A total of 335 patients with schizophrenia spectrum disorders and 198 HVs were available for this study through the University of Iowa Mental Health Clinical Research Center. These subjects participated in various Mental Health Clinical Research Center studies approved by the University of Iowa Institutional Review Board, and all subjects gave written informed consent to participate in these protocols.
Patients were evaluated using a semistructured interview instrument, the Comprehensive Assessment of Symptoms and History,21
from which schizophrenia (n=310), schizoaffective disorder (n=21), delusional disorder (n=2), schizophreniform disorder (n=1), and schizotypal personality disorder (n=1) diagnoses meeting DSM-III-R22
criteria were based. Healthy volunteers (n=198) were recruited from the community through newspaper advertisements. They were initially screened by telephone and further evaluated using an abbreviated version of the Comprehensive Assessment of Symptoms and History instrument to exclude subjects with current or past medical, neurologic, or psychiatric illnesses or with schizophrenia in first-degree relatives.
The DNA was prepared by high-salt extraction from whole blood.24
One ng/µL DNA was amplified using ABI 2720 thermocyclers, an ABI Taqman SNP Genotyping 5′ nuclease assay for rs1347706, and ABI Taqman Universal PCR Master Mix (all, Applied Biosystems). Alleles were called using ABI Prism 7900 end point–read allelic discrimination software (Applied Biosystems). Replicate samples were included on all genotyping plates to ensure accurate allele calling.
MRI ACQUISITION AND IMAGE PROCESSING
High-resolution brain anatomical MRI data in this study were collected from 1 of 2 imaging protocols. Scans performed before calendar year 2000 (termed MR5) were obtained on a 1.5-T GE Signa MR scanner (General Electric Medical Systems). For the MR5 imaging protocol, 3-dimensionalT1-weighted images were acquired in the coronal plane using a spoiled gradient recalled acquisition in steady state sequence. The parameters were echo time (TE)=5 milliseconds, repetition time (TR)=24 milliseconds, numbers of excitations (NEX)=2, rotation angle=45°, field of view (FOV)=26×24×18.8 cm, and matrix=256×192×124. Two-dimensional proton density and T2 sequences were acquired as follows: 3.0- or 4.0-mm-thick coronal slices, TR=3000 milliseconds, TE=36 milliseconds (for proton density) and 96 milliseconds (for T2), NEX=1, FOV=26×26 cm, and matrix=256 × 192.
Scans from 2000 or later (termed MR6) were obtained on a 1.5-T GE CVMRI scanner (General Electric Medical Systems) using T1- and T2-weighted sequences. For the MR6 imaging protocol, the T1 sequence was gathered as a 3-dimensional volume in the coronal plane using a spoiled gradient recalled acquisition in steady state sequence with the following parameters: TE=6 milliseconds, TR=20 milliseconds, flip angle=30°, FOV=160 × 160 × 192 mm, matrix=256 × 256 × 124, and NEX=2. The MR6 T2 images were acquired using a 2-dimensional fast spin-echo sequence in the coronal plane. The parameters were TE=85 milliseconds, TR=4800 milliseconds, slice thickness/gap = 1.8/0.0 mm, FOV=160×160 mm, matrix=256 × 256, NEX=3, number of echoes=8, and number of slices=124.
To enhance MR5 and MR6 data compatibility, MR6 scans were resampled into the same resolution and image size as MR5 scans so as to simulate similar amounts of partial volume effects in voxels that border 2 tissue types. To verify our ability to combine data from the 2 MR protocols, we acquired MR5 and MR6 scans on 60 patients.25
Brain volume differences between the 2 imaging sequences were small (median percent difference, 0.19%). Intraclass correlations were high across regions of interest (median intraclass correlation, 0.97). Hence, MR5 and MR6 data are compatible for combined statistical analyses.
