Dose-related neurocognitive effects of marijuana use

Methods.

Results.

Table 1 shows the demographic and drug use characteristics of the marijuana users. When taken as a whole, the entire group consisted of predominantly heavy marijuana users (median joints per week = 35; range 2 to 117). The group was also divided into light, middle, and heavy users by dividing the group using terciles of joints per week smoked (see table 1). Except for years of education, there were no significant differences for any of the subject characteristics listed in table 1 (see also below).

Table 2 summarizes significant dose-related effects on key outcome variables for the regression analyses. The R² total reflects the overall proportion of the variance accounted for by the model after the last significant variable was entered in the equation. The results show both linear and nonlinear dose-response effects (i.e., as joints per week increase, neurocognitive performance declines; p < 0.05). This was found for tests of verbal memory (RAVLT, delayed recall, F[1,21] = 7.30), visual learning and memory (Symbol-Digit Paired Associate Learning, F[1,21] = 6.57), executive functioning (WCST categories completed, F[1,20] = 7.09), psychomotor speed (simple reaction time [CALCAP], F[1,21] = 8.32; complex reaction time–number correct, F[1,21] = 11.96), and manual dexterity (Grooved pegboard–nondominant hand, F[1,21] = 6.55). A significant dose-related effect in the opposite direction (i.e., as joints per week increased, performance increased) was found for the CALCAP–numbers in sequence, false positive responses (F[1,21] = 4.87). Moreover, the models accounted for a moderate to a large amount of variance (19 to 57%) in neurocognitive performance. Duration of use was associated only with a decrease in performance on one test, a test of visuoperception/visuoconstruction (Rey Osterreith Complex Figure–copy, F[1,21] = 4.38). Finally, the combination of amount and duration was not related to performance on any of the tests.

To illustrate differences in neurocognitive performance between the lightest and heaviest marijuana users, the group was divided into three groups based on the amount of marijuana smoked as noted above (see table 1). The light group smoked a mean of 11 ± 4 joints/week (range 2 to 14), the middle group reported smoking a mean of 42 ± 18 joints/week (range 18 to 70), and the heavy marijuana group reported smoking a mean of 94 ± 15 joints/week (range 78 to 117) (see table 1). The groups did not differ significantly on age, Shipley IQ score, number of women and men, duration of marijuana use, and alcohol use (see table 1). However, because the mean Shipley IQ score was different for the light (102), middle (95), and heavy (91) users, we elected to take a conservative approach and analyze the group differences using an analysis of covariance (ANCOVA) with Shipley IQ score as a covariate. The mean performance scores, adjusted for differences in Shipley IQ, are presented in the online supplementary table (available at www.neurology.org). Differences among the three groups were examined with post-hoc t-tests. Comparison of group means shows the heavy users performing worse than the light users on 24/35 (69%) of the neurocognitive performance measurements; this difference was significant on five of the neurocognitive measures. Significant group differences were also found between the light and middle users on four of the tests, and between the middle and heavy users on two of the tests. When the scores of the heavy users were compared to published age-appropriate normative values for each of the tests, scores considered to be clinically relevant (below the ninth percentile for the general population) were found for the WCST–categories completed, Rey Complex Figure (copy and delayed recall), and Finger Tapping (dominant hand).

Interaction effects.

There were four significant interactions involving joints/week and Shipley IQ. A Shipley IQ × joints per week interaction was found for the Stroop (F[1,21] = 10.31). A Shipley IQ × joints per week² interaction was found for Symbol-Digit Paired Associate Learning (F[1,21] = 8.67), reaction time repetition of numbers-correct (F[1,21] = 5.89), and Grooved Pegboard–nondominant hand (F[1,21] = 8.25) (see table 2). In general, individuals with lower Shipley IQ scores (less than 96) showed decreasing cognitive performance with increasing number of joints smoked/week whereas individuals with higher Shipley IQ scores had fewer decrements and better performance with increasing marijuana use. To visualize the joints smoked × Shipley IQ interaction, joints smoked/week was divided into terciles and the mean for each tercile was used for the joints/week × Shipley IQ adjusted plots (figure, A and B). Shipley IQ groups were formed by splitting the group by the median Shipley IQ score of 96.

