Comparison of percent change in total and regional gray matter (GM)volumes between 23 patients with childhood-onset schizophrenia (COS), 19 patientswho were multidimensionally impaired (MDI), and 38 healthy control subjects(HC) across 2½ years. The P values were obtainedwith 1-way analysis of variance with Tukey Honestly Significantly Differentpost hoc testing. Error bars indicate SD. The MDI and HC groups do not differsignificantly for any measure (all P>.6).
Gogtay N, Sporn A, Clasen LS, Nugent TF, Greenstein D, Nicolson R, Giedd JN, Lenane M, Gochman P, Evans A, Rapoport JL. Comparison of Progressive Cortical Gray Matter Loss in Childhood-OnsetSchizophrenia With That in Childhood-Onset Atypical Psychoses. Arch Gen Psychiatry. 2004;61(1):17-22. doi:10.1001/archpsyc.61.1.17
Recent anatomical brain magnetic resonance imaging (MRI) studies show
a striking postpsychotic progressive loss of cortical gray matter (GM) in
patients with childhood-onset schizophrenia (COS), which appears greater than
that seen for adult patients. However, the diagnostic specificity and the
relationship of these changes to drug treatment and cognitive functioning
remain unclear. We performed a comparative prospective brain MRI study in
patients with COS and pediatric patients with transient psychosis with behavior
problems (psychosis not otherwise specified) provisionally considered multidimensionally
impaired (MDI). We hypothesized that cortical GM loss would occur in patients
with COS but not in adolescents with atypical psychoses.
Anatomical brain MRI was performed at baseline and follow-up in 19 patients
in the MDI group (mean [SD] age of 13.3 [3.1] years); in 23 patients with
COS matched for age, sex, IQ score, and drug treatment (mean [SD] age of 13.9
[2.5] years); and 38 healthy control subjects matched for age and sex (mean
[SD] age of 13.3 [3.1] years). The mean (SD) follow-up was 2.5 (0.8) years.
Volumes of the cerebrum and total and regional GM were obtained by using automated
analysis, and percent change in volume across time was calculated. One-way
analyses of variance with post hoc Tukey Honestly Significantly Different
comparisons were performed to examine group differences in the percent change
in GM across follow-up.
The COS group had significantly greater total, frontal, temporal, and
parietal GM loss than did the MDI or healthy control groups; analysis of variance
post hoc P values ranged from .03 to .001. The MDI
and control groups did not differ significantly from each other.
The cortical GM volume loss in COS appears diagnostically specific;
it was not seen in children and adolescents with atypical psychosis. Because
both patient groups had similar early developmental patterns, cognitive functioning,
medications, and hospitalizations, this progressive loss appears to be intrinsic
to COS. An ongoing neurodevelopmental process and/or brain response specific
to the illness could account for these changes.
Childhood-onset schizophrenia (COS), defined as onset of psychotic symptomsbefore the 13th birthday, is a rare and severe form of the illness, whichis continuous with its adult counterpart.1 Longitudinalmagnetic resonance imaging (MRI) studies of brain changes in adult patientswith schizophrenia showed mixed results, with some researchers finding progressivechanges and others noting no differential progression relative to controlsubjects.2,3 However, researchersin other MRI studies, in which tissue classification was used, noted abnormalvolume loss for the whole brain, total cortical, frontal, and temporal graymatter (GM) and progressive expansion of the lateral ventricles, as comparedwith findings in healthy control subjects.2,4- 8 Prospectivebrain MRI studies in patients with COS reveal a more striking progressiveloss of total cerebral volume and total and regional cortical GM volumes duringadolescence, as compared with findings in healthy control subjects.9,10 Effect size comparison indicates thatthis progression is greater in adolescent patients with COS than that reportedin adult patients with schizophrenia.11
These intriguing findings are confounded by the effects of illness severityand the probably related lower cognitive level of subjects with COS. Prospectiveexamination in patients without schizophrenia who were treated with similarmedications could provide a comparison population with similar antipsychotictreatment and chronic illness to evaluate the diagnostic specificity of thesebrain changes.
To address these confounding variables of chronic stress of illness,cognitive level, and medication exposure, we compared the brain developmentalchanges in patients with COS with the changes in patients without schizophreniawho were matched for age, sex, and IQ score. The latter group had receivedcomparable treatment and had been referred for the COS study but were consideredto have psychosis not otherwise specified. These subjects had transient psychoticsymptoms and behavioral dyscontrol.12,13 Atfollow-up, none became schizophrenic but most continued to have behavioraland affective problems necessitating residential or special school placementand long-term medication management.14
We hypothesized that the progressive brain changes in patients withCOS were specific to the illness and not caused by medication exposure, cognitiveimpairment, or other illness-related confounding variables, and thus the patientswho were multidimensionally impaired (MDI) would not show prospective longitudinalloss of brain tissue.
