Neuropsychological Correlates of Hippocampal Volumes in Patients Experiencing a First Episode of Schizophrenia

Neurodevelopmental models of schizophrenia posit an important role for aberrant hippocampal morphology in the pathophysiology of the disorder. There is considerable evidence from postmortem (1–6) and magnetic resonance imaging (MRI) (7–9) studies that patients with schizophrenia have hippocampal structural abnormalities. Moreover, a meta-analysis of 18 structural MRI studies (10) identified a 4% reduction in bilateral hippocampal volumes in patients with schizophrenia in relation to comparison subjects. Because hippocampal volume reductions have been identified in younger patients experiencing their first episode of schizophrenia (8), these abnormalities do not appear to be an artifact of long-term exposure to antipsychotic medications or to chronicity of the disorder.

Despite evidence for hippocampal pathology in patients with schizophrenia, few studies have examined their neuropsychological correlates using quantitative MRI. Several studies reported no significant associations of hippocampal volumes and neuropsychological functions (11–13). In contrast, Goldberg et al. (14) reported that in a group of monozygotic twins discordant for schizophrenia, intrapair differences in left anterior hippocampal volumes correlated significantly with intrapair differences in logical memory but not other functions. Bilder et al. (15) found that smaller anterior hippocampal volumes were associated with lower scores on measures of executive and motor functions considered sensitive to the integrity of the frontal lobes in patients experiencing a first episode of schizophrenia. This study was interpreted in the context of a neurodevelopmental defect within the medial frontolimbic system involving the anterior hippocampus, cingulate gyrus, and/or dorsal aspects of the premotor or prefrontal cortex. The medial frontolimbic system or dorsal archicortical trend (16) has been hypothesized to be important for redundancy-biased activation of cortical networks for executive functioning and forward planning; an abnormality in these brain structures and/or their connections may, at least in part, constitute the structural basis for frontal lobe dysfunction in schizophrenia (15, 17–20).

Several reviews (21, 22) have highlighted important neuroanatomical and functional differences between posterior and anterior hippocampal entorhinal-hippocampal projections (23). Specifically, the posterior hippocampus receives converging sensory input from posterior cortices, whereas the anterior hippocampus is connected primarily with the striatum and other parts of the limbic system. Empirical evidence for an anterior or posterior distinction of the hippocampus has been observed in functional MRI (24) and proton magnetic resonance spectroscopic imaging (25). In addition, in a meta-analysis of positron emission tomography (PET) studies of experimentally induced changes in blood flow (26), episodic memory encoding was found to be associated with anterior hippocampal activations, whereas episodic memory retrieval was associated with posterior hippocampal activations. A subsequent report (27), however, which included and excluded studies on the basis of criteria from the meta-analysis, reported that encoding activations in PET studies were observed frequently in both posterior and anterior mesiotemporal lobe regions. One possible reason for the apparent lack of a consistent relationship between hippocampal and neurocognitive measures in schizophrenia may be the lack of a distinction between the posterior and anterior hippocampus. Indeed, several studies that did distinguish between these regions implicated abnormalities in the anterior hippocampus compared to other parts of the hippocampus or other limbic structures (15, 28–31).

In this study we examined the neuropsychological correlates of posterior and anterior hippocampal volumes in large groups of men and women experiencing a first episode of schizophrenia. For significant correlations of hippocampal volume and neuropsychological functioning, we examined the pattern of correlations between that neuropsychological domain and individual slice volumes across the long axis of the hippocampus after cubic-spline interpolation to 10 evenly spaced slice positions. Correlations were examined separately by sex, given evidence that men with schizophrenia have been found to have more severe temporal (32, 33) and mesiotemporal (3, 8) lobe abnormalities than women with schizophrenia as well as reports noting sex differences between patient groups in patterns of structure-function relations (20, 34).

