Effects of Surgical Levels of Propofol and Sevoflurane Anesthesia on Cerebral Blood Flow in Healthy Subjects Studied with Positron Emission Tomography

The study protocol was approved by the local ethics committee. After giving written informed consent, 16 healthy (American Society of Anesthesiologists [ASA] physical status I), right-handed, nonsmoking male volunteers aged 23 yr (range, 20–30 yr) and weighing 76 kg (range, 66–83 kg) were enrolled. The volunteers were openly assigned to one of two parallel groups to be anesthetized with either sevoflurane (the first eight subjects) or propofol (the latter eight). All were found in good health on detailed prestudy examination, which included laboratory data and a 12-lead electrocardiogram. All subjects confirmed having no history of drug allergies nor were they currently taking any medications. Subjects restrained from using alcohol for 48 h and fasted at least 8 h before the induction of anesthesia.

In each participant, regional cerebral blood flow (rCBF) was assessed during the awake state and at three escalating drug concentrations: 1 MAC/EC⁵⁰, 1.5 MAC/EC⁵⁰, and 2 MAC/EC⁵⁰. The terms minimum alveolar concentration  (MAC) of volatile anesthetic and effective plasma concentration 50  (EC⁵⁰) of intravenous anesthetic refer to the drug concentration that has been shown to prevent the response to surgical stimulus in 50% of subjects. In the sevoflurane group, the actual target end-tidal (ET) drug concentrations used were 0% (awake), 2% (1 MAC), 3%, and 4%, respectively. Target plasma drug concentrations in the propofol group were 0 μg/ml, 6 μg/ml (1 EC⁵⁰), 9 μg/ml, and 12 μg/ml. 8Each of these pseudo–steady state levels lasted for 30 min. At approximately 20 min after the beginning of each level, a flow scan was performed. The analysis and results of these four scans are described in this article. After completing these flow assessments, at each MAC/EC⁵⁰level, a second flow scan was performed during 70 dB auditory stimuli (0.1 ms clicks at 9.9 Hz for 3.5 min), as part of an additional study intended to examine the impact of auditory stimulation on CBF in the primary auditory cortex. These results will be reported separately. The study setup is illustrated in figure 1.

We have previously reported that two of the subjects in the sevoflurane group showed spontaneous, electroencephalograph (EEG)-verified epileptiform discharges at the deepest level of anesthesia before the auditory activation (fig. 1). These occurred distinctly after the CBF assessments reported in this article. 9No epileptiform discharges were present in the raw EEG during the rCBF scans reported here.

The left radial artery was cannulated for arterial sampling as were two large veins in the right forearm for administration of the anesthetic drugs, H²¹⁵O-tracer bolus doses (see  Appendix, PET Imaging) and 2.5% dextrose + 0.45% NaCl infusion (100 ml/h). The subjects received no premedication. Anesthesia was induced after the awake PET scans. In the sevoflurane group, mask induction was used with 8% sevoflurane (Sevorane, Abbot Oy, Espoo, Finland) in 100% oxygen. In the propofol group, anesthesia was induced with an intravenous target-controlled infusion (TCI) of propofol (Diprivan, 20 mg/ml, AstraZeneca Oy, Masala, Finland), with the primary target venous plasma concentration of 6 μg/ml. Infusion was delivered with a Harvard 22-syringe pump (Harvard Apparatus, South Natick, MA) connected to a portable computer running Stanpump software (Steven L. Shafer, Palo Alto, CA), with pharmacokinetic parameters by Marsh. 10After the loss of eyelid reflex, subjects were given rocuronium, 0.6 mg/kg (Esmeron, 10 mg/ml, Organon Teknika Oy, Vantaa, Finland), intravenously, and a laryngeal mask was placed. In the sevoflurane group, the inhaled anesthetic concentration was immediately thereafter reduced to reach and maintain a 2% ET level. Mechanical ventilation was initiated using a Ventilator 710 (Siemens-Elema Ab, Solna, Sweden) with a semiclosed breathing circuit. An oxygen and air mixture with an inspired oxygen level set at 30% was used as a fresh gas at a flow of 2.5 l/min. Breathing frequency was fixed to 15 breaths/min. The ET concentration of CO²was kept strictly at 4.5% (33.75 mmHg) during anesthesia by adjusting the tidal volume. Muscle relaxation was maintained with bolus doses of 10 mg rocuronium. Electrocardiogram (ECG), noninvasive blood pressure, peripheral oxygen saturation, nasopharyngeal temperature, and breathing gases were monitored with Datex AS3 equipment (Datex-Ohmeda Division, Instrumentarium Corp, Helsinki, Finland). A bispectral index (BIS®) monitor (Aspect Medical Systems, Natick, WA) was used with an active electrode placed on each temple and one reference electrode in the midline of the forehead. BIS values were manually recorded at 1-min intervals throughout the study. 11Four-channel EEG was recorded throughout the PET flow studies as previously described. 9The time from the induction of anesthesia to the beginning of the burst suppression activity in the EEG was measured; when present, suppression-to-burst ratios were determined during each MAC/EC⁵⁰level of anesthesia. After the PET scans, residual neuromuscular block was reversed with a neostigmine and glycopyrrolate combination, after which the subjects were extubated. They were then monitored until vital functions had been stable for at least 1 h. Subjects were allowed to leave the research unit after they met the hospital's standard outpatient discharge criteria.

