OUP user menu

Marijuana use is associated with a reorganized visual-attention network and cerebellar hypoactivation

L. Chang, R. Yakupov, C. Cloak, T. Ernst
DOI: http://dx.doi.org/10.1093/brain/awl064 1096-1112 First published online: 3 April 2006

Summary

Attention and memory deficits have been reported in heavy marijuana users, but these effects may be reversible after prolonged abstinence. It remains unclear whether the reversibility of these cognitive deficits indicates that chronic marijuana use does not alter cortical networks, or that such changes occur but the brain adapts to the drug-induced changes. Blood oxygenation-level dependent (BOLD) functional MRI (fMRI) was performed in 24 chronic marijuana users (12 abstinent and 12 active) and 19 age-, sex- and education-matched control subjects during a set of visual-attention tasks with graded levels of difficulty. Neuropsychological tests were also administered on each subject. The two marijuana user groups showed no significant difference in usage pattern (frequency or duration of use, age of first use, cumulative joints used, averaged >2000 joints) or estimated cumulative lifetime exposure of Δ-9-tetrahydrocannabinol (THC) (mean 168 ± 45 versus 244 ± 135 g). Despite similar task and cognitive test performance compared with control subjects, active and abstinent marijuana users showed decreased activation in the right prefrontal, medial and dorsal parietal, and medial cerebellar regions, but greater activation in various frontal, parietal and occipital brain regions during the visual-attention tasks (all with P ≤ 0.001, corrected, cluster level). However, the BOLD signals in the right frontal and medial cerebellar regions normalized with duration of abstinence in the abstinent users. Active marijuana users, with positive urine tests for THC, showed greater activation in the frontal and medial cerebellar regions than abstinent marijuana users and greater usage of the reserve network (regions with load effect), suggesting a neuroadaptive state. Both earlier age of first use and greater estimated cumulative dose of THC exposure were related to lower BOLD signals in the right prefrontal region and medial cerebellum. The altered BOLD activation pattern in the attention network and hypoactivation of the cerebellum suggest neuroadaptive processes or alteration of brain development in chronic marijuana users. These changes also may be related to marijuana-induced alteration in resting cerebral blood volume/flow or downregulation of cannabinoid (CB1) receptors. The greater activation in the active compared with abstinent marijuana users demonstrates a neuroadaptive state in the setting of active marijuana use, while the long-term chronic effect of marijuana on the altered brain network may be reversible with prolonged abstinence.

  • cannabis
  • cerebellum
  • attention network
  • BOLD = blood oxygenation-level dependent
  • DMP = dorsal medial parietal
  • fMRI = functional MRI
  • IFG = inferior frontal gyrus
  • MFG = middle frontal gyrus
  • PCG = post-central gyrus
  • PET = positron emission tomography
  • rCBF = regional cerebral blood flow
  • ROI = regions of interest
  • SFG = superior frontal gyrus
  • SPM = statistical parametric mapping
  • THC = Δ-9-tetrahydrocannabinol

Introduction

Marijuana is a frequently used illicit drug in the United States and in many countries; however, its effects on brain function are still not well understood. Nearly 50% of twelfth graders in the United States have tried marijuana at least once and 6% use it daily; only nicotine cigarettes have more daily users than marijuana (Monitoring the Future Study, 2004). Recent studies also indicate that marijuana use precedes use of other illicit substances (Fergusson and Horwood, 2000). Whether marijuana use predisposes individuals to drug abuse (as a ‘gateway drug’) or whether it is just the first available illicit drug to at-risk populations is not known. A study of 311 pairs of same-sex twins found that the twins with earlier marijuana use (before age 17 years) were 2–5 times more likely to use other illicit drugs, especially psychostimulants (Lynskey et al., 2003). Therefore, it is likely that marijuana alters brain function or the neural substrate and hence further predisposes the individual to future drug use. Functional neuroimaging studies and cognitive tests could demonstrate how the brain might be affected by chronic marijuana use.

Findings regarding the neuropsychological function of cannabis users have been somewhat controversial. Several studies found deficits in attention and memory in heavy marijuana users within the first 7 days of abstinence (Pope and Yurgelun-Todd, 1996; Harrison et al., 2002; Solowij et al., 2002). However, with prolonged abstinence (after 28 days), some reported normalization of cognitive function (Pope et al., 2001; Harrison et al., 2002) while others observed persistent cognitive deficits (Bolla et al., 2002). A meta-analysis of >1000 subjects also found that chronic cannabis use had little effect on cognitive function except for possible decrements in the ability to learn and remember new information (Grant et al., 2003). It is unclear, however, whether the reversibility of attention and memory deficits indicates that chronic marijuana use does not alter brain networks, or that such changes occur but the brain adapts to the drug-induced changes.

Several neuroimaging studies have been performed to evaluate the effects of marijuana, or Δ-9-tetrahydrocannabinol (THC), the major active ingredient in marijuana, on brain structure and function. The few structural MRI studies have shown contradictory results. Two studies in frequent or long-term marijuana users found no evidence of global or regional cerebral atrophy (Block et al., 2000a; Tzilos et al., 2005), including the hippocampus (Tzilos et al., 2005). However, others reported decreased grey matter but increased white matter density in heavy marijuana users (Matochik et al., 2005), as well as smaller whole brain volumes in those who had started using marijuana before age 17 years compared with those who started using it later (Wilson et al., 2000).

A positron emission tomography (PET) study found decreased glucose metabolism in the cerebellum of chronic marijuana users compared with controls at baseline, but increased metabolism in the cerebellum, frontal and striatal regions after acute intravenous THC administration (Volkow et al., 1996). Several subsequent PET studies that used O-15 water found analogously decreased regional cerebral blood flow (rCBF) in the cerebellum of marijuana users after monitored abstinence (Block et al., 2000b), but increased rCBF in frontal and cerebellar regions after either intravenous THC administration or marijuana smoking (Mathew et al., 1998; O'Leary et al., 2002). Subjects that performed an auditory attention task additionally showed reduced rCBF in the temporal lobes (O'Leary et al., 2002). More recently, functional MRI (fMRI) studies in chronic marijuana smokers were performed, with variable results. Decreased activation in the cingulate region was found with a motor task (Pillay et al., 2004); increased activation in the cingulate, prefrontal and striatal regions with a spatial working memory task (Kanayama et al., 2004); and both increased and decreased blood oxygenation-level dependent (BOLD) signals in prefrontal regions with tasks that evaluated response inhibition (Smith et al., 2004; Gruber and Yurgelun-Todd, 2005).

Since attention is required for all cognitive tasks, but no fMRI studies have been performed to evaluate specifically how attention might be affected, we performed fMRI in both abstinent and active chronic marijuana users during a set of visual-attention tasks and assessed their cognitive function using neuropsychological tests. We also compared brain activation patterns in active users, documented by positive THC metabolites in the urine, with those in abstinent subjects with negative urine tests. On the basis of these recent PET, fMRI and cognitive studies, we hypothesized that (i) both abstinent and active chronic marijuana users would show an altered brain activation network, with decreased activation in the normal attention network and increased activation in compensatory brain regions; (ii) active users would show greater usage of brain regions typically activated with increasing attentional load (i.e. reserve regions) compared with abstinent users; and (iii) both user groups would have relatively normal cognitive function, demonstrating neuroadaptation to chronic marijuana use.

