Neuroanatomy of Patients with Deficit Schizophrenia: An Exploratory Quantitative Meta-Analysis of Structural Neuroimaging Studies

Little is known regarding the neuroanatomical correlates of patients with deficit schizophrenia or persistent negative symptoms. In this meta-analysis, we aimed to determine whether patients with deficit schizophrenia have characteristic brain abnormalities. We searched PubMed, CINAHL and Ovid to identify studies that examined the various regions of interest amongst patients with deficit schizophrenia, patients with non-deficit schizophrenia and healthy controls. A total of 24 studies met our inclusion criteria. A random-effects model was used to calculate a combination of outcome measures, and heterogeneity was assessed by the I2 statistic and Cochran’s Q statistic. Our findings suggested that there was statistically significant reduction in grey matter volume (−0.433, 95% confidence interval (CI): −0.853 to −0.014, p = 0.043) and white matter volume (−0.319, 95% CI: −0.619 to −0.018, p = 0.038) in patients with deficit schizophrenia compared to healthy controls. There is also statistically significant reduction in total brain volume (−0.212, 95% CI: −0.384 to −0.041, p = 0.015) and white matter volume (−0.283, 95% CI: −0.546 to −0.021, p = 0.034) in patients with non-deficit schizophrenia compared to healthy controls. Between patients with deficit and non-deficit schizophrenia, there were no statistically significant differences in volumetric findings across the various regions of interest.


Introduction
The heterogeneity of schizophrenia has long captured the interest of researchers and clinicians alike. Considerable neuroanatomical, neurobiological and neuropsychological research has gone into discriminating between potential subtypes of schizophrenia characterized by the prevalence of symptom domains. In particular, negative symptoms, which may present as a deficit in goal-directed or pleasurable activity, speech and non-verbal expression [1], have been the source of some discussion, with Carpenter et al. [2] proposing the term deficit schizophrenia (DS) to describe the presence of primary and persistent negative symptoms [2].

Data Extraction and Quality Assessment
Once a finalized list of relevant studies had been generated via the database search, initial screening of titles and abstracts was undertaken using a data collection and eligibility checklist sheet (see Appendix A Table A1) to decide which full papers should be included. Following verbal consensus on study inclusion, full-text articles were then collected and the data were extracted and compiled into a series of Excel spreadsheets for both systematic review and meta-analytic consideration. A database for demographic details and ROI examined by each individual study was created (see Appendix A Table A2) to enable gathering information on the number of studies that had examined a specific region of interest. Due to the variability in definitions of specific region of interest, specific quotes from the study outlining the region of interest examined were input into the database and compared. This process ensured that studies were accurately matched for specific region of interests, to prevent over-sampling error and bias. Regions of interests that were examined by more than one study were recruited into the review and statistical analysis. The mean and standard deviation values were then sought out and recorded into an Excel database (see Appendix A Table A3) for meta-analytic considerations. The following data were collected from the studies: the neuroimaging modality employed, i.e., magnetic resonance imaging-region of interest (ROI), voxel-based morphometry (VBM) or diffusion tensor imaging (DTI); the diagnostic instrument used to define deficit schizophrenia, i.e., the Schedule of Deficit Syndrome (SDS), the Persistent Negative Symptoms classification (PNS) or Scale for Assessment of Negative Symptoms (SANS)-although the Schedule of Deficit Syndrome is suggested as the gold standard in diagnosing deficit schizophrenia, studies that used certain proxy diagnostic instruments such as the Scale for Assessment of Negative Symptoms and Persistent Negative Symptoms to diagnose deficit schizophrenia were also accepted in order to increase the number of acceptable studies included in the meta-analysis; the number of deficit schizophrenia patients, non-deficit schizophrenia patients and healthy controls; the number and ratio of males to females in each study; the mean age of each group of DS, NDS and control patients in each study. For studies to be used in meta-analysis, we recorded all the mean and standard deviation values for the matched regions of interests.

