Neurorehabilitation in Patients With Disorders of Consciousness: Multidimensional Ambispective Study on the Impact of Single and Combined Approaches

The protocol is grounded in the increasing survival of patients with severe acquired brain injury due to advances in intensive care, which has led to a growing population with prolonged DOC and substantial socio-economic burden, including lifetime care costs exceeding one million euros in some estimates. Within this context, neurorehabilitation is conceptualized as a multidimensional process targeting arousal, awareness, motor and cognitive function, communication, and autonomy, and the study posits that combining structured sensory stimulation with non-invasive brain stimulation and technologically assisted communication can enhance residual brain plasticity and functional recovery beyond that achieved with standard conservative approaches. The rationale emphasizes that DOC assessment based solely on overt behavioural responses is vulnerable to misclassification, underscoring the importance of integrating advanced neurophysiological, neuroimaging, and biological markers to more accurately quantify residual consciousness and track treatment-related changes.

From an intervention standpoint, the protocol details a spectrum of non-invasive treatments: multimodal SS consists of repeated, simple, emotionally salient, or autobiographical stimuli delivered through auditory, visual, tactile, and olfactory channels, which can be administered singly or in combination to enhance arousal and behavioural responsiveness. AAC systems with eye-tracking enable "active" sensory stimulation by pairing visual and auditory stimuli with graded tasks whose performance is inferred from oculomotor behaviour, thus offering both diagnostic information about cognitive abilities and opportunities for structured engagement with the environment. tDCS is introduced as a low-intensity direct current technique applied repeatedly to modulate resting-state membrane polarization and network connectivity, with advantages over repetitive TMS in terms of safety, logistical simplicity, and potential for inducing longer-lasting neuromodulatory effects; TMS is framed as a complementary, focal method of electromagnetic stimulation targeting cortical regions relevant for consciousness.

The protocol allocates substantial attention to biomarkers and mechanistic readouts, particularly the role of neurofilament proteins as indicators of axonal integrity and neuroaxonal damage. Neurofilament light chain (NfL), owing to its relative specificity compared with traditional markers such as neuron-specific enolase, S-100B, and glial fibrillary acidic protein, is incorporated as a key biochemical outcome, measured from venous blood samples obtained in conjunction with routine clinical draws to enable minimally invasive longitudinal monitoring. The study aims not only to explore NfL as a prognostic indicator of long-term neurological outcome but also as a dynamic marker of treatment impact, hypothesizing that responders will show reductions in serum NfL consistent with attenuated ongoing axonal injury.

Neurophysiological recording protocols are designed to capture both spontaneous and stimulus-locked brain activity as correlates of consciousness and plasticity. EEG and high-density EEG acquisitions are typically performed at the bedside under standardized environmental conditions (moderated lighting, potential use of eye masks to reduce artefacts), and processing pipelines include channel-wise normalization, high-pass filtering to remove slow artefacts such as respiration, and low-pass filtering to suppress high-frequency noise. Beyond conventional spectral analyses, the study focuses on event-related potentials derived from paradigms such as auditory oddball stimulation, in which infrequent deviant sounds interspersed among frequent standard tones elicit P300 responses that have been shown to differentiate between diagnostic subgroups and may index residual conscious processing of salient stimuli, including responses to one's own name.

Autonomic and oculomotor measures complement EEG as additional windows onto arousal and information processing. Cardiac monitoring yields indices such as heart rate and heart rate variability, which can reflect autonomic balance and reactivity, while skin conductance responses quantify sympathetic activation in response to sensory events. Eye-tracking metrics captured through AAC devices include fixation duration and count, saccadic latency, amplitude and velocity, gaze accuracy, and distributions across pre-defined areas of interest, enabling fine-grained characterization of visual exploration and potential signs of goal-directed attention or command following in patients with severely limited motor output.

On the neuroimaging side, the protocol leverages a comprehensive suite of structural and functional modalities acquired predominantly on a 1.5T MRI system equipped with a 32-channel head coil. The standard protocol comprises high-resolution three-dimensional T1-weighted imaging using a fast field-echo sequence, high-resolution T2-weighted imaging with turbo spin-echo, diffusion-weighted imaging with multiple diffusion gradient directions and dual-phase encoding for susceptibility correction, and, when available, functional MRI to assess task- or stimulus-related activation patterns. Additional analyses include PET to quantify cerebral glucose metabolism and VBM as well as surface-based morphometry (e.g. with FreeSurfer) to measure cortical thickness, surface area, and subcortical volumes; preprocessing uses tools such as SPM and VBM toolboxes with bias correction, tissue segmentation, and rigid-body registration into MNI space, applying grey-matter thresholds to mitigate partial volume effects.

The statistical analysis plan operationalizes the ambispective design by first characterizing the study population with appropriate descriptive statistics stratified by exposure type and aetiology, then modelling treatment effects while rigorously addressing confounding and heterogeneity. For the primary endpoint, logistic models incorporate clinically motivated covariates and may include interaction terms to assess temporal consistency of effects across retrospective and prospective cohorts; propensity-score techniques are specified to enhance covariate balance, with diagnostics to document adequacy. For longitudinal and continuous outcomes, mixed-effects models account for within-subject correlation and allow random intercepts and, where appropriate, random slopes to capture individual recovery trajectories, while flexible dose-response modelling is planned to identify clinically relevant thresholds or saturation effects of treatment intensity.

The plan for missing data, multiplicity, and robustness reflects an intention to align with contemporary standards in observational clinical research. Multiple imputation by chained equations is proposed under a missing-at-random assumption, with imputation models incorporating exposures, outcomes, and key predictors, and the protocol mandates sensitivity analyses to evaluate the impact of departures from this assumption, including complete-case analyses and extreme-case imputation strategies for the responder classification. Multiplicity across numerous secondary and exploratory endpoints will be controlled by false discovery rate adjustment using the Benjamini-Hochberg procedure while maintaining the conventional two-sided alpha of 0.05, and exploratory subgroup and misclassification-robustness analyses will probe effect modification by aetiology, baseline severity, temporal window since injury, and potential exposure misclassification.

Ethical and administrative sections specify that all data are pseudo-anonymized via unique numeric study codes, with the linkage key stored separately in a secure repository accessible only to authorized research personnel. Informed consent (verbal and written) is obtained from caregivers or legal representatives before data collection in the prospective cohort, while previously collected clinical data from unreachable or deceased patients in the retrospective cohort may be used in pseudo-anonymized form. The study introduces no additional diagnostic or therapeutic procedures beyond standard care, entails no incremental costs or external funding, does not provide financial compensation to investigators, and relies on existing institutional insurance, with publication plans that assign authorship according to substantive contributions to study conception, data collection, analysis, and manuscript preparation.