Predictive models for motor outcomes from deep brain stimulation in Parkinson's disease: a systematic review

This is a preprint of an unpublished manuscript. Abstract Background Deep brain stimulation (DBS) is an effective treatment for Parkinson’s disease, but the extent of improvement in motor symptoms is variable. A tool which accurately predicts patient outcomes based on information available pre-surgery would be useful for clinical decision-making and patient expectation management. Candidate input data types to such a model include clinical and cognitive measures, brain anatomy, kinematics, functional connectivity, and genetics. Objectives In this systematic review we searched for predictive models of motor outcomes from DBS. We also provide a critical discussion of inputs and outputs of such models, and recommendations to improve validation of models. Methods We searched the databases Web of Science, PubMed and Scopus for primary research articles that tested predictions from models based on pre-surgical data and focused on motor outcomes from DBS for Parkinson’s disease. Results We identified 17 studies fitting these criteria. The studies with high power and generalisability use only clinical data and have limited accuracy, while studies including other types of data are under-powered. Predictions are mostly limited to the first year after surgery. The strongest predictors across studies are age at surgery, duration of symptoms and levodopa-responsiveness. We also provide conceptualization of the prediction task, summarize weak points of current approaches, and discuss predictor modalities. Conclusions Further work is necessary to achieve predictive models that are ready for translation to clinical practice. To achieve this goal, models and datasets should be made publicly available to enable wider validation. Models should also maximise clinical utility by using interpretable inputs and outputs.