Statistical Analyses
Statistical analyses were conducted while using the Comprehensive Meta-Analysis Version (CMA) 3.0 program. A random-effects model was adopted to calculate the continuous outcome measures from chosen studies and 95% confidence intervals (CIs) in view of the expected heterogeneity across the studies. Standard mean differences (SMD) were measured and referred to the Cohen's effect size. Regions of interests with more than one study investigating this particular brain structure were included in statistical evaluation as long as suitable diagnostic instruments were implemented and continuous outcome measurements of means and standard deviation were recorded. The between-study heterogeneity was assessed by calculating the Cochran Q test statistic [18]. To assist with interpretation of between-study heterogeneity, the I 2 statistic was also calculated. The I 2 statistic was equivalent to the proportion of total variation across studies due to heterogeneity [19].
Of the identified region of interests, four brain structures have been examined by three or more independent studies with continuous quantitative data. A total of 12 meta-analytic comparisons took place between the deficit schizophrenia patient group, the non-deficit schizophrenia patient group and the healthy control group.
The demographic data of these 24 studies were entered into a database . The variables of which are summarized in Table 1.
The mean age of patients with deficit schizophrenia ranged from 22.33 to 49.03 years. There was a mean of 23.4 deficit schizophrenia patients, 36.3 non-deficit schizophrenia patients and 47.9 healthy controls per study. This low deficit schizophrenia patient sample size is noted and may suggest the actual lower clinical sample prevalence. It may also indicate a sense of difficulty in diagnosing patients with deficit schizophrenia.
Of the identified region of interests, four brain structures have been examined by three or more independent studies with continuous quantitative data. A total of 12 meta-analytic comparisons took place between the deficit schizophrenia patient group, the non-deficit schizophrenia patient group and the healthy control group.
The demographic data of these 24 studies were entered into a database . The variables of which are summarized in Table 1.
The mean age of patients with deficit schizophrenia ranged from 22.33 to 49.03 years. There was a mean of 23.4 deficit schizophrenia patients, 36.3 non-deficit schizophrenia patients and 47.9 healthy controls per study. This low deficit schizophrenia patient sample size is noted and may suggest the actual lower clinical sample prevalence. It may also indicate a sense of difficulty in diagnosing patients with deficit schizophrenia. In the deficit schizophrenia patient group, the percentage of males was 77%. In all, 23 studies included both males and females, and there was only one paper that comprised of only male patients [29]. Roughly four-fifths of DS subjects were men, suggesting that males are much more commonly diagnosed with deficit schizophrenia than females are.

Comparing Patients with Deficit Schizophrenia to Healthy Controls
Comparisons between the deficit schizophrenia patient group and the healthy controls across the four regions of interest were made and summarized in Table 2. The effect sizes of grey matter and white matter volumes in deficit schizophrenia compared against healthy controls (in bold) were statistically significantly smaller (effect size p-value less than 0.05). Abbreviations: DS = deficit schizophrenia; NDS = non-deficit schizophrenia; NR = not reported; SDS = Schedule for Deficit Schizophrenia; PANSS = Positive and Negative Syndrome Scale; PNS = Persistent Negative Symptoms (PNS) classification; SCOS = Strauss-Carpenter Outcome Scale; PDS = Proxy for the Deficit Syndrome; MRI = stereotaxy-based regional brain volumetry applied to segmented MRI.

Comparing Patients with Deficit Schizophrenia to Healthy Controls
Comparisons between the deficit schizophrenia patient group and the healthy controls across the four regions of interest were made and summarized in Table 2. The effect sizes of grey matter and white matter volumes in deficit schizophrenia compared against healthy controls (in bold) were statistically significantly smaller (effect size p-value less than 0.05).   according to five studies [21][22][23]25,26].
In this particular comparison, there were four different diagnostic instruments used among the five studies. Two studies utilized the SDS [22,25], one study employed PNS [23], one study used PANSS [26] and one study used SANS [21] to define deficit schizophrenia in their patient group. This heterogeneous diagnostic process may affect inter-rater reliability, especially in the three studies that did not use SDS. Compared with controls, patients with deficit schizophrenia had statistically significant smaller white matter volume, with a random effect size of −0.319 (95% CI: −0.619 to −0.018, p = 0.038), according to four studies [21][22][23]26].
Similar to the previous comparison, the diagnostic instruments used in all four studies were different to each other. One study utilized the SDS [22], one study employed PNS [23], one study used PANSS [26] and one study used SANS [21] to define deficit schizophrenia in their patient group. This heterogeneous diagnostic process may affect inter-rater reliability, especially in the three studies that did not use SDS.

