[Skip to Navigation]
Sign In
Figure 1.  Causal Mediation and Structured Equation Modeling Analyses
Causal Mediation and Structured Equation Modeling Analyses

Analyses were performed to study the direct and mediated effects of fractional anisotropy (FA) on working memory (WM) via processing speed (PS). These analyses were performed separately in the full sample, patient, and control groups.

aIndicates statistically significant (P < .05) pathway branch.

Figure 2.  Cognition–White Matter Association
Cognition–White Matter Association

Associations are closely linked to white matter regional vulnerability to schizophrenia. Associations with processing speed (PS) are shown in circles; associations with working memory (WM) are shown in triangles. A, Association between correlation coefficients for 19 regional fractional anisotropy (FA)-PS (y-axis) and 19 regional FA effect sizes of schizophrenia from ENIGMA DTI work group (x-axis; Cohen d). B, Partial correlation coefficients of A after a correction for WM. C, Association between correlation coefficients for regional FA values and WM (y-axis) and regional FA effect sizes of schizophrenia (x-axis). D, Partial correlation coefficients of C after a correction for PS. This was tested separately in the full sample (A-D), patient (E-H), and control (I-L) groups. ACR indicates anterior corona radiata; BCC, body of corpus callosum; CGC, cingulum cortex; CST, corticospinal tract; EC, external capsule; FX, fornix; GCC, genu corpus callosum; IC, internal capsule; IFO, inferior fronto-occipital fasciculus; PCR, posterior corona radiata; PTR, posterior thalamic radiation; SCC, splenium corpus callosum; SCR, superior corona radiata; SFO, superior fronto-occipital fasciculus; SLF, superior longitudinal fasciculus; SS, sagittal stratum; and UNC, uncinate fasciculus.

Figure 3.  Mediation Effect of Processing Speed (PS) on Fractional Anisotropy (FA) to Working Memory (WM) for 19 Regional FA Values
Mediation Effect of Processing Speed (PS) on Fractional Anisotropy (FA) to Working Memory (WM) for 19 Regional FA Values

Values were plotted against the meta-analytical effect sizes (Cohen d) of schizophrenia on regional FA values developed by the ENIGMA DTI working group. Mediation effect (z score) values for regional white matter measurements were plotted against the meta-analytical effect sizes (Cohen d) of schizophrenia on regional FA values developed by the ENIGMA DTI working group. For other abbreviation expansions, see the caption to Figure 2.