• Download figure
• Open in new tab
• Download powerpoint

Figure. (A) Relation between amount of marijuana smoked² and Repetition of Numbers Task, number correct for the high Shipley IQ group (squares) and the low Shipley IQ group (circles). (B) Relation between amount of marijuana smoked² and performance on the Stroop task for the high Shipley IQ group (squares) and the low Shipley IQ group (circles). The lower the IQ score the worse the performance. Both A and B reveal a significant joints per week × Shipley IQ interaction.

Discussion.

In very heavy marijuana users, persistent, negative dose-related effects are found on tests measuring verbal and visual memory, executive functioning, visuoperception, psychomotor speed, and manual dexterity. This effect was nonlinear for some tests, suggesting a threshold effect. Although we find a dose-related association between joints per week smoked of marijuana and cognitive decline, duration of use is only associated with performance on one test and a combination of joints/week × duration is not associated with performance on any test. In contrast to previous findings,4 duration is not strongly related to performance. This is probably because our marijuana group has shorter duration of use (4.8 ± 3.1 years, range 2 to 15 years) compared to other samples of marijuana users (7.1 ± 7.9 years, range 2.7 to 31.7 years).4 Additionally, our findings do not confirm previous reports showing resolution of cognitive effects after 24 days of marijuana abstinence.9 This discrepancy may be due to our approach to estimating marijuana use (i.e., joints per week) in contrast to those of other investigators (i.e., duration and frequency).2-5,9⇓⇓⇓⇓ Indeed, joints smoked per week may be a better estimate of total marijuana intake than frequency or duration of use because a marijuana user smoking 10 joints/day for 10 years would probably show greater neurocognitive effects than a marijuana user smoking one joint/day for 10 years.

Heavy marijuana use was associated with lower performance on tests of memory, executive functioning, and manual dexterity. These findings are similar to the findings of others.2,4⇓ The RAVLT delayed memory test shows a significant association with the amount of marijuana smoked and there is a trend showing that heavy users performed below the light users on all measures of verbal learning and memory. In fact, the magnitude of the difference in mean performance between the heavy and light users is substantial (1.0 to 3.3 SD units). However, because the heavy marijuana group can recognize previously learned material (RAVLT–Recognition), this pattern suggests difficulty with information recall, not problems with acquisition or retention of information. This pattern of memory performance is characteristic of subcortical, prefrontal lobe involvement, and normal aging. Visual learning and memory (Symbol-Digit Paired Associate Learning) are also affected by heavy marijuana use.

There was also an association between increasing marijuana use and decreasing executive cognitive functioning. This is apparent on the WCST and the effect sizes are large (4.1 to 4.2 SD units). Poor performance on the WCST indicates difficulty incorporating feedback to guide and change incorrect response selection. The Stroop test requires suppression of a more habitual response in favor of an atypical one (response inhibition) and involves performance monitoring. Performance on the Stroop is affected by marijuana use but only in individuals with lower cognitive reserves, as illustrated by the significant joints per week × Shipley IQ interaction (see the figure, B). This is consistent with the suggestion that individuals with higher intellectual functions, or cognitive reserve, demonstrate a higher threshold for developing neurocognitive deficits after insults to the brain.31 This argument is supported by observations that individuals exposed to solvents,32 aluminum,33 and MDMA (Ecstasy)12 show similar interactions. Difficulties with executive functions indicate a prefrontal lobe dysfunction. The prefrontal lobe is suspected to play an important role in substance abuse/addiction and dysfunction of this region may be associated with perpetuation of self-destructive drug using behavior and resistance to treatment.34

The heavy marijuana users also showed slower reaction times on a simple reaction time test (CALCAP). However, when presented with more complex reaction time tests, the difference between the heavy and light marijuana users became less pronounced. The reason for this is unclear. In addition, heavier use of marijuana is not associated with less accurate performance except for the Repetition of Numbers task but only for those with lower Shipley IQ (see the figure, A). No dose-related association is found for false positive responses, a measure of impulsivity. Thus, heavy marijuana use appears to be unrelated to decrements in response time, accuracy, or impulsive performance on complex psychomotor speed/reaction time tests. Heavy marijuana use is also associated with lower performance on both manual dexterity measures (Finger Tapping and Grooved Pegboard).

