Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Midazolam decreases cerebral blood flow in the left prefrontal cortex in a dose-dependent fashion

Midazolam decreases cerebral blood flow in the left prefrontal cortex in a dose-dependent fashion Midazolam, a short-lived benzodiazepine producing sedation and reversible anterograde amnesia, was administered intravenously to 14 healthy male volunteers. Regional cerebral blood ¯ow (rCBF) was measured using positron emission tomography (PET) with intravenous H "&O at either a ` high ' midazolam EEG effect (EEG signs of stage 2 sleep), or ` low ' midazolam EEG effect (increase in EEG beta power only). Memory tests administered following PET scans showed signi®cant drug-induced impairment in learning and retrieval at the same drug concentration at which PET images were acquired. Statistical parametric mapping was used to identify regions where rCBF changes after drug administration were signi®cantly different in the high- vs. low- effect groups. Dose¬condition interactions were found in the left dorsolateral prefrontal cortex [Brodmann's areas (BA) 9 and 46], bilateral orbital-frontal cortex (BA 47), the left middle temporal gyrus (BA 22) and the right hippocampus. The predominantly left frontal rCBF decreases occur in a region associated with semantic processing, working memory, and encoding of verbal material, a process preferentially affected by midazolam. Our interpretation is that rCBF changes in the hippocampus are unlikely to mediate the anterograde amnesia produced by midazolam. Although in the present study PET images were acquired during the resting state rather than during memory processing, these results underscore the need for further investigation relating to the interaction of midazolam with speci®c cognitive operations in these brain regions. Received 30 June 1999 ; Reviewed 27 September 1999 ; Revised 20 December 1999 ; Accepted 10 January 2000 Key words : Benzodiazepines : midazolam, regional cerebral blood ¯ow, positron emission tomography, memory, prefrontal cortex, medial temporal lobe. Introduction effects of midazolam are quite speci®c, with encoding of new information affected more than retrieval of old Minor tranquillizers with short half-lives are used in information (Ghoneim and Mewaldt, 1990 ; Lister, 1985). anaesthesia to sedate patients for short surgical pro- The EEG effects of midazolam are similar to those of other cedures. Recovery is rapid and there is less risk of adverse benzodiazepines (Greenblatt et al., 1989), which act on the side-effects than with general anaesthetics. Among the GABA-A receptor throughout the brain (Tanelian et al., benzodiazepine class, midazolam (half-life of 2±4 h), is 1993). But neither electrophysiology nor receptor-binding widely administered for ambulatory surgery using the studies can explain how midazolam produces its effects of technique of conscious sedation. In contrast to general sedation and amnesia. If speci®c regions of the brain anaesthesia, conscious sedation allows the patient to affected by midazolam could be identi®ed, then a better remain awake and cooperative during the procedure. In understanding of how midazolam produces reversible this context, the ability to ablate memory for surgical amnesia may be possible. events is seen as a bene®cial side-effect. A number of studies have shown changes in regional Midazolam has marked effects on episodic memory, cerebral blood ¯ow (rCBF) following the administration of producing profound anterograde amnesia as well as various drugs active in the central nervous system (CNS) signi®cant sedation (Veselis et al., 1997c). The memory (Firestone et al., 1996 ; Friston et al., 1991, 1992 ; Grasby et al. 1992, 1993, 1995 ; Gyulai et al., 1996 ; Jones et al., Address for correspondence : Dr R. A. Reinsel, Department of 1991 ; Parekh et al., 1995 ; Roy-Byrne et al., 1993), Anesthesiology and Critical Care Medicine, Memorial Sloan-Kettering including midazolam (Veselis et al., 1997b). It is interesting Cancer Center, 1275 York Avenue, Box 24, New York, NY 10021, that the neuroanatomical regions demonstrating changes USA. in rCBF in the presence of these drugs frequently subserve Tel. : (212) 639-8111 Fax : (212) 772-8646 functions that are in¯uenced by those drugs. Thus, it is E-mail : reinselr!mskcc.org 118 R. A. Reinsel et al. possible that drug-induced changes in rCBF can identify EEG power spectrum was used to verify the degree of the neuroanatomical structures necessary for the ex- drug effect at the time of drug infusion. Low midazolam pression of the drug's clinical effects. Unfortunately, many effect was de®ned by high frequency, low amplitude EEG of these studies demonstrate changes in rCBF in numerous activity in the beta frequency range (13±20 Hz) and locations in the presence of drug, which we also found represents a serum concentration at which a 50 % with midazolam (Veselis et al., 1997b). Identi®cation of decrement in memory effect is present (Veselis et al., dose-related changes in rCBF in this situation may help 1997c). The high effect was de®ned by the presence of de®ne the neuroanatomical regions worthy of closer ` sleep ' spindles at approx. 14 Hz, and was chosen to attention in further studies of the neuroanatomical basis of represent an EEG effect clearly separable from the low drug effects. midazolam effect (Feshchenko et al., 1997). Following the This report details the investigation of dose-related start of drug infusion a period of 20±30 min elapsed changes in rCBF patterns during midazolam adminis- before subsequent positron emission tomography (PET) tration using an interaction analysis between dose (high scans were obtained to allow time for complete equi- vs. low EEG effect) and the presence}absence of drug. The libration of drug effect. Midazolam infusion occurred for interaction analysis is designed to identify brain regions approx. 60 min. Midazolam concentrations were deter- where the effect of a high dose is signi®cantly different mined by high performance liquid chromatography from a low dose of midazolam. Our hypothesis is that the (Veselis et al., 1997a). neuroanatomical regions thus identi®ed are more likely to represent the neuroanatomical substrate which midazolam Neuropsychologic state during PET scanning acts upon, possibly mediating its clinical effects of amnesia In both control and midazolam conditions subjects and sedation. listened to binaural auditory tones at 1000 Hz, 80 dB, with an interstimulus interval of 1.1 s, presented 30 s Method before isotope injection and continuing throughout the scanning procedure. This was done because the attention This investigation was approved by the Hospital task (scans 2, 4, 6) required a control condition with Institutional Review Board and Radiation Safety Com- subjects hearing tones. This control condition was used mittee before accrual of subjects took place. for comparisons for the attention task, as well as the administration of midazolam. Subjects were instructed to Subjects pay no attention to the tones. The room was dimly lit Fourteen normal male volunteers (ages 21±44 yr ; with minimal background noise. mean³s.d. 28.1³5.8 yr) participated in this study after giving informed consent. All but one was strongly right- Memory testing handed (the one exception being ambidextrous), and no Between scans 2 and 3 (` baseline ') and 7 and 8 (` drug ') volunteer had a history or physical evidence of neurologic subjects were asked to memorize a list of 16 words or psychiatric illness. Blood sampling was accomplished presented through headphones by tape recorder. It should from a radial arterial line. be noted that there was no memory task administered during PET scanning. However, as midazolam concen- Study design trations were held constant during the infusion, the Subjects were randomized to receive either a high or low memory data serve as a measure of the amnesic effect dose of midazolam. All subjects were scanned in three present during PET scanning at these doses of midazolam. conditions on a single day : control (scans 1, 3, 5), an Two equivalent word lists were employed, in counter- attention task (scans 2, 4, 6), and during midazolam balanced order, for baseline and drug conditions. Memory infusion (scans 7, 8, 9). This report concerns data collected was assessed by immediate free recall after each pres- from scans 1, 3, 5 and 7, 8, 9. The results of the attention entation of the list, and again at the end of the study day task have been reported separately (Reinsel et al., 1995). by delayed recognition. Measures analysed for this report After scan 6, midazolam was infused to one of two include trial 4 (the last learning trial), immediate recall concentrations using a computer-assisted continuous after ` interference ', delayed recall some 30±40 min later, infusion device (CACI*), during simultaneous EEG moni- as well as delayed recognition. We predicted that the low- toring (see Veselis et al., 1995, for details of methods). The effect group would have higher recall and recognition scores than the high-effect group. Our memory test is * CACI : Glass PSA and Jacobs J, Duke University Medical Center, modelled on the Rey Auditory Verbal Learning Test (Rey Durham, NC. (See Veselis et al., 1997a, for details of the use of this AVLT) but with a few modi®cations required by the PET device.) Midazolam decreases rCBF in prefrontal cortex 119 scan environment. Word lists were presented four times contiguous slices are obtained. Subjects were positioned for learning and recall in the interval between PET scans. so that the centre of the ®eld of view was approx. 3 cm This took approx. 10 min, at which point the next dose of above the orbito-meatal line. This resulted in some of the isotope was ready for injection. Given the rigorous time superior cortex not being in the ®eld of view, as the schedule imposed by a fast-decaying isotope, it was not scanner could only image up to 8 cm above the orbito- possible to delay the next scan in order to present the meatal line. Subjects spent approx. 3 h in the scanner interference word list. The 2-min PET scan itself and the (mean total time : 192 min, range 112±273 min) and were associated bustle of activity around the volunteer took the receiving midazolam for approximately the last hour of place of the interference list that would normally be that time. presented. Subjects were instructed not to rehearse the words during this interval, and immediate recall was Data analysis obtained immediately after the PET scan. During the drug condition, delayed recall was obtained at the same serum Behavioural data were analysed by non-parametric concentration that was present during learning. Mann±Whitney U test for differences between groups, and by paired t test for within-subject change between conditions. Test results are two-tailed p values unless Study groups otherwise noted. PET image analysis was performed using As the EEG response during PET scanning could possibly statistical parametric mapping (SPM95) developed by the be different than the EEG effect seen just before PET Wellcome Department of Cognitive Neurology, Uni- scanning commenced, subjects were classi®ed into low- versity College London, and implemented in pro matlab and high- effect groups based solely on the EEGs obtained v. 4.2 (Mathworks, New York). Images were realigned to during PET scanning. After accrual of all subjects, the EEG the ®rst scan and transformed into the standard stereo- spectra obtained during scanning were reviewed for tactic space utilized in the human brain atlas of Talairach determination of study groups for data analysis. Three ex- and Tournoux (1988). As a ®nal pre-processing step the perienced encephalographers (R.V., R.R., V.F.) examined images were smoothed using an isotropic Gaussian kernel EEG spectra from four derivations across the scalp [12 mm resulting in a FWHM (x, y, z) smoothing of 18, (Fz, Cz, Pz and O2). Spectra were classi®ed without 25, 21 mm] to accommodate normal variability in func- knowledge of which subject or study group they were tional and gyral anatomy for group analysis. The three associated with. Subjects with a clearly visible peak in the scans obtained for a particular subject in baseline and drug power spectrum centred at 14 Hz during the drug conditions were included as replicated measurements for condition were included in the high midazolam effect that subject in the group analysis. After specifying the group (Feshchenko et al., 1997). Subjects without a clearly appropriate design matrix (multi-study with 2 conditions visible spectral peak were classi®ed to the low midazolam and 3 replications), the condition and subject effects were effect group. After this classi®cation, the high- and low- estimated according to the general linear model at each effect groups consisted of 7 subjects each. This classi- and every voxel using a subject-speci®c ANCOVA model ®cation was congruent with the initial randomization to (Friston, 1994). Mean global CBF was normalized by low- or high-effect groups, except for 3 subjects. Two SPM95 to 50 ml}100 g\min. This analysis generated a were randomized to the high-effect group, but the spectral mean voxel value with associated error variance for each peak was not clearly visible. One subject was randomized condition. Voxel¬voxel comparison between conditions to the low-effect group, but power spectral analysis was then made, and mean voxel values of control and revealed a prominent spectral peak in the spindle range experimental condition scans were compared for (12±15 Hz). signi®cant differences by t test. A group (high, low effect)¬condition (baseline, drug) interaction analysis was used to identify regions where CBF changes in the PET scanning high-effect group were signi®cantly different than the For each scan, 20 mCi of H "&O in 10 cc were infused low-effect group, expressed as Z scores using a threshold over 20 s into the antecubital vein using a constant rate Z value of " 3.09 (corresponding to a p ! 