Year-Long Cannabis Use for Medical Symptoms and Brain Activation During Cognitive Processes

Key Points Question Is there an association between year-long cannabis use for medical symptoms and brain activation during cognitive processes implicated in cannabis use? Findings In a cohort study of adults who newly obtained medical cannabis cards for symptoms of depression, anxiety, pain, or insomnia, functional magnetic resonance imaging measures during working memory, reward, and inhibitory control tasks did not differ statistically from baseline to 1 year and were not associated with changes in cannabis use frequency. Meaning The absence of activation differences in this study suggests that adults using cannabis for medical symptoms over 1 year may not experience significant changes within reward, working memory, or inhibitory control domains.


Study protocol
The study flow diagram shows, for the entire clinical trial, which arm received imaging and how many participants were available for imaging.

eFigure 1. Study Flow Diagram
The study used data that was collected as part of a larger clinical trial of cannabis for medical symptoms.Imaging data was collected in one arm of the trial and in a control group.

Experimental Paradigm Cannabis Metrics
Frequency of cannabis use was assessed using a likert-scale of use throughout the past month (Less than once a week; Less than once a month; 1-2 days a week; 3-4 days a week; 5-6 days a week; Once or more per day).Cannabis use disorder (CUD), was assessed using the Cannabis Use Disorder Identification Test Revised (CUDIT-R).Urinalysis for cannabis metabolites was also conducted N-Back Task In the N-back task, participants were asked to press a button either on the same letter as two letters back (two-back) or on a predetermined letter (zero-back).There were 3 two-back blocks and 3 zero-back blocks, which were presented in an alternating fashion in a single run, starting with a zero-back block.After zero-back blocks there was a 2000ms gap, while after two-back blocks there was a 15000ms rest period.Participants were shown a different pseudorandomized letter sequence in each block and each sequence had 29 letters per round, of which 5 required a button press.Letters were presented for 500ms, followed by a 1500ms blank interstimulus screen and participants could respond at any time during this 2000ms window.Accuracy and reaction time were calculated for the zeroback and two-back trials individually and for the whole task.See eFigure 2 for a graphical overview.

eFigure 2. Schematic of the N-back Task
The N-back task consists of one run with 6 blocks of alternating zero-back and two-back blocks.Participants press a button either on the same letter as two letters back (two-back) or on a predefined letter (zero-back) within a time frame of 2000ms.

MID Task
In the MID task, participants were asked to press a button as fast as possible following a fixation cross of variable length to either win or avoid losing a certain monetary amount.The task consisted of 2 runs with 50 trials each (10 $5 win, 10 $5 loss, 10 $0.20 win, 10 $0.20 loss, 10 neutral).Each trial started with the presentation of the cue slide (2000ms) showing an outcome (win/loss/neutral), followed by a fixation cross lasting a randomized duration (1500-4000ms).Then the participant was given a certain period of time to respond.The response period was initially set at a participant's reaction time and recalculated every 3 trials based on the participant's overall accuracy (between 150ms and 500ms).After each run a score was calculated based on wins and losses and the combined score was paid out to the participant at the end of the second run.See eFigure 3 for a graphical overview.

eFigure 3. Schematic of the MID Task
The MID task consists of two runs in which participants are presented cues to either win or lose a high or low amount of money.Then they have to press a button as fast as possible following a fixation cross of variable length and receive feedback on whether they gain or lose money.The response period is altered based on performance and the monetary reward is paid out to participants at the end of the two runs.

SST Task
In the SST task, participants were asked to press a button upon presentation of an arrow (go trial) unless it was immediately followed by a stop signal (stop trial).The task consisted of 2 runs with 104 trials each (78 go/ 26 stop).A go trial consisted of a 200ms fixation cross followed by an image displayed for 500ms.The image was either related to cannabis or neutral (52 cannabis/ 52 neutral).Following the image, a blank screen was displayed for 500ms, after which an arrow image, pointing either in the left or right direction, was presented.The participant was to respond to this arrow by pressing either a left or right button.The arrow image was released once the press was registered (maximum duration 1000ms).The arrow was followed by an interstimulus fixation cross with variable duration (900-2100ms, mean = 1477ms), which transitioned into the next trial's 200ms starting fixation cross.In stop trials, a stop signal was shown above the arrow following a delay, which indicated to the participant to inhibit the button press.The initial delay time was 250ms.Successful inhibition increased the time by 50ms while failed inhibition decreased the time by 50ms within a range of 0-500ms.Stop signal reaction times, taken as the inferred mean latency between the stop signal and response inhibition, were estimated via an additive multilevel linear model with a participant-varying intercept.See eFigure 4 for a graphical overview.

