Brain Structure in Neuropsychologically Defined Subgroups of Schizophrenia and Psychotic Bipolar Disorder

Abstract

Background:

Neuropsychological impairment is heterogeneous in psychosis. The association of intracranial volume (ICV) and total brain volume (TBV) with cognition suggests brain structure abnormalities in psychosis will covary with the severity of cognitive impairment. We tested the following hypotheses: (1) brain structure abnormalities will be more extensive in neuropsychologically impaired psychosis patients; (2) psychosis patients with premorbid cognitive limitations will show evidence of hypoplasia (ie, smaller ICV); and (3) psychosis patients with evidence of cognitive decline will demonstrate atrophy (ie, smaller TBV, but normal ICV).

Methods:

One hundred thirty-one individuals with psychosis and 97 healthy subjects underwent structural magnetic resonance imaging and neuropsychological testing. Patients were divided into neuropsychologically normal and impaired groups. Impaired patients were further subdivided into deteriorated and compromised groups if estimated premorbid intellect was average or below average, respectively. ICV and TBV were compared across groups. Localized brain volumes were qualitatively examined using voxel-based morphometry.

Results:

Compared to healthy subjects, neuropsychologically impaired patients exhibited smaller TBV, reduced grey matter volume in frontal, temporal, and subcortical brain regions, and widespread white matter volume loss. Neuropsychologically compromised patients had smaller ICV relative to healthy subjects, and neuropsychologically normal and deteriorated patient groups, but relatively normal TBV. Deteriorated patients exhibited smaller TBV compared to healthy subjects, but relatively normal ICV. Unexpectedly, TBV, adjusted for ICV, was reduced in neuropsychologically normal patients.

Conclusions:

Patients with long-standing cognitive limitations exhibit evidence of early cerebral hypoplasia, whereas neuropsychologically normal and deteriorated patients show evidence of brain tissue loss consistent with progression or later cerebral dysmaturation.

Introduction

Neuropsychological impairment is common in psychotic disorders. At the group level, cognitive dysfunction is characterized by a modest deficit in overall cognitive ability punctuated by more severe deficits in processing speed, memory, and executive functioning.^1–5 However, the severity of impairment varies markedly across individuals. Approximately 20%–30% of schizophrenia spectrum patients and 40%–60% of psychotic bipolar patients are neuropsychologically normal, at least when compared to standard normative data, while the remaining patients typically perform 1 to 2 standard deviations below normal.^6–10

Variability in cognitive impairment implies heterogeneity of neuropathology. The well-established association between general cognitive ability and brain volume suggests cognitively impaired patients will have smaller overall brain volumes and more widespread abnormalities in regional tissue volumes.^11,12 There is evidence this is the case. Cognitively impaired patients, and psychosis patients as a whole, demonstrate reduced volume in the frontal and temporal lobes, hippocampus, and subcortical structures such as the thalamus.^13–15 In contrast, brain structure alterations, including regional grey matter volume loss, cortical thinning, and sulcal widening, are relatively limited in neuropsychologically normal patients.^14–16 However, there are some inconsistencies. For example, Wexler et al¹⁷ found similar patterns of cortical grey matter volume loss in both neuropsychologically normal and impaired patients; although neuropsychologically impaired patients also had smaller white matter volumes, reduced subcortical volumes, and ventricular enlargement.¹⁷ The inconsistent findings may be due to the small samples sizes after stratification, especially for the neuropsychologically normal subgroup, and limited neuropsychological testing on which cognitive subgroups were defined.

