Moderate Hyperventilation during Intravenous Anesthesia Increases Net Cerebral Lactate Efflux

HYPOCAPNIA induced by hyperventilation and associated alkalosis have a wide range of physiological effects, including increased cerebrovascular resistance (CVR), decreased cerebral blood flow (CBF), cerebral oxygen delivery (cDO₂), and cerebral metabolism.¹  Despite routine end-tidal carbon dioxide monitoring, periods of inadvertent hyperventilation occur frequently during mechanical ventilation under general anesthesia, which may be associated with unfavorable side effects such as cognitive dysfunction and increased length of hospital stay.² 

Patients with Alzheimer disease are predisposed to postoperative cognitive dysfunction. This group of patients has an increased vasoconstrictive response to hypocapnia and concomitantly a greater increase in oxygen extraction fraction (OEF) than control patients.³  In patients with traumatic injury, vascular disorders, or meningitis, hyperventilation is associated with an impaired aerobic cerebral metabolism, reflected by an increase in net cerebral lactate efflux (cerebral metabolic rate [CMR] of lactate [CMRL]).^4–8 

However, until now there are only few studies describing the interrelation between hyperventilation and CMRL in animals and humans without cerebral diseases, and their results have not been consistent. A recent report about moderate and profound hyperventilation in anesthetized young pigs without cerebral disorder showed a reduction in regional CBF and oxygen availability, resulting in tissue hypoxia as reflected by an increase in markers of anaerobic metabolism.⁹  Similarly, investigations using magnetic resonance spectroscopy or the Kety–Schmidt technique in awake volunteers undergoing hyperventilation showed an increase in net cerebral lactate efflux.^8,10,11 

The anesthetized brain might be less vulnerable to ischemia than the nonanesthetized brain as the induction of anesthesia reduces cerebral electric activity, metabolism, and flow.¹  In humans, induction of intravenous anesthesia even may reduce cerebral lactate efflux.^12–14  But moderate hyperventilation during anesthesia also showed a trend to increase net cerebral lactate efflux though not reaching significance.^13,15,16  The relevance of this finding, however, may be limited, because of the small number of patients, which have been included in these studies.

The interrelation between moderate variations in Paco₂, CVR, CBF, global cDO₂, and cerebral metabolism in patients undergoing intravenous anesthesia is thus not fully understood.

We therefore investigated the effects of variation in arterial carbon dioxide partial pressure on cerebral hemodynamics and metabolism in 30 cardiac surgical patients undergoing intravenous anesthesia. We hypothesized that moderate hyperventilation, when compared with moderate hypoventilation, will reduce CBF and cDO₂ to an extent which might impair cerebral aerobic metabolism.

The prospective study was designed and performed in a controlled, crossover design at the University of Göttingen Medical Center aiming at changes in CBF, CBF velocity (V), and the metabolic effects of hyper- versus hypoventilation in anesthetized patients. Each patient served as his own control. Approval was obtained from the local institutional review board (Medical Ethical Committee of the Georg-August-University of Göttingen, Göttingen, Niedersachsen, Germany; No. 07/09/90). Study period was 27 months (February 20, 1991 until May 10, 1993).

The primary endpoints of the trial were changes in CBF, blood flow velocity of the middle cerebral artery (V_MCA), cDO₂, CMRL, CMR of oxygen (CMRO₂), and CMR of glucose. The secondary endpoints were changes in cerebral zero-flow pressure (ZFP), effective cerebral perfusion pressure (CPP_eff), and CVR.

Due to logistic reasons, we could perform only 1–2 measurements per month. Thus, standard screening procedures could not be applied in this crossover trial. Patients were eligible for inclusion if scheduled for elective coronary surgery. Exclusion criteria were being older than 80 yr of age, female sex, patient refusal, active neurological disease, and a history of cerebrovascular disease, brain injury, or intracranial surgery. All patients were informed of the purpose of the study and provided written informed consent before being enrolled. None of the eligible patients refused inclusion of the trial. There were no dropouts during the study period.

The intersubject and intrasubject variability of CBF and cerebral lactate metabolism has been reported in earlier studies.^10,13,15,16  However, there was a lack of data regarding the variance of the CMRL measurement method in anesthetized patients, which was necessary for an exact sample size calculation for this crossover trial. We expected a 50% difference of CMRL with an estimated effect size of 0.7–0.8. For a statistical power of 0.8–0.9, the sample size had to be between 24 and 30 patients. Therefore, we projected a sample size of 30 patients.

