Correspondence address: YoshiroOkubo MD Section of Clinical Physiology, School of Allied Health Sciences, Faculty of Medicine, Tokyo Medical and Dental University, 1-5-45 Yushima, Bunkyo-ku, Tokyo 113-8519, Japan. Email: firstname.lastname@example.org
Abstract Recent studies of the brain using magnetic resonance imaging (MRI) have suggested progressive structural changes in schizophrenics. However, those studies were conducted over periods of less than 5 years and thus lacked sufficient capacity to determine the course and nature of this process. In this study, MRI scans were obtained in 15 schizophrenics and 12 controls at baseline and after 4- and 10-year follow ups. Volumes of the lateral ventricles were measured. Patients were assessed by the Brief Psychiatric Rating Scale (BPRS) at the same two time points: at baseline and at 10-year follow up. After 10 years, a significant lateral ventricular enlargement was found in patients (mean percentage change: +22.9%) but not in controls (5.1%). Although our results are not in disagreement with the neurodevelopmental hypothesis, they do provide strong evidence that in schizophrenia progressive brain reduction occurs even in its chronic stage.
The first half of the century was witness to a period of vigorous research, from which a plethora of abnormalities occurring in the brains of patients with schizophrenia (or dementia praecox) were identified and described. Schizophrenia was believed to be caused by a new form of progressive neuronal degeneration characterized by earlier onset than that seen with previously described entities, such as Huntington's disease or Alzheimer's disease. In recent years, however, the original Kraepelinian pathogenetic theory of premature progressive neuronal degeneration has been opposed by a pathogenetic model that postulates that schizophrenia results from a non-progressive pre- or perinatal derangement of development.1–3
The essence of the case against a neurodegenerative mechanism is that gliosis, which is regarded as a necessary neuropathological hallmark of neuronal degeneration, has not been found in a number of post-mortem studies in examinations of brains of schizophrenics.4 Furthermore, the hypothesis that schizophrenia is a disorder caused by early and static damage has been supported by computed tomography (CT) studies which demonstrated ventricular enlargement at the onset of schizophrenia and failed to find any progressive structural changes.5
In contrast to the CT studies, some recent longitudinal magnetic resonance imaging (MRI) studies have suggested the existence of progressive structural changes in the brains of schizophrenic patients. The findings include enlargement of the bodies of the lateral ventricles, whole hemispheric volume reduction, and some changes in frontal lobes and the superior temporal gyrus.6–12 These findings have led to the proposal of a ‘two hit’ model of brain structural changes, where the neurodevelopmental event is necessary, but it is still not enough to explain the latter abnormalities after the onset of symptoms.13 However, these longitudinal MRI studies were conducted over periods of no more than 5 years, lacking a sufficient range of follow up to determine the course and nature of this progressive process. Further, the timing of the development of these abnormalities in the course of the illness and their relationship to the evolving morbidity have remained unclear. Consequently, there was a definite need for cohort studies to monitor brain changes over longer periods of time.
In the present study, 15 chronic patients and 12 controls were re-scanned at 4- and 10-year intervals. To our knowledge, this study was based on the hitherto longest follow-up term for the examination of brain morphology in schizophrenia patients and healthy comparison subjects using MRI. The present study aimed to determine whether schizophrenia is associated with progressive structural brain change.