Magnetic resonance images were processed using our locally developed BRAINS2 (Brain Research: Analysis of Images, Networks, and Systems, version 2) software package.26
Detailed descriptions of image analysis methods have been provided elsewhere.27–30
In brief, the T1-weighted images were spatially normalized and resampled so that the anterior-posterior axis of the brain was realigned parallel to the anterior-posterior commissure line, and the interhemispheric fissure was aligned on the other 2 axes. The T2-weighted images were aligned to the spatially normalized T1-weighted image using an automated image registration program.31
These images were then subjected to a linear transformation into standardized stereotaxic Talairach atlas space32
to generate automated measurements of frontal, temporal, parietal, and occipital lobes.29
To further classify tissue volumes into GM, WM, and cerebrospinal fluid, we used a discriminant analysis method of tissue segmentation based on automated training class selection that used data from the T1 and T2 sequences.31
In this study, we examined total and lobar cerebral cortical GM and WM volumes and cerebrospinal fluid volume of the lateral ventricles.
Symptom severity was assessed at the time of brain imaging using the Scale for the Assessment of Positive Symptoms (SAPS)33
and the Scale for the Assessment of Negative Symptoms (SANS).34
All available sources of information were used to assess symptom severity, including patient reports, informant interviews, and medical records. Items corresponding to the SANS/SAPS global items were rated using a score ranging from 0 (none) to 5 (severe) and then grouped into psychotic, disorganized, and negative symptom dimensions that have repeatedly been shown to cluster independently.35
The psychotic dimension summed ratings for hallucinations and delusions; the disorganized dimension included ratings for psychotic formal thought disorder, bizarre/ disorganized behavior, and inappropriate affect; and the negative dimension included ratings for attention, affective flattening, alogia, avolition/apathy, and anhedonia/asociality.
Genotype frequencies of rs1344706 were compared between patients and HVs using χ2 tests of differences.
The relationships between rs1344706 genotype and the quantitative measures were examined using analysis of covariance (ANCOVA). Genotype is a 3-level variable that can be coded in numerous ways. One could presume an additive model in which each copy of the risk allele influences quantitative traits in an incremental fashion, but this would fail to detect recessive/ dominant effects. Many studies, because of power limitations due to small sample sizes, combine minor allele homozygotes with heterozygotes and compare these against major allele homozygotes, but this would fail to detect additive effects or dominant effects of the major allele. Both of these approaches would be tests with 1 df. Given our larger sample, we chose a more conservative, inclusive approach, coding genotype as a 3-level categorical variable free from a priori assumptions, thereby producing a test with 2 df that compared each genotype group against the other 2.
We followed up these analyses with specific tests of replication. Lencz et al,9
for example, found that HV risk allele homozygotes had increased frontal lobe WM volumes compared with other HVs, so we performed a similar test. Donohoe et al12
found that risk allele homozygotes with schizophrenia had increased superior temporal GM volumes compared with others with schizophrenia, so we performed a similar analysis.
For tests of brain structure, MRI volumes were the dependent measures, genotype was the independent predictor, and intracranial volume, age at the time of imaging, sex, lifetime antipsychotic treatment, imaging protocol, and diagnostic grouping were entered as covariates. A genotype×diagnostic group interaction term was also included to assess whether genotype relationships with brain volumes differed across patients and HVs. For tests of symptom measures, genotype was the predictor, the 3-symptom summary scores were independent variables, and age and sex were covariates.
The brain structure tests included a treatment variable because we and others have shown that antipsychotic medications can affect brain structure volumes.25,36
We use a measure that converts all antipsychotic medications into chlorpromazine hydrochloride dose equivalents and calculates a lifetime chlorpromazine equivalent exposure with the unit of measure being years taking chlorpromazine at 100 mg/d.37
The value for HVs was zero, and because many of our schizophrenia subjects were early in the course of illness, this variable was not normally distributed and so was rank ordered before analysis.
Determining an appropriate significance level for tests performed in this study is complicated by a number of factors. First, the brain structure volumes that we test are correlated with each other, and so treating them as independent, as in a Bonferroni correction, would be excessively conservative. Even after removing the variance due to the covariates, the GM measures are correlated with r values of 0.13 to 0.81, while the WM volumes are intercorrelated with r values of 0.41 to 0.90, all of which are highly significant (except for frontal and occipital GM volumes, which are not correlated).This pattern also diminishes enthusiasm for a multivariate ANCOVA test, which is more appropriate for dependent measures that are not highly intercorrelated. Furthermore, as noted earlier, the primary motivation for this study was to test previously identified associations; although we perform additional tests, these are not numerous. In light of this, we choose P=.01 for significance.