Table 1.  Participant Demographics for 3 Cohorts
Participant Demographics for 3 Cohorts
Table 2.  Standardized Mediated, Direct, and Total Effects and Standardized Pathway Coefficients for Mediation Analyses Performed Separately in the Full Sample, Patient, and Control Groups
Standardized Mediated, Direct, and Total Effects and Standardized Pathway Coefficients for Mediation Analyses Performed Separately in the Full Sample, Patient, and Control Groups
1.
Reichenberg  A, Harvey  PD.  Neuropsychological impairments in schizophrenia: integration of performance-based and brain imaging findings.  Psychol Bull. 2007;133(5):833-858.PubMedGoogle ScholarCrossref
2.
Faraone  SV, Seidman  LJ, Kremen  WS, Toomey  R, Pepple  JR, Tsuang  MT.  Neuropsychologic functioning among the nonpsychotic relatives of schizophrenic patients: the effect of genetic loading.  Biol Psychiatry. 2000;48(2):120-126.PubMedGoogle ScholarCrossref
3.
Keefe  RS, Eesley  CE, Poe  MP.  Defining a cognitive function decrement in schizophrenia.  Biol Psychiatry. 2005;57(6):688-691.PubMedGoogle ScholarCrossref
4.
Keefe  RS, Goldberg  TE, Harvey  PD, Gold  JM, Poe  MP, Coughenour  L.  The brief assessment of cognition in schizophrenia: reliability, sensitivity, and comparison with a standard neurocognitive battery.  Schizophr Res. 2004;68(2-3):283-297.PubMedGoogle ScholarCrossref
5.
Dickinson  D, Ramsey  ME, Gold  JM.  Overlooking the obvious: a meta-analytic comparison of digit symbol coding tasks and other cognitive measures in schizophrenia.  Arch Gen Psychiatry. 2007;64(5):532-542.PubMedGoogle ScholarCrossref
6.
Knowles  EE, David  AS, Reichenberg  A.  Processing speed deficits in schizophrenia: reexamining the evidence.  Am J Psychiatry. 2010;167(7):828-835.PubMedGoogle ScholarCrossref
7.
Waxman  SG, Bennett  MV.  Relative conduction velocities of small myelinated and non-myelinated fibres in the central nervous system.  Nat New Biol. 1972;238(85):217-219.PubMedGoogle ScholarCrossref
8.
Felts  PA, Baker  TA, Smith  KJ.  Conduction in segmentally demyelinated mammalian central axons.  J Neurosci. 1997;17(19):7267-7277.PubMedGoogle Scholar
9.
Bartzokis  G.  Neuroglialpharmacology: myelination as a shared mechanism of action of psychotropic treatments.  Neuropharmacology. 2012;62(7):2137-2153.PubMedGoogle ScholarCrossref
10.
Bartzokis  G, Lu  PH, Tingus  K,  et al.  Lifespan trajectory of myelin integrity and maximum motor speed.  Neurobiol Aging. 2010;31(9):1554-1562.PubMedGoogle ScholarCrossref
11.
Salthouse  TA.  Aging and measures of processing speed.  Biol Psychol. 2000;54(1-3):35-54.PubMedGoogle ScholarCrossref
12.
Kochunov  P, Coyle  T, Lancaster  J,  et al.  Processing speed is correlated with cerebral health markers in the frontal lobes as quantified by neuroimaging.  Neuroimage. 2010;49(2):1190-1199.PubMedGoogle ScholarCrossref
13.
Penke  L, Muñoz Maniega  S, Murray  C,  et al.  A general factor of brain white matter integrity predicts information processing speed in healthy older people.  J Neurosci. 2010;30(22):7569-7574.PubMedGoogle ScholarCrossref
14.
Wright  SN, Hong  LE, Winkler  AM,  et al.  Perfusion shift from white to gray matter may account for processing speed deficits in schizophrenia.  Hum Brain Mapp. 2015;36(10):3793-3804.PubMedGoogle ScholarCrossref
15.
Nazeri  A, Chakravarty  MM, Felsky  D,  et al.  Alterations of superficial white matter in schizophrenia and relationship to cognitive performance.  Neuropsychopharmacology. 2013;38(10):1954-1962.PubMedGoogle ScholarCrossref
16.
Pérez-Iglesias  R, Tordesillas-Gutiérrez  D, McGuire  PK,  et al.  White matter integrity and cognitive impairment in first-episode psychosis.  Am J Psychiatry. 2010;167(4):451-458.PubMedGoogle ScholarCrossref
17.
Glahn  DC, Kent  JW  Jr, Sprooten  E,  et al.  Genetic basis of neurocognitive decline and reduced white-matter integrity in normal human brain aging.  Proc Natl Acad Sci U S A. 2013;110(47):19006-19011.PubMedGoogle ScholarCrossref
18.
Karbasforoushan  H, Duffy  B, Blackford  JU, Woodward  ND.  Processing speed impairment in schizophrenia is mediated by white matter integrity.  Psychol Med. 2015;45(1):109-120.PubMedGoogle ScholarCrossref
19.
Turken  A, Whitfield-Gabrieli  S, Bammer  R, Baldo  JV, Dronkers  NF, Gabrieli  JD.  Cognitive processing speed and the structure of white matter pathways: convergent evidence from normal variation and lesion studies.  Neuroimage. 2008;42(2):1032-1044.PubMedGoogle ScholarCrossref
20.
Alloza  C, Cox  SR, Duff  B,  et al.  Information processing speed mediates the relationship between white matter and general intelligence in schizophrenia.  Psychiatry Res. 2016;254(254):26-33.PubMedGoogle ScholarCrossref
21.
Salthouse  TA.  When does age-related cognitive decline begin?  Neurobiol Aging. 2009;30(4):507-514.PubMedGoogle ScholarCrossref
22.
Coyle  TR, Pillow  DR, Snyder  AC, Kochunov  P.  Processing speed mediates the development of general intelligence (g) in adolescence.  Psychol Sci. 2011;22(10):1265-1269.PubMedGoogle ScholarCrossref
23.
Fry  AF, Hale  S.  Relationships among processing speed, working memory, and fluid intelligence in children.  Biol Psychol. 2000;54(1-3):1-34.PubMedGoogle ScholarCrossref
24.
Karlsgodt  KH, Kochunov  P, Winkler  AM,  et al.  A multimodal assessment of the genetic control over working memory.  J Neurosci. 2010;30(24):8197-8202.PubMedGoogle ScholarCrossref
25.
Karlsgodt  KH, Niendam  TA, Bearden  CE, Cannon  TD.  White matter integrity and prediction of social and role functioning in subjects at ultra-high risk for psychosis.  Biol Psychiatry. 2009;66(6):562-569.PubMedGoogle ScholarCrossref
26.
Zeng  B, Ardekani  BA, Tang  Y,  et al.  Abnormal white matter microstructure in drug-naive first episode schizophrenia patients before and after eight weeks of antipsychotic treatment.  Schizophr Res. 2016;172(1-3):1-8.PubMedGoogle ScholarCrossref
27.
Kochunov  P, Ganjgahi  H, Winkler  A,  et al.  Heterochronicity of white matter development and aging explains regional patient control differences in schizophrenia.  Hum Brain Mapp. 2016;37(12):4673-4688.PubMedGoogle ScholarCrossref
28.
Kochunov  P, Thompson  PM, Winkler  A,  et al.  The common genetic influence over processing speed and white matter microstructure: evidence from the Old Order Amish and Human Connectome Projects.  Neuroimage. 2016;125:189-197.PubMedGoogle ScholarCrossref
29.
Kelly  S, Jahanshad  N, Zalesky  A,  et al.  Widespread white matter microstructural differences in schizophrenia across 4,375 individuals: results from the ENIGMA Schizophrenia DTI Working Group.  Mol Psychiatry. In press.Google Scholar
30.
Jahanshad  N, Kochunov  P, Sprooten  E,  et al.  Multi-site genetic analysis of diffusion images and voxelwise heritability analysis: a pilot project of the ENIGMA-DTI working group.  Neuroimage. 2013;81:455-469.PubMedGoogle ScholarCrossref
31.
van Erp  TG, Hibar  DP, Rasmussen  JM,  et al.  Subcortical brain volume abnormalities in 2028 individuals with schizophrenia and 2540 healthy controls via the ENIGMA consortium.  Mol Psychiatry. 2016;21(4):547-553.PubMedGoogle ScholarCrossref
32.
Kochunov  P, Jahanshad  N, Marcus  D,  et al.  Heritability of fractional anisotropy in human white matter: a comparison of Human Connectome Project and ENIGMA-DTI data.  Neuroimage. 2015;111:300-311.PubMedGoogle ScholarCrossref
33.
Weshsler  D.  Wechsler Adult Intelligence Scale. 3rd ed. San Antonio, TX: Psychological Corp; 1997.
34.
Woods  SW.  Chlorpromazine equivalent doses for the newer atypical antipsychotics.  J Clin Psychiatry. 2003;64(6):663-667.PubMedGoogle ScholarCrossref
35.
Kochunov  P, Jahanshad  N, Sprooten  E,  et al.  Multi-site study of additive genetic effects on fractional anisotropy of cerebral white matter: Comparing meta and megaanalytical approaches for data pooling.  Neuroimage. 2014;95:136-150.PubMedGoogle ScholarCrossref
36.
Tingley  D, Yamamoto  T, Hirose  K, Keele  L, Imai  K.  Mediation: R Package for Causal Mediation Analysis.  J Stat Softw. 2014;59(5):1-38. doi:10.18637/jss.v059.i05Google ScholarCrossref
37.
Imai  K, Keele  L, Tingley  D.  A general approach to causal mediation analysis.  Psychol Methods. 2010;15(4):309-334.PubMedGoogle ScholarCrossref
38.
R-Development-Core-Team. The R Project for statistical computing. http://www.R-project.org. Accessed June 2016.
39.
Wolf  EJ, Harrington  KM, Clark  SL, Miller  MW.  Sample size requirements for structural equation models: an evaluation of power, bias, and solution propriety.  Educ Psychol Meas. 2013;76(6):913-934.PubMedGoogle ScholarCrossref
40.
Fritz  MS, Mackinnon  DP.  Required sample size to detect the mediated effect.  Psychol Sci. 2007;18(3):233-239.PubMedGoogle ScholarCrossref
41.
Flechsig  P.  Developmental (myelogenetic) localisation of the cerebral cortex in the human subject.  Lancet. 1901;158(4077):1027-1030. doi:10.1016/S0140-6736(01)01429-5Google ScholarCrossref
42.
Lebel  C, Walker  L, Leemans  A, Phillips  L, Beaulieu  C.  Microstructural maturation of the human brain from childhood to adulthood.  Neuroimage. 2008;40(3):1044-1055.PubMedGoogle ScholarCrossref
43.
Grinberg  F, Maximov  II, Farrher  E,  et al.  Diffusion kurtosis metrics as biomarkers of microstructural development: A comparative study of a group of children and a group of adults.  Neuroimage. 2017;144(Pt A):12-22.PubMedGoogle ScholarCrossref
44.
Makinodan  M, Rosen  KM, Ito  S, Corfas  G.  A critical period for social experience-dependent oligodendrocyte maturation and myelination.  Science. 2012;337(6100):1357-1360.PubMedGoogle ScholarCrossref
45.
Liu  J, Dietz  K, DeLoyht  JM,  et al.  Impaired adult myelination in the prefrontal cortex of socially isolated mice.  Nat Neurosci. 2012;15(12):1621-1623.PubMedGoogle ScholarCrossref
46.
Kochunov  P, Hong  LE.  Neurodevelopmental and neurodegenerative models of schizophrenia: white matter at the center stage.  Schizophr Bull. 2014;40(4):721-728.PubMedGoogle ScholarCrossref
47.
Hennekens  CH, Hennekens  AR, Hollar  D, Casey  DE.  Schizophrenia and increased risks of cardiovascular disease.  Am Heart J. 2005;150(6):1115-1121.PubMedGoogle ScholarCrossref
48.
Kochunov  P, Glahn  DC, Rowland  LM,  et al.  Testing the hypothesis of accelerated cerebral white matter aging in schizophrenia and major depression.  Biol Psychiatry. 2013;73(5):482-491.PubMedGoogle ScholarCrossref
49.
Acheson  A, Wijtenburg  SA, Rowland  LM,  et al.  Combining diffusion tensor imaging and magnetic resonance spectroscopy to study reduced frontal white matter integrity in youths with family histories of substance use disorders.  Hum Brain Mapp. 2014;35(12):5877-5887.PubMedGoogle ScholarCrossref
50.
Acheson  A, Wijtenburg  SA, Rowland  LM,  et al.  Assessment of whole brain white matter integrity in youths and young adults with a family history of substance-use disorders.  Hum Brain Mapp. 2014;35(11):5401-5413.PubMedGoogle ScholarCrossref
Original Investigation
September 2017

