Cognitive impairment is a well-recognized element of schizophrenia, not simply attributable to psychotic symptoms R1215511BABDBDADR1215511BABDHCIF. Subtle impairment can sometimes be detected before the onset of frank psychosis, it is a prominent feature of schizophrenia after the onset of psychosis, and it increases in severity and prevalence with aging R1215511BABEAHCFR1215511BABCCGAA.
Much of the cognitive impairment associated with age in the nonpsychotic population is attributable to Alzheimer’s disease. A number of studies have tested for a high prevalence of neuropathologically verified Alzheimer’s disease in schizophrenia, with inconclusive but mostly negative results R1215511BABDHIEIR1215511BABBGJHB. However, just as individuals with greater educational and occupational attainments may possess a "cognitive reserve" that mitigates the cognitive effects of senile degeneration R1215511BABCIJDJR1215511BABBEDIH, the converse may also hold. In schizophrenia, where intellectual functions may already be impaired, low levels of neuritic senile plaques and neurofibrillary tangles, consistent with "normal aging" in a healthy individual, may be sufficient to cause recognizable cognitive decline.
In the present study we sought to evaluate this possibility by addressing three questions: 1) Is the frequency of neuropathologically defined Alzheimer’s disease higher than normal in schizophrenia? 2) Are levels of neuritic senile plaques and neurofibrillary tangles below the threshold for Alzheimer’s disease related to cognitive impairment in schizophrenia? 3) If so, does the relationship differ from that in other populations?
Psychiatric cases were collected from autopsies performed at seven New York State psychiatric hospitals between 1982 and 1993. An age below 45 was the only exclusion criterion. Patients were divided into four DSM-III-R diagnostic categories, on the basis of a standardized review of medical records (to be described): "schizophrenia" included all patients with a primary diagnosis of schizophrenia or schizoaffective disorder, "mood disorder" included all individuals with a primary diagnosis of major depression or bipolar disorder, "dementia" comprised all individuals with a primary diagnosis of dementia, and "other psychiatric" included all individuals with psychiatric diagnoses that did not fit into the preceding categories.
We also included a nonpsychiatric group culled from autopsies at a general hospital. Clinical records were reviewed in the same manner as for the psychiatric cases. Only those with no psychiatric diagnosis or with adjustment disorder as the only psychiatric diagnosis were included. Summaries of age and sex are given in T1. In all cases, consent for autopsy was obtained.
The brains were preserved in 10% phosphate-buffered formalin before and after routine neuropathological examination. The data in this study were based on a second, standardized neuropathological examination (see the following), typically performed several years after the autopsy, on new paraffin blocks prepared in a vacuum-infiltration tissue processor (Instrumentation Laboratory, Lexington, Mass.). The intervals between death and autopsy and between autopsy and the standardized neuropathological examination are included in T1.
Hospital records were reviewed independently by at least two members of a team of psychiatrists and psychologists, who were blind to autopsy results. Consensus diagnoses were determined after discussion of these reviews by this entire team. All raters were trained in the chart review protocol and the modified Diagnostic Evaluation After Death and were tested for interrater reliability R1215511BABCDHBJ. The diagnostic procedures followed DSM-III-R guidelines, except that cognitive loss in the context of schizophrenia was not diagnosed as a separate dementia.
Review with the modified Diagnostic Evaluation After Death also evaluated cognitive impairment. Cases were rated as showing "definite cognitive impairment" on the basis of straightforward evidence of a decline in cognitive function, without recovery, in the form of 1) psychometric test results, 2) clinical mental status examinations, or 3) a general level of functioning consistent with any dementia diagnosis. In most cases, ratings of definite cognitive impairment were based on physicians’ notes of significantly impaired memory or orientation and on staff notes of declines in several areas of self-care. While meeting the DSM criteria for dementia was not a formal requirement, most subjects classified as having definite cognitive impairment would meet them. We required a period of at least 1 year of impaired functioning without recovery. Cognitive impairment was not diagnosed in the context of general physical decline within the last year of life. A final rating of definite cognitive impairment indicated a consensus that these criteria were definitely satisfied.
