Imaging Human Cerebral Pain Modulation by Dose-dependent Opioid Analgesia: A Positron Emission Tomography Activation Study Using Remifentanil

Seven right-handed male volunteers participated in this experiment. All subjects gave written informed consent acknowledging (1) that they would receive radioactive tracers, (2) that they would experience experimental pain stimuli and receive a potent analgesic in several dosages, (3) that all methods and procedures were clearly explained, and (4) that they were free to withdraw from the experiment at any time. Subjects were studied after all procedures were approved by the local institutional review board and the radiation protection authorities. The pain-free subjects ranged in age from 28 to 38 yr (mean ± SD, 32.7 ± 4.1 yr) and denied any previous or actual neurologic, psychological, and medical problems; history of any other severe disease (American Society of Anesthesiologists physical status I); or history of drug abuse.

The volunteers had fasted for at least 6 h before the study. Electrocardiograms and arterial oxygen saturation were measured and continuously recorded (Capnomac Ultima; Datex, Helsinki, Finland). Noninvasive blood pressure measurements were performed at 5-min intervals (Dinamap 1846 SX; Criticon, Tampa, FL). End-tidal carbon dioxide concentrations were measured using a Capnomac Ultima monitor via  a catheter placed at the nasopharyngeal border. Capillary carbon dioxide was measured immediately after every condition of drug administration by blood samples taken from a warm, nonheated fingertip.

During experimental pain stimulation, a total of three different drug infusion regimens were investigated in respect to regional cerebral blood flow (rCBF): saline (“control”), 0.05 μg · kg⁻¹· min⁻¹remifentanil (“low-dose remifentanil”), and 0.15 μg · kg⁻¹· min⁻¹remifentanil (“moderate-dose remifentanil”). According to its short half-life, remifentanil was delivered by an infusion pump (Combimat 2000; Döring, München, Germany) in a blinded, randomized order with a time interval of more than 30 min between the two remifentanil infusion rates.

To establish steady state plasma concentrations, remifentanil was administered via  a separate intravenous line in a left antecubital vein to avoid bolus effects during ¹⁵O-water injections. All PET scanning sessions were scheduled at similar times of the day in a quiet ambient environment. Subjects were instructed to remain in a supine position with their eyes closed, to concentrate on the pain stimuli, and not to move or say anything until termination of each PET scan.

After each stimulation and completion of each associated PET scan in the control condition and the two remifentanil dose conditions, subjects were asked to rate their individually experienced pain intensity on a visual analog scale (0–100; 0 = no pain, 100 = unbearable pain).

A semirandomized study protocol was used to overcome the problem of different dates of data acquisitions and possible residual remifentanil effects. Each of the seven subjects underwent two separate PET scan sessions, with at least 3 months between the scanning sessions. One group of subjects (n = 3) was first subjected to the “painful/nonpainful heat + control” condition (3 PET scans), whereas “painful/nonpainful heat + low-dose remifentanil” and “painful/nonpainful heat + moderate-dose remifentanil” (3 PET scans each) were performed on a second session. In the second group, subjects (n = 4) were first exposed to the painful/nonpainful heat + low-dose remifentanil and painful/nonpainful heat + moderate-dose remifentanil condition and at a second session to the painful/nonpainful heat + control condition (fig. 1). Therefore, each subject underwent a total of 18 PET scans, 9 during painful heat stimulation and 9 during nonpainful heat stimulation. With this protocol design, we were able to study the effects of remifentanil without the potential bias of interscan variability.

A temperature-controlled contact thermode (surface area 1.6 × 3.6 cm; contact pressure 0.4 N/cm²; PATH-tester MPI 100; PHYWE, Göttingen, Germany) was used for the two stimulus conditions (nonpainful heat, painful heat) in the three drug conditions (control, low-dose remifentanil, moderate-dose remifentanil). The thermode was attached to the right volar forearm, and the position was changed in a clockwise direction after each scan to avoid habituation effects.