eFigure 4. Schematic of the SST Task
In the SST task participants need to press a button as fast as possible when presented with an arrow unless a stop sign appears above the arrow with a slight delay.Some trials are preceded by a neutral and others by a cannabis-related image and there are two runs total.

Pre-Processing
The following is explicitly copied according to the instructions from the fMRIPrep boilerplate generated during the fMRIPrep 23.0.1 preprocessing.Changes made were the deletion of a duplicate paragraph and the changing of the citation style.
Results included in this manuscript come from preprocessing performed using fMRIPrep 23.0.1 (RRID:SCR_016216), 1,2 which is based on Nipype 1.8.5 (RRID:SCR_002502). 3,4eprocessing of B0 inhomogeneity mappings A total of 1 fieldmaps were found available within the input BIDS structure for this particular subject.A B0nonuniformity map (or fieldmap) was estimated based on two (or more) echo-planar imaging (EPI) references with topup (FSL 6.0.5.1:57b01774). 5atomical data preprocessing A total of 1 T1-weighted (T1w) images were found within the input BIDS dataset.The T1-weighted (T1w) image was corrected for intensity non-uniformity (INU) with N4BiasFieldCorrection, 6 distributed with ANTs 2.3.3 (RRID:SCR_004757), 7 and used as T1w-reference throughout the workflow.The T1w-reference was then skullstripped with a Nipype implementation of the antsBrainExtraction.sh workflow (from ANTs), using OASIS30ANTs as target template.Brain tissue segmentation of cerebrospinal fluid (CSF), white-matter (WM) and gray-matter (GM) was performed on the brain-extracted T1w using fast (FSL 6.0.5.1:57b01774,RRID:SCR_002823). 8Brain surfaces were reconstructed using recon-all (FreeSurfer 7.3.2,RRID:SCR_001847), 9 and the brain mask estimated previously was refined with a custom variation of the method to reconcile ANTs-derived and FreeSurfer-derived segmentations of the cortical gray-matter of Mindboggle (RRID:SCR_002438). 10Grayordinate "dscalar" files 11 containing 91k samples were also generated using the highest-resolution fsaverage as an intermediate standardized surface space.Volume-based spatial normalization to two standard spaces (MNI152NLin6Asym, MNI152NLin2009cAsym) was performed through nonlinear registration with antsRegistration (ANTs 2.3.3),using brain-extracted versions of both T1w reference and the T1w template.The following templates were were selected for spatial normalization and accessed with TemplateFlow (23.0.0): 12

