Toward
Ronald
a Brain
Map
of Auditory
Hallucinations
John M. Cleghorn, M.D., Sheryl Franco, R.N., Barbara Szechtman, D. Kaplan, Ph.D., Henry Szechtman, Ph.D., Gregory M. Brown, Claude Nahmias, Ph.D., and E. Stephen Garnett, M.D.
Objective: metabolic
map
This study asks whether of the brain. Method:
auditory Regional
hallucinations brain metabolism
emission tomography with [1SF]fluorodeoxyglucose who experienced auditory hallucinations during
glucose
tients
were
been
were with
reexamined the patients
hallucinations
ofneuroleptics after who had
and a trend Hallucination striatum
free
and
and
1 9 had
never
lower
relative
regions voices
treated
metabolism
anterior
cingulate
regions.
Neuroleptic
of the cortex in a pattern similar but different in that left prefrontal
to that regions
with
schizophrenic 1 0 who did
neuroleptics.
treatment. the patients in auditory
toward higher metabolism in the right hemisphere scores correlated positively and significantly
in striatal metabolism and a reduced frontal-parietal with a decrease in hallucination scores. Conclusions:
reflected in a distinctive was measured by positron
in 12 DSM-III uptake and
1 year to assess effects of neuroleptic did not experience hallucinations,
significantly
are
treatment
B.A., M.D., Ph.D.,
and
resulted
Nine
Results: who did
homologue with relative
patients All pa-
not.
patients
Compared experience
Wernicke’s
regions
ofBroca’s metabolism
region. in the
in a significant
increase
ratio, which was significantly Auditory hallucinations involve
correlated language
seen in normal subjects are not activated. The
listening striatum
to their own plays a critical
role in auditory hallucinations. (Am J Psychiatry 1992; 149:1062-1069)
I
n order to discover the neural basis of psychotic phenomena such as auditory hallucinations, the systems of the living brain involved in their occurrence must be identified. Auditory hallucinations, a salient feature of
(2). We chronic nations
70%
(N=10) (3). We reported that correlations between metabolism in certain language regions were significantly different in the hallucinating group than in the nonhallucinating group and control subjects. Hallucination scores correlated positively and significantly with relative metabolism in the anterior cingulate region. All patients were taking neuroleptic medications. In the present study we examined metabolic activity in language regions of the brain (figure 1 ) with positron emission tomography (PET) using [18F]-fluorodeoxyglucose (FDG) in 22 drug-free schizophrenic patients who experienced auditory hallucinations; 19 of the patients had never been treated with antipsychotic drugs. Twelve of the patients reported hearing voices during the period in which FDG was being taken up in the brain, and 10 did not. Thus, the relative regional metabolism of the brain was pictured and measured during the fortuitous occurrence of hallucinations in one group of patients, while in the other group, the voices were quiescent during the examination. Using unmedicated subjects, we could thus reexamine the hypothesis that hallucinatory events involve the language regions of the brain.
of schizophrenic
psychoses,
may
represent
subvo-
cal speech. The notion that auditory hallucinations might reflect the subjects’ own speech was suggested by the observation that the reported content of hallucinations corresponded with that recorded through a microphone held to the mouth (1 ). The fact that auditory hallucinations can be interrupted by opening the mouth very wide, thus preventing vocalization, suggests that the hallucinations are expressed by the subject’s larynx
Presented the American
at a new research session at the Psychiatric Association, New Received Nov. 6, 1 990; revision received Jan. I 992. From the Departments of Psychiatry McMaster University; the Department of Toronto, Toronto, Ont.; and the Department Chedoke-McMaster Hospitals, Hamilton, quests versity,
to Dr. 1200
Szechtman, Department Main St. West, Hamilton,
143rd annual meeting of York, May 12-17, 1990. 8, 1 992; accepted Feb. 7, and Biomedical Sciences, Psychiatry, University of of Nuclear Medicine, Ont. Address reprint re-
of Psychiatry, Ont., Canada
McMaster UniL8N, 3ZS.
Dr. Cleghorn died June 8, 1992. Supported by grant MA-10214 from the Medical Research Council of Canada, NIMH grant MH-44073, and by Research Associate Awards from the Ontario Mental Health Foundation to Dr. Brown and Dr. Szechtman. Copyright © 1992 American Psychiatric Association.
