The Residual Cognitive Effects of HeavyMarijuana Use in College StudentsHarrison G. Pope, Jr, MD, Deborah Yurgelun-Todd, PhD

Objective.\p=m-\Toassess whether frequent marijuana use is associated withresidual neuropsychological effects.

Design.\p=m-\Single-blindcomparison of regular users vs infrequent users of mari-juana.

Participants.\p=m-\Twosamples of college undergraduates: 65 heavy users, whohad smoked marijuana a median of 29 days in the past 30 days (range, 22 to 30days) and who also displayed cannabinoids in their urine, and 64 light users, whohad smoked a median of 1 day in the last 30 days (range, 0 to 9 days) and whodisplayed no urinary cannabinoids.

Intervention.\p=m-\Subjectsarrived at 2 PM on day 1 of their study visit, thenremained at our center overnight under supervision. Neuropsychological tests wereadministered to all subjects starting at 9 AM on day 2. Thus, all subjects were ab-stinent from marijuana and other drugs for a minimum of 19 hours before testing.

Main Outcome Measures.\p=m-\Subjectsreceived a battery of standard neuropsy-chological tests to assess general intellectual functioning, abstraction ability, sus-tained attention, verbal fluency, and ability to learn and recall new verbal andvisuospatial information.

Results.\p=m-\Heavyusers displayed significantly greater impairment than light us-ers on attentional/executive functions, as evidenced particularly by greater perse-verations on card sorting and reduced learning of word lists. These differences re-mained after controlling for potential confounding variables, such as estimatedlevels of premorbid cognitive functioning, and for use of alcohol and othersubstances in the two groups.

Conclusions.\p=m-\Heavymarijuana use is associated with residual neuropsycho-logical effects even after a day of supervised abstinence from the drug. However,the question remains open as to whether this impairment is due to a residue of drugin the brain, a withdrawal effect from the drug, or a frank neurotoxiceffect of the drug.

(JAMA. 1996;275:521-527)

DURING the last 30 years, more than 40studies have examined the residual ef¬fects of cannabis on neuropsychologicalperformance. We have reviewed thesestudies in detail previously.1 Most .of thestudies produced positive results, butmany failed to find residual impairment incannabis users. These inconsistent find¬ings may reflect several methodologicaldifficulties. First, the term residual effectis ambiguous. It may refer to impairmentsimply due to a drug residue of cannabi-noids lingering in the central nervous sys¬tem (CNS) after acute intoxication hassubsided, or it may refer to actual CNSdamage that persists even after the drug

From the Biological Psychiatry Laboratory (Dr Pope)and the Brain Imaging Center (Dr Yurgelun-Todd),McLean Hospital, Belmont, Mass, and Department ofPsychiatry, Harvard Medical School, Boston, Mass(Drs Pope and Yurgelun-Todd).

Reprint requests to Biological Psychiatry Laboratory,McLean Hospital, 115 Mill St, Belmont, MA 02178 (DrPope).

residue has vanished. Most studies do notdistinguish between the concepts of drugresidue and CNS alteration effects in theiranalyses.1

Second, it is difficult to quantify thetotal amount of cannabis ingested by thesubjects under study. This problem isminimized in laboratory studies, in whicha known amount of cannabis or its activeprinciple, tetrahydrocannabinol, is admin¬istered to subjects, followed by testing atvarious time points after ingestion.1·25However, this design cannot assess theresidual effects (in both senses ofthe term)of months or years of daily cannabis use.Naturalistic studies of actual users avoidthis limitation,1·6"11 but some ofthese stud¬ies have examined users without a strictlysupervised period of abstinence beforetesting.811 Lacking such supervision, it ispossible that some users arrived for test¬ing only a few hours after last using thedrug; therefore, some test results may

have reflected the effects of acute intoxi¬cation rather than residual effects.

Third, comparisons of cannabis userswith nonusers may be influenced by con¬

founding variables. Of particular impor¬tance is the possibility that users mighthave been more impaired than nonuserson neuropsychological functioning evenbefore they began to use cannabis. Onlyone previous, recent study has thoroughlycontrolled for such premorbid differ¬ences.11 Other potential sources of con¬

founding include the possibility that heavycannabis users might display greater ratesof premorbid psychopathology or more

frequent use of other drugs than con¬trols. Also, evidence suggests that heavyusers of cannabis are disproportionatelymale,1 yet we are aware ofonly one studythat has matched users and nonusers onsex11 and of no study that has analyzedthe residual effects of marijuana in menand women separately.1

For editorial comment see 560.

One way to reduce the influence ofconfounding variables, at least to some

extent, is to compare heavy cannabis us¬ers against a control group of light can¬nabis users rather than a control groupof nonusers. Infrequent users might beexpected to differ less from heavy userson some possible confounding variablesthan would control subjects who hadnever used cannabis at all, while still dif¬fering sharply from heavy users on thespecific variable chosen for study, namelyextent of recent cannabis use. We usedthis design, together with additionalmethodological strategies to address theproblems discussed above, to test thehypothesis that frequent marijuana use

might produce residual neuropsychologi-cal effects in college students.

METHODS

SubjectsFrom mid 1991 to late 1993, we ad¬

vertised in student newspapers atcolleges in the Boston, Mass, area forstudents who had smoked marijuana forat least 2 years and who were willing to

by guest on 6 December 2010jama.ama-assn.orgDownloaded from

participate in a 24-hour study for $150compensation. This study was approvedby the McLean Hospital InstitutionalReview Board. Our advertisement in¬dicated that students were eligiblewhether they smoked marijuana fre¬quently or only occasionally. All respon¬dents were then screened on the tele¬phone by the principal investigator, whoassured them of the confidentiality oftheir responses and then asked a seriesof 15 questions about marijuana andother drug use. In fact, only one of thesequestions, the number of days of mari¬juana smoking in the past month, wasused to select subjects; all other ques¬tions were designed simply to mask theactual selection criterion of the study.By this method, we reduced the riskthat students might guess the selectioncriterion and then misrepresent theirextent ofmarijuana use to gain entranceto the study.