Association of White Matter With Core Cognitive Deficits in Patients With Schizophrenia

Author Affiliations
  • 1Maryland Psychiatric Research Center, Department of Psychiatry, University of Maryland School of Medicine, Baltimore
  • 2Department of Psychology, The University of Texas at San Antonio
  • 3Imaging Genetics Center, Keck School of Medicine of the University of Southern California, Marina del Rey
  • 4Worldwide Research and Development, Pfizer Inc, Cambridge, Massachusetts
  • 5Biogen, Cambridge, Massachusetts
JAMA Psychiatry. 2017;74(9):958-966. doi:10.1001/jamapsychiatry.2017.2228
Key Points

Question  Is cerebral white matter associated with 2 core cognitive deficits (reduced information processing speed and impaired working memory) observed in schizophrenia?

Findings  In this cross-sectional study of 166 patients with schizophrenia and 213 healthy control individuals, the strength of the mediating relationships among white matter microstructure, processing speed, and working memory is associated with the vulnerability to schizophrenia in different white matter tracts.

Meaning  White matter deficits in schizophrenia may partially underwrite the 2 core neurocognitive deficits that contribute to functional disability in patients with schizophrenia.

Abstract

Importance  Efforts to remediate the multiple cognitive function impairments in schizophrenia should consider white matter as one of the underlying neural mechanisms.

Objective  To determine whether altered structural brain connectivity is responsible for 2 of the core cognitive deficits in schizophrenia— reduced information processing speed and impaired working memory.

Design, Setting, and Participants  This cross-sectional study design took place in outpatient clinics from August 1, 2004, to August 31, 2015. Participants included 166 patients with schizophrenia and 213 healthy control individuals. These participants were from 3 independent cohorts, each of which had its own healthy control group. No participant had current or past neurological conditions or major medical conditions. Patients were diagnosed with either schizophrenia or schizoaffective disorder as defined by the DSM-IV. Controls had no Axis I psychiatric disorder.

Main Outcomes and Measures  Mediation analyses and structural equation modeling were used to analyze the associations among processing speed, working memory, and white matter microstructures. Whole-brain and regional diffusion tensor imaging fractional anisotropy were used to measure white matter microstructures.

Results  Of the study participants, the 166 patients with schizophrenia had a mean (SD) age of 38.2 (13.3) years and the 213 healthy controls had a mean (SD) age of 39.2 (14.0) years. There were significantly more male patients than controls in each of the 3 cohorts (117 [70%] vs 91 [43%]), but there were no significant differences in sex composition among the 3 cohorts. Patients had significantly reduced processing speed (Cohen d = 1.24; P = 6.91 × 10−30) and working memory deficits (Cohen d = 0.83; P = 1.10 × 10−14) as well as a significant whole-brain fractional anisotropy deficit (Cohen d = 0.63; P = 2.20 × 10−9). In schizophrenia, working memory deficit was mostly accounted for by processing speed deficit, but this deficit remained when accounting for working memory (Cohen d = 0.89; P = 2.21 × 10−17). Mediation analyses showed a significant association pathway from fractional anisotropy to processing speed to working memory (P = 5.01 × 10−7). The strength of this brain-to-cognition pathway in different white matter tracts was strongly associated with the severity of schizophrenia-associated fractional anisotropy deficits in the corresponding white matter tracts as determined by a meta-analysis (r = 0.85-0.94; all P < .001). The same pattern was observed in patients and controls either jointly or independently.

Conclusions and Relevance  Study findings suggest that (1) processing speed contributes to the association between white matter microstructure and working memory in schizophrenia and (2) white matter impairment in schizophrenia is regional tract–specific, particularly in tracts normally supporting processing speed performance.

Introduction

Schizophrenia is characterized by severe cognitive deficits that contribute to functional disability in patients.1 Cognitive deficits, including reduced information processing speed and impaired working memory, are pervasive, precede clinical diagnosis, and form the core of schizophrenia-associated cognitive disabilities.2-6 Processing speed and working memory depend on large-scale, long-distance neural network operations that are supported by myelinated neuronal axonal fibers.7-10 Patients with schizophrenia show deficits in the microstructure of white matter in postmortem studies and in in vivo studies using diffusion tensor imaging (DTI) as indexed by fractional anisotropy (FA). The role of white matter microstructure for maintaining processing speed and other cognitive measures has been reported in healthy patients with psychiatric and neurological disorders.10-20 We hypothesized that white matter alterations mediate core cognitive deficits in schizophrenia in both processing speed and working memory.

