Using biomarkers to predict clinical outcomes in multiple sclerosis

Combination protein biomarkers predict multiple sclerosis diagnosis and outcomes

Establishing biomarkers to predict multiple sclerosis diagnosis and prognosis has been challenging using a single biomarker approach. We hypothesised that a combination of biomarkers would increase the accuracy of prediction models to differentiate multiple sclerosis from other neurological disorders and enhance prognostication for people with multiple sclerosis. We measured 24 fluid biomarkers in the blood and cerebrospinal fluid of 77 people with multiple sclerosis and 80 people with other neurological disorders, using ELISA or Single Molecule Array assays. Primary outcomes were multiple sclerosis versus any other diagnosis, time to first relapse, and time to disability milestone (Expanded Disability Status Scale 6), adjusted for age and sex. Multivariate prediction models were calculated using the area under the curve value for diagnostic prediction, and concordance statistics (the percentage of each pair of events that are correctly ordered in time for each of the Cox regression models) for prognostic predictions. Predictions using combinations of biomarkers were considerably better than single biomarker predictions. The combination of cerebrospinal fluid [chitinase-3-like-1 + TNF-receptor-1 + CD27] and serum [osteopontin + MCP-1] had an area under the curve of 0.97 for diagnosis of multiple sclerosis, compared to the best discriminative single marker in blood (osteopontin: area under the curve 0.84) and in cerebrospinal fluid (chitinase-3-like-1 area under the curve 0.84). Prediction for time to next relapse was optimal with a combination of cerebrospinal fluid[vitamin D binding protein + Factor I + C1inhibitor] + serum[Factor B + Interleukin-4 + C1inhibitor] (concordance 0.80), and time to Expanded Disability Status Scale 6 with cerebrospinal fluid [C9 + Neurofilament-light] + serum[chitinase-3-like-1 + CCL27 + vitamin D binding protein + C1inhibitor] (concordance 0.98). A combination of fluid biomarkers has a higher accuracy to differentiate multiple sclerosis from other neurological disorders and significantly improved the prediction of the development of sustained disability in multiple sclerosis. Serum models rivalled those of cerebrospinal fluid, holding promise for a non-invasive approach. The utility of our biomarker models can only be established by robust validation in different and varied cohorts.

Lessons from immunotherapies in multiple sclerosis

The improved understanding of multiple sclerosis (MS) neurobiology alongside the development of novel markers of disease will allow precision medicine to be applied to MS patients, bringing the promise of improved care. Combinations of clinical and paraclinical data are currently used for diagnosis and prognosis. The addition of advanced magnetic resonance imaging and biofluid markers has been strongly encouraged, since classifying patients according to the underlying biology will improve monitoring and treatment strategies. For example, silent progression seems to contribute significantly more than relapses to overall disability accumulation, but currently approved treatments for MS act mainly on neuroinflammation and offer only a partial protection against neurodegeneration. Further research, involving traditional and adaptive trial designs, should strive to halt, repair or protect against central nervous system damage. To personalize new treatments, their selectivity, tolerability, ease of administration, and safety must be considered, while to personalize treatment approaches, patient preferences, risk-aversion, and lifestyle must be factored in, and patient feedback used to indicate real-world treatment efficacy. The use of biosensors and machine-learning approaches to integrate biological, anatomical, and physiological parameters will take personalized medicine a step closer toward the patient's virtual twin, in which treatments can be tried before they are applied.

Personalized prediction of rehabilitation outcomes in multiple sclerosis: a proof-of-concept using clinical data, digital health metrics, and machine learning

Predicting upper limb neurorehabilitation outcomes in persons with multiple sclerosis (pwMS) is essential to optimize therapy allocation. Previous research identified population-level predictors through linear models and clinical data. This work explores the feasibility of predicting individual neurorehabilitation outcomes using machine learning, clinical data, and digital health metrics. Machine learning models were trained on clinical data and digital health metrics recorded pre-intervention in 11 pwMS. The dependent variables indicated whether pwMS considerably improved across the intervention, as defined by the Action Research Arm Test (ARAT), Box and Block Test (BBT), or Nine Hole Peg Test (NHPT). Improvements in ARAT or BBT could be accurately predicted (88% and 83% accuracy) using only patient master data. Improvements in NHPT could be predicted with moderate accuracy (73%) and required knowledge about sensorimotor impairments. Assessing these with digital health metrics over clinical scales increased accuracy by 10%. Non-linear models improved accuracy for the BBT (+ 9%), but not for the ARAT (-1%) and NHPT (-2%). This work demonstrates the feasibility of predicting upper limb neurorehabilitation outcomes in pwMS, which justifies the development of more representative prediction models in the future. Digital health metrics improved the prediction of changes in hand control, thereby underlining their advanced sensitivity.

Graphical Abstract