Mood Disturbances and Regional Cerebral Metabolic Abnormalities inRecently Abstinent Methamphetamine Abusers

Background  Mood disturbances in methamphetamine (MA) abusers likely influence drug use, but the neurobiological bases for these problems are poorly understood.

Objective  To assess regional brain function and its possible relationships with negative affect in newly abstinent MA abusers.

Design  Two groups were compared by measures of mood and cerebral glucose metabolism ([¹⁸F]fluorodeoxyglucose positron emission tomography) during performance of a vigilance task.

Setting  Participants were recruited from the general community to a research center.

Participants  Seventeen abstaining (4-7 days) MA abusers (6 women) were compared with 18 control subjects (8 women).

Main Outcome Measures  Self-reports of depressive symptoms and anxiety were measured, as were global and relative glucose metabolism in the orbitofrontal, cingulate, lateral prefrontal, and insular cortices and the amygdala, striatum, and cerebellum.

Results  Abusers of MA provided higher self-ratings of depression and anxiety than control subjects and differed significantly in relative regional glucose metabolism: lower in the anterior cingulate and insula and higher in the lateral orbitofrontal area, middle and posterior cingulate, amygdala, ventral striatum, and cerebellum. In MA abusers, self-reports of depressive symptoms covaried positively with relative glucose metabolism in limbic regions (eg, perigenual anterior cingulate gyrus and amygdala) and ratings of state and trait anxiety covaried negatively with relative activity in the anterior cingulate cortex and left insula. Trait anxiety also covaried negatively with relative activity in the orbitofrontal cortex and positively with amygdala activity.

Conclusions  Abusers of MA have abnormalities in brain regions implicated in mood disorders. Relationships between relative glucose metabolism in limbic and paralimbic regions and self-reports of depression and anxiety in MA abusers suggest that these regions are involved in affective dysregulation and may be an important target of intervention for MA dependence.

Abstaining methamphetamine (MA) abusers have mood disturbances¹ that likely reflect neurochemical abnormalities. Animalstudies indicate that MA alters dopaminergic, serotonergic, and nonmonoaminergicsystems,^2-10 andpostmortem tissue from human MA abusers exhibits deficits in striatal dopaminergicmarkers and in orbitofrontal cortical serotonin.¹¹ Invivo studies of abstinent MA abusers also indicate loss of striatal markersfor dopaminergic systems.^12,13

Effects of MA on cerebral metabolism have also been studied. Chronictreatment reduced subcortical glucose metabolism in rats.¹⁴ Inaddition, compared with control subjects, abstinent human MA abusers (abstinentfrom 2 weeks to >2 years) had higher cerebral glucose metabolism,¹⁵ lower levels of N-acetylaspartatein the basal ganglia and frontal white matter, lower total creatinine levelsin the basal ganglia, and higher levels of choline-containing compounds and myo-inositol in the frontal gray matter.¹⁶

The present study aimed to clarify the nature of brain disorder in MAabusers by identifying brain regions whose dysfunction may underlie negativeaffect. Stimulant abusers often enter treatment within their first week ofabstinence. For example, most participants (86/112) entering a research protocolfor treatment of MA dependence provided MA-positive urine samples, indicatingMA use within 72 hours (R.R., unpublished data, September 2002). Because treatmentfor MA abuse almost exclusively involves outpatient methods,¹ thefirst week of abstinence is a crucial determinant of engagement and retentionand, thus, of treatment outcome.¹⁷ We thereforemeasured cerebral metabolism in MA abusers during early abstinence and incomparison subjects using the [¹⁸F]fluorodeoxyglucose (FDG) positronemission tomographic (PET) method.^18,19

Twenty inpatient MA abusers and 22 control subjects gave informed consentand participated in this study, approved by the institutional review boardsof the University of California, Los Angeles, and the Long Beach Departmentof Veterans Affairs Medical Center, Long Beach, Calif. Participants in bothgroups were healthy according to medical history, physical examination findings,and laboratory test results. Use of psychoactive medications and seropositivestatus for human immunodeficiency virus were exclusionary.

