For almost a century, researchers have suggested that schizophrenia is the result of brain-based abnormalities (Kraepelin, 1919). The first structural neuroanatomic studies conducted in patients with schizophrenia were undertaken in the 1930s and 1940s using pneumoencephalography -- a radiologic technique wherein cerebrospinal fluid (CSF) is replaced by air, thereby facilitating X-ray imaging of the brain ventricular system. These early investigations demonstrated lateral ventricle enlargement in some patients with schizophrenia (Haug, 1962). A few investigators found evidence of progression of ventricular enlargement in severely ill patients; however, technical factors involved in image acquisition limited the veracity of the results.
The first computed tomography (CT) scan study in patients with schizophrenia was published in 1976 by Johnstone et al. The researchers found that, in chronically ill patients, ventricle enlargement was common and most pronounced in those patients who were most ill (i.e., had the most negative symptoms). Since that time, dozens of studies have replicated the finding of ventricular enlargement in schizophrenia (as well as findings of cerebral cortical atrophy) (Shelton and Weinberger, 1986), some found evidence for progression (Illowsky et al., 1988; Nasrallah et al., 1986), while others did not (Kemali et al., 1989; Woods et al., 1990). The discrepancy in results is very likely the result of methodologic differences between studies.
Magnetic resonance imaging studies in schizophrenia began to appear in the psychiatric literature in 1984 (Smith et al., 1984). Since then, this powerful imaging technology has allowed researchers to demonstrate a number of neuroanatomic findings, in addition to confirming findings from CT studies. These include reduction in the size of the hippocampus and amygdala (Hirayasu et al., 1998), superior temporal gyrus (Pearlson, 1997), and frontal and temporal lobes (Woodruff et al., 1997), as well as changes in corpus callosum size (Jacobsen et al., 1997) in patients with schizophrenia. Magnetic resonance imaging offers many advantages over CT scans, such as greater spatial resolution, the ability to separately measure gray and white matter volume, and the ability to image the brain in any plane. Advances in MRI technology, accompanied by the availability of increasingly powerful computerized image analysis technology, have allowed highly detailed and reliable measurements of various brain structures, both cross-sectionally and over time. As with CT studies, longitudinal MRI studies conducted in patients with schizophrenia have yielded mixed results in terms of evidence for progressive neuro-anatomic changes (DeLisi, 1999).
Several recent structural imaging studies have demonstrated progressive neuroanatomic changes in some patients with schizophrenia. These studies have methodologic advantages over prior work, such as more homogenous patient populations, larger study populations, and/or use of sophisticated image analysis techniques. The result of these improvements is greater sensitivity to detect subtle changes in neuroanatomy over time.
Giedd and colleagues (1999) studied 42 patients with childhood-onset schizophrenia (COS) and compared them to 74 control subjects drawn from a neuroimaging database used to study normal brain development. Due to its severity, COS -- a particularly virulent form of schizophrenia with multiple similarities to adult-onset schizophrenia -- may be a particularly fruitful setting in which to conduct studies investigating brain changes related to the illness. Patients and controls were carefully screened and underwent MRI imaging at study entry and at approximately two-year intervals for one to three follow-up timepoints. The MRI scans were used to quantify total volume of the cerebrum, lateral ventricles, hippocampus and amygdala. The volume of these structures and a measure of their left-right laterality in patients and controls were compared at each scan time. The severity of patients' illnesses was quantified by rating global psychopathology (Brief Psychiatric Rating Scale [BPRS]), positive symptoms (Scale for the Assessment of Positive Symptoms [SAPS]) and negative symptoms (Scale for the Assessment of Negative Symptoms [SANS]) at baseline and at the time of each follow-up scan. Statistical comparisons were conducted both with and without controlling for total cerebral volume.