1062
group
examined regional brain glucose metabolism in schizophrenic subjects with persistent halluci(N=9) and compared them with a matched
of patients
(N=l0)
Am
and
] Psychiatry
normal
control
1 49:8,
August
subjects
1992
CLEGHORN,
FIGURE 1. Three Slice Levels of the Brain Examined in 22 Schizophrenic Patients With and Without Auditory Hallucinations During PET Scana
FRANCO,
TABLE 1. Demographic Characteristics and Without Auditory Hallucinations
SZECHTMAN,
of Sch izophrenic
Hallucinating Subjects (N=12)
ET AL.
Patients
With
Nonhallucinating Subjects (N=10)
Item
Mean
SD
Mean
SD
Age (years) Education (years)a Length of medicanon (years) Age at onset of
24.2 11.1
7.4 1.8
29.5 13.0
9.7 2.2
0.8
2.6
1.0
2.1
18.8
2.9
23.6
4.5
4.5
5.5
5.9
5.8
1.3 I .2 98.1 81.6
0.6 0.8 14.6 18.8
1.2 1 .0 105.4 99.4
0.6 0.8 12.6 17.2
7.7
2.5
10.9
3.1
1.0
1.9
1.0
schizohrenia (years) Length of illness
(years) Number
aLow:
orbitofrontal-superior
an inferior frontal
part
including
marginal
temporal
of Wernicke’s Broca’s
including
region-inferior region-parietal;
auditory
cortex
and
parietal;
middle:
pre-
high:
prefrontal-supra-
and parietal.
METHOD Patients
sodes Use of street drupC Performance IQ Memory quotiente Picture arranpement score SADS-C positive
symptomscores5 3.0 SADS-C negative symptomscores1’ 4.3 at2.25, df=20, p=O.04 (two-tailed test bt300, df=20, p=0.OO7. Cl .0=prior substance abuse to a mild or criteria. dt,2.10, df=20, et2.31, df=20,
We recruited patients from the emergency and outpatient services who met presumptive DSM-III-R criteria for schizophrenia. They were then admitted to the hospital and followed as outpatients for at least 1 year. Patients with a history of neurological disorder and majon medical illness were excluded. Detailed developmental histories were obtained from family members. Histories included the onset of negative and later positive symptoms; this information, when combined with follow-up data, permitted the diagnosis of schizophrenia by DSM-III criteria. It should be understood that the patients classified as not having hallucinations had experienced hallucinations at various times before the procedure. The mean age of the patients who did experience hallucinations during PET scan was 24.2 years (SD=7.4). Eight were men and four were women. Nine patients were reexammed after 1 year of neuroleptic treatment. Five had hallucinated during the scan, but all had auditory hallucinations before and after the scan. The patients who did not experience hallucinations during PET scan had a mean age of 29.5 years (SD=9.7). Eight were men and two were women. All but three patients were righthanded. The two left-handed patients and one mixedhanded patient were in the group that experienced hallucinations. Only two patients who did not and one patient who did experience hallucinations had in the past been treated with neuroleptic medications, and they had been drug free for many months before the study. All patients had been ill for 6 months or more, some for many years. There was no significant difference between the groups’ length of illness: the mean for patients who did experience hallucinations was 4.5 years
Am]
of epi-
Psychiatry
149:8,
August
1992
t=2.62, t=2.80, ht2.67,
df=20, df=20, df=20,
0.8 2.0 for all comparisons). moderate
degree
0.8
by DSM-III
p=O.OS. p=O.O3. p=O.O2. p=O.Ol. p=O.Ol.
(SD=5.S), and the mean for patients who did not expenience hallucinations was 5.9 years (SD=S.8). However, the mean age at onset for the patients who did expenience hallucinations was significantly younger than that of the patients who did not experience hallucinations during PET scan (p=O.007) (table 1). The patients who experienced hallucinations also had significantly less education and a lower Wechsler perfonmance IQ and memory quotient. Picture arrangement scores were significantly lower in the patients who experienced hallucinations (table 1). None of the 60 additional neuropsychological tests we used discriminated between groups. These measures are described in Cleghorn et al. (4). Type I errors are possible here because of the large number of comparisons. Both positive and negative symptoms were more severe among the patients who experienced hallucinations (table 1). The patients who experienced hallucinations had higher hallucination scores than those who did not experience hallucinations when measured before scanning (mean=4.3, SD=0.8, and mean=3.0, SD=O.7, respectively) as well as during the procedure (mean=4.1, SD=1.2, and mean=0, nations occurred more frequently
did experience ability of their
SD=0). in the
Since hallucipatients who
hallucinations, there was a higher proboccurring during the scanning procedure.
