High-dose Remifentanil Does Not Impair Cerebrovascular Carbon Dioxide Reactivity in Healthy Male Volunteers

After obtaining institutional review board approval (University of Vienna, Vienna, Austria) and written, informed consent, we studied 10 healthy male volunteers. We excluded subjects with a history of drug or alcohol abuse, a surgical procedure in the preceding 6 months, or any detectable abnormalities on physical examination.

To initiate the study, unpremedicated, fasting subjects underwent intravenous fluid loading with 10 ml/kg lactated Ringer's solution. A tightly fitting facemask was applied, and the subjects breathed a mixture of oxygen and air with a fraction of inspired oxygen of 0.5. The subjects maintained normal minute ventilation, guided by end-tidal carbon dioxide (ETco²) values close to 40 mmHg, with blood gas controls of Paco². Mean CBFv by TCD of the left middle cerebral artery before baseline assessment of rCBF by Xe-CT was then performed (NORMO-PRE).

A loading dose of remifentanil at a rate of 10 μg · kg⁻¹· min⁻¹was commenced via  an infusion pump (Perfusor Secura FT; B. Braun, Melsungen, Germany) until loss of consciousness occurred and was then decreased to 2 or 4 μg · kg⁻¹· min⁻¹according to the randomization. Chest wall rigidity, if present, was noted. A propofol bolus of 0.5–1 mg/kg was administered to facilitate insertion of a laryngeal mask. Mechanical ventilation was adjusted sequentially to achieve arterial Paco²values of 35–45 mmHg (NORMO-2), 25–32 mmHg (HYPO-2), and 50–55 mmHg (HYPER-2). Cisatracurium bromide was administered in aliquots of 10 mg intravenously if chest wall rigidity prevented achievement of target Paco²values. At each level of Paco², TCD measurements before Xe-CT were performed to avoid confounding effects of xenon and to obtain data under conditions of normocapnia, hypocapnia, and hypercapnia.

The remifentanil infusion dose was then increased to 4 μg · kg⁻¹· min⁻¹, and additional TCD and Xe-CT measurements were performed during hypercapnia (HYPER-4) and hypocapnia (HYPO-4). Thereafter, the remifentanil infusion was discontinued, and the final TCD measurements and Xe-CT were made during spontaneous ventilation with ETco²and Paco²in the normal range (NORMO-POST).

To avoid the possible influence of the sequence of remifentanil dosing, we actually randomized subjects to start with either the low-dose (2 μg · kg⁻¹· min⁻¹) or high-dose (4 μg · kg⁻¹· min⁻¹) infusion. Thus, in the subjects who were randomized to start with the high-dose infusion, the sequence of measurements was (1) NORMO-PRE, (2) NORMO-4, HYPO-4, HYPER-4, (3) HYPER-2, HYPO-2, and (4) NORMO-POST. We wished to focus our attention on carbon dioxide reactivity rather than absolute changes in CBF during remifentanil infusion. Therefore, we decided to decrease radiation exposure to our subjects by omitting a second normocapnic CBF assessment for each dose. We accepted that this would diminish the statistical power of the normocapnic data by decreasing the number of subjects in this group to five.

At each remifentanil infusion dose adjustment or minute ventilation change, sufficient time for equilibration (minimum 15 min) was allowed. After discontinuation of remifentanil, we recorded the time until recovery of spontaneous ventilation and return of full consciousness with removal of the laryngeal mask. The presence of emergence delirium, nausea, vomiting, or respiratory depression was noted.

To avoid the confounding effects of altered hemodynamics, hypotension was treated with intravenous phenylephrine (0.5–2 μg · kg⁻¹· min⁻¹) and bradycardia with intravenous glycopyrrolate to maintain mean arterial pressure or heart rate within 10% of baseline values. Standard monitoring (Hewlett Packard, Palo Alto, CA) was used, including a radial artery cannula for continuous monitoring of blood pressure and blood gas sampling (AVL Opti, Critical Care Analyzer; AVL, Graz, Austria) to adjust ventilator settings.