Only subjects reporting regular mari¬juana use (a minimum of 22 days of thepast 30 days) and subjects reporting onlyoccasional use (a maximum of 9 days inthe past 30 days) were accepted for thestudy. During the actual recruitmentprocess, we attempted when possible toobtain subjects at the highest and low¬est ends of the frequency range, so thatin practice most of the heavy users re¬cruited had smoked at least 27 days inthe last 30 days, while most light usershad smoked no more than 3 days. Wealso attempted to keep the groups ap¬proximately balanced for sex by tem¬porarily rejecting candidates of one sexif a surplus of that sex developed. Re¬jected callers were simply told that theywere ineligible for unspecified reasons.

Study ProceduresQualifying subjects were asked to ar¬

rive at McLean Hospital by 2 PM on thefirst day of the study. They were in¬formed that they could smoke marijuanaat their usual frequency before this time,but that they would be supervised fol¬lowing their arrival at the hospital toensure that they did not consume anyfurther marijuana or any other drug(aside from their customary amounts oftobacco and caffeine) during the studyperiod. On the afternoon of the first day,after signing informed consent, subjectscompleted a demographic questionnaire,including questions regarding their self-reported high school Scholastic AptitudeTest (SAT) scores, questions regardingsubjective residual effects ofmarijuana,and the questions of the RAND MentalHealth Inventory.12 We then asked sub¬jects about their frequency ofmarijuanause during each year since they firstbegan using the drug and the number ofdays on which they had smoked mari-

juana during the last 30 days. In addi¬tion, we asked the specific days on whichsubjects had smoked and the number ofepisodes of smoking per day in each ofthe last 7 days. Incidents of smokinghad to occur at least 1 hour apart to becounted as separate episodes. We alsoobtained subjects' self-reports of cur¬rent frequency of alcohol, tobacco, andcaffeine use; lifetime use of other typesof illicit drugs; and most recent date ofuse of each other type of illicit drugreported. Finally, we administered theAxis I and Axis II portions of the Struc¬tured Clinical Interview for DSM-III-R(SCID)13 to determine subjects' lifetimehistory of psychiatric disorders by thecriteria of the Diagnostic and Statisti¬cal Manual ofMental Disorders, ThirdEdition, Revised (DSM-III-R).14 Sub¬jects also gave a urine sample, whichwas tested for cannabinoids, as well asfor amphetamines, barbiturates, meth-aqualone, opiates, phencyclidine, cocainemetabolites, and benzodiazepines.

Subjects then spent the remainder ofthat day and night under supervision.They had dinner, watched television,studied, or otherwise occupied them¬selves until bedtime, then slept in a roomthat could be observed by a monitorthrough a one-way mirror to ensure thatthey did not smoke marijuana at night.The following morning they were awak¬ened at about 8:30 AM, permitted to drinktheir customary amount of coffee andhave their customary breakfast, and thenwere administered a battery of neuro-

psychological tests between approxi¬mately 9 AM and 1 PM. Thus, all subjectshad been abstinent from marijuana andother drugs for a minimum of 19 hours,and in practice usually at least 24 hours,before performing their testing.Test Battery

During the testing session, an inves¬tigator blinded to the subject's historyadministered a neuropsychological testbattery. First, the vocabulary subtest ofthe Wechsler Adult Intelligence Test-Revised (WAIS-R)1"' was administeredto obtain an estimate of the subject'sVerbal Intelligence Quotient (VIQ).Thereafter, attention and memory wereassessed with the Stroop test,16 the Wis¬consin Card Sorting Test (WCST),17 theBenton Verbal Fluency Test (BentonVFT),18 the Wechsler Memory Scale(WMS),19 the California Verbal Learn¬ing Test (CVLT),20 and the Rey-Oster-reith Complex Figure Test (ROCF).21 Allscores are reported herein as means andSDs unless otherwise specified.Statistical Analysis

Differences between the heavy andlight users and between other dichoto-

mous groups (ie, heavy vs light male us¬

ers, heavy vs light female users) were

initially assessed using Student's t testfor continuous variables. For RxC con¬

tingency tables, we computed significanceby the exact test,22 and for ordered cat¬egorical data, by the Cochran-Armitagetrend test.21 Subsequently, we performedanalyses of covariance (ANCOVA), inwhich VIQ or SAT scores were enteredinto the model along with frequency ofmarijuana use, to assess the relative sig¬nificance of each of these variables in de¬termining test performance.

RESULTS

We recruited and interviewed 161 stu¬dents. Two were excluded because theydescribed only intermediate marijuana useon interview (approximately 15 days permonth), even though they had reportednearly daily use on prior telephone screen¬

ing. Ofthe 159 subjects accepted, sixwereexcluded from analysis because of pos¬sible confounding variables: one was sus¬

pected of surreptitiously using marijuanadespite supervision during the study; one

reported a prior head injury with pro¬longed loss of consciousness; and four dis¬played current psychiatric disorders likelyto affect neuropsychological test perfor¬mance (two cases of major depression,one case ofbipolar disorder, and one caseof schizotypal personality disorder).

The remaining 153 subjects included85 heavy and 68 light users. However,four of the self-described light users werefound to display cannabinoids in theirurine, and 15 of the self-described heavyusers did not display cannabinoids in theirurine. These 19 subjects were eliminatedfrom the analysis, lest they had misrep¬resented the extent of their recent mari¬juana use. In addition, five of the heavyusers (but none of the light users) dis¬played drugs other than cannabinoids intheir urine (one had traces of phencycli-dine, three had traces of cocaine, and onehad traces ofbenzodiazepines). These sub¬jects were also eliminated, leaving a finaltotal of 65 heavy users (all with canna¬binoids but with no other drugs in theirurine) and 64 light users (none with anydrug in the urine) for comparison.