0.001). A infusion pump with 15 min between scans. PET scans further correction for multiple comparisons was applied, were obtained with a PC4600 NeuroPet scanner, with a resulting in a corrected p ! 0.05 at the level of individual 10 cm axial ®eld of view and a 1.2 cm in-plane resolution. voxels. Signi®cant maxima were mapped using the This scanner consists of ®ve detection rings separated by Talairach and Tournoux atlas to determine the anatomical septa which detect coincident counts in projections location and corresponding Brodmann's area, where orthogonal to the axial plane of the scanner, and thus 9 applicable. 120 R. A. Reinsel et al. Quantitative image analysis 79.5³9.7 kg for all subjects combined). Signi®cant differences were present in total dose of midazolam The PET frame data, the PET central ring data and the administered (7.5³1.7 vs. 9.7³1.3 mg, p ¯ 0.02 by t blood activity data were all decay corrected to the time of test) and average serum concentrations during scanning injection as determined by the PC's clock at the start of PC (74³24 vs. 129³48 ng}ml, p ¯ 0.02 by t test). Sedation data acquistion. The ` head curve ' data were roughly in the low-effect group varied from a small degree of calibrated to activity concentration by plotting the mean subjective sleepiness to requiring verbal stimulation to brain activity concentration, as determined from the respond. Subjects in the high-effect group were heavily reconstructed and calibrated PET images for each frame, sedated and frequently required physical stimulation to vs. the mean count rate of the ` head curve ' over the time respond. Assessment of sedation and stimulation, if period corresponding to the PET frames. The slope of the necessary, were done just prior to the memory test. best least-squares ®tting line was taken to be the conversion factor. A single conversion factor was used for all runs on a given subject. Unfortunately, values derived Memory effects for absolute CBF are biased and therefore unsuitable for inter-subject comparisons. This bias is probably related to Both groups suffered memory loss during midazolam inclusion of non-cerebral structures in the volume of brain administration (see Table 1). Learning was effectively analysed, and will vary among individuals (Matthew et abolished in both groups, with mean recall on trial 4 of al., 1993 ; Ramsay et al., 1993). However, the bias should less than 25 % of the words presented. Immediate and be constant across the different runs for a single subject delayed recall dropped to near zero. Comparing change and therefore paired comparisons of the sedated to the scores (baseline minus drug), the high-effect group baseline condition are valid and accurate (Matthew et al., showed the predicted larger decrement in immediate 1993). Because of the individual variability in CBF values, recall after midazolam (p ! 0.05 one-tailed). From the 16- and of the small, insigni®cant changes seen in the low- word list administered during midazolam infusion, the effect group, all subjects were combined for global CBF number of words recognized at the end of the study day analysis. were 5.0³4.5 (p ! 0.002) and 1.6³1.7 (p ! 0.001) words for the low- and high- effect groups, respectively Results (compared to baseline scores by paired t test, see Table 1). The difference between recognition memory scores Study groups between high- and low-effect groups during midazolam No differences in demographics were observed between sedation had a signi®cance of p ¯ 0.082 one-tailed by the low and high EEG effect groups (mean weight³s.d.: Mann±Whitney U test. The high variance of recognition Table 1. Detailed results of memory testing performed between PET scanning Low effect High effect Baseline Drug Baseline Drug No. words correct on 13.3 (2.5) 3.4 (2.8) 13.3 (2.4) 1.7 (2.9) 4th presentation of word list Immediate recall after 11.4 (3.8) 1.9 (1.8) 12.9 (2.8) 0.9 (1.6) ®nal list presentation Delayed recall (30 min) 12.5 (2.6) 0.6 (1.5) 12.7 (2.9) 0.3 (0.8) after ®nal presentation Recognition at end of 14.7 (1.0) 5.0 (4.5)* 14.6 (0.8) 1.6 (1.7)* study day For both groups combined, the difference between results from the drug vs. baseline conditions was highly signi®cant (p ! 0.001 by paired t test) for all parameters. The only result approaching statistical signi®cance, when the low vs. high midazolam effect groups were compared, was the comparison of word recognition after midazolam (* p ¯ 0.08 one-tailed). All results are presented as means with the standard deviation indicated in parentheses. Midazolam decreases rCBF in prefrontal cortex 121 rCBF decreases rCBF increases Sagittal Coronal Sagittal Coronal R R 0 0 0 0 32 32 VPC VAC 064 VPC VAC 064 –104 68 –104 68 Low VPC VAC VPC VAC dose 0 0 64 64 Transverse Transverse Sagittal Coronal Sagittal Coronal R R 0 0 0 0 32 32 VPC VAC 064 VPC VAC 064 –104 68 –104 68 High VPC VAC VPC VAC dose 0 0 R R 64 64 Transverse Transverse Figure 1. Changes in rCBF related to midazolam administration. Healthy volunteers received midazolam infusion to either 74³24 or 129³48 ng}ml concentrations of midazolam (to produce low and high EEG effects), which were held constant by computer control during memory testing and PET scanning. This ®gure summarizes the main effects of midazolam on rCBF for the two doses administered. Regional CBF changes were identi®ed using statistical parametric mapping (SPM) analysis, and thus all changes are relative to a normalized baseline. Regions of signi®cant change in CBF (p ! 0.001) are shown in a ` glass brain ' projection. Decreases in rCBF are shown in the upper left and lower left panels for the low and high EEG effect groups, respectively. Increases in rCBF are shown in the upper right and lower right panels. The ®eld of view of the PET scanner was not able to cover the entire brain. Quantitative analysis (Veselis et al., 1997b) revealed that regions of ` rCBF increases ' are actually decreases in rCBF, but the decreases are statistically signi®cantly smaller than other blood ¯ow changes. Thus these areas can be considered regions where midazolam has less effect on rCBF. scores in the low-effect group most likely accounts for the between dose and condition. The largest areas of change failure to reach signi®cance in this comparison. in rCBF with higher doses of midazolam occurred in the left prefrontal cortex (PFC) (see Figure 2). The most signi®cant decreases in rCBF occurred at Talairach Dose-related decreases in rCBF coordinates (x, y, z, in mm) ®44, 34, 16 and ®42, 26, 28 As previously reported, the effect of midazolam is not located in the left middle frontal gyrus, close to uniform over the brain. In the high EEG effect group, 22 Brodmann's areas (BA) 46 and 9. These areas are involved different areas of brain demonstrated signi®cant decreases in working memory as well as other functions (D'Esposito in rCBF, but there were only 11 areas common to both the et al., 1998). A smaller area of rCBF decrease was seen low and high EEG effect groups (see Figure 1). Many of bilaterally in the inferior frontal gyrus at ®26, 28, 0 and these regions have been identi®ed as important in 24, 28, 0 (BA 47) located in the frontal operculum (inner processes involving memory, attention and mood. portion of the lateral sulcus, just anterior to the insula). An The reader is referred to Veselis et al. (1995) for speci®c additional area of signi®cantly lower rCBF after results relating to the separate doses of midazolam. We midazolam was found at ®42, ®34, ®4 in the left speci®cally report here on the signi®cant interaction middle temporal gyrus (BA 22). 122 R. A. Reinsel et al. Interaction, rCBF decreases, smoothing 12 mm Sagittal Coronal VPC VAC –104 68 –44, 34, 16 VPC VAC (Z = 4.86, sp = 0.020, ht = 0.001) –42, 26, 28 (Z = 4.85, sp = 0.020, ht = 0.002) –26, 28, 0 (Z = 4.60, sp = 0.020, ht = 0.004) 24, 28, 0 (Z = 4.26, sp = 0.148, ht = 0.017) Transverse –42, –34, 4 (Z = 4.06, sp = 0.161, ht = 0.036) Figure 2. Dose¬condition interaction. Relative rCBF decreases during midazolam administration which were signi®cantly greater (p ! 0.001) in the high than the low EEG effect group. Peak changes in rCBF occurred at ®44, 34, 16 and ®42, 26, 28 in the left middle frontal gyrus (BA 46 and 9, respectively) ; ®26, 28, 0 and 24, 28, 0 in the left and right inferior frontal gyrus (BA 47) ; and ®42, ®34, 4 in the sulcus between the left middle and superior temporal gyrus (BA 22). Z, Z score ; sp, probability based on spatial extent ; ht, probability based on peak height of signal. Dose-related ` increases ' in rCBF (regions of lesser hippocampus (36, ®24, ®12) and parahippocampal decrease in rCBF) gyrus (22, ®48, ®4), and the left lingual gyrus just rostral to the parahippocampal gyrus (®16, ®52, ®4; The results of this analysis represent areas of rCBF which BA 19). Another area was identi®ed in the right posterior are ` increased ' in the high EEG effect group when cerebellum, that is very close to the extreme inferior compared with the low EEG effect, when normalized for extent of the fusiform gyrus (BA 18). global changes. Midazolam in the doses used in this study induced about a 12 % reduction in global CBF (Veselis et al., 1997b). Regions which showed greater ` increases ' in the high EEG effect group varied from ­6 % (left middle Quantitative blood ¯ow values occipital gyrus) to ­9 % (right middle occipital gyrus). As the overall decreases in global CBF were larger than the A major confounding variable in the measurement of localized rCBF effects, regions identi®ed as showing global CBF is the change in p following drug CO signi®cant ` increases ' in rCBF in the SPM analysis are best administration. In the low-effect group it increased interpreted as areas where rCBF did not decrease to the 3.5³1.3 mmHg (p ¯ 0.04 compared with control by same extent as the rest of the brain. One could think of t test), and in the high-effect group it increased these regions as being relatively resistant to the global 5.1³4.1 mmHg (p ¯ 0.03 compared with control by t CBF effects of midazolam. These areas are the right test). Thus, all values for quantitative global CBF were Midazolam decreases rCBF in prefrontal cortex 123 Interaction, rCBF increases, smoothing 12 mm Sagittal Coronal VPC VAC –104 68 VPC VAC –16, –52, –4 (Z = 4.38, sp = 0.079, ht = 0.011) 36, –24, –12 (Z = 4.24, sp = 0.051, ht = 0.018) Transverse 24, –76, 16 (Z = 4.07, sp = 0.145, ht = 0.034) Figure 3. Dose¬condition interaction. Relative rCBF increases signi®cantly greater (p ! 