Processing speed is a cognitive construct that may mediate other cognitive functions,10,21,22 and the mediation effects of processing speed have been observed in aging-related changes in working memory.22,23 Working memory deficits are also directly linked to white matter.15,24-26 However, whether processing speed and working memory deficits are separately or jointly associated with white matter abnormalities in schizophrenia, to date, is unknown. Here, we tested the first hypothesis: White matter abnormalities in schizophrenia underlie cognitive deficits as measured by processing speed and working memory tests; the effects of white matter abnormalities on processing speed are direct; and the working memory deficits are mediated via processing speed. This first hypothesis of a pathway from white matter to processing speed to working memory was tested using whole-brain mean white matter DTI values.

Schizophrenia effects on white matter as well as white matter effects on cognition are region specific.19,27,28 Therefore, the white matter–to–processing speed–to–working memory effect in schizophrenia, if present, is also likely to have regional specificity. To test this second hypothesis, we compared the white matter–to–processing speed–to–working memory mediation effect with known regional FA deficits in patients with schizophrenia. Regional FA deficits in schizophrenia were independently defined by Enhancing Imaging Genetics Meta-Analysis (ENIGMA) DTI, the largest schizophrenia-control meta-analysis of white matter DTI FA.29 The aim of the second hypothesis was to reaffirm the first hypothesis by further demonstrating the anatomic specificity of the white matter–to–processing speed–to–working memory pathway in schizophrenia.

A challenge for testing these 2 hypotheses was the need for a large sample on which to perform mediation analyses. We used a mega-analysis approach developed by the ENIGMA DTI working group30 to pool data from 3 cohorts that were collected at 1 site where the recruitment criteria, the neurocognitive and clinical assessments, and the inclusion of a control group were kept the same across the 3 cohorts. The mega-analysis method developed by the ENIGMA DTI working group28,31,32 allowed for data homogenization by creating a single aggregated sample, which yielded the needed sample size for testing our 2 hypotheses.

Methods
Study Participants

The study involved 166 patients with a mean (SD) age of 38.2 (13.3) years and 213 healthy control individuals with a mean (SD) age of 39.2 (14.0) years from 3 independent cohorts (Table 1). Each cohort had its own healthy control group. There were significantly more male patients than controls in each cohort (Table 1), but there were no significant differences in sex composition among the 3 cohorts (χ2 = 3.1; P = .10). More details about each cohort are in the eMethods in the Supplement. Magnetic resonance images were acquired over the past decade through studies using magnetic resonance imaging scanners (for cohorts A and B, 3-T Trio; Siemens; for cohort C, 3-T Allegra; Siemens); see the eMethods in the Supplement for more details. All participants with schizophrenia were evaluated for their capacity to provide informed consent. All participants gave written informed consent as approved by the University of Maryland, Baltimore, institutional review board, which also approved this study. The study was conducted in outpatient clinics from August 1, 2004, to August 31, 2015.

Uniform cognitive and clinical assessment and exclusion criteria were maintained across the cohorts. No participant had current or past neurological conditions or major medical conditions. Patients were diagnosed with either schizophrenia or schizoaffective disorder as defined by the DSM-IV. Controls had no Axis I psychiatric disorder. Processing speed was assessed with the digit symbol coding task of the Wechsler Adult Intelligence Scale, third edition.33 Meta-analyses of all cognitive domains affected by schizophrenia argued that speed of information processing showed the largest patient-control differences when measured by digit symbol coding.5,6 Working memory was assessed with the digit sequencing test.4 Note that processing speed and working memory are complex constructs and a single task may not capture all aspects. Patients also underwent a psychiatric interview using the Brief Psychiatric Rating Scale (BPRS) (score range: 20-140, with higher scores indicating worse symptoms). Their current medication dose was recorded and converted to chlorpromazine equivalent dose (CPZ).34 The CPZ was not significantly different among the 3 patient cohorts (χ2 = 1.66; P = .20).

Image Processing

The DTI data for all 3 cohorts were processed using the ENIGMA DTI analysis pipeline (https://www.nitrc.org/projects/enigma_dti).30 All data included in the analysis passed the ENIGMA DTI quality assurance or quality control procedures. The per-tract mean values were found by calculating the mean values along tract regions of interest in both hemispheres. Regional white matter FA measurements were generated for 19 tracts (eTable 1 in the Supplement). Whole-brain white matter mean FA values were found by calculating the mean values for the entire white matter skeleton. Further details of the mega-analysis homogenization procedures, developed by the ENIGMA DTI work group30,35 and implemented in the SOLAR Eclipse software (http://www.solar-eclipse-genetics.org) are provided elsewhere27 (eFigure 1 in the Supplement). This mega-analysis technique normalized neuroimaging data from each cohort to remove cohort- and scanner-related biases, including the effects of age,2 sex, and sex × age interactions.35 The outputs of this analysis were normalized values adhering to zero-mean normal distribution. The effect sizes obtained using mega-analysis were nearly identical (r > 0.99) to those obtained using a classic meta-analysis but provided about a 10% improvement in statistical power (eFigure 1 in the Supplement).27 Raw FA values for patients and controls in each cohort are provided in eTable 2 in the Supplement.

Statistical Analysis

Hypothesis testing was performed in 2 steps. First, we tested the overall group effects of whole-brain white matter mean FA, processing speed, and working memory values. We followed this step with mediation analyses to understand the relative effects of whole-brain white matter on processing speed and on working memory, including a structural equation modeling step to estimate the path coefficients of the mediation models. Second, we performed a correlation analysis between regional white matter FA values and processing speed and working memory. We followed this step with mediation analyses to test the hypothesis that processing speed mediated the effect of regional white matter FA values on working memory.

Mediation Effects on Working Memory by Processing Speed

The first hypothesis was tested using mediation analyses36,37 followed by structural equation modeling to estimate pathway coefficients. Fractional anisotropy was the predictor, working memory was the outcome, and processing speed was the mediator (Figure 1). This analysis was tested using whole-brain mean FA values for the full sample and then repeated in patients and controls separately. For the second hypothesis, the same mediation analyses were performed for the regional FA measurements to study the regional specificity of this association. The mediation analysis was performed with R38 (R Project) package mediation and 1000 permutations and bootstrapping to estimate the confidence interval.36 Structural equation modeling was performed with SPSS Amos, version 22 (IBM). Structural equation modeling and maximum likelihood estimation were used to estimate the standardized path coefficients.