Xenon-enhanced computed tomography was performed on a Somatom PLUS scanner (Siemens, Erlangen, Germany) using an optimized 3-min washin/3-min washout protocol. 14End-expiratory xenon and carbon dioxide concentrations were measured throughout the examination. In the washin phase, a mixture of 30% xenon gas and 70% oxygen was inhaled. One axial slice was evaluated at the level of the deep white matter of the left and right hemisphere. Time-dependent xenon concentration within various tissue segments in the brain was used to derive CBF in each tissue volume (voxel) of the computed tomographic image. The slice thickness was 8 mm, and the matrix of the calculated functional maps was 256 × 256 voxels. Maps of rCBF and xenon-enhancement/voxel were calculated after the measurement based on the formulations of Kety and Schmidt. 15,16For the calculation of rCBF, the recorded xenon end-tidal values are converted into an estimated time-dependent computed tomographic enhancement of arterial blood. 17Because of the small pixel size, a single compartment model for further analysis was assumed. 14–17 

For further evaluation of vascular reactivity, rCBF (in ml · 100 g⁻¹· min⁻¹) was traced over the normocapnic, hypocapnic, and hypercapnic measurements and plotted for six regions of interest for the investigated slice. These included left frontal, left parietal, left occipital, right occipital, right parietal, and right frontal regions.

The rCBF reactivity was defined as the rCBF change per mmHg change in Paco²(ml · 100 g⁻¹· min⁻¹· mmHg⁻¹). Relative rCBF reactivity (percent change/mmHg) was determined by the formula:

where hypo = during hypocapnia, and hyper = during hypercapnia.

As a second method to assess cerebral vascular reactivity, 18TCD of the left middle cerebral artery was performed before every Xe-CT using a 2-MHz pulsed Doppler monitoring probe placed at the left transtemporal window (Multidop 10; DWL, Sipplingen, Germany). Insonation of the middle cerebral artery was initiated at a depth of 45 mm. The correct position was determined manually before each measurement by first locating the bidirectional flow pattern typical of the bifurcation of the internal carotid artery into middle cerebral and anterior communicating arteries. Then, the insonation depth was increased to the point of the maximal signal intensity, and a series of readings of middle cerebral artery blood flow velocity was obtained, including mean (Vm), peak systolic (Vs), and peak diastolic (Vd) velocity. A pulsatility index (PI) was calculated by dividing the difference between Vs and Vd by Vm:

Vascular reactivity was defined in terms of mean CBFv as the mean CBFv change per mmHg change in Paco²(cm · s⁻¹· mmHg⁻¹). Relative mean CBFv reactivity was determined by the formula:

where CBFv = mean CBFv, hypo = during hypocapnia and hyper = during hypercapnia.

All values are given as mean ± SD. Data were analyzed by analysis of variance for repeated measurements followed by a Fisher post hoc  test. To assess rCBF and mean CBFv changes produced by different drug doses in a given group, a paired Student t  test was used. A P  value less than 0.05 indicated a significant difference. An a priori  power analysis was performed. This analysis demonstrated that a sample size of 10 patients would provide approximately 80% power at α= 0.05 to detect a CBF change of 30 ml · 100 g⁻¹· min⁻¹from baseline. This information was based on an expected 30% CBF change from baseline for remifentanil 19and a CBF change of 2% per mmHg carbon dioxide. 11 

All volunteers (n = 10; age, 27 ± 6 yr; weight, 71 ± 9 kg; height, 179 ± 4 cm) completed the study without complication. Data for rCBF, hemodynamics, and mean CBFv are summarized in table 1. Hemodynamics (mean arterial pressure and heart rate) remained unchanged through the study. Requirement for phenylephrine or glycopyrrolate was similar at both remifentanil doses as well as the use of cisatracurium to facilitate ventilation during the study (table 2). Two volunteers required a cisatracurium bolus after induction of anesthesia because of chest wall rigidity. The rCBF and mean CBFv values were not significantly different before and after remifentanil infusion but were significantly decreased during the infusion of remifentanil. However, cerebrovascular reactivity to carbon dioxide was preserved.