0.001) in the high than the low EEG effect group. Based on quantitative blood ¯ow analysis, these areas represent regions where rCBF decreases are ` resistant ' to the global CBF-decreasing effect of midazolam. Maximal rCBF change at 36, ®24, ®12 is in the right hippocampus, and at ®16, ®52, ®4 is in the left lingual gyrus (BA 19) just rostral to the parahippocampal gyrus. The posterior maximal rCBF change at 24, ®76, ®16 is in the right cerebellum. Z, Z score ; sp, probability based on spatial extent ; ht, probability based on peak height of signal. corrected by ANCOVA for changes in p . With- regional changes in quantitative CBF could not be CO out correction for changes in p , global CBF did not determined on a group basis. CO change signi®cantly between conditions [36.5³5.2 ml} 100 g\min in the control vs. 36.2³6.2 ml}100 g\min in Discussion the midazolam condition (all subjects combined)]. When the in¯uence of p is taken into account there is approx. Midazolam, as do other benzodiazepines, produces a CO a 12 % decrease in global CBF (39.2³4.1 ml}100 g\min speci®c anterograde amnesic effect by interfering with in the control vs. 34.4³6.1 ml}100 g\min in the acquisition of new material independent of drug-induced midazolam condition, p ! 0.02). However, there is a sedation (Curran et al., 1998 ; Ghoneim and Mewaldt, tremendous amount of variability in the data, and the 1990 ; Lister, 1985 ; Veselis et al., 1997c). The speci®city of strongest factor in¯uencing the change in global CBF the memory effects of midazolam suggests that its action between conditions was the individual subject. The on the brain is speci®c, and likely to be mediated in p ¬condition¬subject interaction was the only signi®- discrete locations. This is supported by the fact that at CO cant source of variation (p ¯ 0.02 by ANCOVA). This higher dose of midazolam memory function tended to be indicates that the effect of p on global CBF between impaired to a greater extent, and signi®cant further CO control and midazolam conditions varies in different changes in rCBF occurred in neuroanatomical regions subjects. Because of this, and the varying individual bias in located in the left and right PFC and the parahippocampal determination of global blood ¯ow (see Methods), and hippocampal areas. The changes identi®ed in the PFC 124 R. A. Reinsel et al. represent changes in rCBF of greater magnitude as dose the parahippocampal gyrus and the fusiform gyrus and increases. These regions are involved in various memory portions of the inferior temporal gyrus, also correlated processes, including encoding of novel information, with successful recognition. The authors suggest that semantic processing, and more complex working memory prefrontal and parahippocampal regions may work tasks. Though this study did not directly study the together to produce successful encoding. interaction of midazolam with memory processes, the Asymmetrical rCBF changes in the frontal regions regions demonstrating dose-related changes have pre- associated with encoding vs. retrieval activity are shown viously been associated closely with memory processes. in a study from Tulving's laboratory using verbal material In spatial extent, the largest changes in rCBF related to (Tulving et al., 1994b). Retrieval activity resulted in a midazolam were seen in the left PFC. The left frontal decrease in rCBF in a region normally associated with cortex has been repeatedly identi®ed as playing an increases in rCBF during encoding activities, and this important role in encoding of verbal material. Replicating location is very close to the region identi®ed in our study numerous studies in the cognitive literature, Kapur and as affected by midazolam. A number of studies show colleagues found that deep compared to shallow encoding similar encoding and retrieval asymmetries depending on led to better memory for nouns, and was associated with the speci®c task performed and the nature of material rCBF increases in the left PFC (Kapur et al., 1994). (Their being processed. Tulving has proposed the ` HERA ' ®gure 1 is strikingly similar to our Figure 2.) Nyberg and hypothesis (hemispheric encoding retrieval asymmetry) collaborators used "&O and PET to study rCBF changes based on these ®ndings (Tulving et al., 1994a). In our associated with encoding and retrieval of aspects of study, even though no encoding task was being per- visually presented stimulus information, including item, formed, the left PFC showed dose-related decreases in location, and time. Left frontal brain regions were rCBF with midazolam during a time when memory primarily involved with encoding, while other regions, encoding was demonstrably impaired. The right PFC, including the left hippoccampus and parahippocampal which was affected to a lesser extent by midazolam, is gyrus, the right inferior parietal lobe and left fusiform more likely to be involved in retrieval of information, a gyrus (extrastriate visual cortex), were also involved, process relatively unaffected by midazolam (Ghoneim and depending on the type of stimulus information involved Mewaldt, 1990 ; Lister, 1985). (Nyberg et al., 1996). Grady and colleagues studied face We have previously shown that midazolam decreases encoding and recognition in young and old subjects the amount of information that can be held in the memory (Grady et al., 1995). Younger subjects showed a network buffer, and that this information is highly vulnerable to of activated brain regions associated with later successful interference effects (Reinsel et al., 1993). This behavioural memory performance, including the left PFC (BA 11}47 effect may be mediated by interference with processes and BA 46) and left inferior temporal gyrus. However, the occurring in Broca's area, a region identi®ed as possibly elderly showed an absence of activity in these areas, and mediating the phonological loop rehearsal processes in their overall performance on the face recognition task was Baddeley's working memory model (Baddeley, 1995 ; impaired. The authors interpreted this memory de®cit as Paulesu et al., 1993). Interference with this rehearsal due to a failure to encode the stimuli adequately, as process would result in an encoding de®cit, as produced re¯ected in the lack of frontal activation during encoding. by midazolam. Of special interest is the recent report from a group Apparent dose-related ` increases ' in rCBF also occurred based at Massachusetts General Hospital (Wagner et al., in the parahippocampal and hippocampal regions and 1998) who identi®ed brain regions associated with right cerebellum, contrary to the existing literature successful incidental encoding of words during semantic regarding quantitative metabolism or blood ¯ow changes and non-semantic tasks. As in the study by Nyberg et al. after benzodiazepine administration (see for example, (1996), semantic processing of the stimuli activated the Mathew and Wilson, 1991 ; Volkow et al., 1995). In the left PFC as well as left parahippocampal and fusiform gyri, SPM analysis, these regions showed relative increases of when compared to a non-semantic task. Activity in these approx. 6±9 % from baseline. Quantitative analysis regions also determined whether words were later showed that global CBF decreased approx. 12 % over the remembered or forgotten. Regions in the frontal cortex entire brain, and thus these apparent ` increases ' represent that predicted subsequent retrieval were in the left inferior regions that show signi®cantly less of a decrease in rCBF frontal gyrus (BA 44}6 and 45}47) and the left frontal than the surrounding brain. operculum (BA 47), similar to the areas showing dose- As medial temporal lobe structures are necessary for related decreases in rCBF after midazolam administration encoding of novel information and transfer of information in the present study. In the experiment by Wagner et al. to long-term memory (Brewer et al., 1998 ; Poldrack and (1998), structures in the medial temporal region, including Gabrieli, 1997 ; Squire et al., 1993 ; Wagner et al., 1998), Midazolam decreases rCBF in prefrontal cortex 125 the relative resistance of these to the CBF effects of only likely to have a small impact on the group analysis. midazolam is surprising. This may indicate that the effects A further consideration is that in our paradigm, instead of of midazolam on memory may be primarily mediated by collecting baseline data during an unstimulated resting impairment of encoding processes occurring in the PFC, state, auditory tones were presented in the control with relatively little effect on the consolidation or transfer condition, as well as during drug administration. The of information to long-term memory. In other words, the stimuli were identical in both conditions and subjects amnesia may be produced by an interruption of processing were instructed to ignore the tones. If this auditory at an earlier stage (e.g. rehearsal processes in the stimulation produced cortical activation differentially in phonological loop), so that less information is presented one condition relative to another, our conclusions might to the medial temporal lobe for transfer to long-term be suspect. Nonetheless, some authorities have recom- memory, which is unimpaired by midazolam. Alterna- mended the use of circumscribed baseline stimulation (e.g. tively, the ®nding that the hippocampus and para- a visual ®xation task) in preference to the uncontrolled hippocampal gyrus show less decrease in blood ¯ow than mental activity of the ` resting ' state in order to reduce the rest of the brain may be explained by procedural within- and between-subject variability (Duara et al., factors. The simplest explanation is that volunteers were 1987 ; Ramsay et al., 1993) Another concern is the degree not performing a memory task during the acquisition of of global CBF change found with the administration of PET images, and thus medial temporal lobe structures midazolam, which was of the order of 12 %. SPM analysis were not activated by task demands and would not be has not been extensively applied in this situation. The expected to show any changes in rCBF after drug assumptions of the ANCOVA model in this SPM analysis administration (Friston et al., 1990). are that whole-brain global blood ¯ow will not differ Regarding the relative resistance of the cerebellum to sign®cantly between the conditions being compared. It is rCBF changes with midazolam, inverse relationships this global invariance that makes it possible to identify between metabolism in the PFC and the contralateral signi®cant regional changes in rCBF (Friston et al., 1990). cerebellum have been demonstrated during verbal tasks However, when global CBF changes of similar magnitude (Barker et al., 1991 ; Junck et al., 1988). Several ®bre tracts are induced by changes in p , the use of SPM analysis CO link the cerebellum with the cognitive areas of frontal seems to be successful (Ramsay et al., 1993) cortex, both ipsilaterally and contralaterally (Leiner et al., In conclusion it is interesting to note the congruence of 1995). There has been an increasing number of reports drug-induced changes in rCBF with neuroanatomical showing the involvement of the cerebellum in memory, regions subserving functions affected by these drugs. attention and language processes (Schmahmann, 1996). Dose-related changes in rCBF associated with a CNS- Signi®cant decreases in rCBF were also identi®ed in the active drug may more clearly help identify the speci®c inferior frontal gyrus (BA 11}47). This region mediates neuroanatomical locations of interest, even when the anxious behaviour (Fredrikson et al., 1997 ; Rauch et al., subjects are tested in a ` resting ' state. However, the fact 1997), and has efferent connections to the hypothalamus that midazolam induces changes in rCBF in regions and the limbic system involved in visceral function and previously demonstrated to be involved with memory emotional and instinctive behaviours (Fuster, 1997). The processing does not necessarily mean that this is the orbito-frontal cortex is postulated to mediate learned mechanism of midazolam's amnesic actions. Further study associations between stimuli and reinforcements (Rolls, is required to support or refute this