Methods

Participants

Forty-three subjects were enrolled and studied: 12 active marijuana users with positive THC urine toxicology (age 27.9 ± 10.8 years, 9 males and 3 females, education: 14.6 ± 1.2 years), 12 abstinent marijuana users with negative THC urine tests (age 29.6 ± 8.7 years, 6 males and 6 females, education: 14.5 ± 2.2 years) and 19 comparison subjects with similar ranges of age and education (age 30.6 ± 8.0 years, 11 males and 8 females, education: 15.2 ± 1.6 years). Each subject was informed of the study and provided a written consent that was approved by our Institutional Review Board. All subjects were carefully screened with neuropsychiatric evaluations and laboratory screening tests to ensure that they fulfilled all study criteria. The inclusion criteria for all subjects were as follows: (i) male or female ages ≥17 years; (ii) healthy and on no medications (except vitamins and/or oral contraceptives); and (iii) willing and able to provide consent for participation in the study. The marijuana users had the following additional inclusion criteria: (i) marijuana use at least 5 days per week for at least 2 years and (ii) negative urine toxicology screen for other drugs of abuse (e.g. cocaine, amphetamines, opiates and benzodiazepines). The comparison subjects were required to have (i) no history of regular marijuana use and (ii) negative urine toxicology screen for drugs of abuse (e.g. cocaine, amphetamines, marijuana, opiates and benzodiazepines). Subjects with any of the following criteria were excluded: (i) HIV-1 positive (tested during screening); (ii) history of co-morbid psychiatric illness that may confound the analysis of the study (e.g. schizophrenia, bipolar disorder, major depression); (iii) confounding neurological disorder [e.g. multiple sclerosis, attention deficit hyperactivity disorder (ADHD), degenerative brain diseases, brain infections, neoplasms or cerebral palsy]; (iv) abnormal screening laboratory tests or abnormal EKG that might affect brain imaging measures (e.g. severe anaemia with haematocrit <32% or significant renal or hepatic dysfunction); (v) current or history of other drug-dependence according to the Diagnostic Statistical Manual (DSM)-IV criteria (including methamphetamine, ecstasy, cocaine, alcohol, opiates and barbiturates, but not nicotine); (vi) history of head trauma with loss of consciousness for >30 min; (vii) for female subjects, pregnancy or breastfeeding; (viii) contraindication for MR studies, such as presence of ferromagnetic substances or electronic implants in the body (e.g. pacemaker, surgical clips, metallic pumps, etc.) or significant claustrophobia; and (ix) less than 8th grade level of English reading skills, to ensure that the subjects could understand the written consent and perform the neuropsychological tests.

Each subject was interviewed for their detailed drug-use history and completed the Addiction Severity Index (ASI). The cumulative THC exposure was estimated from the subject-reported marijuana use, including duration, frequency, amount and route of administration. Marijuana was assumed to contain 5% THC; a joint, bowl or bong was assumed to contain 0.5 g marijuana, while a blunt was considered to contain 1 g of marijuana and a single one ‘hit’ was assumed to contain 15% of the THC of a joint. For example, a subject who reported smoking a half joint on weekends (8–12 days/month) from age 15–22 and who now (and for the last 8 years) takes 2 ‘hits’ from a bong each day would have 96 months × 0.5 joint/use × 10 uses/month × 0.05 g THC/joint = 24 g THC plus 96 months × 0.15 joint × 2 use/day × 30 days/month × 0.05 g THC/joint = 43 g THC for an estimated lifetime total of 67 g of THC. Urine toxicology tests were also performed on the day of the fMRI scan using the DOA-254 Multi-(5) Drug Screen Test Panel (THC sensitivity 50 ng/ml) from Sacks Medical Corp (Evans City, PA).

Neuropsychological tests

All subjects were evaluated with a battery of neuropsychological tests sensitive to functional deficits related to injury in the frontal lobe and basal ganglia (two regions often affected in substance use), as well as the cerebellum. The battery included testing of (i) attention/concentration/working memory (with measures for speed of information processing): California Computerized Assessments Package (CalCAP, Norland Software, Los Angeles, CA, USA; customized expanded version); Paced Auditory Serial Addition Test (PASAT); Trail Making A & B; WAIS-III Digit Span; Letter–Number Sequencing; Arithmetic; (ii) episodic and procedural memory: Rey Auditory Verbal Learning Test; Rey Osterrieth Complex Figure Test; (iii) psychomotor speed: Symbol Digit Modalities; (iv) fine motor speed: Grooved Pegboard; (v) gross motor functioning: Timed Gait; (vi) executive systems functioning: Controlled Oral Word Association (FAS); Ruff Figural Fluency Test; Stroop Colour-Interference Test; (vii) mood: Center for Epidemiologic Studies—Depression Scale (CES-D); Symptoms Checklist (SCL-90); and (viii) estimate of intellectual functioning: New Adult Reading Test Revised (NART-R) (as an estimate of pre-morbid intellectual function). These tests were performed within 1 month of the fMRI studies.

Activation paradigm

Each subject completed the fMRI studies during a set of non-verbal visual-attention tasks with variable levels of difficulty. The tasks required mental tracking of multiple targets (2, 3 or 4 balls) amongst 10 balls moving randomly (in a simulated Brownian motion) and colliding with each other (Culham et al., 2001; Jovicich et al., 2001); see Fig. 1. The movies showing the stimuli were created in MATLAB and presented to the subjects on MRI-compatible LCD goggles connected to a computer, which was also used to record events from a push-button. The response button events were used to determine task performance (% accuracy and reaction times). An MRI trigger pulse was used to synchronize the stimulus display software with the MR acquisition.

Fig. 1

Schematic of the visual-attention paradigm. Subjects are instructed to track 2, 3 or 4 of the 10 moving balls once the target balls are identified (4 are highlighted in the figure), and push a response button if the same target balls are re-highlighted. Each sequence of ball tracking (11.75 s) is repeated 5 times (total 58.75 s). During ‘Do not track’ periods (bottom row), subjects view the balls passively while fixating on the centre cross. All 10 balls move and stop in the same manner as during active tracking for 11.75 × 5 = 58.75 s; however, no balls are highlighted during this period. These block-design sequences alternate from tracking to passive viewing every 60 s (total time: 2 min for each tracking and passive viewing block × 3 blocks = 6 min per task).

MRI acquisition

Before undergoing the MR studies, each subject was briefly trained to ensure that he or she was able to perform the tasks correctly. The studies were performed on a 4 T whole-body Varian/Siemens MRI scanner, equipped with whole-body SONATA gradients. T1- and T2-weighted images were acquired before the functional study, using the 3D-MDEFT (Lee et al., 1995) [echo time (TE)/repetition time (TR) = 7/15 ms, 0.94 × 0.94 × 3 mm3 resolution, axial, 256 × 192 × 48 matrix, 8 min] and hyperecho (Hennig and Scheffler, 2001) (TE/TR = 68/8000 ms, echo train length = 16, 256 × 256 matrix, 28 coronal slices, 0.86 × 0.86 mm in-plane resolution, 5 mm thickness, 1 mm gap, 2 min) sequences. BOLD signal changes were measured using a T2*-weighted single-shot gradient-echo echo-planar imaging sequence (TE/TR = 25/3000 ms, 4 mm slice thickness, 1 mm gap, typically 33 coronal slices, 64 × 64 matrix, 3.1 mm in-plane resolution, 90°-flip angle, 124 time points) covering the whole brain. Padding was used to minimize motion. Task performance and subject motion were determined immediately after each fMRI trial to assure performance accuracy better than 80%, and head motion <1-mm translation and <1° rotation in any direction. (Caparelli et al., 2003)

fMRI data pre-processing

A custom program written in IDL was used for image reconstruction, and the statistical parametric mapping package SPM99 (Welcome Department of Cognitive Neurology, London, UK) was used for subsequent analyses. Images were realigned to the first image in the time-series, to correct for head motion. The realigned datasets were normalized to the Talairach frame (Ashburner et al., 1997) using a 3 × 3 × 3 mm3 voxel size and smoothed with an 8-mm Gaussian kernel.