+
Evaluation of Neuritic Senile Plaques and Neurofibrillary Tangles
Neuritic senile plaques and neurofibrillary tangles were evaluated in the following regions: superior and middle frontal gyri, at the level of the crossing of the anterior commissure; hippocampal formation and adjacent neocortex at the levels of the amygdala, pes hippocampi, and lateral geniculate body; superior and middle temporal gyri at the level of the lateral geniculate body; inferior parietal lobule at the level of the posterior end of the pulvinar; and calcarine and adjacent cortex approximately 1 cm rostral to the caudal tip of the occipital lobe. The single field of highest density was counted on each slide R1215511BABCFGGJR1215511BABCDICJ except the hippocampal slides, where the allocortex and neocortex were treated separately. Only neuritic plaques were counted R1215511BABCDICJ since amyloid plaques without neuritic elements may be very numerous in the absence of dementia R1215511BABDFEACR1215511BABBAHCH. Counts were made with a 25× objective, which afforded a field of 0.38 mm2. (We made appropriate adjustments when applying the Khachaturian criteria R1215511BABCFGGJ, defined for 1 mm2.) Neuritic senile plaques were identified by immunohistochemistry with Alz 50, which stains the neuritic elements in the plaques but not β-amyloid R1215511BABBHCEB. Neurofibrillary tangles were identified by immunohistochemistry with Ab 39, which recognizes paired helical filaments R1215511BABDHBCF. Details of the staining techniques and confirmation of their applicability to archival, formalin-fixed material are described elsewhere R1215511BABCACEE. The right and left hemispheres were randomly alternated for sampling.
All microscopic examinations were performed by investigators who were blind to any clinical information and to the source of the specimen. Neuritic senile plaques and neurofibrillary tangles were counted manually by a single observer (D.L.). For evaluations other than the fulfillment of the Khachaturian criteria for Alzheimer’s disease R1215511BABCFGGJ, counts of neuritic senile plaques and neurofibrillary tangles were averaged separately across the allocortical and neocortical regions for each case.
First, we determined the prevalence of Alzheimer’s disease in each diagnostic group by applying the Khachaturian age-dependent neuropathological criteria R1215511BABCFGGJ.
Second, we compared the numbers of neuritic senile plaques and neurofibrillary tangles in the schizophrenia subjects with and without definite cognitive impairment, by two-tailed t tests.
Third, we used analysis of covariance (ANCOVA) to compare the age-adjusted levels of neuritic senile plaques and neurofibrillary tangles in the schizophrenia subjects without definite cognitive impairment and the nonpsychiatric subjects (all of whom lacked definite cognitive impairment, by definition). This tested for lower tolerance to neuritic senile plaques or neurofibrillary tangles in schizophrenia.
Fourth, we compared the schizophrenia and mood disorder subjects, using chi-square analysis to compare rates of cognitive impairment. We used logistic regression to test the effect of diagnosis on cognitive impairment while controlling for age, neuritic senile plaques, and neurofibrillary tangles.
Definite cognitive impairment was present in 68% of the schizophrenia subjects (N=45), 19% of the mood disorder subjects (N=5), and 59% of the "other psychiatric" subjects (N=10). By definition, definite cognitive impairment was absent in the nonpsychiatric subjects and was present in all of the dementia subjects. As expected, neuritic senile plaques and neurofibrillary tangles were significantly more numerous in the dementia group than in any of the other diagnostic groups (F1).
The various measures of senile degeneration were highly correlated with one another (T2).
+
Frequency of Neuropathological Alzheimer’s Disease in Schizophrenia
Five (8%) of the schizophrenia subjects (aged 73, 80, 80, 82, and 89), all with definite cognitive impairment, and 17 (47%) of the dementia subjects met the Khachaturian neuropathological criteria for Alzheimer’s disease R1215511BABCFGGJ. The counts of neuritic senile plaques and of neurofibrillary tangles were similar for the subjects with Alzheimer’s disease in the two diagnostic groups. Alzheimer’s disease can thus explain cognitive impairment in only 11% of the 45 schizophrenia subjects with definite cognitive impairment.
