The impact of cannabidiol placebo on amygdala-based neural responses to an acute stressor

Introduction

Cannabidiol (CBD), a phytocannabinoid derived from the Cannabis sativa plant, has gained considerable interest for its therapeutic potential (Corroon and Phillips, 2018; Kirkland et al., 2022). CBD is believed to exhibit a wide range of therapeutic effects including but not limited to immunomodulatory, antiemetic, anti-depressant, anti-convulsant, and anxiolytic effects (Bergamaschi et al., 2011; Campos et al., 2016; Martinez Naya et al., 2023). CBD has been shown to have a high margin of safety across a range of doses (Blessing et al., 2015; Kwee et al., 2022b; Madeo et al., 2023) with low abuse potential, making it a potential candidate for the treatment of psychiatric disorders (Kirkland et al., 2022). Indeed, there have been notable increases in the use of CBD for medicinal purposes reported globally (World Health Organization, 2018). Although CBD is not yet established as a pharmacological treatment for psychiatric conditions (Black et al., 2019; World Health Organization, 2018), a primary motive for non-medical CBD use appears to be its purported anxiolytic and stress-dampening effects (Geppert et al., 2023; Gournay et al., 2024). Among healthy individuals, stress and anxiety represent adaptive responses to challenges that arise during daily living, and it is estimated that on average healthy individuals will elicit a stress response once every 3 days (Stawski et al., 2013). Healthy individuals often engage in efforts to dampen stress-related effects when they occur, and there is growing evidence that CBD is commonly used for relaxation and stress relief among the general population (Geppert et al., 2023; Gournay et al., 2024).

The mechanisms of action of CBD are diverse and not yet completely understood (Martinez Naya et al., 2023); however, there is evidence that CBD may impact neural processes implicated in a variety of neuropsychiatric disorders, including anxiety- and stress-related disorders (Campos et al., 2016). Proposed pharmacological mechanisms for CBD’s ability to diminish stress and anxiety include its interactions with targets such as serotonin 5HT-1A receptors, cannabinoid CB1 receptors, and transient receptor potential vanilloid type 1 receptors (Blessing et al., 2015; García-Gutiérrez et al., 2020). Preclinical animal models have demonstrated reduced stress- and anxiety-related responses with CBD administration across a range of stress tasks (for review, see Blessing et al., 2015). Though evidence in humans is limited, there is some support for CBD’s purported anxiolytic effects in both healthy and clinical populations (García-Gutiérrez et al., 2020). An early study by Zuardi et al. (1982) reported that CBD acutely reduced the subjective anxiogenic effects of delta-9-tetrahydrocannabinol (THC) relative to placebo in eight healthy participants. In a second, placebo-controlled study, they reported that CBD reduced subjective anxiety in 10 healthy participants performing a public speaking task (Zuardi et al., 1993). A third study by this group (Bergamaschi et al., 2011) compared the effects of CBD and placebo on subjective stress- and anxiety-related responses in participants diagnosed with social anxiety disorder (SAD) completing a public speaking task. They identified that CBD (vs placebo) significantly reduced subjective ratings of anxiety, alertness, cognitive impairment, and discomfort throughout the task (Bergamaschi et al., 2011).

There is also some evidence that CBD may alter activity in brain regions implicated in anxiety reactivity and stress reactivity (e.g., amygdala, anterior cingulate cortex (ACC), anterior insula (AI)) (Batalla et al., 2021; Bhattacharyya et al., 2010; Fusar-Poli et al., 2009). Preclinical and clinical literature has demonstrated that the amygdala is critical for fear-related emotional processing, and an important mediator of stress-, anxiety-, and worry-related effects on the brain (Chattarji et al., 2015; Makovac et al., 2016; McEwen et al., 2016). Moreover, the ACC is involved in emotion and behavior regulation, as well as other fundamental cognitive processes, such as learning and decision-making (Brockett and Roesch, 2021). The AI is involved in interoceptive awareness, such as bringing information from internal states into subjective awareness. These regions have also been demonstrated to be functionally connected in stress- and anxiety-related contexts. For example, increased connectivity between the amygdala and the ACC has been associated with attentional bias to threatening stimuli in healthy controls (Jenks et al., 2020), and excessive worry states in GAD patients (Makovac et al., 2016). A strong positive correlation between rsFC of the left AI and basolateral amygdala has also been reported with state anxiety (Baur et al., 2013). Moreover, viewing one’s performance (vs viewing that of another person) during the Trier Social Stress Test (Kirschbaum et al., 1993) was associated with increased activity in the bilateral ACC and insula (Lee et al., 2014). Insofar as CBD impacts stress- and anxiety-related responses, one would expect it to impact amygdala functional connectivity with the AI and/or ACC. Indeed, CBD administration has been reported to blunt amygdala and ACC responses to anxiety-related cues and reduce rsFC between these regions at rest (Bhattacharyya et al., 2010; Fusar-Poli et al., 2010), as well as attenuate activity within and between the amygdala and ACC when viewing intensely fearful faces (Fusar-Poli et al., 2009). Together, consistent with observed behavioral effects, evidence suggests that CBD may correspondingly impact stress- and anxiety-related neural substrates.

A growing body of evidence suggests that CBD’s therapeutic effects may be significantly impacted by placebo or expectancy effects. Double-blind trials have yielded mixed findings regarding the effects of pharmacologically active CBD relative to placebo on stress and anxiety, with some results suggesting a therapeutic effect (Bergamaschi et al., 2011; Masataka, 2019), and others reporting null results (Bolsoni et al., 2022; Gournay et al., 2023a; Kwee et al., 2022a). Given that perceived drug assignment in double-blinded trials is often not directly assessed, the possibility of strong placebo effects being obscured in negative trials, or contributing to reductions in stress and anxiety in positive trials remains unclear. Accordingly, open-label trials have uniformly reported significant effects of CBD (Berger et al., 2022; Dahlgren et al., 2022; Elms et al., 2019; Gournay et al., 2023b), suggesting a potential role for drug expectancy in CBD’s purported anxiolytic and stress-reducing effects. Indeed, De Vita et al. (2022) identified that CBD expectancy instructions were associated with increased analgesia and reduced pain unpleasantness in healthy adults (De Vita et al., 2022), and our group has previously reported that CBD placebo alone is sufficient to impact subjective, physiological, and endocrine markers of stress and anxiety (Spinella et al., 2021, 2023; Zhekova et al., 2024); however, the extent to which these effects are associated with altered brain reactivity remains unknown.