Statistical significance was evaluated by unpaired, 2-tailed Sobel tests. Sample size requirement was estimated using simulation analyses,39,40 which showed that 85 participants were sufficient to achieve 80% power for the Sobel tests given the mediation model and effect sizes; thus, the pooled sample should have adequate power. If significant findings on whole-brain FA were identified, additional analysis would be performed on regional measurements. P = .05 was used to indicate statistical significance. For whole-brain white matter tracts, P = .05/19 = .003 was used to indicate statistical significance for regional white matter measurements. All tests were 2-tailed.

Mediation Effects on Schizophrenia-Related White Matter Regions

The ENIGMA DTI working group reported the largest DTI study (http://www.enigma-viewer.org)29 and provided patient-control comparison effect sizes across white matter regions in the brain (eTable 1 in the Supplement). The results provide a definitive and independent assessment of white matter regions most vulnerable to schizophrenia. If the proposed mediation path of white matter–to–processing speed–to–working memory were relevant to schizophrenia, we predicted that specific white matter regions most closely associated with schizophrenia would show the strongest mediation effect.

Clinical Correlates

Effects of symptoms and medication dose on whole-brain FA were examined: Symptom severity was expressed as BPRS total score (coefficient βBPRS), and medication dose was expressed as CPZ (coefficient βCPZ). Models included age and sex effects. Effects of smoking were analyzed with a binary variable 1 for smoker and 0 for nonsmoker.

Results
Whole-Brain FA and Cognitive Deficits

The patients had significantly lower whole-brain mean FA (Cohen d = 0.63; P = 2.20 × 10−9), processing speed (Cohen d = 1.24; P = 6.91 × 10−30) and working memory (Cohen d = 0.83; P = 1.10 × 10−14). Processing speed and working memory were significantly correlated (r = 0.44; P = 2.10 × 10−19), and both processing speed and working memory were correlated with FA (r = 0.34 and r = 0.28; both P < 10−8). Patient-control differences in working memory were diminished after adjusting for processing speed (P value changed from 1.1 × 10−14 to .03), with a significant reduction in Cohen d (from 0.83 to 0.24; P < .001). In comparison, when adjusting for working memory, patient-control differences in processing speed showed only marginally reduced effect sizes (from 1.2 to 0.89; P = .01).

Mediation Effect of Processing Speed on the Whole-Brain FA–Working Memory Association

Processing speed showed significant mediation effect on the association between white matter and working memory in the combined sample (Sobel z = 5.33; P = 5.01 × 10−7), where approximately 55% of the total (FA–working memory) effect was mediated by processing speed (Table 2 and Figure 1). In patients, the finding was similarly strong (Sobel z = 2.81; P = .007); 52% of the FA–working memory association was mediated by processing speed. In controls, the effect was in the same direction and approached significance (Sobel z = 1.51; P = .08), with 20% of the FA–working memory effect mediated by processing speed. Working memory had insignificant mediation effect on the FA–processing speed association in the full sample (Sobel z = 1.2; P = .36).

Regional FA and Cognitive Deficits

To define regions affected by schizophrenia, independently from the present study, we used the patient-control regional effect size of FA by the ENIGMA DTI29 working group. (The regional effect sizes from this megasample showed excellent correlation [r = 0.85] with the effect sizes from the ENIGMA DTI working group [eFigure 2 in the Supplement].) Significant correlations with processing speed were observed for all regions after correcting for 19 comparisons (r > 0.15; P < .003), with the exception of the corticospinal tract and inferior fronto-occipital tract, where the correlations were only nominally significant (r = 0.12; P = .02 vs r = 0.14; P = .006). A similar pattern was observed between regional FA and working memory. This analysis generated 19 correlation coefficients for regional FA–processing speed analyses and 19 correlation coefficients for regional FA–working memory analyses (eTable 3 in the Supplement).

The 19 ENIGMA DTI–based regional effect sizes of FA were significantly correlated with the regional FA–processing speed (r = 0.94; P < .001) (Figure 2A) and regional FA–working memory (r = 0.80; P < .001) correlation coefficients (Figure 2C). When controlling for working memory the ENIGMA DTI–based regional effect sizes remained significant in explaining regional FA–processing speed correlation coefficients (r = 0.86; P < .001) (Figure 2B) but not in explaining regional FA–working memory correlations when corrected for processing speed (r = 0.32; P = .31) (Figure 2D). We used permutation analysis (by permuting the difference on the partial correlation coefficients 1000 times) and found that the difference in these correlation coefficients was significant (P = .05).

Similar trends were observed in patients (Figure 2E-H) and controls (Figure 2I-L) separately. The ENIGMA DTI–based regional effect sizes were significantly associated with the regional FA–processing speed correlations, even after correcting for working memory for patients (r = 0.86; P = .001) and for controls (r = 0.59; P = .005). They were not significant for the FA–working memory correlations in patients (r = 0.42; P = .06) and controls (r = 0.31; P = .21) and were further diminished after correcting for processing speed (Figure 2).

Mediation Effect of Processing Speed on Regional FA–Working Memory Associations

The results for mediation analyses repeated in the full sample and in patients and controls separately are consistent with the results described earlier (eTables 4-6 in the Supplement). In the full sample, the processing speed–mediation effects on regional white matter FA to working memory were nominally significant in all white matter regions except for the corticospinal tract (eTable 4 in the Supplement). The pattern was similar in patients and controls (eTables 5-6 in the Supplement).

The regional z scores of the processing speed mediation power were significantly correlated with the ENIGMA DTI–based regional effect sizes of FA in the full sample (r = 0.94; Figure 3A). It was significant in patients (r = 0.72) and controls (r = 0.77) independently (P < .001; Figure 3B and C).

Clinical Correlates

We detected no significant association of BPRS scores or subscales of BPRS (thought disorder, withdrawal symptoms, anxiety or depression, activation, and hostility) with whole-brain FA (all r < 0.1; all P > .20). There were no significant effects of age-of-onset (r = 0.12; P = .33), duration of the illness (r = 0.08; P = .35), medication dosage (r = 0.05; P = .50), or smoking (r = 0.16; P = .15) on whole-brain FA values.