Evidence that the pathway leading to cognitive impairment varies across patients further suggests there will be differences among cognitively impaired patients. Some cognitively impaired patients have below average premorbid abilities consistent with long-standing limitations in cognition while others have average premorbid intellectual ability suggesting decline from a previously higher level of functioning.^18–26 These groups are often referred to as “compromised” and “deteriorated,” respectively.^18,21 As such, predictions about brain structure abnormalities in cognitively impaired patients must be made within the context of normal brain development. During normal development, total brain volume (TBV) and intracranial volume (ICV) rapidly increase in parallel during early childhood then diverge in late childhood/early adolescence; TBV gradually declines throughout adulthood while ICV remains static.^27–30 Consequently, ICV is often considered a proxy of early or premorbid brain development and the discrepancy between TBV and ICV an indicator of normal age-related atrophy and pathological processes that begin later in life.³¹ For example, prenatal exposure to toxins, such as alcohol, and congenital infections, including rubella and toxoplasmosis, are associated with cognitive impairment and reduced ICV, whereas neurodegenerative disorders are characterized by brain volume loss in the context of normal ICV and premorbid cognitive functioning.^32–35 From this perspective, compromised patients may exhibit a “neurodevelopmental” pattern of abnormalities consisting of smaller ICV, but relatively normal TBV after adjusting for reduced ICV. In contrast, deteriorated patients may demonstrate a “neuroprogressive” pattern of structural brain abnormalities characterized by reduced TBV in the context of normal ICV.

To address the limitations of previous studies and test the novel hypotheses described above, we investigated the relationship between brain structure and cognition in individuals with a psychotic disorder grouped according to their estimated premorbid and current cognitive abilities. First, we tested the hypothesis that neuropsychologically impaired patients will exhibit reduced TBV and qualitatively more extensive regional tissue volume abnormalities. Second, we tested the hypothesis that the compromised subgroup of neuropsychological impaired patients will demonstrate a “neurodevelopmental” profile characterized by smaller ICV and TBV, but relatively normal TBV after adjusting for smaller ICV. Finally, we hypothesized that deteriorated patients will demonstrate a “neuroprogressive” pattern consisting of reduced TBV, but normal ICV.

Methods

Participants

One hundred thirty-one patients with a psychotic illness comprised of schizophrenia/schizoaffective disorder (n = 101) and bipolar disorder with psychotic features (n = 30), and 97 healthy subjects were included in the current investigation. An additional 56 healthy subjects with neuropsychological test data only were also included to establish the neuropsychological test score cut-offs used to define the psychosis subgroups (described in detail in the following section). This study was approved by the Vanderbilt University Institutional Review Board and all study subjects provided written informed consent prior to contributing data to the repository. Subjects were included in the current investigation if they: (1) had a diagnosis of either schizophrenia/schizoaffective disorder or bipolar disorder with psychotic features, or were a healthy subject; (2) completed the brief neuropsychological battery described below; and (3) participated in an MRI (magnetic resonance imaging) scanning session that included acquisition of a T1-weighted structural scan. Study participants were excluded if they met any of the following exclusion criteria: age less than 16 or greater than 65, estimated premorbid intellect less than 70 and/or history of intellectual disability, presence of a central nervous system disorder (eg, multiple sclerosis, epilepsy) that would affect study participation, reported pregnancy or lactation, history of significant head trauma, psychotropic drug use (healthy subjects only), active substance abuse within the past 1 month, and MRI contraindications (eg, metal implants, claustrophobia). Clinical diagnoses were confirmed in all patients, and ruled out in healthy subjects, using the Structured Clinical Interview for DSM Disorders.³⁶ Additionally, the severity of psychotic symptoms was assessed in patients using the Positive and Negative Syndrome Scale (PANSS).³⁷

Neuropsychological Testing

Participants completed the Wechsler Test of Adult Reading,³⁸ a single-word reading measure of estimated premorbid intellectual functioning, and the Screen for Cognitive Impairment in Psychiatry (SCIP).³⁹ The SCIP consists of several subtests, including (1) a verbal list learning (VLL) test; (2) a working memory test (WMT) similar to the Auditory Consonant Trigrams test; (3) a phonemic verbal fluency test (VFT) of executive functioning; and (4) a processing speed test (PST). The VLL subtest consists of 3 learning trials of a 10-item word list, with total recall as the primary outcome measure, and a delayed recall trial administered after the WMT and VFT subtests. The WMT consists of 8 recall trials of 3 consonants, equally distributed among 4 conditions; no delay, and delays of 3, 9, or 18 s. During the delay period, subjects count backwards aloud to prevent rehearsal. The VFT consists of two 30 s trials during which subjects must generate as many words beginning with the letters C and L. The PST is a novel task developed from 6 letters of the Morse code. The corresponding Morse code dots and dashes are printed underneath each letter in a response key. Below the key is a 4×9 grid of boxes containing a letter in the top half of each box and subjects are given 30 s to translate as many letters to Morse code by filling in the bottom portion of the box with the correct dots and dashes. The SCIP has good to excellent test-retest reliability,^39,40 is sensitive to cognitive impairment in schizophrenia and psychotic bipolar disorder,^41,42 and correlates with other brief screening measures and composite scores derived from a comprehensive neuropsychological assessment.^43,44