hypothesis, and in the 1996), and speci®cally avoidance behaviour (Heilman, case of midazolam the interaction of the drug with 1997), functions which may underlie the observed changes concurrent memory processes needs to be studied, as has in cerebral activation in anxiety disorders. It seems been attempted using scopolamine (Grasby et al., 1995). reasonable to assume that rCBF changes in this region Before drugs with speci®c effects on memory can be used may correlate with midazolam's anxiolytic effects. as tools in the investigation of normal memory function Several methodological factors in our study deserve using functional neuroimaging techniques, the direct special consideration and may in¯uence the generality of interaction of memory processes with drug administration the results. A possible doubt in this study is that 1 subject must be clari®ed. was included who was ambidextrous, and not purely right-handed. Although stimuli were not presented in Acknowledgements lateralized fashion, there is a possibility that the laterality of the ®ndings may have been affected by the inclusion of We are grateful to Dr Roger Wilson, Dr Steven Larson, Dr this ambidextrous individual. If anything, there would be Ronald Blasberg and Dr Homer Macapinlac of Memorial a tendency to have less lateralization of ®ndings, but in a Sloan-Kettering Cancer Center for their generous advice group of 14 subjects, 1 subject with a differing response is and support of this research. In addition, we thank Dr 126 R. A. Reinsel et al. Osama Mawlawi and Mr Martin Duff for their assistance Fuster JM (1997). The Prefrontal Cortex : Anatomy, Physiology and Neuropsychology of the Frontal Lobe (3rd edn). in the collection and analysis of PET data. Philadelphia : Lippincott±Raven. Ghoneim MM, Mewaldt SP (1990). Benzodiazepines and References human memory : a review. Anesthesiology 72, 926±938. Grady CL, McIntosh AR, Horwitz B, Maisog JM, Ungerleider Baddeley A (1995). Working memory. In : Gazzaniga MS LG, Mentis MJ, Pietrini P, Schapiro MB, Haxby JV (1995). (Ed.), The Cognitive Neurosciences (pp. 755±764). Cambridge, Age-related reductions in human recognition memory due MA. : MIT Press. to impaired encoding. Science 269, 218±221. Barker WW, Yoshii F, Loewenstein DA, Chang JY, Apicella Grasby PM, Friston KJ, Bench C, Cowen PJ, Frith CD, Liddle A, Pascal S, Boothe TE, Ginsberg MD, Duara R (1991). PF, Frackowiak RS, Dolan RJ (1992). Effect of the 5-HT1A Cerebrocerebellar relationship during behavioral activation : partial agonist buspirone on regional cerebral blood ¯ow in a PET study. Journal of Cerebral Blood Flow and Metabolism man. Psychopharmacology 108, 380±386. 11, 48±54. Grasby PM, Friston KJ, Bench CJ, Cowen PJ, Frith CD, Liddle Brewer JB, Zhao Z, Desmond JE, Glover GH, Gabrieli JD PF, Frackowiak RS, Dolan RJ (1993). The effect of the (1998). Making memories : brain activity that predicts how dopamine agonist, apomorphine, on regional cerebral blood well visual experience will be remembered. Science 281, ¯ow in normal volunteers. Psychological Medicine 23, 1185±1187. 605±612. Curran HV, Pooviboonsuk P, Dalton JA, Lader MH (1998). Grasby PM, Frith CD, Paulesu E, Friston KJ, Frackowiak RS, Differentiating the effects of centrally acting drugs on Dolan RJ (1995). The effect of the muscarinic antagonist arousal and memory : an event-related potential study of scopolamine on regional cerebral blood ¯ow during the scopolamine, lorazepam and diphenhydramine. performance of a memory task. Experimental Brain Research Psychopharmacology (Berlin) 135, 27±36. 104, 337±348. D'Esposito M, Ballard D, Aguirre GK, Zarahn E (1998). Greenblatt DJ, Ehrenberg BL, Gunderman J, Locniskar A, Human prefrontal cortex is not speci®c for working Scavone JM, Harmatz JS, Shader RI (1989). Pharmacokinetic memory : A functional MRI study. NeuroImage 8, 274±282. and electropharmacographic study of intravenous Duara R, Gross-Glenn K, Barker WW, Chang JY, Apicella A, diazepam, midazolam and placebo. Clinical Pharmacology Loewenstein D, Boothe T (1987). Behavioral activation and and Therapeutics 45, 356±365. the variability of cerebral glucose metabolic measurements. Gyulai FE, Firestone LL, Mintun MA, Winter PM (1996). In Journal of Cerebral Blood Flow and Metabolism 7, 266±271. vivo imaging of human limbic responses to nitrous oxide Feshchenko VA, Veselis RA, Reinsel RA (1997). Comparison inhalation. Anesthesia and Analgesia 83, 291±298. of the EEG effects of midazolam, thiopental, and propofol : Heilman KM (1997). The neurobiology of emotional the role of underlying oscillatory systems. experience. Journal of Neuropsychiatry and Clinical Neuropsychobiology 35, 211±220. Neuroscience 9, 439±448. Firestone LL, Gyulai F, Mintun M, Adler LJ, Urso K, Winter Jones AK, Friston KJ, Qi LY, Harris M, Cunningham VJ, Jones PM (1996). Human brain activity response to fentanyl T, Feinman C, Frackowiak RS (1991). Sites of action of imaged by positron emission tomography. Anesthesia and morphine in the brain [letter]. Lancet 338, 825. Analgesia 82, 1247±1251. Junck L, Gilman S, Rothley JR, Betley AT, Koeppe RA, Fredrikson M, Fischer H, Wik G (1997). Cerebral blood ¯ow Hichwa RD (1988). A relationship between metabolism in during anxiety provocation. Journal of Clinical Psychiatry 58, frontal lobes and cerebellum in normal subjects studied 16±21. with PET. Journal of Cerebral Blood Flow and Metabolism 8, Friston KJ (1994). Statistical parametric mapping. In : Thatcher 774±782. RW, Hallett M, Zeffiro T, John ER, Huerta M (Eds), Kapur S, Craik FI, Tulving E, Wilson AA, Houle S, Brown Functional Neuroimaging : Technical Foundations (pp. 79±93). GM (1994). Neuroanatomical correlates of encoding in San Diego : Academic Press. episodic memory : levels of processing effect. Proceedings of Friston KJ, Frith CD, Liddle PF, Dolan RJ, Lammertsma AA, the National Academy of Sciences USA 91, 2008±2011. Frackowiak RS (1990). The relationship between global and Leiner HC, Leiner AL, Dow RS (1995). The underestimated local changes in PET scans. Journal of Cerebral Blood Flow cerebellum. Human Brain Mapping 2, 244±254. And Metabolism 10, 458±466. Lister RG (1985). The amnesic action of benzodiazepines in Friston KJ, Grasby PM, Bench CJ, Frith CD, Cowen PJ, Liddle man. Neuroscience and Biobehavioral Reviews 9, 87±94. PF, Frackowiak RS, Dolan R (1992). Measuring the neuromodulatory effects of drugs in man with positron Mathew RJ, Wilson WH (1991). Evaluation of the effects of emission tomography. Neuroscience Letters 141, 106±110. diazepam and an experimental anti-anxiety drug on Friston KJ, Grasby PM, Frith CD, Bench CJ, Dolan RJ, Cowen regional cerebral blood ¯ow. Psychiatry Research 40, PJ, Liddle PF, Frackowiak RS (1991). The neurotransmitter 125±134. Matthew E, Andreason P, Carson RE, Herscovitch P, basis of cognition : psychopharmacological activation Pettigrew K, Cohen R, King C, Johanson CE, Paul SM studies using positron emission tomography. Ciba Foundation Symposium 163, 76±87 (discussion 87±92). (1993). Reproducibility of resting cerebral blood ¯ow Midazolam decreases rCBF in prefrontal cortex 127 measurements with H (15)O positron emission Schmahmann JD (1996). From movement to thought : tomography in humans. Journal of Cerebral Blood Flow and Anatomic substrates of the cerebellar contribution to Metabolism 13, 748±754. cognitive processing. Human Brain Mapping 4, 174±198. Squire LR, Knowlton B, Musen G (1993). The structure and Nyberg L, McIntosh AR, Cabeza R, Habib R, Houle S, organization of memory. Annual Review of Psychology 44, Tulving E (1996). General and speci®c brain regions 453±495. involved in encoding and retrieval of events : what, where Talairach J, Tournoux P (1988). Co-Planar Stereotaxic Atlas of and when. Proceedings of the National Academy of Sciences the Human Brain. Stuttgart : Georg Thieme Verlag. USA 93, 11280±11285. Tanelian Dl, Kosek P, Mody I, MacIver MB (1993). The role Parekh PI, Spencer JW, George MS, Gill DS, Ketter TA, of the GABA-A receptor}chloride channel complex in Andreason P, Herscovitch P, Post RM (1995). Procaine- anesthesia. Anesthesiology 78, 757±776. induced increases in limbic rCBF correlate positively with Tulving E, Kapur S, Craik FI, Moscovitch M, Houle S (1994a). increases in occipital and temporal EEG fast activity. Brain Hemispheric encoding}retrieval asymmetry in episodic Topography 7, 209±216. memory : positron emission tomography ®ndings. Paulesu E, Frith CD, Frackowiak RS (1993). The neural Proceedings of the National Academy of Sciences USA 91, correlates of the verbal component of working memory. 2016±2020. Nature 362, 342±345. Tulving E, Kapur S, Markowitsch HJ, Craik FI, Habib R, Poldrack RA, Gabrieli JDE (1997). Functional anatomy of Houle S (1994b). Neuroanatomical correlates of retrieval in long-term memory. Journal of Clinical Neurophysiology 14, episodic memory : auditory sentence recognition. 294±310. Proceedings of the National Academy of Sciences USA 91, Ramsay SC, Murphy K, Shea SA, Friston KJ, Lammertsma 2012±2015. AA, Clark JC, Adams L, Guz A, Frackowiak RS (1993). Veselis RA, Glass P, Dnistrian A, Reinsel R (1997a). Changes in global cerebral blood ¯ow in humans : effect on Performance of computer-assisted continuous infusion at regional cerebral blood ¯ow during a neural activation task. low concentrations of intravenous sedatives. Anesthesia and Journal of Physiology 471, 521±534. Analgesia 84, 1049±1057. Veselis RA, Reinsel RA, Beattie BJ, Mawlawi OR, Feshchenko Rauch SL, Savage CR, Alpert NM, Fischman AJ, Jenike MA VA, DiResta GR, Larson SM, Blasberg RG (1997b). (1997). The functional neuroanatomy of anxiety : a study of Midazolam changes cerebral blood ¯ow in discrete brain three disorders using positron emission tomography and regions : an H (15)O positron emission tomography study. symptom provocation. Biological Psychiatry 42, 446±452. Anesthesiology 87, 1106±1117. Reinsel R, Veselis R, Wronski M, Marino P, Heino R, Veselis RA, Reinsel RA, Feshchenko VA, Di Resta G, Alagesan R (1993). Memory impairment during conscious Mawlawi O, Beattie B, Silbersweig D, Stern E, Blasberg R, sedation : A comparison of midazolam, propofol and Macapinlac H, Finn R, Goldsmith S, Larson S (1995). thiopental. In : Sebel PS, Bonke B, Winograd E (Eds.), Cerebral blood ¯ow changes during midazolam sedation Memory and Awareness in Anesthesia (pp. 127±140). using O-15 positron emission tomography (PET) [Abstract]. Englewood, N.J. : Prentice-Hall. Anesthesiology 83, A154. Reinsel RA, Veselis RA, Feshchenko VA, Di Resta GR, Veselis RA, Reinsel RA, Feshchenko VA, Wronski M (1997c). Mawlawi O, Beattie B, Silbersweig D, Stern E, Blasberg R, The comparative amnestic effects of midazolam, propofol, Macapinlac H, Finn R, Goldsmith S, Larson S (1995). thiopental, and fentanyl at equisedative concentrations. Target detection and the prefrontal cortex. A PET scan Anesthesiology 87, 749±764. study of the P300 event-related potential. Annals of the Volkow ND, Wang GJ, Hitzemann R, Fowler JS, Pappas N, New York Academy of Sciences 769, 393±397. Lowrimore P, Burr G, Pascani K, Overall J, Wolf AP (1995). Rolls ET (1996). The orbitofrontal cortex. Philosophical Depression of thalamic metabolism by lorazepam is Transactions of the Royal Society of London (B : Biological associated with sleepiness. Neuropsychopharmacology 12, Sciences) 351, 1433±1443. 123±132. Roy-Byrne P, Fleishaker J, Arnett C, Dubach M, Stewart J, Wagner AD, Schacter DL, Rotte M, Koutstaal W, Maril A, Radant A, Veith R, Graham M (1993). Effects of acute and Dale AM, Rosen BR, Buckner RL (1998). Building chronic alprazolam treatment on cerebral blood ¯ow, memories : remembering and forgetting of verbal memory, sedation, and plasma catecholamines. experiences as predicted by brain activity. Science 281, Neuropsychopharmacology 8, 161±169. 1188±1191. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png International Journal of Neuropsychopharmacology Oxford University Press

Midazolam decreases cerebral blood flow in the left prefrontal cortex in a dose-dependent fashion

Loading next page...
 
/lp/oxford-university-press/midazolam-decreases-cerebral-blood-flow-in-the-left-prefrontal-cortex-q4tTK74UxP

References (0)

References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.