+
Comparison of Schizophrenia Subjects With and Without Definite Cognitive Impairment
The schizophrenia subjects with and without definite cognitive impairment were similar in terms of duration of illness (without impairment: mean=46.2 years, SD=12.0, range=27–72; with impairment: mean=54.8, SD=8.9, range=25–72) and history of somatic treatments (T3). The schizophrenia subjects with cognitive impairment were significantly older (mean age=80.6 years, SD=10.3) than those without cognitive impairment (mean=73.3, SD=9.8) (t=2.72, df=64, p=0.008); all but one of the 17 schizophrenia subjects aged 85 or older were cognitively impaired. With the subjects aged 85 or older eliminated, the mean ages of the schizophrenia groups with cognitive impairment (mean=75.2, SD=8.3) and without cognitive impairment (mean=72.6, SD=9.4) were similar (t=1.01, df=47, p=0.32), while all measures of neuritic senile plaques and neurofibrillary tangles were greater in the group with definite cognitive impairment. The differences were statistically significant for all measures except neocortical neurofibrillary tangles. When the schizophrenia subjects with neuropathological Alzheimer’s disease were excluded, the difference in neuritic senile plaques remained significant (F2). With schizophrenia subjects of all ages included, the differences were larger.
The most specific neuropathological predictor of definite cognitive impairment in schizophrenia was neocortical neuritic senile plaques. All but three of the schizophrenia subjects with any neocortical neuritic senile plaques had definite cognitive impairment, and each of these three had only one neuritic senile plaque per field in only one section. However, neocortical neuritic senile plaques were not a sensitive predictor: 13 of the 30 schizophrenia subjects below age 85 with no neocortical neuritic senile plaques had definite cognitive impairment. Among these 30 cases without neocortical neuritic senile plaques, those with and without definite cognitive impairment did not differ significantly on any of the other measures of neuritic senile plaques or neurofibrillary tangles.
There were 14 schizophrenia subjects, all with definite cognitive impairment, who did not have neuropathological Alzheimer’s disease but who had more than one neocortical neuritic senile plaque per field in at least one section or had one neocortical neuritic senile plaque per field in more than one section. One had frontal and parietal neuritic senile plaques, seven had only temporal neocortical neuritic senile plaques, and six had temporal neocortical neuritic senile plaques plus neuritic senile plaques in the frontal (N=3), parietal (N=3), and/or occipital (N=1) neocortex.
+
Comparison of Schizophrenia Subjects Without Definite Cognitive Impairment and Nonpsychiatric Subjects
Mean age-adjusted counts of neuritic senile plaques and neurofibrillary tangles were lower for the schizophrenia subjects without definite cognitive impairment than for the nonpsychiatric group, by factors ranging from about one-half (0.551 for allocortical neurofibrillary tangles) down to about one-thirtieth (0.035 for neocortical neuritic senile plaques). All differences were statistically significant except for allocortical neurofibrillary tangles (F3).
+
Comparison of Schizophrenia and Mood Disorder Subjects
Schizophrenia and mood disorder subjects below age 85 had similar mean ages (mean=74.1, SD=8.7, and mean=70.5, SD=7.5, respectively) (t=1.66, df=68, p=0.10) and similar levels of allocortical and neocortical neuritic senile plaques and neurofibrillary tangles (not shown; similar to levels for all ages, F1). Both groups had chronic illnesses (schizophrenia: mean duration=48.9 years, SD=9.6, range=25–63; mood disorder: mean=27.7 years, SD=6.0, range=5–55), and they received various somatic treatments with similar frequencies (T3). However, definite cognitive impairment was over six times as prevalent in the schizophrenia group (29 of 49) as in the mood disorder group (two of 21) (χ2=14.7, df=1, p=0.0001). When we adjusted for age, neurofibrillary tangles, and neuritic senile plaques, the odds ratio for definite cognitive impairment in schizophrenia versus mood disorders was 12.7 (95% confidence interval=2.1–76.7). Thus, cognitive function appears to be more vulnerable to the effects of neuritic senile plaques and neurofibrillary tangles in schizophrenia than in mood disorders.