Comparing Patients with Deficit Schizophrenia to Patients with Non-Deficit Schizophrenia
Comparisons between the deficit schizophrenia patient group and the non-deficit schizophrenia patient group across the four ROIs were made and summarized in Table 3. There appear to be no statistically significant differences in the effect sizes across the four regions of interest between patient groups of DS and NDS. As a result, we are not able to make any conclusions about the brain structural correlation changes between these two patient groups. Compared with controls, patients with deficit schizophrenia had statistically significant smaller grey matter volumes, with a random effect size of −0.433 (95% CI: −0.853 to −0.014, p = 0.043), according to five studies [21][22][23]25,26].
In this particular comparison, there were four different diagnostic instruments used among the five studies. Two studies utilized the SDS [22,25], one study employed PNS [23], one study used PANSS [26] and one study used SANS [21] to define deficit schizophrenia in their patient group. This heterogeneous diagnostic process may affect inter-rater reliability, especially in the three studies that did not use SDS.
Similar to the previous comparison, the diagnostic instruments used in all four studies were different to each other. One study utilized the SDS [22], one study employed PNS [23], one study used PANSS [26] and one study used SANS [21] to define deficit schizophrenia in their patient group. This heterogeneous diagnostic process may affect inter-rater reliability, especially in the three studies that did not use SDS.

Comparing Patients with Deficit Schizophrenia to Patients with Non-Deficit Schizophrenia
Comparisons between the deficit schizophrenia patient group and the non-deficit schizophrenia patient group across the four ROIs were made and summarized in Table 3. There appear to be no statistically significant differences in the effect sizes across the four regions of interest between patient groups of DS and NDS. As a result, we are not able to make any conclusions about the brain structural correlation changes between these two patient groups.

Comparing Patients with Non-Deficit Schizophrenia to Healthy Controls
Comparisons between the non-deficit schizophrenia patient group and the healthy controls across the four regions of interest were made and summarized in Table 4. There were statistically significant findings in the total brain volume and white matter volume during comparison between patients with NDS versus healthy controls. Graphical representations of the statistically significant comparisons are plotted on the Forest plots in Figures 4 and 5. Abbreviations: DS, deficit schizophrenia; NDS, non deficit schizophrenia; CI, confidence interval, TBV, total brain volume; GM, gray matter; WM, white matter; CSF, cerebrospinal fluid.

Comparing Patients with Non-Deficit Schizophrenia to Healthy Controls
Comparisons between the non-deficit schizophrenia patient group and the healthy controls across the four regions of interest were made and summarized in Table 4. There were statistically significant findings in the total brain volume and white matter volume during comparison between patients with NDS versus healthy controls. Graphical representations of the statistically significant comparisons are plotted on the Forest plots in Figures 4 and 5.   Compared with controls, patients with non-deficit schizophrenia had statistically smaller total brain volume, with an effect size of −0.212 (95% CI: −0.384 to −0.041, p = 0.015), according to seven studies [20][21][22][23][24][25]27].  Compared with controls, patients with non-deficit schizophrenia had statistically smaller total brain volume, with an effect size of −0.212 (95% CI: −0.384 to −0.041, p = 0.015), according to seven studies [20][21][22][23][24][25]27].

Deficit Schizophrenia versus Healthy Controls
In patients with deficit schizophrenia compared with healthy controls, we identified statistically significant reduced grey matter volume and reduced white matter volume.