Since group differences reflect the severity of any
in brain metabolic of the symptoms
data might or neurocog-
1063
AUDITORY
HALLUCINATIONS
nitive impairments rather than the presence or absence of hallucinations during the scan, correlations between these variables and hallucination scores during scanning were calculated. None of the symptom scale scores, the mean positive and negative symptom scores, or neurocognitive measures were significantly related to hallucination scores during the scanning procedure or before it. Thus, these hallucinatory events are independent of the other measures but might well be related to factors underlying severity of illness. Since these factors might be reflected in regional brain metabolism, we examined the relation of all symptom scores to regional brain metabolism. One clear difference between groups is that they did or did not hallucinate during the FDG uptake period before the scan. Hallucinations occur intermittently, while the other symptoms are more constant.
Procedure Informed consent, following a full explanation of all procedures employed in the study, was obtained from all subjects. An experienced psychiatric nurse cared for the subjects and inserted a number 18 angiocatheter into a vein of the forearm. Subjects were injected with FDG, S mCi, while lying on a comfortable stretcher. They were instructed to keep their eyes closed and were not spoken to. The only noise was due to the air handling system of the building. Warming the arm produced arterialized samples of venous blood that were taken from the arm opposite that used for FDG injection for the determination of plasma glucose and FDG concentrations. Scanning was started 45 minutes after FDG injection. Local rates of cerebral glucose metabolism were calculated according to the method of Brownell et al. (5). The procedures and methods have been more fully described elsewhere (3, 6). The Schedule for Affective Disorders and Schizophrenia-Change Version (SADS-C) (7) was administered immediately before the glucose uptake period by the nurse so that ratings of positive and negative symptoms, initially obtained on admission, could be checked. Scales for the assessment of five negative symptoms (Scale for the Assessment of Negative Symptoms) and four positive symptoms (Scale for the Assessment of Positive Symptoms) were 6-point scales in which each level of intensity was defined in behavioral terms (8) (table I ). The ratings are based on interview data and on behavioral observations by staff and family members. The patients were well-known to the highly experienced research nurse (S.F.). Immediately before being placed in the tomograph, patients were asked to report any auditory hallucinations they had during the uptake period for FDG. These reports were recorded and reviewed together with an experienced clinician (J.M.C.), and a consensus rating was made. The mean hallucination score for the hallucinating patients was 4.1 (SD=L2). A rating of 3 describes the hallucination as vivid, occasional, and bothersome; a rating of 4, as vivid and frequent.
1064
Local
Cerebral
Glucose
Metabolism
Regional distribution of FDG was measured with the McMaster PET, which has a spatial resolution of 8 mm (full width at half maximum) in the plane and 12 mm in the axial direction (9). It can therefore resolve voxels of less than I cm3. At least 106 counts were collected, in 3 minutes, for each slice. The subject’s head was held in a Perspex restraining device. Alignment in the left-toright axis was ensured by inspection of a preliminary scan taken at the level of the thalamus. Typically, 16 slices of the brain (on a plane S to the orbitomeatal line) were obtained sequentially from the level of the hippocampal gyni up to the level of cortical vertex. Although each slice was 15 mm thick, it overlapped its predecessor by 10 mm. With this procedure, the slice with greatest volume of gray matter of interest for each subject can be identified and selected. For this study, three slices were first selected for analysis (figure 1). The lowest slice (midthalamic-midstriatal slice) contamed the largest amount of temporal cortex, thalamic, and striatal tissue; auditory cortex is defined as segments 10 and I I (figure 2). The procedure for dividing the cortex into segments is shown in figure 2. The middle slice (midcallosal slice) passed above the basal ganglia through the midportion of the dorsoventral extent of the corpus callosum. Language regions of the brain were approximately located by transposing brain stimulation maps onto the PET image, as described by us (3), and are shown in figures 1 and 2. Data
Analysis
In order to test the hypothesis that language areas of the brain are involved in auditory hallucinations, the groups of patients who did and did not hallucinate duning FDG uptake were compared. A two-way analysis of variance (ANOVA) was used to examine group and laterality differences and their interactions in each region of interest, followed by t tests (two-tailed) for post hoc comparisons. Frontal and parietal relative metabolism was later examined by using the same methods.
