Concussion As a Multi-Scale Complex System: An Interdisciplinary Synthesis of Current Knowledge

Abstract

Traumatic brain injury (TBI) has been called "the most complicated disease of the most complex organ of the body" and is an increasingly high-profile public health issue. Many patients report long-term impairments following even "mild" injuries, but reliable criteria for diagnosis and prognosis are lacking. Every clinical trial for TBI treatment to date has failed to demonstrate reliable and safe improvement in outcomes, and the existing body of literature is insufficient to support the creation of a new classification system. Concussion, or mild TBI, is a highly heterogeneous phenomenon, and numerous factors interact dynamically to influence an individual's recovery trajectory. Many of the obstacles faced in research and clinical practice related to TBI and concussion, including observed heterogeneity, arguably stem from the complexity of the condition itself. To improve understanding of this complexity, we review the current state of research through the lens provided by the interdisciplinary field of systems science, which has been increasingly applied to biomedical issues. The review was conducted iteratively, through multiple phases of literature review, expert interviews, and systems diagramming and represents the first phase in an effort to develop systems models of concussion. The primary focus of this work was to examine concepts and ways of thinking about concussion that currently impede research design and block advancements in care of TBI. Results are presented in the form of a multi-scale conceptual framework intended to synthesize knowledge across disciplines, improve research design, and provide a broader, multi-scale model for understanding concussion pathophysiology, classification, and treatment.

Keywords: complex; concussion; models of injury; multi-scale; networks; systems science; traumatic brain injury.

Figure 1

Multi-scale framework for concussion. Factors influencing concussion pathophysiology and recovery are shown across four nested emergent scales: cellular, network, experiential, and social. Endogenous factors—those that affect and are affected by other factors in the system—are included inside the scale boxes. Exogenous system drivers that act upon the system are shown at the perimeter. On the left are exogenous factors present at the time of injury (e.g., injury phenomena and biomechanics, personal characteristics, and injury context), while interventions on the right and top margins impact the system dynamically during the recovery process. Aspects of the ongoing environment influence factors at all scales. Feedback exists within and also between scales. Medium gray arrows indicate cross-scale interactions. Factors show emergence, increasing size, and longer time-scale moving up from the cellular to social levels. A team of systems scientists produced this diagram based on literature review, interviews with researchers and clinicians, and iterative review by subject matter experts.

Figure 2

Concussion at the network scale. When investigating how information passes through a network (A,B), a node may represent neuronal activations, or network properties of gray or white matter in a given location. Connections between nodes may represent structural white matter tracts (i.e., groups of axons), temporal synchronization, or other functional relationships. Major hubs [shown in panel (B)] refer to regions that play a central role in connectivity and information processing. For any given structural or functional connectivity network, major hubs may be contrasted with less important hubs and peripheral nodes and edges. Comparison of magnetic resonance imaging scans from mild and severe TBI patients is shown in panel (C). The smaller the site of injury, the less likely the damage is to significantly interrupt information processing within a given network. In the concussed patient, if the injury impacts a peripheral small node, there might only be a minimal disruption in function, especially if the functional connectivity needed for recovery is restored via rerouting. In the severe TBI case, such workarounds will not be possible as the damage is too extensive. A network approach illustrates the possibility that severity or prognosis may be based on extent of network damage. Panels (A,B) were borrowed from Ref. (138), and panel (C) was borrowed from Ref. (124). Used with permission.

Figure 3

Multi-scale map of factors influencing recovery from concussion in two cases. In this comparative hypothetical example, two individuals present with sleep disruption of the same magnitude at the same time point after injury. Additional variables differentiate the cases. Factors are mapped across four nested scales and are linked with arrows indicating causal influence. The minus sign in the diagram for person A signifies a reduction in depression. While both persons show the same symptom (i.e., sleep disturbance), outcomes differ between the two cases according to distinct profiles of feedback relationships with other variables. Disrupted sleep will result in clinical depression for person A if left untreated, placing her at increased risk of dementia, given her medical history and age. Person B might develop persistent post-concussive symptoms if the compounding effects of stress and headache are not addressed. Default mode network is abbreviated DMN. This example illustrates the utility of considering scale and feedback in the clinical care of concussion.

Figure 4

Identifying patterns of variables across multiple scales in concussion. Three subgroups are shown using dashed, solid, and dotted lines to indicate relationships between variables in concussion, mapped across four nested scales. This framework can be utilized for synthesizing existing data and knowledge in concussion, based on systematic review or big data analytics. Relationships between variables can be organized to visualize a “landscape” of factors, within which patterns or subgroups might be identified. This subgrouping would allow better phenotyping for the design, recruitment, and analysis of clinical trials and might enable reanalysis of failed drug trials to distinguish responders from non-responders. The ultimate goal of such a model is better clinical prognostication of outcomes following concussion and therefore more personalized treatments.