Deficit Schizophrenia versus Non-Deficit Schizophrenia
In patients with deficit schizophrenia compared with those with non-deficit schizophrenia, there appeared to be no statistically significant differences in the effect sizes across the four brain regions investigated.

Non-Deficit Schizophrenia versus Healthy Control
In patients with non-deficit schizophrenia compared with healthy controls, we identified reduced total brain volume and decreased white matter volume.

Strengths and Limitations
The main strength of this study is that it is the first study to attempt to examine brain structural correlates in patients with deficit schizophrenia using a meta-analytic approach. With limited numbers of relevant studies so far, it is particularly important to ensure all related studies are considered. A methodical systematic approach to include all relevant studies was undertaken and achieved using a thorough and comprehensive search strategy.
Despite the strength of inclusion of relevant papers, the study has a number of significant limitations that should be taken into account prior to serious interpretation of the study findings. Regarding study design limitations, it became apparent during the data collection phase that the number of available and relevant neuroimaging studies that specifically addressed questions about the neuroanatomy of patients with deficit schizophrenia is relative scarce. For instance, the recent literature search revealed 24 studies relevant to deficit schizophrenia, whereas the systematic review study in 2001 by Shenton et al. [44] produced 180 studies. The sample in the Shenton et al. study [44] was mostly patients with chronic schizophrenia. The existing studies of patients with deficit schizophrenia tended to have a smaller patient sample. The existing average of 23.4 patients in this study is almost one third lower than the average of 33 patients per study reported in the systematic review by Shenton et al., 2001 [44]. In our meta-analysis, one out of the four region-of-interest comparisons that suggested statistical significance have three studies' sample size. The other region-of-interest comparisons have between four and seven studies. The low number of studies, which translates to a small patient sample, per brain structure evaluated reduces the power of the analysis. An inadvertent limitation due to the small number of studies included in the meta-analysis would be that it is not possible to determine for publication bias, which may occur. The current lack of consensus among comparisons between studies studying the same region of interest is likely to reflect the generally low power of studies. In addition, for the four meta-analyses, the p values ranged from 0.015 to 0.043, and they would probably not be statistically significant if they were adjusted for multiple comparison.
Most brain volumetric studies included in this systematic review employed a region-of-interest approach (13 out of 24 studies). In this ROI approach, brain regions are outlined in an exacting manner, using pre-set operationalized procedures [45]. Due to this precise nature, it can create errors, because other region-of-interest outlines may not fulfil the specific description in another study. This error become magnified when a large volume of other similar but not exactly precise ROI were gathered together, in the case of a systematic review of many studies. In this present study, the specific description of brain regions of interest can differ between different authors and their papers. The lack of cohesion in describing the regions of interest measured between studies make direct comparisons of reported outcome measures difficult. This issue is compounded when authors use dissimilar labels to describe the same brain area [46,47]. Voxel-based morphometry studies are theoretically more favorable for meta-analytic processing. As group differences are described in standardized coordinates, meta-analytic techniques can be applied effectively. Although our search uncovered seven VBM studies, we were not able to utilize them for quantitative analysis.
There were also clinical limitations faced in this study. Deficit schizophrenia is described as "a set of primary, enduring negative symptoms of schizophrenia". However, there often exists a complicated heterogeneity between primary and secondary symptoms of negative schizophrenia. The mean age of the patient group with deficit schizophrenia in this study was 33.4 years old. However, the age group ranged from 24 to 40 years old. This wide age range may introduce confounding factors that may affect the accuracy of diagnosis of deficit schizophrenia. For example, an older patient with deficit schizophrenia is more likely to develop negative symptoms secondary to the use of antipsychotics or become affected by psychosocial circumstances. The diagnostic instrument used in defining deficit schizophrenia has not been singularly standardized. Experts differ in their opinions regarding these scales. Some recommend the Schedule of Deficit Schizophrenia (SDS) as the current gold standard for diagnosing deficit schizophrenia. However, only 14 of the 24 studies (58.3%) in our review used the Schedule of Deficit Schizophrenia as a diagnostic tool. In the comparisons involving patients with deficit schizophrenia and healthy controls, there was heterogeneity in the diagnostic instrument used to diagnose patients with deficit schizophrenia, thereby affecting inter-rater reliability, especially in studies that did not use SDS. For patients with schizophrenia, both the deficit and the non-deficit form, one typical scenario is that they will be rapidly started on some form of neuroleptic medication soon after diagnosis. Different types and dosages of medications will be prescribed, presenting with significant treatment heterogeneity. Questions should be asked about the timing as well as the cause of brain volume changes, particularly in studies that show statistically significant findings. Volumetric changes occurring for reasons other than those related to the pathophysiology of deficit schizophrenia are likely to cause Type 1 errors or false-positive outcomes. Older patients are more likely to develop volumetric changes due to secondary causes of negative symptoms (for example, antipsychotic medications). In younger patients with true deficit schizophrenia, the rate of volumetric loss may be insufficient for detection by either the MRI or during analytical cutoffs in this study.
Lastly, we also encountered imaging limitations in this study. Since this study involved only the MRI modality, it is important to discuss potential pitfalls and difficulties with the use of MRI volumetric measuring methods [48]. Different MRI software or machinery operation can lead to volumetric changes of up to 5% [49]. Calculation errors, both manual and computerized, occurring during neuroimaging processing can average approximately 1.5%, and even though this inaccuracy can be adjusted for, neglecting its adjustment can lead to systematic error, and ultimately reduce the level of agreement amongst the various studies. In MRI studies employing voxel-based morphometry, imprecision due to the misclassification of voxels occurring during brain segmentation is one of the more common causes of imaging error [50]. Grey matter proximity with cerebrospinal fluid can lead to a poorly defined edge and cause volume estimation errors [50]. Poor positioning of the head or of the imaging slab can cause inaccuracy in brain volume measurements [48]. The only way to resolve this issue fully is to aim for full-brain coverage during an examination. One of the last but important imaging limitations likely to be encountered in MRI volumetric measurements is random mistakes or miscalculations, that are often out of the control of the technician. These are non-systemic errors [51] and may be significant.