This study has a number of limitations. Despite making multiple comparisons, we used a p value of 0.05 in order to detect small adverse effects of marijuana on neurocognitive functioning. More adverse associations were found than could be accounted for by chance alone. In addition, although our ability to detect more effects might have been limited by the relatively small sample size, significant effects were found on several measures. Moreover, in the regression analyses, the sizes of the effects were moderate to large (R² = 0.22 to 0.57). Likewise, the heavy users performed two standard deviations or more below the light users on 8/35 (23%) of the measures; this is not a trivial effect. Furthermore, heavy users showed clinically abnormal scores on four of our test measures. Although we use a different estimate of marijuana use (i.e., joints per week) in general, our findings show decrements on similar tests of neuropsychological functioning.4 Those decrements are not secondary to concomitant use of other drugs because participants were excluded for a current or past history of significant use of other substances including alcohol. Although the presence of a dose-related response strengthens the ability to make causal inferences, no definitive statements about causality can be made. This can only be determined with a prospective study of controlled marijuana administration, an approach that would be ethically untenable. Finally, because our primary interest is the determination of a dose-related effect of marijuana on neurocognitive function, we did not include a group of nonusers. We agree with others that a comparison between light users and heavy users is less influenced by confounding variables (i.e., background differences) than a comparison between marijuana users and nonusers.2

It may be difficult to generalize these findings to all users of marijuana because of our strict selection criteria. For example, comorbid psychiatric disorders (i.e., anxiety disorders, major depression) and heavy alcohol use are common in substance abusers. However, we excluded individuals with these disorders to avoid any possible confounding effects on neurocognitive functioning. Finally, it could be argued that the self-reports of marijuana use are inaccurate. The finding of a biologically plausible dose-response suggests that the estimates of drug use were accurate, although this cannot be proven definitively.

The neurocognitive functions most negatively affected were memory, executive function, and manual dexterity. The hippocampus, prefrontal cortex, and cerebellum play a major role in these functions. All of these regions are dense with cannibinoid receptors,35 and these results are biologically plausible because tetrahydrocannabinol (THC) has been shown to cause deleterious effects on these brain regions. Our observations in humans are consistent with studies in laboratory animals that find learning and memory impairments after administration of Δ⁹-THC.36 In rats, morphologic changes are found in the CA1 region of the hippocampus with acute administration of a synthetic THC analogue.37 Damage to the CA1 is also seen after ischemia,38 toxin exposure,39 or traumatic brain injury.40 Therefore, cannabinoids may exert changes in the hippocampus that are similar to those found with other types of brain injury. Changes in CB1 receptors in the hippocampus are also observed in rats after THC administration and are associated with selective deficits in working memory.41 These animal studies provide strong evidence that hippocampal changes might indeed underlie the memory deficits in the current report.

Following marijuana administration, brain images show lower rCBF in the human motor cortex and superior temporal gyrus and higher rCBF in paralimbic brain regions during a dichotic listening task.42 The authors suggest that the increases in rCBF may modulate the intoxicating and mood-related effects of marijuana whereas reductions in task-related rCBF in the temporal lobe regions may account for impaired cognition associated with marijuana intoxication. In 26-hour abstinent marijuana abusers, lower rCBF was found during a resting condition in the ventral prefrontal cortex, bilaterally.43 Thus, when taken together with the evidence of THC-induced hippocampal damage in animals and with the THC-associated neurophysiologic alterations in humans, our current data suggest that THC may exert a significant negative impact on the human brain.

Finally, whereas heavy use of marijuana is associated with decrements in neurocognitive performance, except for a few tests, performance was not clinically abnormal. However, the average age of our group was only 22 years. Given the large extent of the effects, very heavy continuous use of marijuana could produce progressive declines in performance that might reach clinical significance. In fact, because the pattern of performance on the learning and memory tests is consistent with normal age-related declines in the elderly, continued heavy marijuana abuse might result in premature cognitive decline.