INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

INDIRECT EVIDENCE SUPPORTS a central nervous feed-forward mechanism of importance for cardiovascular regulation during exercise ("central command"). One model is to compare heart rate and blood pressure during handgrip before and after regional anesthesia of the arm. After regional anesthesia, when handgrip strength is reduced (sometimes to zero) and the subjects find it more difficult to perform a given task, both heart rate and blood pressure responses to this "attempted handgrip" are enhanced, although the block attenuates or eliminates neural signals from the limb to the brain (16, 20). Similarly, during partial (21, 23, 24) and complete neuromuscular blockade (17), the cardiovascular responses are pronounced. Even though these results argue for the existence of a feed-forward mechanism, it has not been possible to detect areas in the brain responsible for this central drive. The application of single-photon-emission computerized tomography shows that the increase in regional cerebral blood flow (rCBF) in the motor sensory area and premotor and supplementary motor areas (SMAs) during exercise vanishes after regional anesthesia of the arm (12, 13). Equally, the exercise-mediated increase in the transcranial Doppler-determined flow velocity in the middle cerebral artery is eliminated after the arm is blocked by regional anesthesia (20). These studies suggest that the increase in rCBF during exercise reflects neuronal integration of afferent input rather than of central command.

We reevaluated patterns of rCBF during exercise before and after regional anesthesia of the arm by positron emission tomography (PET). Oxygen-15-labeled water (H₂¹⁵O) was used as a flow tracer to indicate neuronal activation (11, 22). We hypothesized that areas involved in cardiovascular control [such as the sensory motor, medial prefrontal, and insular cortices (4, 26)] would be activated during both actual and attempted handgrip. Available statistical software allows for detection of small but consistent changes in the flow distribution, which may not have been appreciated in previous studies (14). Also, PET offers a better spatial resolution and field of view than is available with the applied single-photon-emission computerized tomography (7, 12, 13).

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Four female and four male normal volunteers, aged 26 yr (23-30 yr; median and range) performed rhythmic handgrip. All subjects were right-handed as evaluated by the Edinburgh inventory (25). Informed consent was given, and the study was approved by the Ethical Committee of Copenhagen (journal no. 01-421/93).

Handgrip strength was determined before the scan by a strain gauge transducer connected to a Caspersen and Nielsen measuring bridge (Copenhagen, Denmark). Afterward, the subjects were placed supine with pillows under the lower legs and were carefully instructed and trained in the procedures, especially in performing handgrip while keeping other muscles as relaxed as possible. The head was immobilized with thermally molded foam, the eyes were covered, and ambient noise was kept to a minimum.

On a single day, each subject underwent an imaging session before and after regional anesthesia of the arm. Each of the two sessions consisted of a transmission scan followed by four emission scans carried out in random order. Emission scans were performed with the subject at rest (condition Rest), at rest while listening to a metronome (condition Metronome), and during handgrip squeezing of a force transducer with the right or left hand while paced by the metronome (conditions Right and Left).

After the first session, subjects were placed supine. Bupivacaine (0.5%; Frederiksberg Apotek, Copenhagen, Denmark) was used for regional anesthesia of the left arm: 37 ml for an axillary block and 3 ml for a radial block to secure anesthesia on both the radial and ulnar sides of the arm. Evaluation of the applied regional anesthesia included determination of the maximum voluntary contraction force and performance of simple neurological tests, such as querying the subject about her or his sensation of touch, cold, and heat. The answers were not recorded or quantified but were used to determine when to continue the experiment. Unsatisfactory blockade led to exclusion of an additional, ninth subject. Usually ~1 h after the injections, there was no further progression of the block. Another head holder was made; the subject was placed in the scanner followed by the second imaging session comprising a transmission scan and the above-mentioned four conditions in randomized order.

Scan Conditions

Condition Rest. The subjects were instructed to remain relaxed, and after 10 min the scan was started. Two minutes before the start of the scan, blood pressure and heart rate were determined.