Conclusions
The most statistically significant volumetric findings in our study of patients suggest that compared with healthy normal controls, patients with deficit schizophrenia have reduced grey and white matter volumes (Table 2), while patients with non-deficit schizophrenia have reduced total brain volume and white matter volume (Table 4). Between patients with deficit and non-deficit schizophrenia, there were no statistically significant differences in volumetric findings across the four brain regions (Table 3).
However, these observed measure outcomes of brain structural changes should not be conclusive due to significant limitations on the study design, particularly in the areas of small sample sizes and limited studies examining the neuroanatomy of deficit schizophrenia. Inconsistencies of imaging technique and the likelihood of a less homogeneous patient sample also contribute to this caution.
This review is an exploratory first-investigation into this topic. It re-affirms the need for further research into the neuroanatomy of deficit schizophrenia. Perhaps with the relatively low level of involvement so far, this is an area of promise.
However, the traditional complexities and barriers that turn away prospective researchers needed to be addressed first: The first lies in diagnosing deficit schizophrenia in the patient. A gold standard diagnostic instrument, currently the Schedule of Deficient Syndrome (SDS), should be used whenever possible because it has the highest level of inter-rater reliability, which ultimately aids research and subsequent reviews. The process of using the SDS is tedious, but the rewards would be worthwhile. The second lies in meticulous study design to improve the power of the study and minimize confounders. Recruitment of larger independent samples and careful sampling criteria to focus on a more homogeneous group of patients with primary negative symptoms by controlling risk factors for secondary negative symptoms such as old age, long duration of mental illness, antipsychotic medications, etc. should also be employed. Funding: This publication fee was supported by academic grant from the National University of Singapore.

Conflicts of Interest:
The authors declare no conflict of interest.