Since 1991, through review of more than 1500 case records and in-personscreening of more than 230 subjects, 72 patients met DSM-IV criteria for schizophrenia with the onset of psychosis before the13th birthday.13,15 Patients witha history of substantial medical problems, substance abuse, or an IQ scorelower than 70 prior to the onset of psychotic symptoms were excluded. Furtherdetails of patient selection are described elsewhere.15 Allsubjects with COS participated in a prospective brain MRI developmental study.16,17
This subgroup of subjects excluded from the COS study had transientpsychotic symptoms but were primarily impaired by disruptive behaviors. Thesepatients were considered to have a psychotic disorder not otherwise specified(psychosis not otherwise specified; 298.9 in DSM-IV)and were considered MDI according to 2 child psychiatrists (N.G., A.S., J.L.R.)using clinical and structured interviews, including the Schedule for AffectiveDisorders and Schizophrenia for School-Age Children.12,13,18
Although these patients reported hallucinations and delusions, theirsymptoms were brief (ie, lasting for only a few minutes) and infrequent (occurringa few times a month), usually in response to stress. Their psychotic symptomswere distressing, but their impairment was mostly attributable to affectiveinstability and associated aggressive behavior. These patients had all beentreated with antipsychotic medications, at doses similar to those used inpatients with COS, before the screening at the National Institute of MentalHealth, Bethesda, Md.12 Although these childrenand adolescents did not fulfill criteria at baseline for schizophrenia orany other major Axis I disorder apart from attention-deficit/hyperactivitydisorder, their morbidity, including recurrent lengthy psychiatric hospitalizations,and early developmental profiles were similar to those in patients with COS.In parallel with the COS group, the patients in the MDI group underwent assessmentclinically and neurobiologically, including MRI, at their entry into the studyand were observed prospectively.
A group of healthy children and adolescents, all of whom underwent MRIat least twice, was selected from a larger prospective study of normal braindevelopment.17 Control subjects were free oflifetime medical or psychiatric disorders as determined by means of clinicalexamination and standardized interview. Psychiatric illness in a first-degreerelative was also exclusionary.
From the 2 patient groups, we were able to match for sex, age, and medicationstatus 19 patients who were MDI (mean [SD] age of 13.3 [3.1] years) and 23patients with COS (mean [SD] age of 13.9 [2.5] years) who had undergone 2prospective MRI studies. Thus, the only basis for exclusion of subjects withCOS was the failure to match with the MDI group and/or the lack of 2 prospectiveMRI studies. Thirty-eight healthy community volunteers (mean [SD] age of 13.3[3.1] years) matched for age and sex who had undergone 2 prospective MRI studieswere selected as control subjects. Sample characteristics are given in Table 1.
This study was approved by the institutional review board of the NationalInstitute of Mental Health. Parents or legal guardians of all participantsprovided written informed consent, and all patients gave written assent fortheir participation.
All images were acquired with the same 1.5-T imager (Signa; GE MedicalSystems, Milwaukee, Wis) located at the National Institutes of Health ClinicalCenter, Bethesda. A 3-dimensional spoiled gradient-recalled echo in the steady-statesequence designed to optimize discrimination of GM, white matter, and cerebrospinalfluid was used to acquire 124 contiguous 1.5-mm-thick sections in the axialplane and 124 contiguous 2.0-mm-thick sections in the coronal plane. Imagingparameters were as follows: repetition time, 24 milliseconds; echo time, 5milliseconds; flip angle, 45°; acquisition matrix, 256 × 192; numberof signals acquired, 1; and field of view, 24 cm.
Total and regional GM and white matter volumes were generated by a fullyautomated system that classifies tissue according to voxel intensity by usinga probabilistic atlas and provides lobar (frontal, parietal, temporal, andoccipital) parcellation of cortical GM volumes. Total cerebral volume wasdefined as the algebraic sum of all GM voxels and white matter pixels, excludingthe cerebellum and brainstem. Anatomical boundaries for classification ofbrain voxels were based on 4000 points selected by a trained neuroanatomistfor GM, white matter, cerebrospinal fluid, fat, and background by using referencespace and the coordinate system from Talairach space.19,20 Additionalinformation about anatomical boundaries is available at request from the authors21 and on our Web site at http://intramural.nimh.nih.gov/chp/.