Determination of the thermal pain threshold was accomplished by an adjustment procedure, in which the subjects used a heating and a cooling button to adjust the temperature to what they perceived as just being barely painful starting from a baseline temperature of 37°C. Seven consecutive trials were performed, and the average temperature of the last six trials was considered as the pain threshold. This procedure for the detection of the individual pain threshold was performed twice (24 h and 1 h before the PET session), and the average value was used for the PET experiment.

Series of heat pulses were applied with a frequency of 0.6 Hz for the H²¹⁵O-PET activation studies. From the individual pain threshold (mean ± SD, 45.11°± 0.73°C; range, 43.98°–46.2°C), the pulses changed between a maximum of 1°C above the pain threshold to a minimum of 0.3°C below the pain threshold for the painful heat stimulation (amplitude 1.3°C). For the nonpainful heat stimulation, the temperature undulated between a maximum of 1°C below and a minimum of 2.3°C below the individual pain threshold (amplitude 1.3°C). Each thermal stimulation was continued for 5 min; the PET scans were taken during the last 50 s of the painful or nonpainful heat stimulation. This kind of thermal stimulation was chosen to avoid skin damages.11 

Positron emission tomography was performed using a Siemens 951 R/31 PET scanner (CTI, Knoxville, TN) in three-dimensional mode with a total axial field of view of 10.5 cm and no interplane dead space. The patient’s heads were positioned parallel to the canthomeatal line with the primary sensorimotor cortex covered within the field of view. Attenuation was corrected using a transmission scan (two-dimensional) with an external 68Ge/68Ga ring source before the tracer injection. For each PET scan, a semibolus of 7 mCi ¹⁵O-water was administered intravenously via  a second intravenous line in a left antecubital vein over 35 s using an infusion pump (SP22; Harvard Apparatus, South Natick, MA). The PET scan was initiated when the tracer bolus entered the brain, as indicated by an abrupt increase in the coincidence-counting rate of the tomograph. After correction for randoms, dead time, and scatter, images were three dimensionally reconstructed by filtered back-projection with a Hanning filter (cutoff frequency 0.4 cycles per projection element), resulting in 31 slices with a 128 × 128 pixel matrix (pixel size 2.0 mm) and interplane separation of 3.375 mm.

For observer-independent determination of changes in rCBF, images were preprocessed and statistically analyzed using the statistical parametric mapping approach (SPM99; Wellcome Department of Imaging Neuroscience, Institute of Neurology, University College London, London, United Kingdom). The emission scans were intraindividually realigned before transformation into a reference space according to the Montreal Neurological Institute template of SPM99 by normalization. This template has been determined from 305 magnetic resonance imaging scans of healthy subjects at the Montreal Neurological Institute (Montreal, Quebec, Canada).12As a final preprocessing step, the images were smoothed using an isotropic gaussian kernel (12-mm full-width at half-maximum).

Categorical comparisons were performed across conditions for each drug concentration between painful versus  nonpainful heat conditions (painful heat > nonpainful heat for control, low-dose remifentanil, and moderate-dose remifentanil). All statistical parametric maps of the categorical comparisons were thresholded at P < 0.05, corrected for multiple comparisons with the false discovery rate approach.

Furthermore, analysis of antinociceptive effects of remifentanil (associated with increases in rCBF) on cerebral pain processing was conducted (P < 0.001, uncorrected). Thereby, an exclusive mask was applied to the subtraction analysis of rCBF increases due to remifentanil dose increase during painful heat [(0.15 μg · kg⁻¹· min⁻¹remifentanil and pain) minus (0.05 μg · kg⁻¹· min⁻¹remifentanil and pain)]. As an exclusive mask, the remifentanil-induced activation during nonpainful heat [(0.15 μg · kg⁻¹· min⁻¹remifentanil and no pain) minus (0.05 μg · kg⁻¹· min⁻¹remifentanil and no pain)] was used (P < 0.01 as masking threshold).

