Brain heterogeneity in 1,792 individuals with schizophrenia: effects of illness stage, sites of origin and pathophysiology

Abstract

ImportanceSchizophrenia is characterized with greater variability beyond the mean differences in brain structures. This variability is often assumed to be static, reflecting the presence of heterogeneous subgroups, but this assumption and alternative explanations remain untested.ObjectiveTo test if gray matter volume (GMV) variability is more less in later stages of schizophrenia, and evaluate if a putative ‘spreading pattern’ with GMV deficits originating in one part of the brain and diffusing elsewhere explain the variability of schizophrenia.Design, settings, and participantsThis study evaluated the regional GMV variability using MRI of 1,792 individuals with schizophrenia and 1,523 healthy controls (HCs), and the association of GMV variability with neurotransmitter and transcriptomic gene data in the human brain.Main outcomes and measuresRegional variability was evaluated by comparing the relative variability of patients to controls, using the relative mean-scaled log variability ratio (lnCVR). A network diffusion model (NDM) was employed to simulate the possible processes of GMV alteration across brain regions.ResultsCompared with HCs, greater lnCVR (pFDR<0.05) was found in 50 regions in the whole patient group (n=1792; 762 females; mean[SD] age, 29.9[11.9] years), at a much greater frequency (p=5.0×10−13) in the first-episode drug-naïve subsample (73 regions) (n=478; mean[SD] illness duration, 0.548[0.459] years), compared to the chronic medicated subsample (28 regions) (n=398; mean[SD] illness duration, 14.0[10.4] years). The average lnCVR across all regions was greater in the first-episode than chronic subsample (t=10.8,p=1.7×10−7). The areas with largest lnCVR were located at frontotemporal cortex and thalamus (first-episode), or hippocampus and caudate (chronic); there was a significant correlation with case-control mean difference (r=0.367,p=6.7×10−4). We determined a gene expression map that correlated with the lnCVR map in schizophrenia (r=0.491,p=0.003). The NDM performed consistently (72.1% patients,pspin<0.001) in replicating GMV changes when simulated and observed values were compared.Conclusion and relevanceBrain-based heterogeneity is unlikely to be a static feature of schizophrenia; it is more pronounced at the onset of the disorder but reduced over the long term. Differences in the site of ‘origin’ of GMV changes in individual-level may explain the observed anatomical variability in schizophrenia.Key PointsQuestionNo two individuals with schizophrenia have the same anatomical change in the brain. Is this variability a fixed feature of schizophrenia or does it become more pronounced at later stages? Is this variability explained by a putative ‘spreading pattern’ of gray matter deficits originating in one part of the brain and diffusing elsewhere?FindingsIn 1,792 individuals with schizophrenia, neuroanatomical variability is not a fixed feature; it is more pronounced at the illness onset but less prominent in later stages. The neuroanatomical variability is associated with various molecular and neurobiological processes implicated in the neurodevelopmental etiology of schizophrenia. Differences in the site of ‘origin’ of gray matter deficits in each individual with schizophrenia explains most of the observed variability.MeaningOur work finds support for a space-time interaction along a shared pathophysiological continuum (network-based trans-neuronal diffusion), as a possible explanatory model for inter-subject variability. These findings contribute to the understanding that inter-individual variability in schizophrenia may arise from a common cohesive process that varies in its state (across time) and space (across brain regions). This also raises the question of what dynamic processes contribute to the reducing heterogeneity over time in schizophrenia. Answering this question will be a key test to the neurobiological validity of the concept of schizophrenia.