Personalised profiling to identify clinically relevant changes in tremor due to multiple sclerosis

David G Western,Simon A Neild,Rosemary Jones,Angela Davies-Smith

doi:10.1186/s12911-019-0881-1

Abstract

BackgroundThere is growing interest in sensor-based assessment of upper limb tremor in multiple sclerosis and other movement disorders. However, previously such assessments have not been found to offer any improvement over conventional clinical observation in identifying clinically relevant changes in an individual’s tremor symptoms, due to poor test-retest repeatability.MethodWe hypothesised that this barrier could be overcome by constructing a tremor change metric that is customised to each individual’s tremor characteristics, such that random variability can be distinguished from clinically relevant changes in symptoms. In a cohort of 24 people with tremor due to multiple sclerosis, the newly proposed metrics were compared against conventional clinical and sensor-based metrics. Each metric was evaluated based on Spearman rank correlation with two reference metrics extracted from the Fahn-Tolosa-Marin Tremor Rating Scale: a task-based measure of functional disability (FTMTRS B) and the subject’s self-assessment of the impact of tremor on their activities of daily living (FTMTRS C).ResultsUnlike the conventional sensor-based and clinical metrics, the newly proposed ’change in scale’ metrics presented statistically significant correlations with changes in self-assessed impact of tremor (maxR2>0.5,p<0.05 after correction for false discovery rate control). They also outperformed all other metrics in terms of correlations with changes in task-based functional performance (R2=0.25 vs. R2=0.15 for conventional clinical observation, both p<0.05).ConclusionsThe proposed metrics achieve an elusive goal of sensor-based tremor assessment: improving on conventional visual observation in terms of sensitivity to change. Further refinement and evaluation of the proposed techniques is required, but our core findings imply that the main barrier to translational impact for this application can be overcome. Sensor-based tremor assessments may improve personalised treatment selection and the efficiency of clinical trials for new treatments by enabling greater standardisation and sensitivity to clinically relevant changes in symptoms.

Highlights

There is growing interest in sensor-based assessment of upper limb tremor in multiple sclerosis and other movement disorders
The proposed metrics achieve an elusive goal of sensor-based tremor assessment: improving on conventional visual observation in terms of sensitivity to change
The plots for mean change and scale change peak erratically in the range 1 < N < 7, beyond which they exhibit no substantial correlation (R2 < 0.1). These results suggest a particular strength of the scale2 metric; by exaggerating the influence of the most tremulous joints on a continuous scale, it reduces the dependence of the number of joints included

Summary

Introduction

There is growing interest in sensor-based assessment of upper limb tremor in multiple sclerosis and other movement disorders Previously such assessments have not been found to offer any improvement over conventional clinical observation in identifying clinically relevant changes in an individual’s tremor symptoms, due to poor test-retest repeatability. The translational impact of this research has been limited by the fact that, despite the precision and sensitivity of available sensors, the recorded data exhibit considerable test-retest variability. This limitation was highlighted recently by a task force of the International Parkinson and Movement Disorder Society (MDS) [14]: In most clinical studies, investigators are interested in changes in tremor amplitude or occurrence that exceed random variability. There is no evidence that the minimum detectable change is smaller when transducers are used

Methods

Results

Discussion

Conclusion