At study entry, patients and controls were similar in age, gender, handedness and height; patients' weights were significantly greater than controls. At initial scan, patients had significantly smaller total cerebral volume and larger lateral ventricle volume than control subjects. Volumes of the hippocampus and amygdala did not differ between groups. In the patients with COS, but not in the controls, there was a significant decrease in total cerebral and hippocampal volume and an increase in lateral ventricle volume during follow-up. Amygdala volume did not change with time. In the control group, none of the structures assessed changed significantly with time. There were no group differences in laterality for any structure. The volume-by-time curves for cerebral, hippocampus and lateral ventricle volumes were significantly nonlinear, tending to asymptote during adolescence. Change in negative symptom severity correlated positively with change in ventricle volume, and positive symptom change correlated inversely with change in hippocampal volume. In other words, patients with greater decline in hippocampal volume experienced less improvement in positive symptoms over time, whereas greater increases in lateral ventricle volume were associated with greater improvement in negative symptoms during the course of the study. This finding may stem from the curvilinear relationship between change in ventricle volume and time or other statistical anomaly.
The authors pointed out that the rate of change of cerebral, lateral ventricle and hippocampus volume could not continue in a linear fashion without producing gross neurologic abnormalities. The fact that the volume-by-time curves appear to level off during late adolescence point to this developmental period as potentially important in the pathophysiology of schizophrenia. During this time, synaptic pruning is intense and important metabolic changes occur, highlighting the importance of this accelerated developmental epoch. These investigators also pointed out that their finding of progressive neuroanatomic changes during the course of illness is not inconsistent with neurodevelopmental theories of schizophrenia.
In contrast to the relatively young, early-onset patients (Giedd et al., 1999), Mathalon and colleagues (2001) examined brain changes occurring over time in patients with schizophrenia. The 24 patients with schizophrenia and the 25 controls were matched on handedness, age and inter-scan interval. Baseline MRI scan data were compared with data obtained from follow-up scans conducted a mean of 3.6 years after initial scans in patients and after a mean of 4.2 years in controls. Patients were drawn from inpatient units in a U.S. Department of Veterans Affairs hospital associated with Stanford University. Control individuals were community volunteers involved in clinical and normal-aging neuroimaging studies, during which they were rescanned at various intervals over time. Both groups were systematically screened utilizing structured diagnostic interviews.
Patients were initially treated in the VA system, and many received follow-up care there as well. However, patients were not followed systematically in terms of documenting symptom severity or treatment variables as part of the study. Patients' symptom severity (assessed via the BPRS) and other clinical data (e.g., number of hospitalizations) were obtained by the investigators near the time of each scan. The mean of the baseline and follow-up BPRS scores was used as an index of illness severity during the inter-scan interval.
Magnetic resonance imaging scans were obtained at both time points utilizing identical scan acquisition protocols. The images were used to quantify region of interest (ROI) volumes representing intracranial contents; lateral ventricles; pre-frontal, frontal, temporal, parietal and occipital lobes; and the superior temporal gyrus. Change in ROI volume over time was quantified by determining the rate of change (slope) of each ROI volume over time. These slopes were compared between patients and controls to assess differences in rates of structural change with time, controlling for group differences in premorbid IQ and age.
The authors found that left lateral ventricle and right frontal sulcal (CSF) volume increased and gray matter volume decreased faster in patients with schizophrenia than in control subjects. Additionally, in patients with schizophrenia there were more rapid bilateral increases in the volume of prefrontal and posterior superior temporal sulci (CSF) and more decreases in posterior superior temporal gray matter volume than there were in controls. In 14 of 18 ROIs examined, patients with schizophrenia demonstrated faster progression than controls, although not all comparisons were statistically significant. Furthermore, these results remained the same whether or not the analysis statistically controlled for baseline ROI volume. Comparing symptom severity at baseline and follow-up scan times, patients were significantly less ill at follow-up. Higher BPRS total scores were significantly correlated with greater frontal sulcal expansion and anterior temporal gray matter decline. Higher positive symptom scores correlated with more rapid frontal sulcal expansion and faster decline in left frontal gray matter volume. Negative symptom severity correlated with faster decline of prefrontal gray matter volume and expansion of right prefrontal sulci.