Individual medications were continued until surgery. Anesthesia was induced by intravenous administration of 7 µg/kg of fentanyl, 0.2 mg/kg of midazolam, and 0.1 mg/kg of pancuronium. Anesthesia was maintained with 10 µg·kg⁻¹·h⁻¹ of fentanyl and 150 µg·kg⁻¹·h⁻¹ of midazolam. The anesthesia procedure, the details of mechanical ventilation, and the methods of insertion of catheters have been described in detail in a previous report.¹⁴ 

Cerebral blood flow was measured with the use of modified Kety–Schmidt inert gas saturation technique with argon as a tracer gas.^14,17,18  The wash-in period was 10 min. Blood samples were obtained simultaneously from the arterial and jugular bulb catheters at a constant rate of 0.5 ml/min by a high-precision aspiration pump with gas-tight Hamilton glass syringes. The withdrawal rate for probes of the argon end-concentration was 5 ml/20 s. A brain–blood partition coefficient of 1.10 was used to calculate CBF.^19,20 

Blood flow velocity in the proximal (M1) segment of the MCA was measured by transcranial Doppler sonography as extensively described in earlier reports.^14,21,22  Because transcranial Doppler measurements of V_MCA from the transtemporal window fail with above average incidence in elder female patients, we included only male patients in this study.^23,24 

Measurements were performed at two different Paco₂ levels, approximately 50 and 30 mmHg, in a randomized sequence before surgery. All measurements were performed during hemodynamic and respiratory steady-state conditions. The time interval between the measurements was 20 min. Blood samples were drawn twice, at the beginning and end of each argon wash-in period, to measure hemoglobin concentration, blood gas analysis (ABL; Radiometer, Copenhagen, Denmark), and blood glucose and lactate concentrations (enzymatic tests kids; Boehringer, Mannheim, Germany). For comparisons with CBF measurements, V_MCA was averaged during the 10-min period of each argon wash-in maneuver. End-expiratory concentrations of carbon dioxide were continuously recorded to ensure a stable Paco₂ during argon saturation.

Cerebral ZFP was calculated at the beginning and end of each CBF measurement from two simultaneous 10-s recordings (two breathing cycles) of the V_MCA envelope and arterial pressure curves. During each 10-s period, we first averaged diastolic, mean, and systolic data of arterial blood pressure (ABP) and V_MCA to obtain a pressure–flow velocity plot. Cerebral ZFP was then extrapolated by linear regression analysis of the ABP–V_MCA relationship. The ABP axis intercept of the regression line determines the ZFP.^22,25  The cerebral ZFP was used as a measure of the effective downstream pressure of the cerebral circulation. Consequently, CPP_eff and CVR were calculated as CPP_eff = mean ABP − ZFP and CVR = CPP_eff·CBF⁻¹, respectively. Cerebrovascular CO₂ reactivity was calculated from the slope of the linear regression line of the relationship between CBF and Paco₂ as well as V_MCA and Paco₂. Relative CO₂ reactivity was calculated as the percentage of change in CBF or V_MCA per mmHg change in Paco₂.²⁶ 

CMRO₂, CMR of glucose, and CMRL were calculated based on the reversed Fick principle, multiplying CBF by the difference in arterio-jugular venous content of oxygen (AJVDO₂), lactate (AJVDL), and glucose.⁸  By definition, positive CMR values indicate consumption or net influx, and negative values indicate production or net efflux. For AJVDL and CMRL, we thus expected negative values in case of cerebral lactate production.

The OEF was calculated as the ratio between AJVDO₂ and arterial oxygen content. The lactate–oxygen index (LOI), which relates cerebral lactate production to oxygen extraction, was defined as LOI [] = AJVDL [mm] AJVDO₂⁻¹ [mm]. The lactate–glucose index (LGI), that is, the fraction of glucose consumption that is metabolized and excreted from brain as lactate, was defined as the ratio between the arterio-jugular venous content differences of lactate and glucose. For these calculations, contents of oxygen and glucose concentrations have been converted in to mm. Because one molecule of glucose is metabolized to two molecules of lactate, an LGI of, for example, −0.1 indicates, that 5% of the cerebral glucose consumption have been metabolized to lactate.^5,8,27 

The results presented in tables are expressed as mean (SD) unless otherwise stated. To provide an estimate of the effect of hypocapnia and its clinical meaningfulness, we calculated mean differences (MD) and their 95% CIs (MD; 95% CI, lower bound, upper bound; P value).²⁸  The difference between hyperventilation and hypoventilation was calculated using t tests for paired data or Welch test and nonparametric Wilcoxon signed-rank tests, if indicated. To prevent from type I error inflation, all primary endpoints were tested by one-way ANOVA for repeated measurements followed by Bonferoni multiple comparison tests. All statistical analyses were performed two-sided, and a P value of less than 0.05 was considered to be significant.