RESULTS The groups did not differ significantly in absolute whole brain glucose metabolic rate: the mean for patients who did experience hallucinations was 34.50 j.tmol/100 g/min (SD=l0.l0), and the mean for patients who did not experience hallucinations was 32.40 mol/ 100 g/min (SD=6.70). There were no significant differences between the means of tomographic slices within or between groups. Relative
Regional
Metabolism
To test the hypothesis are involved in auditory compared by two-way
Am
and
Laterality
that
language areas of the brain hallucinations, groups were ANOVA (Group by Laterality)
] Psychiatry
1 49:8,
August
1992
CLEGHORN,
FIGURE Auditory
2. Cortical Rim, Divided Into 36 Segments Hallucinations During PET Scana
Each Spanning
Low Slice Left
an Arc of 100, Examined
Middle
Right
in 22 Schizophrenic
Slice
Left
Frontal
FRANCO,
SZECHTMAN,
Patients
ET AL.
With and Without
High Slice
Right
Frontal
Left Right Frontal Clngulate
Broca’s
Anterior
Temporal Auditory
Posterior
Parletal
Parietal aThe
borders
were
defined
Auditory cortex is defined center of the brain. Broca’s rim of the cortex.
by an edge-detecting
algorithm.
as segments 1 0 and 1 1 and region is defined as segments
In the low
Wernicke’s region 6-8 in that slice
and by t tests in each region in the temporal and supramarginal areas and Broca’s regions in frontal cortex (figure 1 ). Since the striatum has a role in language, it was also included in the analysis. The results show trends toward higher relative metabolism in the right hemisphere regions homologous for Broca’s and lower relative metabolism in primary auditory areas and in the right superior temporal region, which includes part of auditory association cortex, in the patients who did experience hallucinations. For the right hemisphere homologue of Broca’s, the mean relative metabolism in the patients who did experience hallucinations was 1.04 (SD=0.04), and the mean in patients who did not experience hallucinations was 1.03 (SD=O.0S) (t=1.77, df=20, p=O.O9). For the right auditory, the mean in patients who did experience hallucinations was 1.03 (SD=0.05), and the mean in patients who did not experience hallucinations was 1.07 (SD=0.0S) (t=l.82, df= 20, p=O.O8). For the right superior temporal, the mean in patients who did experience hallucinations was 1.01 (SD=0.04), and the mean in patients who did not experience hallucinations was 1.05 (SD=0.04) (t=l.91, df= 20, p=O.O7). The differences reach significance in the posterior half of the superior temporal gyrus, the mean of regions 9-14 on the left and 23-28 on the right (figure 2) (F=6.59, df=l, 20, p=O.O2), with the patients who did experience hallucinations having lower relative metabolism (mean=l.03, SD=O.02) than patients who did not experience hallucinations (mean= 1 .05, SD=0.02). When the left-handed patients were removed from the analysis, the posterior temporal values were significantly lower in the patients who did experience halluci-
Am
J
Psychiatry
1 49:8,
August
1992
slice
central
structures
as segments
(see text).
are 9-1
corpus
3. In the
striatum middle
and, slice
In the high slice the cingulate
to the posterior,
no structures
gyrus
are
is visible,
thalamus. visible
in the
as well as the
nations (mean=1.03, SD=0.02) than in the patients who did not (mean=1.05, SD=O.02) (F=8.74, df=1, 17, p= 0.009). The mean of right and left auditory regions was also significantly lower in the patients who did experience hallucinations than in those who did not (mean= 1.01, SD=0.03, N=9, versus mean=1.0S, SD=0.04, N=
10) (F=6.42,
df=1,
17, p=0.02).
There was also a significant Group by Laterality interaction in the anterior temporal region (F=4.57, df=l, 20, p=O.OS). When the three left-handed patients were removed from this analysis, the interaction became nonsignificant. An interaction between group and laterality in the striatum was significant whether or not left-handers were deleted (F=4.78, df=1, 20, p=O.O4); in patients who did experience hallucinations, the mean left relative metabolism was 0.98 (SD=0.09) and the mean right was 0.98 (SD=0.09); in patients who did not experience hallucinations the mean left relative metabolism was 0.94 (SD=0.08) and the mean right was 0.98
(SD=O.0S). There were no significant differences between groups in relarive metabolism in supramarginal, frontal, or parietal regions or in the anterior or posterior cingulate gyrus or thalamus, as depicted in figures 1 and 2. Correlations Between Regional Metabolism Hallucination
Symptom
Scores
and
Relative
correlated with stniatum (r= one-tailed, Pearson product-moment correlation), left stniatum (r=0.69, p
Ronald
a Brain
Map
of Auditory
Hallucinations
John M. Cleghorn, M.D., Sheryl Franco, R.N., Barbara Szechtman, D. Kaplan, Ph.D., Henry Szechtman, Ph.D., Gregory M. Brown, Claude Nahmias, Ph.D., and E. Stephen Garnett, M.D.