To identify pain processing regions that show decreasing activation during remifentanil analgesia in a dose-dependent manner, negative covariation analysis (remifentanil dosage as covariate of interest) was performed (P < 0.001, uncorrected).

The minimal cluster extension (number of activated voxels) was set at 15 or more contiguous voxels passing the significance threshold for all analyses. Small volume correction was applied on the PAG according to the hypothesis of opioidergic pain modulation by PAG activation and its relatively small spatial extension.

Cardiorespiratory parameters showed no significant differences during all conditions and are presented in table 1. Especially the end-tidal and capillary carbon dioxide values did not show a statistically detectable difference, nor did an oxygen desaturation occur.

All volunteers rated the nonpainful heat stimulation during the control condition on the 0–100 visual analog scale as 0, whereas the painful heat stimulus was rated as 68 ± SEM 5.

Remifentanil significantly reduced the subjective perception of pain: Volunteers rated the painful heat stimulus on the visual analog scale during the low-dose remifentanil condition as 42 ± SEM 7 and during the moderate-dose remifentanil condition as 29 ± SEM 6 (fig. 2). All visual analog scale rating changes were statistically significant (paired t  test) across the three conditions (control vs . low-dose remifentanil, P < 0.0003; control vs.  moderate-dose remifentanil, P < 0.00002; and low- vs . moderate-dose remifentanil, P < 0.007).

Painful heat stimulation during saline infusion induced brain activation in areas that have been previously described to be activated during experimental painful heat (thalamus, insula, ACC, S2, frontal cortex; fig. 3Aand table 2), whereas remifentanil administration at both dosages suppressed all detectable activations at the chosen threshold (P < 0.05, false discovery rate corrected; fig. 3B).

With increasing remifentanil dosage [(0.15 μg · kg⁻¹· min⁻¹remifentanil and pain) minus (0.05 μg · kg⁻¹· min⁻¹remifentanil and pain)], increases in rCBF during painful heat stimulation were detected in the PAG (with small volume correction) and cingulofrontal cortex. Detailed results of this voxel-wise statistical analysis are depicted in figure 4and table 3.

The thalamus, prefrontal cortex, S2 cortex, insula, temporal cortex, basal ganglia, and parahippocampal and occipital cortex showed decreases in rCBF during painful heat stimulation while remifentanil dosage was increased (covariation analysis, fig. 5and table 4).

Remifentanil had subjective analgesic effects and changed the pain-related rCBF pattern in human volunteers. In fact, we observed that brain regions were decreasingly activated by pain stimulation during remifentanil analgesia in a dose-dependent manner. Furthermore, the ACC and PAG seemed to be increasingly activated by remifentanil analgesia during a painful versus  a nonpainful stimulus.

Neuronal activity is reflected by rCBF changes and can be investigated in vivo  in the awake human brain by PET. Experimental noxious stimuli alter rCBF in a number of cortical and subcortical regions.13–17Thereby the ACC; the prefrontal, insular, inferior parietal, and somatosensory cortices; and the thalamus were most consistently activated in previous studies.18,19 

Analgesia is a dose-dependent phenomenon, but the neuronal correlate of this clinical observation has not been investigated so far. Therefore, we used experimental painful heat and PET to analyze the multifocal activity of supraspinal pain processing brain regions including the descending inhibitory system in response to increasing remifentanil analgesia.

Regarding data interpretation, it is appreciable that some brain regions of the complex pain network might contribute to the individual pain perception, whereas others are rather involved in pain modulation. It might thereby be expected that remifentanil suppresses activity in brain areas promoting the different aspects contributing to the complex sensation of pain. Indeed, we found decreases in activation in a number of brain areas such as the thalamus and somatosensory cortex (S2), which have been previously reported to process the pain experience.20 

Contrarily, brain areas that would rather contribute to pain modulation would not necessarily be expected to decrease their activity, but to differentially change their activity to jointly suppress pain-related activity in the various regions. In fact, such an activity pattern was observed in the PAG and cingulofrontal cortex.