Discussion

This study advances in 3 ways our understanding of white matter contribution to the cognitive deficits observed in schizophrenia. First, impaired whole-brain white matter microstructure was found to significantly underwrite processing speed and working memory, which are two of the fundamental cognitive deficits in schizophrenia. Second, the FA–to–working memory effects were significantly mediated by processing speed, using whole brain or regional white matter analyses. Third, white matter regions with deficits observed in schizophrenia tended to have the strongest processing speed mediation effects in the pathway of FA, processing speed, and working memory. The overlap between independently determined regional effect sizes for schizophrenia and regional variances in the association between FA, processing speed, and working memory in both patients and controls passes the stringent test for the anatomic specificity for the processing speed mediation effect and its relevance to schizophrenia pathophysiological mechanism.

The neurobiological underpinnings of the FA, processing speed, and working memory association and its links to schizophrenia are complex and likely involve interactions between brain maturation and genetic and environmental etiologies for schizophrenia. White matter maturation is regionally heterochronic, with areas supporting motor and sensory function such as the corticospinal tract fully developing by the first decade of life.41 In contrast, the associative and commissural white matter responsible for higher-level cognitive functions, such as the anterior corona radiata and the genu of corpus callosum, continues to develop into the third and fourth decades of life.41-43 The prolonged development of associative white matter may stem from the need to support higher cognitive functions, including processing speed and working memory particularly in adolescence to early and middle adulthood.44,45

Schizophrenia is both a neurodevelopmental and a neurodegenerative disorder.46 It emerges in late adolescence and young adulthood and may lead to premature cerebral aging and shorter lifespan.47,48 The prolonged development of associative white matter regions may make young people (causally or consequentially) more vulnerable to the genetic and environmental risks for schizophrenia,27 which, in turn, may affect both regional associations of FA and processing speed (Figure 2) and FA, processing speed, and working memory (Figure 3) in schizophrenia patients. The associations are generally less robust in controls than in patients (Figure 2). Schizophrenia-related disease progress may affect white matter, processing speed, and working memory more profoundly than normal conditions might and may accentuate their association. The interpretation of these findings was aided by the replication of FA, processing speed, and working memory mediation in controls who neither had the disorder nor took antipsychotic medications. The regional pattern of the schizophrenia-associated white matter deficits was correlated with the association between FA and processing speed and working memory even in the controls, indicating that the processing speed effects in schizophrenia are likely driven by reduced white matter FA values that are not secondary effects of antipsychotic medications or a chronic disease status.

Mediation analysis was used to measure the strength of the white matter–to–processing speed–to–working memory pathway. The white matter tracts most vulnerable to schizophrenia (as determined by an independent sample) were also those that show the strongest FA–to–processing speed–to–working memory mediation effect. The results suggest that the hypothesized white matter–to–cognition mediation pathway effects have neuroanatomic specificity associated with schizophrenia white matter deficits. One could argue that the correlation between FA and processing speed is secondary to FA and processing speed deficits in schizophrenia and thus results in a strong but trivial correlation finding. This argument can be refuted given that this observation was also present in controls. Instead, the strong associations among specific white matter regions, processing speed, working memory, and schizophrenia should be viewed as shared vulnerabilities between schizophrenia and the FA–to–processing speed–to–working memory pathway. Analysis on regional FA supports this interpretation. For example, anterior corona radiata was the region showing the strongest correlations with both processing speed and working memory in healthy controls (Figure 2I and K). Anterior corona radiata contains associative fibers interconnecting frontal lobe areas that develop later in life and are thus likely related to cognition.28,49,50 In patients, the strongest association between regional FA and neurocognitive measures were observed for the body of the corpus callosum that contains interhemispheric motor and sensory fibers (Figure 2E and G). Speculatively, normal cognitive functions supported by anterior corona radiata are disrupted in patients, and alternative circuitry may take a more prominent role in cognitive performance. Other observations are more straightforward to interpret. For example, tracts not involved in working memory, such as the corticospinal tract, showed consistently low associations with neurocognitive measures regardless of diagnosis (Figure 2A and L).

Limitations

This study has limitations. The positive correlations among processing speed and working memory measurements limit how specific the interpretation of findings can be. However, while these tasks were correlated, our analyses suggested that their associations with the white matter were meaningfully different. We collected only 2 neurocognitive measurements that robustly separated patients with schizophrenia from healthy controls, and this approach may not capture all aspects of these complex constructs. The selection of the 2 tasks was based on the considerations of collecting a large sample of imaging and cognition data from a single center. Diffusion parameters besides FA (eg, axial, radial, and mean diffusivities) were not explored; FA was selected because it is more sensitive to schizophrenia deficits than are other parameters.29 The combined sample was not balanced on sex ratio, which may bias the findings, although no significant sex difference in FA was observed within patients or controls in any of the 3 cohorts or in the combined cohort. Repeating the mediation analysis replicated significant overall effect for the mean FA–to–processing speed–to–working memory model separately in men and women. Despite the insignificant correlations between the CPZ and FA (all P > .30), we cannot fully rule out effects from chronic antipsychotic exposure. Both processing speed and working memory are complex cognitive constructs involving extensive contributing factors and multiple alternative measurement approaches. Our approach may miss some aspects of working memory and processing speed as well as other cognitive constructs relevant to schizophrenia.

Conclusions

Findings from this study indicate that schizophrenia-related white matter deficits are a primary contributor to 2 core cognitive deficits associated with the disorder: reduced information processing speed and impaired working memory. The strong link between the schizophrenia-related effects on white matter and the cognition effects through specific white matter regions, even in normal controls, is remarkable, but the underlying mechanism is likely complex. Our hope is that these findings may elevate research interest in white matter–directed pharmacological interventions to enhance cognition for patients with schizophrenia.

Back to top
Article Information

Corresponding Author: Peter Kochunov, PhD, Maryland Psychiatric Research Center, Department of Psychiatry, University of Maryland School of Medicine, Tawes Court, Baltimore, MD 21228 (pkochunov@mprc.umaryland.edu).

Accepted for Publication: June 11, 2017.

Published Online: August 2, 2017. doi:10.1001/jamapsychiatry.2017.2228

Author Contributions: Drs Kochunov and Hong had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Kochunov, Hong.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Kochunov, Hong.

Critical revision of the manuscript for important intellectual content: Kochunov, Thompson, Hong, Coyle, Rowland.

Statistical analysis: Kochunov, Hong, Coyle, Shuo Chen.

Obtained funding: Hong, Kochunov, O’Donnell, Schubert.

Administrative, technical, or material support: All authors.

Study supervision: Kochunov, Hong.

Conflict of Interest Disclosures: Dr Hong reported receiving or planning to receive research funding or consulting fees from Mitsubishi, Your Energy Systems LLC, Neuralstem, Sound Pharma, Heptares, Taisho Pharmaceutical, and Pfizer. No other disclosures were reported.

