Cannabidiol (CBD) constitutes up to 40% of Cannabis sativa (Grlie, 1976) and has quite different psychological effects to the plant's best known constituent, Δ9-tetrahydrocannabinol (Δ9-THC) (Perez-Reyes et al, 1973; Zuardi et al, 1982). In particular, in animal studies CBD has effects similar to anxiolytic drugs in conditioned emotional paradigms (Zuardi and Karniol, 1983), the Vogel conflict test (Musty et al, 1984), and the elevated plus maze test (Guimaraes et al, 1990; Onaivi et al, 1990). Using the latter test, anxiolytic effects were also reported for three derivates of CBD, HU-219, HU-252, and HU-291 (Guimaraes et al, 1994). In humans, oral administration of CBD in healthy volunteers attenuates the anxiogenic effect of Δ9-tetrahydrocannabinol (Δ9-THC) (Zuardi et al, 1982). This effect does not seem to involve any pharmacokinetic interactions (Agurell et al, 1981; Zuardi et al, 1982), and CBD does not bind to the central known cannabinoid receptor, CB1, (Bisogno et al, 2001; Mechoulam et al, 2002) and hence cannot be a competitive antagonist (Howlett et al, 1992). CBD may thus possess inherent anxiolytic properties unrelated to THC-type activity. This is consistent with its anxiolytic effect on anxiety elicited by simulated public speaking (Zuardi et al, 1993a).

As the receptors that mediate the psychological effects of CBD are unknown, its mechanism of action on the brain is unclear. The aim of the present study was to use functional neuroimaging to investigate this. In view of its anxiolytic effect, we tested the hypothesis that CBD would affect neural activity in areas that normally mediate anxiety. We compared the effects of CBD and placebo on resting cerebral regional blood flow (rCBF) in healthy volunteers in a double-blind, cross-over design. Based on previous functional imaging studies of anxiety (Maddock and Buonocore, 1997; Fischer et al, 1996; Liotti et al, 2000; Ketter et al, 1996), we predicted that, relative to placebo, CBD would modulate rCBF in limbic and paralimbic areas: the orbitofrontal, cingulate and medial temporal cortex, and the insula.



A total of 10 healthy male postgraduate students were studied. None had undergone rCBF SPECT examinations or other nuclear medicine procedures before. All were right-handed as assessed by the Edinburgh Handedness Inventory (Oldfield, 1971), and were nonsmokers (of tobacco). Their mean age was 29.8 years (range 25–42 years, SD=5.1), their mean weight was 74.1 kg (67–85 kg, SD=6.05), and their body mass index ranged between 21 and 25 kg/m2. The subjects had not taken any medicines for at least 3 months before the study (Mathew et al, 1992). No subject had a history of head trauma, neurological, or major medical illnesses, based on a semistandardized medical questionnaire and physical examination. Neither the subjects (based on the Structured Clinical Interview for DSM-IV, First et al, 1997) nor their first-degree relatives (based on subjects' report) had a history of psychiatric illness. No subject had used marijuana more than five times in their lives (nor in the last year), and none had used any other illegal drug. The experiment was conducted with the understanding and consent of each subject, following approval by the local ethical committee.


CBD in powder, approximately 99.9% pure (supplied by THC-Pharm, Frankfurt, Germany), was dissolved in corn oil (Zuardi et al, 1993a, 1995). The same amount of corn oil was used as a placebo. The drug and placebo were packed inside identical gelatin capsules.

Self-Rating Scale

Subjective states were evaluated by means of the Visual Analogue Mood Scale (VAMS) of Norris (1971), translated into Portuguese by Zuardi and Karniol (1981b). It consists of 16 analogue scales to measure drug effects, which were arbitrarily divided by Norris into four factors: anxiety, physical sedation, mental sedation, and other feelings and attitudes. A factor analysis with the Portuguese version of this scale (Zuardi et al, 1993a) extracted four factors that can be identified with those of the Norris proposal. Prior to the experiment, each volunteer had performed a training session completing this scale.