Publisher
Oxford University Press
Copyright
© 2000 Collegium Internationale Neuropsychopharmacologicum
ISSN
1461-1457
eISSN
1469-5111
DOI
10.1017/S1461145700001814
Publisher site
See Article on Publisher Site

Abstract

Midazolam, a short-lived benzodiazepine producing sedation and reversible anterograde amnesia, was administered intravenously to 14 healthy male volunteers. Regional cerebral blood ¯ow (rCBF) was measured using positron emission tomography (PET) with intravenous H "&O at either a ` high ' midazolam EEG effect (EEG signs of stage 2 sleep), or ` low ' midazolam EEG effect (increase in EEG beta power only). Memory tests administered following PET scans showed signi®cant drug-induced impairment in learning and retrieval at the same drug concentration at which PET images were acquired. Statistical parametric mapping was used to identify regions where rCBF changes after drug administration were signi®cantly different in the high- vs. low- effect groups. Dose¬condition interactions were found in the left dorsolateral prefrontal cortex [Brodmann's areas (BA) 9 and 46], bilateral orbital-frontal cortex (BA 47), the left middle temporal gyrus (BA 22) and the right hippocampus. The predominantly left frontal rCBF decreases occur in a region associated with semantic processing, working memory, and encoding of verbal material, a process preferentially affected by midazolam. Our interpretation is that rCBF changes in the hippocampus are unlikely to mediate the anterograde amnesia produced by midazolam. Although in the present study PET images were acquired during the resting state rather than during memory processing, these results underscore the need for further investigation relating to the interaction of midazolam with speci®c cognitive operations in these brain regions. Received 30 June 1999 ; Reviewed 27 September 1999 ; Revised 20 December 1999 ; Accepted 10 January 2000 Key words : Benzodiazepines : midazolam, regional cerebral blood ¯ow, positron emission tomography, memory, prefrontal cortex, medial temporal lobe. Introduction effects of midazolam are quite speci®c, with encoding of new information affected more than retrieval of old Minor tranquillizers with short half-lives are used in information (Ghoneim and Mewaldt, 1990 ; Lister, 1985). anaesthesia to sedate patients for short surgical pro- The EEG effects of midazolam are similar to those of other cedures. Recovery is rapid and there is less risk of adverse benzodiazepines (Greenblatt et al., 1989), which act on the side-effects than with general anaesthetics. Among the GABA-A receptor throughout the brain (Tanelian et al., benzodiazepine class, midazolam (half-life of 2±4 h), is 1993). But neither electrophysiology nor receptor-binding widely administered for ambulatory surgery using the studies can explain how midazolam produces its effects of technique of conscious sedation. In contrast to general sedation and amnesia. If speci®c regions of the brain anaesthesia, conscious sedation allows the patient to affected by midazolam could be identi®ed, then a better remain awake and cooperative during the procedure. In understanding of how midazolam produces reversible this context, the ability to ablate memory for surgical amnesia may be possible. events is seen as a bene®cial side-effect. A number of studies have shown changes in regional Midazolam has marked effects on episodic memory, cerebral blood ¯ow (rCBF) following the administration of producing profound anterograde amnesia as well as various drugs active in the central nervous system (CNS) signi®cant sedation (Veselis et al., 1997c). The memory (Firestone et al., 1996 ; Friston et al., 1991, 1992 ; Grasby et al. 1992, 1993, 1995 ; Gyulai et al., 1996 ; Jones et al., Address for correspondence : Dr R. A. Reinsel, Department of 1991 ; Parekh et al., 1995 ; Roy-Byrne et al., 1993), Anesthesiology and Critical Care Medicine, Memorial Sloan-Kettering including midazolam (Veselis et al., 1997b). It is interesting Cancer Center, 1275 York Avenue, Box 24, New York, NY 10021, that the neuroanatomical regions demonstrating changes USA. in rCBF in the presence of these drugs frequently subserve Tel. : (212) 639-8111 Fax : (212) 772-8646 functions that are in¯uenced by those drugs. Thus, it is E-mail : reinselr!mskcc.org 118 R. A. Reinsel et al. possible that drug-induced changes in rCBF can identify EEG power spectrum was used to verify the degree of the neuroanatomical structures necessary for the ex- drug effect at the time of drug infusion. Low midazolam pression of the drug's clinical effects. Unfortunately, many effect was de®ned by high frequency, low amplitude EEG of these studies demonstrate changes in rCBF in numerous activity in the beta frequency range (13±20 Hz) and locations in the presence of drug, which we also found represents a serum concentration at which a 50 % with midazolam (Veselis et al., 1997b). Identi®cation of decrement in memory effect is present (Veselis et al., dose-related changes in rCBF in this situation may help 1997c). The high effect was de®ned by the presence of de®ne the neuroanatomical regions worthy of closer ` sleep ' spindles at approx. 14 Hz, and was chosen to attention in further studies of the neuroanatomical basis of represent an EEG effect clearly separable from the low drug effects. midazolam effect (Feshchenko et al., 1997). Following the This report details the investigation of dose-related start of drug infusion a period of 20±30 min elapsed changes in rCBF patterns during midazolam adminis- before subsequent positron emission tomography (PET) tration using an interaction analysis between dose (high scans were obtained to allow time for complete equi- vs. low EEG effect) and the presence}absence of drug. The libration of drug effect. Midazolam infusion occurred for interaction analysis is designed to identify brain regions approx. 60 min. Midazolam concentrations were deter- where the effect of a high dose is signi®cantly different mined by high performance liquid chromatography from a low dose of midazolam. Our hypothesis is that the (Veselis et al., 1997a). neuroanatomical regions thus identi®ed are more likely to represent the neuroanatomical substrate which midazolam Neuropsychologic state during PET scanning acts upon, possibly mediating its clinical effects of amnesia In both control and midazolam conditions subjects and sedation. listened to binaural auditory tones at 1000 Hz, 80 dB, with an interstimulus interval of 1.1 s, presented 30 s Method before isotope injection and continuing throughout the scanning procedure. This was done because the attention This investigation was approved by the Hospital task (scans 2, 4, 6) required a control condition with Institutional Review Board and Radiation Safety Com- subjects hearing tones. This control condition was used mittee before accrual of subjects took place. for comparisons for the attention task, as well as the administration of midazolam. Subjects were instructed to Subjects pay no attention to the tones. The room was dimly lit Fourteen normal male volunteers (ages 21±44 yr ; with minimal background noise. mean³s.d. 28.1³5.8 yr) participated in this study after giving informed consent. All but one was strongly right- Memory testing handed (the one exception being ambidextrous), and no Between scans 2 and 3 (` baseline ') and 7 and 8 (` drug ') volunteer had a history or physical evidence of neurologic subjects were asked to memorize a list of 16 words or psychiatric illness. Blood sampling was accomplished presented through headphones by tape recorder. It should from a radial arterial line. be noted that there was no memory task administered during PET scanning. However, as midazolam concen- Study design trations were held constant during the infusion, the Subjects were randomized to receive either a high or low memory data serve as a measure of the amnesic effect dose of midazolam. All subjects were scanned in three present during PET scanning at these doses of midazolam. conditions on a single day : control (scans 1, 3, 5), an Two equivalent word lists were employed, in counter- attention task (scans 2, 4, 6), and during midazolam balanced order, for baseline and drug conditions. Memory infusion (scans 7, 8, 9). This report concerns data collected was assessed by immediate free recall after each pres- from scans 1, 3, 5 and 7, 8, 9. The results of the attention entation of the list, and again at the end of the study day task have been reported separately (Reinsel et al., 1995). by delayed recognition. Measures analysed for this report After scan 6, midazolam was infused to one of two include trial 4 (the last learning trial), immediate recall concentrations using a computer-assisted continuous after ` interference ', delayed recall some 30±40 min later, infusion device (CACI*), during simultaneous EEG moni- as well as delayed recognition. We predicted that the low- toring (see Veselis et al., 1995, for details of methods). The effect group would have higher recall and recognition scores than the high-effect group. Our memory test is * CACI : Glass PSA and Jacobs J, Duke University Medical Center, modelled on the Rey Auditory Verbal Learning Test (Rey Durham, NC. (See Veselis et al., 1997a, for details of the use of this AVLT) but with a few modi®cations required by the PET device.) Midazolam decreases rCBF in prefrontal cortex 119 scan environment. Word lists were presented four times contiguous slices are obtained. Subjects were positioned for learning and recall in the interval between PET scans. so that the centre of the ®eld of view was approx. 3 cm This took approx. 10 min, at which point the next dose of above the orbito-meatal line. This resulted in some of the isotope was ready for injection. Given the rigorous time superior cortex not being in the ®eld of view, as the schedule imposed by a fast-decaying isotope, it was not scanner could only image up to 8 cm above the orbito- possible to delay the next scan in order to present the meatal line. Subjects spent approx. 3 h in the scanner interference word list. The 2-min PET scan itself and the (mean total time : 192 min, range 112±273 min) and were associated bustle of activity around the volunteer took the receiving midazolam for approximately the last hour of place of the interference list that would normally be that time. presented. Subjects were instructed not to rehearse the words during this interval, and immediate recall was Data analysis obtained immediately after the PET scan. During the drug condition, delayed recall was obtained at the same serum Behavioural data were analysed by non-parametric concentration that was present during learning. Mann±Whitney U test for differences between groups, and by paired t test for within-subject change between conditions. Test results are two-tailed p values unless Study groups otherwise noted. PET image analysis was performed using As the EEG response during PET scanning could possibly statistical parametric mapping (SPM95) developed by the be different than the EEG effect seen just before PET Wellcome Department of Cognitive Neurology, Uni- scanning commenced, subjects were classi®ed into low- versity College London, and implemented in pro matlab and high- effect groups based solely on the EEGs obtained v. 4.2 (Mathworks, New York). Images were realigned to during PET scanning. After accrual of all subjects, the EEG the ®rst scan and transformed into the standard stereo- spectra obtained during scanning were reviewed for tactic space utilized in the human brain atlas of Talairach determination of study groups for data analysis. Three ex- and Tournoux (1988). As a ®nal pre-processing step the perienced encephalographers (R.V., R.R., V.F.) examined images were smoothed using an isotropic Gaussian kernel EEG spectra from four derivations across the scalp [12 mm resulting in a FWHM (x, y, z) smoothing of 18, (Fz, Cz, Pz and O2). Spectra were classi®ed without 25, 21 mm] to accommodate normal variability in func- knowledge of which subject or study group they were tional and gyral anatomy for group analysis. The three associated with. Subjects with a clearly visible peak in the scans obtained for a particular subject in baseline and drug power spectrum centred at 14 Hz during the drug conditions were included as replicated measurements for condition were included in the high midazolam effect that subject in the group analysis. After specifying the group (Feshchenko et al., 1997). Subjects without a clearly appropriate design matrix (multi-study with 2 conditions visible spectral peak were classi®ed to the low midazolam and 3 replications), the condition and subject effects were effect group. After this classi®cation, the high- and low- estimated according to the general linear model at each effect groups consisted of 7 subjects each. This classi- and every voxel using a subject-speci®c ANCOVA model ®cation was congruent with the initial randomization to (Friston, 1994). Mean global CBF was normalized by low- or high-effect groups, except for 3 subjects. Two SPM95 to 50 ml}100 g\min. This analysis generated a were randomized to the high-effect group, but the spectral mean voxel value with associated error variance for each peak was not clearly visible. One subject was randomized condition. Voxel¬voxel comparison between conditions to the low-effect group, but power spectral analysis was then made, and mean voxel values of control and revealed a prominent spectral peak in the spindle range experimental condition scans were compared for (12±15 Hz). signi®cant differences by t test. A group (high, low effect)¬condition (baseline, drug) interaction analysis was used to identify regions where CBF changes in the PET scanning high-effect group were signi®cantly different than the For each scan, 20 mCi of H "&O in 10 cc were infused low-effect group, expressed as Z scores using a threshold over 20 s into the antecubital vein using a constant rate Z value of " 3.09 (corresponding to a p ! 