Funding/Support: This research was supported in part by grants U01MH108148, 2R01EB015611, R01MH112180, R01DA027680, R01MH085646, P50MH103222, U54 EB020403, and T32MH067533 from the National Institutes of Health; by contract M00B6400091 from the State of Maryland; by grant 1620457 from the National Science Foundation; and by a research grant from Pfizer.

Role of the Funder/Sponsor: The funders provided support to enable the design and conduct of the study and collection, management, and analysis of the data. The funders had no role in the interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

References
1.
Reichenberg  A, Harvey  PD.  Neuropsychological impairments in schizophrenia: integration of performance-based and brain imaging findings.  Psychol Bull. 2007;133(5):833-858.PubMedGoogle ScholarCrossref
2.
Faraone  SV, Seidman  LJ, Kremen  WS, Toomey  R, Pepple  JR, Tsuang  MT.  Neuropsychologic functioning among the nonpsychotic relatives of schizophrenic patients: the effect of genetic loading.  Biol Psychiatry. 2000;48(2):120-126.PubMedGoogle ScholarCrossref
3.
Keefe  RS, Eesley  CE, Poe  MP.  Defining a cognitive function decrement in schizophrenia.  Biol Psychiatry. 2005;57(6):688-691.PubMedGoogle ScholarCrossref
4.
Keefe  RS, Goldberg  TE, Harvey  PD, Gold  JM, Poe  MP, Coughenour  L.  The brief assessment of cognition in schizophrenia: reliability, sensitivity, and comparison with a standard neurocognitive battery.  Schizophr Res. 2004;68(2-3):283-297.PubMedGoogle ScholarCrossref
5.
Dickinson  D, Ramsey  ME, Gold  JM.  Overlooking the obvious: a meta-analytic comparison of digit symbol coding tasks and other cognitive measures in schizophrenia.  Arch Gen Psychiatry. 2007;64(5):532-542.PubMedGoogle ScholarCrossref
6.
Knowles  EE, David  AS, Reichenberg  A.  Processing speed deficits in schizophrenia: reexamining the evidence.  Am J Psychiatry. 2010;167(7):828-835.PubMedGoogle ScholarCrossref
7.
Waxman  SG, Bennett  MV.  Relative conduction velocities of small myelinated and non-myelinated fibres in the central nervous system.  Nat New Biol. 1972;238(85):217-219.PubMedGoogle ScholarCrossref
8.
Felts  PA, Baker  TA, Smith  KJ.  Conduction in segmentally demyelinated mammalian central axons.  J Neurosci. 1997;17(19):7267-7277.PubMedGoogle Scholar
9.
Bartzokis  G.  Neuroglialpharmacology: myelination as a shared mechanism of action of psychotropic treatments.  Neuropharmacology. 2012;62(7):2137-2153.PubMedGoogle ScholarCrossref
10.
Bartzokis  G, Lu  PH, Tingus  K,  et al.  Lifespan trajectory of myelin integrity and maximum motor speed.  Neurobiol Aging. 2010;31(9):1554-1562.PubMedGoogle ScholarCrossref
11.
Salthouse  TA.  Aging and measures of processing speed.  Biol Psychol. 2000;54(1-3):35-54.PubMedGoogle ScholarCrossref
12.
Kochunov  P, Coyle  T, Lancaster  J,  et al.  Processing speed is correlated with cerebral health markers in the frontal lobes as quantified by neuroimaging.  Neuroimage. 2010;49(2):1190-1199.PubMedGoogle ScholarCrossref
13.
Penke  L, Muñoz Maniega  S, Murray  C,  et al.  A general factor of brain white matter integrity predicts information processing speed in healthy older people.  J Neurosci. 2010;30(22):7569-7574.PubMedGoogle ScholarCrossref
14.
Wright  SN, Hong  LE, Winkler  AM,  et al.  Perfusion shift from white to gray matter may account for processing speed deficits in schizophrenia.  Hum Brain Mapp. 2015;36(10):3793-3804.PubMedGoogle ScholarCrossref
15.
Nazeri  A, Chakravarty  MM, Felsky  D,  et al.  Alterations of superficial white matter in schizophrenia and relationship to cognitive performance.  Neuropsychopharmacology. 2013;38(10):1954-1962.PubMedGoogle ScholarCrossref
16.
Pérez-Iglesias  R, Tordesillas-Gutiérrez  D, McGuire  PK,  et al.  White matter integrity and cognitive impairment in first-episode psychosis.  Am J Psychiatry. 2010;167(4):451-458.PubMedGoogle ScholarCrossref
17.
Glahn  DC, Kent  JW  Jr, Sprooten  E,  et al.  Genetic basis of neurocognitive decline and reduced white-matter integrity in normal human brain aging.  Proc Natl Acad Sci U S A. 2013;110(47):19006-19011.PubMedGoogle ScholarCrossref
18.
Karbasforoushan  H, Duffy  B, Blackford  JU, Woodward  ND.  Processing speed impairment in schizophrenia is mediated by white matter integrity.  Psychol Med. 2015;45(1):109-120.PubMedGoogle ScholarCrossref
19.
Turken  A, Whitfield-Gabrieli  S, Bammer  R, Baldo  JV, Dronkers  NF, Gabrieli  JD.  Cognitive processing speed and the structure of white matter pathways: convergent evidence from normal variation and lesion studies.  Neuroimage. 2008;42(2):1032-1044.PubMedGoogle ScholarCrossref
20.
Alloza  C, Cox  SR, Duff  B,  et al.  Information processing speed mediates the relationship between white matter and general intelligence in schizophrenia.  Psychiatry Res. 2016;254(254):26-33.PubMedGoogle ScholarCrossref
21.
Salthouse  TA.  When does age-related cognitive decline begin?  Neurobiol Aging. 2009;30(4):507-514.PubMedGoogle ScholarCrossref
22.
Coyle  TR, Pillow  DR, Snyder  AC, Kochunov  P.  Processing speed mediates the development of general intelligence (g) in adolescence.  Psychol Sci. 2011;22(10):1265-1269.PubMedGoogle ScholarCrossref
23.
Fry  AF, Hale  S.  Relationships among processing speed, working memory, and fluid intelligence in children.  Biol Psychol. 2000;54(1-3):1-34.PubMedGoogle ScholarCrossref
24.
Karlsgodt  KH, Kochunov  P, Winkler  AM,  et al.  A multimodal assessment of the genetic control over working memory.  J Neurosci. 2010;30(24):8197-8202.PubMedGoogle ScholarCrossref