MATERIALS AND METHODS
Fifteen patients and 12 controls underwent MRI scans at the time of the initial examination from 1988 through 1989 and then after 4- and 10-year follow ups, after written informed consent was obtained from each of the participating subjects.14 The characteristics of the subjects are summarized in Table 1. Fifteen patients met DSM-III-R criteria (American Psychiatric Association, 1987) for the diagnosis of schizophrenia (one paranoid type, three residual type, 10 disorganized type, one undifferentiated type) and were selected from patients at Asai Hospital, Togane, Chiba, Japan. These patients consisted of nine males and six females. The mean age at initial scan was 37.5 ± 8.9 years. They were in the chronic stage of illness. They had no history of neurologic disorders, metabolic disorders, drug abuse or alcoholism, and had not received electroconvulsive therapy. The patients who showed the symptoms of water intoxication and had hyponatraemia were excluded. All of them were hospitalized at the initial scan. Nine of them had been discharged during the follow-up term, but they had been readmitted. Thus, all of patients were hospitalized at the 10-year follow-up scan. They had received neuroleptics continuously over the follow-up interval. Compliance was assessed by medical records. Twelve control subjects (seven males and five females; mean age at initial scan 37.1 ± 4.2 years) recruited from the surrounding community who were healthy, showed no signs of any psychiatric or personality disorder, and had no physical illness, were used as a control group. They were matched as closely as possible to the total patient group by age, gender and the duration of education (Table 1). Average daily dose was quantified as the haloperidol equivalent mg/kg of bodyweight units.
Table 1. Demographic and clinical characteristics*
Patients (9 men, 9 women)
Controls (7 men, 5 women)
Data are reported as mean (SD). –, not applicable.
Education (no. of years)
Age at onset (years)
Patients were clinically assessed using the Brief Psychiatric Rating Scale (BPRS) at two time points: in 1988–1989 (baseline) and in 1998–1999 (10-year follow up). In addition to the total scores, the sum of the three negative symptom subscale scores (blunted affect, emotional withdrawal, motor retardation) and that of three positive symptom subscale scores (suspicion, hallucination, unusual thoughts) were derived from BPRS. Interrater reliabilities (ICC), based on ratings of 10 patients by two sets of psychiatrists at the two time points in the study (one rater consistent across both assessments), ranged between 0.92 and 0.99 for the total scores, the negative symptom subscale scores, and the positive symptom subscale scores.
All baseline and follow-up scans were obtained with the same 0.2-T scanner (HITACHI Medico KK, Chiba, Japan) and performed with the same parameters and slice levels. The T1-weighted coronal images were obtained as follows: slice thickness, 9 mm with 1-mm gaps; field of view, 240 mm; matrix 256 × 256; TR (repetition time), 1500; TE (echo time), 38 ms; and TI (inversion time), 500 ms.
To standardize the levels and angles of MRI slices and to increase the test–retest reliability, we used the following scanning procedure: (i) at first, parasagittal slices including the mamillary body were acquired; (ii) using the parasagittal slice as a guide, one reference slice was designated in the coronal plane that was parallel to the bottom line of the fourth ventricle and ran through the centre of the mamillary body; (iii) parallel to the reference slice, three or more contiguous coronal slices were obtained; and (iv) finally, we confirmed that the reference slice perfectly contained left and right mamillary bodies and had no deviation.
In addition, 10-mm contiguous slices in the coronal and sagittal planes covering the whole cerebrum were also obtained so as to exclude any subjects that had an intracranial pathological lesion at 10 years of follow up.
Although more slices were acquired at the subsequent scans, we allocated the three coronal slices consisting of one coronal slice containing the mamillary bodies and two contiguous slices 10 mm on either side for the measurements, because only these three coronal slices were obtained at the initial scan.
Measurements of the left and right lateral ventricles were obtained from the coronal slices. These volumes were quantified from these three slices by one of the authors (TS), slice by slice, without knowledge of the year of the scan or patient/control status, using a supervised thresholding technique available in the Image program software from the National Institutes of Health.15 To correct for differences in head size, we used a method that divided the ventricular volume by the intracranial space that was derived from the three contiguous slices at each time point.
Test–retest reliability was checked for volume measurements on 10 randomly selected normal individuals. They were scanned and measured twice at an interval of 1 month or more. Intraclass correlation (ICC) was 0.99 for left and right lateral ventricles.
Comparison with the entire volume obtained by 3-D method
To ascertain that the measurements on the three coronal slices reflect the volume of the lateral ventricles, we compared the data from measuring three coronal slices with the data from measuring all slices covering the entire head. We collected T1-weighted coronal images of the entire head in eight patients and four controls by three-dimensional (3-D) method as follows: slice thickness 3 mm with no gap, field of view 240 mm, matrix 256 × 256, TR 100, TE 23 ms, and flip angle 90°.16 The entire volume of the lateral ventricles obtained by 3-D method was significantly correlated (r = 0.97, P = 0.0001) with the data obtained by measuring three coronal slices. Thus, we regarded the data obtained in this study by measuring three coronal slices to reflect the entire volume of the lateral ventricles.
