Article Text

Download PDFPDF

Spontaneous hyperventilation and brain tissue hypoxia in patients with severe brain injury
  1. Emmanuel Carrera1,
  2. J Michael Schmidt1,
  3. Luis Fernandez1,
  4. Pedro Kurtz1,
  5. Maxwell Merkow2,
  6. Morgan Stuart2,
  7. Kiwon Lee1,2,
  8. Jan Claassen1,2,
  9. E Sander Connolly2,
  10. Stephan A Mayer1,2,
  11. Neeraj Badjatia1,2
  1. 1Department of Neurology, Division of Neurocritical Care, Columbia University Medical Center, New York, USA
  2. 2Department of Neurosurgery, Columbia University Medical Center, New York, USA
  1. Correspondence to Dr Neeraj Badjatia, Columbia University Medical Center, 177 Fort Washington Avenue, MHB 8-300, New York, NY 10032, USA; nbadjatia{at}


Background Hyperventilation has been shown to be associated with cerebral vasoconstriction and increased risk of infarction. Our aim was to determine whether spontaneous reduction in end-tidal CO2 (EtCO2) was associated with an increased in brain tissue hypoxia (BTH).

Method We studied 21 consecutive patients (mean age 50±16 years; 15 women) undergoing continuous monitoring for brain tissue oxygenation (PbtO2), intracranial pressure (ICP), cerebral perfusion pressure (CPP) and EtCO2; mean values were recorded hourly BTH was defined as brain tissue oxygen tension (PbtO2) <15 mm Hg.

Results Diagnoses included subarachnoid haemorrhage (67%), intracranial haemorrhage (24%) and traumatic brain injury (10%). Overall, BTH occurred during 22.5% of the study period (490/2179 hourly data). The frequency of BTH increased progressively from 15.7% in patients with normal EtCO2 (35–44 mm Hg) to 33.9% in patients with EtCO2<25 mm Hg (p<0.001). The mean tidal volume and minute ventilation were 7±2 ml/kg and 9±2 1/min, respectively. Hypocapnia was associated with higher measured-than-set respiratory rates and maximal minute ventilation values, suggestive of spontaneous hyperventilation. Using a generalised estimated equation (GEE) and after adjustment for GCS, ICP and core temperature, the variables independently associated with BTH events were EtCO2 (OR: 0.94; 95% CI 0.90 to 0.97; p<0.001) and CPP (OR: 0.98; 95% CI 0.97 to 0.99; p=0.004).

Conclusion The risk of brain tissue hypoxia in critically brain-injured patients increases when EtCO2 values are reduced. Unintentional spontaneous hyperventilation may be a common and under-recognised cause of brain tissue hypoxia after severe brain injury.

  • Hyperventilation
  • end-tidal CO2
  • hypercapnia
  • hypocapnia
  • brain tissue oxygen pressure
  • brain injury
  • coma
  • mechanical ventilation
  • head injury
  • intensive care
  • subarachnoid haemorrhage

Statistics from


Monitoring of local brain tissue oxygen pressure (PbtO2) has been shown to be a reliable marker of brain tissue perfusion and is considered to reflect the balance between oxygen delivery and demand.1 2 Recent studies have shown that a decrease in PbtO2 under a threshold of 15 mm Hg is associated with an increased risk of brain ischaemia and poor outcome in brain-injured patients.3 4

In patients with severe brain injury, reduction in arterial carbon dioxide pressure (PaCO2) with hyperventilation is often utilised as a treatment for elevated intracranial pressure (ICP).5–8 Hypocapnia induces vasoconstriction leading to a decrease in cerebral blood flow (CBF) and ICP. However, reductions in PaCO2 may also induce significant reductions in CBF leading in turn to a reduction in oxygen supply to the brain parenchyma.9 The potential benefits of hyperventilation on ICP could be offset by these reductions in brain tissue oxygenation.10–13 For these reasons, clinicians are usually careful to avoid excessive induced hyperventilation when treating elevated ICP.