The heavy and light users were rea¬

sonably well balanced in sexual distribu¬tion (light=31 men, 33 women; heavy=38men, 27 women). As intended by the studydesign, the heavy and light groups were

widely separated in their frequency ofrecent marijuana use. The median (range)number of days on which subjects hadsmoked in the last 30 days was as follows:heavy-using men, 28 (22 to 30); heavy-using women, 29 (22 to 30); light-usingmen, 2 (0 to 9); and light-using women, 1(0 to 9) (P<.001 between heavy and light

by guest on 6 December 2010jama.ama-assn.orgDownloaded from

groups within each sex; P=.96 betweensexes in the light-using group and .80 inthe heavy-using group). No significantdifferences were found between the over¬all heavy and light groups on any demo¬graphic variables, including mean age,year in college, area of residence in theUnited States, ethnicity, stated religiousor political preferences, or type of sec¬

ondary school attended (Table 1). Al¬though marital and domiciliary statuswere not formally assessed, virtually allsubjects in both groups were single andwere living in dormitories or off-campusapartments. Although some subjectsknew one another socially, no subject ineither group was a biological relative ofanother. The two subject groups did notdiffer significantly on most measures offamily background, including parents'marital status, political preference, or oc¬

cupations. However, the heavy users re¬

ported that they came from more afflu¬ent families (P=.01 by trend test) (Table1). No significant differences on any ofthese variables were observed when com¬

parisons were made within the sexes sepa¬rately, save for a significant relation be¬tween heavy use and higher family incomeamong the men (P=.02 by trend test).

Heavy and light users did not differsignificantly on self-reported scores onthe verbal portion of the SAT in highschool (heavy, 489±98, vs light, 512±93;P=.21). However, the heavy users re¬

ported significantly lower scores than lightusers on the SAT quantitative portion(494±96 vs 543±111; P=.02) and on thetotal SAT (983+130 vs 1055+174; P=.02).

No significant differences were foundbetween heavy and light users on life¬time prevalence of any DSM-III-R AxisI or Axis II psychiatric disorder. Nine¬teen (29%) of the heavy users vs 15 (23%)of the light users displayed a history ofother substance dependence or abuse.In a majority of cases, this history wasa period ofheavy alcohol use lasting lessthan 6 months. Only one user displayedcurrent dependence on any other sub¬stance (alcohol dependence in one heavyuser). On the RAND Mental Health In¬ventory, no differences were found onseven of the eight composite scores gen¬erated by this instrument (anxiety, de¬pression, loss of behavioral/emotionalcontrol, emotional ties, life satisfaction,psychological distress, and psychologi¬cal well-being). However, on the remain¬ing scale, general positive affect, whichproduces a range ofscores from 10 to 60,heavy users scored slightly higher thanlight users (40.6±11 vs 37.3+7; P=.04).

In a separate publication,24 we havecompared in detail the heaviest users(those who had smoked daily for 2 yearsor more; n=45) and the lightest users(those who had never smoked more than

Table 1.—Demographic Features of Heavy vs Light Marijuana Users*

Demographic FeatureLight Users

(n=64)Heavy Users

(n=65)Age, y

Median 21

Range 18-28 18-24White, No. (%) 59 (92) 58 (89)Public high school, No. (%) 49 (77) 43 (66)Political orientation, No. (%)

Conservative-moderate 32 (52) 29 (47)Liberal-very liberal 30 (48) 33 (53)

Parents' marital status, No. (%)Married 37 (59) 36 (55)Divorced 18(29) 22 (34)Other (13) 7(11)

Family annual income, No. (%)f<$15000 5(8)$15 000-$30 000 7(11) 7(11)$30 000-$50 000 16(25) 14(22)$50000-$100 000 25 (39) 19(29)>$100 000 11(17) 25 (38)"The number of subjects with missing data is as follows: for political orientation, two light users and three heavy

users; for parents' marital status, one light user.tP=01 by Cochran-Armitage trend test.

10 times in any 1-month period in theirlives; n=44), drawn from our entire origi¬nal sample of 159 subjects. Even thesetwo extreme groups, who were notsubject to any of the exclusion criteriaused in the present study, exhibited al¬most no significant differences on de¬mographic measures, prevalence ofpsy¬chiatric disorders other than substanceabuse, and scores on the RAND MentalHealth Inventory.

The heavy and light users differedslightly on VIQs based on the WAIS-Rvocabulary subscale (heavy users,100.6 ±12.0, vs light users, 104.8 ±13.5;P=.06). Because of this finding, the sta¬tistical significance ofdifferences betweengroups on all subsequent neuropsycho-logical variables was assessed usingANCOVA controlling for VIQ. However,means and SDs for the two groups pre¬sented herein and in the tables are raw

figures; they would be altered only slightlyif adjusted for the covariate.

Looking first at digit span, variouslyconsidered a measure of span of atten¬tion,2'5 a measure of auditory sequentialprocessing,26 or an indirect measure ofanxiety,27 the two groups proved virtu¬ally identical, both in the analysis of thetotal sample and on separate analysis ofthe male subjects and female subjects(total sample: heavy users, 11.0±2.4; lightusers, 10.8±2.3; P=.41). Similarly, on theStroop test, which measures sustainedattention with and without an interfer¬ence condition, heavy and light users againshowed no significant differences on theword reading, color naming, or interfer¬ence subtests. After separate analysis ofmale and female users, no significant dif¬ferences were found between the heavy-

and light-using women, but differencesappeared among the men: male heavyusers were slower to name 100 colorsthan light users (62.5+11.5 vs 56.7±10.5seconds; P=.05), but they also madeslightly fewer errors on this task (medianof one error in each group; range of zeroto four errors in male heavy users andzero to six errors in male light users;P=.05). Male heavy users were also slowerthan light users on the interference por¬tion of the Stroop test (111.7±24.9 vs97.7±22.8 seconds; P=.04).

