25.3 IMAGING ACCELERATED AGING MACHINE-LEARNING BASED MULTIVARIATE PATTERN ANALYSIS OF STRUCTURAL MRI IN SCHIZOPHRENIA AND ULTRA-HIGH-RISK SUBJECTS. (9th April 2019)

• Record Type: Journal Article
• Authors: Nenadic, Igor
• Is Part Of: Schizophrenia bulletin. Volume 45(2019)Supplement 2
• Page Start: S130
• Page End: S130
• Publication Date: 2019-04-09
• Subjects: Schizophrenia -- Periodicals; Schizophrenia -- Research -- Periodicals; 616.898005
• DOI: 10.1093/schbul/sbz022.103
• Languages: English
• ISSNs: 0586-7614

Abstract

Abstract: Background: Brain structural changes in schizophrenia parallel some age-associated changes. Previous studies were often limited to cross-sectional analyses of interacting effects of diagnosis and age, while pointing to prefrontal and (superior) temporal cortical alterations being most prone to progressive brain structural changes associated with age. The use of machine-learning based approaches has allowed the estimation of age in individual brain scans based on comparison to normative data sets. One of these strategies, the BrainAGE (brain age estimation gap) approach provides a metric describing the deviance of an individual's brain from the physiological age trajectory, thus yielding an indicator of accelerated brain aging based on multivariate pattern analysis (rather than univariate statistics of single brain regions). In this presentation, we include a systematic evaluation in samples with chronic schizophrenia, first-episode schizophrenia, ultra-high-risk (UHR) subjects for schizophrenia, as well as phenotypic (schizotypy) and genetic (polygenic risk) factors in large healthy cohorts. Methods: We include four different data sets to test the hypothesis of accelerated brain aging in schizophrenia, based on high-resolution anatomical 3T MRI scans. We used the BrainAGE algorithm (Franke et al., 2011; Gaser et al. 2013) applying machine-learning based on a separate training data set to calculate the gap between estimated and chronological brain age. This includes a pilot BrainAGE study on a total of 70 subjects with chronic schizophrenia patients compared to healthy controls and bipolar patients, and a comparison data set (n=72) with major depression patients and matched healthy controls; a 112 subject study comparing ultra-high risk (UHR) subjects with first-episode schizophrenia patients and healthy controls, and finally two analyses from the FOR2107 multi-center cohort study on the association of BrainAGE with psychometric schizotypy and polygenic risk for schizophrenia. Results: First, we find that BrainAGE scores are elevated in chronic schizophrenia (compared to bipolar disorder and healthy controls), and that BrainAGE is not elevated in major depression. Secondly, we find that BrainAGE is elevated also in first-episode schizophrenia and with a trend in UHR. Third, we find that subgroups within UHR differ on BrainAGE scores: those with genetic risk show elevated BrainAGE, while those with UHR status based only on attenuated psychotic symptoms do not show elevated BrainAGE. Currently, additional replication studies are being performed to corroborate genetic findings (also being extended to polygenic risk). Conclusions: We find elevated BrainAGE scores, indicative of accelerated brain aging, in schizophrenia (both chronic and first-episode), but not other major psychoses, as well as UHR subjects with genetic risk. These findings suggest that surrogate markers of accelerated aging are not (only) an epiphenomenon of disease duration or medication but might reflect an inherent and possibly genetically influenced acceleration or progression of structural changes in schizophrenia. While ongoing replication studies are also aiming at extending our findings to molecular genetic markers, our findings demonstrate a consistent pattern across multiple stages of the disease. At the same time, progress is being made to understand a potential genetic overlap of (physiological) processes regulating brain maturation and aging and those implicated in schizophrenia.