The WTAR and SCIP were administered according to published guidelines and scored using published normative data. Briefly, WTAR raw scores were converted to standard scores which were then used to estimate Full Scale IQ using Appendix D from the WTAR manual. For the SCIP, raw scores for each subtest were converted to z scores using published norms and averaged to create a composite z score of overall cognitive functioning.³⁹

Neuropsychological Classification of Psychosis Patients

Psychosis patients were divided into neuropsychologically normal and impaired groups on the basis of both their estimated premorbid intellect and discrepancy between premorbid intellect and current cognitive functioning (ie, SCIP composite z score). Psychosis patients were classified as neuropsychologically normal if: (1) their estimated premorbid intellect was above the 10th percentile of the healthy subjects distribution (n = 153; 97 subjects with imaging data plus an additional 56 with only cognitive data); and (2) their current cognitive abilities (ie, SCIP composite z score) were consistent with expectations based on their estimated premorbid intellect. The latter was tested using discrepancy analysis, an approach commonly used in clinical neuropsychology for improving sensitivity to detect cognitive impairment.^45,46 For each psychosis patient we generated a predicted SCIP composite z score and compared the discrepancy between their predicted and actual scores to base rates in the healthy subjects group. The equation used to predict SCIP composite z scores was derived from the healthy subjects sample (n = 153) by regressing SCIP composite z scores on age, sex, and WTAR estimated premorbid IQ. The overall regression model was highly significant (F(3,152) = 28.50, P < .001), with age (t(152) = 6.77, P < .001), sex (t(152) = 2.84, P = .005), and WTAR estimated premorbid IQ (t(152) = 7.57, P < .001) each being a significant predictor of SCIP composite z score. Psychosis patients with a discrepancy above the 10th percentile of the distribution for healthy subjects were considered neuropsychologically normal. This corresponded to a 0.80 SD difference between their actual and predicted SCIP global z score. The remaining psychosis patients (ie, those with an estimated premorbid IQ less than 10th percentile or SCIP composite z score more than 0.80 SDs below their predicted level) were considered neuropsychologically impaired. Using a similar approach and same nomenclature as Weickert et al¹⁸ neuropsychologically impaired psychosis patients were further divided into “deteriorated” and “compromised” subgroups based on whether or not their estimated premorbid IQ was above or below the 10th percentile of the healthy subjects distribution. The 10th percentile was used as the cut-point for premorbid IQ and the discrepancy analysis because: (1) it corresponds closely to the conventional cutoff between “low average” and “borderline” IQ ranges⁴⁷; and (2) according to the authors of the WTAR, a premorbid IQ/current neuropsychological functioning discrepancy below the 10th percentile is considered a “moderate” indicator of dysfunction.³⁸

Neuroimaging

A high resolution (1mm isotropic voxel size) T1-weighted fast field echo anatomical scan (170 sagital slices, matrix = 256×256, TR = 8.0ms, TE = 3.7ms) was collected on each subject on a Philips Intera-Achieva 3T scanner located at the Vanderbilt Institute of Imaging Science. Structural scans were segmented into grey, white, and cerebrospinal fluid (CSF) tissue classes using the Voxel-Based Morphometry (VBM) toolbox, version 8 (http://dbm.neuro.uni-jena.de/vbm/) for Statistical Parametric Mapping, version 8 (http://www.fil.ion.ucl.ac.uk/spm). Total grey, white, and CSF volumes were calculated as was TBV (TBV = grey + white matter volume) and ICV (ICV = grey + white + CSF). Following segmentation, each tissue class was warped to a template image comprised of 550 subjects included with the VBM8 toolbox using the high dimensional DARTEL normalization method.⁴⁸ The tissue class images were modulated during DARTEL normalization to preserve the original volumes of the tissue classes. Since we were interested in examining changes on a voxel-wise level after accounting for individual differences in TBV, we elected to save the modulated images after correcting for nonlinear warping only. Thus, volume changes induced by linear warping to account for differences in TBV were removed from the modulation step. Images were smoothed with a 4mm kernel prior to statistical analysis.