The volumes of the lateral ventricles were analysed using repeated-measures analysis of variance (ANOVA) with the Greenhouse–Geisser correction. Diagnosis was treated as a between-subjects variable, and side and time as within-subject (repeated measures) variables. Time by diagnosis interaction was investigated to evaluate the diagnostic group difference in structural changes over 10 years. When there was a significant interaction between time and diagnosis, Scheffé's post-hoc test after two factor (time × side) ANOVA was conducted separately for the diagnostic groups to determine the time points at which the brain-volume changes occurred.
Within the schizophrenic group, exploratory correlation analysis (P < 0.05) was conducted to examine the relationship between structural change and clinical changes. Change rates were calculated for MRI and clinical measurements as follow-up measurements minus initial measurements. Medication effects were examined by correlating the daily neuroleptic dose for the period between initial and follow up with the structural changes.
The average and standard deviation (SD) of the volume of the lateral ventricles at each time point are presented in Table 2. A significant main effect for time (F2,50 = 31.44, P = 0.0001) and a significant time-by- diagnosis interaction (F2,50 = 8.96, Greenhouse–Geisser P = 0.0007) were confirmed, while time-by-side interaction (F2,50 = 0.34, P = 0.67) was not significant. This means that the rate of ventricular enlargement did not significantly differ between left and right hemispheres, but differed significantly between the diagnostic groups. This change was greater in patients (mean percentage change: +22.9%) than in controls (mean percentage change: +5.1%) as shown in Fig. 1. The post-hoc test after ANOVA, which was calculated separately for the diagnosis, showed significant increases in the 4–10-year and in the 0–10-year follow-up terms for the patients (Fig. 2), whereas the controls showed no significant increase after the initial scan.
Table 2. Lateral ventricles volume% of 15 patients and 12 controls at baseline and 4- and 10-year follow ups
Volume at baseline (%)
Volume at 4-year follow up (%)
Volume at 10-year follow up (%)
Patients Mean (SD)
Controls Mean (SD)
Patients Mean (SD)
Controls Mean (SD)
Patients Mean (SD)
Controls Mean (SD)
Diagnosis-by-time interaction: F(2,50) = 8.96, P = 0.0007.
There was a trend toward correlation between changes in the volume of lateral ventricles of the patients during the follow-up interval and a worsening of BPRS negative symptom subscale scores (r = 0.43, P < 0.1) (Fig. 3), while no significant association was found between structural changes and BPRS total or positive symptom subscale scores.
Daily haloperidol equivalent exposure dose over the follow-up interval was not associated with ventricular enlargement for patients (r = 0.35, P = 0.20).
There was no time by sex interaction in any MRI measurement. Age of onset and duration of illness were not significantly correlated with the structural brain changes within the schizophrenic group.
To standardize the methods for the volume measurements at the different time points, we necessarily used the MRI protocol that was in use 10 years ago. Thus, it should be considered that the thick slices (9-mm thickness with 1-mm gap) of the old scan protocol might contribute to errors from partial volume effects and difficulties in the standardization of scanning planes, and might also reduce the test–retest reliability of the measurements. To minimize these potential errors, we measured only the ventricles because of their clear contrast with the surrounding cerebral matter, and we standardized the level and angle of the reference slice with the mamillary bodies as indexes in each subject. Consequently, we established good test–retest reliability for the measurements of the ventricles. The small number of slices used for the measurements might also be considered a limitation. For all of the different time points, only three coronal slices were identical and available for comparison. Thus, we compared the ventricular measurements on the three coronal slices and the entire ventricular volume measurements on the 3 mm contiguous slices obtained by 3-D method, and a good correlation between the data was obtained. This means that the data from the three slices in the present study reflect the entire volume of the lateral ventricle.