To our knowledge, no prior studies have directly examined the effects of CBD expectancy on stress- and anxiety-related neural processing. Several studies have investigated the neurological basis of anxiolytic placebo effects in the context of selective serotonin reuptake inhibitors (SSRIs) used for the treatment of anxiety disorders (Huneke et al., 2022). For example, in a pair of studies by one group (Faria et al., 2012, 2014), 72 patients with SAD were randomized to receive either an SSRI or placebo under double-blind conditions, and regional cerebral blood flow (rCBF) was assessed during a public speaking task before and after 6–8 weeks of treatment. In the first study, no significant differences in rCBF of the amygdala were identified between placebo responders and those who received the active SSRI (Faria et al., 2012). In a follow-up analysis (Faria et al., 2014), both SSRI and placebo responders demonstrated decreased rCBF in the amygdala, which was accompanied by reduced activation of the dorsal anterior cingulate cortex (dACC). Taken together, these findings suggest that placebo effects may act on the same brain regions as the drug’s pharmacological effects to produce therapeutic responses. This raises the possibility that CBD-related placebo effects may also act on the same brain regions as other placebo anxiolytic agents; however, this has not been directly tested to date.

The present study used a between-subject, repeated measures design to assess the independent effects of CBD expectancy on anxiety- and stress-related neural activity in a sex-balanced sample of community-recruited healthy adults. This study is the first to examine the neural substrates of anxiolytic placebo effects in the context of CBD. Findings will help clarify the role of expectancy effects in relation to CBD’s purported stress- and anxiety-relieving properties among the general population and could lead to an improved understanding of the neurobiological mechanisms underlying CBD placebo effects.

Our primary aim was to determine the extent to which CBD expectancy effects are associated with altered amygdala rsFC with the ACC and AI in response to an acute stressor and the anticipation of a subsequent stressor. Consistent with CBD’s reported effects on the brain (Batalla et al., 2021; Fusar-Poli et al., 2009, 2010), we expected to see lower amygdala rsFC with the ACC and AI in the told CBD condition, relative to the told CBD-free condition after acute stress exposure, while anticipating a subsequent stressor. A secondary aim of this study was to replicate findings from our prior work (Spinella et al., 2021; Zhekova et al., 2024) characterizing the effects of CBD placebo on subjective responses to acute stress exposure, as well as stress-related anticipation and recovery effects. Consistent with findings from Spinella et al. (2021) and Zhekova et al. (2024), we anticipated that participants in the Told CBD condition would report increased subjective sedation following oil administration, and lower subjective ratings of stress and anxiety following oil administration, and during stress anticipation and recovery, relative to those in the Told CBD-free condition.

Method

Procedure

Once eligibility was confirmed via telephone screening, participants were invited to schedule an initial 1-h baseline session at Dalhousie University. During the baseline session, participants provided written consent and their weight was measured, which they were told would determine the dose of oil they would receive during the experimental session. As part of the consent process, participants were informed that they would be randomly assigned to receive either a CBD-containing oil or a CBD-free hempseed oil during their experimental session and that they would be told the CBD content of their assigned oil on the day of their experimental session, prior to consuming it. All participants were given the same instructions, regardless of their condition assignment. Participants were also told that an independent blinder, who is not directly involved in data collection, would come in to administer their assigned oil so that the experimenter’s observations would not be biased. Next, they completed a series of baseline measures (i.e., demographic characteristics, cannabis use history, beliefs about the effects of CBD and THC, trait anxiety, and perceived life stress) and their experimental session was scheduled. Participants were then randomized to receive either accurate or inaccurate information regarding the oil they would receive in the experimental session. Using stratified sampling, participants were divided into two separate blinding lists according to their biological sex. Within each sex stratum, participants were randomly assigned to one of two expectancy conditions (Told CBD vs Told CBD-free) in a 1:1 ratio using an online list randomizer.

A timeline of the experimental session protocol is presented in Figure 1. The 3-h experimental session took place at the Nova Scotia Health Biomedical Translational Imaging Center (BIOTIC), following a minimum of 12 h of abstinence from cannabis, alcohol, tobacco, and illicit drugs. Participants were also required to abstain from caffeine for a minimum of 2 h before the experimental session to minimize confounding effects of recent caffeine use on study outcomes, but not to induce significant caffeine withdrawal symptoms which emerge 12–24 h after acute abstinence among regular caffeine users (Juliano and Griffiths, 2004). Abstinence was verified via self-report, and a CO breath analyzer was administered as a bogus pipeline to enhance compliance with the cannabis and tobacco abstinence requirements. Participants then completed an MRI safety screening questionnaire, which was approved by the MRI technologist prior to initiating the baseline MRI scans. After providing baseline ratings of acute subjective state (i.e., stress, anxiety, sedation, energy) (T1; Baseline) to account for individual differences, participants were set up in the MRI scanner and underwent a series of anatomical MRI scans (~15 min). Next, they completed a second assessment of acute subjective state while inside the scanner (T2; in the scanner), followed by an 8-min baseline resting state MRI scan.

OPEN IN VIEWER

Figure 1.

Experimental session timeline.