Not all hyperventilation is intentional, however. Central (or neurogenic) hyperventilation is a common consequence of brain injury, but very little is known about its clinical epidemiology or effects on brain oxygenation. Most reports of central hyperventilation refer to accounts of extreme hypocapnia in alert patients with brain tumours.14–16 Hyperventilation is also a common manifestation of paroxysmal dysautonomia, a syndrome that also involves hypertension, tachycardia, fever and dystonic posturing.17

The conventional method for determining the extent of hyperventilation is assessment of PaCO2 using blood gas analysis; this method provides an accurate but only intermittent measure of PaCO2. By contrast, end-tidal CO2 (EtCO2) monitoring provides continuous monitoring at the bedside and has been shown to be useful to monitor the impact of hyperventilation on PbtO2 in animal models.18–20 In humans, the role of continuous EtCO2 monitoring in the management of patients with severe brain injury is largely unknown, despite its common use. In this study, we investigated the relationship between changes in EtCO2 and brain tissue oxygenation in comatose patients with severe brain injury.


Patient population

We prospectively collected hourly data in 21 consecutive patients with severe brain injury who underwent monitoring with PbtO2, intraparenchymal ICP and EtCO2 admitted to our Neuro ICU between August 2004 and May 2008. Intracranial monitoring was performed as a standard of care in comatose patients with a Glasgow Coma Scale score (GCS) ≤8 who required continuous ICP monitoring for management of their condition. Intracranial monitoring was discontinued at the discretion of the attending neurointensivist based on the overall clinical course of the patient and perceived need for continued monitoring. Continuous data were prospectively collected and recorded using a high-resolution data-acquisition system (Bed Master, Excel Medical, Jupiter, Florida) All procedures were performed according to the Declaration of Helsinki. This prospective observational cohort study was approved by the hospital Institutional Review Board.

Intracranial monitoring

PbtO2 was measured with a flexible polarographic Clark-type probe (Licox mbH, Kiel, Germany; Integra Neurosciences, Plainsboro, New Jersey) and ICP with an intraparenchymal pressure gauge (Ventrix or Camino System, Kiel, Germany; Integra Neurosciences). After a burr hole was made at the bedside, the catheters were inserted using a three-lumen bolt. The intracranial catheters were placed into either the right or left frontal lobe, and directed into the most symptomatic hemisphere into regions deemed at highest risk for secondary injury based on the consensus opinion of the attending neurointensivist and neurosurgeon. In patients with diffuse and non-lateralised brain injury, monitors were placed into the right frontal lobe.

Physiological variables

Physiological variables, including heart rate (HR), mean arterial blood pressure, respiratory rate (RR), core temperature (bladder) and oxygen saturation (SO2), were continuously monitored in all patients and recorded hourly. EtCO2 was continuously measured (breath by breath) using an infrared capnometer (Respironics). EtCO2 values were displayed ‘on line’ at the bedside. Hourly ICP, mean arterial pressure (MAP), PbtO2, SO2 and EtCO2 data were prospectively recorded as part of the standard of care. Cerebral perfusion pressure (CPP) was calculated as CPP=MAP–ICP.

Clinical management

Patient care for subarachnoid and intracerebral haemorrhages and traumatic brain injury conformed to guidelines established by the American Heart Association and Brain Trauma Foundation.10 21–24 Haemodynamic and fluid management was targeted to maintain a minimum CPP >60 mm Hg and ICP <20 mm Hg. Fever was aggressively treated using surface (Arctic Sun Cooling System, Medivance, Louisville, Colorado) or intravascular (Celsius Control System, Innercool Therapies, San Diego, California) cooling systems. Sedation was given in the form of propofol (20–50 μg/kg/min), dexmedetomidine (0.3–0.7 mg/kg/h), fentanyl (1–3 μg/kg/h) or remifentanyl (0.03–0.25 mg/kg/min) as needed to attain comfort and eliminate patient–ventilator dyssynchrony; higher doses were used only in order to treat ICP crisis or promote oxygenation in patients with acute lung injury or acute respiratory distress syndrome.