Condition Metronome. A metronome set at 1 Hz was started 4 min before the scan. After 2 min, heart rate and blood pressure were determined.

Conditions Left and Right. A transducer was placed in the respective hand, and the subjects performed handgrip at 1 Hz with maximum strength followed by complete relaxation. After regional anesthesia of the arm, left rhythmic handgrip was performed with the transducer taped to the hand to secure a constant position. Handgrip was started 4 min before the scan. After 2 min, heart rate and blood pressure were determined.

PET Scanning

Each subject was exposed to two 10-min transmission scans and eight intravenous bolus injections of H₂¹⁵O, with each amounting to 200 MBq (18, 19). The tracer was administered via the right brachial vein over 2-4 s followed by 10 ml of isotonic saline for flushing. Data acquisition began concomitantly and lasted for 90 s. During conditions Left and Right, rhythmic handgrip was stopped after 60 s of data acquisition. This strategy improves counting statistics by reducing isotope washout and optimizes the signal-to-noise ratio from activated areas (5, 29). Between repeated emission scans, there was an interval of at least 10 min to allow for isotope decay.

Image Processing and Statistical Analysis

Foci of activated areas were assessed on a voxel-by-voxel level by using the t-statistic with the appropriate contrasts between conditions. They were calculated at an omnibus level of P < 0.001, comparing the expected and observed number of pixels above the threshold (14). Changes are reported in Z scores (number of SDs) after the statistical maps were transformed to the unit Gaussian distribution with the use of a probability integral transform. The level of significance corresponded to Z > 3.09. Not to be mislead, we checked for trends by rerunning the same analysis at a level of P < 0.01 (Z > 2.33).

The location of activation during stimuli was defined as the coordinates of the peak in a confluent area that exceeded the chosen threshold and its spatial extent, defined as the number of voxels above the threshold. The calculated statistical maps had an effective resolution of ~18 × 19 × 19 mm³, giving 201 resolution elements.

Statistical evaluation was improved by taking conditions Rest and Metronome together as the baseline, which then was compared with the activated states (conditions Right and Left) for the respective session (before or after regional anesthesia of the arm). An interaction analysis was performed to compare activation before the block with that after the block. To describe the interactions between block and handgrip, comparisons were made to reveal areas that became differently activated because of the block when right (or left) handgrip was performed, compared with the control comparison before regional anesthesia. Because SPM only provides a one-sided test, increases and decreases need to be calculated separately; i.e., a total of four comparisons had to be carried out. In addition, two simple comparisons (increases during right and left handgrip before regional anesthesia) were performed. In SPM, this corresponds to the contrasts listed in Table 1. For comparison of blood pressure, heart rate, and force, a paired t-test was used at a threshold of P < 0.05. 

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

After regional anesthesia, attempted left handgrip was performed with 25% (range 0-50%) of the strength before anesthesia, whereas the right (control) handgrip was not affected. Sense of touch and cold was blocked in all subjects. During rhythmic handgrip, heart rate and mean arterial pressure increased by 13 ± 3 (SE) beats/min and 12 ± 5 mmHg, respectively, and these increases did not change after the block.

Before regional anesthesia, rhythmic handgrip caused an increase in regional counts with the peak activation in an area in the contralateral cortex near the postcentral sulcus but also covering the cortex of the postcentral gyrus, the central sulcus, and the inferior parietal lobe (Figs. 1 and 2, Table 2). This area could correspond to the primary sensory motor hand area. Additionally, activation was found in the cortex of the contralateral medial frontal gyrus, which may correspond to the SMA (2, 27), and in the ipsilateral cerebellar cortex. For right handgrip, the peak activation in the SMA reached a significance level of only Z = 3.08. 