The authors also noted that their results are similar to previously published works on several fronts. Their findings support the theory of progressive degenerative neuroanatomic changes, involving several brain regions, in patients with schizophrenia during the course of the illness. Furthermore, because greater brain changes over time were associated with greater symptom severity and poorer clinical course, neuroanatomic changes may reflect the neurotoxic effects of psychotic illness. The pathophysiologic mechanism(s) underlying psychosis-related neurotoxicity include abnormal regulation of excitatory amino acid and/or neuroprotective systems. The authors further noted that their findings suggested that neurodevelopmental theories do not account for all neuroanatomic abnormalities seen in the brains of patients suffering from schizophrenia. Due to the fact that they had only two data points, they could not fully describe the pattern and rate of change of neuroanatomic structures in schizophrenia over time. Other limitations included using only male subjects, geometrically defined (as opposed to anatomically defined) ROIs, and lack of detailed follow-up data related to symptoms, resource use and medication compliance.
The two previously described studies looked at brain changes over time involving volume (size) of various brain structures. Gharaibeh and colleagues (2000) studied the change in brain shape independent of the size of the structures examined. This investigation utilized shape analysis techniques, developed by Fred Bookstein, Ph.D., at University of Michigan, that involved averaging the shape of a set of neuroanatomic landmarks on MRI scans of both patients and controls. The analysis mathematically eliminates the effects of size and rotation on the landmark set and statistically tests differences in landmark shape between groups under study. It can also be used to produce averaged MRI images that pictorially display group differences.
In the study, 55 patients with a first episode of non-affective psychosis (the majority were diagnosed with schizophrenia when reassessed over time) and 22 controls (matched with patients on age, gender and socioeconomic status) were examined. All subjects were carefully screened for medical, neurological and psychiatric exclusion criteria. Magnetic resonance imaging scans were obtained at baseline and after a period of time ranging from 33 to 64 months using identical image acquisition protocols at both time points. Only the mid-sagittal image was used in the analysis. A set of landmarks clearly and reliably identified on this image focused the analysis primarily on sub-cortical structures, including cerebellum, corpus callosum and pons. The statistical analyses controlled for age, gender, diagnosis and number of comparisons in assessing landmark shape changes over time.
These investigators found no statistically significant landmark shape differences between patients and controls at baseline or follow-up, although the between-group shape differences were very similar to those seen in previous studies. At baseline and follow-up there were significant gender differences in shape across groups; these shape differences tended to be less pronounced in the schizophrenia sample. However, the splenium was thicker and farther from the cerebellum in women than in men, regardless of diagnostic group. The change in shape over time was significantly different between patients and controls. For example, the analysis found a relative increase in the distance between the superior colliculus and the splenium of the corpus collosum in patients. Controls' landmark shape changed little over time. The authors suggested that their findings are consistent with differential shape changes over time between patients with schizophrenia and controls. These shape changes seem to be tied to onset of symptoms and may be neurodegenerative in nature. The mechanism underlying the differential shape changes may be related to abnormalities in neuroplasticity in schizophrenia that may be impacted in a positive way by treatment with antipsychotic medication. They also pointed out, similar to Giedd et al. (1999) and Mathalon et al. (2001), that their results are not inconsistent with a neurodevelopmental hypothesis of schizophrenia.
The results of these three studies that examined very different patient populations and/or utilized different quantification and analysis techniques produced a similar conclusion: There are progressive neuroanatomic changes in the brains of patients with schizophrenia. This limited review of recent research is not meant to suggest that all longitudinal neuroimaging studies have found evidence for progressive brain changes, because there are negative studies in the literature. Research findings suggesting progression of neuroanatomic abnormalities in schizophrenia have important implications for theories of the pathophysiology of psychosis and do not preclude neurodevelopmental theories of the illness. Indeed, the neuropathophysiologic processes that are at work prior to the onset of illness may continue after the onset of frank psychotic symptoms, perhaps muted by consistent treatment with antipsychotic medication.
The studies by Giedd et al. (1999) and Mathalon et al. (2001) both found a relationship between symptom severity and clinical course and change in brain structure over time in patients with schizophrenia. These results, coupled with research findings suggesting the possibility of normalization of neuroanatomic abnormalities in schizophrenia over time, suggest that early and consistent treatment with atypical antipsychotic medication may not only improve symptom severity and overall functioning, but it may also limit neurodegeneration and chronicity. Future prospective, longitudinal neuroimaging studies, carefully documenting symptom severity, medication treatment and resource utilization, will be required to clarify these relationships.