Statistical Analysis

Total Brain Volumes.

ICV and TBV were examined using between groups ANOVA with age and sex entered as covariates. Given our hypothesis that: (1) TBV in compromised patients, while smaller in absolute terms, would be relatively normal in size after adjusting for ICV and (2) reduced TBV in deteriorated patients would persist after adjusting for ICV; we repeated the analysis of TBV including ICV as an additional covariate. Planned contrasts were carried out within the omnibus ANCOVAs to test the following hypotheses: (1) TBV will be reduced in neuropsychologically impaired psychosis patients compared to healthy subjects and neuropsychologically normal patients; (2) ICV and TBV will be smallest in compromised patients; however, after correcting for ICV, TBV will be relatively normal in this group; and (3) deteriorated patients will demonstrate reduced TBV, but normal ICV, compared to healthy subjects and neuropsychologically normal patients.

Voxel-Based Morphometry.

Because it’s possible to have normal total tissue volume, but regionally specific volume loss (ie, decreased volume of a specific structure), we followed up the primary analysis of total cerebral volumes with a VBM analysis. Regional differences in grey, white, and CSF volumes were examined on a voxel-wise basis by entering the nonlinear modulated tissue class images into an ANOVA with planned contrasts comparing each neuropsychologically defined patient group to healthy subjects. Consistent with the analysis of total tissue volumes, we first compared neuropsychologically normal and impaired psychosis patient groups to healthy subjects then compared the compromised and deteriorated groups to healthy subjects. We hypothesized that regional tissue volume loss would be qualitatively more extensive in neuropsychologically impaired patients. Age, sex, and ICV were included as covariates. All contrasts were thresholded at the whole-brain cluster-level corrected family-wise error rate = 0.05 for voxel-wise P = .005.

Results

Demographic, cognitive, and clinical data are presented in table 1. Aside from a modest difference in age, due to the fact that neuropsychologically normal patients were slightly younger than healthy subjects and neuropsychologically impaired patients, the groups were demographically similar. The neuropsychologically impaired group had a greater proportion of schizophrenia spectrum patients compared to the neuropsychologically normal group (81% vs 63%); however, this difference was not statistically significant (x ²(1) = 2.61, P = .105). One hundred twenty-one out of 131 patients were taking an antipsychotic medication (second generation: n = 111; first generation: n = 3; first and second generation = 7). The proportion of medicated patients did not differ between neuropsychologically normal and impaired groups (90.2% vs 93.3%; x ²(1) = 0.38, P = .537). Average daily antipsychotic dose, calculated in chlorpromazine equivalents according to Gardner et al⁴⁹ was higher in neuropsychologically impaired patients compared to the neuropsychologically normal group (502.3±285.3mg/d vs 319.4±143.9mg/d; t(117) = 3.64, P < .001). Clinical symptoms were also slightly more severe in neuropsychological impaired patients; although the differences were modest ranging from 2 to 4 points on the PANSS subscales. Demographic data, after subdividing the neuropsychologically impaired patient group into deteriorated and compromised subgroups is presented in supplementary table 1. Prior to examining structural brain differences in cognitive subgroups of patients, we first compared the entire cohort of psychosis patients to healthy subjects (supplementary figures 1 and 2).

Table 1.