Mechanical ventilation

All patients were intubated and mechanically ventilated with Puritan-Bennett 2000 ventilators during the course of intracranial monitoring. Patients were initially ventilated with the continuous mandatory ventilation (CMV, ie, assist-control mode) with a positive end expiratory pressure (PEEP of 5 mm Hg). Selection of volume- versus pressure-cycled ventilation in CMV mode was determined by the attending neurointensivist based on the need to control peak and plateau airway pressures. Respiratory rate and tidal volume were adjusted to ensure a minimum minute ventilation of 5 l/min. The fraction of inspired oxygen was started at 40% and adjusted as necessary to maintain a PaO2 >90 mm Hg. Lung protective ventilation (tidal volume 4–6 ml/kg) was employed in patients with acute lung injury (ARI)/Acute Respiratory Distress syndrome (ARDS). No efforts were made to routinely provide prophylactic hyperventilation to pCO2 levels below 30 mm Hg. When sedation and osmotherapy alone were not sufficient to control ICP, mild hyperventilation (target pCO2 28–32 mm Hg) was initiated for brief periods of time. Spontaneous hyperventilation was treated with intravenous sedation only when it was associated with distress due to increased work of breathing, severe agitation or unstable cardiovascular haemodynamics. To ensure homogeneity of the PbtO2 data, hourly measurements with a fraction of inspired oxygen (FiO2) >50% were excluded from the present analysis.

Statistical analysis

We defined brain tissue hypoxia (BTH) as PbtO2 <15 mm Hg. This threshold was used because it has been associated with poor outcome in previous reports.3 4 We then determined the frequency of brain tissue hypoxia within five prespecified EtCO2 ranges: (1) EtCO2≤24 mm Hg, (2) EtCO2 between 25 and 29 mm Hg, (3) EtCO2 between 30–34 mm Hg, (4) EtCO2 between 35 and 44 mm Hg (normal values) and (5) EtCO2≥45 mm Hg. The results in the normal range (EtCO2 35–44) were then compared with each of the four other ranges using the χ2 test. Finally, we performed a generalised estimated equations (GEE) analysis using multivariable logistic regression link function and modelling within-subject dependencies with the autoregressive process (AR-1) to account for within-subjects and between-subjects variations over time and to establish independent correlations between variables. Brain tissue hypoxia (PbtO2<15 mm Hg) was entered in the model as a dichotomised variable. EtCO2, GCS, ICP, CPP and core temperature were entered in the model as covariates. The statistical analyses were performed using SPSS 15 software (SPSS, Chicago, Illinois). p Values <0.05 were considered statistically significant.


Patient population and recording characteristics

Twenty-one patients admitted between August 2004 and May 2008 were included in the study. There were 15 women (71%) and six men (29%). The mean age was 50±16 years. The most frequent diagnosis was SAH, followed by ICH and TBI (table 1).

Table 1

Baseline characteristics

The median Glasgow Coma Scale (GCS) at the time of admission was 7 (IQR 4-8), and the Acute Physiology and Chronic Health Enquiry II score was 7 (IQR 5 to 9). Overall, we analysed 2179 hourly data points. The median interval between onset of injury and the start of monitoring was 3 days (IQR 1 to 5), and the median duration of recording was 98 h (IQR 48 to 139). Median physiological values during the entire recording are presented in table 2.

Table 2

Physiological characteristics during the study period

A significant correlation between EtCO2 values and PaCO2 was found, when EtCO2 values were compared with 227 concomitant PaCO2 values obtained by blood gases analysis (Pearson coefficient r=0.66; p<0.001). Additionally the comparison of individual mean of EtCO2 and PaCO2 revealed no significant differences (31.8±6.3 vs 33.1± 4.6; p=0.19).

Ventilation settings

Ventilator settings were available for 93% (2008/2179) of the hourly data points. During recording, the most frequent modes of ventilation were assist-control mode (volume-control (VC) or pressure-control ventilation (PCV)), followed by continuous positive airways pressure (CPAP) and synchronised intermittent mandatory ventilation (SIMV). The details of the ventilation settings during the monitoring are presented in table 3. Respiratory rates were higher, and tidal volumes were lower on continuous positive airway pressure (CPAP) compared with CMV and synchronised intermittent mandatory ventilation (SIMV) (table 3). EtCO2 was negatively correlated with both minute ventilation (r=−0.173, p<0.001) and tidal volume/kg (r=−0.281, p<0.001) during the monitoring period (figure 1).