Method

Subjects

Patients with schizophrenia were participants in our larger Prospective Study of Psychobiology in First-Episode Psychosis and were recruited from patients consecutively admitted to the inpatient service at Hillside Hospital. Detailed information regarding the ascertainment, treatment, and clinical characteristics of the overall study group has been published elsewhere (35, 36). Inclusion criteria included admittance for a first episode of psychotic illness as defined as requiring hospital admission and having a current rating of 4 or more on at least one of the psychotic items of the Schedule for Affective Disorders and Schizophrenia (SADS), a diagnosis of definite or probable schizophrenia or of schizoaffective disorder, mainly schizophrenia, and fewer than 12 weeks cumulative (lifetime) antipsychotic treatment. Patients were excluded if there was any medical contraindication to the use of antipsychotic medications or if there was any history of serious neurologic or endocrine disorder. All procedures were approved by the institutional review board of Long Island Jewish Medical Center, and written informed consent was obtained from all patients.

The 75 patients (43 men, 32 women) described in this study were a subset of the 118 patients who participated in our larger study who had MRI scans available for measurement of the hippocampus and whole brain and who had completed at least one comprehensive neuropsychological examination after meeting operational criteria for clinical stabilization. Seventeen individuals in the present study were included in a prior study (15), but these patients were remeasured blindly as part of a larger series that included patients and healthy comparison subjects. The average number of weeks from study entry to the neuropsychological and MRI examinations is illustrated in Table 1. The men and women included in this study were not significantly different, respectively, from the male and female patients in the larger study who were not included in this study in terms of age, education, parental social class, baseline clinical global impression rating, age at onset of psychotic symptoms, and number of weeks of psychotic symptoms before study entry (all p>0.05). Diagnoses of patients in this study were based on the SADS interview and supplemental information from family members obtained by using the Research Diagnostic Criteria (37) and included schizophrenia (N=56; subtypes were paranoid [N=45], disorganized [N=4], and undifferentiated [N=7]) and schizoaffective disorder (N=19).

MRI Procedures

MRI scans were completed as soon as it was practical for patients after study entry; in most cases patients were receiving antipsychotic medication at the time of the MRI scan. The average numbers of weeks from the administration of antipsychotic medication to the MRI examination and from the MRI examination to the neuropsychological examination are reported in Table 1. MRI scans were acquired in the coronal plane by using a three-dimensional, gradient echo, fast low-angle-shots sequence, with a 50° flip angle, 40-msec TR, and 15-msec TE (38) on a 1.0-T whole-body superconducting imaging system (Siemens Magnetom, Erlangen, Germany). This sequence produced 63 contiguous coronal slices (slice thickness=3.1 mm) through the whole head in about 11 minutes, with a nominal in-plane resolution of 1.17 mm × 1.17 mm in a 256×256 matrix. Before the scan, head tilt was adjusted by using external landmarks. In addition, a two-dimensional fast low-angle-shots scan was used to position the head so that the floor of the fourth ventricle was parallel to the inferior-superior plane. The patients typically received 200–400 mg of sodium amytal orally before the procedure. Each of the scans was reviewed by a neuroradiologist and a member of the research team. Any scan with significant artifacts was repeated.

Measurement Criteria

All measurements were completed by using a semiautomated computerized mensuration system described previously (38). Before all measurements, scans of patients with schizophrenia were mixed together with those of healthy comparison subjects and were flipped randomly. No identifying information was available from the scan. Measurements were thus completed by an operator who was blind to group membership and hemisphere. Whole-brain measurements included both hemispheres, from the frontal to the occipital poles, the brainstem (i.e., mesencephalon, pons, and medulla), cerebellum, and ventricles but excluded the sulcal CSF. Interrater reliability (as assessed by intraclass correlations) between two operators regarding whole-brain measurements for 10 patients was 0.96. Measurement of the hippocampus complex was on the basis of operational criteria from postmortem histological work (2, 39). The anatomic regions used for measurement are illustrated elsewhere (8) and have been used in our previous MRI research (7, 8, 15). Two contiguous portions of the hippocampal complex in each hemisphere were analyzed: 1) the anterior hippocampus (including the mammillary body slice and the slice immediately posterior to this [postmammillary body slice]) and 2) the posterior hippocampus (from the slice posterior to the postmammillary slice to the slice at which the ascending fornix is interrupted by the surrounding pulvinar). Measurements of the hippocampal formation included all CA segments (CA1, CA2, CA3, and CA4), the dentate gyrus, the alveus, and the subicular region, which could not be separated in the scans. Intraclass correlations between two operators on scans for 15 patients ranged from 0.69 to 0.87 for the right and left posterior and anterior hippocampus and were 0.82 and 0.78 for the total posterior and total anterior hippocampus, respectively.