25.
Karlsgodt  KH, Niendam  TA, Bearden  CE, Cannon  TD.  White matter integrity and prediction of social and role functioning in subjects at ultra-high risk for psychosis.  Biol Psychiatry. 2009;66(6):562-569.PubMedGoogle ScholarCrossref
26.
Zeng  B, Ardekani  BA, Tang  Y,  et al.  Abnormal white matter microstructure in drug-naive first episode schizophrenia patients before and after eight weeks of antipsychotic treatment.  Schizophr Res. 2016;172(1-3):1-8.PubMedGoogle ScholarCrossref
27.
Kochunov  P, Ganjgahi  H, Winkler  A,  et al.  Heterochronicity of white matter development and aging explains regional patient control differences in schizophrenia.  Hum Brain Mapp. 2016;37(12):4673-4688.PubMedGoogle ScholarCrossref
28.
Kochunov  P, Thompson  PM, Winkler  A,  et al.  The common genetic influence over processing speed and white matter microstructure: evidence from the Old Order Amish and Human Connectome Projects.  Neuroimage. 2016;125:189-197.PubMedGoogle ScholarCrossref
29.
Kelly  S, Jahanshad  N, Zalesky  A,  et al.  Widespread white matter microstructural differences in schizophrenia across 4,375 individuals: results from the ENIGMA Schizophrenia DTI Working Group.  Mol Psychiatry. In press.Google Scholar
30.
Jahanshad  N, Kochunov  P, Sprooten  E,  et al.  Multi-site genetic analysis of diffusion images and voxelwise heritability analysis: a pilot project of the ENIGMA-DTI working group.  Neuroimage. 2013;81:455-469.PubMedGoogle ScholarCrossref
31.
van Erp  TG, Hibar  DP, Rasmussen  JM,  et al.  Subcortical brain volume abnormalities in 2028 individuals with schizophrenia and 2540 healthy controls via the ENIGMA consortium.  Mol Psychiatry. 2016;21(4):547-553.PubMedGoogle ScholarCrossref
32.
Kochunov  P, Jahanshad  N, Marcus  D,  et al.  Heritability of fractional anisotropy in human white matter: a comparison of Human Connectome Project and ENIGMA-DTI data.  Neuroimage. 2015;111:300-311.PubMedGoogle ScholarCrossref
33.
Weshsler  D.  Wechsler Adult Intelligence Scale. 3rd ed. San Antonio, TX: Psychological Corp; 1997.
34.
Woods  SW.  Chlorpromazine equivalent doses for the newer atypical antipsychotics.  J Clin Psychiatry. 2003;64(6):663-667.PubMedGoogle ScholarCrossref
35.
Kochunov  P, Jahanshad  N, Sprooten  E,  et al.  Multi-site study of additive genetic effects on fractional anisotropy of cerebral white matter: Comparing meta and megaanalytical approaches for data pooling.  Neuroimage. 2014;95:136-150.PubMedGoogle ScholarCrossref
36.
Tingley  D, Yamamoto  T, Hirose  K, Keele  L, Imai  K.  Mediation: R Package for Causal Mediation Analysis.  J Stat Softw. 2014;59(5):1-38. doi:10.18637/jss.v059.i05Google ScholarCrossref
37.
Imai  K, Keele  L, Tingley  D.  A general approach to causal mediation analysis.  Psychol Methods. 2010;15(4):309-334.PubMedGoogle ScholarCrossref
38.
R-Development-Core-Team. The R Project for statistical computing. http://www.R-project.org. Accessed June 2016.
39.
Wolf  EJ, Harrington  KM, Clark  SL, Miller  MW.  Sample size requirements for structural equation models: an evaluation of power, bias, and solution propriety.  Educ Psychol Meas. 2013;76(6):913-934.PubMedGoogle ScholarCrossref
40.
Fritz  MS, Mackinnon  DP.  Required sample size to detect the mediated effect.  Psychol Sci. 2007;18(3):233-239.PubMedGoogle ScholarCrossref
41.
Flechsig  P.  Developmental (myelogenetic) localisation of the cerebral cortex in the human subject.  Lancet. 1901;158(4077):1027-1030. doi:10.1016/S0140-6736(01)01429-5Google ScholarCrossref
42.
Lebel  C, Walker  L, Leemans  A, Phillips  L, Beaulieu  C.  Microstructural maturation of the human brain from childhood to adulthood.  Neuroimage. 2008;40(3):1044-1055.PubMedGoogle ScholarCrossref
43.
Grinberg  F, Maximov  II, Farrher  E,  et al.  Diffusion kurtosis metrics as biomarkers of microstructural development: A comparative study of a group of children and a group of adults.  Neuroimage. 2017;144(Pt A):12-22.PubMedGoogle ScholarCrossref
44.
Makinodan  M, Rosen  KM, Ito  S, Corfas  G.  A critical period for social experience-dependent oligodendrocyte maturation and myelination.  Science. 2012;337(6100):1357-1360.PubMedGoogle ScholarCrossref
45.
Liu  J, Dietz  K, DeLoyht  JM,  et al.  Impaired adult myelination in the prefrontal cortex of socially isolated mice.  Nat Neurosci. 2012;15(12):1621-1623.PubMedGoogle ScholarCrossref
46.
Kochunov  P, Hong  LE.  Neurodevelopmental and neurodegenerative models of schizophrenia: white matter at the center stage.  Schizophr Bull. 2014;40(4):721-728.PubMedGoogle ScholarCrossref
47.
Hennekens  CH, Hennekens  AR, Hollar  D, Casey  DE.  Schizophrenia and increased risks of cardiovascular disease.  Am Heart J. 2005;150(6):1115-1121.PubMedGoogle ScholarCrossref
48.
Kochunov  P, Glahn  DC, Rowland  LM,  et al.  Testing the hypothesis of accelerated cerebral white matter aging in schizophrenia and major depression.  Biol Psychiatry. 2013;73(5):482-491.PubMedGoogle ScholarCrossref
49.
Acheson  A, Wijtenburg  SA, Rowland  LM,  et al.  Combining diffusion tensor imaging and magnetic resonance spectroscopy to study reduced frontal white matter integrity in youths with family histories of substance use disorders.  Hum Brain Mapp. 2014;35(12):5877-5887.PubMedGoogle ScholarCrossref
50.
Acheson  A, Wijtenburg  SA, Rowland  LM,  et al.  Assessment of whole brain white matter integrity in youths and young adults with a family history of substance-use disorders.  Hum Brain Mapp. 2014;35(11):5401-5413.PubMedGoogle ScholarCrossref
×