0.001). A infusion pump with 15 min between scans. PET scans further correction for multiple comparisons was applied, were obtained with a PC4600 NeuroPet scanner, with a resulting in a corrected p ! 0.05 at the level of individual 10 cm axial ®eld of view and a 1.2 cm in-plane resolution. voxels. Signi®cant maxima were mapped using the This scanner consists of ®ve detection rings separated by Talairach and Tournoux atlas to determine the anatomical septa which detect coincident counts in projections location and corresponding Brodmann's area, where orthogonal to the axial plane of the scanner, and thus 9 applicable. 120 R. A. Reinsel et al. Quantitative image analysis 79.5³9.7 kg for all subjects combined). Signi®cant differences were present in total dose of midazolam The PET frame data, the PET central ring data and the administered (7.5³1.7 vs. 9.7³1.3 mg, p ¯ 0.02 by t blood activity data were all decay corrected to the time of test) and average serum concentrations during scanning injection as determined by the PC's clock at the start of PC (74³24 vs. 129³48 ng}ml, p ¯ 0.02 by t test). Sedation data acquistion. The ` head curve ' data were roughly in the low-effect group varied from a small degree of calibrated to activity concentration by plotting the mean subjective sleepiness to requiring verbal stimulation to brain activity concentration, as determined from the respond. Subjects in the high-effect group were heavily reconstructed and calibrated PET images for each frame, sedated and frequently required physical stimulation to vs. the mean count rate of the ` head curve ' over the time respond. Assessment of sedation and stimulation, if period corresponding to the PET frames. The slope of the necessary, were done just prior to the memory test. best least-squares ®tting line was taken to be the conversion factor. A single conversion factor was used for all runs on a given subject. Unfortunately, values derived Memory effects for absolute CBF are biased and therefore unsuitable for inter-subject comparisons. This bias is probably related to Both groups suffered memory loss during midazolam inclusion of non-cerebral structures in the volume of brain administration (see Table 1). Learning was effectively analysed, and will vary among individuals (Matthew et abolished in both groups, with mean recall on trial 4 of al., 1993 ; Ramsay et al., 1993). However, the bias should less than 25 % of the words presented. Immediate and be constant across the different runs for a single subject delayed recall dropped to near zero. Comparing change and therefore paired comparisons of the sedated to the scores (baseline minus drug), the high-effect group baseline condition are valid and accurate (Matthew et al., showed the predicted larger decrement in immediate 1993). Because of the individual variability in CBF values, recall after midazolam (p ! 0.05 one-tailed). From the 16- and of the small, insigni®cant changes seen in the low- word list administered during midazolam infusion, the effect group, all subjects were combined for global CBF number of words recognized at the end of the study day analysis. were 5.0³4.5 (p ! 0.002) and 1.6³1.7 (p ! 0.001) words for the low- and high- effect groups, respectively Results (compared to baseline scores by paired t test, see Table 1). The difference between recognition memory scores Study groups between high- and low-effect groups during midazolam No differences in demographics were observed between sedation had a signi®cance of p ¯ 0.082 one-tailed by the low and high EEG effect groups (mean weight³s.d.: Mann±Whitney U test. The high variance of recognition Table 1. Detailed results of memory testing performed between PET scanning Low effect High effect Baseline Drug Baseline Drug No. words correct on 13.3 (2.5) 3.4 (2.8) 13.3 (2.4) 1.7 (2.9) 4th presentation of word list Immediate recall after 11.4 (3.8) 1.9 (1.8) 12.9 (2.8) 0.9 (1.6) ®nal list presentation Delayed recall (30 min) 12.5 (2.6) 0.6 (1.5) 12.7 (2.9) 0.3 (0.8) after ®nal presentation Recognition at end of 14.7 (1.0) 5.0 (4.5)* 14.6 (0.8) 1.6 (1.7)* study day For both groups combined, the difference between results from the drug vs. baseline conditions was highly signi®cant (p ! 0.001 by paired t test) for all parameters. The only result approaching statistical signi®cance, when the low vs. high midazolam effect groups were compared, was the comparison of word recognition after midazolam (* p ¯ 0.08 one-tailed). All results are presented as means with the standard deviation indicated in parentheses. Midazolam decreases rCBF in prefrontal cortex 121 rCBF decreases rCBF increases Sagittal Coronal Sagittal Coronal R R 0 0 0 0 32 32 VPC VAC 064 VPC VAC 064 –104 68 –104 68 Low VPC VAC VPC VAC dose 0 0 64 64 Transverse Transverse Sagittal Coronal Sagittal Coronal R R 0 0 0 0 32 32 VPC VAC 064 VPC VAC 064 –104 68 –104 68 High VPC VAC VPC VAC dose 0 0 R R 64 64 Transverse Transverse Figure 1. Changes in rCBF related to midazolam administration. Healthy volunteers received midazolam infusion to either 74³24 or 129³48 ng}ml concentrations of midazolam (to produce low and high EEG effects), which were held constant by computer control during memory testing and PET scanning. This ®gure summarizes the main effects of midazolam on rCBF for the two doses administered. Regional CBF changes were identi®ed using statistical parametric mapping (SPM) analysis, and thus all changes are relative to a normalized baseline. Regions of signi®cant change in CBF (p ! 0.001) are shown in a ` glass brain ' projection. Decreases in rCBF are shown in the upper left and lower left panels for the low and high EEG effect groups, respectively. Increases in rCBF are shown in the upper right and lower right panels. The ®eld of view of the PET scanner was not able to cover the entire brain. Quantitative analysis (Veselis et al., 1997b) revealed that regions of ` rCBF increases ' are actually decreases in rCBF, but the decreases are statistically signi®cantly smaller than other blood ¯ow changes. Thus these areas can be considered regions where midazolam has less effect on rCBF. scores in the low-effect group most likely accounts for the between dose and condition. The largest areas of change failure to reach signi®cance in this comparison. in rCBF with higher doses of midazolam occurred in the left prefrontal cortex (PFC) (see Figure 2). The most signi®cant decreases in rCBF occurred at Talairach Dose-related decreases in rCBF coordinates (x, y, z, in mm) ®44, 34, 16 and ®42, 26, 28 As previously reported, the effect of midazolam is not located in the left middle frontal gyrus, close to uniform over the brain. In the high EEG effect group, 22 Brodmann's areas (BA) 46 and 9. These areas are involved different areas of brain demonstrated signi®cant decreases in working memory as well as other functions (D'Esposito in rCBF, but there were only 11 areas common to both the et al., 1998). A smaller area of rCBF decrease was seen low and high EEG effect groups (see Figure 1). Many of bilaterally in the inferior frontal gyrus at ®26, 28, 0 and these regions have been identi®ed as important in 24, 28, 0 (BA 47) located in the frontal operculum (inner processes involving memory, attention and mood. portion of the lateral sulcus, just anterior to the insula). An The reader is referred to Veselis et al. (1995) for speci®c additional area of signi®cantly lower rCBF after results relating to the separate doses of midazolam. We midazolam was found at ®42, ®34, ®4 in the left speci®cally report here on the signi®cant interaction middle temporal gyrus (BA 22). 122 R. A. Reinsel et al. Interaction, rCBF decreases, smoothing 12 mm Sagittal Coronal VPC VAC –104 68 –44, 34, 16 VPC VAC (Z = 4.86, sp = 0.020, ht = 0.001) –42, 26, 28 (Z = 4.85, sp = 0.020, ht = 0.002) –26, 28, 0 (Z = 4.60, sp = 0.020, ht = 0.004) 24, 28, 0 (Z = 4.26, sp = 0.148, ht = 0.017) Transverse –42, –34, 4 (Z = 4.06, sp = 0.161, ht = 0.036) Figure 2. Dose¬condition interaction. Relative rCBF decreases during midazolam administration which were signi®cantly greater (p ! 0.001) in the high than the low EEG effect group. Peak changes in rCBF occurred at ®44, 34, 16 and ®42, 26, 28 in the left middle frontal gyrus (BA 46 and 9, respectively) ; ®26, 28, 0 and 24, 28, 0 in the left and right inferior frontal gyrus (BA 47) ; and ®42, ®34, 4 in the sulcus between the left middle and superior temporal gyrus (BA 22). Z, Z score ; sp, probability based on spatial extent ; ht, probability based on peak height of signal. Dose-related ` increases ' in rCBF (regions of lesser hippocampus (36, ®24, ®12) and parahippocampal decrease in rCBF) gyrus (22, ®48, ®4), and the left lingual gyrus just rostral to the parahippocampal gyrus (®16, ®52, ®4; The results of this analysis represent areas of rCBF which BA 19). Another area was identi®ed in the right posterior are ` increased ' in the high EEG effect group when cerebellum, that is very close to the extreme inferior compared with the low EEG effect, when normalized for extent of the fusiform gyrus (BA 18). global changes. Midazolam in the doses used in this study induced about a 12 % reduction in global CBF (Veselis et al., 1997b). Regions which showed greater ` increases ' in the high EEG effect group varied from ­6 % (left middle Quantitative blood ¯ow values occipital gyrus) to ­9 % (right middle occipital gyrus). As the overall decreases in global CBF were larger than the A major confounding variable in the measurement of localized rCBF effects, regions identi®ed as showing global CBF is the change in p following drug CO signi®cant ` increases ' in rCBF in the SPM analysis are best administration. In the low-effect group it increased interpreted as areas where rCBF did not decrease to the 3.5³1.3 mmHg (p ¯ 0.04 compared with control by same extent as the rest of the brain. One could think of t test), and in the high-effect group it increased these regions as being relatively resistant to the global 5.1³4.1 mmHg (p ¯ 0.03 compared with control by t CBF effects of midazolam. These areas are the right test). Thus, all values for quantitative global CBF were Midazolam decreases rCBF in prefrontal cortex 123 Interaction, rCBF increases, smoothing 12 mm Sagittal Coronal VPC VAC –104 68 VPC VAC –16, –52, –4 (Z = 4.38, sp = 0.079, ht = 0.011) 36, –24, –12 (Z = 4.24, sp = 0.051, ht = 0.018) Transverse 24, –76, 16 (Z = 4.07, sp = 0.145, ht = 0.034) Figure 3. Dose¬condition interaction. Relative rCBF increases signi®cantly greater (p ! 