View larger version (58K): [in this window] [in a new window]   Fig. 1.   Activation during left handgrip before regional anesthesia compared with baseline. Voxels in statistical parametric map that lie above significance threshold P < 0.001 (Z > 3.09) are shown as orthogonal projections in standard stereotactic space (29). Activated areas are gray scale coded with black corresponding to maximum Z score. VPC, vertical plane through posterior commissure; VAC, vertical plane through anterior commissure; R, right. Nos. are coordinates in stereotaxic space.

View larger version (56K): [in this window] [in a new window]   Fig. 2.   Activation during right handgrip before anesthesia compared with baseline. See Fig. 1 legend for details.

There were no positive interactions between regional anesthesia and left handgrip, not even at the lower threshold (Table 2). Negative interactions (i.e., fewer counts during left handgrip due to regional anesthesia of the arm) were found in two areas deep in the cortex close to the left subparietal sulcus and in the cortex close to the posterior section of the superior temporal sulcus (Fig. 3 and Table 2). This interaction seems to be located in white matter, but, because of the 80% cutoff threshold, it is more likely to result from partial volume and smoothing effects from the signal in nearby gray matter. There were no interactions in or close to the sensory motor area or SMA.

View larger version (55K): [in this window] [in a new window]   Fig. 3.   Negative interactions of regional anesthesia and left handgrip, i.e., areas that are less activated during left handgrip as a consequence of application of regional anesthesia. See Fig. 1 legend for details.

Neither positive nor negative interactions between regional anesthesia and right handgrip reached the significance level. Searching for trends at the lower threshold (Z = 2.33), we found a positive interaction located in the cerebellar vermis (132 voxels), two negative interactions in the cerebellar cortex (~40 voxels each), and one in the cortex of the ipsilateral superior temporal gyrus (50 voxels). There were no interactions in or close to the sensory motor area, SMA, or the insula.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This study examined effects of regional anesthesia of the arm on changes in rCBF patterns during rhythmic handgrip to localize areas in the brain responsible for a feed-forward mechanism of importance for cardiovascular regulation. We found no evidence for an effect of regional anesthesia on the regional increase in gray matter counts caused by rhythmic handgrip, despite reduced sensory feedback to the brain, compromised motor performance, and an elevated cardiovascular response. Areas with increased activation during attempted handgrip were not found.

Activation in the sensory motor cortex is attributed to sensory feedback from the moving limbs. Weiller et al. (31) compared active with passive elbow movements and found little difference between activation patterns. Interestingly, they saw almost identical activation in the contralateral sensory motor cortex during both conditions and concluded that the activation during motor tasks is located in the postcentral sulcus. During passive movements, they observed weaker activation in the SMA. The present results show that reduced sensory feedback in the presence of motor commands causes equal activation in both the sensory motor cortex and SMA. The two studies together suggest that a mismatch between afferent and efferent signals (sensory feedback but no motor command as in the case of passive movement, or motor command but reduced sensory feedback as investigated here) can elicit neural activity similar to that caused by an actually performed, voluntary movement. This might reflect transcortical servo-loops balancing efferent with afferent signals or corollary discharges in the sensory motor cortex (8).

Activation was maintained in the sensory motor cortex during attempted handgrip, thereby contradicting results obtained with similar models but based on other methods. The increase in rCBF to both static (12) and rhythmic handgrip (13) is eliminated after regional anesthesia of the arm as is the transcranial Doppler-derived increase in flow velocity (20). These studies were performed with lidocaine, which induces a feeling of general "sleepiness" and markedly different techniques concerning field of view, spatial resolution, and statistical evaluation. This may explain the differences we observed. The present data were evaluated on a voxel-by-voxel basis and not by using large, predefined regions of interest (ROIs) (or even unspecified ROIs as in Doppler studies). A drawback of this design is that the 3D data sets were spatially normalized to the confinements of a "standard" brain before statistical evaluation (30). Whereas this approach improves the statistical sensitivity, it essentially acts as a filter on the images to accommodate the anatomic differences among individuals and thereby compromises the spatial resolution. Improved anatomic localization can be obtained when each rCBF PET image is superimposed on a corresponding anatomic image from a second modality.