Hippocampus Slice Interpolation

The hippocampal slice volumes were resampled so that each individual’s measurements comprised 10 slices per hemisphere. This was implemented in MATLAB (40) by performing a cubic-spline interpolation of the measured hippocampal slice areas and resampling the spline curve at 10 equally spaced slice positions.

Neuropsychological Procedures

Because of the possibility that neuropsychological assessments could be confounded by factors typically associated with a first episode of schizophrenia, these tests were conducted after patient remission or when patients had a stable level of residual symptoms for the preceding 2 weeks, as measured by rating scale assessments (41). The patients were therefore receiving medication at the time of neuropsychological testing. The average number of weeks from the administration of antipsychotic medication to the neuropsychological examination is shown in Table 1. Administration of the neuropsychological tests was counterbalanced and took place only if the patients were clinically stable. The comprehensive test battery has been used extensively in our previous research (15, 20, 42) and includes 41 tests selected to characterize six domains of neuropsychological functioning: language, attention, memory, executive, motor, and visuospatial (Appendix 1).

Scaled scores for each domain for the patients in this study are from Bilder et al. (42) and were computed by averaging the z scores for each contributing test variable. A combined group of 36 healthy comparison subjects (24 men and 12 women) was used to generate z scores for the patients on the contributing neuropsychological variables. There were significant performance differences between the healthy men and women on the Finger Tapping test; thus z scores for the patients on these tests were computed by using a same-sex comparison group rather than the combined group. Functional scales were computed so that higher values indicated better performance. Additional information regarding the use of transformations to stabilize variances and the validity of a priori assignments of variables to scales by confirmatory factor analysis is provided elsewhere (42).

Handedness was assessed by using a modified version of the Edinburgh Inventory (43) consisting of 20 items. Total number of right- and left-hand items were scored, and the laterality quotient was computed according to the following formula: (total right – total left)/(total right + total left). This yielded a total laterality quotient for each participant that ranged from 1.00 (totally dextral) to –1.00 (totally nondextral). Subjects with a laterality quotient greater than 0.70 were classified as dextral and the rest as nondextral (44). Information regarding handedness classifications is provided in Table 1.

Statistical Analyses

To minimize type I errors, we made several predictions regarding the direction of the correlations between the neuropsychological domains and hippocampal volumes on the basis of prior empirical and theoretical research. Specifically, we hypothesized that smaller anterior hippocampal volume would be significantly correlated with lower scores on the executive and motor functional domains. Therefore, although the main hypotheses focused on the associations of anterior hippocampal volume with executive and motor performance, we examined all the neuropsychological correlates of both posterior and anterior hippocampal volumes to determine specificity of the findings. Given that distributions of variables were judged to have a normal shape, Pearson’s product-moment correlations (two-tailed; alpha=0.05) were used for investigation of structure-function relations. Partial correlation analysis was used to investigate the possible effects of age at onset and parental social class on significant findings. For significant correlations of hippocampal volume and neuropsychological functioning, we tested the hypothesis that there is a linear relationship between performance on that neuropsychological domain and individual slice volumes across the long axis of the hippocampus after cubic-spline interpolation to 10 evenly spaced slice positions. Curve estimation was used to model the pattern of correlations between structure and function across the long axis of the hippocampus; hippocampal slice number was the independent variable, and the structure-function correlation was the dependent variable.