During remifentanil infusion at 2 and 4 μg · kg⁻¹· min⁻¹, rCBF and mean CBFv increased as the Paco²increased. The average slope of the rCBF reactivity was 0.56 ± 0.27 and 0.49 ± 0.28 ml · 100 g⁻¹· min⁻¹per mmHg Paco²change, at 2 and 4 μg · kg⁻¹· min⁻¹remifentanil, respectively. This corresponded to a relative CBF reactivity of 1.9 ± 0.8 and 1.6 ± 0.5% change per mmHg, respectively (not significant).

The average slopes of the mean CBFv reactivity were 1.61 ± 0.95 and 1.54 ± 0.83 cm/s per mmHg Paco²change at 2 and 4 μg · kg⁻¹· min⁻¹remifentanil, respectively. This corresponded to relative CBF reactivities of 1.86 ± 0.59 and 1.79 ± 0.59% change per mmHg Paco², respectively (not significant).

Bivariate calculations (Pearson) between Xe-CT and TCD on raw data and relative carbon dioxide reactivities to identify a correlation slightly missed significance because of variance and a very small sample size (n = 10). Nonetheless, rCBF measured with Xe-CT and mean CBFv measured with TCD showed a consistent and significant response to alterations in Paco², which is demonstrated in figure 1, where individual CBF data of the left parietal region of interest, the brain region corresponding to the left middle cerebral artery, and mean CBFv are shown. The results were similar in all other regions of interest (data not shown).

The mean duration of anesthesia was 102 ± 12 min, and the total study duration was 4.3 ± 0.3 h. Emergence from anesthesia was rapid. Patients breathed spontaneously an average of 2.3 ± 1.7 min after discontinuation of remifentanil. The laryngeal mask was removed after a mean of 7.4 ± 2.9 min, with no incidence of respiratory depression.

This study demonstrates that human CBF reactivity to arterial carbon dioxide tension remains intact under remifentanil–air anesthesia. Unlike previous studies, our findings are based on the simultaneous direct (Xe-CT) and indirect (TCD) assessment of CBF, which both confirmed that carbon dioxide reactivity is preserved independently of the dose (2 and 4 μg · kg⁻¹· min⁻¹) of remifentanil.

Similar observations were made by Baker et al.  11with 0.35 μg · kg⁻¹· min⁻¹remifentanil and background N²O anesthesia, but they noted greater CBF reactivity: 3.6 ± 1.2% per mmHg carbon dioxide compared with our findings of 1.9 ± 0.8 and 1.6 ± 0.5% per mmHg carbon dioxide (at 2 and 4 μg · kg⁻¹· min⁻¹remifentanil infusion, respectively). In another study from the same group, 20carbon dioxide reactivity for remifentanil–N²O anesthesia was 2.6 ± 3.6% per mmHg carbon dioxide and was similar to an equipotent fentanyl–N²O anesthesia (4.5 ± 2.2% per mmHg carbon dioxide). The quantitative difference in CBF reactivity between the study of Baker et al.  and our study might be because we used 5- to 10-fold greater doses of remifentanil and made measurements over a different Paco²range. Also, CBF reactivity to carbon dioxide has traditionally been described by a linear model, whereas the true response curve is sigmoidal. This brings up an intriguing question of whether or not remifentanil dosing causes a shift in CBF reactivity. CBF increases from hypocapnia to normocapnia during low-dose remifentanil but not during high-dose remifentanil (table 1). However, there is no difference in CBF reactivity from hypocapnia to hypercapnia between the two doses of remifentanil (fig. 1). This suggests that the cerebrovasoconstrictor effect of high-dose remifentanil overrides normocapnia but not hypercapnia. Unfortunately, because of the small sample size of normocapnic patients, these data did not reach statistical significance, and the difference in response slopes could be an artifact of where on the sigmoidal carbon dioxide response curve CBF reactivity was plotted. Furthermore, we have not compared CBF reactivity to the awake state, which is necessary to identify a quantitative shift under remifentanil anesthesia.