Calculations were performed using SPSS 17 (IBM SPSS Statistics, Armonk, NY), and graphs were made using Prism 6.0c (GraphPad Software, La Jolla, CA).

Sample size calculation was done with G*Power 3 (University of Düsseldorf, Department of Psychology, Düsseldorf, Germany).²⁹ 

A total of 30 male patients were included in the study. The mean age of the patients was 56(8) yr (median, 58; range, 41–78 yr), mean height 173(6) cm, and mean body weight 77(9) kg. In one patient, CBF could not be measured during hypoventilation, because of technical problems during jugular venous blood sampling. Hemodynamic and metabolic data are presented in table 1.

Due to the controlled adjustment of mechanical ventilation, the variability of Paco₂ at both target levels was small. The blood temperature of the patients was effectively kept constant. Only one of our patients had diabetes mellitus. None of the patients showed increased levels of blood glucose.

The effects of ventilatory changes on the cerebral circulation were substantial. Hyperventilation reduced CBF by 60%, and V_MCA by 41%, when compared with hypoventilation. This reduction was predominantly caused by increased CVR (MD, 0.95; 95% CI, 0.75–1.11 mmHg·ml⁻¹·min⁻¹·100 g⁻¹; P < 0.001) and decreased CPP_eff (−14%). The decrease in CPP_eff during hyperventilation occurred because of a significant increase in the cerebral ZFP (MD, 13; 95% CI, 9–16 mmHg; P < 0.001), which exceeded the small increase in mean arterial pressure (MD, 5; 95% CI, 2–8 mmHg; P = 0.003). The decrease in CBF and cDO₂ (by 58%) during hyperventilation was associated with a pronounced decrease in the venous jugular bulb oxygen saturation (MD, 31; 95% CI, 34–28%; P < 0.001) and venous jugular bulb partial pressure of oxygen (PjvO₂; MD, 21; 95% CI, 19–22 mmHg; P < 0.001; fig. 1). The cDO₂/CMRO₂ ratio changed from 4.5:1 to 1.8:1; that is, the OEF during hyperventilation markedly increased from 0.24 to 0.57.

Mean arterial-jugular venous difference of oxygen, glucose, and lactate changed significantly during moderate hyperventilation when compared with hypoventilation (table 1). The mean cerebral efflux of lactate significantly increased, by 2.0 µmol·min⁻¹·100 g⁻¹, whereas mean CMRO₂ and CMR of glucose remained constant. LOI and LGI significantly decreased, that is, became more negative.

Cerebrovascular CO₂ reactivity of CBF was 2.02 (1.18) ml·min⁻¹·100^–1·mmHg⁻¹, corresponding to a relative change of 2.79 (0.77) %·mmHg⁻¹. The cerebrovascular CO₂ reactivity of mean V_MCA was 1.18 (0.48) cm·s⁻¹·mmHg⁻¹, corresponding to a relative change of 2.03 (0.50) %·mmHg⁻¹, which was significantly lower than the relative CO₂ reactivity of CBF (MD, −0.76; 95% CI, −0.96 to −0.56; P < 0.001).

We investigated the effects of variations in arterial carbon dioxide partial pressure on cerebral hemodynamics and metabolism in cardiac surgical patients undergoing fentanyl or midazolam anesthesia. Compared with hypoventilation, moderate hyperventilation was associated with a significant reduction in CBF, cDO₂, and PjvO₂. The mean cerebral efflux of lactate significantly increased, whereas mean CMRO₂ and CMR of glucose remained constant.

Hyperventilation reduces Paco₂ and decreases extracellular H⁺, leading to cerebral vasoconstriction and consecutively to reduced CBF and cDO₂.¹  An associated increase in net cerebral efflux of lactate at low Paco₂ levels in principle may be explained by different mechanisms:

• Dissociation of oxygen-binding curve to the left as a result of the respiratory alkalosis (Bohr effect)^30–32 

• Alkalosis induced change of redox systems of lactate/ pyruvate and NADH/NAD⁺,^16,33–35  and