Objective: metabolic
map
This study asks whether of the brain. Method:
auditory Regional
hallucinations brain metabolism
emission tomography with [1SF]fluorodeoxyglucose who experienced auditory hallucinations during
glucose
tients
were
been
were with
reexamined the patients
hallucinations
ofneuroleptics after who had
and a trend Hallucination striatum
free
and
and
1 9 had
never
lower
relative
regions voices
treated
metabolism
anterior
cingulate
regions.
Neuroleptic
of the cortex in a pattern similar but different in that left prefrontal
to that regions
with
schizophrenic 1 0 who did
neuroleptics.
treatment. the patients in auditory
toward higher metabolism in the right hemisphere scores correlated positively and significantly
in striatal metabolism and a reduced frontal-parietal with a decrease in hallucination scores. Conclusions:
reflected in a distinctive was measured by positron
in 12 DSM-III uptake and
1 year to assess effects of neuroleptic did not experience hallucinations,
significantly
are
treatment
B.A., M.D., Ph.D.,
and
resulted
Nine
Results: who did
homologue with relative
patients All pa-
not.
patients
Compared experience
Wernicke’s
regions
ofBroca’s metabolism
region. in the
in a significant
increase
ratio, which was significantly Auditory hallucinations involve
correlated language
seen in normal subjects are not activated. The
listening striatum
to their own plays a critical
role in auditory hallucinations. (Am J Psychiatry 1992; 149:1062-1069)
I
n order to discover the neural basis of psychotic phenomena such as auditory hallucinations, the systems of the living brain involved in their occurrence must be identified. Auditory hallucinations, a salient feature of
(2). We chronic nations
70%
(N=10) (3). We reported that correlations between metabolism in certain language regions were significantly different in the hallucinating group than in the nonhallucinating group and control subjects. Hallucination scores correlated positively and significantly with relative metabolism in the anterior cingulate region. All patients were taking neuroleptic medications. In the present study we examined metabolic activity in language regions of the brain (figure 1 ) with positron emission tomography (PET) using [18F]-fluorodeoxyglucose (FDG) in 22 drug-free schizophrenic patients who experienced auditory hallucinations; 19 of the patients had never been treated with antipsychotic drugs. Twelve of the patients reported hearing voices during the period in which FDG was being taken up in the brain, and 10 did not. Thus, the relative regional metabolism of the brain was pictured and measured during the fortuitous occurrence of hallucinations in one group of patients, while in the other group, the voices were quiescent during the examination. Using unmedicated subjects, we could thus reexamine the hypothesis that hallucinatory events involve the language regions of the brain.
of schizophrenic
psychoses,
may
represent
subvo-
cal speech. The notion that auditory hallucinations might reflect the subjects’ own speech was suggested by the observation that the reported content of hallucinations corresponded with that recorded through a microphone held to the mouth (1 ). The fact that auditory hallucinations can be interrupted by opening the mouth very wide, thus preventing vocalization, suggests that the hallucinations are expressed by the subject’s larynx
Presented the American
at a new research session at the Psychiatric Association, New Received Nov. 6, 1 990; revision received Jan. I 992. From the Departments of Psychiatry McMaster University; the Department of Toronto, Toronto, Ont.; and the Department Chedoke-McMaster Hospitals, Hamilton, quests versity,
to Dr. 1200
Szechtman, Department Main St. West, Hamilton,
143rd annual meeting of York, May 12-17, 1990. 8, 1 992; accepted Feb. 7, and Biomedical Sciences, Psychiatry, University of of Nuclear Medicine, Ont. Address reprint re-
of Psychiatry, Ont., Canada
McMaster UniL8N, 3ZS.
Dr. Cleghorn died June 8, 1992. Supported by grant MA-10214 from the Medical Research Council of Canada, NIMH grant MH-44073, and by Research Associate Awards from the Ontario Mental Health Foundation to Dr. Brown and Dr. Szechtman. Copyright © 1992 American Psychiatric Association.