Ventricular enlargement over 10 years
Some recent longitudinal MRI studies have provided evidence that excessive brain volume loss occurs and continues to progress after the onset of overt illness.6–8,11,17 However, these studies were conducted on follow ups of no more than 5 years, and some inconsistencies were apparent. In other words, several studies measuring ventricular volume failed to find continued ventricular enlargement.5,11,18–21 It is clear then, that if the range of follow up is insufficient to determine the course and nature of this progressive process, longer follow-up periods are needed for their more precise and reliable evaluation. Thus, we used perhaps the longest follow-up term to date to examine brain morphology in schizophrenic patients and healthy comparison subjects. As a result, we found a significant difference between schizophrenic patients and controls in the rate of brain structural changes over the long term, with schizophrenic patients showing significantly greater ventricular enlargements. The rates of lateral ventricular change per year calculated from the previous studies, ranged between –4.9% and 13% for patients and between –0.6% and 3.3% for controls.6–8,18 Although the methodological difference makes it difficult to compare these values with ours, our data calculated from the same method [i.e. 2.29% (= 22.9%/10 years) for patients and 0.51% for controls] were within the ranges reported in the previous studies.
Many researchers have proposed the neurodevelopmental hypothesis as the causation of schizophrenia.1,2 Although our results are not discordant with the neurodevelopmental hypothesis, they do provide strong evidence that progressive brain reduction occurs in schizophrenia even in the chronic stage of the illness. As for the supposition that these findings may be a direct consequence of schizophrenia's underlying pathophysiology and characteristics of the disease process itself, it must be considered to be of a speculative nature. One possible pathogenetic model is that of excessive neuronal apoptosis.22 Although cell death can occur by apoptosis, it does not lead to inflammatory changes and gliosis. If postnatal pathological neuronal loss could result from non-gliotic apoptosis, the absence of gliosis would no longer limit the time of occurrence of that loss. Aberrant neuronal pruning, which increases neuronal density without cell loss, is also believed to contribute to the brain volume change.23,24
Several lines of neuroimaging studies have demonstrated associations between brain structural abnormalities and schizophrenic symptoms.13 However, there have been few studies investigating symptom changes relevant to progressive structural changes in schizophrenia. Two studies have reported that there was no significant correlation between clinical change measured by BPRS and lateral ventricular change.6,7 One study has suggested volume reduction in the frontal and temporal lobes was correlated with less improvement in negative and some positive symptoms measured by the Scale for the Assessment of Negative Symptoms and the Assessment of Positive Symptoms.11 Our exploratory analysis of the patients revealed a trend toward correlation between ventricular enlargement and worsening of BPRS negative symptom subscale scores. It is likely that symptom changes of schizophrenic patients over time are related to changes in brain morphology. Further study with a larger sample of patients is needed to clarify the alteration in symptoms in relation to brain structural changes over time.
Our results do not argue against the neurodevelopmental hypothesis, but they do provide strong evidence that, in schizophrenia, progressive brain reduction occurs even in the chronic stage of the illness. It may be possible to explain that the clinical symptoms of schizophrenic patients are not the same with the passage of time. In the future, large cohort studies to monitor brain changes over time using up-to-date MRI techniques with a fine resolution for specifying brain structures, which progressively change in the course of the illness, will need to be carried out. Such approaches are likely to result in significant gains in the clarification of the characteristics and pathophysiology of schizophrenia.
Our 10-year longitudinal MRI study revealed progressive ventricular enlargements in patients with schizophrenia but not in controls. Our findings provide compelling evidence that progressive brain reduction occurs in schizophrenia even in the chronic stage of illness.
Professor Toru Nishikawa of the Section of Psychiatry and Behavioral Science, Graduate School of Tokyo Medical and Dental University and Yasutomo Taiji of the Asai Hospital are greatly acknowledged. This work was supported by a Grant-in-Aid for Scientific Research from the Ministry of Education and the research grant for nervous and mental disorders from the Ministry of Health and Welfare of Japan (11B-3).