Positron emission tomography is based on 3D imaging of the tissue distribution of physiologically insignificant doses of tracer substances. The biochemical behavior of the tracer in the body is identical to an unlabeled molecule. In this study, the rCBF image was obtained using oxygen-15– labeled water (H²¹⁵O). As a freely diffusing substance, water washes into the brain tissue in relation to blood flow. The PET camera is able to localize the positron-emitting isotope because each positron creates two γ-rays that can be detected in an array of crystals outside the body. For details, see  Appendix, PET Imaging.

All PET studies were performed in a dimly lit room with no sudden loud noises. Subjects wore earplugs, and a plastic head holder was used to minimize head movement. A GE Advance PET scanner (General Electric Medical Systems, Milwaukee, WI) was used to assess tracer tissue concentration in 35 slices parallel to the canthomeatal line, spanning a distance of 150 mm below the cranial vertex. The 3D (septa-retracted) transaxial spatial resolution full width at half maximum (FWHM) at a radius of 10 cm in mid-planes (used in the analysis) was 6 mm in the radial and 5 mm in the tangential direction. The axial resolution of the GE Advance PET scanner was 6.5 mm (FWHM). 12For anatomic reference, individual magnetic resonance imaging (MRI) brain scans were acquired with a 1.5 Tesla scanner (Siemens Magnetom SP63, Erlangen, Germany) on a separate occasion.

At the end of each concentration level, a 5-ml arterial sample was collected for determination of arterial plasma propofol concentrations (in the propofol group only). Plasma was immediately separated and kept frozen at −70°C until analyzed with high-performance liquid chromatography (HPLC) at Yhtyneet Laboratoriot (Helsinki, Finland). The intraassay coefficient of variation was 7%. 13Before completing the anesthesia, arterial blood gas analysis and acid base balance were determined in eight subjects.

The tissue tracer activity images were first computed into quantitative images by using arterial tracer activity data and the appropriate tracer model (in vivo  autoradiography; see  Appendix, Quantitative rCBF). Each subject's MR and PET images were aligned and resliced using either the surface-fit 14or the amir-fit method 15to achieve matching image planes. Individual bilateral regions of interest (ROIs) were drawn to outline the frontal (on nine image planes), parietal (three planes), temporal (five planes), and occipital gray matter (three planes), and the thalamus (three planes), caudate (three planes), putamen (three planes), and cerebellum (three planes). For quantification of global CBF, individual ROIs enclosing the whole brain were drawn on 23 image planes. ROIs were transferred to the parametric PET images to obtain individual flow values for each structure.

Statistical Parametric Mapping (SPM) software version 99 (SPM99, the Wellcome Department of Cognitive Neurology, University College, London, England) 16is an analysis tool used in functional neuroimaging studies. It is routinely used in activation experiments to detect brain areas in which the given task causes an increase or a decrease in local neuronal activity reflected as a statistically significant change in rCBF. It has also previously been used to study the neurofunctional effects of anesthetics and analgesics. 3,5,17–19Whereas quantitative ROI analysis is restricted to anatomic regions manually defined for each person, SPM explores every voxel for a statistically significant change at the group level in regional flow normalized for global changes and enables localization according to a common stereotactic atlas. We used this tool for two separate purposes: (1) to examine how the flow distribution pattern changed in relation to different concentrations of the two studied anesthetic agents, and (2) to locate brain areas with a statistically significant group level correlation between each subject's relative flow and BIS values at all levels of anesthesia. For details, see  Appendix, SPM Analyses.