On the WCST, a test of mental flex¬ibility and abstraction ability,28 an over¬all comparison of heavy vs light usersshowed a significant difference in cat¬egories achieved on deck 1 (Table 2). Oninspection, this difference was causedby a significantly higher number of er¬rors among the heavy users, which inturn proved attributable to a markedlygreater tendency for perseveration.Heavy users produced significantly more

perseverations outside the category (anincorrect response, based on a previ¬ously correct sorting principle, follow¬ing the completion of the sort) andslightly more perseverations inside thecategory (an incorrect response that im¬mediately followed an identical incor¬rect response).29 Combining the totalnumber of perseverations of both typesfor both decks 1 and 2 of the WCST(Figure 1), the heavy users displayedstrikingly more perseverations than thelight users. These findings remainedsimilar when heavy and light users were

compared separately within each sex.On the Benton VFT, scores repre¬

sented the total number of words be¬ginning with the letters F, A, and S

by guest on 6 December 2010jama.ama-assn.orgDownloaded from

Table 2.—Neuropsychological Findings: Wisconsin Card Scoring Test*

Findings

Light Users

Men(n=31)

Women(n=33)

All(n=64)

Men(n=38)

Heavy Users

Women(n=27)

All(n=65)

Deck 1No. of sets completed 4.4 (0.8) 4.5 (0.7) 4.4 (0.8) 4.1 (0.8) 4.2 (0.Í 4.1 (0.8)tItems sorted correctly 53.1 (3.5) 53.5 (3.7) 53.3 (3.6) 50.6 (4.4)t 52.3 (4.6) 51.3(4.5)1"Perseverations outside,

median (range) 3(0-16) 3(0-10) 3(0-16) 5(0-18)}: 4(0-14)} 4(0-18)§Perseveratlons inside,

median (range) 0 (0-5) 0 (0-4) 0 (0-5) 1 (0-6) 1 (0-10) 1 (0-10)Deck 2

No. of sets completed 4.8 (0.6) 4.8 (0.5) 4.8 (0.5) 4.6 (0.6) 4.7 (0.5) 4.6 (0.5)Items sorted correctly 56.6 (2.3) 56.0(2.2) 56.3 (2.2) 55.1 (2.4)t 56.2 (2.3) 55.5 (2.4)Perseverations outside,

median (range) 2 (0-5) 2 (0-7) 2 (0-7) 2 (0-6)} 2 (0-6) 2 (0.6)tPerseverations inside,

median (range) 0(0-1) 0 (0-3) 0 (0-3) 0 (0-2) 0 (0-2) 0 (0-2)

*Number of sets completed and number of items sorted correctly are shown as means with SDs In parentheses. Perseverations are shown as medians with ranges inparentheses. See text for definitions of perseverations "outside" and "inside" the category. Significance is reported for all differences between comparable groups of heavy vs

light users (¡e, men vs men, women vs women, all vs all) by analysis of covariance controlling for Verbal Intelligence Quotient.}P<.05.}P<.01.§P<.001.

generated by the subject when allowed60 seconds per letter. Scores were ad¬justed for age, sex, and education by themethod of Benton and Hamsher.18 Ascore of 31 to 44 is considered normal,and 30 is the threshold of low normal.30Five (8%) of the heavy users, but noneof the light users, scored at or below thisthreshold, a nearly significant difference(P=.06). On ANCOVA, a significant in¬teraction (P=.009) was found betweenVIQ and marijuana use (heavy vs light)in their effect on VFT scores. On analy¬sis of this interaction, we found thatamong subjects with the lowest VIQs,those who were heavy users producedfewer words than those who were lightusers. However, heavy and light userswith average to high VIQs exhibitedvirtually no difference in verbal fluency.This interaction between VIQ and de¬gree of marijuana use was the only suchinteraction attaining a significance of .05or less among any of the overall com¬

parisons between heavy and light usersassessed by ANCOVA in this article.

Turning to measures of memory, thememory quotient on the WMS did notdiffer significantly between the groupswhile controlling for VIQ. Looking at spe¬cific submeasures of the WMS, no sig¬nificant differences were found betweengroups on immediate recall for prose pas¬sages, word pairs, or figures. Delayed re¬call ofprose passages and word pairs wasalso similar between groups, but delayedrecall of figures was significantly reducedamong heavy users. Looking at the sexes

separately, this difference was contrib¬uted entirely by men (heavy users,10.3±3.2 elements recalled; light users,12.2+1.5; P=.007); women exhibited nodifference (heavy users, 11.7±2.3; lightusers, 11.8+2.5; P=.87).

Further examining verbal memory on

the CVLT, heavy and light users dif¬fered significantly on recall of the firstadministration of the word list (Table 3).After five presentations of this list,heavy users still lagged significantly be¬hind light users in their ability to learn,with fewer items recalled in each trialand in the sum ofall five trials (Figure 2).Furthermore, after an interference con¬dition that included the presentation ofa different word list, the heavy users

produced fewer correct words from theoriginal list despite the short delay be¬tween the presentation of the originallist and the recall condition (shown inTable 3 as "short-delay free recall"). Af¬ter a 40-minute delay, heavy users againdisplayed a trend (P=.07) to recall fewerwords from the original list (shown inTable 3 as "long-delay free recall"). Thedifference between heavy and light us¬ers in long-delay recall persisted evenafter cuing (see "long-delay cued recall"in Table 3), suggesting that heavy usersalso experienced more difficulty in re¬trieval of new verbal information. Ofnote, however, there was no decay ofrecall between the short-delay and long-delay conditions in either group. Allthese findings remained similar (thoughwith reduced statistical power) whenthe male and female subgroups were

analyzed separately.Finally, looking at visuospatial memory

with the ROCF, no differences were foundbetween heavy and light users on copy¬ing the complex figure, indicating no fun¬damental impairment ofvisuospatial func¬tioning (Table 4). However, on immediaterecall of the figure, heavy-using menscored significantly lower than light-us¬ing men on the number ofelements of thefigure reproduced. No similar differenceswere observed among women. No fur¬ther loss of information appeared in ei-

Figure 1.—Total perseverations on the WisconsinCard Sorting Test. The error bars represent the 95%confidence interval for each mean value shown.The significance of differences was computed byanalysis of covariance controlling for Verbal Intelli¬gence Quotient. For all heavy users vs all light us¬ers, P<.001; for male heavy users (n=38) vs malelight users (n=31), P=.0O2; for female heavy users

(n=27) vs female light users (n=33), P=.01.

ther sex on delayed copying of the figureafter approximately 40 minutes.