Among the schizophrenia subjects under age 85, those with cognitive impairment had fewer years of schooling (7.8 years versus 10.2 years for those without cognitive impairment) (t=2.47, df=47, p=0.02). The mood disorder subjects under age 85 had a slightly higher mean educational level than the schizophrenia subjects under age 85 (10.2 versus 8.8 years) (t=1.67, df=67, p=0.10). When education was included in the logistic regression, the odds ratio for cognitive impairment in schizophrenia versus mood disorders was reduced to 10.7 (95% confidence interval=1.6–70.2).
+
Other Neuropathological Explanations for Dementia
Only six schizophrenia subjects (aged 76–81) had other neuropathological conditions that sometimes cause cognitive impairment: four had Parkinson’s disease, one had hydrocephalus of unknown etiology, and one had a small (<1 cm diameter) focus of idiopathic fibrosis in the left frontal lobe. Exclusion of these cases produced little change in the number of neuritic senile plaques or neurofibrillary tangles in the schizophrenia subjects with or without cognitive impairment.
Among the 36 dementia subjects, there were 17 cases of Alzheimer’s disease, seven cases of cerebral infarct, two each of frontal lobe dementia, infection, and metabolic disorders, one case of Huntington’s disease, and one case of neuroaxonal dystrophy. The remaining four cases (11%) showed no significant pathology except for Alzheimer-type changes of insufficient severity to meet the age-dependent criteria for a histological diagnosis of Alzheimer’s disease R1215511BABCFGGJ.
Our study yielded four main findings.
While definite cognitive impairment was present in 68% of a group of elderly, chronically institutionalized individuals with schizophrenia, neuropathological Alzheimer’s disease was present in only 8%.
Schizophrenia subjects with definite cognitive impairment had more neuritic senile plaques (even after exclusion of cases with neuropathological Alzheimer’s disease) than did schizophrenia subjects without cognitive impairment.
Schizophrenia subjects without definite cognitive impairment had fewer neuritic senile plaques and neurofibrillary tangles than did cognitively intact nonpsychiatric patients.
Cognitive impairment was much more common among schizophrenia subjects than among chronically institutionalized mood disorder patients, while the two groups were comparable in terms of age, Alzheimer-type changes, and somatic treatments.
A parsimonious interpretation of these findings is that schizophrenia creates a vulnerable state in which even mild senile degeneration, particularly in the form of neocortical neuritic senile plaques, is sufficient but not necessary to cause definite cognitive impairment. In contrast, mild levels of senile degenerative change appear to be better tolerated in subjects without preexisting psychiatric disease. For example, four of 16 normal subjects had at least four neuritic senile plaques per field in at least one neocortical region. However, while the nonpsychiatric subjects clearly lacked the behavioral manifestations of cognitive impairment that were present in the subjects with schizophrenia, neuropsychological testing or detailed clinical evaluation may elicit subtle cognitive differences between elderly subjects with and without mild Alzheimer-type changes R1215511BABBAHCHR1215511BABCIDBAR1215511BABBGFIA. Individuals who can tolerate such cognitive loss without behavioral manifestations must posses either compensatory abilities or a surfeit of cognitive ability. We suggest that these are lacking in chronically institutionalized subjects with schizophrenia.
Our findings are similar to those of El-Mallakh et al. R1215511BABDIAJD, who found more frontal and hippocampal senile plaques in 10 intellectually impaired schizophrenia subjects (mean age=84 years) than in seven intellectually intact schizophrenia subjects (mean age=62). In striking similarity to our study, the mean plaque counts for the intellectually intact schizophrenia subjects were extremely low, approximately one-tenth of the mean value for seven normal subjects (mean age=73).