Quantitative rCBF and physiologic variables were analyzed with repeated measures analysis of variance (RM-ANOVA) having drug concentration (0, 1, 1.5, and 2 MAC/EC⁵⁰) as a within-factor and the drug (sevoflurane vs.  propofol) as a between-factor. When a significant drug or drug-by-concentration effect was detected, the analysis was continued with paired comparisons using linear contrasts in the same model. Linear regression models were used to evaluate relationships between changes in CBF and BIS, mean blood pressure, and plasma propofol concentration within each concentration level. In addition, the Pearson correlation coefficients were calculated. Statistical analyses were conducted with SAS (Version 8.01, SAS Institute Inc., Cary, NC). A two-sided P  value of < 0.05 was considered statistically significant. P  values reported for paired comparisons were Bonferroni correction adjusted to maintain the type I error rate of 0.05. The possible differences between the two drugs in the time course of occurrence of burst suppression in EEG were analyzed with the Cox proportional hazard model. Data are presented as mean ± SD if not otherwise stated.

All flow assessments except for the awake scan of the last subject in the propofol group (because of invalid arterial sampling) were successful. This scan was omitted from the SPM analysis whereas in the ROI analysis the adjacent auditory activation scan was used instead after confirming that there were no differences in any of the ROI areas between the awake-silent and awake-activation scans in either group (data not presented). Compared with the target propofol concentrations (6, 9, and 12 μg/ml), actual plasma levels (mean ± SD) were somewhat higher (7.4 ± 1.7, 12.3 ± 2.6, and 18.3 ± 5.0 μg/ml;fig. 2A). The highest individual concentration was 27.5 μg/ml.

Both drugs induced a concentration-dependent reduction in mean blood pressure (P < 0.001;fig. 2B). The decreases in the sevoflurane group were greater at all concentration levels, revealed by a significant main drug effect in the RM-ANOVA (P = 0.022). Maximal decreases were 40 and 26 mmHg in the sevoflurane and propofol groups, respectively. At the 1 EC⁵⁰level, heart rate increased by 25 beats/min in the propofol group, and statistical analysis showed a significant drug-by-concentration interaction (P = 0.0053;fig. 2C). Paco²results matched well with ETco²values. Nasopharyngeal temperatures remained unchanged throughout the study.

In both groups, BIS was reduced in a concentration-dependent manner down to 22 ± 18 and 1 ± 2 at 2 MAC sevoflurane and 2 EC⁵⁰propofol, respectively (P < 0.001 for both;fig. 2D). At concentrations higher than 1 MAC/EC⁵⁰, the recorded BIS values were lower in the propofol group, and a significant drug-by-concentration interaction was seen in the statistical analysis (P = 0.0012). The burst suppression pattern appeared in the EEG significantly earlier in the propofol group compared with the sevoflurane group (P = 0.011). All subjects in the propofol group showed burst suppression at 1.5 MAC/EC⁵⁰, whereas only 50% of the subjects receiving sevoflurane had burst suppression EEG at this level. Also the mean suppression-to-burst ratio was higher in the propofol group (95%) compared with the sevoflurane group (78%) at 2 MAC/EC⁵⁰(P = 0.021); at this deepest level of anesthesia, two subjects in the sevoflurane group still had continuous EEG activity.

Compared with awake values, all concentrations of both drugs caused significant (P < 0.001 for all) global and regional reduction of absolute blood flow in all studied areas (fig. 3and table 1). Absolute rCBF was decreased from awake values more in the propofol group than in the sevoflurane group at all levels of anesthesia, and significant drug or drug-by-concentration interactions were seen in the overall statistical analyses (P < 0.001 for all). Compared with the awake state, at the 1 MAC level, rCBF values were decreased by 36–53% and 62–70% in the sevoflurane and propofol groups, respectively. At concentration levels above 1 MAC, changes between adjacent levels were relatively small. In the statistical analyses, significant concentration effects were, however, seen in several brain areas.