As established by the Structured Clinical Interviewfor DSM-IV Axis I Disorders–Patient Edition (SCID-IP, version 2.0),²⁰ current Axis I diagnoses of dependencies on substancesother than MA or nicotine and lifetime Axis I diagnoses unrelated to drugabuse were exclusionary for MA abusers. The same criteria applied for controlsubjects, but MA dependence was not allowed and restrictions on drug dependenceincluded lifetime diagnoses. Personality disorders, other than antisocialpersonality disorder, which was evaluated using the StructuredClinical Interview for DSM-IV Axis II Personality Disorders (SCID II): User'sGuide,²¹ were not exclusionary. Antisocialpersonality disorder was detected in 3 MA abusers.

Brain imaging data from 17 MA abusers and 18 control subjects were analyzedusing statistical parametric mapping (SPM). Seven subjects were excluded forvarious reasons: motion artifact, no magnetic resonance (MR) image, sleepingafter FDG injection, and an inconsistent drug history report. Fourteen MAabusers and 13 control subjects provided measures of absolute global cerebralglucose metabolic rate (CMRglc). Nine were excluded for technical reasons:no arterial catheter, plasma glucose levels higher than 150 mg/dL (8.3 mmol/L),and improper instrument calibration.

The groups did not differ significantly in age or mother's education,but MA abusers had fewer years of education than control subjects (group differencein SPM analysis, t₃₅= 2.90; P = .006; group difference in absolute CMRglc analysis, t₂₅ = 1.85; P = .08) Table 1). Although the groups were similar in handedness²² and sex distribution, the control group includeda larger proportion of non-Hispanic white individuals, whereas white MA abusersreported Hispanic ethnicity more frequently.

Both groups were allowed light alcohol use (equivalent to <7.5 drinksper week). For MA abusers, self-reports of spending $100 or more on MA duringthe month before screening and MA use within 3 days of enrollment (verifiedby urine sample) were required. Control subjects provided urine samples negativefor illicit drugs.

Participants completed self-report questionnaires about drug use (intakequestionnaire and drug use survey) and were administered the Addiction SeverityIndex²³Table 2). The MA abusers had used the drug, on average, for morethan 8 years, consumed about 2 to 4 g/wk, and had used MA on most of the 30days before entering the study. The MA and control groups reported similaralcohol and marijuana use and were generally matched on use of illicit drugsother than MA. Most of the MA abusers smoked cigarettes, but none of the controlsubjects did.

Participants completed the Addiction Severity Index when entering thestudy. Self-ratings of depressive symptoms (Beck Depression Inventory [BDI])²⁴ and anxiety (State-Trait Anxiety Inventory)²⁵ were obtained on the day of PET scanning. Abusersof MA rated their MA craving (visual analog scale)²⁶ within48 hours of scanning.

Structural MR images (3-T; GE Medical Systems, Waukesha, Wis) includedT1-weighted volumetric scans (spoiled gradient-recalled acquisition, 256 ×256 matrix, echo time [TE] = 4 milliseconds, repetition time [TR] = 24 milliseconds,angle = 35°, 1.22-mm slice thickness) that were used for coregistrationwith PET data (see "PET Procedure" subsection). The findings from T2-weighted(spin echo) scans (256 × 256 matrix, TE = 34 milliseconds, TR = 5500seconds, 1 number of excitations, 2-mm slice thickness) from all subjectsincluded were read as clinically normal.

Methamphetamine abusers were tested when abstinent for 4 to 7 days.Catheters were inserted into the antecubital vein and contralateral radialartery for infusing FDG and blood sampling, respectively. The participantwas positioned in the scanner gantry and fitted with a plastic face mask (ScryptonSystems, Annapolis, Md) to minimize head motion. A 3-minute ⁶⁸Getransmission scan verified proper positioning, and a 20-minute ⁶⁸Getransmission scan provided data for attenuation correction.

