Alterations in the Functional Anatomy of Working Memory in Adult Attention Deficit Hyperactivity Disorder

Attention deficit hyperactivity disorder (ADHD) is the most common childhood psychiatric disorder (DSM-IV) and persists into adulthood for 30%–50% of affected children. Neuropsychological studies of ADHD (1) suggest that it involves a dysfunction in working memory.

Brain imaging studies of ADHD point to abnormalities in regional brain structure (2) and function (3). Positron emission tomography (PET) with [¹⁸F]fluorodeoxyglucose imaging in adults (e.g., reference 3) has identified alterations in neural activity-linked metabolism related to ADHD traits rather than states. An alternate approach for defining the pathophysiology of ADHD incorporates brain activation tasks to explore cognitive deficits, such as defects in working memory. The goal of the present study was to compare regional cerebral blood flow (rCBF) changes related to working memory in adults with and without ADHD.

METHOD

Six unmedicated, right-handed male adults diagnosed with ADHD who had current and childhood histories of inattention, impulsivity, and hyperactivity but no other lifetime axis I diagnoses were studied. The comparison group comprised six healthy men matched in age and general intelligence who had no lifetime axis I diagnosis. The mean ages of the men with ADHD was 28.5 years (SD=8.9), and the mean age of the men without ADHD was 25.7 (SD=5.2). Screenings included the ADHD Rating Scale (4); review of developmental, family, educational, and social characteristics, drug and alcohol use (drug use was confirmed by urinalysis), and occupational histories; SCL-90-R; neurological examination; Wechsler Adult Intelligence Scale—Revised; and Wide Range Achievement Test—Revised. The men with ADHD fulfilled at least eight of 14 items of the DSM-III-R criteria during both childhood and at the time of imaging. Diagnoses were confirmed by using scores on the ADHD Rating Scale (4), which was completed by participants, their parents, and significant others, who also rated the subjects’ current and/or childhood functioning (5). The men with ADHD endorsed at least 10 of 14 items rating current behavior (4), and all had scores of 12 or greater when rated by a parent (or a family friend for one subject) on the Conners Abbreviated Symptom Questionnaire (5). Participants with learning disabilities were excluded from both groups. Three of the men with ADHD had a childhood ADHD diagnosis. Two men with ADHD had taken methylphenidate during childhood but had been methylphenidate-free for 2 and 10 years, respectively, before the study. All subjects provided written informed consent.

The Paced Auditory Serial Addition Task (6) was administered to engage working memory. Single-digit numbers were presented binaurally every 2.4 seconds. Subjects added each number to the preceding number. Subjects also vocalized random numbers during a self-paced number-generation control condition to correct for the contribution of arousal, auditory stimulation, verbal production, and short-term (nonworking) memory to the rCBF changes related to the auditory addition task.

PET imaging occurred on 2 consecutive days; conditions of day 1 were replicated on day 2. Subjects completed two resting, number-generation, and addition task conditions per day, for a total of 12 scans. A Siemens 921 PET scanner acquired 90-second scans in three-dimensional mode after bolus intravenous administration of 25 mCi of[¹⁵O]H₂O. T₁-weighted transaxial spin echo magnetic resonance images (repetition time=650 msec, echo time=20 msec) of brain structure were obtained for coregistration to PET images (7).

Scans were fit to an atlas centered in Talairach and Tournoux coordinates. PET images were smoothed to a final isotropic resolution of 11.8 mm full width at half maximum and normalized by proportionate scaling of global activity. Alterations in rCBF were identified in a pixel-by-pixel comparison using a four-way analysis of variance (ANOVA) to generate t statistic images accounting for differences between factors of day, task, subject, and repetitions.

The difference (p<0.005) between number-generation versus addition task images isolated the neural correlates of working memory. Differences over time among the four addition task images were compared in an omnibus F test with a threshold of p<0.005.

RESULTS

There were no significant differences in rate of number generation between the men without ADHD (mean=0.63/second, SD=0.03) and those with ADHD (mean=0.62/second, SD=0.28) (t=0.05, df=10, p=0.96). Men without ADHD performed better over the averaged four addition task trials (mean=97% correct, SD=6.8%) than men with ADHD (mean=83% correct, SD=15.8%) (F=5.06, df=1, 10, p<0.05, repeated measures ANOVA).

Table 1 summarizes significant increases in rCBF for the addition task condition compared with the number-generation condition. The largest activations for the men without ADHD were in the left superior temporal and right lateral frontal gyrus. The largest relative activations for the men with ADHD were in the precuneus and inferior parietal lobe.

Significant changes in the rCBF for the four repetitions of the addition task scans (omnibus F=6.5, df=3, 15, p<0.005) were compared in men with and without ADHD. Men without ADHD demonstrated significant time-related rCBF increases in the anterior cingulate and medial frontal regions (Brodmann area 32/10) and decreases in the left middle frontal regions (Brodmann area 9). Men with ADHD showed significant time-related decreases in the left middle (Brodmann area 21) temporal lobe and increases in the right lenticulate, left parahippocampal gyrus (Brodmann area 35/36), and bilateral cerebellum.

DISCUSSION

The use of a working memory task indicated striking differences and similarities in the recruitment of brain regions by men with and without ADHD. The activations in the men without ADHD are consistent with models of working memory implicating the involvement of right frontal regions in a central executive system (8) and the recruitment of subsidiary regions. Frontal activations may reflect retrieval of previous information enhanced by preparatory attention. Temporal activations are consistent with the engagement of a phonological loop for subvocal rehearsal of the previously presented digit.

In contrast, the men with ADHD exhibited a more diffuse pattern of rCBF activations during the Paced Auditory Serial Addition Task. The relative lack of task-related frontal activations suggests a diminished ability of retrieval mechanisms. Extrastriate activations suggest the primary use of visual imagery to perform the addition task; postimaging interviews of the men with ADHD confirmed their use of visual strategies during the addition task. Use of visual strategies by subjects with ADHD may reflect inadequate use of frontal regions to recruit more effective subsidiary regions subserving subvocal rehearsal strategies or may reflect superior compensatory visuospatial skills. Lenticulate findings are consistent with brain volumetric evidence (2) supporting a role for the basal ganglia in ADHD pathophysiology.

These findings need confirmation in a larger study group. A strength of this study was the inclusion of rigorously selected subjects without comorbid psychiatric or learning diagnoses and carefully matched comparison subjects. This study suggests that people with ADHD recruit novel neural pathways while organizing less efficient strategies to solve working memory tasks.

Presented in part at the annual meeting of the Society for Neuroscience, New Orleans, Oct. 25–30, 1997. Received Dec. 10, 1998; revisions received May 27 and July 22, 1999; accepted Aug. 2, 1999. From the Departments of Psychiatry and Behavioral Sciences, Neurology, and Radiology and The Emory University Center for Positron Emission Tomography, Emory University School of Medicine, Atlanta. Address reprint requests to Dr. Schweitzer, Maryland Psychiatric Research Center, University of Maryland, P.O. Box 21247, White Bldg., Redwood Circle, Baltimore, MD 21228; [email protected] (e-mail). Supported in part by the University Research Committee of Emory University and by NIMH grants MH-01053 and MH-55550 (Dr. Schweitzer). The authors thank the PET technologists at Emory University Hospital and Ms. Lara Kay for technical assistance.

TABLE 1

Enlarge table