Compared with 1 MAC of sevoflurane, 1.5 MAC increased the absolute rCBF by 7% in the frontal lobe (P = 0.049), 9% in the thalamus (P = 0.0021), and by 16% in the cerebellum (P < 0.001). Between 1 and 1.5 EC⁵⁰levels in the propofol group, flow was significantly decreased in most of the cortical areas (15–24%;P < 0.001 for all except the occipital lobe) and in the thalamus, caudate, and putamen (13–32%;P < 0.001 for all). In the sevoflurane group, the last increase in concentration had a dual effect. Compared with the 1.5 MAC level, rCBF decreased at 2 MAC by 23% in the frontal lobe (P = 0.011), whereas a 38% flow increase was noted in the cerebellum (P = 0.011). In the propofol group, no statistically significant changes were seen between the 1.5 and 2 EC⁵⁰levels in any of the ROI areas.

Individual percentage decreases in frontal cortical or global CBF between the awake state and the 1 MAC/EC⁵⁰level did not correlate with concurrent BIS values. Individual reductions in global rCBF between the awake state and the three concentration levels did not correlate with concomitant changes in mean blood pressure in either group. Similarly, individual changes in rCBF did not correlate with measured propofol concentrations.

Compared with flow distribution at the awake state, 1 MAC sevoflurane caused a statistically significant relative reduction in the posterior medial cortex, the thalamus, cerebellar tonsils bilaterally, and in the right inferior parietal lobule (fig. 4A). In the propofol group, the awake to 1 EC⁵⁰step caused a widespread relative reduction in parts of all cerebral lobes, the cerebellum, and the midbrain (fig. 4B). At 1.5 MAC in the sevoflurane group, relative rCBF was reduced compared with the 1 MAC level in a small parietal area (fig. 4C). Correspondingly (1 vs.  1.5 EC⁵⁰), in the propofol group, rCBF decreased in the right anterior cingulate and bilaterally in the inferior frontal cortex (fig. 4D). At 2 MAC in the sevoflurane group, relative rCBF was reduced compared with 1.5 MAC in most of the frontal cortex bilaterally, and a concomitant increase was noted in the whole cerebellum, as in the absolute rCBF analysis (fig. 4E). In the propofol group, flow distribution remained unaffected between the 1.5 and 2 EC⁵⁰levels (fig. 4F). Stereotactic coordinates for the areas of significant reductions between the awake state and the 1 MAC/EC⁵⁰level in relative rCBF are presented in tables 2 and 3.

A comparison was also made between the awake state and the 2 MAC/EC⁵⁰levels within both groups. Compared with the awake state, a relative flow reduction was seen in the large frontal, parietal, and temporal areas in both groups at the 2 MAC/EC⁵⁰level (figs. 4G and H). At 2 MAC sevoflurane caused a decrease (compared with awake) also in the posterior cingulate, whereas 2 EC⁵⁰propofol affected the whole cingulate.

The SPM correlation analysis showed areas in which subjects’ rCBF–BIS correlations were statistically significant at the group level. In the sevoflurane group, these areas were located bilaterally in the posterior cingulate, precuneus, cuneus, and parietal lobe, and unilaterally in the right frontal lobe and the left head of caudate (fig. 5). In the propofol group, the areas covered most of the frontal cortex and parts of the cingulate gyrus bilaterally, and the right head of caudate.

The primary finding of this study was that both anesthetics caused a marked global reduction of rCBF at 1 MAC/EC⁵⁰. At deeper levels of anesthesia, only minor further changes were observed. However, at 2 MAC sevoflurane anesthesia caused noticeable redistribution of rCBF.

Our finding of a generalized reduction of absolute rCBF in both groups during anesthesia is consistent with most previous animal and human studies on sevoflurane 20–24and propofol. 6,7,25–30The regional effects of sevoflurane on absolute flow in pigs have previously been studied ex vivo  using microspheres. 21,22Apart from reporting a nonsignificant flow reduction in the cerebellum in the former study, the overall pattern is comparable with ours, with a general decrease at 1 MAC and then an increase in the cerebellum and thalamus and brainstem at 1.5 MAC. In a PET study with monkeys, isoflurane–N²O anesthesia also produced changes resembling our results. Compared with ketamine baseline, the 0.8% ET isoflurane caused a general decrease in regional flow excluding the cerebellum, and at 1.8% isoflurane, the flow reverted to awake values or even above (cerebellum, striatum). 7Previous animal studies, therefore, indicate that with volatile anesthetics, rCBF may be less reduced in the cerebellum than in the cerebrum, but our report includes the first quantification of this phenomenon in humans. The effect of volatile anesthetics on metabolism may be more uniform because in humans, 0.4 ET% isoflurane also decreased regional cerebral metabolic rate (rCMR) in the cerebellum. 2 