Table 3

Comparison between set and measured ventilatory parameters

Figure 1

Correlation between end-tidal CO2 and tidal volume/kg (r=−0.281, p<0.001) (a) and between end-tidal CO2 minute volume (b) (r=−0.173, p<0.001).

Frequency of brain tissue hypoxia (PbtO2<15 mm Hg) according to EtCO2

Overall, brain ischaemia (PbtO2<15 mm Hg) occurred in 490 hourly data (22.5%) of the hourly measures (figure 2). Twenty of 21 patients had at least one episode of brain tissue hypoxia. In patients with hypocapnia, the percentage of brain tissue hypoxia was 33.9% (78/230) in patients with EtCO2<25 mm Hg, 27.7% (128/460) with EtCO2 between 25 and 29 mm Hg and 25.8% with EtCO2 between 30 and 34 mm Hg (160/620). During normocapnia (EtCO2 35–44 mm Hg), the percentage hourly data with brain tissue hypoxia were 15.7% (118/753), whereas 5.2% (6/116) of hourly data with EtCO2>44 mm Hg were associated with brain tissue hypoxia. Compared with normocapnia, the percentage of hourly data with PbtO2<15 mm Hg was significantly higher for patients with EtCO2<25 mm Hg (p<0.001), between 25 and 29 mm Hg (p<0.001) and between 30 and 34 mm Hg (p<0.001). Additionally, patients with hypercapnia (EtCO2 >44) had significantly fewer episodes of brain tissue hypoxia (p<0.001). An example of the correlation between EtCO2 and brain tissue hypoxia in a single patient is presented in figure 3. The Pearson correlation between EtCO2 and PbtO2 coefficient was stronger in patients with TBI (r=0.64; p<0.001) and with ICH (p=0.30; p<0.001) compared with SAH (r=0.27; p<0.001).

Figure 2

Incidence of brain tissue hypoxia (brain tissue oxygen pressure <15 mm Hg) events depending on end-tidal CO2 values. The discontinuous line represents the percentage of hourly episodes of brain tissue hypoxia in the entire cohort.

Figure 3

Correlation between end-tidal CO2 and brain tissue oxygen pressure in a 40-year-old man suffering from left/right frontal haemorrhage. The figure represents the 322 hourly data collected in this patient (r=0.452, p<0.001). PbtO2, brain tissue oxygen pressure.

Predictors of brain tissue hypoxia

Using a longitudinal multivariable logistic regression model that accounted for within-subject and between-subject variations over time (GEE analysis) and after adjusting for intracranial pressure, core temperature and Glasgow Coma Scale score, both lower EtCO2 levels and CPP levels were associated with an increased risk of brain tissue hypoxia (table 4).

Table 4

Physiological variables associated with cerebral ischaemia (brain tissue oxygen pressure <15 mm Hg)


In a diverse population of critically brain-injured patients, we found that a reduction in EtCO2 during continuous monitoring was independently associated with the development of brain tissue hypoxia defined as PbtO2<15 mm Hg. Eleven per cent of 2179 ETCO2 measurements were below 24 mm Hg, and another 21% were between 25 and 29 mm Hg. The frequency of tissue hypoxia progressively increased from 16% when EtCO2 was normal (35–44 mm Hg) to 34% when EtCO2 was ≤24 mm Hg. Our management protocol did not call for intentional prophylactic hyperventilation: in the majority of cases, hypocapnia occurred in the setting of higher measured-than-set respiratory rates and minute ventilation. These findings suggest that spontaneous central hyperventilation may be a common and largely unrecognised cause of brain tissue hypoxia after severe brain injury.

Previous studies performed in patients suffering from TBI or SAH have shown that episodes of induced hyperventilation may lead to a reduction in cerebral blood flow and brain tissue oxygenation.18–20 25 26 Recognition of the potentially harmful effects of excessive hyperventilation has led the Brain Trauma Foundation recently to recommend hyperventilation only as a temporising expedient for the reduction in ICP (Level III of evidence) with concomitant use of SjO2 or PbtO2 monitoring (Level III) and that PaCO2 levels below 29 mm Hg should be avoided during the first 24 h after injury.10 However, it is generally assumed that most hyperventilation in critically ill neurological patients is intentional; the possibility that central hyperventilation might compromise cerebral perfusion to a significant degree has received little attention to date.