Participants then exited the scanner to administer their assigned oil. All participants received a CBD-free hempseed oil (0.3 mg/kg body weight); however, verbal instructions about the CBD content of the oil varied across participants, based on their assigned expectancy condition (Told CBD vs Told CBD-free). The experimenter was required to leave the room while the oil was being administered and an independent blinder, who was not directly involved in the stress task or data collection, was assigned to administer the oil to participants sublingually (i.e., under the tongue). Participants received standardized CBD content instructions (Told CBD vs Told CBD-free) immediately before administering the oil and were asked not to share this information with the experimenter so that their observations would not be biased. To enhance the believability of the expectancy manipulation, the oil was presented to participants in packaging that was consistent with their assigned condition (i.e., commercial CBD packaging in the told CBD condition, hemp seed oil packaging in the told CBD-free condition). The oil was delivered via a syringe and participants were instructed to keep the oil under their tongues for 60-s before swallowing. Oil administration was followed by a 10-min sham absorption period. To standardize participants’ beliefs about the timing of CBD effects, all participants in the told CBD condition were informed that the effects of CBD can emerge in the brain within 10-min of sublingual administration. At the end of the 10-min sham absorption period, participants completed a third acute subjective assessment (T3; post-oil) and were then provided with detailed instructions about the task they would complete inside the MRI scanner.

Participants then re-entered the scanner and were instructed to begin the 4-min stress-induction “counting task” by performing serial subtraction in units of 13 starting from a 4-digit number (e.g., 2043), as quickly and accurately as possible. Scripted verbal instructions were provided to participants who made a mistake, made no mistakes, slowed down, and/or stopped counting. The scanner was paused during the task to avoid head-motion artifacts associated with verbal responding (Van Dijk et al., 2012). Immediately after the stress induction, participants provided ratings of perceived task difficulty and completed a fourth assessment of acute subjective state (T4; post-stressor), followed by a second 8-min resting state MRI scan. Once the scan ended, participants completed a fifth assessment of acute subjective state (T5; anticipation), after which they were informed that they would no longer be required to complete the second trial of the task and were permitted to exit the scanner. Ten minutes later, they completed a final assessment of the acute subjective state (T6; recovery). At this point, participants were asked to report the CBD content of the oil they received during the session, with these response options: “CBD oil,” “CBD-free hempseed oil,” or “unsure.” This served as a manipulation check to determine whether participants believed the information provided by the independent blinder about the CBD content of the oil they received (told CBD vs told CBD-free). All experimenters and blinders involved in this study were required to follow a detailed protocol checklist to ensure that all aspects of the protocol and associated instructions were provided in a standardized manner to all participants. Full debriefing about the nature and aims of the study, and the use of deception, was delayed until data collection for the study was complete to ensure that the deceptive nature of the study was not revealed to prospective participants.

Materials and measures

Carbon monoxide measurement: A breath carbon monoxide (CO) analyzer (Vitalograph, UK) was used to “verify” 12-h abstinence from tobacco and cannabis smoking. Given that there is no reliable CO cutoff for cannabis abstinence, the CO analyzer served as a bogus pipeline (i.e., a technique to increase the truthfulness of self-reports) to enhance compliance with study abstinence requirements. The bogus pipeline helped minimize potential confounding effects associated with recent substance use and/or acute drug-related intoxication, as well as limiting resource costs associated with re-scheduling MRI scans in the event of non-compliance.

Cannabis use history: Current and historical patterns of cannabis use were assessed using an author-compiled questionnaire, adapted from the Daily Sessions, Frequency, Age of Onset, and Quantity of Cannabis Use Inventory (DFAQ-CU; Cuttler and Spradlin, 2017). Items included age at first cannabis use, cannabis use frequency, preferred method of use, and perceptions of the relative THC and CBD concentrations in the cannabis participants use.

Trait anxiety: The psychometrically validated 20-item Trait version of the State-Trait Anxiety Inventory (STAI-T; Spielberger, 1983) was used to measure trait-level anxiety at baseline. Participants rated each statement (e.g., “I feel nervous and restless,” “I am cool, calm, and collected”) based on how they usually feel, using a 4-point scale from 1 (“Not at all”) to 4 (“Very much”). Total scores can range from 20 to 80, with higher scores indicating greater levels of trait anxiety.

Perceived life stress: The 10-item Perceived Stress Scale (PSS; Cohen et al., 1993) was used to measure participants’ perceptions of stress over the past month. Participants rated how often various life situations were appraised as stressful on a 5-point scale ranging from 0 (“Never”) to 4 (“Very often”). Items tap into how unpredictable, uncontrollable, and overloaded participants perceive their lives to be. Total scores can range from 0 to 40, with higher scores indicating greater levels of perceived life stress.

Beliefs about CBD’s effects: A priori beliefs about CBD’s effects on stress and anxiety (i.e., reduces stress, reduces anxiety) were assessed using an author-compiled rating scale ranging from 1 (“Not at all”) to 10 (“Completely”). For relative specificity, beliefs about THC’s effects on stress and anxiety were assessed using the same scale.

CBD-free hemp seed oil: A CBD-free hemp seed oil (Manitoba Harvest: Manitoba, Canada) was administered sublingually to all participants at a dose of 0.3 mg/kg to mimic the CBD doses that reportedly produce anxiolytic effects in humans (MacCallum and Russo, 2018). Oil was delivered via a syringe and participants were instructed to hold the oil under their tongues for 60 s before swallowing. Importantly, Manitoba Harvest hemp seed oil is derived from industrial hemp seed, which contains only trace amounts of CBD, with possible concentrations estimated at <0.002%. Manitoba Harvest Hemp Foods also certifies that their products contain <0.001% THC and will not produce a psychoactive effect (Manitoba Harvest, 2022). This is consistent with other evidence suggesting hemp seed oil is pharmacologically inactive and free of psychoactive properties (Hazekamp et al., 2010).