There are, however, data suggesting that volatile anesthetics increase also global CBF (sevoflurane, 31isoflurane, 32halothane 33). Part of this controversy could be the result of methodologic shortcomings. It is interesting that, in our study, identical tracer bolus doses created systematically higher arterial and tissue tracer concentrations at 1.5–2 MAC sevoflurane anesthesia than during the awake state. Further, this phenomenon was not seen in the propofol group. Without arterial activity correction, one would have assumed that sevoflurane increases global CBF compared with the awake state. However, it turned out that arterial tracer concentrations had increased even more, and after quantification, a marked global CBF decrease was seen in both groups. One possible explanation for this result could be a drug-induced centralization of the circulation, which causes a higher percentage of cardiac output to be directed to the brain 21and thereby increases the concentration of the tracer substance. Such effects may disturb the validity of quantification if the method depends on identical concentrations of the tracer or contrast media. Arterial sampling used in the present study effectively corrects the tissue data for such errors.

The effects of propofol have been quantified ex vivo  in rats using labeled microspheres. Compared with a N²O–fentanyl control, flow was reduced more profoundly in the cortex than in the subcortex or midbrain, 26which differs from our more homogenous finding. A more even reduction, similar to ours, was seen in a PET study by Enlund et al.  7, when they compared the effects of propofol–N²O anesthesia to ketamine baseline in monkeys. We are not aware of any previous human studies of propofol that would present regional data in absolute terms. However, the effects of propofol on metabolism have been studied using rCMR PET and ROI analysis. At the level of unresponsiveness, propofol caused a global, uniform reduction, although the effect was greater in cortical (58% reduction) than subcortical (48% reduction) structures. 1 

The aforementioned studies report quantitative values that are imperative as they depict tissue perfusion per se  during deep anesthesia. The fundamental question is what can be concluded from the changes that these drugs induce on rCBF. Assuming that flow and neuronal activity are coupled, 34rCBF imaging could, in principle (1) locate direct sites of action for drugs, or (2) at least reveal which cortical functions are lost.

Although local relative metabolic and flow responses have been shown to correlate with the distribution of various neurotransmitter systems in humans 35and animals, 36these findings do not conclusively prove that anesthetics would affect the synapses located exactly in the flow response area. Irrespective of this, regional deactivation patterns could, at least in theory, elucidate drug-specific suppression profiles of cortical functions. Therefore, we wanted to explore the usefulness of SPM software, which is a specifically developed tool for localizing neuronal activation and deactivation areas in the brain. In a way, we deliberately tested “the limits” of this method, knowing that it is initially designed for settings with stable global CBF. Because of the large absolute decrease in global flow, only findings of relative reductions may be meaningful between the awake state and the 1 MAC/EC⁵⁰level in our study; the local flow decrease is significantly greater than the reduction in the whole brain. A relative increase would only mean that flow is better preserved than in the whole brain (which was actually true in large portions of white matter). Therefore, we presented (fig. 4) the relative increases only for the 2 versus  1.5 MAC sevoflurane comparison because these findings were located in gray matter and were also evident in the quantitative analysis.

