The relationship between schizophrenia and intelligence is complex: decrease in IQ can reflect impaired central nervous system development, cognitive deterioration, and/or clinical state. Although there is consensus about an initial cognitive deterioration before and after onset of schizophrenia, further decline in intellectual performance with chronic illness remains controversial, with most studies finding no postpsychotic decline (1).
Most published studies do not examine non-age-corrected (raw) IQ scores, a more valid index of possible dementia than age-corrected scores. Change in raw scores is more likely to reflect dementia because it allows direct comparison of performance without penalty for increasing age.
Longitudinal assessment of cognitive deterioration in childhood-onset schizophrenia (onset of psychosis by age 12 years) has yet to be reported. This is of particular interest because childhood/adolescence is a time of continued brain reorganization (2, 3) and new learning. To address this, a prospective longitudinal study was carried out in which postpsychotic IQ testing was repeated at approximately 2-year intervals.
Our initial hypothesis was that postpsychotic decline in full-scale IQ reflected a true loss of general knowledge and commonsense knowledge (as measured by raw information and comprehension scores, respectively).
Patients aged 6 to 18 who had been diagnosed as schizophrenic with onset of psychotic symptoms by age 12 were recruited for an inpatient trial of the atypical neuroleptic drugs clozapine or olanzapine. Recruitment and diagnostic procedures have been described elsewhere (4). Parents and patients provided written informed consent or assent for participation in the study.
As of April 1999, 48 patients with childhood-onset schizophrenia participated in the study. Thirty-six (75%) of these patients had more than one valid postpsychotic full-scale IQ score as rated by the administering psychologist. Thirty-one (86%) of these 36 patients had more than one subtest raw score from the same test version.
The scaled and raw subtest scores of these 31 patients form the basis for this study. The group included 21 (68%) boys and 10 (32%) girls; their mean age at onset of psychosis was 10.0 years (SD=1.8), and they had a mean of 24.4 months (SD=19.9) of exposure to typical or atypical neuroleptics before their admission to our facility. At the time of final IQ testing, 27 (87%) were receiving an atypical neuroleptic, two (7%) were receiving a typical neuroleptic, and two (7%) were unmedicated. Their mean scores on the Scale for Assessment of Positive Symptoms (SAPS) and the Scale for Assessment of Negative Symptoms (SANS) were 37.4 (SD=20.9) and 55.0 (SD=26.9), respectively. These 31 patients did not differ significantly from the rest of the patients with childhood-onset schizophrenia with respect to demographic, clinical, or neurobiological measures.
Raw scores from all subtests obtained after the onset of psychosis were examined for cognitive decline. For patients younger than 17, either the WISC-R or the WISC-III was administered to match their previous testing. For patients older than 16, the age-appropriate WAIS-R was administered, and, in addition, either the WISC-R or WISC-III information and comprehension subtests were repeated for consistency with previous testing. Information and comprehension subtests were chosen because they are thought least affected by brain damage and may approximate premorbid functioning (5); therefore, a drop in the raw scores of either of these subtests might be more likely to indicate a dementing process.
The mean age at first testing was 12.3 (SD=2.6). The mean number of test administrations was 2.6 (SD=0.9, range=2–5), with a mean of 2.9 years (SD=1.4, range=0.8–6.3) between tests.
All initial and follow-up magnetic resonance imaging (MRI) scans were obtained on the same GE 1.5-T Signa magnetic resonance scanner as described in detail elsewhere (2, 6). Both ventricular and hippocampal values were available.
Linear regression was used to determine the slope for all IQ scores for each subject. The Wilcoxon one-sample test was used to determine whether these regression lines differed significantly from zero.
The relationships between information and comprehension raw score slopes and sex, socioeconomic status, age at onset of psychosis, months of previous hospitalization, months of neuroleptic exposure, SAPS score, and SANS score were examined with Spearman correlations.
Spearman correlations were also used to examine change in lateral ventricular and hippocampal volume between MRI scans (mean=4.2 years, SD=2.3) and change in information and comprehension raw scores over the same time period. Sixteen patients had IQ tests administered during the same weeks as the two MRI scans, which limited this analysis.
A significant decline was found in postpsychotic full-scale IQ scores: mean slope=–0.21, SD=0.40 (Wilcoxon T=115, N=36, p=0.001). Three of 11 age-scaled subtest scores declined significantly: picture arrangement (Wilcoxon T=58, N=28, p=0.003), information (Wilcoxon T=93, N=30, p=0.02), and block design (Wilcoxon T=83, N=29, p=0.01).
The raw scores for all postpsychotic subtests were then examined to determine if a decline was present in any measured area of functioning. Raw scores did not increase as would be expected for healthy children, but the group did not show an absolute decline for any subtest. Slopes of the subtests of greatest interest (information and comprehension) were close to zero and suggested no change. For information, mean slope=0.03, SD=0.10, range=–0.22–0.30 (Wilcoxon T=245, N=31, p=0.18); for comprehension, mean slope=0.05, SD=0.21, range=–0.31–0.67 (Wilcoxon T=223, N=30, p=0.42).
Brain imaging analysis revealed a significant correlation between decrease in hippocampal volume and smaller increment in information raw score (rs=0.49, N=16, p=0.05).
There was no significant relationship between information and comprehension raw score slopes and any clinical measure except that patients with longer hospitalizations showed less increase in the comprehension raw score (rs=–0.45, N=30, p=0.01).
No significant relation was found between initial full-scale IQ and subtest slope scores, indicating that ceiling and floor effects were unlikely.
The striking decrease in postpsychotic full-scale IQ in our group of severely ill patients with childhood-onset schizophrenia stands in contrast to the majority of studies in adult-onset schizophrenia (1). Possible explanations include the scoring of IQ during school years, the greater progression in brain abnormality seen with these early-onset cases (6, 7), or greater severity of illness in these cases. Although no relationship was found between clinical status and IQ change scores, almost all patients were severely ill, limiting interpretation of this analysis.
The significant relationship between decrease in hippocampal volume and smaller increase in information raw score may be due to excessive synaptic pruning in the hippocampal region or other trophic change, as discussed in detail elsewhere (6).
Limitations of this study include the fact that the subjects were very ill and treatment resistant and the use of published norms for IQ comparisons. However, converging data suggest that adolescence provides the optimal window for investigating cognitive and brain changes characteristic of schizophrenia (6).
Received Dec. 1, 1998; revision received May 3, 1999; accepted May 7, 1999From the Child Psychiatry Branch, NIMH. Address reprint requests to Dr. Rapoport, Child Psychiatry Branch, NIMH, 9000 Rockville Pike, 10/3N202, Bethesda, MD 20892.The authors thank Dr. Terry Goldberg of the Clinical Brain Disorders Branch, NIMH, for his help with this project.