We then performed additional analy¬ses to assess for possible confoundingeffects in the study. We first consideredthe important question of whether thedifferences in observed neuropsychologi¬cal performancebetweenheavy and lightusers might be due to premorbid dif¬ferences between these groups in cog¬nitive function, rather than to an effectof marijuana use.

The lower VIQ scores of the heavyusers, as well as their lower self-reportedSAT scores, raises this possibility. How¬ever, as discussed herein, when we per¬formed ANCOVA controlling for VIQ,we nevertheless continued to find equallyor more strongly significant differences

by guest on 6 December 2010jama.ama-assn.orgDownloaded from

Table 3.—Neuropsychological Findings: California Verbal Learning Test (Selected Items)*

Findings

Light Users

Men(n=31)

Women(n=33)

All(n=64)

Men(n=38)

Heavy Users

Women(n=27)

All(n=65)

Items recalled, trial 1 7.9(1.9) 8.3 (2.6) 1.1 (2.3) 6.6(1.8)} 7.3(1.8) 6.9(1.8)}Items recalled, trial 5 13.7(2.1) 14.3(2.2) 14.0 (2.2) 12.5(2.1) 14.2(1.4) 13.2(2.0)}Short-delay free recall 12.4(2.9) 13.7(2.2) 13.0(2.6) 11.4(2.3) 12.6(2.5) 11.9(2.4)}Long-delay free recall 12.6(3.2) 13.9(2.5) 13.3(2.9) 11.6(2.4) 13.3(2.2) 12.3(2.4)Long-delay cued recall 15.1 (1.3) 15.6 (0.Í 15.3(1.1) 14.8(1.3) 15.0(1.7) 14.9(1.5)}

All data shown are means with SDs in parentheses. Significance is reported for differences between comparable groups of heavy vs light users (ie, men vs men, womenvs women, all vs all) by analysis of covarlance controlling for Verbal Intelligence Quotient.

tP<01.}P<.05.

16

i 14CD co

COS -go|Ò £

Ü 12

10-

• Light Users (n=64)D Heavy Users (n=65)

6Trials 1

Significance P=.001 .04

3

.05

4

.20

5

.05

S

Total ofAll Trials

.007

7065 SS

-605550

CL

> a

Figure 2.—Number of items recalled on successive trials of the California Verbal Learning Test. The error

bars represent the 95% confidence interval for each mean value shown. The significance of differences was

computed by analysis of covariance controlling for Verbal Intelligence Quotient, comparing light users vs

heavy users for each trial and for the sum of all trials.

between groups. We also performedANCOVA examining all the neuropsy¬chological test results while controllingfor verbal, quantitative, and total SATscores. These latter calculations wouldbe expected to produce, if anything, an

overly conservative estimate of the neu¬

ropsychological effects ofmarijuana, sincelower SAT scores may have representedin part a consequence of heavy use inhigh school, rather than a consequence oflower premorbid cognitive abilities. Nev¬ertheless, ANCOVA controlling for SATscores yielded essentially the same re¬sults as that controlling for VIQ. Theseobservations argue against the possibil¬ity that the impairment observed in theheavy users was attributable to some

premorbid deficiency in this group.Also relevant to this question is our

previously published analysis, cited ear¬

lier,24 which found few demographic or

psychiatric differences even when com¬

paring the 45 heaviest users and the 44lightest users in our overall sample of 159subjects. These results also argue againstan innate or premorbid difference betweengroups and thus further suggest that thereduced test performance among heavyusers in the present study was an effectof marijuana itself.

The more frequent abuse ofother sub¬stances among heavy users representsanother potentially confounding vari¬able: greater alcohol use, history of de¬pendence on another substance, or re¬cent use of another substance might allbe expected to contribute to impairedneuropsychological performance. To testfor this possibility, we performed com¬

parisons of heavy vs light users afterdeleting (1) all subjects with a history ofabuse of or dependence on any othersubstance (this yielded a comparison of46 heavy vs 49 light users); (2) all sub¬jects who reported that they currentlyconsumed more than 12 alcoholic drinksper week (yielding 36 heavy vs 52 lightusers); and (3) all subjects who reporteduse of any illicit substance other thanmarijuana within the past month (yield¬ing 35 heavy vs 57 light users). Finally,we excluded all subjects who met any ofthese three criteria, creating particu¬larly "purified" samples of 16 heavy us¬ers vs 36 light users. In all these exer¬

cises, the differences between heavy andlight users remained similar or even wid¬ened. As shown in Table 5, on three offour measures previously found to dif¬ferentiate heavy users from light users,the difference between the "purified"

groups actually exceeds that found be¬tween the original groups.