Despite the above limitations for using SPM in situations with marked global changes, the relative rCBF results were surprisingly unanimous with the quantitative analysis. The 1 MAC/EC⁵⁰versus  awake patterns of either drug also agree with previous studies. 3,6In addition, many of the anatomic regions found to be most intensively deactivated have previously been suspected to be important in producing or relaying the anesthetic state. The SPM analysis of the awake to 1 MAC/EC⁵⁰step revealed that some of the areas where rCBF was especially reduced, like the cuneus, precuneus, and posterior limbus, were common to both drugs. Propofol affected the flow distribution in addition in the parietotemporal and frontal cortices. The areas affected by both drugs at 1 MAC/EC⁵⁰have numerous functions. The cuneus and precuneus are involved in secondary processing and somatosensory associations of visual input. The cingulate gyrus is a part of the limbic system that controls emotions and emotional behavior, sensation of pain, and visceral functions. The cerebellum takes part in the control of muscle tone and equilibrium and some cognitive functions. Sevoflurane reduced relative rCBF in the thalamus, which, as a part of the reticular system, is the main sensory gateway to the cortex. Propofol affected the globally normalized flow distribution in the insula involved in autonomic functions, and in the frontal and temporal cortical areas controlling complex voluntary motor functions and behavior, language, and auditory associative functions. The present results could form a platform for future studies on the molecular mechanisms of anesthetic action. The effects of general anesthetics on various neurotransmitter systems could be studied even in humans using various receptor ligands currently available for PET. 35,37 

As far as we know, there are no previous studies on the dose-dependent effects of sevoflurane on rCBF at the voxel level. Effects of 20% N²O have been studied, showing neuronal activation in the anterior cingulate and deactivation in the posterior cingulate, hippocampus, parahippocampus, and precuneus. 19We are not aware of any previous human studies assessing the effects of sevoflurane on rCMR with which our results could be compared. Alkire et al.  2have studied 0.5% isoflurane in humans with rCMR PET in absolute terms, but the ROI-based analysis did not reveal any drug-specific deactivation pattern. Their more recent study on halothane, on the other hand, was made with voxel level SPM and displayed significantly reduced metabolism in the basal forebrain, thalamus, limbic system, cerebellum, and occipital lobe, 3a finding resembling ours at 1 MAC/EC⁵⁰sevoflurane anesthesia.

The relative rCBF effects of propofol have also been recently assessed with PET. The results of a three-level sedation study by Fiset et al.  6showed that drug concentrations correlated to the regional flow decrease in the thalamus, cuneus, precuneus, posterior cingulate, and the orbitofrontal and right angular gyri, matching our SPM results. According to a quantitative sedation versus  awake study from Alkire, 1with relative analysis of ROI-derived regions, relative decrease in rCMR was most profound in the parietal and occipital lobes and also significant in the frontal lobe, a pattern roughly resembling our flow findings. Compared with the effects of a volatile agent (isoflurane), propofol reduced glucose metabolism more in the cortex and less in the midbrain, thalamus, and basal ganglia, 2which also parallels our SPM results.

Drug-induced decrease in metabolism or blood flow could also relate to the level of consciousness. Concerning metabolism, such relation is supported by recent data, as the percentage of propofol- and isoflurane-induced reduction in whole brain metabolism was found to correlate significantly with BIS. 38We were, however, not able to demonstrate such a correlation between BIS and CBF decrease in either of our groups. This result is probably because the first level of anesthesia in the present study was already deep. Interestingly, a statistically significant correlation through all levels of anesthesia was, however, found with SPM analysis between BIS and relative CBF in the frontal cortex in the propofol group (fig. 5). This finding is logical because the BIS signal originates from this area. In the sevoflurane group, such a correlation was not as evident in the frontal lobe, and the relevance of the correlation seen in the posterior cingulate and parietal lobe remains unclear.

Despite our objective to create highly standardized conditions free from confounding factors like surgical pain, stress responses, and effects of other drugs, a significant error in the applied study design is obvious. To comply with normal clinical practice, our subjects were breathing spontaneously at awake, whereas a slight hyperventilation was maintained during anesthesia. This approach presents a flaw in the study setup because of the direct relationship between Paco²levels and CBF. Therefore, the absolute rCBF decreases presented in this study are undoubtedly somewhat exaggerated. Unfortunately, we did not measure Paco²or ETco²levels before the induction of anesthesia, but assuming a 6-mmHg difference (i.e. , 34 mmHg during anesthesia vs.  40 mmHg at baseline), we can estimate that the error is at most 8.7 ml ·100 g⁻¹· min⁻¹(1.5 ml ·100 g⁻¹· min⁻¹for each 1 mmHg of Paco²). 30,39,40This would account for up to 22–46% and up to 14–23% of the absolute rCBF reductions in the sevoflurane and propofol groups, respectively. These calculations are based on the presumption that cerebrovascular Paco²reactivity is preserved during anesthesia, as has been demonstrated for sevoflurane 20,23,24,31,39,41and propofol. 7,30,40,42,43The effects of hyperventilation on the SMP results are probably negligible because hyper- or hypocapnia does not seem to affect relative CBF distribution in the brain. 44,45 