Demographics

	Healthy Subjects		Patients With Psychosis
			NP Normal		NP Impaired
Sample Size	97		41		90
Variable	Mean	SD	Mean	SD	Mean	SD	F/t/x 2	P	Post Hoc a
Age	33.3	11.5	28.9	10.3	34.6	12.9	3.32	.038	NP I>NP N
Education	5.2	1.5	4.4	1.6	3.5	1.2	34.59	<.001	HS>NP N, NP I
Sex (M/F)	52/45		29/12		50/40		3.68	.159
Race (C/AA/O)	62/29/6		32/7/2		55/29/6		3.82	.431
Diagnosis (BP/SCZ)	—	—	13/28		17/73		2.62	.105
Premorbid IQ	106.7	8.8	105.9	7.0	96.2	10.9	32.02	<.001	HS, NP N>NP I
Current NP functioning	−0.06	0.84	−0.17	0.68	−2.08	0.92	149.56	<.001	HS, NP N>NP I
Age at illness onset	—	—	21.9	4.7	22.1	8.3	0.20	.840	—
Duration of illness	—	—	7.0	8.8	12.3	12.0	2.53	.012	—
PANSS positive	—	—	16.9	8.1	19.9	6.4	2.30	.023	—
PANSS negative	—	—	12.3	5.7	14.7	7.0	1.87	.064	—
PANSS general	—	—	28.3	7.3	32.4	7.8	4.10	.005	—
APD dose (CPZ equiv)	—	—	319.4	143.9	502.3	285.3	3.64	<.001	—

Note: AA = African American; APD = antipsychotic drug; BP = bipolar disorder; C = Caucasian; CPZ = chlorpromazine; F = female; HS = healthy subjects; NP = neuropsychologically; I = impaired; N = normal; M = male; O = other; PANSS=Positive and Negative Syndrome Scale; SCIP = Screen for Cognitive Impairment in Psychiatry; SCZ = schizophrenia. Education coded as follows: 1 = Grade 6 or less; 2 = Grade 7–12; 3 = high school; 4 = part college; 5 = graduated 2-year college; 6 = graduated 4-year college; 7 = part graduate/professional school; 8 = completed graduate/professional school.

^aBonferroni corrected for multiple comparisons.

Structural Brain Changes in Neuropsychologically Defined Psychosis Patient Subgroups

Total Brain Volumes.

Main effects of group were detected for ICV (F(3,222) = 4.15, P = .007), absolute TBV (F(3,222) = 4.47, P = .005), and ICV adjusted TBV (F(3,221) = 3.15, P = .026). As shown in the top panel of figure 1, the results were consistent with our first hypothesis; TBV was reduced in neuropsychologically impaired patients compared to both healthy subjects (P = .001) and neuropsychologically normal patients (P = .051). There was no difference between healthy subjects and neuropsychologically normal patients (P = .507). After adjusting for ICV, the reduction in TBV in neuropsychologically impaired patients remained significant compared to healthy subjects (P = .037), despite the fact that mean ICV within the impaired group was smaller compared to healthy subjects (P = .005) and neuropsychologically normal patients (P = .015). However, contrary to expectations, after adjusting for ICV, neuropsychologically normal patients also had smaller TBVs compared to healthy subjects (P = .007).

Fig. 1.

Neuropsychological functioning and brain volumes in individuals with a psychotic disorder. Top panel: Individuals with psychosis were classified into neuropsychologically (NP) normal and NP impaired groups based on their premorbid intellect and current NP functioning, and compared to healthy subjects on intracranial volume (ICV), total brain volume (TBV), and ICV adjusted TBV. ICV and TBV were reduced in NP impaired patients. After adjusting for ICV, mean TBV was lower in both NP impaired and NP normal patients. Bottom panel: Follow-up analysis after dividing the NP impaired group into deteriorated (ie, average premorbid IQ, impaired current NP functioning) and compromised (ie, below average premorbid IQ, impaired current NP functioning) groups revealed that reduced ICV in NP impaired patients was specific to the compromised subgroup; deteriorated patients did not differ from healthy subjects and NP normal patients. Absolute TBV was lowest in compromised patients; however, after adjusting for ICV, mean TBV in the compromised group did not differ from healthy subjects. In contrast, relative TBV was reduced in deteriorated patients compared to healthy subjects. Error bars = standard error of the mean.