Stress induction: An adapted version of the Trier Social Stress Test (Kirschbaum et al., 1993) was used to induce state anxiety and stress during the study. Participants were required to perform serial subtraction in increments of 13, starting from a 4-digit number (i.e., 2043) as quickly and accurately as possible while inside an MRI scanner. During the task, participants were prompted for faster performance and were required to restart if they made an error. Scripted evaluative feedback (e.g., “please count faster,” “that is incorrect, start again from 2043”) was delivered by the experimenter via a two-way MRI communication system. To emphasize the social evaluative component of the task, participants were informed that their task performance would be compared to other participants in the study. Participants were also erroneously informed that there would be a second, more difficult, trial of the task after measuring their brain activity. This adaptation was incorporated to minimize potential confounding effects of relief associated with knowing the task had ended (Smeets et al., 2012; Spinella et al., 2021) as well as to induce an element of anticipatory stress/anxiety during the post-stressor resting state MRI scan. The task was piloted outside of the MRI scanner in a separate, laboratory-based study by our group, and was found to elicit significant increases in subjective stress and anxiety (Zhekova et al., 2024). A similar stress induction paradigm has also been shown to induce reliable changes in subjective, physiological (i.e., cortisol response), and neural markers of stress and anxiety (Wang et al., 2005, 2007).

Acute subjective state: A numerical rating scale was used to measure participants’ acute subjective stress, anxiety, sedation, and energy throughout the experiment. Participants rated the extent to which they felt “stressed,” “anxious,” “sedated,” and “energized” on a scale from 1 (“Not at all”) to 10 (“Extremely”). Elevated scores on the “stress” item are thought to be indicative of non-specific arousal and associated difficulties with relaxation, agitation, and irritability (DASS; Lovibond and Lovibond, 1995), while elevated scores on the “anxiety” item are believed to be indicative of apprehension, tension, nervousness, and worry (STAI-S; Spielberger, 1983). The “energized” and “sedated” items were derived from the Brief Biphasic Alcohol Effects Scale (B-BAES; Rueger and King, 2013) to measure potential stimulating (e.g., elated, excited, stimulated) and sedative (e.g., slow thoughts, sluggish) drug effects. The use of single-item descriptors enabled us to conduct brief subjective assessments in the neuroimaging context. Similar rating scales have been demonstrated to have sound psychometric properties in the assessment of stress (e.g., Karekla et al., 2017; Lesage et al., 2012) and anxiety (e.g., Davey et al., 2007; Rossi and Pourtois, 2012).

Perceived task difficulty: A single-item rating scale was used to measure participants’ perceptions of task difficulty immediately after task completion. Participants rated the extent to which they found the “counting task” difficult on a scale from 1 (“Not at all”) to 10 (“Extremely”). This enabled us to test for group differences in perceived task difficulty across the two expectancy conditions, as the level of difficulty could impact participants’ appraisals of and associated responses to the cognitive stressor (Allen et al., 2014).

MRI data acquisition and preprocessing

A detailed description of MRI data acquisition and preprocessing parameters is found in Supplemental File 1. Briefly, anatomical and resting state functional MRI data were collected using a 3.0 Tesla GE MR750 scanner with a 32-channel radiofrequency head coil. The head coil was positioned with a support to minimize head motion and earplugs were provided to limit the effect of noise. Pulse sequences and parameters closely matched those used in the Human Connectome Project (Glasser et al., 2013; see Supplemental File 1 for details).

Preprocessing of anatomical and functional MRI data was conducted using fMRIPrep 22.0.1 (Esteban et al., 2019; Esteban, Blair, et al., 2018; RID:SCR_016216), which is based on Nipype 1.8.4 (Gorgolewski et al., 2011; Gorgolewski et al., 2018; RRID:SCR_002502; see Supplemented File 1 for details). Briefly, preprocessing of T1-weighted anatomical data included field inhomogeneity bias correction, skull stripping, brain tissue segmentation, and spatial normalization to a standard Montreal Neurological Institute (MNI) 152 template. For each of the two resting state functional runs (i.e., baseline, post-stressor), a BOLD reference volume was generated, followed by head motion estimation, spatiotemporal filtering, field inhomogeneity-induced distortion correction, and slice-timing correction. The BOLD reference was then co-registered to the high-resolution T1-weighted anatomical reference using boundary-based registration. Confound time series were calculated based on head motion estimates (i.e., Framewise Displacement, DVARS), global signals extracted within the cerebrospinal fluid (CSF), white matter (WM), and whole-brain global signal (GS). Preprocessed BOLD time series were then resampled to standard MNI152 template space (152NLin2009cAsym; https://nist.mni.mcgill.ca/icbm-152-nonlinear-atlases-2009/).

Data analyses

fMRI data: Resting-state functional connectivity (rsFC) was assessed using a seed-based analysis, implemented through Nilearn version 0.9.0 (Nilearn, RRID: SCR_001362). Pairwise ROIs were located bilaterally in the amygdala (Left: −23.5, −6.73, −19.7; Right: 24.1, −8.05, −20.06), dorsal anterior cingulate cortex (dACC; Left: −6.8, 21.32, 28.29; Right: 7.35, 19.98, 29.06), and ventral anterior insula (vAI; Left: −38.97, 12, −8.11; Right: 44.54, 13.44, −7.51) (see Figure 2). These ROIs are interconnected and have each been implicated in stress- and anxiety-related processing (e.g., Baur et al., 2013; Brooks and Stein, 2015; Jenks et al., 2020; Wang et al., 2005, 2007), as well as CBD-related effects (e.g., Batalla et al., 2021; Fusar-Poli et al., 2009, 2010; Wall, 2019). Seed ROIs were defined using coordinates from the Multi-resolution Intrinsic Segmentation Template (MIST; Urchs et al., 2019), a publicly available parcellation of large functional brain networks and localized functional regions. Center coordinates for each ROI were replicated directly from the MIST and are reported in standard MNI152 template space. To maximize the likelihood of extracting signal from the voxels most likely to be impacted by our experimental manipulations, MIST coordinates for each ROI that was used in the present analyses were compared with MNI coordinates (when available) or the general locale of previously reported effects of stress, anxiety, and CBD administration (e.g., Baur et al., 2013; Fusar-Poli et al., 2009; Jenks et al., 2020). ROIs were defined as a 6 mm sphere centered on the coordinates specified above, created using NiftiSphereMasker in Nilearn 0.9.0.

OPEN IN VIEWER

Figure 2.