After subjects were removed from the scanner, they performed a continuousperformance task (CPT) (version 2.26; Sunrise Systems, Pembroke, Mass) usinga laptop computer. The task required discrimination of a target tone (higherpitch) from a sequence (interstimulus interval = 2 seconds) of nontarget tones(lower pitch). Pressing the X signified hearing atarget tone. With the CPT underway, FDG (≤5 mCi [≤185 MBq]) was administeredintravenously. The CPT was stopped 30 minutes later, and the subject was repositionedin the scanner.

Arterial blood samples were taken at 10-minute intervals for 90 minutesafter the FDG injection, and plasma from these samples was assayed for radioactivity(Cobra II Auto-Gamma; Packard Instruments, Downers Grove, Ill) and glucose.The plasma activity curve was fit according to an analytic procedure to determinethe integral of plasma-specific activity²⁷ foruse in an operational equation.²⁸

Brain images were acquired in 3-dimensional mode (Siemens ECAT EXACTHR+ tomograph; CTI, Knoxville, Tenn) for 30 minutes, starting 50 minutes afterthe FDG injection. We reconstructed 128 × 128-pixel images using a Hannfilter (cut-off frequency = 0.5 cycles per pixel). The average transverseresolutions at 1 and 10 cm from the center of the field of view (measuredusing an ¹⁸F line source) were 6.52 and 7.16 mm (full width athalf maximum), and the average axial resolutions were 3.72 and 5.64 mm at0 cm and 10 cm from the center of the field of view, respectively.

The CMRglc was calculated from the modeled images using MEDx software(Sensor Systems, Sterling, Va). An edge-detection algorithm defined the braincontour in each transaxial plane. A 3-dimensional contour was then derivedfrom the sum of the 2-dimensional slices, and mean CMRglc was calculated.Voxels exhibiting metabolic rates less than or equal to 4 mg of glucose per100 g per minute were excluded to minimize contribution from cerebrospinalfluid.

We conducted group comparisons of demographic variables and CMRglc usingthe t test. When a continuous measurement variabledid not meet the assumption of homogeneity of variance, it was assessed usinga separate variance t test (Statistical Package forthe Social Sciences; SPSS Inc, Chicago, Ill). Pearson product moment correlationanalyses tested relationships between CMRglc and MA use (duration, numberof days in the last month, grams per week in the last month). We used d′(discriminability statistic) to assess CPT performance and evaluated groupdifferences in d′ using a t test. For theseanalyses, the statistical threshold was P<.05,uncorrected for multiple comparisons.

Group comparisons of self-reports of depressive symptoms and anxietywere performed by multivariate analysis of variance (SPSS). Post hoc t tests determined the contribution of each variable tothe significant results of an omnibus test.

Group comparisons of brain activity, indicating relative regional cerebralglucose metabolism, were performed by SPM (SPM99; Wellcome Department of CognitiveNeurology, London, England).^29,30 ThePET images (decay-corrected counts) were coregistered to the correspondingstructural MR images using automated image registration,³¹ andthe MR images were used to normalize the PET data spatially by transformationsthat warped them into a standard coordinate system (MNI space; Montreal NeurologicalInstitute, Toronto, Ontario).²⁹ Normalizedimages were smoothed with an 8-mm (full width at half maximum) isotropic Gaussiankernel, and effects of global activity were removed by proportional scaling.

In SPM, a parametric statistical model, assumed at each voxel, describesvariability in the data in terms of experimental and confounding effects andresidual variability. The model provides that for each group of subjects,the activity in each voxel is normally distributed with homogeneous varianceabout a group mean. The hypothesis that the group means for each voxel werehomogeneous was assessed with a t test, giving animage SPM(t) whose voxel values were t statistics. The multiple comparisons problem of simultaneously assessingall the voxel statistics is addressed by modeling the image as a sample ofa continuous Gaussian random field. For each voxel, the corrected P value is the probability of finding at least 1 voxel in the searchvolume with a greater or equal t value. For eachcluster of contiguous voxels where t exceeds a presetthreshold, the corrected P value is the probabilityof finding at least 1 cluster that is at least that large.