Cognitive and brain volumes data after breaking the neuropsychologically impaired group into compromised and deteriorated subgroups are shown in the bottom panel of figure 1. Consistent with our hypothesis that compromised patients would demonstrate a “neurodevelopmental” profile, the compromised subgroup had smaller mean ICV compared to healthy subjects (P = .001), neuropsychologically normal psychosis patients (P = .003), and deteriorated psychosis patients at the trend significance level (P = .057). Critically, mean TBV, while smallest in absolute terms, did not differ from healthy subjects after adjusting for ICV (P = .209) indicating that TBV among neuropsychologically compromised patients was relatively normal in size after accounting for their smaller mean ICV. In contrast, deteriorated patients did not differ from healthy subjects in mean ICV (P = .204); however, mean TBV was reduced both before and after adjusting for ICV (P = .055 and P = .030), which is consistent with our hypothesis that this group would exhibit evidence of neuroprogression. We performed several additional analyses to confirm that the selective reduction in ICV observed in neuropsychologically compromised patients was not due to the greater proportion of schizophrenia spectrum patients in this group, or group differences in symptoms and antipsychotic dose. The results remained largely unchanged (supplementary material).

Follow-up analyses of specific tissue volumes were performed to determine if the group differences in TBV were due to selective reduction in total grey or white matter volume. These analyses focused on ICV adjusted tissue volumes because there were group differences in ICV. Main effects of group were detected for white matter (F(3,221) = 3.80, P = .011) and CSF (F(3,221) = 3.15, P = .026), but not grey matter (F(3,221) = 0.13, P = .945). As shown in the top panel of figure 2, the reduction in ICV adjusted TBV observed in neuropsychologically normal and impaired patients was due to a reduction in total white matter volume (P = .005 and P = .035, respectively) and increased CSF (P = .007 and P = .037). Moreover, as shown in the bottom panel of figure 2, the reduction in adjusted total white matter volume and increased CSF was specific to deteriorated patients (P = .009 and P = .030); compromised did not differ from healthy subjects on adjusted white matter (P = .353) and CSF (P = .199) volumes.

Fig. 2.

Brain tissue volumes in neuropsychologically (NP) defined groups of individuals with psychosis. Top panel: After adjusting for intracranial volume (ICV), total white matter volume was decreased and CSF volume increased in both NP normal and NP impaired patient groups. Bottom panel: Dividing the NP impaired group into deteriorated and compromised subgroups revealed that white matter volume reduction and increased CSF was present only in the deteriorated group; compromised patients did not differ from healthy subjects on any tissue volume. †Tissue volume adjusted for ICV. Error bars = standard error of the mean.

Given evidence that even psychosis patients with average overall cognitive ability may exhibit deficits in specific neuropsychological domains,⁵⁰ we repeated the analysis after further limiting the neuropsychologically normal patient group to only those patients that performed no worse than 1 SD below healthy subjects on every subtest of the SCIP. Twenty-two of 41 psychosis patients classified as neuropsychologically normal met this threshold. Even this highly restricted subset of neuropsychologically normal psychosis patients exhibited smaller adjusted TBV and white matter volume, adjusted for ICV, compared to healthy subjects (TBV: 1056.2±4.7ml vs 1073.0±2.9ml, P = .015, main effect of group F(4,220) = 2.42, P = .049; white matter volume: 489.9±5.3ml vs 502.4±2.5ml, P = .033, main effect of group F(4,220) = 2.84, P = .025).

Voxel-Based Morphometry.

Results of the VBM analyses are summarized in figure 3 and on serial axial slices covering the entire brain in supplementary figures 3 and 4. Compared to healthy subjects, neuropsychologically normal patients did not exhibit any localized changes in grey matter volume. White matter volume was reduced in a region of the frontal lobe beneath the superior and middle frontal gyri, genu of the corpus callosum, and in an area of the left medial parietal lobe adjacent to the posterior horn of the left lateral ventricle. In contrast, neuropsychologically impaired patients exhibited reduced grey matter volume in the left precentral gyrus corresponding to Brodmann’s area 6, medial parieto-occipital sulcus, bilateral middle occipital gyrus corresponding to Brodmann’s area 19, left superior temporal gyrus, left anterior insula/inferior frontal gyrus, thalamus, amygdala, hippocampus/parahippocampal gyrus, and cerebellum. Neuropsychologically impaired patients also exhibited widespread reductions in white matter, particularly in the centrum semiovale, sub-cingulate white matter, anterior and posterior periventricular regions, body and splenium of the corpus callosum, and brainstem. Lateral and third ventricle enlargement was present in both neuropsychological normal and neuropsychological impaired patients.