Although it varied in magnitude, rCBF significantly decreased in every region of interest during both doses of remifentanil infusion (2 and 4 μg · kg⁻¹· min⁻¹;fig. 1).

After discontinuation of remifentanil, rCBF was not significantly different from baseline values before anesthesia, presumably reflecting remifentanil's rapid elimination in plasma. In other descriptions of cerebral hemodynamics during remifentanil–N²O 11,20or fentanyl–N²O anesthesia 10with doses lower than in our study, under comparable Paco²conditions, CBF was reported as 31 ± 7, 36 ± 11, and 31 ± 4 ml · 100 g⁻¹· min⁻¹, respectively. In contrast, the values for cumulative CBF that we measured in five normocapnic volunteers in our study were 21 ± 2 and 16 ± 1 ml · 100 g⁻¹· min⁻¹for remifentanil at 2 and 4 μg · kg⁻¹· min⁻¹, respectively.

These differences could be accounted for by various reasons. Higher CBF values in studies using N²O might be attributed to its potent cerebral vasodilating effects, which caused a significantly more pronounced increase of CBF in all gray matter regions compared with remifentanil in conscious humans. 12 

Lower CBF values in our study likely result from a direct constrictive effect of high-dose opioids on cerebral blood vessels. We attempted in our study to minimize the influence of autoregulation on CBF by maintaining constant blood pressure with a phenylephrine infusion, which itself has no direct effect on cerebral vasculature. 21,22Therefore, decreased rCBF points toward opioid-induced vasoconstriction. Furthermore, opioid administration is generally believed to suppress brain activity. For example, in animal 23and human 24studies, infusion of sufentanil induces a significant decrease in cerebral metabolic rate for oxygen coupled with changes in cerebral hemodynamics. Although we did not measure cerebral metabolic rate, with intact metabolic coupling, reduced CBF might reflect diminished metabolic demands under the high-dose remifentanil administration in our study.

However, some investigations on human brain revealed significant rCBF increases and failed to demonstrate a global suppression of neuronal activity after the administration of fentanyl 3or remifentanil 12,25at a low dose. In these studies, both drugs induced changes in relative rCBF in areas involved in pain processing as assessed with positron emission tomography 3,25or contrast-enhanced magnetic resonance imaging. 12Although one could argue that agonist actions of remifentanil on neuron-located opioid receptors can result in an increased synaptic energy demand, 26the more than 20-fold higher remifentanil dose in our study makes it unlikely that we would have observed effects other than related to vasoconstriction or brain activity suppression of opioids.

Our findings are in keeping with others that have indicated that remifentanil exerts a dose-related suppression of mean CBFv while maintaining vascular carbon dioxide reactivity and flow-metabolism coupling. In a study in cardiac patients ventilated with oxygen without background anesthesia, 13CBFv was decreased 9% (not significant) by 1 μg · kg⁻¹· min⁻¹remifentanil and by 31% (P < 0.05) by 3 μg · kg⁻¹· min⁻¹remifentanil. The authors interpreted these data to suggest that 1 μg · kg⁻¹· min⁻¹may be the threshold above which remifentanil progressively depresses mean CBFv. A similar dose- response effect has been reported for sufentanil, 4with a 27–30% decrease in mean CBFv noted at 6 μg/kg but not at 3 μg/kg. In our study, the missing dose–response relation of remifentanil 2 and 4 μg · kg⁻¹· min⁻¹with mean CBFv suggests a near maximum response of the drug. The decreased pulsatility index we saw with hypercapnia represents decreased vascular resistance and also suggests that the bulk of change in CBF is in response to increases in Paco².