ROIs in MNI152 template space. (a) Bilateral amygdala (R: 24.1, −8.05, −20.06; L: −23.5, −6.73, −19.7). (b) Bilateral dACC (R: 7.35, 19.98, 29.06; L: −6.8, 21.32, 28.29). (c) Bilateral AI (R: −38.97, 12, −8.11; L: 44.54, 13.44, −7.51).

First-level (individual run) analyses were performed on spatially normalized fMRIPrep functional derivatives using Nilearn’s GLM. Each functional run was spatially smoothed with FWHM = 7, detrended, standardized, and temporally filtered with high-pass = 0.008 Hz, before denoising with confounds. Confound regressors of no interest included Global Signal, WM, CSF, FD, 6 head motion parameters, plus each of their temporal derivatives, squares, and quadratic terms (i.e., a total of 24 motion regressors plus 12 region-wise global signals plus 4 FD signals). In addition, a scrubbing method was employed to censor volumes where FD > 0.5 or STD_DVARS > 1.5. Each censored volume added one regressor column to the confounds to “ignore” that volume. Confound regressors remove variance associated with known sources of noise, allowing us to focus on the signal variance unique to each ROI. Participants were excluded from the group-level analysis if head motion exceeded a threshold of 30% of total censored volumes on either the baseline or post-stressor rs MRI runs.

Model coefficients (i.e., r correlations) from the individual-run analyses were entered into a series of group-level analyses of covariance (ANCOVA) in SPSS version 28.0 (SPSS Inc., Chicago, IL, USA). Correlation coefficients derived from the individual-run analyses represent the full (marginal) connectivity between the amygdala and pairwise ROIs located bilaterally in the dACC and vAI. For each amygdala-ROI pairing, the model structure was the same, with expectancy condition (Told CBD vs Told CBD-free) as a between-subject factor. Estimates of baseline (pre-stressor) rsFC between each amygdala-ROI pairing were entered as covariates to control for individual differences in neural activity that were unrelated to our experimental manipulations.

Subjective data: Subjective data were analyzed with marginal linear models using the linear mixed model function of SPSS version 28 (SPSS Inc., Chicago, IL, USA). To select the optimal covariance structures, model simplicity and likelihood ratio tests were conducted. The main outcomes were subjective ratings of stress, anxiety, sedation, and energy. Time (i.e., baseline, scanner, post-oil, post-stressor, anticipation, recovery) was specified as a fixed and repeated factor, expectancy condition (Told CBD, Told CBD-free) was specified as a fixed factor, and baseline ratings were entered as time-varying covariates in each model, to control for individual variability across participants.

Between-group independent t-tests were also conducted to determine whether the two expectancy conditions differed on any individual characteristics that could systematically bias responses to the oil administration or the acute stressor (i.e., trait anxiety and perceived life stress scores, a priori beliefs about CBD’s effects, task difficulty ratings, current cannabis use frequency; see Table 1).

OPEN IN VIEWER

Table 1.

Sample characteristics and between-group comparisons.

Variable	Entire sample	Told CBD	Told CBD-free	t (df)	p
	Mean (SD)	Mean (SD)	Mean (SD)
Age	23.67 (8.30)	22.0 (3.02)	25.7 (10.9)	1.35 (30)	0.19
STAI-T	38.21 (8.82)	39.53 (10.34)	37.06 (7.37)	−0.79 (30)	0.44
PSS	15.03 (6.07)	14.27 (4.76)	15.71 (7.11)	0.66 (30)	0.51
CBD beliefs
Reduces stress	7.31 (1.47)	7.13 (1.60)	7.47 (1.37)	0.64 (30)	0.53
Reduces anxiety	7.03 (1.56)	7.07 (1.79)	7.00 (1.37)	−0.12 (30)	0.91
Past-month cannabis use (days)	5.69 (7.85)	6.30 (7.72)	5.15 (8.15)	−0.41 (30)	0.68
CO (ppm)	4.62 (3.34)	4.40 (3.02)	4.82 (3.70)	0.35 (30)	0.73
Task difficulty	6.80 (1.67)	6.67 (1.50)	6.91 (1.85)	0.41 (30)	0.69
Baseline rsFC
L Amyg—L dACC	−0.06 (0.22)	−0.11 (0.20)	−0.02 (0.24)	1.14 (30)	0.27
L Amyg—R dACC	−0.04 (0.21)	−0.02 (0.11)	−0.07 (0.27)	−0.60 (22)	0.56
R Amyg—L dACC	−0.04 (0.20)	−0.06 (0.21)	−0.02 (0.19)	0.54 (30)	0.59
R Amyg—R dACC	−0.02 (0.16)	0.02 (0.15)	−0.06 (0.15)	−1.41 (30)	0.17
L Amyg—L AI	0.03 (0.20)	0.04 (0.18)	0.02 (0.22)	0.44 (30)	0.66
L Amyg—R AI	0.02 (0.15)	0.01 (0.14)	0.03 (0.16)	0.78 (30)	0.44
R Amyg—L AI	−0.03 (0.19)	−0.05 (0.18)	−0.02 (0.20)	−0.24 (30)	0.81
R Amyg—R AI	−0.01 (0.19)	−0.04 (0.17)	0.01 (0.20)	0.34 (30)	0.74
	Entire sample	Told CBD	Told CBD-free	X2 (df)	p
	N (%)	N (%)	N (%)
Sex (female)	15 (46.9)	7 (46.7)	8 (47.1)	0.00 (1)	0.98
Ethnicity
White (European)	23 (71.9)	11 (73.3)	12 (70.6)	4.27 (5)	0.51
Middle Eastern	3 (9.4)	1 (6.7)	2 (11.8)
South Asian	2 (6.3)	—	2 (11.8)
Black	1 (3.1)	1 (6.7)	—
Mixed ethnicity	3 (9.2)	2 (13.4)	1 (5.9)

STAI-T: state-trait anxiety inventory (Spielberger, 1983); PSS: perceived stress scale (Cohen et al., 1993); L: left; R: right; rsFC: resting-state functional connectivity, measured as r correlations between pairwise ROIs; dACC: dorsal anterior cingulate cortex; AI: anterior insula; STAI-T scores have possible values from 20 to 80; PSS scores have a possible value of 0–40; Past-month Cannabis use is defined as the number of days cannabis was used in the past month; CBD beliefs and task difficulty were rated on a scale from 1 (not at all) to 10 (completely or extremely).