Statistical analyses

Statistical tests on non-imaging data were performed with StatView (SAS Institute Inc., Cary, NC, USA). Corrections for multiple comparisons were performed on neuropsychological tests, and log transformation was performed on clinical variables that were not normally distributed. BOLD fMRI data were analysed using SPM99. First, activation maps for each subject and condition were calculated using a fixed effects model with a box-car design convolved with a canonical haemodynamic response function (HRF), and low- and high-pass filters (1/246 Hz cut-off frequency). Next, random-effects analyses were performed for each task (using the individual BOLD activation maps as input) to determine activation patterns for each task and group (Woods, 1996; Friston et al., 1999). Additionally, a random-effects repeated measures analysis of variance (ANOVA) model was created that included marijuana use status as a between-variable and the attentional load (number of balls tracked) as a repeated measure. This model was used to evaluate the effects of attention (independent of load) and load (independent of attention). The attention effect for each group was defined as the average BOLD signal across all three tasks, while the load effect was defined as the increase in the BOLD signal with task difficulty (from tracking 2 to 3 to 4 balls). Contrasts involving differences in the attention effect between any two groups, as well as contrasts involving the load effect within each group, were limited to a mask based on voxels that showed a significant (uncorrected voxel level P < 0.05) attention effect in either group. Likewise, contrasts involving an interaction between marijuana-use status and load were limited to voxels that showed a significant (uncorrected voxel level P < 0.05) load effect in either group. For all group comparisons, only clusters with size >25 voxels and P < 0.05 (corrected for multiple comparisons) were considered significant (Friston et al., 1994).

To further validate the voxel-by-voxel comparisons in SPM, additional regions-of-interest (ROI) analyses were performed in brain regions that showed group differences (with cluster size >50 voxels on SPM). The ROI volumes were 0.729 cm3 (cube of 27 SPM voxels), with the centre of each ROI at the cluster maximum. Subsequent statistical analyses for the ROI measurements were performed in StatView (SAS Institute) using repeated measure ANOVA; P = 0.05 was considered significant.

Correlation analyses between BOLD signals in brain regions that showed group differences and estimated cumulative drug use (measured both in equivalent lifetime joints and grams of THC), age of first marijuana use and length of abstinence (log transformed) were also performed, using random-effects simple regression analyses in SPM (without bias and corrected for multiple comparisons). ROI values from all regions that showed significant correlations with estimated cumulative drug use and age of first drug use were extracted for exploratory analyses of covariance (ANCOVA) in StatView, using marijuana-use status as a categorical independent variable, and estimated cumulative drug use and age of first marijuana use as continuous independent variables.

Results

Subject characteristics and marijuana usage

Both groups of marijuana users had similar education and estimated verbal intelligence quotient as the comparison subjects (Table 1). These subjects also had relatively normal mental function on Mini-Mental State Examination and minimal depressive symptoms as assessed by the CES-Depression score (Table 1). Haematocrit was similar amongthe abstinent marijuana users (42.5 ± 1.2%), the active marijuana users (43.9 ± 1.0%) and comparison subjects (42.0 ± 0.7%).

View this table:
Table 1

Neuropsychological test scores in marijuana subjects relative to controls

Mean (SEM)ANOVA P-values
Controls (n = 19)Marijuana (THC negative) (n = 12)Marijuana (THC positive) (n = 12)
Age (years)30.57 ± 1.8329.63 ± 2.5027.91 ± 3.13n.s.
Education (years)15.16 ± 0.3614.50 ± 0.6214.58 ± 0.36n.s.
NART-estimated VIQ116.72 ± 2.08112.34 ± 2.70114.23 ± 1.92n.s.
CES-depression scale6.90 ± 0.989.42 ± 1.9212.67 ± 2.710.08*
MMSE29.47 ± 0.1628.33 ± 0.1928.58 ± 0.340.001
Timed gait (s)8.73 ± 0.328.76 ± 0.568.43 ± 0.86n.s.
Complex Figure Test
    Copy35.00 ± 0.3935.50 ± 0.3635.21 ± 0.39n.s.
    Immediate19.92 ± 1.3924.17 ± 1.9022.04 ± 1.95n.s.
    Delayed19.71 ± 1.1623.33 ± 1.8522.17 ± 1.37n.s.
Grooved Pegboard
    Dominant hand (s)59.42 ± 1.8961.92 ± 1.3962.08 ± 2.72n.s.
    Non-dominant hand (s)67.95 ± 3.5767.67 ± 3.7672.83 ± 4.36n.s.
Symbol Digit Modalities57.67 ± 1.6656.42 ± 2.5155.50 ± 4.43n.s.
Auditory Verbal Learning Test
    Immediate (no. of words)7.63 ± 0.457.67 ± 0.617.00 ± 0.72n.s.
    After five trials (no. of words)13.32 ± 0.2513.17 ± 0.5112.58 ± 0.61n.s.
    Following interference (no. of words)11.98 ± 0.3112.63 ± 0.4311.92 ± 0.560.09*
    Delayed (no. of words)11.68 ± 0.5112.17 ± 0.5911.25 ± 0.73n.s.
Trail Making A (s)27.05 ± 2.0330.67 ± 2.3231.75 ± 3.71n.s.
Trail Making B (s)56.42 ± 4.5759.58 ± 4.2652.75 ± 3.61n.s.
Stroop Colour Interference Test
    Colour (s)61.05 ± 3.0359.33 ± 4.2157.42 ± 2.40n.s.
    Word (s)43.79 ± 1.8044.17 ± 1.9243.00 ± 2.26n.s.
    Interference (s)101.90 ± 4.1597.92 ± 4.62106.92 ± 8.14n.s.
Paced Auditory Serial Addition Testa
    2.4′ pacing42.47 ± 2.6843.00 ± 1.8147.46 ± 2.74n.s.
    2.0′ pacing40.42 ± 2.7337.09 ± 2.5542.73 ± 2.28n.s.
Verbal Fluency (total FAS correct)45.05 ± 2.7945.08 ± 3.6953.92 ± 3.70n.s.
WAIS-III
    Digit span forward11.58 ± 0.4411.08 ± 0.6311.75 ± 0.39n.s.
    Digit span backward8.16 ± 0.498.08 ± 0.618.33 ± 0.53n.s.
    Digit span total19.74 ± 0.8419.17 ± 1.1620.08 ± 0.69n.s.
    Arithmetic16.32 ± 0.8714.75 ± 1.2615.08 ± 0.86n.s.
    Letter–number sequenceb11.83 ± 0.6011.50 ± 0.6611.67 ± 0.69n.s.
Ruff Figural Fluency Test
    Unique designs91.58 ± 5.4184.42 ± 9.7686.42 ± 9.91n.s.
    Preservative errors3.95 ± 1.446.00 ± 3.245.50 ± 2.61n.s.
  • * P ≤ 0.1 indicates trends for significance.

  • a One marijuana user and one comparison subject did not complete this test;

  • b one comparison subject did not complete this test.