Functional data preprocessing
For each of the 6 BOLD runs found per subject (across all tasks and sessions), the following preprocessing was performed.First, a reference volume and its skull-stripped version were generated using a custom methodology of fMRIPrep.Head-motion parameters with respect to the BOLD reference (transformation matrices, and six corresponding rotation and translation parameters) are estimated before any spatiotemporal filtering using mcflirt (FSL 6.0.5.1:57b01774). 15The estimated fieldmap was then aligned with rigid-registration to the target EPI (echoplanar imaging) reference run.The field coefficients were mapped on to the reference EPI using the transform.BOLD runs were slice-time corrected to 0.722s (0.5 of slice acquisition range 0s-1.45s) using 3dTshift from AFNI (RRID:SCR_005927). 16The BOLD reference was then co-registered to the T1w reference using bbregister (FreeSurfer) which implements boundary-based registration. 17Co-registration was configured with six degrees of freedom.Several confounding time-series were calculated based on the preprocessed BOLD: framewise displacement (FD), DVARS and three region-wise global signals.FD was computed using two formulations following Power (absolute sum of relative motions) 18 and Jenkinson (relative root mean square displacement between affines). 15FD and DVARS are calculated for each functional run, both using their implementations in Nipype (following the definitions by Power et al. 2014).The three global signals are extracted within the CSF, the WM, and the whole-brain masks.Additionally, a set of physiological regressors were extracted to allow for component-based noise correction (CompCor). 19Principal components are estimated after high-pass filtering the preprocessed BOLD time-series (using a discrete cosine filter with 128s cut-off) for the two CompCor variants: temporal (tCompCor) and anatomical (aCompCor).tCompCor components are then calculated from the top 2% variable voxels within the brain mask.For aCompCor, three probabilistic masks (CSF, WM and combined CSF+WM) are generated in anatomical space.The implementation differs from that of Behzadi et al. in that instead of eroding the masks by 2 pixels on BOLD space, a mask of pixels that likely contain a volume fraction of GM is subtracted from the aCompCor masks.This mask is obtained by dilating a GM mask extracted from the FreeSurfer's aseg segmentation, and it ensures components are not extracted from voxels containing a minimal fraction of GM.Finally, these masks are resampled into BOLD space and binarized by thresholding at 0.99 (as in the original implementation).Components are also calculated separately within the WM and CSF masks.For each CompCor decomposition, the k components with the largest singular values are retained, such that the retained components' time series are sufficient to explain 50 percent of variance across the nuisance mask (CSF, WM, combined, or temporal).The remaining components are dropped from consideration.The head-motion estimates calculated in the correction step were also placed within the corresponding confounds file.The confound time series derived from head motion estimates and global signals were expanded with the inclusion of temporal derivatives and quadratic terms for each. 20Frames that exceeded a threshold of 0.5 mm FD or 1.5 standardized DVARS were annotated as motion outliers.Additional nuisance timeseries are calculated by means of principal components analysis of the signal found within a thin band (crown) of voxels around the edge of the brain, as proposed by. 21The BOLD timeseries were resampled into standard space, generating a preprocessed BOLD run in MNI152NLin6Asym space.First, a reference volume and its skull-stripped version were generated using a custom methodology of fMRIPrep.The BOLD time-series were resampled onto the following surfaces (FreeSurfer reconstruction nomenclature): fsaverage.Automatic removal of motion artifacts using independent component analysis (ICA-AROMA) 22 was performed on the preprocessed BOLD on MNI space time-series after removal of non-steady state volumes and spatial smoothing with an isotropic, Gaussian kernel of 6mm FWHM (full-width half-maximum).Corresponding "non-aggresively" denoised runs were produced after such smoothing.Additionally, the "aggressive" noiseregressors were collected and placed in the corresponding confounds file.Grayordinates files 11 containing 91k samples were also generated using the highest-resolution fsaverage as intermediate standardized surface space.All resamplings can be performed with a single interpolation step by composing all the pertinent transformations (i.e.head-motion transform matrices, susceptibility distortion correction when available, and co-registrations to anatomical and output spaces).Gridded (volumetric) resamplings were performed using antsApplyTransforms (ANTs), configured with Lanczos interpolation to minimize the smoothing effects of other kernels. 23Non-gridded (surface) resamplings were performed using mri_vol2surf (FreeSurfer).
Many internal operations of fMRIPrep use Nilearn 0.9.1 (RRID:SCR_001362), 24 mostly within the functional processing workflow.For more details of the pipeline, see the section corresponding to workflows in fMRIPrep's documentation.

Smoothing and Scaling
Data in grayordinate format used in the surface-based analysis was smoothed using the Human Connectome Project (HCP) workbench toolbox with the surface and volume kernel both set to 1.6986 (corresponds to a full width at half maximum (fwhm) of 4mm).Smoothing of voxels for the volume-based analysis (fwhm set to 4mm) and general linear model (GLM) fitting were conducted using the Python package Nilearn version 0.9.2.Post smoothing and pre GLM, the time series was scaled by multiplying by a constant, 1000, and dividing by the median of the grayordinates across time and space, after subtracting the minimum value of the time series across time and space from both the time series and the median.