Results

Sample characteristics: Of the 38 participants, one male (randomized to the told CBD condition) opted to have their data removed from the study after debriefing, and one female (randomized to the told CBD-free condition) was excluded because of a panic reaction during their baseline MRI scan, resulting in excessive head motion artifact. This led to a final sample of N = 36 (18 Told CBD, 18 Told CBD-free; 17 males, 19 females). An additional four cases (3 told CBD; 1 told CBD-free) were excluded from the group-level rsFC analysis because of excessive head motion (i.e., >30% total censored volumes) on either the baseline or post-stressor resting state MRI scans. The final sample (N = 32) was sex-balanced (54% female), with a mean age of 24.0 years (SD = 8.3). Participants were relatively infrequent cannabis users, with an average reported cannabis use frequency of 6 days over the past month (SD = 7.8). The mean STAI-T score was 38 (SD = 8.8), which is indicative of moderate levels of trait anxiety, and the mean PSS score was 15 (SD = 6.1), which reflects relatively low levels of perceived life stress. A complete summary of participant characteristics is presented in Table 1. Two participants in this sample (5.3%) reported concurrent (i.e., past-month) nicotine use during eligibility screening, with N = 1 concurrent nicotine user randomly assigned to each of the two expectancy conditions. Both concurrent nicotine users received a total score of 0 on the Fagerstrom Test for Nicotine Dependence (Heatherton et al., 1991), which is indicative of non-dependent use and would not be sufficient to induce significant withdrawal symptoms following 12-h of nicotine abstinence. No significant differences were detected between the two expectancy groups (told CBD vs Told CBD-free) on measures of trait anxiety, perceived life stress, baseline rsFC, or other individual characteristics that could systematically bias results (e.g., a priori beliefs about CBD’s therapeutic properties, task difficulty ratings, CO levels, concurrent nicotine use). All participants reported believing the CBD content information they received during the experimental session, suggesting the CBD manipulation was successful.

Amygdala—dACC rsFC: A significant main effect of CBD expectancy was observed for rsFC between the L Amygdala and R dACC (F(1, 30) = 4.53, p = 0.042, η_p² = 0.14 (large effect)). The told CBD group had significantly lower L amygdala-R dACC rsFC following stress exposure, compared to the told CBD-free group while controlling for individual differences at baseline (see Figure 3). There was no significant main effect of CBD expectancy on rsFC between the other amygdala-dACC pairings. Test statistics for all amygdala-dACC ROI pairings are presented in Table 2.

OPEN IN VIEWER

Figure 3.

Group differences in post-stressor rsFC between the Amygdala and pairwise ROIs.

OPEN IN VIEWER

Table 2.

ANCOVA results for the main effect of CBD expectancy condition on rsFC between the amygdala and pairwise ROIs.

Pairwise ROIs	Mean (SE)		F	p	η 2	Covariate
	Told CBD	Told CBD-free				F	p
Amygdala-dACC
L Amyg—L dACC	−0.06 (0.05)	−0.08 (0.05)	0.16	0.69	0.01	0.18	0.68
L Amyg—R dACC	−0.15 (0.05)	−0.02 (0.04)	4.53	0.04	0.14	0.35	0.56
R Amyg—L dACC	−0.08 (0.05)	−0.01 (0.05)	0.92	0.35	0.03	0.13	0.72
R Amyg—R dACC	−0.07 (0.05)	−0.03 (0.05)	0.37	0.55	0.01	2.80	0.11
Amygdala-AI
L Amyg—L AI	0.07 (0.05)	0.07 (0.04)	0.00	0.97	0.00	0.47	0.50
L Amyg—R AI	−0.06 (0.05)	−0.06 (0.05)	0.00	0.99	0.00	2.15	0.15
R Amyg—L AI	−0.02 (0.04)	0.04 (0.03)	1.36	0.25	0.05	0.02	0.89
R Amyg—R AI	−0.03 (0.04)	−0.06 (0.04)	0.21	0.65	0.01	3.12	0.09

ROI: region of interest; L: left, R: right; dACC: dorsal anterior cingulate cortex; AI: anterior insula; Covariate represents baseline rsFC between pairwise ROIs.

*

p < 0.05.

Amygdala—anterior insula rsFC: No significant main effect of CBD expectancy was observed for rsFC of any amygdala-AI pairings (p-values > 0.05; see Figure 3). Test statistics for all amygdala-anterior insula ROI pairings are presented in Table 2.

Subjective effects: Marginal linear models were used to examine Time, Expectancy, and Time by Expectancy effects for subjective stress, anxiety, sedation, and energy. Estimated marginal means and standard errors for stress and anxiety are presented in Figure 4. A significant main effect of time was identified for stress (F(4, 28) = 26.27; p < 0.001), anxiety (F(4, 28) = 13.80; p < 0.001), and energy (F(4, 28) = 3.05; p < 0.001), as well as a trend-level main effect of time for sedation (F(4, 28) = 2.42; p = 0.070). Notably, subjective stress and anxiety increased significantly immediately following the stressor (post-stressor), followed by a significant decrease during anticipation and again at recovery (see Figure 4).

OPEN IN VIEWER

Figure 4.

Estimated marginal means (±standard error) for subjective outcomes. Subjective results presented in panels (a–d) correspond to a more restricted head motion sample of N = 32. Stress, anxiety, sedation, and energy were rated on a scale from 1 (not at all) to 10 (extremely). T1–T6 refer to the time-points for subjective assessments (i.e., T1: Baseline; T2: in scanner; T3: post-oil; T4: post-stressor; T5: anticipation; T6: recovery). All panels show the pairwise breakdown of subjective ratings by time, within each of the two expectancy conditions (Told CBD vs Told CBD-free). For panels (a) and (b), time point 4 was significantly different from all other time points within each expectancy condition, indicating that the counting task effectively increased subjective stress and anxiety. Time-varying covariates specified in each model are shown as baseline values.