Two recent studies of elderly, chronic inpatients had findings somewhat different from ours. One study R1215511BABEACGG showed levels similar to ours for neocortical neuritic senile plaques in nonpsychiatric subjects and cognitively impaired schizophrenia subjects but a higher value for cognitively unimpaired schizophrenia subjects, close to the value for the nonpsychiatric subjects. All ages were similar, and we cannot attribute this difference to staining or counting procedures, since the values were similar for the first two groups, nor to overdiagnosis of definite cognitive impairment on our part, since Purohit et al. R1215511BABEACGG classified 87 of 100 schizophrenia subjects as cognitively impaired, while we classified 68% as having definite cognitive impairment. Purohit et al. counted neuritic senile plaques in the orbitofrontal cortex, while we did not. If their higher value was attributable to neuritic senile plaques in this region, it is possible that these are better tolerated than neuritic senile plaques in the temporal neocortex. In another study R1215511BABCJHHH there was no correlation between any neuropathological measure and any cognitive measure in 23 elderly schizophrenia subjects. Since that study did not include counts of neuritic plaques, it is difficult to compare with ours.
Enhanced sensitivity of cognition to neuritic senile plaques was not apparent in our group of chronically institutionalized patients with mood disorder, although this group was similar to the schizophrenia group in terms of age, presence of psychosis, institutionalization, and somatic treatments. The subjects in the two groups died during the same period and in the same institutions; hence, one should expect similar reporting of their symptoms in the medical records. Thus, from the available data, we cannot attribute the high rate of cognitive impairment in schizophrenia to institutionalization, somatic treatments, or reporting biases of clinicians.
There are several potential limitations to this study.
The tissues had been in formalin for several years. However, we verified previously that immunoreactivity with the antibodies we used is preserved R1215511BABCACEE, and at least one other study of immunohistochemistry for senile degenerative changes has used such tissue R1215511BABBHIBC. Furthermore, our rates of neuropathological Alzheimer’s disease among inpatients with schizophrenia or primary dementia are comparable to those in other studies R1215511BABEAHGDR1215511BABEACGG.
In order to apply standardized, age-related neuropathological criteria for Alzheimer’s disease R1215511BABCFGGJ, we counted neuritic senile plaques and neurofibrillary tangles only in the field of highest density within each region. Thus, the counts are not necessarily representative of average densities over broad regions. However, we were very consistent in the selection of regions for examination, and the counts represent the maximum severity of pathology in each region. Hence, comparability across cases was maintained. This strategy allowed us to make the critical distinction between cases with low levels of neocortical neuritic senile plaques and those with none.
Since "definite cognitive impairment" represents rater certainty of cognitive impairment, inadequate documentation could theoretically result in a false negative rating. This seems unlikely, however, since the records were extensive and detailed, and our rate of definite cognitive impairment (73% in schizophrenia patients aged 65 and over) is comparable to the rate of dementia (83%) in a series of live schizophrenia patients from one of the contributing institutions R1215511BABBBCID. False negative classifications would anyway work against our findings by inflating counts of neuritic senile plaques and neurofibrillary tangles in the group without definite cognitive impairment.
All of the schizophrenia subjects died in state institutions after 1981. Therefore, they may have had unusually severe forms of the illness.
+
"Cognitive Reserve" in Schizophrenia
Our findings suggest that schizophrenia and senile degeneration are synergistic in producing cognitive decline. This interpretation is consistent with theories of "cognitive reserve" protecting against dementia, first proposed by Katzman et al. R1215511BABDGIHD. These investigators found that the brains of nondemented elderly individuals with large numbers of neuritic senile plaques but no dementia were heavier and contained more neurons than those of elderly individuals with dementia. Evidence that higher educational and occupational levels protect against the dementing effects of neuritic senile plaques and neurofibrillary tangles has been interpreted as further support for the cognitive reserve hypothesis R1215511BABCIEDBR1215511BABBEDIH. The observation that neocortical synaptic density provided the best neuroanatomic correlate for cognitive function in Alzheimer’s disease R1215511BABCAIHBR1215511BABCGIAA has led to the hypothesis that the protective effect of education resides in increased synaptic density R1215511BABCIEDB. Under this model, we postulate that senile impairment of cognition in schizophrenia and nonschizophrenia subjects alike results from a continuous, age-related process that is correlated with the appearance of neuritic senile plaques. The threshold at which this process produces clinical evidence of cognitive change is hypothetically lower in individuals with schizophrenia than in others.