0.001) in the high than the low EEG effect group. Based on quantitative blood ¯ow analysis, these areas represent regions where rCBF decreases are ` resistant ' to the global CBF-decreasing effect of midazolam. Maximal rCBF change at 36, ®24, ®12 is in the right hippocampus, and at ®16, ®52, ®4 is in the left lingual gyrus (BA 19) just rostral to the parahippocampal gyrus. The posterior maximal rCBF change at 24, ®76, ®16 is in the right cerebellum. Z, Z score ; sp, probability based on spatial extent ; ht, probability based on peak height of signal. corrected by ANCOVA for changes in p . With- regional changes in quantitative CBF could not be CO out correction for changes in p , global CBF did not determined on a group basis. CO change signi®cantly between conditions [36.5³5.2 ml} 100 g\min in the control vs. 36.2³6.2 ml}100 g\min in Discussion the midazolam condition (all subjects combined)]. When the in¯uence of p is taken into account there is approx. Midazolam, as do other benzodiazepines, produces a CO a 12 % decrease in global CBF (39.2³4.1 ml}100 g\min speci®c anterograde amnesic effect by interfering with in the control vs. 34.4³6.1 ml}100 g\min in the acquisition of new material independent of drug-induced midazolam condition, p ! 0.02). However, there is a sedation (Curran et al., 1998 ; Ghoneim and Mewaldt, tremendous amount of variability in the data, and the 1990 ; Lister, 1985 ; Veselis et al., 1997c). The speci®city of strongest factor in¯uencing the change in global CBF the memory effects of midazolam suggests that its action between conditions was the individual subject. The on the brain is speci®c, and likely to be mediated in p ¬condition¬subject interaction was the only signi®- discrete locations. This is supported by the fact that at CO cant source of variation (p ¯ 0.02 by ANCOVA). This higher dose of midazolam memory function tended to be indicates that the effect of p on global CBF between impaired to a greater extent, and signi®cant further CO control and midazolam conditions varies in different changes in rCBF occurred in neuroanatomical regions subjects. Because of this, and the varying individual bias in located in the left and right PFC and the parahippocampal determination of global blood ¯ow (see Methods), and hippocampal areas. The changes identi®ed in the PFC 124 R. A. Reinsel et al. represent changes in rCBF of greater magnitude as dose the parahippocampal gyrus and the fusiform gyrus and increases. These regions are involved in various memory portions of the inferior temporal gyrus, also correlated processes, including encoding of novel information, with successful recognition. The authors suggest that semantic processing, and more complex working memory prefrontal and parahippocampal regions may work tasks. Though this study did not directly study the together to produce successful encoding. interaction of midazolam with memory processes, the Asymmetrical rCBF changes in the frontal regions regions demonstrating dose-related changes have pre- associated with encoding vs. retrieval activity are shown viously been associated closely with memory processes. in a study from Tulving's laboratory using verbal material In spatial extent, the largest changes in rCBF related to (Tulving et al., 1994b). Retrieval activity resulted in a midazolam were seen in the left PFC. The left frontal decrease in rCBF in a region normally associated with cortex has been repeatedly identi®ed as playing an increases in rCBF during encoding activities, and this important role in encoding of verbal material. Replicating location is very close to the region identi®ed in our study numerous studies in the cognitive literature, Kapur and as affected by midazolam. A number of studies show colleagues found that deep compared to shallow encoding similar encoding and retrieval asymmetries depending on led to better memory for nouns, and was associated with the speci®c task performed and the nature of material rCBF increases in the left PFC (Kapur et al., 1994). (Their being processed. Tulving has proposed the ` HERA ' ®gure 1 is strikingly similar to our Figure 2.) Nyberg and hypothesis (hemispheric encoding retrieval asymmetry) collaborators used "&O and PET to study rCBF changes based on these ®ndings (Tulving et al., 1994a). In our associated with encoding and retrieval of aspects of study, even though no encoding task was being per- visually presented stimulus information, including item, formed, the left PFC showed dose-related decreases in location, and time. Left frontal brain regions were rCBF with midazolam during a time when memory primarily involved with encoding, while other regions, encoding was demonstrably impaired. The right PFC, including the left hippoccampus and parahippocampal which was affected to a lesser extent by midazolam, is gyrus, the right inferior parietal lobe and left fusiform more likely to be involved in retrieval of information, a gyrus (extrastriate visual cortex), were also involved, process relatively unaffected by midazolam (Ghoneim and depending on the type of stimulus information involved Mewaldt, 1990 ; Lister, 1985). (Nyberg et al., 1996). Grady and colleagues studied face We have previously shown that midazolam decreases encoding and recognition in young and old subjects the amount of information that can be held in the memory (Grady et al., 1995). Younger subjects showed a network buffer, and that this information is highly vulnerable to of activated brain regions associated with later successful interference effects (Reinsel et al., 1993). This behavioural memory performance, including the left PFC (BA 11}47 effect may be mediated by interference with processes and BA 46) and left inferior temporal gyrus. However, the occurring in Broca's area, a region identi®ed as possibly elderly showed an absence of activity in these areas, and mediating the phonological loop rehearsal processes in their overall performance on the face recognition task was Baddeley's working memory model (Baddeley, 1995 ; impaired. The authors interpreted this memory de®cit as Paulesu et al., 1993). Interference with this rehearsal due to a failure to encode the stimuli adequately, as process would result in an encoding de®cit, as produced re¯ected in the lack of frontal activation during encoding. by midazolam. Of special interest is the recent report from a group Apparent dose-related ` increases ' in rCBF also occurred based at Massachusetts General Hospital (Wagner et al., in the parahippocampal and hippocampal regions and 1998) who identi®ed brain regions associated with right cerebellum, contrary to the existing literature successful incidental encoding of words during semantic regarding quantitative metabolism or blood ¯ow changes and non-semantic tasks. As in the study by Nyberg et al. after benzodiazepine administration (see for example, (1996), semantic processing of the stimuli activated the Mathew and Wilson, 1991 ; Volkow et al., 1995). In the left PFC as well as left parahippocampal and fusiform gyri, SPM analysis, these regions showed relative increases of when compared to a non-semantic task. Activity in these approx. 6±9 % from baseline. Quantitative analysis regions also determined whether words were later showed that global CBF decreased approx. 12 % over the remembered or forgotten. Regions in the frontal cortex entire brain, and thus these apparent ` increases ' represent that predicted subsequent retrieval were in the left inferior regions that show signi®cantly less of a decrease in rCBF frontal gyrus (BA 44}6 and 45}47) and the left frontal than the surrounding brain. operculum (BA 47), similar to the areas showing dose- As medial temporal lobe structures are necessary for related decreases in rCBF after midazolam administration encoding of novel information and transfer of information in the present study. In the experiment by Wagner et al. to long-term memory (Brewer et al., 1998 ; Poldrack and (1998), structures in the medial temporal region, including Gabrieli, 1997 ; Squire et al., 1993 ; Wagner et al., 1998), Midazolam decreases rCBF in prefrontal cortex 125 the relative resistance of these to the CBF effects of only likely to have a small impact on the group analysis. midazolam is surprising. This may indicate that the effects A further consideration is that in our paradigm, instead of of midazolam on memory may be primarily mediated by collecting baseline data during an unstimulated resting impairment of encoding processes occurring in the PFC, state, auditory tones were presented in the control with relatively little effect on the consolidation or transfer condition, as well as during drug administration. The of information to long-term memory. In other words, the stimuli were identical in both conditions and subjects amnesia may be produced by an interruption of processing were instructed to ignore the tones. If this auditory at an earlier stage (e.g. rehearsal processes in the stimulation produced cortical activation differentially in phonological loop), so that less information is presented one condition relative to another, our conclusions might to the medial temporal lobe for transfer to long-term be suspect. Nonetheless, some authorities have recom- memory, which is unimpaired by midazolam. Alterna- mended the use of circumscribed baseline stimulation (e.g. tively, the ®nding that the hippocampus and para- a visual ®xation task) in preference to the uncontrolled hippocampal gyrus show less decrease in blood ¯ow than mental activity of the ` resting ' state in order to reduce the rest of the brain may be explained by procedural within- and between-subject variability (Duara et al., factors. The simplest explanation is that volunteers were 1987 ; Ramsay et al., 1993) Another concern is the degree not performing a memory task during the acquisition of of global CBF change found with the administration of PET images, and thus medial temporal lobe structures midazolam, which was of the order of 12 %. SPM analysis were not activated by task demands and would not be has not been extensively applied in this situation. The expected to show any changes in rCBF after drug assumptions of the ANCOVA model in this SPM analysis administration (Friston et al., 1990). are that whole-brain global blood ¯ow will not differ Regarding the relative resistance of the cerebellum to sign®cantly between the conditions being compared. It is rCBF changes with midazolam, inverse relationships this global invariance that makes it possible to identify between metabolism in the PFC and the contralateral signi®cant regional changes in rCBF (Friston et al., 1990). cerebellum have been demonstrated during verbal tasks However, when global CBF changes of similar magnitude (Barker et al., 1991 ; Junck et al., 1988). Several ®bre tracts are induced by changes in p , the use of SPM analysis CO link the cerebellum with the cognitive areas of frontal seems to be successful (Ramsay et al., 1993) cortex, both ipsilaterally and contralaterally (Leiner et al., In conclusion it is interesting to note the congruence of 1995). There has been an increasing number of reports drug-induced changes in rCBF with neuroanatomical showing the involvement of the cerebellum in memory, regions subserving functions affected by these drugs. attention and language processes (Schmahmann, 1996). Dose-related changes in rCBF associated with a CNS- Signi®cant decreases in rCBF were also identi®ed in the active drug may more clearly help identify the speci®c inferior frontal gyrus (BA 11}47). This region mediates neuroanatomical locations of interest, even when the anxious behaviour (Fredrikson et al., 1997 ; Rauch et al., subjects are tested in a ` resting ' state. However, the fact 1997), and has efferent connections to the hypothalamus that midazolam induces changes in rCBF in regions and the limbic system involved in visceral function and previously demonstrated to be involved with memory emotional and instinctive behaviours (Fuster, 1997). The processing does not necessarily mean that this