Subjects were told not to consume any alcohol for 24 h and caffeine for at least 4 h before each visit to the laboratory (Mathew et al, 1999). Subjects who reported having less than 6 h of sleep the previous night were excluded. After at least 8 h of fasting, subjects were instructed to have a light, standardized breakfast 2 h before the experiment. They were randomly divided into two groups of five subjects. Each subject was evaluated on two different occasions, 1 week apart. In the first session, after a 30-min period of adaptation, subjects were given a single dose of oral CBD (400 mg) or placebo, in a double-blind procedure. The sessions were held in the morning (between 0800 and 1200) to minimize the effects of circadian variation. SPECT image acquisition was performed 110 min after drug ingestion. Subjective ratings on the VAMS were made 30 min before drug ingestion, at the time of drug ingestion, and at 60 min and at 75 min afterwards. In the second session, an identical procedure was followed except that the other drug was administered (ie those given CBD in the first session received placebo in the second; and vice versa). Subjects were informed that they would receive CBD and placebo, but they were not told in which order. The investigators were also blind to the content of the capsules.


Subjects had a venous cannula inserted into their right arm, and rested supine with minimal environmental sensory stimulation. They were instructed to keep their eyes closed under eye pads and to relax for 15 min without falling asleep. Their ears were unplugged. VAMS ratings were made just before and 15 min after insertion of the venous cannula. At 30 min after insertion of the venous cannula, 740 MBq (20 mCi) of ethyl-cisteinate dimer (ECD) labeled with technetium-99 m (99mTc-ECD) was injected. Subjects rested for an additional period of 5 min postinjection, after which the venous cannula was removed.

Image acquisition started 20 min after the 99mTc-ECD injection, using a double-detector SOPHA® DST system (Sophy Medical Vision, Twinsburg, USA). High-resolution low-energy collimators were used, with 128 views acquired on a 128 × 128 matrix (30 s per view), with a total acquisition time of 30 min, and approximately 75 000 counts/frame/head. Raw images were prefiltered with a Butterworth filter (order number 4, cutoff frequency 0.16), and reconstructed by filtered back-projection as transaxial slices parallel to the long axis of the temporal lobe. Attenuation correction was performed considering a pixel size of 2.55 mm and using the first-order algorithm of Chang (coefficient 0.12/cm).

Image Processing and Analysis

Images were analyzed using Statistical Parametric Mapping software (SPM99) (Friston et al, 1995). Reconstructed transaxial datasets were transferred to a PC (Pentium IV, 2.2 GHz, 512 Mb RAM), converted to Analyze format and reoriented to neurological convention (ie left=left). Placebo images were realigned to CBD images using sinc interpolation. Linear (translations and rotations) and nonlinear (7 × 8 × 7 nonlinear basis functions) deformations were used to register images to the SPM SPECT template, which is based on the Montreal Neurological Institute (MNI) template (Mazziotta et al, 1995). Finally, an isotropic Gaussian filter of 12 mm was applied to diminish interindividual differences, and to conform data to the theory of Gaussian Random Fields (Friston et al, 1995), in order to allow the subsequent application of parametric statistical tests.

Between-condition (CBD vs placebo) comparisons of regional tracer uptake were performed on a voxel-by-voxel basis using paired t-tests. Before statistical testing, the regional ECD uptake of every voxel in each subject was standardized to the mean global uptake of the image in that subject, using proportional scaling. Only voxels with signal intensities above a threshold of 0.8 of the global mean (calculated using the standardized values) entered the statistical analysis. The resulting statistics at each voxel were transformed to Z-scores, thresholded at Z=2.33 (corresponding to p<0.01, one-tailed), and displayed as 3-D statistical parametric maps (SPM). These maps were first inspected for the presence of voxel clusters of significant difference in the regions where effects of CBD had been predicted a priori (medial temporal, cingulate, orbitofrontal, and insular cortices). Clusters in these regions were considered as significant if they included voxels with Z-scores of 3.09 or greater (corresponding to one-tailed p<0.001), and contained more than 20 voxels. Levels of p<0.001, uncorrected for multiple comparisons, have been frequently used in previous SPM analyses of positron emission tomography (PET) (Dougherty et al, 1999; Bremner et al, 1999b) and SPECT (Blackwood et al, 1999; Busatto et al, 2000) data, and are considered to provide good protection against false-positive results when there are clear hypotheses as to the location of findings. The SPMs were also inspected for differences in other, unpredicted regions. These areas were reported as significant if they survived a correction for multiple comparisons based on Gaussian random field theory (p<0.05) (Friston et al, 1995).