These analyses suggest that the dif¬ferences observed in this study wereindeed due to residual effects of mari¬juana use, rather than to premorbid dif¬ferences between groups or to the ef¬fects of other substance use. However,this observed impairment might repre¬sent merely a temporary effect, due toeither a drug residue of cannabinoidslingering in the brain1'11 or an abruptwithdrawal from heavy use.3234 Alter¬natively, the impairment might repre¬sent a lasting CNS alteration effect dueto some actual damage to the CNS re¬lated to overall lifetime exposure. In aneffort to distinguish between these pos¬sibilities, we performed two analyses.First, we compared the 13 light userswho at a previous time in their lives hadbeen heavy users (smoking 25 or more

days per month) with the 32 light userswho had never smoked more than ninetimes per month at any time in theirlives. Unexpectedly, the former groupdisplayed a much higher VIQ (114±11vs 103±14; P=.01). We then chose onlythe 13 always-light users with the high¬est VIQs, thereby matching the twogroups on VIQ. On comparing these twogroups of 13, no differences approach¬ing significance were found on any neu¬

ropsychological variables—although itmust be recognized that the sample sizesin this comparison were small.

As a second test, we divided the en¬tire group of 64 light users into threeequal subgroups on the basis of lifetimeuse ofmarijuana (group 1=17 to 70 jointssmoked in their lifetimes, n=21 subjects;group 2=80 to 400 joints in lifetime, n=21;and group 3=400 to 4500 joints inlifetime, n=22). We then performedANCOVA examining each neuropsycho¬logical test score with group status (1,2,or 3) and VIQ entered as covariates.Although many of the test scores were

significantly associated with VIQ in thisanalysis (as might be expected), on noneof the variables was performance sig¬nificantly related to total lifetime con¬

sumption of marijuana. Thus, theseanalyses, though of low statistical power,

by guest on 6 December 2010jama.ama-assn.orgDownloaded from

Table 4.—Neuropsychological Findings: Rey-Osterreith Complex Figure Test*

Findings

Light Users

Men(n=31)

Women(n=33)

All(n=64)

Men(n=38)

Heavy Users

Women("=27)

All(n=65)

Elements of figure correctly identifiedInitial copying 34.6(1.6) 34.0(1.8) 34.3(1.7) 34.2 (2.0) 33.7 (2.2) 34.0(2.1)Immediate recall 23.8 (4.7) 19.5(6.5) 21.6(6.1) 20.1 (6.2)} 20.6 (4.8) 20.3 (5.6)Delayed recall 23.2 (4.5) 19.5(6.3) 21.3(5.8) 20.2(6.1) 20.6 (6.0) 20.4 (6.0)*AII data shown are means with SDs in parentheses. Significance is reported for differences between comparable groups of heavy vs light users (ie, men vs men, women

vs women, all vs all) by analysis of covariance controlling for Verbal Intelligence Quotient.}P<.05.

Table 5.—Neuropsychological Findings: Total Sample vs 'Pure' Cases*

Light Users Heavy Users

All Cases "Pure" Cases Ail Cases "Pure" Cases_Findings_(n=64)_(n=36)_(n=65)_(n=16)Stroop test

Interference subtest, time in seconds_99,6(19.6) 99.6(17.7) 106.8(23.5) 110.5(32.1)Wisconsin Card Sorting Test

Total perseverations, median (range) 5(0.18) 4(0-18) 9(0-28)} 12.5(0.25)}Wechsler Memory Scale

Delayed recall of figures, elements recalled 12.0 (2.1 ) 11.8 (2.6) 10.9 (2.9)} 10.0 (3.9)California Verbal Learning Test

Total items remembered on 5 trials combined 59.2 (8.7) 58.5 (9.0) 55.0 (8.0)§ 56.9 (7.3)*"Pure" cases represent individuals who (1) displayed no current or past history of any substance abuse or

dependence (other than cannabis) by the criteria of the Diagnostic and Statistical Manual of Mental Disorders, ThirdEdition Revised, (2) currently reported drinking fewer than 12 alcoholic drinks per week; and (3) reported no useof any illicit drug other than marijuana during the past month. All data shown are means with SDs in parentheses,with the exception of total perseverations on the Wisconsin Card Sorting Test, which are shown as medians withthe ranges In parentheses. Significance is reported for differences between comparable groups of heavy vs lightusers (ie, all cases vs all cases, "pure" cases vs "pure" cases) by analysis of covarlance controlling for VerbalIntelligence Quotient.

fP<001.}P<.05.§P<.01.

weigh tentatively against a CNS alter¬ation hypothesis.COMMENT

We compared 65 college students whosmoked marijuana regularly with 64 stu¬dents who smoked only occasionally. Theheavy users had smoked a median of 29days (range, 22 to 30 days) in the past 30days, and all displayed cannabinoids intheir urine. Light users had smoked amedian of 1 day (range, 0 to 9 days) inthe past 30 days, and none displayedurinary cannabinoids. Students display¬ing a current psychotic disorder, schizo-typal personality disorder, major mooddisorder, or history of severe head in¬jury with loss of consciousness were ex¬

cluded, as were students currently tak¬ing any medication with psychotropiceffects and those displaying any sub¬stance other than cannabinoids in theirurine. We also attempted to recruitroughly equal numbers of male and fe¬male students in each group so that re¬sults could be analyzed within each sex

separately, as well as for the heavy us¬ers vs the light users as a whole. All 129students received a battery of neuro¬

psychological tests after a supervisedovernight period of abstinence frommarijuana and other drugs of abuse.

The findings on this battery of testssuggest that heavy marijuana use wasassociated with reduced function of the

attentional/executive system, as exhib¬ited by decreased mental flexibility andincreased perseveration on the WCST,and reduced learning, as seen on theCVLT. In addition, verbal fluency onthe Benton VFT was impaired withheavy use, but only in subjects withlower VIQs, perhaps because their flu¬ency was more vulnerable to attentionalimpairment. However, the ability to re¬tain newly learned information after a

temporal delay appeared to remain rela¬tively intact in the heavy users, as ex¬hibited by the absence of decay on thedelayed recall conditions of the CVLTand ROCF. In short, although marijuanause may produce some compromise inmemory function, it appears that theprincipal effect is in the attentional/ex¬ecutive system, with recall memory func¬tions per se remaining relatively unaf¬fected. Our data suggest that multiplebrain systems may be affected by mari¬juana, but the most pronounced effectsmay be in sustained attention, mediatedby brain stem structures, and in thecapacity to shift attention, associatedwith prefrontal cortical regions.