We compared correlated correlation coefficients using the method described by Meng et al. (45). In this method two correlations obtained in a single group of subjects can be compared where each correlation is between one predictor variable (X₁ or X₂) and a single common dependent variable (Y). Its purpose, therefore, is to test the null hypothesis that two correlation coefficients are not significantly different and not to test whether any single correlation is statistically significant. Tests of the difference between correlation coefficients employed r-to-z transformations to test the null hypothesis that a structure-function correlation obtained for the male patients did not differ significantly from the same structure-function correlation obtained for the female patients.

Differences in group characteristics, brain volumes, and neuropsychological performance were examined by using either independent-groups t tests or chi-square analyses (Table 1). Differences in the average number of weeks between the various time points in the study were examined by using nonparametric Mann-Whitney U tests because of violations in the homogeneity-of-variance assumption (Table 1). To control for nonspecific size differences among study participants, anterior and posterior hippocampal volumes and hippocampal slices were adjusted for total brain volume through linear regression; these analyses were conducted separately for male and female patients.

Results

Group characteristics and the numbers of weeks between various time points in the study are reported in Table 1. Male and female patients did not differ significantly at the time of the MRI or neuropsychological examinations in age, years of education, parental social class, handedness, or performance in any of the neuropsychological domains. In addition, there were no significant differences between male and female patients in the number of weeks between the various time points in the study (i.e., study entry, administration of antipsychotic medications, MRI examination, and neuropsychological examination). The groups did differ significantly, however, in age at onset of psychotic symptoms and in racial or ethnic composition (Table 1). We also compared demographic variables for the 75 patients in this study and the 36 healthy comparison subjects who were used to standardize patient scores on the neuropsychological domains. Similar to our previous study (42), the patient and healthy comparison groups were well matched on distributions of sex, age, and handedness (all p>0.05) but differed in parental social class (χ²=6.2, df=1, p=0.01). Possible effects of this difference were investigated in subsequent analyses.

The correlations between total posterior and total anterior hippocampal volumes and the six neuropsychological domains are illustrated in Figure 1 and Figure 2. Figure 1 shows that posterior hippocampal volume was not significantly correlated with functioning on any of the neuropsychological domains among men or women with a first episode of schizophrenia (all p>0.05). Anterior hippocampal volume was significantly correlated only with executive (r=0.33, df=43, p=0.03) and motor (r=0.38, df=42, p=0.01) functioning among men; none of the correlations between anterior hippocampal volume and neuropsychological functioning was statistically significant among women with a first episode of schizophrenia (Figure 2). Figure 3 and Figure 4 provide the scatter plots for the relation of anterior hypocampal volume to executive or to motor functioning among men with a first episode of schizophrenia. The use of Spearman’s rank-order correlations did not alter any of the significant (or nonsignificant) findings.

Additional analyses examined the possible effects of racial or ethnic group composition by comparing the coefficients of determination (r²) for the Caucasian subjects, the only racial or ethnic group that was large enough for analysis, and the entire group of male patients. The coefficients of determination for the correlations of anterior hippocampal volume and executive functioning and anterior hippocampal volume and motor functioning were comparable between groups. We also examined the possible effects of parental social class and age at onset (by using partial correlations) on the observed findings; these analyses did not differ substantively from the original analyses.

Tests of the difference between correlated correlation coefficients indicated that among the male patients (N=43), anterior hippocampal volume was significantly more strongly correlated with executive functioning than with either memory (difference: z=1.98, p=0.02) or language (difference: z=2.47, p=0.007) functioning. Similarly, among the male patients (N=42), anterior hippocampal volume was significantly more strongly correlated with motor functioning than with either memory (difference: z=2.03, p=0.02) or language (difference: z=2.11, p=0.02) functioning. Neither executive nor motor functioning was more strongly correlated with anterior hippocampal volume than with posterior hippocampal volume (p>0.05). Anterior hippocampal volume was more strongly correlated with motor functioning among male (N=42) than female (N=31) patients (difference: z=2.34, p=0.01).