The underlying vulnerability in schizophrenia may be developmental; several investigators have reported low cognitive capacity in young subjects R1215511BABEAHCFR1215511BABDJIBDR1215511BABCBJEH. Vulnerability could also be associated with some progressive process, yet unidentified, that is specifically related to schizophrenia. Theoretically, such a process could produce cognitive impairment by itself when sufficiently advanced, or at an earlier stage if neocortical neuritic senile plaques are also present.
Processes contributing to diminished cognitive reserve in schizophrenia have yet to be identified. In contrast to Alzheimer’s disease, where loss of synaptic density R1215511BABCAIHBR1215511BABCGIAA and cholinergic markers R1215511BABDICGG are prominent features and correlated with cognitive impairment, diminished synaptic density has not been reported in schizophrenia, and cholinergic markers are normal R1215511BABDIAJDR1215511BABDICGG. However, reported abnormalities in cortical synaptic proteins R1215511BABDJIBBR1215511BABCHEDF, microtubule-associated proteins R1215511BABDJCHDR1215511BABCFCEE, neuronal size R1215511BABCGJIFR1215511BABEADGB, and neuropil volume R1215511BABEBBAA suggest the possibility of impaired synaptic function or remodeling.
Dopaminergic deficits are associated with dementia in a variety of conditions R1215511BABEBDEDR1215511BABCDHJA. Although the positive symptoms of schizophrenia have generally been related to subcortical dopaminergic hyperactivity, many authors have suggested that in schizophrenia, some aspects of dopaminergic activity are suppressed R1215511BABBDHJHR1215511BABEAFCD, which could contribute to cognitive impairment R1215511BABEAHCF. Low R1215511BABDDEFG and normal R1215511BABBIHHD prefrontal immunoreactivity for tyrosine hydroxylase have both been reported. If dopaminergic reserves are even focally diminished in schizophrenia, further decline with age (normally beginning in adolescence R1215511BABDGGAC and progressing through senescence R1215511BABDDJEI) could impair cognition or increase vulnerability to the effects of other aging processes.
Neuroleptic drugs could, of course, play a part. They are antagonists at a variety of receptors for dopamine, acetylcholine, serotonin, norepinephrine, and epinephrine. The authors of a report of low tyrosine hydroxylase activity in the deep layers of the anterior cingulate cortex noted the absence of this abnormality in two neuroleptic-free patients R1215511BABBIHHD, raising the possibility that neuroleptic drugs influence the distribution of cortical dopamine terminals. However, the relatively low incidence of cognitive impairment in the neuroleptic-treated mood disorder patients in our study indicates that neuroleptics alone do not account for the greater sensitivity to neuritic senile plaques and neurofibrillary tangles that we find in schizophrenia.
Presented in part at the 51st annual meeting of the Society of Biological Psychiatry, New York, May 1–5, 1996; the 72nd annual meeting of the American Association of Neuropathologists, Vancouver, June 11–16, 1996; and the 26th annual meeting of the Society for Neuroscience, Washington, D.C., Nov. 16–21, 1996Received Sept. 9, 1997; ; revision received April 24, 1998; accepted May 7, 1998. From the Divisions of Neuropathology and Brain Imaging, Department of Neuroscience, and the Departments of Brain Imaging and Clinical Psychobiology, New York State Psychiatric Institute; and the Departments of Pathology, Psychiatry, Neurology, and Radiology, Columbia University, New York. Address reprint requests to Dr. Dwork, Division of Neuropathology, Unit 62, New York State Psychiatric Institute, 722 West 168th St., New York, NY 10032; ajd6@columbia.edu (e-mail). Supported by grants AG-10638 and AG-08702 from the National Institute on Aging and by grant MH-50727 from NIMH. Abbott Laboratories provided Alz 50 antibody.The authors thank Dr. Shu-Hui Yen for antibody to paired helical filaments.