There were no significant differences between the active (THC+) and abstinent (THC−) marijuana users on measures of marijuana use: duration or total months used (P = 0.83), frequency or days/month used (P = 0.50), estimated cumulative use in lifetime joints (P = 0.27) or lifetime grams of THC (P = 0.60), log estimated lifetime grams of THC (P = 0.71) or age of first use (P = 0.47). The THC(−) marijuana users smoked 26.7 ± 1.4 (20–30) days per month, for 138.8 ± 24.4 months (48–276 months), which yielded an estimated cumulative lifetime exposure of 2708 ± 553 (924–6900; median, 2400) joints and estimated 168 ± 45 (14–518; median, 118) grams of lifetime THC exposure. The THC(+) marijuana users smoked 27.9 ± 1.1 (20–30) days per month, for 147.6 ± 33.7 months (36–448 months), which yielded an estimated cumulative lifetime exposure of 6047 ± 2908 (522–36 500; median, 2700) joints and estimated 244 ± 135 (13–1680; median, 88) grams of lifetime THC exposure. Age of first marijuana use was 14.7 ± 0.4 (12–16) years for the THC(−) subjects and 15.5 ± 0.9 (9–20) years for the THC(+) subjects. Subjects who were THC(−) last used the drug 38 ± 18 (0.5–156) months before they were assessed, while all the THC(+) subjects used marijuana within 4–24 h of their fMRI scans. None of the THC(+) marijuana users showed signs of intoxication at the time of their evaluations (neuropsychological tests or fMRI).

Three controls, five THC(−) and five THC(+) subjects smoked nicotine cigarettes daily; all were asked to abstain from smoking marijuana for at least 4 h and nicotine for at least 2 h before the fMRI. Other drugs used recreationally (<10 times in life) included opiates [one THC(−) marijuana user], stimulants [amphetamines or cocaine: three controls, six THC(−) and four THC(+) marijuana users], hallucinogens [LSD, mescaline, or psilocybin: one control, seven THC(−) and six THC(+) marijuana users], and club drugs [ketamine, GHB or inhalants: zero control, four THC(−) and three THC(+) marijuana users]. Most subjects used alcohol recreationally; only one control and one THC(+) subject did not have any alcohol consumption, while one control and two THC(−) marijuana subjects had regular alcohol use (>3 days/week). However, none of the subjects were dependent on these drugs (except for nicotine) according to the DSM-IV criteria, and marijuana was the primary drug of choice and was the only illicit drug used on a regular basis by all marijuana subjects. All subjects had negative urine toxicology for other drugs covered by the screening test (see Methods), and only the 12 current marijuana users had positive THC urine assays.

Addiction severity and neuropsychological function

As a group, the marijuana use affected their general function minimally as evaluated on the ASI. THC(+) subjects had the following composite ASI scores: medical, 0.03 ± 0.03; economic, 0.37 ± 0.05; alcohol, 0.07 ± 0.02; drug, 0.08 ± 0.01; legal, 0.01 ± 0.01; family, 0.14 ± 0.04; and psychological 0.08 ± 0.04, while the THC(−) marijuana users had composite ASI scores of medical, 0.01 ± 0.01; economic, 0.37 ± 0.09; alcohol, 0.07 ± 0.02; drug, 0.04 ± 0.01; legal, 0.00 ± 0.00; family, 0.06 ± 0.03; and psychological 0.06 ± 0.03. As expected, only the drug score was significantly different between these two groups (current users greater than past users; P = 0.002).

All three groups had similar performance on neuropsychological tests (Tables 1 and 2). The THC(−) marijuana users showed only trends for better performance on one of the verbal memory tests (AVLT after interference), while the THC(+) marijuana user group showed better performance on the response reversal/visual scanning tests and a trend for faster reaction times on the choice reaction task (Table 2). However, no significant differences were observed on any of these tests after corrections for multiple comparisons.

View this table:
Table 2

CalCAP test scores in marijuana subjects and non-drug users (mean ± SEM) and ANOVA P-values

TasksControl (n = 19)Marijuana, THC negative (n = 12)Marijuana, THC positive (n = 12)ANOVA P-values for reaction times
Reaction times (ms)True positives (%)False positives (%)Reaction times (ms) (% difference)True positives (%)False positives (%)Reaction times (ms) (% difference)True positives (%)False positives (%)
Simple reaction295 ± 13n.a.n.a.331 ± 33 (+12)n.a.n.a.294 ± 25 (0)n.a.n.a.n.s.
Choice reaction (‘7’) Single digit recognition393 ± 699.6 ± 0.40.5 ± 0.2381 ± 9 (−3)99.4 ± 0.60.5 ± 0.2371 ± 9 (−6)100 ± 00.4 ± 0.30.13
1-back cued response (‘X’ only after ‘A’)387 ± 1397.4 ± 1.11.0 ± 0.3373 ± 15 (−4)96.7 ± 1.31.0 ± 0.3368 ± 24 (−5)97.9 ± 1.01.0 ± 0.3n.s.
Sequential # (1-back)543 ± 2088.7 ± 2.71.5 ± 0.3524 ± 21 (−4)92.1 ± 3.60.7 ± 0.3494 ± 23 (−9)94.6 ± 1.91.4 ± 0.4n.s.
Sequential # (1-increment)618 ± 2467.1 ± 3.14.1 ± 0.5592 ± 36 (−4)67.5 ± 7.02.8 ± 0.8586 ± 40 (−5)72.5 ± 5.33.2 ± 0.8n.s.
Sequential # (2-back)720 ± 2678.2 ± 2.92.7 ± 0.6622 ± 31 (−13)89.2 ± 2.91.5 ± 0.5677 ± 56 (−6)79.2 ± 5.72.3 ± 0.4n.s.
Degraded words with distracters504 ± 1099.6 ± 0.42.2 ± 0.7495 ± 14 (−2)98.3 ± 1.71.1 ± 0.3476 ± 15 (−5)98.3 ± 1.21.2 ± 0.3n.s.
Response reversal/visual scanning**597 ± 1391.9 ± 1.71.2 ± 0.5628 ± 29 (+5)82.8 ± 4.50.3 ± 0.2573 ± 24 (−4)95.6 ± 2.40.7 ± 0.4n.s.
Form discrimination662 ± 2183.2 ± 3.24.5 ± 1.1682 ± 21 (+3)80.4 ± 2.13.9 ± 0.9630 ± 34 (−5)84.6 ± 4.74.1 ± 1.4n.s.
  • ** Significant difference was observed for true positives ANOVA P = 0.01, post hoc Control versus MJ(−) P = 0.02; MJ(−) versus MJ(+) P = 0.005.

fMRI

Subject performance

Marijuana users performed better while tracking 2 balls [ANOVA P = 0.04; THC(−) 99.5 ± 0.5%, THC(+) 98.9 ± 0.7%] than comparison subjects (95.2 ± 1.6%) but there were no significant differences in accuracy for tracking 3 balls [THC(−) 93.7 ± 2.7%, THC(+) 94.6 ± 2.1%, comparison subjects 97.4 ± 0.9%] or 4 balls [THC(−) 93.7 ± 2.7%, THC(+) 94.6 ± 2.1%, comparison subjects 92.2 ± 1.6%]. There were no significant differences between groups on reaction times [THC(+) users: tracking 2 balls 836 ± 79 ms; 3 balls 913 ± 62 ms; 4 balls 948 ± 124 ms; THC(−) users: tracking 2 balls 789 ± 118; 3 balls 687 ± 93; 4 balls 893 ± 76; comparison subjects: tracking 2 balls 951 ± 71 ms, 3 balls 938 ± 72 ms, 4 balls 847 ± 72 ms].

Attention effect in abstinent or active marijuana users and non-drug users

The ball-tracking tasks produced robust (T-scores ranged from 7 to 30 or corrected-P ≤ 6.8 × 10−9 to P < 10−16) and characteristic activation in the visual-attention network in all three groups (Fig. 2). Brain regions that were activated across all three ball-tracking tasks (‘attention effect’) include a network comprising bilateral (right greater than left) dorsal medial and lateral prefrontal cortices (PFC), the parietal cortices, occipital regions and the cerebellum (Fig. 2, left). Furthermore, some regions within the attention network, including dorsal and lateral parietal and dorsal prefrontal regions, showed increased activation with increasing number of balls tracked (‘load effect’) (Fig. 2, right).