1062
group
examined regional brain glucose metabolism in schizophrenic subjects with persistent halluci(N=9) and compared them with a matched
of patients
(N=l0)
Am
and
] Psychiatry
normal
control
1 49:8,
August
subjects
1992
CLEGHORN,
FIGURE 1. Three Slice Levels of the Brain Examined in 22 Schizophrenic Patients With and Without Auditory Hallucinations During PET Scana
FRANCO,
TABLE 1. Demographic Characteristics and Without Auditory Hallucinations
SZECHTMAN,
of Sch izophrenic
Hallucinating Subjects (N=12)
ET AL.
Patients
With
Nonhallucinating Subjects (N=10)
Item
Mean
SD
Mean
SD
Age (years) Education (years)a Length of medicanon (years) Age at onset of
24.2 11.1
7.4 1.8
29.5 13.0
9.7 2.2
0.8
2.6
1.0
2.1
18.8
2.9
23.6
4.5
4.5
5.5
5.9
5.8
1.3 I .2 98.1 81.6
0.6 0.8 14.6 18.8
1.2 1 .0 105.4 99.4
0.6 0.8 12.6 17.2
7.7
2.5
10.9
3.1
1.0
1.9
1.0
schizohrenia (years) Length of illness
(years) Number
aLow:
orbitofrontal-superior
an inferior frontal
part
including
marginal
temporal
of Wernicke’s Broca’s
including
region-inferior region-parietal;
auditory
cortex
and
parietal;
middle:
pre-
high:
prefrontal-supra-
and parietal.
METHOD Patients
sodes Use of street drupC Performance IQ Memory quotiente Picture arranpement score SADS-C positive
symptomscores5 3.0 SADS-C negative symptomscores1’ 4.3 at2.25, df=20, p=O.04 (two-tailed test bt300, df=20, p=0.OO7. Cl .0=prior substance abuse to a mild or criteria. dt,2.10, df=20, et2.31, df=20,
We recruited patients from the emergency and outpatient services who met presumptive DSM-III-R criteria for schizophrenia. They were then admitted to the hospital and followed as outpatients for at least 1 year. Patients with a history of neurological disorder and majon medical illness were excluded. Detailed developmental histories were obtained from family members. Histories included the onset of negative and later positive symptoms; this information, when combined with follow-up data, permitted the diagnosis of schizophrenia by DSM-III criteria. It should be understood that the patients classified as not having hallucinations had experienced hallucinations at various times before the procedure. The mean age of the patients who did experience hallucinations during PET scan was 24.2 years (SD=7.4). Eight were men and four were women. Nine patients were reexammed after 1 year of neuroleptic treatment. Five had hallucinated during the scan, but all had auditory hallucinations before and after the scan. The patients who did not experience hallucinations during PET scan had a mean age of 29.5 years (SD=9.7). Eight were men and two were women. All but three patients were righthanded. The two left-handed patients and one mixedhanded patient were in the group that experienced hallucinations. Only two patients who did not and one patient who did experience hallucinations had in the past been treated with neuroleptic medications, and they had been drug free for many months before the study. All patients had been ill for 6 months or more, some for many years. There was no significant difference between the groups’ length of illness: the mean for patients who did experience hallucinations was 4.5 years
Am]
of epi-
Psychiatry
149:8,
August
1992
t=2.62, t=2.80, ht2.67,
df=20, df=20, df=20,
0.8 2.0 for all comparisons). moderate
degree
0.8
by DSM-III
p=O.OS. p=O.O3. p=O.O2. p=O.Ol. p=O.Ol.
(SD=5.S), and the mean for patients who did not expenience hallucinations was 5.9 years (SD=S.8). However, the mean age at onset for the patients who did expenience hallucinations was significantly younger than that of the patients who did not experience hallucinations during PET scan (p=O.007) (table 1). The patients who experienced hallucinations also had significantly less education and a lower Wechsler perfonmance IQ and memory quotient. Picture arrangement scores were significantly lower in the patients who experienced hallucinations (table 1). None of the 60 additional neuropsychological tests we used discriminated between groups. These measures are described in Cleghorn et al. (4). Type I errors are possible here because of the large number of comparisons. Both positive and negative symptoms were more severe among the patients who experienced hallucinations (table 1). The patients who experienced hallucinations had higher hallucination scores than those who did not experience hallucinations when measured before scanning (mean=4.3, SD=0.8, and mean=3.0, SD=O.7, respectively) as well as during the procedure (mean=4.1, SD=1.2, and mean=0, nations occurred more frequently
did experience ability of their
SD=0). in the
Since hallucipatients who
hallucinations, there was a higher proboccurring during the scanning procedure.