For each voxel cluster showing significant between-condition differences, estimates were calculated for the mean, median, and maximal percentages of ECD count rate change (and their variances) (Table 1). These indices were obtained by partitioning the Student's t-test value of each voxel into its main components, with the numerator of the t statistic used as an approximation of the magnitude of the signal change for each contrast (placebo>CBD or CBD>placebo), and the denominator (the standard error) used to calculate the variances. The MNI coordinates for the voxels of maximal statistical significance for each anatomical brain region included in a given cluster were converted to the Talairach and Tournoux (1988) system using the method described by Brett et al (2002).

Table 1 Limbic and Paralimbic Areas of Significant rCBF Differences in CBD Compared to Placebo Condition

The four VAMS factors were submitted to an ANOVA for repeated measures in both CBD and placebo sessions. The differences between CBD and placebo in each phase of the experimental session (−30, 0, 1′00, 1′15) were analyzed by t-tests. Correlations between the regional tracer uptake and each of the VAMS factors scores were also investigated with SPM99, at the same statistical significance levels as described above for the between-condition rCBF comparisons. The last point in which the VAMS was applied (75′ after drug intake) was chosen for these correlations due to its proximity to the injection of the SPECT tracer. Moreover, this is the point where CBD is expected to have its maximum anxiolytic effect among all the time points chosen for assessment during the experimental session (Zuardi et al, 1993a). The choice for this time point was also based on previous studies, which have shown that the plasma peak of an oral dose of CBD usually occurs between 1 and 2 h after ingestion (Agurell et al, 1981).


Visual Analogue Mood Scale

The administration of CBD was associated with significantly decreased subjective anxiety (F(3,27)=18.56, p<0.001) and increased mental sedation (F(3,27)=42.85, p<0.001), while placebo was not (F(3,27)=1.86, p=0.16 and F(3,27)=2.24, p=0.11, respectively) (Figure 1). In addition, an analysis at each time point indicated the following: (i) CBD was associated with significantly decreased anxiety at cannula insertion (60′ after drug intake, t=2.95, p=0.009) and resting phases (75′ after drug intake, t=5.50, p<0.001) as compared to placebo; (ii) CBD was associated with significantly increased feelings of mental sedation at cannula insertion (60′ after drug intake, t=−3.91, p=0.001) and resting phases (75′ after drug intake, t=−3.67, p=0.002) as compared to placebo.

Figure 1
figure 1

Effect of CBD and placebo (PLCB) on the four factors of the VAMS. Points are means (±SEM) of 10 healthy subjects in the following phases of the experiment: predrug (−30), drug intake (0), prestress (60), and adaptation (75). Asterisk (*) indicates significant difference from placebo in each phase. Triangle () indicates ANOVA significant changes.

Between-Condition rCBF Comparisons

The SPM showing increases in ECD uptake in the CBD relative to placebo condition revealed only one cluster (>20 voxels) that surpassed the initial Z=2.33 statistical cutoff (Figure 2). This cluster, which achieved statistical significance at the p<0.001 level (uncorrected for multiple comparisons), was located in the medial temporal cortex, where the effects of CBD had been predicted a priori, and involved the left parahippocampal gyrus, extending inferiorly to encompass the left fusiform gyrus (Table 1).

Figure 2
figure 2

The brain region where there was significantly increased rCBF in healthy volunteers (n=10) during CBD vs placebo has been overlaid on coronal sections (−18, −16, −14, −12) and sagittal sections (−28, −30, −32, −34) of a reference brain, imaged with structural MRI and spatially normalized into an approximation to the Tailarach and Tornoux (1988) stereotactic atlas. The results are displayed in neurological convention (ie left=left). The numbers associated with each frame represent the standard coordinate in the y- (for the coronal frames) and x-axis (for the sagittal frames). The voxel cluster shown was located in the left parahippocampal gyrus extending inferiorly to encompass the left fusiform gyrus (peak Z-score=3.69, coordinatesx,y,z=−30, −14, −30; p<0.0001 uncorrected for multiple comparisons; 114 voxels).