We tested the hypothesis that there would be a linear relationship between executive and motor performance and hippocampal volume across its long axis (after cubic-spline interpolation to 10 evenly spaced slice positions). The correlations between executive and motor functioning and volumes of individual hippocampal slices are illustrated in Figure 5 and Figure 6, respectively. Results of these analyses, however, indicated that when examined across its long axis, the relationships between hippocampal volume and executive performance (F=1.37, df=1, 8, p=0.28) and hippocampal volume and motor performance (F=1.05, df=1, 8, p=0.34) were not linear.

Discussion

The main finding of this study is that anterior hippocampal volume correlates significantly with neuropsychological functions in male patients experiencing a first episode of schizophrenia who were studied soon after illness onset and before extensive pharmacological intervention. Among male patients, anterior hippocampal volume was more strongly correlated with both executive and motor functioning than with either memory or language functioning. These findings also suggest that the pattern of correlations between structure and function differ between male and female patients.

Despite evidence for hippocampal abnormalities in patients with schizophrenia, few MRI studies have investigated their neuropsychological correlates. In one such study, Hoff et al. (11) reported no significant associations of hippocampal volume with neuropsychological measures in a sample of patients experiencing a first episode of schizophreniform illness. Similarly, DeLisi et al. (12) and Nestor et al. (13) reported no significant neuropsychological correlates of hippocampal volumes in their patient samples, although significant associations were identified with the parahippocampal gyrus. In contrast to these studies, Goldberg et al. (14) implicated the left anterior hippocampus in the inability of patients to repeat stories read to them. Differences between our study findings and those of previous reports may be related to selection and diagnostic issues and methodological differences in measuring the hippocampus, including our distinction between the posterior and anterior hippocampus. In addition, if neuropsychological deficits are more strongly tied to anterior hippocampal abnormalities, studies investigating the entire hippocampus might not detect these associations.

The finding that smaller anterior hippocampal volume was associated with deficits in performance of executive and motor tasks considered sensitive to the integrity of frontal lobe functioning may seem paradoxical in light of research that has implicated the hippocampus in memory formation and consolidation (46, 47). Moreover, we did not observe any significant correlation between hippocampal volume and memory functioning in patients. It should be noted, however, that although studies have identified memory deficits as the most conspicuous result of hippocampal lesions, such findings were often obtained following normal development in healthy humans after they sustained some focal lesion or in temporal lobe epilepsy. In schizophrenia, hippocampal abnormalities are more likely the result of some neurodevelopmental abnormality rather than the result of an acquired lesion, and therefore, the adult-lesion model may not be applicable (48, 49). Animal studies have indicated that early developmental lesions to the hippocampal formation may not yield evidence of memory deficits; rather, these animals demonstrated both pharmacological and behavioral abnormalities that are more consistent with frontal lobe lesions in adult animals (50–53). Particularly noteworthy is that these animals with developmental lesions demonstrated evidence of “frontal dysfunction” only when they matured into adolescence or adulthood. Our results are therefore consistent with the hypothesis that some neurodevelopmental abnormality affecting the anterior hippocampus may later disrupt frontolimbic functioning in adulthood, thereby yielding a pattern of frontal lobe dysfunction. Thus, the adult-lesion model (with its implication that memory deficits are strongly tied to hippocampal integrity) may therefore be less relevant to our understanding of structure-function relations in schizophrenia, given that neurodevelopmental mechanisms are hypothesized to play a significant role in the pathophysiology of the disorder.

Given the extensive anatomic connections between frontal and mesiotemporal regions, a defect in connectivity between these regions may therefore be implicated in the pathophysiology of schizophrenia (17, 26). Csernansky et al. (54) found that the superior and lateral aspects of the hippocampal head, which have strong connections with the medial prefrontal regions, had the greatest shape abnormalities in patients with schizophrenia. Similarly, Weinberger et al. (55) found that differences in anterior hippocampal volume computed between monozygotic twins who were discordant for schizophrenia were significantly correlated with the difference between the twins in regional CSF to the prefrontal cortex during performance of the Wisconsin Card Sorting Test. It is therefore plausible that neurodevelopmental abnormalities in the medial frontolimbic system—which includes the dorsal aspects of the premotor and prefrontal cortex, cingulate gyrus, and anterior hippocampus—may comprise the structural basis for the observation of executive and motor dysfunction in schizophrenia. The medial frontolimbic system, in which the frontal lobes and the limbic system are linked by the cingulate bundle (17), is isomorphic with the medial/dorsal archicortical system and can be distinguished from the ventral or lateral paleocortical system that comprises the olfactory cortex and the peri-insular and ventral neocortices, including the orbital frontal cortex (16). The dorsal archicortical system has been hypothesized to be important for projectional control of behavior (56), including the types of executive and motor functions that correlated significantly with anterior hippocampal volume in this study (57).