For whole-brain SPM analysis, the multiple comparisons correction isbased on all gray matter voxels in the brain and the possibility of deviationfrom the null hypothesis. To test the effect of group, we set an initial voxelheight threshold of P = .05 (uncorrected) for inclusionin clusters. We considered individual clusters to indicate a significant differenceonly when P<.05 (corrected) also for spatial extent.

Statistical parametric mapping allows for multiple comparisons correctionon the basis of a restricted set of hypotheses, namely, consideration of limitedcontiguous regions within the cerebral gray matter. Such an analysis is appropriatewhen prior work identifies particular brain regions as relevant.

We tested certain regions of interest (ROIs) because of their implicationin negative affective states: orbitofrontal cortex (OFC) (medial [gyrus rectusand medial orbital gyrus; Brodmann area (BA) 11] and lateral [lateral andposterior orbital gyri, orbital portion of the inferior frontal gyrus; BA47 and 11]), cingulate gyrus (infragenual [BA 25 and 32], perigenual [BA 24,32, and 33], and posterior [BA 31, 23, and 30]), lateral prefrontal cortex(middle frontal [BA 8, 9, and 46] and inferior frontal gyri [BA 44, 45, 46,and 47]), insula (BA 13), and amygdala. Other than the posterior cingulategyrus, activities of these regions have been linked to depressive disorders,depressed mood, or sadness.^32-36 Activityof the posterior cingulate gyrus was increased during presentation of aversiveand anxiogenic stimuli^37,38 andwas related to the anxiety component of depressive symptoms.³⁹ Inaddition, activities of the OFC and insula have been related to anxiety,⁴⁰ and pharmacologically induced anxiety was associatedwith activation of the amygdala, insula, and anterior cingulate cortex (ACC).⁴¹ Lastly, the ventral and dorsal striatum and cerebellarvermis were tested because their activities were increased or associated withdrug craving.^26,42-46 TheROIs were drawn on the structural MR template provided in SPM99, using MEDx.Bilateral sampling, except for the cerebellar vermis, provided data on 23ROIs and 46 comparisons (positive and negative contrasts). Statistical significancewithin each ROI was determined according to the SPM model described earlier,using a voxel height threshold of P = .05 (uncorrected)for inclusion in clusters. An ROI was considered to show a significant groupdifference if it contained a cluster with P<.05for spatial extent (corrected). In each ROI that showed a significant groupdifference using these criteria, the probability associated with the peakvoxel height (corrected for ROI search volume) was also noted. Further, weidentified ROIs whose group differences maintained statistical significance(P≤.001) after applying the Bonferroni correctionfor number of comparisons (46) and the correction for search volume.

Relationships between relative regional CMRglc (rCMRglc) and depressivesymptoms (BDI) and anxiety (State-Trait Anxiety Inventory) were tested usingcovariate analysis. Statistical significance of the effect of each covariatewas assessed within the 18 ROIs preselected because of their implication innegative affective states (9 bilateral regions). Separate SPM analyses wereperformed for each covariate, yielding SPM(t)s. Themultiple comparisons correction within each ROI and the Bonferroni correctionfor number of ROIs were applied. An ROI was considered to show a significantcovariate effect if it contained a cluster with P<.05for spatial extent (corrected). In each ROI that showed significant covarianceusing these criteria, the probability associated with the peak voxel height(corrected for search volume) was also noted. Further, we identified thoseROIs whose covariance maintained statistical significance (P<.001) after applying the Bonferroni correction for number of comparisons(ie, 18 ROIs × 2 = 36, testing positive and negative covariance) andthe correction for search volume.