Fig. 2

Statistical parametric maps of BOLD signals in abstinent marijuana users (THC−), active marijuana users (THC+) and non-drug user comparison subjects. Surface maps demonstrate the effect of attention (independent of load, left) and the effect of attentional load (increasing difficulty from tracking 2 to 3 to 4 balls, right) on repeated measures ANOVA (random-effects analyses) for each group. See Table 3 for P-values and T-scores for brain regions with attention effects, and Table 4 for regions with load effects. DLPFC: dorsal lateral prefrontal cortex; IFG: inferior frontal gyrus; MFG: middle frontal gyrus; SFG: superior frontal gyrus; PPC: posterior parietal cortices; DMPL: dorsomedial parietal lobule; MT/V5: motion detection area.

Compared with control subjects, both abstinent and active marijuana users showed less activation in significant portions of the attention network (Fig. 3, red clusters; Table 3). Abstinent marijuana users showed less activation (attention effect) in large regions in the right prefrontal [middle frontal gyrus (MFG)] and left prefrontal cortex [superior frontal gyrus (SFG) and inferior frontal gyrus (IFG)], the right dorsal medial parietal (DMP) lobe and the cerebellum (declive), totaling 5350 voxels (Fig. 3, top left; Table 3), but greater attention effect in several small regions spanning the frontal, parietal, temporal [superior temporal gyrus (STG)] and occipital brain regions, totaling only 207 voxels (Fig. 3, blue clusters in top left; Table 3).

Fig. 3

Left: Surface rendered maps showing significant group differences on the attention effect (cluster level: P-corrected < 0.005, cluster size ≥ 25 voxels; voxel level T-scores > 4.72; (P-uncorrected < 0.0001 within the significant clusters). Both marijuana groups activate less within the normal attention network, especially in the dorsal parietal regions, right dorsal and inferior lateral PFC and the medial cerebellum (red regions). Instead, the marijuana users activate more than non-drug users in several small brain regions outside the normal attention network (blue areas for the abstinent subjects and green for the active users). See also Table 3 for P-values, T-scores and coordinates in the cluster maximum of these regions. Right: Bar graphs showing ROI measurements (group mean ± SEM) in brain regions that showed group differences (P < 0.05) in attention effect on both the SPM and ROI analyses (repeated measure ANOVA, with number of balls tracked as a within-subject measure and the group status as a between-subject variable). In each bar graph, the first three sets show regions where marijuana (MJ) users (blue or green bars) have less activation than controls (red bars) while the last three sets show regions where the MJ users have greater activation than controls. 2b, 3b, and 4b denote tracking 2, 3, or 4 balls; R = right; L = left; MFG: middle frontal gyrus; IFG: inferior frontal gyrus; Cere: cerebellum (posterior declive); PCG: post-central gyrus; SFG: superior frontal gyrus; OL: occipital lingual; PSG: parietal subgyral; DMP: dorsal medial parietal; Prec: precuneus; DPC: dorsal parietal cortex; LimU: limbic uncus; FSG: frontal subgyral; STG: superior temporal gyrus.

View this table:
Table 3

Brain regions showing attention effects across all three visual-attention tasks that showed group differences (repeated measures ANOVA, T ≥ 4.72 and voxel threshold ≥ 25 voxels or 0.67 ml)

Brain region(s) in clusterTalairach coordinateaCluster levelVoxel level T-score
x, y, z (mm)Corrected P-valueNo. of voxels
Attention effect differences: Controls > THC−
L Superior frontal gyrus−21, −3, 636.80E−091597.00
L Inferior frontal gyrus−36, 21, −33.33E−06836.76
R Middle frontal gyrus27, 3, 51<E−16176613.60
R Dorsal medial parietal lobe3, −48, 69<E−1699913.16
R Cerebellum (posterior declive)30, −60, −18<E−1623439.09
Attention effect differences: THC− > Controls
L Frontal subgyral−27, −6, 307.15E−04325.77
R Parietal subgyral30, −48, 331.38E−05687.02
R Superior temporal gyrus39, 15, −274.18E−05578.25
L Occipital lingual gyrus (BA 18)−18, −78, −68.79E−05509.31
Attention effect differences: Controls > THC+
R Superior frontal gyrus12, 60, 9<E−16132210.93
Left dorsal parietal cortex−27, −54, 63<E−1669810.84
R cerebellum27, −54, −51<E−1612629.52
L Parietal post-central gyrus (BA 40)−39, −33, 542.71E−04405.69
Attention effect differences: THC+ > Controls
L Parietal precuneus−21, −75, 362.26E−05638.24
L Occipital lingual gyrus−15, −75, −61.85E−05657.62
L Limbic uncus−21, 0, −243.40E−05596.18
Attention effect differences: THC− > THC+
R Parietal subgyral24, −42, 421.68E−041056.38
L Medial frontal gyrus−12, 60, 95.21E−04875.56
Attention effect differences: THC+ > THC−
R Superior frontal gyrus6, 12, 51<E−1687610.09
Dorsal medial parietal cortex6, −54, 661.22E−051518.25
L Parietal precuneus−18, −75, 367.32E−051196.78
R Posterior cerebellum (declive)21, −66, −121.95E−082856.36
  • a Talairach coordinates indicate the location(s) of the maximum %BOLD signal within each cluster.

Active marijuana users also showed less activation than control subjects in large areas of the right frontal, right and left dorsal parietal, and right cerebellar regions, total of 3322 voxels (Fig. 3, bottom left, red clusters; Table 3) and greater activation only in several small regions, including left precuneus, left occipital lingual gyrus (OLG) and the left limbic uncus, total of 187 voxels (Fig. 3, bottom left, green clusters; Table 3).

Active marijuana users had greater activation than abstinent marijuana users in the right SFG, the parietal cortex (dorsal medial and left precuneus) and the right cerebellum (declive), totaling 1431 voxels, and only minimal increase in activation in two small parietal and frontal regions, total of 192 voxels (Fig. 4, left; Table 3).

Fig. 4

Left: Surface maps showing brain regions with greater activation in marijuana (MJ) users with positive urine for THC (THC+) than those with negative urine toxicology (THC−) (green areas, T-scores 3 to 10, P-values range from <10−5 to <10−16) and a small region where the abstinent users had greater activation (blue). Right: Bar graphs of %BOLD signal changes (group mean ± SEM) in brain regions that showed significant group differences during the ball-tracking tasks (using repeated measure ANOVA). The abstinent MJ users (blue bars) show higher activation in the right parietal subgyral (RPSG) region and the left medial frontal gyrus (LMFG) while the active MJ users show higher activation in the other four regions. RSFG: right SFG; DMPa: dorsal medial parietal; LPrec: left precuneus; RCereb: right cerebellum (declive). 2b, 3b, and 4b indicate tracking 2, 3 or 4 balls.

These SPM results were confirmed by ROI analyses (bar graphs in Figs 3 and 4). Abstinent marijuana users had reduced BOLD signals during attention tasks (repeated measure ANOVA) compared with non-drug users in all regions that were found to be reduced on SPM, except for three regions: MFG (27, −3, 51), left SFG (21,−3, 63), the right STG (39, 15, −27); see Fig. 3, right upper bar graphs. Similarly, all but one region [left parietal post-central gyrus (PCG) at −39, −33, 54] that showed significantly reduced attention effects on SPM in active marijuana users compared with controls were also reduced on ROI analyses (Fig. 3, bottom right bar graphs). The ROI analyses also confirmed all group differences between abstinent and active marijuana users identified in SPM (Fig. 4).