The present study, conducted with large groups of men and women with a first episode of schizophrenia, suggested sex differences in the functional correlates of anterior hippocampal volumes. Specifically, anterior hippocampal volume was more strongly correlated with motor dysfunction in male than in female patients. It is important to acknowledge a possible selection bias, however, in that female patients who were functioning better overall may have dropped out of the study earlier than women who were not functioning as well and, therefore, did not complete the neuropsychological assessment (41). It is difficult to compare our results with prior research because few studies have investigated sex differences in structure-function relations in patients with schizophrenia. In one such study, Flaum et al. (34) found that full-scale IQ was uncorrelated with brain regions of interest for their entire group of patients with schizophrenia. When patients were divided by sex, however, female patients had a pattern of correlations similar to that observed in normal comparison subjects, while no such relationship was evident among the male patients. In another study, Szeszko et al. (20) found that smaller anterior cingulate gyrus volume correlated significantly with worse executive functioning in male, but not female, patients experiencing a first episode of schizophrenia. Future research would benefit from the investigation of possible sex differences in analyses of structure-function relations in schizophrenia, as this may have implications for better understanding the pathophysiology of the disorder. For example, it may be that male patients have more pronounced abnormalities in the medial frontolimbic system, and this may be reflected in their neuropsychological test performance.

Although executive and motor functioning correlated significantly with anterior hippocampal volume and not with posterior hippocampal volume in the current study, these correlations were not statistically different from one another. This lack of anatomic specificity was most evident in the slice-by-slice analysis of hippocampal volumes in the male patients, which revealed that the relationship between executive and motor performance and hippocampal volume along its long axis did not fit a linear model. It was not possible to determine in this study, however, whether correlations involving the posterior hippocampal slices reflected morphologic abnormalities in the hippocampus or possibly in adjacent subcortical structures. Moreover, hippocampal tissue posterior to where the crus of the fornix is interrupted by the surrounding pulvinar was not included in the posterior hippocampal volume measurement, and consequently, we were unable to determine whether the most posterior part of the hippocampal formation would also show a similar pattern of correlations.

It is important to note several limitations of our hippocampal delineation methods that may have influenced the observed pattern of correlations. First, although we used the MRI scan that contained the mammillary bodies to assist in making the distinction between the posterior and anterior hippocampus before analysis, the use of this landmark to differentiate these regions is arguably somewhat arbitrary and may, therefore, have only approximated true functional differences between these regions. Second, because precise separation of the anterior hippocampus from the amygdala was not possible in these MRI scans, the anterior hippocampal volume measurement may have included the most caudal parts of the amygdala, thereby lowering the correlations of anterior hippocampal volume with the executive and motor domains.

In summary, this study indicated that there are neuropsychological correlates of hippocampal volumes in male patients experiencing a first episode of schizophrenia and that this relationship may not be linear along the long axis of the hippocampus. The results of the present study also suggest that our prior findings (15) were confirmed in a larger group of patients, even after independent remeasurement of the hippocampal formation. Strengths of this study include the large groups of male and female patients who were studied soon after illness onset, our assessment with a comprehensive neuropsychological test battery, and the distinction between the posterior and anterior hippocampus. Future studies should employ more precise methods for separation of the anterior hippocampus from the amygdala to better elucidate the functional correlates of these brain structures.

TABLE 1

Enlarge table

Enlarge table