First Level Modeling
First level modeling using a multiple linear model included as nuisance regressors six rigid head motion parameters, single-volume motion and non-steady state outliers, discrete cosine-basis regressors, the first six anatomical CompCor regressors and framewise displacement.Noise was modeled using a first-order autoregressive model.Stimuli regressors used in the first level modeling included regressors for the zero-back and the two-back blocks of the N-back task, cue, hit and miss regressors for the high/low reward, high/low loss, and neutral events of the MID task, and cannabis and neutral-primed regressors for the go, successful stop, and unsuccessful go events of the SST task.These were convolved with a hemodynamic response function from the SPM dispersion derivative model.
For the N-back task, a single image contrast of the two-back vs zero-back stimuli was calculated.For the MID task, given the number of possible contrasts, this study focused on those from two recent publications. 25,26Anticipation contrasts included high reward vs neutral anticipation, low reward vs neutral anticipation, reward vs neutral anticipation, high reward vs low reward anticipation, high reward vs implicit baseline, high loss vs neutral anticipation, low loss vs neutral anticipation, and high loss vs low loss anticipation.Feedback contrasts included high reward vs neutral hit feedback, reward vs missed reward feedback, high loss vs neutral hit feedback and loss vs avoided loss feedback.For the SST task, contrasts calculated include correct inhibition (successful stop vs go), incorrect inhibition (unsuccessful stop vs go), and unsuccessful inhibitory control (unsuccessful stop vs successful stop).Two runs were collected for the MID and SST tasks, which were combined using a second linear model prior to group modeling.

Group Level Analysis
The effect sizes of group contrasts were taken to be the average (for group level results at a single time point), average difference (for differences in the same participants across time) and difference of averages (for differences across the control and MCC participant groups at baseline) of the individual contrasts of the participants.These were estimated using an ordinary least squares multiple regression model and significance of the group contrasts was assessed in a two-sided manner.Imaging results were FDR controlled at 0.05.Effect sizes at the group level were standardized for visualization purposes using the unbiased Hedges estimator.Covariates in the group level model included sex, age, and past-month cannabis use frequency, mean-centered for numerical variables.Of note, pastmonth cannabis use frequency, originally collected as a categorical variable, was re-coded as a numerical variable representing the approximate fraction of days in a month where cannabis was used.In the supplemental sensitivity analysis of using an FD cutoff of 0.3 and no outlier removal, an additional subject had to be excluded due to a missing past-month cannabis use frequency value at baseline.

Quality Control
Runs were excluded if they were statistical outliers at a study time point based on their signal-to-noise ratio, temporal signal-to-noise ratio, ghost-to-signal ratio along the two phase encoding directions, if their mean framewise displacement (FD) was above 0.2 (unless <30% of a scan had a FD above the FD threshold), or if more than 30% of their volumes were motion outliers as determined by MRIQC. 27Since we made the decision to exclude data on a run-by-run basis, there can be varying sample sizes across task analyses even though the initial number of participants scanned was the same.For example, a participant's N-back run may be excluded at a time point due to having too high a FD value, but at the same time their SST and MID data may be kept due to lower FD values.We also ran our analyses with a more relaxed FD cutoff of 0.2, keeping all other cutoffs the same, and without excluding any outliers to assess the effect of quality control on the neuroimaging results.
a MCC baseline corresponds to the participants of the medical cannabis group, who had imaging collected at both time points, at baseline b MCC 1 year corresponds to the participants of the medical cannabis group, who had imaging collected at both time points, at baseline c CUD = cannabis use disorder The following results expand upon those in the main text.For each task, we show the results from the same contrasts displayed in the main text but derived from the grayordinate-based analysis.Of note, the findings, including the comparisons between HC and MCC participants at baseline and between baseline and 1 year time points of the MCC participants, did not differ between the analysis conducted in volumetric vs in grayordinate space.We also show results from additional contrasts of the MID and SST tasks derived from both the volumetric and grayordinatebased analyses.For these, no significant differences between HC and MCC participants at baseline and between baseline and 1 year time points of the MCC participants were observed.The following results are showing the same contrasts as in the main article but the scans removed due to being quality control outliers are varied.Shown below are the results when no outlier scans were removed and when the framewise displacement (FD) cutoff of 0.2 from the main results was relaxed to 0. Associations between brain activation and longitudinal factors not related to cannabis use, such as age-related changes, cannot be separated from cannabis-specific associations.Thus, future studies should compare brain imaging from a non-using control group across two time points as well.Additionally, the small behavioral changes we observed may have resulted from differences between cohorts (SSRT) or from the impact of task exposure (twoback reaction time).To better interpret these findings, further studies with a more diverse selection of participants are necessary.
We focused on the frequency of cannabis use, rather than on amount of use per occasion or quantitative measures of exposure such as urine or saliva metabolites, as these measures can have significant limitations, and participants are often inaccurate when asked about chemical contents of products. 28,29The choice to focus on frequency, however, may have affected our ability to detect fine-grained associations with the amount of use.
Finally, tasks included in this study only capture a snapshot of cognition, probing aspects of working memory, reward processing, and inhibitory control.Brain activation during tasks that probe other cognitive processes could be affected by year-long cannabis use for medical symptoms.While this study selected classic tasks probing areas of cognition known to be affected by recreational cannabis use, it cannot ascertain that no facet of cognition was associated with cannabis use, and future studies should explore additional tasks or functional connectivity independent of a specific task.