Fig. 3.

Regional brain volume abnormalities in individuals with a psychotic disorder classified on the basis of neuropsychological (NP) functioning. Top panel: Grey and white matter volume loss was qualitatively more extensive in NP impaired psychosis patients. Specifically, NP impaired patients exhibited grey matter volume loss in the superior temporal gyrus, frontal lobe, thalamus, hippocampus/entorhinal cortex, and cerebellum, and extensive white matter volume loss, particularly in periventricular regions and body/splenium of the corpus callosum. Brain tissue volume loss in NP normal patients was restricted to anterior and posterior periventricular areas, and genu of the corpus callosum. Ventricular enlargement was detected in both NP normal and impaired psychosis patients. Bottom panel: Among NP impaired patients, grey matter volume loss was present in both groups; however, white matter volume loss was more widespread in the deteriorated subgroup. Ventricular enlargement was detected in both deteriorated and compromised subgroups.

Among neuropsychologically impaired patients, regional changes in brain structure were qualitatively more extensive in neuropsychologically deteriorated patients (figure 2 and supplementary figure 4). Both deteriorated and compromised patients exhibited grey matter volume loss in the thalamus and medial temporal lobe structures. However, precentral gyrus volume loss was more pronounced in neuropsychological deteriorated patients, whereas volume loss in the left superior temporal gyrus and cerebellum was more pronounced in neuropsychologically compromised patients. White matter volume loss was more extensive in cognitively deteriorated patients and included the centrum semiovale, periventricular regions, especially white matter regions adjacent to anterior horns of lateral ventricles, and corpus callosum. In contrast, white matter volume loss in neuropsychologically compromised patients was restricted to posterior periventricular regions, splenium of the corpus callosum, cerebral peduncles, and pons. Enlargement of the lateral and third ventricles was present in both neuropsychologically deteriorated and compromised groups.

Relationship Between Brain Volumes, Clinical Symptoms, and Antipsychotic Dose.

After adjusting for age and sex, PANSS positive, negative, and general symptoms did not significantly correlate with ICV, TBV, and total tissue volumes (all partial correlations < |.14|, P > .123). Similarly, daily dose of antipsychotic medication did not correlate with ICV, TBV, and total tissue volumes (all partial correlations < |.14|, P > .129).

Discussion

We examined brain structure in psychosis patients grouped according to neuropsychological functioning in order to better understand the relationship between cognitive impairment and neuropathology in psychosis. As expected, neuropsychologically impaired patients exhibited reduced TBV, extensive white matter volume loss, and localized grey matter volume reductions in the frontal lobe, temporal lobe, including lateral and medial temporal lobe structures, thalamus, and cerebellum. Qualitatively, grey and white matter volume loss was considerably more extensive in neuropsychologically impaired patients compared with cognitively normal patients. These results are consistent with most prior investigations which have found greater reductions in cerebral volume in neuropsychologically impaired psychosis patients.^14,15,17

Unexpectedly, neuropsychologically normal patients also exhibited reduced TBV, after adjusting for ICV, due to lower total white matter volume. The presence of white matter volume loss in both neuropsychologically normal and impaired patients suggests that white matter pathology may be a core feature of psychosis that is partially independent of cognitive impairment. However, there is an important caveat to this interpretation; the possibility that neuropsychologically “normal” patients actually have some degree of cognitive impairment. Prior studies have found that even schizophrenia patients with average intellectual functioning have deficits in specific cognitive domains, including executive functioning and processing speed,¹⁸ leading some to argue that psychotic illnesses should be redefined as cognitive disorders.^51,52 It’s possible our cognitive battery was not comprehensive enough to detect impairment. Alternatively, standard population-based normative data may not be sensitive enough to detect impairment in all patients. For instance, even psychosis patients within the “normal” range based on standard population normative data perform worse than expected when compared to their monozygotic twin or their performance predicted by maternal education.^53,54 Thus, it’s possible that neuropsychologically “normal” patients, like their deteriorated counterparts, have experienced some degree of cognitive decline. If correct, it is perhaps not surprising that the 2 groups share some similarities in brain structure abnormalities. It is noteworthy in this regard that the reduction in TBV in neuropsychologically normal patients was only detected after adjusting for ICV, which was slightly larger than healthy subjects.