In general, these findings were some¬what more prominent among men thanwomen, even though male heavy usersdid not differ from female heavy usersin their reported recent or lifetime mari¬juana consumption. Nevertheless, thepossibility remains that men were ac-

tually ingesting a greater total dose ofcannabinoids during their individual epi¬sodes of smoking.

The differences between heavy usersand light users did not appear to be dueto fundamental demographic or psycho¬logical differences between these groups,as evidenced by our comparisons in thepresent study and our previously pub¬lished comparison of the heaviest vs thelightest users from our larger originalsample.24 Similarly, the differences ob¬served did not appear attributable topremorbid differences between heavyand light users in neuropsychologicalfunction—since the differences in testscores remained significant, or even in¬creased in significance, when we per¬formed ANCOVA controlling for VIQor for self-reported verbal or quantita¬tive high school SAT scores. Finally, thedifferences did not appear attributableto use of other drugs or large amountsof alcohol, since analyses excluding in¬dividuals with these possible confound¬ing factors failed to change the differ¬ences observed, and in some instanceseven widened them.

Despite these analyses, however, cer¬tain methodological limitations must beacknowledged. First, it must be recog¬nized that a large number of individualcomparisons were performed; therefore,some results may have achieved statis¬tical significance by chance alone. Onthe other hand, the trend of the datawas consistent: virtually every differ¬ence, whether significant or not, was inthe direction of greater impairmentamong heavy users.

Second, it is conceivable that heavyand light users did differ on certainpremorbid cognitive abilities and thatthe analyses controlling for VIQ and SATscores did not fully correct for such dif¬ferences. Only a standardized test, ad¬ministered to all subjects in a uniformmanner before their first marijuana use,would fully address this issue. Only one

published study, to our knowledge, hasutilized such test results11 and, despitethis control, nevertheless found residualimpairment in heavy users similar tothat found in our study.

Third, it might be argued that non-users would have represented a better

by guest on 6 December 2010jama.ama-assn.orgDownloaded from

comparison group than light users. How¬ever, we reasoned that light users mightbe less likely to differ from heavy userson possible confounding variables, suchas aspects of lifestyle, study habits, or

background, than would nonusers. Thus,our comparison, while perhaps produc¬ing a slightly conservative estimate ofthe effects of marijuana, may also beless vulnerable to confounding.

Fourth, most historical data, includ¬ing subject's SAT scores, prior use ofother substances, and even prior mari¬juana use, were obtained from self-re¬ports without external validation. How¬ever, inaccurate self-reports appearunlikely to have imposed marked bias,since subjects would have derived no

particular benefit from distorting theirreports in this study. Further, our con¬clusions would be seriously affected onlyin the unlikely event that subjects' self-reports were distorted in some system¬atic manner that was correlated withneuropsychological test scores.

Fifth, several studies have suggestedthat abrupt discontinuation of heavycannabis use, as was required by our

supervised abstinence period, may pre¬cipitate a withdrawal syndrome char-

acterized by insomnia, restlessness, andirritability.3234 Such a syndrome mighthave influenced our findings; only a studywith a much longer abstinence periodcould fully resolve this issue. Of studiesusing longer abstinence periods, onefound few neuropsychological deficitsamong 37 heavy users after 3 to 4 days,35but another found memory impairmentin 10 users after 6 weeks9 and anotherfound deficits on an auditory selectiveattention task in 28 users after a meanabstinence of about 2 years.36 These lat¬ter two findings suggest a CNS alter¬ation effect of marijuana, not explain¬able by drug withdrawal or by a drugresidue in the CNS.

Sixth, it should be recognized thatour study might have underestimatedthe potential residual effects of mari¬juana because of the young age of thesample. Although many students hadsmoked marijuana daily for 2 years or

more, as noted earlier, none had smokedregularly for more than a decade. Thus,a comparison ofmiddle-aged heavy mari¬juana users, with longer histories ofuse,might produce more robust findings.

In short, the findings of our study,together with those ofother recent, well-

controlled investigations,w1·36 raise sev¬eral questions that must be pursued.These questions include how long mari¬juana-associated neuropsychological im¬pairment lasts, whether it is related tocumulative lifetime use, whether it in¬teracts with normal aging, and to whatextent, if any, it compromises the actualfunctioning of users in their daily lives.Specifically, although marijuana use maynot cause overt psychopathology or se¬rious academic impairment,24·37 subtledrug-induced deficits in sustaining andshifting attention may neverthelesscause important difficulties in adaptingto intellectual and interpersonal tasks.Only by further study of these issuescan it be judged whether neuropsycho¬logical impairment associated with mari¬juana use should be considered a publichealth problem.

This study was supported in part by NationalInstitute of Drug Abuse grant ROI DA06522.

We wish to express our gratitude to DeepVarma for assistance with the design of this study,to Harlyn Aizley, EdM, and Staci Gruber, EdM, fortheir assistance in performing neuropsychologicaltesting on the subjects, to Paul Oliva and DavidAmato, PhD, for assistance with the statisticalanalysis of the data, and to Robert Block, PhD, forhis commentary on the manuscript.

