The Enduring Allure of Explanatory Reductionism in Schizophrenia

From its earliest description, the human brain has been conceptualized as a collection of interconnected regions that process and exchange information about the internal and external environment. Over the last 2 decades, methodological advances in neuroimaging acquisition and analysis have increased our understanding of the organization of the human brain. Network neuroscience theories and computational tools have been instrumental in the comprehensive mapping of the structural and functional connections between neural elements. These efforts have given rise to the concept of the “connectome,” which has been characterized principally at a macroscale level, focusing on anatomical and functional connections between brain regions. Major efforts are now directed toward the identification of mechanisms that link and shape the structural and functional architecture of the human connectome. Of particular interest are metrics that provide information about what influences both local and long-raging connectivity and how regional connectivity profiles may shape the trajectories of information flow across the brain (1).

The challenge of understanding the interactions between the structural and functional connectome is amplified in the case of psychiatric disorders. Schizophrenia is the most studied disorder in terms of neuroimaging and the first for which the concept of dysconnectivity was applied to explain the brain basis of its clinical and cognitive symptoms (2). The dysconnectivity model of schizophrenia is based on evidence of anatomical and functional changes in patients compared with healthy individuals; notably, these include widespread reductions in brain volume, cortical thickness, and the integrity of white matter tracts and abnormal connectivity of the major functional networks (3–5).

In this issue of Biological Psychiatry, Shafiei et al. (6) tackle the question of the relationship between the structural and functional connectome in schizophrenia. Specifically, they tested whether volumetric abnormalities in schizophrenia clustered within regions that were structurally and functionally connected to each other. To achieve this, they used structural magnetic resonance imaging data from 2 independent samples (n₁ and n₂) to calculate regional changes in brain tissue volume density in patients with chronic schizophrenia (n₁ = 133 and n₂ = 108) compared with healthy individuals (n₁ = 133 and n₂ = 68). Estimates of the structural and functional connectivity of these regions were obtained from a third sample comprising only healthy individuals for whom diffusion and resting-state functional magnetic resonance imaging data were available. Graph theory metrics were then used to compute the degree of anatomical and functional “connectedness” of each brain region to all other regions. Brain regions were also assigned to each of the 7 major resting-state networks as defined by the Yeo atlas and to each of 7 cytoarchitectonic classes as defined by von Economo. Shafiei et al. (6) found that regions in close spatial proximity with direct anatomical connections tended to show similar schizophrenia-related regional tissue density changes. Disease-related tissue density changes were widespread and involved multiple resting-state systems and cytoarchitectonic classes. The pattern of case-control differences within regions assigned to resting-state networks showed a gradation, with effect sizes being larger for regions in the default mode network and cognitive control networks (ventral attention and frontoparietal) and smallest for orbitofrontal and temporal pole regions assigned to the limbic network. Similarly, there was a gradation of case-control differences in cytoarchitectonically defined clusters, with effect sizes being larger and comparable for the cingulate, insular, and primary sensory cortices. The authors went to admirable lengths to test that the findings were reproducible across datasets and at different scales of structural parcellation and were robust to sex, age, and medication.

To a great extent, the findings of the study confirm what is already known about the brain structural correlates of schizophrenia (4,5). The spatial pattern of case-control differences in brain tissue density reported by Shafiei et al. (6) is widely distributed but shows some evidence of regional variation in the degree of disease-related deviance. An interesting feature of the study is the use of resting-state networks and cytoarchitectonic classes as organizing principles for reporting the results; it makes the findings relatable to other levels of the brain’s architecture, namely the level of cognitive systems, which are supported by resting-state networks, and of micro-structural characteristics that define the von Economo classes. The latter has additional heuristic value because it could be used to generate testable hypotheses regarding the cellular underpinnings of schizophrenia.

This study presents an interesting approach to describing the spatial pattern of brain structural changes in schizophrenia, but the results do not provide any mechanistic insights. This observation applies to most, if not all, of the current brain imaging literature in psychiatry. To date, the main achievement of neuroimaging is to have demonstrated that psychiatric disorders involve the brain and not just the “mind.” However, questions relating to why and how these brain alterations come about remain largely unanswered. Much to the frustration of researchers in the field, there is ongoing uncertainty about the role of brain alterations for psychiatric symptom dimensions or traditional syndromal classifications. The current state of affairs contrasts with societal pressures to find the “cause” of psychiatric disorders in a similar fashion that the rest of medicine has found the cause of infectious diseases and a host of other diseases. This desire for explanatory reductionism—to explain psychiatric morbidity in terms of limited and specific biological abnormalities—is pervasive and persistent. It is understandable that the authors ventured into mechanistic inferences by electing to focus on the anterior cingulate cortex based on the observation that this region had the highest schizophrenia-related deviance in tissue density in general but also among its neighboring regions. The researchers characterized the anterior cingulate as an “epicenter” of brain structural pathology in schizophrenia, suggesting that deviance in this region may play a key role in this disorder. This formulation brings to mind older explanatory theories for schizophrenia that were largely inspired by the “lesion model” and focused mostly on hippocampal or prefrontal pathology (7,8). At the same time, the anterior cingulate cortex has consistently been implicated, not just in schizophrenia but across many psychiatric disorders (9,10). Similar transdiagnostic involvement has been reported mainly for the insula, the precuneus, the ventral and dorsal prefrontal regions, the hippocampus, and the thalamus (9,10). These regions perform mostly integrative functions within the brain and are thus engaged in response to multiple and diverse cognitive demands. It could be argued that collectively they define a space of vulnerability to mental dysfunction. Within this space, impairments could result from numerous distinct deficits that may vary at the level of individuals, symptom dimensions, or clinical syndromes. Teasing out these underlying mechanisms is beyond the resolution of macro-level measures of brain morphometry or connectivity. Studies using such measures are unlikely to provide mechanistic insights or to substantially enhance our understanding of the pathogenesis of psychopathology. The concept of a space of vulnerability can be used as a spotlight to direct research efforts toward the identification of biological or environmental mechanisms that may increase or mitigate the structural and functional integrity of these vulnerable regions. Such efforts can be deployed at different levels ranging from interventions aiming to increase adaptive plasticity within these regions, to genetic and molecular characterization of biological pathways that shape their development and function.