The hemodynamic responses to remifentanil we encountered are well known. Opioid-induced decreases in heart rate and blood pressure required the administration of glycopyrrolate and phenylephrine to prevent bradycardia and hypotension. In fact, the cardiovascular support required in this study was modest, with some volunteers requiring no support at all. It was not significantly different at the two dosing regimens, suggesting that a maximal hemodynamic effect had been achieved. Other side effects were a moderate chest wall rigidity due to the high induction dose of remifentanil, but only two patients needed a 10-mg bolus of cisatracurium after induction to facilitate ventilation. All other volunteers received cisatracurium, a drug with no influence on cerebral hemodynamics, 27,28during the study if chest wall rigidity prevented achievement of target Paco²values.

This study has certain methodologic limitations. Inhaled xenon, an inert gas used for measurements of CBF, is known to influence CBF itself in a dose-dependant fashion. Autoradiographic studies in rats revealed a doubling of rCBF in some neocortical structures after 80% xenon inhalation for 1- and 2-min periods but no effect on rCBF after 40% xenon inhalation for 1 min. 29Microsphere studies in baboons revealed a 17% increase in CBF when a 35–42% xenon–oxygen mixture was inhaled for more than 2 min, 30and in humans, 30–35% xenon inhaled for 4.5 min was associated with CBF increases as much as 35% above baseline. 31 

Obrist et al.  32attempted to quantify xenon-induced cerebral blood flow activation during xenon computed tomography. They developed a computer simulation model based on repeated TCD measurements of blood velocity during 4.5 min of stable xenon inhalation. In contrast to the peak 35% increase in blood flow velocity encountered during xenon inhalation, a computer-assisted analysis revealed only minor (3–5%) increases in calculated CBF. 32In our study, we attempted to minimize activation bias by keeping xenon inhalations brief (3 min) and the inspired xenon concentration low (30%). It is reasonable to expect that at these levels, exposure to xenon would have exerted little influence on the validity of our CBF measurements. Under conditions of reduced global CBF, the situation is different. The concentration of diffusible tracer in tissue, which is necessary for measurement of enhancement over time, especially in low-flow compartments, may be comparably low. Therefore, a greater error with underestimation of real rCBF than in normal flow conditions mainly in the white matter is possible. This might explain why almost ischemic-looking rCBF values during remifentanil exposure in hypocapnia unlikely were ischemic. First, the expected ischemic range of 15–20 ml · 100 g⁻¹· min⁻¹refers to gray matter. 33Because our regions of interest represent an average of both gray and white matter, lower values will be tolerable. Second, in a study about rCBF measurements in patients with acute ischemia rCBF, values in the range 11–20 ml · 100 g⁻¹· min⁻¹were found in at least 20% of the area of the nonischemic hemisphere. 34In the same study, in follow-up examinations, no infarct could be found in those areas. Third, both studies were performed in awake patients while under high-dose opioid anesthesia; because of additional suppression of brain activity, decreased rCBF values may also reflect decreased metabolic demand. Finally, all volunteers required ultrashort and predictable time to regain consciousness after anesthesia or to remove the laryngeal mask. None of the volunteers showed any neurologic sign indicative of a preceding ischemic period.

In conclusion, we have demonstrated that infusion of high-dose remifentanil (2 and 4 μg · kg⁻¹· min⁻¹) in healthy male volunteers decreases CBF but leaves cerebrovascular reactivity to acute changes in Paco². Our results are based on both direct (Xe-CT) and indirect (TCD velocity) measurements of CBF, which both significantly increased with an increase in Paco².

The authors thank Claudia Matter, Doctor of Law, and Sieglinde Garscha (Regional Representatives, Glaxo-Wellcome, Vienna, Austria) for providing the study drugs.