The distinct role of the most important regions evidencing altered activity in the context of various degrees of remifentanil analgesia will be discussed in greater detail in the following sections.

Our finding of an activation of the PAG by remifentanil demonstrates the importance of descending inhibition of nociceptive transmission as part of the “brain defense system.”21,22Its potential to significantly act in an antinociceptive way has been demonstrated in experimental as well as clinical settings.23Implantation of electrodes and electrical stimulation of the PAG induces inhibition of nociceptive dorsal horn neurons and profound analgesia in humans and animals.24–26This analgesic effect is thought to derive from a release of endogenous opioids, because the effects are reversible by the administration of the opioid antagonist naloxone.24,27 

The PAG controls nociceptive transmission indirectly by means of connections through neurons in the rostral ventromedial medulla and the dorsolateral pontine tegmentum. These two regions project to the spinal cord dorsolateral funiculus and control pain by selectively influencing primary afferent nociceptor terminals and somata of dorsal horn neurons responding to noxious stimulation. Supraspinal input to the PAG originates from the hypothalamus and from the limbic forebrain (ACC), including several regions of the frontal neocortex and the central nucleus of the amygdalae, while projecting to spinothalamic pathways. Together with the perigenual ACC and the orbitofrontal cortex, the PAG seems to play a key role in pain modulation during distraction as an experimental pain stimulus and a simultaneous distraction task reduces pain while an increase in activation in the PAG was observed.1,28The functional interaction between cingulofrontal areas and the midbrain/PAG and posterior thalamus has been underlined using functional magnetic resonance imaging and functional connectivity analysis.28Moreover, Petrovic et al .29detected functional interactions on covariation analysis between the rostral ACC and the brainstem/PAG during both opioid and placebo analgesia.

The endogenous opioid that mediates these antinociceptive effects has not been identified. Results of experiments using microinjections of μ-opioid receptor agonists whose analgesic effects were reversed by μ-opioid receptor antagonists implicate that the μ-opioid receptor and enkephalins, as endogenous agonists, play a fundamental role in this native endogenous pain control system.21Furthermore, the importance of the μ-opioid receptor is underlined by its presence in the nuclei of pain-modulating circuits.

Henderson et al .30have revealed the existence of strong ventrolateral PAG projections to cardiovascular depressor regions within the caudal medulla which likely contribute to ventrolateral PAG-mediated hypotension and bradycardia. Hypotension and bradycardia are well-known phenomena accompanying remifentanil administration. Especially after rapidly changing infusion rates and bolus injections of remifentanil, these cardiovascular effects are frequent. Because we omitted these factors in conjunction with only moderate infusion rates to counteract respiratory depression, we did not see these cardiovascular changes in our experiment, and PAG activity can therefore be explained by its antinociceptive action.

Positron emission tomography studies using the opioidergic ligands diprenorphine and carfentanil were able to demonstrate a high opiate receptor density in the cingulofrontal region.8,31With the anatomical linkage of the ACC to the PAG, the former is anatomically closely connected to the opioid-mediated pain-modulatory circuit.32 

Although opioid analgesia specifically attenuates cerebral responses to painful stimulation, the cingulofrontal cortex was increasingly activated in our study during painful stimulation and increased remifentanil administration. As noted above, participation of this area in pain modulation was recently supposed due to the results of functional magnetic resonance imaging studies investigating effects of distraction and placebo on pain perception and processing.28,33Taken together with the activation patterns of other neuroimaging studies and opioidergic ligand PET results, the antinociceptive effect of exogenous opioids as well as placebo analgesia is likely to be mediated via  opioidergic neurotransmission in the ACC.2,9,29,34,35 

The properties of the cingulofrontal region are however not limited to pain processing, but an involvement in the processing and modulation of emotional contents, such as fear and anxiety, is well recognized.36In view of these complex functions, attentional and also emotional control of pain processing is suggested as a major role of the cingulofrontal cortex.37,38 