Significantly decreased (p<0.001, uncorrected for multiple comparisons), ECD uptake in the CBD relative to the placebo condition was evident in two regions where effects of CBD had been predicted a priori (Table 1). One cluster included the medial portion of the left amygdala–hippocampal complex and uncus, as well as the hypothalamus. The other was located in the superior portion of the left posterior cingulate gyrus (Brodmann area—BA31), extending towards the paracentral lobule (BA5/6). At the p<0.001 uncorrected level of significance, this SPM showed additional, unpredicted foci of decreased rCBF in the ECD relative to the placebo condition (>20 voxels) in the right cerebellum, medial occipital cortex, left inferior temporal, and posterior lateral frontal cortex, but none of these retained significance after correction for multiple comparisons.

Figure 3 displays, for each subject, the magnitude of tracer uptake changes between the CBD and placebo conditions at the voxel of maximal statistical significance in the regions where ECD uptake differences were observed (as summarized in Table 1). All 10 subjects showed greater ECD uptake values in the CBD condition relative to the placebo condition in the left parahippocampal/fusiform gyri. Of the 10 subjects, eight showed lower ECD activity in the CBD condition relative to the placebo condition in the left amygdala–hippocampal complex; eight in the left hypothalamus; and nine in the left posterior cingulate cortex/paracentral lobule (Figure 2). Similar patterns across individual subjects were observed when we used the mean tracer uptake values of all voxels included in the clusters of significant difference between the CBD and placebo conditions (data not shown).

Figure 3
figure 3

Tracer uptake values during the CBD (filled circle) and PLCB (hollowed circle) conditions are plotted for the 10 subjects, using the voxel of maximal significant difference of each of the four regions reported in Table 1. From left to right and top to bottom: −4, −27, 47 (left posterior cingulate cortex/paracentral lobule), −6, −6, −8 (left hypothalamus), −16, −11, −21 (left amygdala–hippocampal complex), and −30, −15, −24 (left parahippocampal/fusiform gyri). Individual values were normalized to the global ECD uptake for each subject and condition. The graphs show that the large majority of individual subjects showed lower ECD activity in the CBD condition relative to the placebo condition in the left amygdala–hippocampus complex, hypothalamus, and posterior cingulate cortex/paracentral lobule, while all subjects had greater ECD uptake in the CBD condition relative to the placebo condition in the left parahippocampal/fusiform gyri.

Correlations with Subjective Status Ratings

No correlations were observed between subjective anxiety ratings and ECD uptake in the brain areas where the effects of CBD had been predicted a priori (p<0.001, uncorrected), or in other unpredicted areas after correction for multiple comparisons.


When undergoing neuroimaging procedures, such as PET or SPECT, subjects often report increased anxiety before scanning, which is greater than that during or after image acquisition (Grey et al, 2000; Gur et al, 1987; Giordani et al, 1990; Malizia, 1999). The results of the present study showed that a single dose of CBD induced significant decreases in state anxiety before SPECT scanning. Our data thus suggest that this compound has anxiolytic properties, consistent with the results from previous studies in both laboratory animals (Zuardi and Karniol, 1983; Musty et al, 1984; Guimaraes et al, 1990; Onaivi et al, 1990) and humans (Zuardi et al, 1982, 1993a).

The anxiolytic effects found in the present study were detected before the anxiety-evoking situation (the tracer injection and scanning procedure), indicating that CBD can affect anticipatory anxiety. In a previous study (Zuardi et al, 1993a), the anxiolytic effect of CBD was evident after the stress of public speaking. These antianxiety effects are in contrast to the anxiogenic effects of high doses of Δ9-THC (Malit et al, 1975; Zuardi et al, 1982; Mathew et al, 1999), and may help to reconcile apparently conflicting findings obtained with Cannabis sativa in relation to anxiety (Johns, 2001; Tournier et al, 2003).