The effects of PaCO2 on the cerebral vasculature with the development of vasoconstriction after hypocapnia and vasodilatation after hypercapnia are well known.27 Additionally, it is known that increases in PaCO2 are associated with a increased hydrogen ion (H+) concentrations, promoting vasodilation and the release of oxygen from haemoglobin to the brain tissue.28 This most likely explains our finding of a reduced number of brain tissue hypoxia events with higher EtCO2. We found that for every 1 mm Hg decrease in CPP, the risk for brain hypoxia increased by 2%, which suggests that a reduction in CPP can expose the brain to hypoxic damage, not only at extreme values but also within the therapeutic range of CPP. These findings taken together suggest that a reduction in EtCO2 and CPP in combination most likely creates the greatest risk for brain ischaemia.

The fact that both CPP and EtCO2 were independently associated with the occurrence of brain tissue hypoxia underlines their distinct effects on PbtO2. Indeed, since the arterial response to a maximal reduction in CPP is extreme vasodilatation, one could expect a reduced or abolished residual vasoreactivity to changes in CO2. However, our results show that at least part of the CO2 reactivity is preserved, even with lower CPP, since EtCO2 is independently associated with episodes of brain tissue hypoxia. This is consistent with previous animal studies, which suggest a decreased but preserved CO2 reactivity at a low mean arterial pressure.19 20 29

The most common ventilator mode used in our study cohort was CMV or assist control, which delivers a minimum number of mandatory volume- or pressure-cycled breaths per minute, and additional breaths whenever spontaneous breathing above this rate are detected, with minimal work-of-breathing for the patient. More aggressive use of opioid sedation to reduxe the spontaneous breathing rate, combined with ventilator modes such as SIMV that can limit maximal minute ventilation more effectively, may be useful for reducing the frequency of hyperventilation-induced brain tissue hypoxia after severe brain injury. In addition, close attention to EtCO2 monitoring may be to be a useful adjunct for detecting brain tissue hypoxia related to unintentional hyperventilation.

Several limitations of this study deserve attention. First, we used EtCO2 instead of PaCO2, which is only an indirect measure of PaCO2. EtCO2 tends to underestimate concurrent PaCO2 measurements, and although a reasonably good correlation can usually be demonstrated, EtCO2 may not provide a stable reflection of PaCO2 in some patients.30 However, EtCO2 can be measured continuously, whereas PaCO2 measurements are conventionally intermittent. Moreover, the gradient between EtCO2 and PaCO2 may be increased by acute pulmonary embolism, pneumonia or pulmonary oedema. Second, PbtO2 levels are profoundly influenced by increases in FiO2.31 To reduce these biases, we excluded data points when FiO2 exceeded >50%. Third, this study is based on a small number of subjects presenting with different pathologies. Although we used a statistical analysis (GEE) which accounted for within-subjects and between-subjects variations during time, further studies of larger patient cohorts are needed to determine which patients are most likely to present with brain tissue hypoxia after hyperventilation, since overall the odd ratio for EtCO2 to be associated with PbtO2<15 mm Hg remains close to 1. These data suggest that the development of physiological events such as vasospasm after SAH may impair vasoreactivity to CO2 and therefore affect the correlation between EtCO2 and PbtO2. Finally, since this is an observational study, we did not systematically control modes of ventilation and target minute ventilation to a uniform target, and so our results may be biased.

Our results underline the potential dangers of hyperventilation in patients with severe acute brain injury. In routine clinical care, EtCO2 monitoring appears to be a useful adjunct for avoiding brain tissue hypoxia in comatose patients. Further studies are needed to confirm our findings, and to determine if control of induced and spontaneous hyperventilation through the implementation of novel sedation and ventilation strategies can minimise the frequency of PBtO2 reductions in critically brain-injured patients.



  • Funding This work was supported by research grants from the Swiss National Science Foundation (EC) (PBLAB-119620), Bern, Switzerland, the SICPA Foundation, Lausanne, Switzerland (EC) and the Charles A. Dana Foundation (SAM). NB received funding from a K 12 Career Development Award (RR024157; PI: H Ginsberg) from the National Center for Research Resources (NCRR), a component of the National Institutes of Health (NIH).

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.