An omnibus multivariate analysis of variance of measures of depressionand anxiety demonstrated a significant group difference (F_3,25 =7.60; Wilks Λ = 0.52; P<.001) Table 3). The groups differed on each measure.Although only 1 MA abuser reported being depressed within the month beforetesting on the Addiction Severity Index (data not shown), MA abusers had higherBDI scores than control subjects (t₁₈ =−3.88; P = .001; n = 35, degrees of freedomadjusted for separate variance test). As the BDI data did not meet the assumptionof homogeneity of variance, separate variance t testsassessed this variable. Some MA abusers (4/17) but no control subjects reportedanxiety symptoms on the Addiction Severity Index. The MA abusers also hadhigher scores of both state (t₂₇= −2.91; P = .007) and trait (t₂₇= −3.82; P = .001) anxiety on the State-TraitAnxiety Inventory. In tests of correlation between these scores and MA use(duration of use, number of days used in the past month, grams per week inthe past month), the only significant finding was a positive correlation betweenBDI score and recent use (grams per week) (r₁₄ = 0.70; P = .006).

The MA abusers reported drug craving (mean ± SEM visual analogscale scores, 4.06 ± 0.94). Craving was correlated with frequency ofMA use (number of days in the last 30) (r₁₇ = 0.597; P = .01).

We used CPT data from 15 control subjects and 16 MA abusers. Data from5 subjects were excluded because of equipment failure (n = 1), lack of responsefor 20 to 40 seconds (n = 2), and responses to all stimuli (including nontargets)during the first few minutes (not understanding the instructions) (n = 2).Performance during the 15 minutes after FDG injection (when most of the brainuptake of FDG occurs) indicated no group differences in reaction time (meansfor the control and MA groups, 0.6257 and 0.6252 seconds, respectively) orpercentage of correct responses (98.42 in the control group and 95.85 in theMA group). There was no group difference in d′ (t₃₃ = 1.08; P = .29; mean (SE) controld′ = 5.11 [0.21]; mean (SE) MA d′ = 4.80 [0.26]).

Global glucose metabolism during performance of the CPT did not differbetween the groups (mean ± SEM, 10.1 ± 0.52 and 10.2 ±0.36 mg/100 g per minute in the control and MA groups, respectively [13 controlsubjects and 14 MA abusers]). Drug use measures were not significantly correlatedwith CMRglc in the MA abusers.

Whole-brain analysis revealed 1 cluster, extending from the middle tothe posterior portions of the dorsal cingulate gyrus (4604 voxels; P<.001), with higher activity in MA abusers than control subjects.The peak voxel (6, −16, 46; MNI coordinates as defined in the legendfor Table 4), which was in BA24 or 31, also had significantly more activity in MA abusers by the criterionof peak height after correction for whole-brain search volume (t = 5.75; P = .02). There were no clustersof significantly higher activity in control subjects than MA abusers.

All regions tested, except for those in the lateral OFC and lateralprefrontal cortex, exhibited clusters with group differences in relative rCMRglc(in 1 or both hemispheres) Table 4 and Figure 1). Control subjects had greater activitythan MA abusers bilaterally in the infragenual ACC, the left perigenual ACC,and the right insula. The MA abusers had higher activity bilaterally in thelateral OFC, right middle and posterior cingulate, amygdala, ventral striatum,and cerebellar vermis. After correction for the number of regions compared,MA abusers still had significantly lower relative rCMRglc than control subjectsbilaterally in the infragenual ACC and higher activity in the bilateral ventralstriatum.

Unlike absolute CMRglc, some relative rCMRglc measures showed relationshipsto recent MA use (which was related to measures of mood [see the "AffectiveStates" subsection]). Activity in the left insula covaried negatively withMA use (grams per week) (cluster of 673 voxels; corrected P<.001), as did activity in the left infragenual ACC (cluster of70 voxels; corrected P = .02). Activity in the cerebellarvermis (cluster of 267 voxels; corrected P = .003)covaried positively with this measure. In addition, activity in the rightinsula (cluster of 158 voxels; corrected P = .03)showed negative covariance with frequency of MA use (number of days in thepast 30). Only the association of activity in the left insula with amountof intake retained statistical significance after correction for the numberof brain regions tested.