Possible Criticisms

Statistical power. Our findings of no evidence for a difference in the activation responses before and after regional anesthesia might have been masked by a large type 2 error, . To estimate , we performed a simple power analysis (1) based on the peak response derived from the main effect of handgrip before anesthesia [Talairach coordinate (x, y, z) = (42, 32, 44)]. During handgrip, the SPM-reported increases in adjusted counts at this location were 8.6 ± 1.9 (means ± SD) before and 6.9 ± 3.2 after regional anesthesia. It should be noted that the values at this coordinate represent a weighted average of a roughly spherical ROI with a diameter the size of the filter (15 mm). Setting our sensitivity threshold for a difference between the two responses ("expected effect") at 4.3, i.e., ~50% of the observed response during handgrip before anesthesia, our SPM analysis has a power (1  ) of 95% with  = 5%. We expect this to be a rough estimate of the power only, because of the data processing performed by the SPM software before the determination of significance (e.g., removal of subject effects, estimation of a pooled variance including all conditions, correction for multiple comparisons, etc.). However, we take our simple power analysis as an indication of the validity of our results.

Other muscle groups. Exercise with a blocked arm is an unusual experience. Consequently, subjects might contract other muscles apart from those normally used for handgrip (i.e., the forearm muscles), and both the motor and sensory signals corresponding to these muscles might cause activation in the sensory motor area. This activation in turn would be misinterpreted as being caused by attempted handgrip. The somatopy in the cortex, however, should have disclosed the involvement of other muscles because the extent of the activation would be moved accordingly; e.g., shoulder movements can be distinguished from hand and finger movements (6). Such a shift of the activated area was not observed. During attempted (left) handgrip, contraction of muscles in the nonblocked (right) arm should give rise to activation in the left sensory motor cortex. This was not observed and indicates that the subjects did not recruit the muscles in the unblocked right arm. In addition, weak muscle contractions do not influence heart rate, and very weak muscular work might actually reduce it slightly (28). Because we have seen an increase in heart rate during attempted handgrip, the dominating response must be from the attempted contractions in the anesthetized arm.

Incomplete regional anesthesia. A concern is the degree of block achieved with regional anesthesia. Unaffected sensory feedback could explain the observed findings. The blocking effect of bupivacaine depends on the thickness of the myelin sheath around the nerves. Somatosensory functions like temperature and pain mediated by unmyelinated and thinly myelinated fibers are blocked before motor function, which again is blocked before proprioceptive sensory function, depending on the most thickly myelinated fibers (3). A semicomplete motor blockade removes the signal from most, if not all, thermoreceptors and nociceptors, whereas proprioceptive feedback conducted by type Ia and Ib nerve fibers from joints and tendons to the brain can at the most be expected to be blocked to the same extent as the motor nerves. This would leave partial sensory feedback activating the sensory cortex. The only available estimate of the effect on proprioceptive fibers was the reduction in force, which was comparable with reported values: 21% (range 0-78%; Ref. 13) and 2% (range 0-96%; Ref. 20), respectively. The contribution of the proprioceptive fibers to the total activation in the sensory motor cortex is not known, but the data indicate that they do not play a dominating role, because the activation in the right sensory motor area does not seem to depend on the different levels of anesthesia achieved in the individual subjects (Fig. 4). Passive movement with intact sensory feedback does not activate the SMA as intensely as active movement does (31). Although we cannot know if all sensory feedback was eliminated, we believe that the unchanged activation in the SMA must, at least partly, originate from motor commands.

View larger version (22K): [in this window] [in a new window]   Fig. 4.   Normalized signal for each subject in right sensory motor cortex at Talairach coordinate (x, y, z) = (42, 32, 44). See Table 2 legend for details. Baseline condition and left handgrip before and after regional anesthesia are shown. Each symbol represents 1 subject. Responses are identical and, therefore, do not give rise to concerns about the different levels of anesthesia.