Since group differences reflect the severity of any
in brain metabolic of the symptoms
data might or neurocog-
1063
AUDITORY
HALLUCINATIONS
nitive impairments rather than the presence or absence of hallucinations during the scan, correlations between these variables and hallucination scores during scanning were calculated. None of the symptom scale scores, the mean positive and negative symptom scores, or neurocognitive measures were significantly related to hallucination scores during the scanning procedure or before it. Thus, these hallucinatory events are independent of the other measures but might well be related to factors underlying severity of illness. Since these factors might be reflected in regional brain metabolism, we examined the relation of all symptom scores to regional brain metabolism. One clear difference between groups is that they did or did not hallucinate during the FDG uptake period before the scan. Hallucinations occur intermittently, while the other symptoms are more constant.
Procedure Informed consent, following a full explanation of all procedures employed in the study, was obtained from all subjects. An experienced psychiatric nurse cared for the subjects and inserted a number 18 angiocatheter into a vein of the forearm. Subjects were injected with FDG, S mCi, while lying on a comfortable stretcher. They were instructed to keep their eyes closed and were not spoken to. The only noise was due to the air handling system of the building. Warming the arm produced arterialized samples of venous blood that were taken from the arm opposite that used for FDG injection for the determination of plasma glucose and FDG concentrations. Scanning was started 45 minutes after FDG injection. Local rates of cerebral glucose metabolism were calculated according to the method of Brownell et al. (5). The procedures and methods have been more fully described elsewhere (3, 6). The Schedule for Affective Disorders and Schizophrenia-Change Version (SADS-C) (7) was administered immediately before the glucose uptake period by the nurse so that ratings of positive and negative symptoms, initially obtained on admission, could be checked. Scales for the assessment of five negative symptoms (Scale for the Assessment of Negative Symptoms) and four positive symptoms (Scale for the Assessment of Positive Symptoms) were 6-point scales in which each level of intensity was defined in behavioral terms (8) (table I ). The ratings are based on interview data and on behavioral observations by staff and family members. The patients were well-known to the highly experienced research nurse (S.F.). Immediately before being placed in the tomograph, patients were asked to report any auditory hallucinations they had during the uptake period for FDG. These reports were recorded and reviewed together with an experienced clinician (J.M.C.), and a consensus rating was made. The mean hallucination score for the hallucinating patients was 4.1 (SD=L2). A rating of 3 describes the hallucination as vivid, occasional, and bothersome; a rating of 4, as vivid and frequent.
1064
Local
Cerebral
Glucose
Metabolism
Regional distribution of FDG was measured with the McMaster PET, which has a spatial resolution of 8 mm (full width at half maximum) in the plane and 12 mm in the axial direction (9). It can therefore resolve voxels of less than I cm3. At least 106 counts were collected, in 3 minutes, for each slice. The subject’s head was held in a Perspex restraining device. Alignment in the left-toright axis was ensured by inspection of a preliminary scan taken at the level of the thalamus. Typically, 16 slices of the brain (on a plane S to the orbitomeatal line) were obtained sequentially from the level of the hippocampal gyni up to the level of cortical vertex. Although each slice was 15 mm thick, it overlapped its predecessor by 10 mm. With this procedure, the slice with greatest volume of gray matter of interest for each subject can be identified and selected. For this study, three slices were first selected for analysis (figure 1). The lowest slice (midthalamic-midstriatal slice) contamed the largest amount of temporal cortex, thalamic, and striatal tissue; auditory cortex is defined as segments 10 and I I (figure 2). The procedure for dividing the cortex into segments is shown in figure 2. The middle slice (midcallosal slice) passed above the basal ganglia through the midportion of the dorsoventral extent of the corpus callosum. Language regions of the brain were approximately located by transposing brain stimulation maps onto the PET image, as described by us (3), and are shown in figures 1 and 2. Data
Analysis
In order to test the hypothesis that language areas of the brain are involved in auditory hallucinations, the groups of patients who did and did not hallucinate duning FDG uptake were compared. A two-way analysis of variance (ANOVA) was used to examine group and laterality differences and their interactions in each region of interest, followed by t tests (two-tailed) for post hoc comparisons. Frontal and parietal relative metabolism was later examined by using the same methods.