The BDI score in the MA abusers covaried directly with relative rCMRglcin the bilateral infragenual and perigenual ACC and amygdala Table 5 and Figure 2). This relationship retained significance in the bilateralperigenual ACC and right amygdala after Bonferroni correction. Because theamount of MA consumption (grams per week) was correlated with BDI score, wereassessed the association of BDI score with relative rCMRglc after removingthe effects of MA consumption by declaring it a nuisance variable. The BDIscore activity retained a strong association with activity in the bilateralperigenual ACC as well as in the right amygdala (P<.001for all). While BDI score for the control subjects also covaried with relativerCMRglc in 3 ROIs, none of these results retained significance after Bonferronicorrection.

The relationship of anxiety to relative rCMRglc was assessed separatelyfor state and trait anxiety measures Table6). There were no significant covariates of anxiety in control subjects.In MA abusers, state anxiety covaried negatively with rCMRglc in the leftinfragenual ACC and bilateral insula (Figure 3). The result in the left insularetained significance after Bonferroni correction. The pattern was more widespreadfor trait anxiety, which covaried negatively with activity in the OFC, infragenualand perigenual ACC, left posterior cingulate gyrus, and left insula. Negativecovariance of trait anxiety with activity in the infragenual anterior cingulategyrus (bilateral), medial (bilateral) and left lateral OFC, and left insularetained statistical significance after Bonferroni correction. Although therewas evidence for the opposite pattern (high anxiety accompanied by high metabolism)in the bilateral amygdala and right insula, none of these results retainedsignificance after Bonferroni correction.

This study identified regional brain dysfunction that may underlie affectivedeficits in MA abusers during early abstinence. Abnormalities were observedin limbic and paralimbic regions that have been implicated in emotional processingby many investigations. These include neuroimaging studies in which mood inductionaccompanied activation of the amygdala, insula, and ACC.^40,47-51 Themost robust group differences in relative rCMRglc were in the infragenualACC, where MA abusers showed relative hypoactivity compared with control subjects,and the ventral striatum, where MA abusers showed relative hyperactivity.

The BDI scores in MA abusers covaried positively with relative activityin regions that have been linked with negative affect. In line with previousevidence for dysfunction of the subgenual and perigenual portions of the ACCin clinical depression,^33,34,52 relativerCMRglc in the right perigenual ACC was robustly associated with BDI scorein MA abusers. This association is also consistent with a report that thisregion was affected (activated) when autobiographical scripts induced sadnessin healthy subjects.⁴⁰ Nonetheless, the positiveassociations appear to be paradoxical because higher ratings of depressivesymptoms accompanied a lower relative rCMRglc in the infragenual and, lessrobustly, the perigenual ACC compared with control subjects. Thus, while theACC appears dysfunctional in MA abusers, the relationship between depressivesymptoms and ACC function may differ between MA users and healthy subjects.

After correction for multiple comparisons, the perigenual (not infragenual)ACC retained a significant positive correlation between BDI score and relativeactivity bilaterally, whereas the infragenual (not perigenual) ACC showedlower relative rCMRglc bilaterally in MA abusers than in control subjects.These subregions of the ACC are functionally distinct. While the infragenualregion is part of BA 25 and functionally linked to the limbic system, thesuperior perigenual region has been linked to diverse functions includingattention,⁵³ conflict resolution,⁵⁴ andanalgesia.^55,56

Another strong finding was the direct covariance of BDI score with relativerCMRglc in the amygdala in MA abusers. Along with the higher relative activityof the amygdala in MA abusers (vs control subjects), this effect suggestedthat amygdalar dysfunction contributes to depressed mood in abstinent MA abusers.Prior studies have shown elevated amygdalar blood flow or rCMRglc in depressedsubjects with familial major depressive disorder or bipolar I and II disordersand have demonstrated elevated amygdalar metabolism rates during major depressiveepisodes (see Drevets³⁴). One report notedthat resting blood flow and relative rCMRglc were higher in the amygdala insubjects with depression than in control subjects and that relative activityin the left amygdala was correlated with the severity of depressive symptoms.⁵⁷ Similar associations in MA abusers but not in patientswith panic disorder, obsessive-compulsive disorder, phobic disorders, or schizophrenia³⁶ suggest a unique similarity between MA abusers andindividuals with primary mood disorders.