Attentional load effects in abstinent or active marijuana users versus non-drug users

Compared with non-drug users, abstinent marijuana users showed minimal load effects in several small regions (total 97 voxels) during these attention tasks (Table 4, Figs 2 and 5). In contrast, active marijuana users showed similar load-dependent increases in brain activation as the controls in various brain regions within the attention network, including bilateral frontal, parietal and occipital regions, as well as cerebellum and medulla, with 994 voxels in the controls and 931 voxels in the active marijuana users (Fig. 5, top left; Table 4). Regions with a load effect in the active marijuana users were more numerous and less confluent (cluster sizes were 29–209 voxels) compared with control subjects, which showed load-dependent changes in fewer regions, including two larger clusters (419 and 525 voxels) comprising primarily the parietal cortices and small regions of the frontal lobe (Fig. 5, top left; Table 4).

Fig. 5

Left top: Glass brain views showing significant attentional load effect (independent of attention) in each group [cluster level: P-corrected < 0.001, and voxel level T-scores > 3.22 (P-corrected < 0.05) within significant clusters; cluster size > 25 voxels]. Left bottom: Brain regions showing group differences in load effects [repeated measures ANOVA; cluster level: P-corrected < 0.05, and voxel level T-scores > 2 (P-uncorrected < 0.05) within the significant clusters; cluster size > 25 voxels]. See also Table 4 for P-values and T-scores corresponding to these regions. Right: Bar graphs of ROI measurements (group mean ± SEM) in brain regions that showed group differences in load-dependent increases in brain activation on the SPM analyses. Each of the brain regions shown in the bar graphs had significant group differences (P < 0.05) on load-dependent changes in activation. The significance level of P < 0.05 reflects the significance of the interaction between the number of balls tracked (load, repeated measure variable) and the group status (between-subject variable) for each brain region, using a repeated measure ANOVA. SPL: superior parietal lobule; IPL: inferior parietal lobule; MFG: middle frontal gyrus; CereP: cerebellar pyramis; PCG: post-central gyrus; IFG: inferior frontal gyrus; PSG: parietal subgyral; OLG: occipital lingual gyrus; CereV: cerebellar vermis; SFG: superior frontal gyrus.

View this table:
Table 4

Brain regions that showed significant load effects (voxel T ≥ 3.21 or P ≤ 0.001 and voxel threshold K ≥ 25 voxels or 0.67 ml) and regions that showed group differences in the load effects (from repeated measure ANOVA, voxel T ≥ 2 and voxel threshold ≥ 25 voxels)

Brain region(s) in clusterTalairach coordinateax, y, z (mm)Cluster levelVoxel level T-score
Corrected P-valueNo.  of voxels
Regions showing load effect in Controls
R Superior parietal lobule27, −57, 57<E−165255.74
L Parietal subgyral−24, −45, 515.55E−164195.08
R Frontal subgyral21, 6, 573.49E−04504.09
Regions showing load effect in THC− subjects
R Superior frontal gyrus12, 18, 548.65E−03253.89
R Inferior parietal lobule (BA40)45, −33, 362.50E−03343.82
L Posterior cerebellar tonsil−33, −57, −391.49E−03383.62
Regions showing load effect in THC+ subjects
R Superior frontal gyrus6, 12, 511.31E−03394.55
R Middle frontal gyrus30, 3, 544.90E−03293.53
R Inferior frontal gyrus48, 9, 244.41E−04484.25
L Middle frontal gyrus−27, 0, 571.16E−03404.37
R Parietal subgyral33, −72, 307.21E−081395.54
R Superior parietal lobule (BA7)33, −48, 602.13E−05765.02
L Parietal post-central gyrus (BA5)−27, −42, 663.37E−102094.26
R Occipital lingual gyrus (BA18)12, −72, −32.36E−05754.31
L Occipital subgyral−39, −66, −33.91E−071194.69
R Middle occipital gyrus45, −63, −34.41E−04484.00
L Posterior cerebellar pyramis−30, −63, −271.93E−03363.63
R Medulla9, −42, −482.89E−05734.24
Load-effect differences: Controls > THC−
R Superior parietal lobule (BA7)24, −54, 600.00012542.91
L Inferior parietal lobule (BA40)−39, −42, 540.00231712.70
Load-effect differences: THC− > Controls
L Medial frontal gyrus (BA10)−6, 60, 120.00431543.49
R Cerebellar pyramis39, −66, −330.00421553.27
R Inferior parietal lobule45, −36, 330.00341602.73
Load-effect differences: Controls > THC+
R Post-central gyrus42, −21, 300.00141832.47
Load effect differences: THC+ > Controls
L Inferior frontal gyrus−42, 15, −120.0031643.51
R Parietal subgyral33, −72, 300.00251683.94
L Occipital lingual gyrus−15, −75, −60.00062083.91
R Cerebellar vermis15, −78, −390.00331613.53
Load-effect differences: THC− > THC+
R Inferior parietal lobule36, −36, 272E−053103.63
R Superior frontal gyrus (BA9)15, 54, 360.00062063.38
Load-effect differences: THC+ > THC−
L Occipital lingual gyrus−12, −75, −35E−052843.64
  • a Talairach coordinates indicate the location(s) of the maximum %BOLD signal within each cluster.

Voxel-by-voxel comparisons showed significant load-effect differences between the three subject groups (Fig. 5, bottom left; Table 4). ROI measurements were also performed in these brain regions to validate and evaluate these differences in load effects (Fig. 5, right bar graphs). While controls showed load-dependent increase in the superior parietal lobe (SPL) and minimal to no increase in the MFG and the right cerebellum, the abstinent marijuana users showed load-dependent decreases in the SPL, but greater load-dependent increases in the MFG and right cerebellum. The active marijuana users showed opposite load-dependent effects compared with controls in many brain regions. While the controls showed load-dependent increase in the right PCG, and load-dependent decreases in four other regions (left inferior frontal gyrus, left parietal cortex, right OLG and right cerebellar vermis), the active marijuana users showed the opposite, with load-dependent decrease in the right PCG and load-dependent increases in these four regions. Load-effect differences were also observed between the active and the abstinent marijuana users (Fig. 5, bottom right bar graph).

Effects of duration of abstinence, age of first use and estimated amount of THC exposure on brain activation

SPM analyses demonstrated that age of first marijuana use correlated positively with brain activation only in the prefrontal and dorsal parietal regions within the normal attention network on almost all tasks, in both active and abstinent marijuana users. Figure 6 shows a positive correlation between age of first use and BOLD signals in the right prefrontal region (30, 54, 21), which is within the normal attention network. In addition, age of first marijuana use correlated inversely with BOLD signals in the cerebellar culmen (12, −60, −9; cluster size, 120 voxels; cluster P-value-corrected = 0.002), which is within a region that showed greater activation in the active marijuana users. Furthermore, estimated cumulative lifetime marijuana exposure in both marijuana user groups correlated inversely with %BOLD signal during 3-ball tracking in the cerebellar vermis and adjacent right medial cerebellar region (18, −48, −24; cluster size, 337 voxels; corrected-P = 2 × 0−7) in both marijuana groups (using active versus abstinence as a covariate). Similar inverse correlations between cumulative THC exposure and brain activation were observed while tracking 2 balls in the same right cerebellum region [(12, −54, −21); r = 0.74; cluster size, 531 voxels; corrected-P = 4 × 10−10; Fig. 6].

Fig. 6

Top: SPM maps (left) and scatterplot (right) showing a positive correlation between %BOLD signals in the right MFG and age of first use of marijuana in both abstinent and active users while tracking 2-balls. Middle: SPM maps and scatterplot showing inverse correlation between the estimated lifetime THC used and %BOLD signals in the right cerebellar dentate region during 2-ball tracking. The graphs show the corresponding individual data points from the cluster maximum, but the significance remains at the cluster level and with ROI analyses. Similar results were observed for correlation of cumulative lifetime THC used and brain activation during 3-ball tracking (data not shown; see Results section). All regression analyses were performed with ANCOVA, using drug-use status as a covariate. Bottom: SPM maps and scatterplot showing a positive correlation between BOLD signals activated in the cerebellar dentate and duration of abstinence (log transformed) during 4-ball tracking. The red line indicates the mean value of the BOLD signals, in the same region, from the control subjects.