Consistent with an anxiolytic effect, we found that CBD significantly modulated resting activity predominantly in limbic and paralimbic cortical areas, which are usually implicated in the pathophysiology of anxiety (Gray, 1982; Graeff, 1994). Thus, between-condition activity differences were detected in a left medial temporal cluster, which included portions of the amygdala and the hippocampus, as well as the hypothalamus, the left posterior cingulate gyrus, and the left parahippocampal gyrus.

The only brain region that showed significantly increased activity in the CBD relative to the placebo condition was the left parahippocampal gyrus. Deactivation of the parahippocampal region in healthy volunteers has been reported after panic attacks induced by lactate (Reiman et al, 1989) and CCK-4 (Javanmard et al, 1999), and with anxiety induced by presentation of combat-related images (Bremner et al, 1999b) and autobiographical memory scripts (Liotti et al, 2000). In addition, the abnormal asymmetry of resting activity in the parahippocampal gyri has been associated with panic disorder and with vulnerability to lactate-induced panic (Reiman et al, 1984, 1986; Nordahl et al, 1990, 1998; Bisaga et al, 1998; De Cristofaro et al, 1993). These studies suggest that anxiety can be associated with reduced parahippocampal activity, consistent with an anxiolytic effect of CBD and the increased activity in this region that we observed.

In contrast, activity in the left amygdala–hippocampal complex, hypothalamus, and posterior cingulate cortex decreased with CBD relative to placebo. The amygdala is thought to play a key role in mediating fear and anxiety (Deakin and Graeff, 1991; LeDoux, 1998; Gorman et al, 2000), being activated during fear conditioning (Furmark et al, 1997; Morris et al, 1998; LaBar et al, 1998; Buchel et al, 1998), while processing anxious faces (Breiter et al, 1996; Morris et al, 1996; Whalen et al, 1998; Hariri et al, 2002) and during pharmacologically induced anxiety (Ketter et al, 1996; Benkelfat et al, 1995; Servan-Schreiber et al, 1998). Functional and structural changes in the amygdala have also been reported in PTSD (Pitman et al, 2001; Rauch et al, 1996; Shin et al, 1997; Rauch et al, 2000; Liberzon et al, 1999), panic disorder (Uchida et al, 2003; Bystritsky et al, 2001), generalized anxiety disorder (Thomas et al, 2001; De Bellis et al, 2000), and in social (Birbaumer et al, 1998; Tillfors et al, 2001; Furmark et al, 2002) and simple phobias (Wik et al, 1997). The reduction in amygdala activity that we observed with CBD is thus consistent with the anxiolytic effect that it had in our subjects. The hippocampus has also been implicated in the processing of anxiety. Functional neuroimaging studies have shown increased activity in the hippocampus in association with anxiety in OCD (McGuire et al, 1994), panic disorder (Bisaga et al, 1998; Bystritsky et al, 2001; Boshuisen et al, 2002), PTSD (Osuch et al, 2001), and in social phobia (Schneider et al, 1999). However, other studies have reported either decreased or no difference in activity in the hippocampus in association with normal anxiety or anxiety disorders (Schuff et al, 2001; Schneider et al, 1999; Bremner et al, 1997; Fredrikson et al, 1997; Fischer et al, 1996; Paradiso et al, 1997; Liotti et al, 2000).

The hypothalamus is a major component of the central autonomic nervous system, and is often involved in mediating the effects of stress and anxiety (Afifi and Bergman, 1998). Functional imaging studies during fear and anxiety induction in healthy subjects (Fredrikson et al, 1995b; Javanmard et al, 1999) and in panic disorder patients (Boshuisen et al, 2002) have reported increases in the activity of the hypothalamic region, and hypothalamic–pituitary–adrenal axis abnormalities have been commonly reported in anxiety disorders (Hageman et al, 2001). The reduced hypothalamic activity that we observed is thus consistent with the anxiolytic effect of CBD.