Although the mechanism by which depressive symptoms may be linked tothe activity of the amygdala in MA abusers or patients with primary mood disordersis not known, Drevets³⁶ discussed how thisrelationship might reflect a role of the amygdala being to organize emotionaland stress responses. He noted that electrical stimulation of the amygdalain humans produces dysphoria (see Brothers⁵⁸)and that excessive amygdalar transmission to the periaqueductal gray mattermay produce depressive signs.

Although several regions had relative rCMRglc that covaried with anxietymeasures (state or trait), only the left insula exhibited a significant associationwith both state and trait anxiety after Bonferroni correction. This inverserelationship is consistent with the negative correlation between anxiety scoresof patients with depression with relative rCMRglc in the left but not theright insula.⁵¹ Covariance of trait anxietywith relative activity of left infragenual ACC, left OFC, and medial OFC (bilaterally)also retained significance after correction. Our observations that the MAabusers had lower relative rCMRglc in the insula and infragenual ACC and higheranxiety self-ratings were also internally consistent. The negative associationof trait anxiety with relative activity in the lateral OFC, where MA abusershad higher activity than control subjects, is less clear.

A report⁵⁹ on the relationships betweencortical glucose metabolism rates, blood flow, and anxiety in control subjectsis relevant to the present findings in the insula. That report indicated alinear inverse correlation between global anxiety score and cortical rCMRglcbut an inverted U relationship between anxiety and cortical blood flow, measuredby the less stressful xenon Xe 133 inhalation technique. Cortical blood flowincreased with anxiety in subjects with low anxiety but decreased with anxietyin subjects with high anxiety. In light of these findings and a report ofinsular activation in healthy subjects by scripts that induce anxiety,⁴⁰ the negative associations between insular metabolismand anxiety measures in MA abusers may reflect a similar curvilinear relationship.Because MA abusers exhibited higher scores on the State-Trait Anxiety Inventorythan the control group, it is feasible that their relationship between insularactivity and anxiety are at the upper end (negative slope) of such a curvilinearfunction.

Sufficient data were not collected on the day of PET scanning to testrelationships between relative rCMRglc and MA craving. In keeping with findingsof individuals who abused drugs other than MA, MA abusers had a higher relativerCMRglc than control subjects in some regions (lateral OFC, posterior cingulategyrus, amygdala, ventral striatum, and cerebellar vermis) whose activitieswere previously positively related to drug cravings.^{26,42,46,60-64} Theydid not, however, show a higher relative rCMRglc of other regions (lateralprefrontal cortex, ACC, and insula), which also were positively associatedwith drug craving in previous studies.^{26,45,46,60-67}

Some abnormalities in rCMRglc of MA abusers may be due to effects ondopaminergic systems. Of the regions selected for planned comparisons, theOFC, ACC, insula, amygdala, and dorsal and ventral striatum have dopaminergicinnervation.^68-70 Eachof these regions contained a cluster with a group difference in relative rCMRglc.The direction of the difference varied across the regions. To the extent thatgroup differences may result from insult to dopaminergic systems, this variationmay reflect differential effects of such deficits.

