the pigs. TTM was achieved using a heat exchanger connected to the ECMO circuit. Body temperature was controlled immediately after ROSC and for all the experiment. The timeline of the experiment is shown in the Fig. 1.

Fig. 1.

Protocol timeline.

MNM = multimodal neuromonitoring; EEG = electroencephalography; PbtO₂ = brain oxygen pressure; ICP = intracranial pressure; MD = microdialysis; ECPR = extracorporeal cardiopulmonary resuscitation; MAP = mean arterial pressure; PaCO₂ = partial pressure of carbon dioxide; T-1 = baseline; T0 = ROSC.

Blood, brain interstitial fluid and cerebral tissue sampling

Blood gas analyses were obtained at baseline (T-1), ROSC (T0), and every hour till the end of the experiment (from T1 to T12). Furthermore, arterial blood samples were collected at T-1, T0, T1, T6 and T12. Samples were then immediately centrifuged to obtain plasma and stored at – 80°C. Microdialysis samples were obtained at baseline and hourly after ROSC.

For all microdialysis samples, glucose, lactate, pyruvate, glutamate and glycerol were measured. After the sacrifice of the animal, the skull was opened, the dura mater was dissected using a scalpel and cerebral tissues (frontal and parietal lobes) were harvested on each side. Tissues were immediately frozen in liquid nitrogen and stored at – 80°C.

Study outcomes

The primary outcome of this study was the longitudinal trajectories of PbtO₂ over time in the two groups. All other neurological, hemodynamic, metabolic and molecular parameters were therefore analysed as secondary exploratory outcomes.

Data analysis and statistical analysis

The data analysis and statistical analysis are described in the Supplemental Appendix.

Results

Baseline characteristics

Twelve pigs were included in the analysis (n = 6 in HC group, and n = 6 in NC group); in particular, 4 males and 2 females were included in the HC group, and 3 males and 3 females in the NC group. All the animals achieved ROSC. Baseline characteristics of the study groups are shown in the Table 1; there were no significant differences in body weight and in body temperature between the two groups. Respiratory parameters (PaO₂, PaCO₂), hemodynamical parameters (e.g. cardiac output, mean arterial pressure, blood lactate levels) and neurological parameters (e.g. cerebral temperature, intracranial pressure and brain oxygen pressure) were also comparable between the two groups at baseline.

Table 1.

Baseline characteristics of the study groups.

	NC group (n = 6)	HC group (n = 6)	P value
Body weight (kg)	53.5 ± 3.4	56.2 ± 1.7	0.12
Body temperature (°C)	36.5 ± 0.8	36.3 ± 0.6	0.61
Cardiac output (L/min)	6.5 ± 0.9	7.4 ± 1.6	0.25
Mean arterial pressure (mmHg)	73 ± 4	76 ± 9	0.47
PaO2 (mmHg)	119 ± 30	128 ± 22	0.56
PaCO2 (mmHg)	38 ± 5	41 ± 1	0.11
Blood lactate (mmol/L)	1.2 ± 0.2	1.1 ± 0.1	0.65
Blood glucose (mg/dL)	98 ± 28	106 ± 20	0.56
Cerebral temperature (°C)	36.8 ± 0.5	36.4 ± 0.6	0.29
PbtO2 (mmHg)	51.3 ± 12.1	47.3 ± 9.3	0.54
ICP (mmHg)	9.6 ± 3.1	11.4 ± 3.7	0.38

Data are presented as count (%) or mean (±SD). NC = normocapnia; HC = hypercapnia; PbtO₂ = brain tissue oxygen pressure; ICP = intracranial pressure. p < 0.05 was considered as statistically significant.

CPR related variables

No statistically significant difference was observed between the HC group and the NC group in terms of number of defibrillation attempts (median value: 4 vs. 2; p = 0.4) and end-tidal CO₂ values throughout the CPR (Fig. S1 in the Supplemental Appendix).

Sedation, analgesia and respiratory support parameters

After ROSC, no statistically significant difference was observed between the HC group and the NC group in terms of ECMO sweep gas settings, sedative and analgesic doses over time, and depth of anaesthesia. Minute ventilation was significantly lower in the HC group (p = 0.02). All the results are provided in the Supplemental Appendix (Fig. S2).

Physiological and metabolic variables

After ROSC, HC group was associated with a significantly higher value of PaCO₂ compared with NC group (PaCO₂ in HC group: 55.4 mmHg, PaCO₂ in NC group: 40.8 mmHg; mean difference 14.67, 95% CI 12.5–16.8; p < 0.0001). pH was significantly lower in HC group (pH in HC group: 7.31, pH in NC group: 7.43; mean difference 0.12, 95% CI 0.04–0.20; p = 0.006). No statistically significant differences were observed between the NC group and the HC group in terms of PaO₂ and body temperature (Fig. 2). Plasma glucose values over time were comparable between the two groups throughout the experiment (p group × time interaction = 0.18) (Fig. S3).