RESULTS The groups did not differ significantly in absolute whole brain glucose metabolic rate: the mean for patients who did experience hallucinations was 34.50 j.tmol/100 g/min (SD=l0.l0), and the mean for patients who did not experience hallucinations was 32.40 mol/ 100 g/min (SD=6.70). There were no significant differences between the means of tomographic slices within or between groups. Relative
Regional
Metabolism
To test the hypothesis are involved in auditory compared by two-way
Am
and
Laterality
that
language areas of the brain hallucinations, groups were ANOVA (Group by Laterality)
] Psychiatry
1 49:8,
August
1992
CLEGHORN,
FIGURE Auditory
2. Cortical Rim, Divided Into 36 Segments Hallucinations During PET Scana
Each Spanning
Low Slice Left
an Arc of 100, Examined
Middle
Right
in 22 Schizophrenic
Slice
Left
Frontal
FRANCO,
SZECHTMAN,
Patients
ET AL.
With and Without
High Slice
Right
Frontal
Left Right Frontal Clngulate
Broca’s
Anterior
Temporal Auditory
Posterior
Parletal
Parietal aThe
borders
were
defined
Auditory cortex is defined center of the brain. Broca’s rim of the cortex.
by an edge-detecting
algorithm.
as segments 1 0 and 1 1 and region is defined as segments
In the low
Wernicke’s region 6-8 in that slice
and by t tests in each region in the temporal and supramarginal areas and Broca’s regions in frontal cortex (figure 1 ). Since the striatum has a role in language, it was also included in the analysis. The results show trends toward higher relative metabolism in the right hemisphere regions homologous for Broca’s and lower relative metabolism in primary auditory areas and in the right superior temporal region, which includes part of auditory association cortex, in the patients who did experience hallucinations. For the right hemisphere homologue of Broca’s, the mean relative metabolism in the patients who did experience hallucinations was 1.04 (SD=0.04), and the mean in patients who did not experience hallucinations was 1.03 (SD=O.0S) (t=1.77, df=20, p=O.O9). For the right auditory, the mean in patients who did experience hallucinations was 1.03 (SD=0.05), and the mean in patients who did not experience hallucinations was 1.07 (SD=0.0S) (t=l.82, df= 20, p=O.O8). For the right superior temporal, the mean in patients who did experience hallucinations was 1.01 (SD=0.04), and the mean in patients who did not experience hallucinations was 1.05 (SD=0.04) (t=l.91, df= 20, p=O.O7). The differences reach significance in the posterior half of the superior temporal gyrus, the mean of regions 9-14 on the left and 23-28 on the right (figure 2) (F=6.59, df=l, 20, p=O.O2), with the patients who did experience hallucinations having lower relative metabolism (mean=l.03, SD=O.02) than patients who did not experience hallucinations (mean= 1 .05, SD=0.02). When the left-handed patients were removed from the analysis, the posterior temporal values were significantly lower in the patients who did experience halluci-
Am
J
Psychiatry
1 49:8,
August
1992
slice
central
structures
as segments
(see text).
are 9-1
corpus
3. In the
striatum middle
and, slice
In the high slice the cingulate
to the posterior,
no structures
gyrus
are
is visible,
thalamus. visible
in the
as well as the
nations (mean=1.03, SD=0.02) than in the patients who did not (mean=1.05, SD=O.02) (F=8.74, df=1, 17, p= 0.009). The mean of right and left auditory regions was also significantly lower in the patients who did experience hallucinations than in those who did not (mean= 1.01, SD=0.03, N=9, versus mean=1.0S, SD=0.04, N=
10) (F=6.42,
df=1,
17, p=0.02).
There was also a significant Group by Laterality interaction in the anterior temporal region (F=4.57, df=l, 20, p=O.OS). When the three left-handed patients were removed from this analysis, the interaction became nonsignificant. An interaction between group and laterality in the striatum was significant whether or not left-handers were deleted (F=4.78, df=1, 20, p=O.O4); in patients who did experience hallucinations, the mean left relative metabolism was 0.98 (SD=0.09) and the mean right was 0.98 (SD=0.09); in patients who did not experience hallucinations the mean left relative metabolism was 0.94 (SD=0.08) and the mean right was 0.98
(SD=O.0S). There were no significant differences between groups in relarive metabolism in supramarginal, frontal, or parietal regions or in the anterior or posterior cingulate gyrus or thalamus, as depicted in figures 1 and 2. Correlations Between Regional Metabolism Hallucination
Symptom
Scores
and
Relative
correlated with stniatum (r= one-tailed, Pearson product-moment correlation), left stniatum (r=0.69, p