Finally, in the abstinent marijuana users, we observed a positive correlation between the duration of abstinence and activated BOLD signal in the right prefrontal region and the cerebellum, which are parts of the normal attention network (Fig. 6, left bottom SPM maps). Simple linear regression analyses of ROI data confirmed these SPM analyses. In particular, there was a significant positive correlation between the duration of abstinence (log transformed) with activated BOLD signals in the right prefrontal region (33, 51, 3; cluster size, 212 voxels; corrected-P = 0.01) and in the cerebellum (Fig. 6, right bottom graph). Compared with the mean BOLD activation of the control subjects, THC subjects with shorter abstinence periods had reduced BOLD signals, while BOLD signals increased (normalized) with longer abstinence (Fig. 6, bottom right).

Additionally, ROI data were tested with ANCOVA models that included age of first marijuana use, cumulative THC use and abstinence status. On these ANCOVAs, the relationship between the BOLD signal in the right MFG and ‘age of first marijuana use’ (Fig. 6, top right) did not demonstrate interactions with cumulative THC exposure or with abstinence status. Conversely, the BOLD signal in the cerebellum (Fig. 6, centre right) demonstrated an inverse correlation both with cumulative THC exposure (P = 0.01) and with ‘age of first use’ (P = 0.036). There was an interaction between these two variables, but not with ‘abstinence status’. Specifically, those with earlier age of first use had a steeper decline of cerebellar BOLD signal with cumulative THC exposure than subjects with later age of first use. However, since both factors showed inverse correlations and because of our limited power to assess these interactions, we illustrate only simple regression lines of the BOLD signal in the cerebellum on cumulative THC exposure (Fig. 6, centre right). These results remained significant on a (non-parametric) Spearman correlation (rho = −0.55, P = 0.009). Importantly, the age of first use did not correlate with the cumulative THC exposure (r = 0.067).

Discussion

Chronic marijuana users, both active and abstinent, demonstrated an altered brain activation network compared with non-drug users, consistent with our first hypothesis. Specifically, marijuana subjects showed decreased activation in the normal attention network (right prefrontal region, DMP region and especially the medial cerebellum) and greater activation in several smaller, possibly compensatory, brain regions throughout the frontal, posterior parietal, occipital and cerebellar regions. Furthermore, chronic marijuana users showed normal performance on a battery of neuropsychological tests and normal task performance during the fMRI, despite the altered brain activation. This finding suggests neuroadaptation to chronic marijuana use as predicted by our third hypothesis.

Regarding hypothesis 2, active users indeed demonstrated greater load-dependent changes in many more brain regions (i.e. reserve regions) than the abstinent marijuana users. However, compared with the non-drug users, the abstinent users had less, rather than an intermediate level of, load-dependent activation in brain regions normally involved in attentional modulation, such as the dorsomedial, superior and inferior parietal brain regions (Jovicich et al., 2001; Tomasi et al., 2004). Both marijuana groups showed load-dependent increases in activation in brain regions not associated with load effects in the controls, including the left frontal, right inferior parietal and the right cerebellum regions. These findings further demonstrate that chronic marijuana use, regardless of abstinence status, is associated with abnormal brain activation, both in terms of amplitude and location. Active marijuana users showed load-dependent changes in more extensive regions in bilateral frontal, parietal, occipital and cerebellar brain regions, while the abstinent marijuana users showed only minimal load-dependent changes in small regions in the right frontal, right parietal and the left cerebellum. These findings suggest that both groups of marijuana users are less able to recruit the normal neural network with increasing attentional load, and that compensatory and adaptive processes might occur to maintain function. These processes lead to altered activation patterns that are different between the active and abstinent users. The lower BOLD signals in the abstinent users might also reflect withdrawal effects since the majority of them were within the first 2 months of abstinence. Furthermore, it remains unclear whether this compensatory process could accommodate even higher levels of cognitive load (i.e. tasks that require even higher levels of attention).

One of these compensatory brain regions includes the precuneus, which has been shown to activate with tasks that required shifting attention (Wenderoth et al., 2005) and in Parkinson patients who had more difficulty in achieving automaticity for a motor task (Wu and Hallett, 2005). Conversely, decreased activation in the precuneus occurs when the subjects performed the same task repeatedly with practice (Schumacher et al., 2005). Activation of the precuneus is also related to reduction in reaction times on serial reaction time tasks (Oishi et al., 2005), and it increased quadratically with verbal working memory load (Kirschen et al., 2005), presumably owing to greater attentional requirement. Taken together, greater activation in the precuneus with increasing visual attentional load in both groups of marijuana users suggests that these subjects require greater attentional modulation in this brain region than the non-drug users. This increased usage of the reserve network might lead to a decreased capacity to perform even more cognitively demanding tasks.

Brain regions with decreased brain activation across both marijuana groups include right lateral prefrontal, DMP and medial cerebellar brain regions. These regions are part of the normal visual-attention network (Jovicich et al., 2001; Chang et al., 2004). The cerebellum, in particular, has been shown to be involved in various attention-requiring tasks (Allen et al., 1997; Le et al., 1998). In particular, some regions of the cerebellar vermis have a high density of dopaminergic receptors and transporters (Melchitzky and Lewis, 2000), which have direct relevance to attention (Anderson et al., 2002). Likewise, the parietal regions are critical for attentional processes (Grosbras et al., 2005; Vandenberghe et al., 2005). An alternative mechanism for decreased BOLD activation in the marijuana users may be an alteration in resting perfusion or function since prior PET studies consistently demonstrated changes in resting blood flow and glucose metabolism in various brain regions, especially the cerebellum, of chronic marijuana users (Volkow et al., 1996; Mathew et al., 1998), even during monitored abstinence (Block et al., 2000b). A recent transcranial Doppler study of marijuana users after 1 month of monitored abstinence further documented increased systolic blood flow velocity and pulsatility index in the middle cerebral and anterior cerebral arteries, which were thought to reflect increased vascular resistance possibly due to altered autoregulation in the distal vessels (Herning et al., 2005). Since our marijuana users showed abnormal brain activation within all three major vascular territories (anterior, middle and posterior cerebral arteries), the abnormal activation may in part be due to the effects of chronic marijuana use on the cerebral vasculature. Several studies have shown that BOLD signal changes during brain activation on fMRI are tightly coupled to cerebral blood flow changes (Hoge et al., 1999; Uludag et al., 2004; Ito et al., 2005), although the relationship between resting blood flow and BOLD activation is less clear. However, it is possible that decreased BOLD signals are partly related to increased resting blood flow; hence a reduced dynamic range for activation in these chronic marijuana users.

The reduced cerebellar activation in our marijuana users is consistent with the finding of attenuated cerebellar activation in a prior fMRI study that evaluated response inhibition using a Go/No-Go task in a cohort of young adults who had prenatal marijuana exposure (Smith et al., 2004). In this earlier fMRI study, however, it is unclear whether the decreased cerebellar activation was associated with chronic marijuana use or with prenatal exposure, since 29 of the 31 participants (16 with prenatal exposure and 15 without) had also used or tried marijuana and 13 subjects had tested positive for cannabis in their urine. It is likely that the decreased cerebellar activation in this earlier study reflected the effects of chronic marijuana use rather than prenatal effects since the prenatal exposure group was using marijuana 6.36 ± 2.7 joints/week as compared with the unexposed group that used marijuana only 0.93 ± 0.67 joints/week (Smith et al., 2004). The finding of cerebellar abnormality in marijuana users is not surprising given that cannabinoid (CB1) receptors are localized in inhibitory interneurons that regulate Purkinje cells (Ashton et al., 2004), which are the principal cerebellar neurons and sole output from the cerebellar cortex. Since both input and output of the cerebellum connect to the contralateral cerebral cortices either through the contralateral brainstem nuclei (in the pons or inferior olive) or via the contralateral red nucleus and the thalamus, it is possible that the abnormal activation in the right cerebellum may be related to the abnormal activation in the left prefrontal region in our marijuana users. The inverse correlation between BOLD signal changes in the cerebellar vermis and estimated lifetime THC exposure supports the interpretation that chronic marijuana use may lead to downregulation of cerebellar CB1 receptors.

