Diffusion Tensor MR Imaging Evaluation of Callosal Abnormalities in Schizophrenia: A Meta-Analysis

Widespread white matter (WM) abnormalities have been found in patients with schizophrenia. Corpus callosum (CC) is the key area that connects the left and right brain hemispheres. However, the results of studies considering different subregions of the CC as regions of interest in patients with schizophrenia have been inconsistent. To obtain a more consistent evaluation of the diffusion characteristics change of the corpus callosum (CC) related to schizophrenia. A meta-analysis involving fractional anisotropy (FA) values in the CC of 729 schizophrenic subjects and 682 healthy controls from 22 studies was conducted. Overall FA values in the CC of the schizophrenic group were less than that of the healthy control group [weighted mean difference (WMD) = -0.021,P< 0.001]. So were the FA values in the genus region (WMD = -0.019, P< 0.001) and the splenium region (WMD = -0.020, P< 0.001) of the CC respectively. The FA reduction was also significant in subjects with chronic schizophrenia (WMD = -0.032, P< 0.001) and first-episode schizophrenia (WMD = -0.014, P = 0.001). In present study, we demonstrated an overall FA decrease in the CC of schizophrenic patients. In the two subgroup analyses of the genu vs splenium region and chronic vs first-episode schizophrenia, the decrease of all groups was significant. Further studies with more homogenous populations and standardized DTI protocols are needed to confirm and extend these findings.


Introduction
More than two hundred million people around the world are suffering from schizophrenia. Schizophrenia is an incapacitating psychiatric disorder characterized by disturbances in perception, cognition, and emotion, as well as an altered sense of self and abnormal behavior patterns. People with schizophrenia experience delusions and auditory hallucinations [1]. The risk of early mortality among schizophrenic patients is more than double that of the general population. It is attributed to a variety of conditions, including sequelae of pathogen infections, cardiovascular diseases and metabolic disorders (http://www.who.int/). Physiologically, schizophrenia is associated with an impaired white matter (WM) connectivity [1]. It has been suggested that some of the symptoms of schizophrenia may be due to observed abnormalities of the corpus callosum (CC) [2], a massive bundle of WM tracts that mediates the cross-talk between the cerebral hemispheres. Indeed, patients with schizophrenia have been found to have a smaller CC than healthy subjects [3].
Diffusion tensor imaging (DTI) provides a convenient way of studying brains of schizophrenic patients and healthy controls. DTI is a non-invasive magnetic resonance (MR) method of assessing water diffusion tendencies in WM structures [4]. The diffusivity isrelated to boundaries created by neural tracts and myelin sheaths and thus can be used as a marker for WM tract integrity [5].Many properties can be calculated from the data obtained by DTI, such as mean diffusivity (MD), apparent diffusion coefficient(ADC),Fractional Anisotropy (FA) and the axial/radial diffusivity [6].
The findings of a recent meta-analysis of DTI studies suggested that schizophrenia may involve an abnormality of the splenium region of the CC in particular and thus perhaps altered posterior interhemispheric connectivity [7].Widespread WM abnormalities have been found in patients with first-episode or chronic schizophrenia [1,[8][9][10][11][12].
In this study, we conducted a meta-analysis of WM diffusion alterations in the CC of schizophrenic patients, both chronic and first-episode cohorts, compared with observations in healthy control subjects. Fractional anisotropy (FA) values of collated studies were submitted to statistical analysis.

Materials and Methods
The protocol for this meta-analysis supporting checklist and flow diagram are available as supporting information (S1 File, S1 Checklist)

Data sources and searches
We performed a comprehensive search on electronic database including PubMed, Embase and Cochrane for relevant studies published from the earliest indexing year to June 15 th , 2016. Document type was restricted to article and publication language was restricted to English.
In order not to miss out papers that did not mention the CC in the title or abstract but covered analysis as part of the white matter or the whole brain in the full text, we extended our search to include white matter and brain. Eventually, the key words used to define relevant articles included: diffusion tensor imaging, schizophrenia, corpus callosum, white matter and brain. Besides the subject terms, some other related free terms were also used. See S2 File for detailed search strategy.

Study inclusion criteria
Studies were evaluated for inclusion and the criteria were the followings: (1) subjects with schizophrenia were compared to unrelated healthy controls; (2) the diagnostic criteria for schizophrenia were clearly stated; (3) they obtained DTI data with a 1.5-T or 3.0-T scanner (sufficient power to meet advanced technical standards for clinical brain imaging); (4) at least a single FA value in the whole CC and/or the genu or splenium of the CC was reported. variables (schizophrenia type, age, gender, handedness, medication use, duration of illness) and dependent variables (FA means, FA standard deviations).