Comparisons of patients and control subjects for baseline and follow-upage and imaging interval were examined with 1-way analysis of variance (ANOVA).Sex and handedness for the 3 groups were compared by using χ2 tests,while clinical functioning, full-scale IQ score, and antipsychotic medicationuse in the 2 patient groups were compared by using t tests.Percent change in total cerebral and total and regional GM volumes was calculatedand compared between the 3 groups. One-way ANOVA was performed for percentchange differences in total cerebral volume and total and regional (parietal,frontal, temporal, and occipital) GM volumes between the 3 groups, followedby post hoc Tukey Honestly Significantly Different testing for significantgroup differences.22 The SPSS 10.0 statisticalpackage for Windows (SPSS Inc, Chicago, Ill) was used for all analyses; a2-tailed P value of .05 was considered to indicatestatistical significance.
As seen in Table 1, the3 groups did not differ significantly with respect to their age at baselineor follow-up, the length of follow-up, sex, or handedness. The 2 patient groupshad lower IQ scores than did control subjects but did not differ significantlyfrom each other. The patients had developed psychotic symptoms at a youngage; in patients with COS, mean (SD) was 10.13 (2.1) years, and in patientsin the MDI group, mean (SD) was 7.8 (1.9) years. The patients also had hadconsiderable exposure to antipsychotic medications during the course of illness(COS, 3.8 [2.0] years; MDI, 5.5 [3.0] years), with most patients having had2 or more years of continuous antipsychotic medication exposure at baseline.Also at baseline, typical and atypical antipsychotic medication exposure wasapproximately equal between the 2 patient groups. At follow-up, most patientsin both groups continued to receive atypical antipsychotic medications, although15 (65%) in the COS group were receiving clozapine, while none of the patientsin the MDI group was. The use of mood stabilizers was more frequent in theMDI group, as compared with findings in the COS group both at baseline (MDI,10 [53%]; COS, 5 [22%]) and at follow-up (MDI, 10 [53%]; COS, 2 [9%]).
At baseline and at follow-up, both patient groups were significantlyclinically ill as assessed by means of the Clinical Global Impression Scale,Brief Psychiatric Rating Scale, Scale for the Assessment of Positive Symptoms,and Scale for the Assessment of Negative Symptoms, but the COS group was relativelymore impaired. In both patient groups, there was no significant change inIQ score at follow-up. Clinical improvement measures in the COS group at follow-updid not correlate significantly with the percent change in GM loss (Pearsoncorrelations all >0.1).
Table 2 shows total andregional GM volumes for all 3 groups at MRI 1 and MRI 2 and the percent changeacross follow-up. As seen in Table 2,the reduction in GM volumes for the COS group was greater than for both theMDI and control groups.
The GM loss (percent reduction) for the 3 groups is plotted graphicallyin Figure 1. One-way ANOVA comparisonsof percent change in total and regional GM volumes between the 3 groups wereall significant (F2,77 = 2.7 to 7.7, P =.03 to .001). Post hoc Tukey Honestly Significantly Different tests, whichaccount for multiple comparisons, revealed that the percent reductions forthe total and regional GM volumes were significantly greater in the COS groupthan in the MDI and control groups (Figure1). The MDI and control groups did not differ significantly fromeach other.
The progressive loss of cortical GM previously reported in patientswith COS appears to be diagnostically specific. The comparison group of adolescentpatients with transient psychotic symptoms and chronic affective and behavioraldisorders (the MDI group) did not show progressive cortical loss. The MDIgroup as a whole (n = 29) at follow-up of 2 to 8 years exhibited several clinicaltrajectories, with approximately half (n = 15) having a diagnosis of chronicaffective disorder or psychotic disorder, while the others (n = 14) retainedthe diagnosis of psychosis not otherwise specified.14 Itis thus probable that these illnesses have different neurodevelopmental trajectoriesdistinct from that of schizophrenia. This finding is in agreement with thoseof Pantelis et al23; in their study, subjectsat high risk who were not receiving medication and developed affective illnessdid not show progressive loss of brain tissue, while the ones progressingto first episode did. It is, however, important to note that although no progressivechanges were seen for MDI in our study, these are crude whole-lobe measuresof GM volume. It is probable that more localized regional measures will indicatesubstantial changes in select cortical regions for the MDI group (eg, medialbrain structures), which may show abnormal trajectories with the onset ofaffective psychosis, as seen in the neuroimaging studies of mood or affectivedisorders.24,25
At initial MRI, despite longer exposure to antipsychotic medications(mean [SD] 5.5 [3.0] years for MDI and 3.8 [2.0] years for COS), the MDI groupdid not have smaller GM volumes, as compared with those in healthy controlsubjects (unlike findings in patients with COS who had smaller GM volumes,as compared with findings in control subjects at initial imaging), which suggeststhat the GM reduction is specific to schizophrenia. Although the MDI groupwas comparable with the COS group for medication and hospitalization at firstMRI, differences in medication status were seen at follow-up.