• Severe cerebral hypoperfusion with tissue hypoxia.^9,36,37 

Recent investigations on lactate kinetics and oxygenation using lactate isotopes demonstrate simultaneous lactate uptake and release in the brain.³⁸  In addition to glucose and ketone bodies, lactate is also known to be an essential part of cerebral energy metabolism. Recent trials have shown that the glucose taken up by astrocytes is converted to lactate, and that the lactate released from astrocytes may be taken up by neurons and used as energy, especially in activated neurons, referred to as the astrocyte–neuron lactate shuttle hypothesis.³⁹  Thus, partial metabolic compartmentalization appears to exist between astrocytes and neurons, with astrocytes feeding the neurons with lactate generated from glycolysis upon cerebral activation.⁴⁰ 

The magnitude of change in mean CMRL of our patients was 2.0 µmol·min⁻¹·100 g⁻¹, which was greater than expected. Previous studies have shown that absolute levels of CMRL may considerably vary depending on the group of patients and the level of consciousness.^14,15,41,42  Absolute values of net cerebral lactate efflux in our patients thus have to be interpreted with care. A slight lactate efflux at hypocapnia may not necessarily indicate tissue hypoxia. However, the increase in AJVDL and net cerebral lactate efflux associated with hypocapnia and the concomitant decrease in CBF might be suspicious for anaerobic metabolism in relatively ischemic brain regions.

It seems unlikely that the increase of net cerebral lactate efflux at low Paco₂ levels might solely be caused by alkalosis-induced enzymatic effects. There are no oxygen stores in the brain in contrast to myoglobin which stores oxygen in the muscle. Thus, the rate of oxygen delivery from the blood to brain tissue critically depends on the vessel-to-tissue oxygen partial pressure (P_tio₂) gradient and the efficiency of oxygen transfer from the capillary bed.^43,44  A definite ischemic threshold for brain tissue oxygenation has not yet been defined. Jones et al.⁴⁵  demonstrated that CBF less than 18 ml·min⁻¹·100 g⁻¹ in awake monkeys results in irreversible brain tissue infarction. Michenfelder et al.⁴⁶  reported critical CBF values of approximately 10–20 ml·min⁻¹·100 g⁻¹ in patients with ischemic changes in electroencephalography during carotid endarterectomy. In awake humans under normocapnic conditions, the cDO₂/CMRO₂ ratio is approximately 3:1 corresponding to an OEF of 0.33.^8,47  An OEF of greater than 0.4 in patients with traumatic head injury corresponded to a critically increased microdialysis lactate/pyruvate ratio which might reflect a mismatch between substrate demand and delivery on a cellular level.⁴⁸  In our patients, the OEF considerably exceeded the threshold of 0.4 during hyperventilation, mainly caused by a reduction of cDO₂ while CMRO₂ remained unchanged.

Most investigators have considered jugular venous PO₂ below 20 mmHg and tissue PO₂ values below 10 mmHg as pathological. Clausen et al.⁹  showed that even moderate hyperventilation (Paco₂ = 30 mmHg) leads to a critical reduction of regional CBF below 18 ml·min⁻¹·100 g⁻¹ in 22% of observed pigs; the tissue oxygen pressure decreased below 10 mmHg in 30% of the animals undergoing moderate hyperventilation. Furthermore, recent investigations showed that the final diffusion gradient from the microcirculation to the mitochondria is quite small.⁴⁹  Then, oxygen tension might play an essential role in mitochondrial cellular oxygen sensing and oxygen-regulated gene expression in clinical situations of low cDO₂.^50,51 

Although net cerebral lactate efflux during hyperventilation increased in our patients, CMRO₂ did not significantly change. Similarly, several previous studies found a reduction of tissue or cerebral venous oxygen tensions during moderate hyperventilation, but no decrease in CMRO₂ could be demonstrated.^{8,10,15,16,52,53} 

In contrast, other investigations could demonstrate that hyperventilation leads to a substantial decrease in CMRO₂^35,54  or tissue oxygen pressure.⁹  However, investigations on CBF and cerebral metabolism with variations of Paco₂ in anesthetized patients or volunteers without cerebral disease are scarce and also showed an unchanged CMRO₂ at moderately low Paco₂ levels.^8,13,15,55  The Kety–Schmidt method measures only global CBF and metabolism. In case of regional hypoperfusion with increased CMRO₂ in other regions of the brain, global CMRO₂ may be unaffected. Only when global oxygen availability decreases below oxygen demand, CMRO₂ will decrease. The results of our report demonstrate that moderate hyperventilation, when compared with hypoventilation, leads to a significant decrease in venous jugular bulb oxygen saturation, PjvO₂, AJVDO₂, CBF, and thus cDO₂. Net cerebral lactate efflux increased, which was associated with more negative LOI and LGI. Thus, our data do not indicate a severely disordered energy metabolism after a short phase of moderate hyperventilation, but it suggests that cerebral aerobic metabolism may partly be impaired. However, the clinical significance of these findings remains unclear, because in humans without cerebrovascular or traumatic brain injury detrimental effects of hyperventilation in terms of morphologic or histologic changes have not yet been demonstrated.