Statistical analysis
Meta-analyses of continuous outcomes (FA values) were performed with weighted mean differences (WMDs)in Stata software (Version 14.0 Stata Inc., College Station, TX). The DerSimonian and Laird random-effects model was employed with a 95% confidence interval (CI) to account for measurement variability among the included studies. We pooled and compared FA data from schizophrenic subjects versus from healthy controls. Additionally, we pooled and statistically compared FA data of genu region with splenium region and data of first-episode schizophrenia subjects with chronic schizophrenia subjects. The random-effects model was used, and inconsistency was evaluated by the I-squared index. The resultant data are presented in forest plots. Publication bias was assessed by Egger's method. For all comparisons, P values < 0.05 was considered the criterion of statistical significance.

Included studies and characteristics
Our search of the three databases identified 678 citations after excluding duplicates with computer assistance and manually. We scanned the title and abstract or the full text of articles and excluded 610 papers obviously unrelated to our purpose. We examined the remaining 68 articles carefully and 22 meeting our criteria were included in this meta-analysis finally (Fig 1).
The study group characteristics are summarized in Table 1. The meta-analysis included a total of 729 schizophrenia subjects [FA values from 1229 volumes of interest (VOIs)] and 682 healthy controls (1314 VOIs) from 22 studies (42 datasets). In general, the number of male subjects is larger than that of female subjects. Most of the subjects are right handed. The medication use of Schizophrenic patients is different among the studies. Most of the studies include patients with medications and some focus on drug-naive patients.
The imaging parameters of the included studies are summarized in Table 2. Briefly, a 3-T MRI scanner was used in 12 of the included studies and a 1.5-T MRI scanner was used in 10 of the studies. A maximum b factor of 1000 s/mm 2 was most common. The DTI directions varied considerably, ranging from 6to 64. Scan thicknesses were mostly in the range of 2.5-5 mm.
Overall See S3 File for detailed FA data extracted from all included studies. The overall analysis revealed lower FA values for WM in the CC of subjects with schizophrenia than of unrelated healthy controls (WMD = -0.020, 95%CI = -0.025--0.016, Z = 8.55, P< 0.001,I 2 = 88.0%) and the details was presented in Fig 2. The result of Egger's test (Fig 3) indicated that there was no publication bias (t = 0.6, P = 0.55).

Discussion
The present study demonstrated reduced FA values in the CC of both first-episode and chronic schizophrenia patients, relative to healthy controls.  Our findings are consistent with previous studies showing FA abnormalities in the entire CC of schizophrenic patients [19,35]. However, the results of studies considering different subregions of the CC as regions of interest have been inconsistent [36]. Nevertheless, findings of WM changes in schizophrenic patients are consistent with other prior findings of schizophrenia-associated loss of WM integrity in principle. Notably, Kunimatsu and colleagues have described reduced expression of genes related to myelin/oligodendrocyte function, oligodendrocyte hypodensity and lamellar abnormalities of myelin as well as abnormal distribution patterns of interstitial neurons in schizophrenic brains [25].  Although schizophrenia presents in late adolescence to early adulthood, it has been theorized to have an extended neurodevelopmental etiology, with brain pathophysiological processes beginning long before the overt manifestation of clinical symptoms [37].Schizophreniaassociated differences in FA values may have embryonic, structural and functional causes. The genu region is formed by axons projecting from the cingulate cortex under the guidance of glia cells, whereas the splenium region develops under the guidance of axons passing from the hippocampal commissure without glial guidance [38].
This meta-analysis study had several limitations that should be acknowledged. Firstly, the study samples were not homogenous. The considerable heterogeneity across the compared datasets may be due to differential patient selection procedures across the included studies. The analyses were further complicated by the fact that imaging methods varied across the analyzed studies. A standardization of the acquisition protocol for DTI in schizophrenia studies would be helpful for alleviating this complication. Secondly, although DTI-derived radial and axial diffusivity are also informative regarding WM integrity, few articles report these data. Consequently, to enable statistical analysis, we focused only on FA data for the CC.
In conclusion, the present DTI findings provide a clear demonstration of a significant overall FA decrease in the WM of the CC in schizophrenic patients. Anatomically, the reduction appeared to be significant in both the genu region and the splenium region. Furthermore, the FA difference of the CC as a whole was pronounced in both chronic schizophrenia patients and those experiencing their first bout with schizophrenia. Further studies with more homogenous populations and standardized DTI protocols are needed to confirm and extend these findings.