Another design issue is that both drugs in this study reduced blood pressure in a concentration-dependent manner. We adopted a non–blood pressure-controlled design to avoid the confounding effects of vasoactive drugs. Animal and human data indicate that compensation for changes in cerebral perfusion pressure is preserved during propofol 25–27and 1.5–2% sevoflurane anesthesia. 23,39As assumed, the mean blood pressure remained within the range of cerebral autoregulation in each subject. Our data also indicated that the within-group variations in global CBF and mean blood pressures were not correlated at any concentration level. We presume that the decrease in systemic blood pressure had a minor effect on the quantitative results.

The third design-related issue concerns our aim to compare equipotent doses of the anesthetics. Our target concentrations were based on surgical MAC/EC⁵⁰rather than hypnotic endpoints because we wanted to study surgical levels of anesthesia, not effects of sedation per se . Intravenous anesthetics are less effective analgesic agents than volatile agents. At “equiMAC” concentrations, propofol is likely to induce a greater reduction in CBF than sevoflurane because of its more profound hypnotic effect. Further, the factual propofol concentrations were higher than targeted (although even higher estimates for propofol EC⁵⁰values have been published 46). Despite these flaws, BIS values were surprisingly found to be identical at the first stage of anesthesia in the two groups, indicating a comparable hypnotic depth of anesthesia (fig. 2D). At the 1.5 MAC/EC⁵⁰and 2 MAC/EC⁵⁰levels, however, BIS values and burst suppression analyses indicated deeper anesthesia in the propofol group. It should also be acknowledged that because of our study design (escalating doses instead of randomization), we cannot exclude possible order effects. Randomization was not considered feasible because it would have lengthened the anesthesia considerably in the propofol group. The study conditions were, however, rigorously standardized, and we consider significant order effects extremely unlikely.

Further, anesthetic drugs may also have direct effects on the cerebral vasculature. Especially volatile anesthetic agents are thought to induce cerebral vasodilatation and to attenuate local reactivity of blood flow to metabolism, i.e. , to cause “uncoupling” or “luxury perfusion.” Also, sevoflurane may affect the flow–metabolism coupling, as recent animal studies have demonstrated that metabolism is decreased more than flow at 1 MAC, 24or flow increases, whereas metabolism remains unchanged when deepening anesthesia from 0.7 to 1.3 MAC. 31Although changes in local reactivity could explain the redistribution phenomenon during 1.5 to 2 MAC sevoflurane, this is only speculation because brain metabolism was not assessed in the present study. Although global level autoregulation is maintained with the two studied anesthetic agents, the marked reduction in MAP could cause changes in relative flow during sevoflurane anesthesia as a result of vasodilatation. 47–49On the other hand, sevoflurane does not seem to completely abolish the reactivity, as blood flow velocity in the middle cerebral artery has been shown to increase during EEG bursts in animals anesthetized with sevoflurane. 33Similarly, the flow images obtained during the epileptiform discharges in the two subjects of the present study (and presented earlier as a case report) 9showed that perfusion responds regionally to increased electrical activity even during 4% ET sevoflurane anesthesia. The question of flow–metabolism coupling (or uncoupling?) during various anesthetic regimens should be further studied with PET using parallel regional assessments of blood flow and metabolism in humans. Such studies are currently underway in our laboratory.

Both anesthetics caused a global decrease of rCBF (propofol > sevoflurane). Most of the reduction was already seen at 1 MAC/EC⁵⁰, and a ceiling effect was observed at higher concentrations. However, deep sevoflurane anesthesia caused noticeable flow redistribution. Despite the marked global changes, SPM analysis enabled detailed localization of regions with greatest relative decreases. Relative flow in the frontal cortex correlated significantly with BIS, especially during propofol anesthesia.

The computer program STANPUMP was developed by Steven L. Shafer, M.D. It is available free of charge from him at Anesthesiology Service (112A), PAVAMC, 3801 Miranda Avenue, Palo Alto, California 94394. The authors thank Berner Oy, Helsinki, Finland for providing the BIS® monitor for this study. The authors also thank Riitta Heino, M.D., District Hospital of Turunmaa, Turku, Finland, for her help in planning the neurophysiologic assessments.