For the analysis of our data, different methodological aspects have to be considered.

First, the a priori sample size calculation was based on estimation of the effect size because of a lack of data regarding CMRL. A post hoc calculation, however, showed a statistical power of 97% (n = 30, α error probability = 0.05; effect size of 0.73).

Furthermore, the type of anesthesia may have potential influence on the results of our study. Induction of anesthesia with fentanyl and midazolam leads to a moderate but proportional reduction in CBF and cerebral metabolism.^12,14,56  The cerebrovascular CO₂ reactivity in our patients favorably compared with data in conscious patients. Although we thus have no reason to assume that intravenous anesthesia with fentanyl and midazolam per se may have affected the PCO₂-induced changes in CMRL in our patients, the results of this study cannot a priori be extrapolated to other types of anesthesia.

Similarly, the external validity of our data could be limited by the fact that our patients were suffering from coronary artery disease and concomitant asymptomatic cerebrovascular disease cannot completely be excluded despite normal cerebrovascular CO₂ reactivity. Therefore, the conclusions from our study results should be limited to this patient population.

The changes in CBF induced by hyperventilation and hypoventilation are related to changes in CVR and CPP. The calculation of these variables commonly requires measurements of intracranial pressure. We used a minor invasive method to estimate cerebral ZFP and CPP_eff by extrapolating pressure–flow velocity plots using recordings of ABP and V_MCA. This technique, however, is well established and may even provide a more meaningful quantification of the cerebral downstream pressure than intracranial pressure, particularly in the absence of intracranial hypertension.^6,22,57,58  The finding that moderate hyperventilation leads to a small but significant reduction of CPP_eff due to a significant increase in ZFP is in accordance with previous trials investigating the effects of hyperventilation on the effective downstream pressure of the cerebral circulation.^6,22,57,58 

The Kety–Schmidt method has been considered a reference method for estimating global CBF.^59–61  In contrast to positron emission tomography or single-photon emission computed tomography, the Kety–Schmidt technique is largely independent of assumptions regarding tracer kinetics, but might slightly overestimate global CBF in case of incomplete cerebral tracer saturation. Because of the crossover design of our study, a slight systematic overestimation of CBF has minor importance for the interpretation of hemodynamic and metabolic changes.

In our investigation, we studied CBF and metabolism aiming at Paco₂ levels of 30 and 50 mmHg. We chose these Paco₂ levels because they roughly reflect the range of unintended variations of Paco₂ that often occur in routine clinical practice. The lack of data during normocapnia is a potential limitation of our study. Individual extrapolation of CBF at a Paco₂ of 40 mmHg gave a mean CBF under normocapnia of 48 (10) ml·min⁻¹·100 g⁻¹. The absolute and relative CO₂ reactivity of CBF is in accordance with previous studies on the cerebral circulation during general anesthesia. This underlines the validity of our hemodynamic measurements.^{8,13,18,55,62}  Interestingly, the relative cerebrovascular CO₂ reactivity of CBF was significantly higher than cerebrovascular CO₂ reactivity of V_MCA. The most probable explanation is that changes in Paco₂ do not only cause changes in vascular diameter at the arteriolar level but might also cause minor changes in diameter of the MCA, resulting in a slight systematic difference between relative changes in flow and flow velocity. A similar phenomenon previously had been described during hypothermic cardiopulmonary bypass.¹⁴ 

In conclusion, moderate hyperventilation, when compared with hypoventilation, increased net cerebral lactate efflux and decreased LOI and LGI in cardiovascular patients undergoing fentanyl–midazolam anesthesia. These metabolic changes suggest that the observed decrease in CBF, cDO₂, and PjvO₂ may partly impair cerebral aerobic metabolism at clinically relevant levels of hypocapnia.

The authors gratefully thank Gerhard H. Visser, M.D., Ph.D. (Heemstede, The Netherlands), for his helpful comments on the interpretation of data, and Sanne Hoeks, Ph.D., Erasmus MC (Rotterdam, The Netherlands), for her expert advice in statistical analysis.

Support was provided solely from institutional and/or departmental sources.

The authors declare no competing interests.