A MULTICENTER HARMONIZED DIFFUSION TENSOR IMAGING STUDY ON THE ASSOCIATION OF WHITE MATTER STRUCTURE AND CLINICAL FUNCTIONING
Johanna Seitz,Suheyla Cetin Karayumak,Joanne Wojcik,Amanda Lyall,James Levitt,Ofer Pasternak,Madhura Baxi,Sinead Kelly,John Sweeney,Petra Viher,Katharina Stegmayer,Sebastian Walther,Jungsun Lee,Timothy J. Crow,Anthony James,Aristotle Voineskos,Robert Buchanan,Philip Szeszko,Anil Malhotra,Martha Shenton,Yogesh Rathi,Matcheri Keshavan,Marek Kubicki
DOI: https://doi.org/10.1093/SCHBUL/SBAA031.223
2020-01-01
Schizophrenia Bulletin
Abstract:Abstract Background The association of white matter (WM) abnormalities with clinical variables in schizophrenia (SCZ) is poorly understood. We investigated the clinical correlates of WM impairments using imaging data of 597 patients with SCZ and 490 healthy controls (HC). We focused on lifelong changes of WM (measured by Fractional Anisotropy [FA]) in SCZ and compared it to that of HC. We investigated how age, duration of illness, and medication influence WM trajectories, and examined how structural impairments are related to symptoms and cognition. Last, we tested for the role of sex in structure-function interactions. Methods Diffusion-weighted images and clinical measurements were collected as part of 13 independent studies, and data was harmonized across all sites. We registered images to the IIT Human Brain Atlas and averaged FA for forceps major, forceps minor, cingulum, inferior fronto-occipital fasciculus, inferior longitudinal fasciculus, superior longitudinal fasciculus (SLF) and uncinate fasciculus. First, we modeled the FA age trajectory of each tract in HC and used it to regress out the effect of age in patients. The residuals in the FA values were utilized for further analyses. We conducted mediator regression analyses with FA as the dependent variable, sex as a covariate, and age and duration of illness as the independent variables. Next, patients were grouped based on Chlorpromazine equivalent dosage (CPZ) and CPZ group was added to the regression model. To examine the association between structure and function, a structural equation model (SEM) was used, with cognition as a mediator. Last, all analyses were repeated for males and females separately. Results Regression analyses revealed a significant influence of duration of illness on the forceps major (T=3.24, p<.001), forceps minor (T=3.40, p<.001), and SLF (T=3.83, p<.0001). When adding both age and duration of illness, duration of illness mediated the influence of age on FA. Adding CPZ to the model displayed a significant effect of medication on forceps major (T=3.93, p<.0001). For SEM, FA of all tracts was used to represent structural impairment, and symptom scores were used to reflect functional impairment. All data paths were calculated with an asymptotically distribution-free model. The overall model fit was acceptable (RMSEA =.09) with a medium-strong effect of structural impairment on functional impairment (standardized estimate=.44). When separating sexes, males showed a significant association of duration of illness and FA. Females displayed a stronger influence of structural impairment on functional impairment (standardized estimate: males = .29, females =.52), and cognition had an impact on this relationship for females only. Discussion The effect of duration of illness on specific WM tracts suggests a progressive, neurodegenerative pathology, which might contribute to the devastating effects of chronicity. Medication seems to have only a small, localized effect on corpus callosum WM integrity. However, since our analysis used cross-sectional CPZ measures, future studies should include measurements of lifetime dosage. The observed impact of WM abnormalities on function further highlights the importance of WM for SCZ pathology. The association of structure and function seems sex-specific. Men show a stronger effect of chronicity. For females, cognition mediates the effect of structure on function, suggesting the role of cognitive reserve. Our approach provides the first step towards evidence-based medicine, by demonstrating the apparent relationship between structural pathology and functional outcome and suggesting possible mediators of this relationship, including duration of illness, cognition, medication, and sex.