The posterior cingulate cortex is strongly linked to temporolimbic structures (Vogt et al, 1992; Maddock, 1999; Afifi and Bergman, 1998), and is thought to play a central role in emotion and anxiety (MacLean, 1993; Maddock, 1999). Increased activity in the posterior cingulate gyrus has been associated with watching anxiety-provoking videos (Fischer et al, 1996, Fredrikson et al, 1995a), and with experimentally provoked obsessions and anxiety in patients with obsessive–compulsive disorder (OCD) (McGuire et al, 1994). Untreated patients with OCD show increased metabolism in the posterior cingulate (Perani et al, 1995) that decreases with treatment, with the change in posterior cingulate rCBF correlated with symptomatic improvement (Rauch et al, 2001, 2002). There have also been reports of increased posterior cingulate activation during symptom provocation in post-traumatic stress disorder (Bremner et al, 1999a) and panic disorder (Bystritsky et al, 2001). However, anxiety induction in phobic patients has been associated with deactivation in the posterior cingulate region (Wik et al, 1993) and Busatto et al (2000) reported a negative correlation between rCBF in the left posterior cingulate cortex and severity of symptoms in OCD.

We did not observe a correlation between the severity of anxiety and rCBF in the areas where activity was modulated by CBD, but this may have been difficult to detect because there was a 15-min gap between the points when the ratings were made and the SPECT tracer was injected.

While the areas where we found modulatory effects of CBD are thus implicated in mediating anxiety, and have also been associated with the anxiolytic effects of diazepam (Di Piero et al, 2001), citalopram (Van der Linden et al, 2000; Furmark et al, 2002), sertraline, and desipramine (Hoehn-Saric et al, 2001), these effects of CBD could be related to an effect other than on anxiety. For instance, we also observed sedative effects of CBD, confirming former findings in animals (Pickens, 1981; Monti, 1977; Colasanti et al, 1984; Zuardi et al, 1981a, 1991) and humans (Carlini et al, 1979; Zuardi et al, 1982, 1993b). This effect has been reported to be dose-related (Pickens, 1981) and CBD has also been shown to decrease wakefulness (Monti, 1977) and to cause longer sleep duration in insomniacs (Carlini and Cunha, 1981). Thus, the reduced hypothalamic activity observed after CBD use in our study could equally be related to sedative effects of CBD, as suggested to occur with other sedative compounds (Tung et al, 2001).

Other pharmacological effects of CBD have been reported in studies in laboratory animals and humans, such as anti-inflammatory (Malfait et al, 2000), anticonvulsant (Carlini et al, 1973; Izquierdo et al, 1973; Cunha et al, 1980), neuroprotective (Hampson et al, 1998), and hormonal effects (Zuardi et al, 1984, 1993a). In addition, the pharmacological profile of CBD is similar to that of clozapine, an ‘atypical’ antipsychotic drug (Zuardi et al, 1991, 1995), and both CBD and clozapine induce c-fos expression in the prefrontal cortex and lateral septal nucleus in rats (Zuardi et al, 2001). The mechanism(s) of action whereby CBD produces all these effects remains obscure. This is largely in contrast with the effects of Δ9-THC, which mimics the endogenous cannabinoids in many of its actions. CBD does not act through the known cannabinoid receptors, but the stereospecificity previously observed may indicate that CBD binds to another type of receptor in the brain (Mechoulam et al, 2002).

In conclusion, our results suggest that CBD has anxiolytic effects that are mediated through an action on limbic and paralimbic areas of the brain. However, the findings need to be seen as preliminary, given the limitations of the study. Firstly, it would have been desirable to measure plasma levels of CBD and relate them to the magnitude of change in rCBF. Without a dose–response curve, uncertainty about the regional cerebral effects of CBD remains. Nevertheless, it should be pointed out that it is not clear whether there is a relation between plasma levels of cannabinoids—especially CBD—and their clinical effects (Agurell et al, 1986). In addition, the subject sample was modest and the use of SPECT limited the study's statistical power. Finally, given the limited spatial resolution of the SPECT technique and the smoothing procedure, the interpretation of large foci of tracer uptake changes as involving different brain structures of small size (such as the amygdala, hippocampus, and hypothalamus) should be made with caution. These limitations could be overcome by examining a larger sample and using functional magnetic resonance imaging, which would permit the acquisition of greater numbers of images with a better spatial and temporal resolution.