The interpretation of our results bases on general assumptions that underlie changes in rCBF in a H²¹⁵O-PET activation study. These underlying mechanisms are complex and related to factors acting in parallel as well as in series.39Among these factors, the following three possibilities seem to be fundamental for changes in rCBF: (1) rCBF might be related to increases in lactate concentration (released by astrocytes), (2) it might be triggered by products of neuronal spiking, and (3) the blood vessels themselves might be involved in the rCBF changes.40,41These considerations emphasize the difficulties in interpreting changes in PET signal although considerable advances have been made during recent years. Finally, a spatial mismatch between the actual μ-opioid receptor effect and the source of the PET signal cannot be excluded.2 

Starting PET scanning for each condition after an interval of greater than 30 min of continuous infusion of remifentanil guaranteed steady state blood concentration of remifentanil.42,43This zero-order infusion and the avoidance of an additional bolus prolonged the experimental duration but provided a safe setting and omitted possible adverse and confounding reactions of the volunteers, such as nausea and vomiting. Furthermore, no cardiovascular side effects or respiratory depression were noticed. This may be caused by the relatively small increments of remifentanil over a relatively long period, thus allowing physical adaptation. Because residual nonanalgesic but psychomimetic effects persist as long as 60 min after remifentanil infusion, we used a semirandomized study protocol with a step-up infusion rate to overcome the problem of extended study time.44 

The pharmacodynamic effects of the μ-opioidergic drug within the chosen concentrations cannot be blinded in practice. Therefore, the identification of the different experimental conditions was relatively simple for the volunteers. Because we believe that the condition with infusion of saline was for this reason not a true placebo condition, we chose to term the saline condition as “control” instead of “placebo.” However, we acknowledge that a partial placebo effect might have contributed to the results of the control condition.

The highest dose of remifentanil used in our study was 0.15 μg · kg⁻¹· min⁻¹, providing adequate analgesia in our as well as other experimental and clinical pain conditions.45–47Higher remifentanil infusion rates would be of further theoretical interest. This is inherent with the need of additional scans; either reduction of tracer activity or a reduction of scans per condition due to radiation protection would result in less robust statistics and would therefore not be beneficial. Furthermore, in a study paradigm with spontaneous breathing volunteers, the maximal dose of remifentanil is limited because the occurrence of unacceptable side effects (e.g ., respiratory depression, nausea and vomiting) would interfere with the interpretation of our findings.45,48 

We clearly showed that remifentanil induced dose-dependent decreases in multiple supraspinal brain areas of the pain neuromatrix (fig. 5). However, one could wonder whether these dose-dependent effects might be in contradiction to the categorical analyses, where both remifentanil dosages abolished all detectable pain-induced brain activations. In our view, this discrepancy is a consequence of the statistical thresholding. We chose thresholding with correction for multiple comparisons (false discovery rate) for the categorical comparisons to reduce the occurrence of false-positive results. If we would have chosen a less strict threshold resulting in limited reliability (e.g ., 0.05 uncorrected for multiple comparisons), the categorical comparisons would evidence the dose-dependent nature of the remifentanil-induced decrease in brain activation, which was also clearly reflected by the clinical pain ratings (fig. 2).

In conclusion, our data reveal the neuroanatomical targets of the μ-opioidergic receptor agonist remifentanil during an experimental pain condition by PET. The specific modulation of pain processing structures by increasing dosages of remifentanil provides further insight into the cerebral mechanisms of exogenous opioid analgesia. On the basis of the presented data, especially the role of the brainstem and the cingulofrontal cortex in remifentanil analgesia is underlined.

The authors thank the staff at the Radioisotope Delivery System 112 cyclotron unit for reliable supply of ¹⁵O-water, the technical positron emission tomography staff for excellent technical assistance, and Doris Droese (Medical Laboratory Technician, Klinik für Anaesthesiologie, Technische Universität München, Klinikum rechts der Isar, München, Germany) for support and helpful advice. The authors thank the volunteers whose participation made this study possible.