Although most patients in the MDI group at follow-up continued to receiveatypical antipsychotic medications, none was receiving clozapine, while mostpatients with COS (15 [65%]) were receiving clozapine. The effect of clozapineon cortical GM has not been well studied either in animals or humans. In animalstudies, clozapine can increase the striatal volume in rats undergoing long-termtreatment,26,27 and in primates,antipsychotic treatment, including clozapine, can result in increased glialdensity, which can result in cortical hypertrophy.28 Resultsin human studies support the increase in striatal volume with typical antipsychotictreatment, which is normalized with atypical antipsychotic medications.29,30
However, there is no evidence that clozapine, as compared with otheratypical agents, differentially influences either the degree or the rate ofprogression of the GM volume changes seen in schizophrenia. When the slopesof GM loss in patients with schizophrenia receiving clozapine (n = 13) werecompared with those in patients not receiving clozapine (n = 25) from ourlarger sample of patients with COS (n = 36) studied longitudinally, therewas no significant difference between the slopes (t36 = 0.8, P = .4).31 Similarly,for our larger COS cohort (n = 36), slopes of GM loss for patients receivingclozapine (n = 19) did not differ significantly from those for patients receivingolanzapine (n = 13) for 2 to 4 years (t30 =0.77, P = .4).31
In agreement with our findings, other investigators have also foundno significant association between clozapine treatment and prospective brainchange32 (Daniel Mathalon, MD, oral communication,November 2002). Lieberman et al33 in a multicenterstudy reported potential protective effects of olanzapine vs haloperidol usefor 3 months to 2 years; however, we do not find evidence for a protectiveeffect of olanzapine vs clozapine for COS at long-term follow-up.
More patients in the MDI group were receiving mood stabilizers as comparedwith patients in the COS group at baseline and follow-up. Mood stabilizersare claimed to have neuroprotective and neurotropic effects, which could causehypertrophic changes in cortical GM.34,35 However,the mean slopes for patients receiving mood stabilizers in the MDI group (n= 10) did not differ significantly from the mean slopes for patients in theMDI group not receiving these medications (n = 9; t16 = 0.75, P = .46). Again, within the largersample of patients with COS (n = 36) studied longitudinally, no significantassociation was found between GM reduction or slope, in loss of millilitersper year, and the groups receiving medication, including the 8 subjects receivingmood stabilizers (t30 = 0.06, P = .9).31 Thus, the progressive GMloss seen in patients with COS appears to be independent of medication status.
The mechanism underlying GM loss in COS remains unknown. Either a progressiveneurodevelopmental process and/or a specific plastic brain response to schizophreniacould account for this loss. Contrary to our prior findings,36 alarger sample (n = 50) of patients with COS showed no postpsychotic cognitivedecline at 2 to 8 years (P. Gochman, MA, unpublished data, November 2003).Thus, the pronounced GM loss does not appear to be directly related to theloss of cognitive functioning. We speculate that the GM loss represents inpart a plastic response of the brain to dysfunctional synaptic processing.Further studies will be necessary to test this hypothesis and its relationshipto clozapine or other treatment.
In summary, patients with atypical nonschizophrenic childhood-onsetpsychotic disorders did not show the exaggerated longitudinal GM loss seenin a matched group of patients with COS. These results suggest that the progressiveGM loss seen in COS during adolescence is specific to the illness and unrelatedto medication exposure. Ongoing longitudinal MRI studies of full siblingsof patients with COS are addressing whether this abnormal development maybe a trait marker.
Corresponding author: Nitin Gogtay, MD, Child Psychiatry Branch,Bldg 10, Room 3N 202, 10 Center Dr, MSC 1600, Bethesda, MD 20892-1600 (e-mail: firstname.lastname@example.org).
Submitted for publication February 24, 2003; final revision receivedJune 9, 2003; accepted June 12, 2003.