There were no significant main effects of Expectancy or Expectancy by Time interactions for any of the subjective outcomes. However, a breakdown of pairwise comparisons within each expectancy condition shows a significant decrease in subjective stress (MD = −0.83, p = 0.017) and anxiety (MD = 0.70, p = 0.021) from the baseline scan to post-oil in the told CBD condition, but not the told CBD-free condition (stress: MD = −0.44, p = 0.16; anxiety: MD = −0.47, p = 0.09). Furthermore, a significant decrease in subjective ratings of stress was also observed from anticipation to recovery in the told CBD condition (MD = −0.60, p = 0.024), but not the told CBD-free condition (MD = 0.235, p = 0.329; see Figure 4); the same group difference was not observed for subjective ratings of anxiety, which significantly decreased from anticipation to recovery in both expectancy conditions (Told CBD: MD = −0.60, p = 0.017; Told CBD-Free: MD = −0.50, p = 0.033). Raw means and standard errors for all subjective outcomes are found in Supplemental File 2.

Additional post hoc exploratory analyses were conducted to test each of the models using a full sample (N = 36) that included participants who were previously removed for excessive head motion in their rsFC data. Post hoc analyses revealed similar, but more robust results for both subjective stress and anxiety, which suggests that interactive effects in the primary analyses may have been limited due to lack of power. Consistent with our primary analyses, there were no significant time-by-expectancy interaction effects for subjective ratings of sedation or energy. Results from all post hoc subjective analyses using the full sample are presented in Supplemental File 3.

Discussion

This study is the first to directly assess the independent effects of CBD expectancy on amygdala-based neural responses to a cognitive stressor in a sample of healthy adults. We found that CBD expectancy effects alone were sufficient to reduce rsFC between the L amygdala and R dACC following an acute stressor. CBD placebo effects were not associated with reliable changes in rsFC between any of the other brain regions examined, which is consistent with prior observations that L amygdala-ACC rsFC may be particularly sensitive to CBD-related effects (e.g., Fusar-Poli et al., 2009, 2010). For example, actual CBD administration has been associated with decreased rsFC between the amygdala and ACC (Batalla et al., 2021). CBD’s anxiolytic effects have also been associated with an attenuated BOLD response in the L amygdala and ACC in healthy adults exposed to fearful stimuli (Fusar-Poli et al., 2009), as well as lower effective connectivity between the L amygdala and ACC (Fusar-Poli et al., 2010). Functional connections between the amygdala and ACC are implicated in how the brain perceives, processes, and regulates emotions, particularly in the context of stress and anxiety (Brooks and Stein, 2015; Fiddick, 2011; Jenks et al., 2020; Robinson et al., 2014). The amygdala is responsible for the detection of threats, as well as signaling to other brain regions about the emotional significance of a stimulus (Brooks and Stein, 2015; Fiddick, 2011). The ACC plays a key role in monitoring and regulating emotional responses driven by the amygdala (Brooks and Stein, 2015; Fiddick, 2011), and increased BOLD activation and associated connectivity between the amygdala and ACC has previously been linked with acute states of anxiety (Brooks and Stein, 2015; Dedovic et al., 2009; Jenks et al., 2020). In the present study, CBD placebo effects alone appeared to blunt rsFC between the amygdala and ACC, which may be the result of a diminished or more regulated stress and anxiety response among those who were led to believe they had received a CBD oil prior to stress exposure. Indeed, blunted amygdala and ACC responses have also been observed among placebo responders in an SSRI trial for social anxiety (Faria et al., 2012, 2014), suggesting a potential shared neural mechanism underlying pharmacologically and placebo-driven anxiolysis. This raises the possibility that expectancy effects may, at least in part, account for anxiolytic effects in the brain that have been attributed to CBD’s pharmacological actions in prior research (e.g., Batalla et al., 2021; Bhattacharyya et al., 2010). These findings also suggest a potential neurobiological basis for CBD placebo effects that have been previously reported by our group using subjective, endocrine, and physiological measures of stress and anxiety (Spinella et al., 2021, 2023; Zhekova et al., 2024).

Although our stress task was found to significantly increase subjective ratings of stress and anxiety in the present study, no statistically significant interactions involving “expectancy condition” and “time” were evident in subjective ratings. However, examination of a priori planned comparisons revealed that consistent with our previous findings (Spinella et al., 2021, 2023; Zhekova et al., 2024), CBD expectancy dampened stress anticipation and/or facilitated stress recovery but did not significantly impact the magnitude of the stress response elicited by the stressor itself. In the present study, there was evidence of decreased subjective stress and anxiety following oil administration in the CBD expectancy condition, relative to a subjective assessment taken during the baseline scan. However, because the MRI environment can serve as a stressor (Madl et al., 2022; Muehlhan et al., 2011) and the post-oil assessment took place between two MRI scans, it is not possible to tease apart the effects on damped stress anticipation versus enhanced stress recovery from the present data. It is noteworthy, however, that there was no significant impact of CBD expectancy on subjective ratings of stress or anxiety during the in-scanner stress anticipation phase of the study, but such an impact was evident during the out-of-scanner recovery assessment. Insofar as the scanner environment served as a stressor itself, it is possible that it obscured the assessment of the impact on anticipation per se, which we have observed out of the scanner using a similar protocol (Zhekova et al., 2024). Future research is necessary to delineate CBD-placebo-specific effects on stress anticipation and recovery. Furthermore, our group previously found that CBD expectancy was associated with increased subjective sedation (Spinella et al., 2021; Zhekova et al., 2024), which was not replicated in the present study. Since our previous studies did not incorporate a neuroimaging component, it is possible that the potentially anxiogenic and/or stimulating scanner environment may have obscured anticipated CBD expectancy effects on sedation in the present study.