Fig. 2.

Time-course of physiological parameters, metabolic variables, hemodynamical parameters, PbtO₂ and ICP levels throughout the experiment.

PaO₂ = partial pressure of oxygen; PaCO₂ = partial pressure of carbon dioxide; MAP = mean arterial pressure; PbtO₂ = brain tissue oxygen pressure; ICP = intracranial pressure. p group × time interaction < 0.05 was considered as statistically significant.

Hemodynamical parameters

After ROSC, no statistically significant difference was observed between the HC group and the NC group in terms of cardiac output, mean arterial pressure, norepinephrine dose and lactate levels (Fig. 2). Fluid balance was also not significantly different between HC group (11,868 ± 1724 mL) and NC group (8573 ± 758 mL; p = 0.12).

Primary outcome and cerebral hemodynamics

After ROSC, no statistically significant difference was observed between the NC group and the HC group regarding PbtO₂ values over time (Fig. 2). Mean PbtO₂ was 49.9 mmHg in the HC group and 36.4 mmHg in the NC group, corresponding to a non-significant mean difference of −13.55 mmHg (95% CI −31.61 to +4.49).

Also, ICP values were similar between groups (Fig. 2). Mean ICP was 19.9 mmHg in the NC group and 22.6 mmHg in the HC group, corresponding to a non-significant mean difference of −2.68 mmHg (95% CI −9.8 to +4.42).

Cerebral metabolism

After ROSC, no statistically significant difference was observed between the HC group and the NC group regarding lactate, pyruvate, glycerol, glutamate or lactate/pyruvate ratio measured by cerebral microdialysis. Mean glucose concentration was 1.6 mmol/L in the NC group and 3.9 mmol/L in the HC group, corresponding to a significant mean difference of −2.3 mmol/L (95% CI −4 to −0.63 – Fig. 3).

Fig. 3.

Time-course of cerebral microdialysis parameters throughout the experiment.

p group × time interaction < 0.05 was considered as statistically significant.

Plasma biomarkers

Circulating levels of haemoglobin, white blood cells, platelets, urea nitrogen, creatinine, GOT, GPT, LDH, total and direct bilirubin, troponin I and CPK were similar between the two groups throughout the experiment. Concerning biomarkers of acute brain injury, circulating levels of GFAP and NfL did not differ significantly between the two groups at T-1, T1, T6 and T12 (p group × time interaction = 0.38 and p group × time interaction = 0.28, respectively) (Fig. 4). Most of NSE values were below the lower limit of detection at baseline; however, no statistically significant difference was detected between the two groups at T12.

Fig. 4.

GFAP and NFL levels at baseline (T-1), T1, T6 and T12 post ROSC.

GFAP = glial fibrillary acid protein; NfL = neurofilament light chain protein. p group × time interaction < 0.05 was considered as statistically significant. *p < 0.05.

Cortical gene expression

There were no statistically significant differences in the expression of MAP2, GFAP, CD11ß, PECAM-1, caspase-3 and HO-1 between the two groups, whether it is at the level of the frontal lobe or for the parietal lobe. Gene expression of caspase-8 in the parietal lobe was significantly higher in the NC group (p = 0.04). All the results are provided in the Supplemental Appendix (Figs. S4 and S5).

Cerebral activity

Analysis of continuous EEG demonstrated that mean amplitude, mean standard deviation, kurtosis and skewness did not differ between the two groups after ROSC. All the results are provided in the Supplemental Appendix (Fig. S6).

Discussion

Our study showed that, compared with systemic NC, induction of mild systemic HC in a refractory CA model treated with ECPR was not associated with a significant increase in brain oxygenation over time, neither improved recovery of cortical electrical activity nor lower concentrations of circulating biomarkers of brain injury. The intervention was also associated with a significantly increased cerebral glucose concentrations and a lower gene expression of caspase-8 in the parietal lobe, compared to normocapnia.

In comatose patients resuscitated after CA, HIBI remains a major determinant of poor neurological outcomes and mortality. Consequently, considerable efforts have been made to identify strategies that could limit secondary brain damages, and various therapeutic strategies, such as induced hypothermia and pharmacological interventions, have been investigated. Due to its significant impact on cerebral perfusion and its antioxidant properties,25, 12, 13, 14 the effect of systemic HC on brain injury reduction after CA have been studied. To the best of our knowledge, except for a few retrospective studies,27, 28, 29 the effect of systemic HC on brain functions in ECPR patients has not been evaluated yet.