eFigure 6 .
Brain Activation for the MID Task's High Reward Cue vs Baseline Contrast Across Groups and Time Points, the Same Contrast as in the Main Article Analyzed in Grayordinate Space eFigure 7. Brain Activation for Additional Contrasts of the MID Task Across Groups and Time Points From the Volumetric Analysis eFigure 8. Brain Activation for Additional Contrasts of the MID Task Across Groups and Time Points From the Grayordinate Analysis eFigure 9. Brain Activation for the SST Task's 2 Stop vs Go Contrasts Across Groups and Time Points, the Same Contrasts as in the Main Article Analyzed in Grayordinate Space eFigure 10.Brain Activation for an Additional Contrast of the SST Task Across Groups and Time Points From the Volumetric Analysis eFigure 11.Brain Activation for an Additional Contrast of the SST Task Across Groups and Time Points From the Grayordinate Analysis © 2024 Burdinski D et al.JAMA Network Open.

eAppendix 2 .
Quality Control and Neuroimaging Results

3 .
Panels labeled (a) refer to the analysis conducted in volumetric space and panels labeled (b) refer to the analysis conducted in grayordinate space.N-back Task Imaging Results eFigure 12. Brain Activation for the N-back Task's 2-back vs 0-back Contrast Across Groups and Time Points With no Outliers Removed eFigure 13.Brain Activation for the N-back Task's 2-back vs 0-back Contrast Across Groups and Time Points With the FD Cutoff Relaxed to 0.3 MID Task Imaging Results eFigure 14.Brain Activation for Various Contrasts of the MID Task Across Groups and Time Points With no Outliers Removed eFigure 15.Brain Activation for Various Contrasts of the MID Task Across Groups and Time Points With the FD Cutoff Relaxed to 0.3 eFigure 16.Brain Activation for the SST Task's 2 Stop vs Go Contrasts Across Groups and Time Points With no Outliers Removed eFigure 17.Brain Activation for the SST Task's 2 Stop vs Go Contrasts Across Groups and Time Points With the FD Cutoff Relaxed to 0.3 eAppendix 3. Supplementary Limitations eTable 1. Statistically Significant Difference in Cannabis Use-Related Metrics After 1 Year of Cannabis Use for Medical Symptoms 2. No Statistically Significant Difference in N-Back Behavioral Measures Compared With Control or Across Time Except in 2-back Reaction Time Across Time HC vs MCC at baseline corresponds to comparing the imaging control group with the medical cannabis group at baseline b MCC at baseline vs 1 year corresponds to comparing the participants of the medical cannabis group with imaging at both time points across time c BH = Benjamini-Hochberg eTable 3. Significantly Faster SSRT in Healthy Controls Compared With MCC Participants at Baseline, but no Significant Difference Across Time in MCC Participants HC vs MCC at baseline corresponds to comparing the imaging control group with the medical cannabis group at baseline b MCC at baseline vs 1 year corresponds to comparing the participants of the medical cannabis group with imaging at both time points across time c SSRT = stop signal reaction time eAppendix 1.Additional Imaging Results eTable a a