Drug responses in humans are known to arise from a combination of the drug’s direct pharmacological effects and other non-pharmacological factors, which comprise the placebo effect (Kirsch, 1997, 2018; Schlagintweit et al., 2019), and the clinical effectiveness of a compound may depend on the presence of both pharmacological and placebo effects. Indeed, open-label trials that have assessed the use of CBD for the treatment of anxiety disorders tend to suggest a significant clinical benefit (Berger et al., 2022; Dahlgren et al., 2022), while the outcomes of randomized double-blind studies assessing CBD’s potential anxiolytic effects when compared to a placebo control tend to be more mixed. For example, although CBD has been reported to acutely blunt subjective anxiety in some double-blinded placebo-controlled studies (Bergamaschi et al., 2011; Crippa et al., 2011), other studies found no effect of CBD versus placebo on various indices of anxiety (Arndt and de Wit, 2017; Gournay et al., 2023a; Hundal et al., 2018; Stanley et al., 2023). Because these studies did not assess perceived drug assignment, it is not clear to what extent strong placebo effects obscure drug effects in negative trials and/or contribute to putative therapeutic effects in positive trials. Given the evidence that patients in double-blind drug trials often make accurate guesses about their treatment assignment and that treatment guesses have been associated with clinical outcomes, above and beyond actual treatment assignment (e.g., Baethge et al., 2013; Chen et al., 2011; Dar and Barrett, 2014; Kirsch, 2019), it will be important to routinely assess and quantify the effects of perceived drug assignment and other expectancy-related factors in double-blind trials of CBD. This would help improve our interpretation of trial outcomes to ensure that the pharmacological effects of CBD are not over-estimated. In this study, CBD placebo was observed to have a large effect on anxiety-related neural responses, which is consistent with placebo effect sizes observed for other psychiatric interventions (e.g., Jones et al., 2021). Our findings also emphasize the need for experimental research that directly manipulates CBD-related expectancies across a range of modalities, outcomes, and user characteristics, as this could help establish a more reliable reference point for quantifying CBD placebo effect sizes in future clinical trials. This emerging body of work has important implications for CNS drug development, informing efforts to leverage expectancies in clinical practice as a means of improving therapeutic outcomes, and providing more accurate public messaging about CBD’s potential therapeutic benefits so that individuals can make more informed decisions about their use of CBD (e.g., Gournay et al., 2024).

The present findings should be interpreted with the following methodological considerations in mind. First, this study used a between-subject design with each subject assigned to only one of the two expectancy conditions and thus, we were only adequately powered to detect large effects. Although our analytic approach enabled us to control for some between-subject variability by covarying baseline estimates of rsFC and subjective ratings of stress and anxiety, we were underpowered to detect subtle effects or explore potential moderators (e.g., sex, individual beliefs about CBD effects, perceived task difficulty). For example, positive beliefs about CBD’s anxiolytic properties have been shown to enhance the magnitude of CBD-related anxiolytic placebo responses in our prior work (Spinella et al., 2021). This raises the possibility that expectancy effects may have been more robust in this study if we were adequately powered to control for individual variability in participants’ beliefs about the anxiolytic properties of CBD. Future studies using larger sample sizes and within-subject contrasts are warranted to maximize statistical power and explore additional moderators of interest. Second, our sample was relatively homogeneous, predominantly comprised of healthy university students of European descent. It will be important to replicate and extend these findings with more diverse samples, including those with anxiety- or stress-related conditions. Third, immediately after the oil administration, participants underwent a brief sham absorption period and those in the “Told CBD” group were informed via standardized instructions that the effects of CBD can emerge in the brain within 10 min of sublingual administration. It is possible that this phrasing may have led to suboptimal timing of maximal expectancy effects. In addition, the verbal requirements of the stress task precluded us from measuring rsFC during the stress task itself, due to the potential of inducing head-motion artifact. Thus, we lack neural data that directly correspond to the application of the stressor itself. Next, our study utilized a seed-based approach to analyze our rsFC data. Although our seed ROIs were selected based on reliable functional parcellations (MIST) and our seed coordinates were compared to those reported in previous research that found effects of stress/anxiety or CBD in the same regions (e.g., Baur et al., 2013; Jenks et al., 2020), we cannot be certain that our seeds were placed in the locale of peak effects. Furthermore, the 12-h cannabis abstinence may not have been sufficient to control for the residual effects of THC, which is highly lipophilic and can remain pharmacologically active for several weeks as it is released from adipose tissues (Lucas et al., 2018). Participants in this study were pre-screened for drug dependence (including cannabis) and no significant differences in past-month cannabis use frequency were identified between the two expectancy conditions (Told CBD vs Told CBD-free), which helps mitigate this concern. Our pre-post design and analytic approach also enabled us to control for individual differences at baseline; however, because the recency of THC exposure was not directly assessed in the present study, we cannot rule out residual THC-related effects as a possible confound. Lastly, our study did not employ a fully balanced placebo design, which does not allow us to make inferences about CBD placebo effects either in combination with, or in comparison to, CBD pharmacology.

Conclusion

This is the first study to explore the neural substrates of CBD-related placebo effects in the context of stress and anxiety. Although the present study was underpowered to detect subtle effects, our preliminary results suggest that CBD expectancy alone may be sufficient to alter neural responses relevant to its purported anxiolytic and stress-relieving properties in healthy adults. Findings from this study are consistent with and extend, previous evidence that CBD can attenuate amygdala responses to fearful stimuli and is associated with decreased connectivity between the amygdala and ACC. Results are also consistent with CBD placebo effects previously reported by our group across a range of modalities (i.e., cortisol responses, heart rate variability, subjective state), and neural substrates of placebo effects associated with other anxiolytic drugs (e.g., SSRIs), which supports the reliability of these findings. Additional research is needed to replicate these results, to extend them to clinically relevant populations, and to determine whether there are interactive effects of CBD expectancy and pharmacology on anxiety reactivity and stress reactivity in the brain. This emerging body of work has the potential to yield more reliable benchmarks for quantifying CBD placebo effects in randomized clinical trials of CBD and may have important implications for clinical practice, drug development, and consumer education.

Supplemental Material

Supplemental Material

Supplemental Material