Adequate cerebral perfusion is fundamental to both normal brain function and prevention of secondary brain injury in CA patients. Although the brain accounts for only 2–3% of total body mass, it receives approximately 15% of cardiac output and consumes around 20% of total body oxygen under physiological conditions. Its high metabolic demand, combined with limited energy reserves, makes CBF essential not only for oxygen and substrate delivery but also for limiting ischemic and metabolic insults and facilitating the clearance of toxic by-products.²⁵ CBF is tightly regulated by several physiological determinants, among which PaCO₂ values and subsequent changes in perivascular pH are a major modulator.²⁶ In healthy individuals, CBF increases by approximately 1–2 mL/100 g/min for each 1 mmHg rise in PaCO₂.11, 12 In our study, although non-significant, PbtO₂ levels tended to be higher in the HC group than in the NC group, suggesting enhanced cerebral perfusion. This interpretation is supported by the concomitant increase in cerebral microdialysis glucose concentrations, consistent with improved substrate delivery. These findings should be interpreted with caution. Although HC was associated with a modest increase in PbtO₂, the magnitude of this effect was small, raising questions about its physiological and clinical relevance. More importantly, this improvement was not accompanied by evidence of enhanced cerebral oxidative metabolism. The lactate-to-pyruvate ratio, a surrogate marker of cerebral redox state and mitochondrial function, was comparable between groups, whereas a meaningful improvement in tissue perfusion would have been expected to reduce this ratio. Similarly, plasma glucose concentrations did not differ between groups, whereas cerebral extracellular glucose levels were higher in the HC group. Rather than indicating improved substrate delivery, this pattern may alternatively reflect reduced cerebral glucose utilisation. Such an interpretation is consistent with the concomitant reductions in neuronal electrical activity on cortical EEG and lower extracellular glutamate concentrations, suggesting decreased neuronal metabolic demand rather than enhanced metabolic recovery. However, this hypothesis remains speculative. A genuine reduction in cerebral metabolic activity would also be expected to influence other microdialysis-derived metabolites, including absolute lactate and pyruvate concentrations, which did not show a coherent pattern supporting this mechanism.

Another important issue is that ICP levels did not differ between groups. This apparent dissociation raises important considerations. First, the increase in CBF might have remained within the compensatory capacity of intracranial compliance mechanisms, preventing a measurable rise in ICP.³⁰ Second, the observed increase in PbtO₂ may not only reflect augmented perfusion but also changes in oxygen diffusion or haemoglobin affinity. Actually, HC and a lower pH might have contributed to increased PbtO₂ through a rightward shift of the haemoglobin–oxygen dissociation curve (i.e. Bohr effect³¹), thereby facilitating oxygen unloading at the tissue level independently of changes in CBF. Future studies should incorporate direct measurements of CBF, which were not performed in the present study, to better define the respective contributions of perfusion, oxygen diffusion, and cerebral glucose metabolism. Taken together, these findings do not provide convincing evidence that moderate HC substantially improves cerebral hemodynamics or metabolism in this experimental model. At most, the observed increase in PbtO₂ appears to reflect a limited physiological effect on cerebral oxygenation that was not consistently accompanied by parallel improvements in metabolic or electrophysiological markers.

The expected effect of HC on cortical activity is not straightforward, resulting from two competing mechanisms: improved cerebral perfusion may promote earlier recovery of electrical activity, while HC and acidosis are known to suppress neuronal excitability, notably through inhibition of NMDA receptor activity and attenuation of glutamate-mediated excitotoxicity.14, 32 In our study, sEEG analysis showed a trend to a reduced mean amplitude associated with higher kurtosis and skewness in the HC group, along with a slower evolution toward a discontinuous suppression–burst pattern. In parallel, cerebral microdialysis revealed a trend of lower glutamate concentrations in the HC group. These findings are consistent with attenuated cortical excitability under HC but should be interpreted with caution given the lack of statistical significance.