All of our marijuana users had used the drug regularly (5–7 days/week) and began their marijuana use in childhood and adolescence (ages 9–20 years). Therefore, it is possible that the reorganized brain network may be partly due to altered brain development since the most dorsal aspects of the frontal and parietal regions, where marijuana users showed decreased activation, demonstrate the greatest age-related decreases in grey matter density during adolescence and early adulthood (Sowell et al., 2003). We hence explored whether age of first marijuana use might be related to abnormal brain activation and found that those who first used the drug later during adolescence had greater activation in the right prefrontal region, which is part of the normal attention network. In contrast, the marijuana users that first used the drug at a younger age had greater activation in the compensatory cerebellar region. Since the cerebellum shows greater age-related growth during early childhood and adolescence than the cerebrum (Castellanos et al., 2002), it might be more susceptible to the effects of THC. The medial cerebellum or the vermis may be especially vulnerable since lesions in this region can lead to increased neuropsychological and psychiatric problems (Steinlin et al., 2003). These findings suggest that marijuana use during early childhood and adolescence might alter brain development.

We further explored whether age of first use is related to the cumulative THC exposure but did not find a correlation. However, the estimated cumulative THC exposure did correlate inversely with BOLD activation, and age of first use also interacted synergistically with cumulative THC exposure on BOLD activation in the medial cerebellum, regardless of the abstinence status. These findings suggest that greater cumulative THC exposure leads to lower brain activation in the medial cerebellum, and this effect is more severe in individuals who first use marijuana at an earlier age.

Compared with abstinent marijuana users who had negative urine toxicology, active marijuana users with positive urine toxicology for cannabinoids showed greater activation in medial frontal regions (left and right MFG, right SFG), parietal regions (dorsal medial and right precuneus) and the right medial cerebellum. These findings suggest that recent marijuana use has residual effects, or alters the network involved in the executive and attention systems. These are also regions that showed load effects in controls, suggesting that active marijuana users activate more of the reserve network. However, subjects with positive and negative urine THC did not differ in their neuropsychological performance, although the sample sizes are small for such a comparison. The prior fMRI study that evaluated chronic marijuana using adolescents with prenatal marijuana exposure did not find brain activation differences in those with positive urine toxicology compared with those without (Smith et al., 2004). However, our findings are analogous to prior PET studies that found decreased baseline blood flow or metabolism in the cerebellum (Volkow et al., 1996; Block et al., 2000b), but increased resting blood flow or metabolism after acute intravenous administration of THC in the cerebellum, frontal lobes (orbitofrontal cortex, prefrontal cortex) and basal ganglia of chronic marijuana users (Volkow et al., 1996; Mathew et al., 1997, 1998). One PET study evaluated the effects of marijuana smoking during an auditory attention task and similarly found increased rCBF after smoking marijuana cigarettes compared with placebo cigarettes in the frontal lobes (orbital, mesial, anterior cingulated), insula, temporal lobes and cerebellum (O'Leary et al., 2002). Furthermore, reduced rCBF was found in temporal lobe auditory regions, visual cortex and regions related to the attention network (parietal, frontal and thalamus) (O'Leary et al., 2002). Another study evaluated the time course of THC effects on rCBF and found persistent elevation of rCBF at 30 and at 120 min (Mathew et al., 2002). Our finding of greater BOLD signals in active marijuana users despite being abstinent for at least 4 h before the fMRI compared with abstinent users suggests a residual effect from recent marijuana use, or a different neuroadaptive state in the active users. The findings that those with longer abstinence had similar BOLD signals to the controls, and that there was a positive correlation between duration of abstinence and the activated BOLD signals in the prefrontal and cerebellar regions, suggest that some of these effects might normalize with prolonged abstinence. However, it would be important to reproduce these findings in a larger cohort and in a longitudinal study. Future studies measuring plasma THC levels and evaluating the same subjects longitudinally during early and late abstinence could provide more insight into the relationships between THC levels, changes in BOLD activation, brain perfusion and the reversibility of these effects.

The current study involved a larger sample size than previous fMRI studies in marijuana users (Kanayama et al., 2004; Pillay et al., 2004; Smith et al., 2004; Gruber and Yurgelun-Todd, 2005) and was performed at higher field strength (4 T compared with 1.5 or 3 T), which generally yielded stronger BOLD signals during task performance and higher significance on statistical tests. Therefore, the findings from the current study clarified some of the previous contradictory results. For example, one study found minimally decreased cingulate and prefrontal activation with a motor task (Pillay et al., 2004), while another study found increased cingulate, prefrontal and basal ganglia activation with a spatial working memory task (Kanayama et al., 2004), although both studies evaluated subjects during early abstinence (4–36 h). One study also found no difference in brain activation, except for lack of deactivation in the hippocampus, between adolescent marijuana users and controls (Jacobsen et al., 2004). Only one study found both regionally increased as well as decreased BOLD signals in the dorsal lateral prefrontal and cingulate brain regions of heavy cannabis smokers during a Stroop Colour Interference Test (Gruber and Yurgelun-Todd, 2005), which is a test of executive function and inhibition, but also requires significant attention. However, the group difference for the brain activation in this Stroop test study was small, perhaps owing to the smaller sample size (nine subjects in each group) and possible motion associated with verbal responses during the task performance. Our current study did not require verbal responses and was performed with careful motion monitoring (Caparelli et al., 2003). Therefore, findings from the current study demonstrate a greater group effect for the brain activation and provide a more complete view of how the neural network is altered during visual attention in marijuana users.

In summary, the altered pattern of brain activation during visual attention in chronic marijuana users and greater activation in a reserve brain network in active marijuana users suggest neuroadaptation in the attention network due to chronic marijuana exposure. However, we cannot rule out the possibility that this abnormal pattern is a pre-morbid condition that existed before marijuana use. The cumulative THC dose-dependent decreases in %BOLD signal in the cerebellar vermis further suggest chronic effects of THC on CB1 receptors in the cerebellum. The decreased BOLD activation may be related to marijuana-induced alteration in resting cerebral perfusion or downregulation of cannabinoid receptors and the downstream effects on other neuroreceptors. The apparent normalization of BOLD signals in the right prefrontal and medial cerebellar regions in those with longer duration of abstinence suggests some reversibility of these brain changes. Future perfusion and receptor binding studies are needed to evaluate the relationship between resting and activated blood flow, specific receptors (e.g. dopaminergic, cannabinoid) and BOLD signals, especially in longitudinal studies of subjects during abstinence.

Acknowledgments

This work was supported in part by funds from NIH (K24-DA16170; K02-DA16991 and T32 DA07316) and the Department of Energy (Office of Environmental Research). We are grateful to the research subjects who participated in this study. We also thank L. Zimmerman for subject recruitment; K. Leckova and K. Warren for subject evaluation; S. Arnold, D. Tomasi and E. Caparelli for some of the fMRI data acquisition; and C. Lozar for some of the image processing.

References

View Abstract