Oxygen-15 was produced with a low-energy deuteron accelerator Cyclone 3 (IBA, Ion Beam Applications Inc., Louvain-la-Neuve, Belgium). H²¹⁵O was manufactured by a dialysis technique in a continuously working water module. 50The 300 MBq (8 mCi) tracer bolus of H²¹⁵O was administered as a 15-s manual intravenous injection. Tissue activity was assessed with a dynamic 3D scan consisting of two frames (60 s and 30 s). The scan was initiated automatically at a rapid increase in coincidence detection rate, marking the tracer entrance to the brain. Simultaneously, arterial radioactivity concentration was assessed using a cross-calibrated, two-channel coincidence detection system (GEMS, Uppsala, Sweden) on an arterial line connected to a peristaltic roller pump.

All H²¹⁵O scan data were corrected for detector dead time, tracer decay, and measured tissue photon attenuation, and reconstructed into a 128 × 128 matrix using a 3D transaxial Hann filter with a 4.6-mm cutoff and an axial Ramp filter with a 8.5-mm cutoff. The field-of-view was 30 cm, resulting in pixel size of 300 mm/128 = 2.34 mm. The correction for photon attenuation in tissue was determined by transmission scans using two external circulating ⁶⁸Ge rod sources. A minimum of 10 million counts/slice was collected. Two separate transmission scans were needed for each subject because of changes in positioning at the induction of anesthesia and were performed during the stabilization periods of awake and 1 MAC/EC⁵⁰levels.

The estimate for actual cerebral arterial activity curve was achieved by correcting the measured blood activity for delay and dispersion according to Iida et al.  51rCBF was quantified with in vivo  autoradiography, as previously described. 52The method is based on a one-compartment model, in which the tissue tracer concentration (Ci) depends on three things: arterial concentration (Ca), venous concentration (Cv), and flow (f), according to Equation 1:

The venous concentration cannot be measured, but it can be estimated. As water diffuses freely between tissue and blood, equilibrium is achieved rapidly enough to be assumed immediate and complete. It is also assumed that the ratio of relative water contents between tissue and blood is constant during the measurement, represented by partition coefficient (p), with a value of 0.8 in normal brain. 53With these assumptions, the venous concentration is at all times constantly related to tissue concentration as depicted in Equation 2:

The tissue concentration in Equation 1can then be expressed (Equation 3) using only arterial concentration and flow:

This equation cannot be solved per se . However, using the measured arterial blood activity curve, the equation can be used to create a lookup table of calculated tissue tracer activity values for a range of imaginary flow values. Using this conversion table pixel by pixel, the tissue activity scans can then be transformed into quantitative (parametric) flow images. 52 

The SPM99 preprocessing includes three routine steps. All flow images were first intraindividually realigned to minimize the error caused by subject movement between consecutive scans. Next, all scans were converted into a standard stereotactic space 54by transforming them to match a PET template image. This procedure minimizes the effects of interindividual differences in gross brain anatomy. Finally, all images were smoothed using an isotropic gaussian kernel (12 mm FWHM) to counteract the interindividual variance in detailed functional anatomy. In all analysis models, proportional scaling was applied for balancing inter- and intraindividual differences in global mean flow. It was considered superior to analysis of covariance because of the large quantitative global changes between conditions (see Results, Absolute Regional Cerebral Blood Flow). In all SPM analyses, the level of significance was defined at P < 0.05 (corrected for multiple comparisons), and the minimum cluster size (cluster extent threshold) was set to 30 voxels. For the localization, the coordinates resulting from statistical analysis were converted to Talairach coordinates 54using “mni2tal” conversion software (##Brett 2001), and localized using Talairach Daemon Software. 55 

Relative flow changes between two anesthesia levels were located for both groups separately using contrast (subtraction analysis in multisubject, multiple conditions model (two models, each with eight subjects and four scans per subject). The changes in the relative (i.e. , scaled) rCBF values between two drug conditions were tested for statistical significance in each voxel using t statistics.

Areas with correlation between relative voxel level flow and concomitant BIS values were located for both groups separately using regression analysis in multisubject, covariates-only model. A subject-by-covariate interaction model with least squares fitting of the subject-specific regression lines was applied. The results show areas in which the average multisubject correlation is statistically significant.