References1. Pope HG Jr, Gruber AJ, Yurgelun-Todd D. Theresidual neuropsychological effects of cannabis: thecurrent status of research. Drug Alcohol Depend.1995;38:25-34.2. Yesavage JA, Leirer VO, Denari M, HollisterLE. Carry-over effects of marijuana intoxicationon aircraft pilot performance: a preliminary report.Am J Psychiatry. 1985;142:1325-1329.3. Chait LD. Subjective and behavioral effects ofmarijuana the morning after smoking. Psychophar-macology. 1990;100:328-333.4. Heishman SJ, Huestis MA, Henningfield JE,Cone EJ. Acute and residual effects of marijuana:profiles of plasma THC levels, physiological, sub-jective, and performance measures. Pharmacol Bio-chem Behav. 1990;37:561-565.5. Leirer VO, Yesavage JA, Morrow DG. Mari-juana carry-over effects on aircraft pilot perfor-mance. Aviat Space Environ Med. 1991;62:221-227.6. Varma VK, Malhotra AK, Dang R, Das K,Nehra R. Cannabis and cognitive functions: a pro-spective study. Drug Alcohol Depend. 1988;21:147\x=req-\152.7. Mendhiratta SS, Varma VK, Dang R, MalhotraAK, Das K, Nehra R. Cannabis and cognitive func-tions: a re-evaluation study. Br J Addict. 1988;83:749-753.8. Page JB, Fletcher J, True WR. Psychosociocul-tural perspectives on chronic cannabis use: the CostaRican follow-up. J Psychoactive Drugs. 1988;20:57-65.9. Schwartz RH, Gruenewald PJ, Klitzner M, Fe-dio P. Short-term memory impairment in cannabis\x=req-\dependent adolescents. AJDC. 1989;143:1214-1219.10. Solowij N, Michie PT, Fox AM. Differentialimpairments of selective attention due to frequencyand duration of cannabis use. Biol Psychiatry. 1995;37:731-739.11. Block RI, Ghoneim MM. Effects ofchronic mari-juana use on human cognition. Psychopharmacol-ogy. 1993;110:219-228.12. RAND. Scoring Manual: Adult Health Status

and Patient Satisfaction Measure Used in RAND'sHealth Insurance Experiment. Santa Monica, Calif:RAND; 1988.13. Spitzer RL, Williams JBW, Gibbon M, First M.Structured Clinical Interview for DSM-III-R, Pa-tient Edition. Washington, DC: American Psychi-atric Press; 1990.14. American Psychiatric Association. Diagnosticand Statistical Manual ofMental Disorders, ThirdEdition, Revised. Washington, DC: American Psy-chiatric Association; 1987.15. Wechsler D. WechslerAdult Intelligence Scale:Revised Manual. Cleveland Ohio: PsychologicalCorp; 1981.16. MacLeod C. Half a century of research on theStroop effect: an integrative review. Psychol Bull.1991;109:163-203.17. Heaton RK. Wisconsin Card Sorting TestManual. Odessa, Fla: Psychological AssessmentResources; 1981.18. Benton AL, Hamsher KdeS. Multilingual Apha-sia Examination. Iowa City: University of Iowa;1978.19. Wechsler D.A standardized memory scale forclinical use. J Clin Psychol. 1945;19:87-95.20. Delis DC, Kramer JH, Kaplan E, Ober BA.California Verbal Learning Test Manual, Version1. San Antonio, Tex: Psychological Corp; 1987.21. Loring DW, Marin RC, Meador KS. Psycho-metric construction of the Rey-Osterreith complexfigure. Arch Clin Neuropsychol. 1990;5:1-14.22. Pagano M, Taylor-Halvorsen K. An algorithmfor finding the exact significance levels in R\m=x\Ccontingency tables. J Am Stat Assoc. 1981;76:931\x=req-\934.23. Mehta CR, Patel NR, Tsiatis AA. Exact sig-nificance test to establish treatment equivalencefor ordered categorical data. Biometrics. 1984;40:819-825.24. Kouri E, Pope HG Jr, Yurgelun-Todd D, Gru-ber S. Attributes of heavy vs. occasional marijuanasmokers in a college population. Biol Psychiatry.

1995;38:475-481.25. Spitz HH. Note on immediate memory for dig-its: invariance over the years. Psychol Bull. 1972;78:183-185.26. Mesulam MM. Principles of Behavioral Neu-rology. Philadelphia, Pa: FA Davis Co; 1985.27. Mueller JE. Test anxiety and the encoding andretrieval of information. In: Sarason IG, ed. TestAnxiety: Theory, Research, and Applications.Hillsdale, NJ: Lawrence Erlbaum Assoc; 1979.28. Milner B. Effects of different brain lesions oncard sorting. Arch Neurol. 1963;9:90-100.29. Levin S, Yurgelun-Todd D, Craft S. Contribu-tions of clinical neuropsychology to the study ofschizophrenia. JAbnorm Psychol. 1989;98:341-356.30. Benton AL, Hamsher KdeS, Varney N, Spreen0. Contributions to Neuropsychological Assess-ment. New York, NY: Oxford University Press;1983.31. Seth R, Sinha S. Chemistry and pharmacologyof cannabis. Prog Drug Res. 1991;36:71-115.32. Jones RT, Benowitz NL, Herning RI. Clinicalrelevance of cannabis tolerance and dependence.J Clin Pharmacol. 1981;21(suppl):143S-152S.33. Mendelson JH, Mello NK, Lex BW, Bavli S.Marijuana withdrawal syndrome in a woman. AmJ Psychiatry. 1984;141:1289-1290.34. RohrJM, Skowlund SW, Martin TE. Withdrawalsequelae to cannabis use. IntJ Addict. 1989;24:627\x=req-\631.35. Rubin V, Comitas L. Psychological assessment.In: Rubin V, Comitas L. Ganja in Jamaica: A Medi-cal-Anthropological Study of Chronic MarijuanaUse. The Hague, the Netherlands: Mouton; 1975:111-119.36. Soiowij N. Do cognitive impairments recover

following cessation of cannabis use? Life Sci. 1995;56:2119-2126.37. Pope HG Jr, Ionescu-Piggia M, Aizley HG,Varma DK. Drug use and life-style among collegeundergraduates in 1989: a comparison with 1969and 1978. Am J Psychiatry. 1990;147:998-1001.

by guest on 6 December 2010jama.ama-assn.orgDownloaded from