Previous experimental studies have reported antiapoptotic effects of HC, notably involving reduced caspase-3 expression. In a rat focal cerebral ischemia/reperfusion model, Tao et al.³³ reported that HC reduced infarct volume, improved neurological outcomes, and modulated mitochondrial apoptotic pathways, including decreased caspase-3 expression. Similarly, in experimental model of ventilator-induced lung injury, Yang et al.¹³ reported reduced cellular apoptosis due to hypercapnia via among others attenuation of caspase-3 activation. Interestingly, in our study, gene expression of caspase-8 in the parietal lobe was significantly lower in the HC group. Although experimental evidence supports an antiapoptotic effect of HC, it has primarily been demonstrated in relation to caspase-3, whereas our results involve caspase-8. Therefore, direct comparisons should be made with caution. Nonetheless, our findings may suggest that HC could influence apoptotic signalling at different levels of the caspase cascade, although this interpretation remains hypothetic. Moreover, this difference in our study may be interpreted cautiously and may reflect a type I error.

Several biomarkers have been investigated for neuroprognostication after CA, including NSE, a cytoplasmic enzyme expressed in neuronal cell bodies and currently the only biomarker recommended for this purpose,⁶ as GFAP, a key protein predominantly found in astrocytes, and NfL. Notably, peak serum concentrations of these biomarkers are typically observed beyond the early phase following CA – around 12 and 48 h for GFAP and often later than 72 h for NSE and NfL.³⁴ In our study, circulating biomarkers of brain injury remained comparable between groups throughout the experimental period. Given that the observation period in our study was limited to 12 h, the absence of differences between groups is not unexpected.

Our research has numerous limitations. First, it took place in a controlled environment, which might not accurately reflect real clinical practice. Second, the observation period was restricted to 12 h, limiting interpretation to the early reperfusion stage and hindering any evaluation of delayed neurological damage or long-term consequences of HC. Third, no histological or immunohistochemical evaluations were conducted in this study. Fourth, only two levels of PaCO₂ were evaluated in our protocol. Fifth, Animals were randomised on the day of the experiment using a simple randomization scheme without stratification by sex. Consequently, sex balance between treatment groups was not guaranteed and should be considered when interpreting the findings. Sixth, the reduced sample size limited the precision of effect estimates, decreased study power and may have prevented detection of small or clinically relevant differences. As a result, interpretation of non-significant trends across perfusion, metabolic, and biomarker measures should be made with caution. Seventh, no direct measure of CBF was made in our study. Further studies should include this measure to better define the respective contributions of perfusion and oxygen diffusion. Eighth, post-ROSC care was managed by the first two authors of this manuscript and was therefore not blinded. Ninth, this study cannot distinguish the physiological effects of HC from those of the accompanying acidemia. Elevations in PaCO₂ inevitably altered systemic and cerebral pH, precluding attribution of the observed changes in cerebral perfusion, tissue oxygenation, metabolism, and electrophysiological activity to either stimulus alone. A mechanistic design incorporating a normocapnic acidotic control group or selective buffering of hypercapnic acidosis, using a non-CO₂-generating buffer such as tromethamine (THAM), would be necessary to isolate the respective roles of CO₂ and pH. Finally, we deliberately focused on quantitative EEG measures, reflecting cerebral functional dynamics rather than on conventional prognostic EEG classifications, including background continuity or burst suppression burden. Indeed, although these variables have well-established prognostic value, they address a different clinical question.

Conclusions

In this clinically relevant model of refractory cardiac arrest supported with ECPR, mild hypercapnia did not produce a coherent pattern of neurophysiological or metabolic benefit. No consistent effects were observed across cerebral oxygenation, hemodynamics, electrophysiological activity, or markers of neuronal injury. The observed increase in cerebral extracellular glucose and reduction in caspase-8 expression may indicate biological effects deserving further investigation.

Ethics and patient consent

The Institutional Review Board for Animal Care of the Université libre de Bruxelles (ULB, Belgium) approved all experimental procedures (Ethical Committee approval number: 731N).

CRediT authorship contribution statement

Anthony Moreau: Writing – original draft, Methodology, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization. Fuhong Su: Writing – review & editing, Formal analysis, Data curation. Lorenzo Ferlini: Writing – review & editing, Formal analysis, Data curation. Nicolas Gaspard: Writing – review & editing, Formal analysis, Data curation. Francesca Pischiutta: Writing – review & editing, Data curation. Elisa Zanier: Writing – review & editing, Formal analysis, Data curation. Jacques Creteur: Writing – review & editing, Conceptualization. Filippo Annoni: Writing – review & editing, Methodology, Formal analysis, Data curation, Conceptualization. Fabio Silvio Taccone: Writing – review & editing, Validation, Supervision, Methodology, Investigation, Formal analysis, Data curation, Conceptualization.

Funding

AM and FST are supported by the Fonds Erasme pour la Recherche Médicale.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary material

The following are the Supplementary material to this article:

Supplementary Data 1

The Supplementary material includes the methods, the ARRIVE 2.0 checklist and supplementary tables / figures.