is the orbito-frontal cortex is postulated to mediate learned mechanism of midazolam's amnesic actions. Further study associations between stimuli and reinforcements (Rolls, is required to support or refute this hypothesis, and in the 1996), and speci®cally avoidance behaviour (Heilman, case of midazolam the interaction of the drug with 1997), functions which may underlie the observed changes concurrent memory processes needs to be studied, as has in cerebral activation in anxiety disorders. It seems been attempted using scopolamine (Grasby et al., 1995). reasonable to assume that rCBF changes in this region Before drugs with speci®c effects on memory can be used may correlate with midazolam's anxiolytic effects. as tools in the investigation of normal memory function Several methodological factors in our study deserve using functional neuroimaging techniques, the direct special consideration and may in¯uence the generality of interaction of memory processes with drug administration the results. A possible doubt in this study is that 1 subject must be clari®ed. was included who was ambidextrous, and not purely right-handed. Although stimuli were not presented in Acknowledgements lateralized fashion, there is a possibility that the laterality of the ®ndings may have been affected by the inclusion of We are grateful to Dr Roger Wilson, Dr Steven Larson, Dr this ambidextrous individual. If anything, there would be Ronald Blasberg and Dr Homer Macapinlac of Memorial a tendency to have less lateralization of ®ndings, but in a Sloan-Kettering Cancer Center for their generous advice group of 14 subjects, 1 subject with a differing response is and support of this research. In addition, we thank Dr 126 R. A. Reinsel et al. Osama Mawlawi and Mr Martin Duff for their assistance Fuster JM (1997). The Prefrontal Cortex : Anatomy, Physiology and Neuropsychology of the Frontal Lobe (3rd edn). in the collection and analysis of PET data. Philadelphia : Lippincott±Raven. Ghoneim MM, Mewaldt SP (1990). Benzodiazepines and References human memory : a review. Anesthesiology 72, 926±938. Grady CL, McIntosh AR, Horwitz B, Maisog JM, Ungerleider Baddeley A (1995). Working memory. In : Gazzaniga MS LG, Mentis MJ, Pietrini P, Schapiro MB, Haxby JV (1995). (Ed.), The Cognitive Neurosciences (pp. 755±764). Cambridge, Age-related reductions in human recognition memory due MA. : MIT Press. to impaired encoding. Science 269, 218±221. Barker WW, Yoshii F, Loewenstein DA, Chang JY, Apicella Grasby PM, Friston KJ, Bench C, Cowen PJ, Frith CD, Liddle A, Pascal S, Boothe TE, Ginsberg MD, Duara R (1991). PF, Frackowiak RS, Dolan RJ (1992). Effect of the 5-HT1A Cerebrocerebellar relationship during behavioral activation : partial agonist buspirone on regional cerebral blood ¯ow in a PET study. Journal of Cerebral Blood Flow and Metabolism man. Psychopharmacology 108, 380±386. 11, 48±54. Grasby PM, Friston KJ, Bench CJ, Cowen PJ, Frith CD, Liddle Brewer JB, Zhao Z, Desmond JE, Glover GH, Gabrieli JD PF, Frackowiak RS, Dolan RJ (1993). The effect of the (1998). Making memories : brain activity that predicts how dopamine agonist, apomorphine, on regional cerebral blood well visual experience will be remembered. Science 281, ¯ow in normal volunteers. Psychological Medicine 23, 1185±1187. 605±612. Curran HV, Pooviboonsuk P, Dalton JA, Lader MH (1998). Grasby PM, Frith CD, Paulesu E, Friston KJ, Frackowiak RS, Differentiating the effects of centrally acting drugs on Dolan RJ (1995). The effect of the muscarinic antagonist arousal and memory : an event-related potential study of scopolamine on regional cerebral blood ¯ow during the scopolamine, lorazepam and diphenhydramine. performance of a memory task. Experimental Brain Research Psychopharmacology (Berlin) 135, 27±36. 104, 337±348. D'Esposito M, Ballard D, Aguirre GK, Zarahn E (1998). Greenblatt DJ, Ehrenberg BL, Gunderman J, Locniskar A, Human prefrontal cortex is not speci®c for working Scavone JM, Harmatz JS, Shader RI (1989). Pharmacokinetic memory : A functional MRI study. NeuroImage 8, 274±282. and electropharmacographic study of intravenous Duara R, Gross-Glenn K, Barker WW, Chang JY, Apicella A, diazepam, midazolam and placebo. Clinical Pharmacology Loewenstein D, Boothe T (1987). Behavioral activation and and Therapeutics 45, 356±365. the variability of cerebral glucose metabolic measurements. Gyulai FE, Firestone LL, Mintun MA, Winter PM (1996). In Journal of Cerebral Blood Flow and Metabolism 7, 266±271. vivo imaging of human limbic responses to nitrous oxide Feshchenko VA, Veselis RA, Reinsel RA (1997). Comparison inhalation. Anesthesia and Analgesia 83, 291±298. of the EEG effects of midazolam, thiopental, and propofol : Heilman KM (1997). The neurobiology of emotional the role of underlying oscillatory systems. experience. Journal of Neuropsychiatry and Clinical Neuropsychobiology 35, 211±220. Neuroscience 9, 439±448. Firestone LL, Gyulai F, Mintun M, Adler LJ, Urso K, Winter Jones AK, Friston KJ, Qi LY, Harris M, Cunningham VJ, Jones PM (1996). Human brain activity response to fentanyl T, Feinman C, Frackowiak RS (1991). Sites of action of imaged by positron emission tomography. Anesthesia and morphine in the brain [letter]. Lancet 338, 825. Analgesia 82, 1247±1251. Junck L, Gilman S, Rothley JR, Betley AT, Koeppe RA, Fredrikson M, Fischer H, Wik G (1997). Cerebral blood ¯ow Hichwa RD (1988). A relationship between metabolism in during anxiety provocation. Journal of Clinical Psychiatry 58, frontal lobes and cerebellum in normal subjects studied 16±21. with PET. Journal of Cerebral Blood Flow and Metabolism 8, Friston KJ (1994). Statistical parametric mapping. In : Thatcher 774±782. RW, Hallett M, Zeffiro T, John ER, Huerta M (Eds), Kapur S, Craik FI, Tulving E, Wilson AA, Houle S, Brown Functional Neuroimaging : Technical Foundations (pp. 79±93). GM (1994). Neuroanatomical correlates of encoding in San Diego : Academic Press. episodic memory : levels of processing effect. Proceedings of Friston KJ, Frith CD, Liddle PF, Dolan RJ, Lammertsma AA, the National Academy of Sciences USA 91, 2008±2011. Frackowiak RS (1990). The relationship between global and Leiner HC, Leiner AL, Dow RS (1995). The underestimated local changes in PET scans. Journal of Cerebral Blood Flow cerebellum. Human Brain Mapping 2, 244±254. And Metabolism 10, 458±466. Lister RG (1985). The amnesic action of benzodiazepines in Friston KJ, Grasby PM, Bench CJ, Frith CD, Cowen PJ, Liddle man. Neuroscience and Biobehavioral Reviews 9, 87±94. PF, Frackowiak RS, Dolan R (1992). Measuring the neuromodulatory effects of drugs in man with positron Mathew RJ, Wilson WH (1991). Evaluation of the effects of emission tomography. Neuroscience Letters 141, 106±110. diazepam and an experimental anti-anxiety drug on Friston KJ, Grasby PM, Frith CD, Bench CJ, Dolan RJ, Cowen regional cerebral blood ¯ow. Psychiatry Research 40, PJ, Liddle PF, Frackowiak RS (1991). The neurotransmitter 125±134. Matthew E, Andreason P, Carson RE, Herscovitch P, basis of cognition : psychopharmacological activation Pettigrew K, Cohen R, King C, Johanson CE, Paul SM studies using positron emission tomography. Ciba Foundation Symposium 163, 76±87 (discussion 87±92). (1993). Reproducibility of resting cerebral blood ¯ow Midazolam decreases rCBF in prefrontal cortex 127 measurements with H (15)O positron emission Schmahmann JD (1996). From movement to thought : tomography in humans. Journal of Cerebral Blood Flow and Anatomic substrates of the cerebellar contribution to Metabolism 13, 748±754. cognitive processing. Human Brain Mapping 4, 174±198. Squire LR, Knowlton B, Musen G (1993). The structure and Nyberg L, McIntosh AR, Cabeza R, Habib R, Houle S, organization of memory. Annual Review of Psychology 44, Tulving E (1996). General and speci®c brain regions 453±495. involved in encoding and retrieval of events : what, where Talairach J, Tournoux P (1988). Co-Planar Stereotaxic Atlas of and when. Proceedings of the National Academy of Sciences the Human Brain. Stuttgart : Georg Thieme Verlag. USA 93, 11280±11285. Tanelian Dl, Kosek P, Mody I, MacIver MB (1993). The role Parekh PI, Spencer JW, George MS, Gill DS, Ketter TA, of the GABA-A receptor}chloride channel complex in Andreason P, Herscovitch P, Post RM (1995). Procaine- anesthesia. Anesthesiology 78, 757±776. induced increases in limbic rCBF correlate positively with Tulving E, Kapur S, Craik FI, Moscovitch M, Houle S (1994a). increases in occipital and temporal EEG fast activity. Brain Hemispheric encoding}retrieval asymmetry in episodic Topography 7, 209±216. memory : positron emission tomography ®ndings. Paulesu E, Frith CD, Frackowiak RS (1993). The neural Proceedings of the National Academy of Sciences USA 91, correlates of the verbal component of working memory. 2016±2020. Nature 362, 342±345. Tulving E, Kapur S, Markowitsch HJ, Craik FI, Habib R, Poldrack RA, Gabrieli JDE (1997). Functional anatomy of Houle S (1994b). Neuroanatomical correlates of retrieval in long-term memory. Journal of Clinical Neurophysiology 14, episodic memory : auditory sentence recognition. 294±310. Proceedings of the National Academy of Sciences USA 91, Ramsay SC, Murphy K, Shea SA, Friston KJ, Lammertsma 2012±2015. AA, Clark JC, Adams L, Guz A, Frackowiak RS (1993). Veselis RA, Glass P, Dnistrian A, Reinsel R (1997a). Changes in global cerebral blood ¯ow in humans : effect on Performance of computer-assisted continuous infusion at regional cerebral blood ¯ow during a neural activation task. low concentrations of intravenous sedatives. Anesthesia and Journal of Physiology 471, 521±534. Analgesia 84, 1049±1057. Veselis RA, Reinsel RA, Beattie BJ, Mawlawi OR, Feshchenko Rauch SL, Savage CR, Alpert NM, Fischman AJ, Jenike MA VA, DiResta GR, Larson SM, Blasberg RG (1997b). (1997). The functional neuroanatomy of anxiety : a study of Midazolam changes cerebral blood ¯ow in discrete brain three disorders using positron emission tomography and regions : an H (15)O positron emission tomography study. symptom provocation. Biological Psychiatry 42, 446±452. Anesthesiology 87, 1106±1117. Reinsel R, Veselis R, Wronski M, Marino P, Heino R, Veselis RA, Reinsel RA, Feshchenko VA, Di Resta G, Alagesan R (1993). Memory impairment during conscious Mawlawi O, Beattie B, Silbersweig D, Stern E, Blasberg R, sedation : A comparison of midazolam, propofol and Macapinlac H, Finn R, Goldsmith S, Larson S (1995). thiopental. In : Sebel PS, Bonke B, Winograd E (Eds.), Cerebral blood ¯ow changes during midazolam sedation Memory and Awareness in Anesthesia (pp. 127±140). using O-15 positron emission tomography (PET) [Abstract]. Englewood, N.J. : Prentice-Hall. Anesthesiology 83, A154. Reinsel RA, Veselis RA, Feshchenko VA, Di Resta GR, Veselis RA, Reinsel RA, Feshchenko VA, Wronski M (1997c). Mawlawi O, Beattie B, Silbersweig D, Stern E, Blasberg R, The comparative amnestic effects of midazolam, propofol, Macapinlac H, Finn R, Goldsmith S, Larson S (1995). thiopental, and fentanyl at equisedative concentrations. Target detection and the prefrontal cortex. A PET scan Anesthesiology 87, 749±764. study of the P300 event-related potential. Annals of the Volkow ND, Wang GJ, Hitzemann R, Fowler JS, Pappas N, New York Academy of Sciences 769, 393±397. Lowrimore P, Burr G, Pascani K, Overall J, Wolf AP (1995). Rolls ET (1996). The orbitofrontal cortex. Philosophical Depression of thalamic metabolism by lorazepam is Transactions of the Royal Society of London (B : Biological associated with sleepiness. Neuropsychopharmacology 12, Sciences) 351, 1433±1443. 123±132. Roy-Byrne P, Fleishaker J, Arnett C, Dubach M, Stewart J, Wagner AD, Schacter DL, Rotte M, Koutstaal W, Maril A, Radant A, Veith R, Graham M (1993). Effects of acute and Dale AM, Rosen BR, Buckner RL (1998). Building chronic alprazolam treatment on cerebral blood ¯ow, memories : remembering and forgetting of verbal memory, sedation, and plasma catecholamines. experiences as predicted by brain activity. Science 281, Neuropsychopharmacology 8, 161±169. 1188±1191.

Journal

International Journal of NeuropsychopharmacologyOxford University Press

Published: Jun 1, 2000

There are no references for this article.