In contrast to the similarities between neuropsychologically normal and deteriorated patients, neuropsychologically compromised psychosis exhibited a unique pattern of brain structure abnormalities suggesting that the trajectory of brain dysmaturation or disease process is different for this group. Neuropsychologically compromised patients had the smallest ICV and absolute TBV among psychosis patients, the latter being consistent with the fact that they were also the most cognitively impaired. Critically, and in contrast to the other patient groups, compromised patients had relatively normal TBV and tissue volumes after adjusting for the reduction in ICV. As discussed earlier, ICV and TBV develop in parallel until late childhood/early adolescence when they begin to diverge; TBV gradually declines over time whereas ICV remains stable.^27–30 Reduced ICV in cognitively compromised patients suggests dysruption of brain development prior to late childhood/early adolescence or underdevelopment (ie, hypoplasia). Prenatal exposure to toxins and maternal infections, including rubella and toxoplasmosis, are associated with reduced ICV.^32,33 It’s possible that individuals with long-standing cognitive limitations and concomitant reduction in ICV have higher rates of prenatal infections and/or exposure to teratogens.⁵⁵ Conversely, reduced TBV in the context of normal ICV suggests derailment of late brain maturation, likely during later adolescence, and/or progressive volume loss in neuropsychologically normal and deteriorated patients. This is consistent with a recent longitudinal investigation by Meier and colleagues which found that IQ remained relatively stable between ages 7 and 13, but declined markedly between age 13 and when study subjects were retested in adulthood at age 38.⁵⁶

There are several limitations of this investigation in addition to those discussed earlier. Recent findings linking antipsychotic medications to progressive brain volume loss poses an interpretational challenge.⁵⁷ However, if our results were confounded by medication effects we would have expected the opposite of what we found; cognitively compromised patients, which had a marginally longer duration of illness and more severe symptomatology suggesting greater lifetime antipsychotic drug exposure, should have demonstrated the smallest relative TBV, not the largest, among the 3 cognitive groups. Additionally, the distinction between deteriorated and compromised groups relies on the assumption that a discrepancy between estimated premorbid and current cognitive abilities represents an actual decline in cognitive functioning. A recent longitudinal cohort study confirmed that WTAR scores acquired in healthy old age correlate with childhood cognitive ability, thereby validating the WTAR as a measure of premorbid cognitive functioning.⁵⁸ In psychosis, both cross-sectional^59–61 and longitudinal^62–64 studies have found evidence of developmental lag or possibly cognitive decline. Nonetheless, the absence of actual premorbid cognitive data is a limitation.

In conclusion, we found that patients with neuropsychological impairment and below average premorbid cognitive functioning exhibit evidence of early cerebral hypoplasia. In contrast, patients with evidence of cognitive decline (ie, average premorbid cognition, but impaired current neuropsychological functioning) and neuropsychologically normal patients show evidence of brain tissue loss consistent with progression or later cerebral dysmaturation. Longitudinal studies assessing both cognition and brain structure before and after illness onset will be required to confirm cognitive subgroups in psychosis and validate associations between cognitive trajectories and brain structure abnormalities.

Supplementary Material

Supplementary material is available at http://schizophreniabulletin.oxfordjournals.org.

Funding

Jack Martin, MD, Research Professorship (held by N.D.W.); the Vanderbilt Psychiatric Genotype/Phenotype Project; the Brain and Behavior Research Fund (National Alliance for Research on Schizophrenia and Depression Young Investigator Award awarded to N.D.W.); National Institute of Mental Health (RO1-MH070560 awarded to S.H.); the Vanderbilt Institute for Clinical and Translational Research (1-UL1-RR024975 National Center for Research Resources/National Institutes of Health).