Notably, the infragenual ACC, which showed the most robust deficit ofrelative rCMRglc compared with the control group, receives a dense dopaminergicinnervation.⁷¹ A dopaminergic deficit in thisregion therefore may produce the local metabolic defect. Alternatively, defectiverCMRglc in the OFC may reflect striatal dopaminergic deficiency as demonstratedby correlation between striatal dopamine D₂ receptor availabilitywith orbitofrontal rCMRglc.⁷² OrbitofrontalrCMRglc abnormalities in MA abusers may also reflect a serotonergic deficitbecause low levels of serotonin were measured in postmortem samples of OFCfrom MA abusers.¹¹

Absolute CMRglc in MA abusers in early abstinence did not differ fromCMRglc in control subjects. Because MA abusers who were abstinent for 2 weeksto 35 months previously exhibited higher CMRglc than in control subjects,¹⁵ rCMRglc may change with sustained abstinence, unmaskinga hypermetabolic condition. Longitudinal studies that include early abstinencemay resolve this question.

This study has limitations, including the fact that although all ofthe regions tested have been shown to influence mood, they contribute to otherbehavioral states not addressed in the current study. Another limitation isthat while the groups were similar in most categories, most of the MA abusersbut none of the control subjects were tobacco smokers. Smoking status wasnot considered in prior studies of MA abusers, and differences in cerebralmetabolism associated with nicotine dependence have not been described butwarrant further study. One potentially confounding variable was craving associatedwith abstinence in nicotine-dependent smokers. Although craving for cigarettescan begin within minutes of smoking, and smokers were required to abstainfrom smoking for at most 2 hours before the PET measurement, the pattern ofabnormalities observed in the MA abusers' relative rCMRglc compared with controlsubjects did not resemble relative rCMRglc findings in smokers during cigarettecraving.⁶⁵

Another issue was the subjects' state during relative rCMRglc measurement.The participants performed an auditory CPT, and this simple task may affectrelative rCMRglc differently in MA abusers than in control subjects becausesubjects in the 2 groups may have had different levels of regional activation.The group differences, accordingly, may be state dependent. Finally, our studycould not discriminate between effects of early abstinence, effects of chronicMA abuse that are unchanged by abstinence, and factors that predated MA abuseon relative rCMRglc and mood.

With these caveats, we conclude that MA abusers in early abstinencehave dysfunction in the limbic and paralimbic regions that have been linkedwith negative affective states. Depressive symptoms showed positive covariancewith perigenual ACC and amygdalar activity in MA abusers. In contrast, anxietywas negatively associated with activities in all of the regions, except forthe dorsal striatum, where MA abusers exhibited CMRglc deficits compared withcontrol subjects. The findings identified brain substrates of affective dysregulationas potential targets for therapeutic intervention during early abstinenceand withdrawal in MA abusers.

Corresponding author and reprints: Edythe D. London, PhD, NeuropsychiatricInstitute, University of California Los Angeles, 760 Westwood Plaza, PO Box175919, Los Angeles, CA 90024-1759 (e-mail: elondon@mednet .ucla.edu).

Submitted for publication December 30, 2002; final revision receivedJune 13, 2003; accepted June 24, 2003.

This study was supported by contract 1 Y01 DA 50038 (Drs London andLing), grants MOI RR 00865 and T32 DA 07272 (Dr Rawson), and the IntramuralResearch Program of the National Institute on Drug Abuse, Bethesda, Md (DrMatochik and Mr Kurian).

This article was presented in part at the annual meeting of the Societyfor Neuroscience; November 7, 2000; New Orleans, La; the annual meeting ofthe Society for Neuroscience; November 12, 2001; San Diego, Calif; the annualmeeting of the American College of Neuropsychopharmacology; December 13, 2000;San Juan, Puerto Rico; the annual meeting of the American College of Neuropsychopharmacology;December 13, 2001; Waikoloa, Hawaii; and the annual meeting of the Collegeon Problems of Drug Dependence; June 18, 2001; Scottsdale, Ariz.

We thank Joshua Mogy, BA; Laurence Rosenthal, BA; Arby Hayrapetian,BA; Eileen Callahan, BA; and Jennifer M. Dacey, BA, for excellent contributionsto data collection and organization, and Bradley Voytek, BA, for data analysis.