Expert Opinion in the Design of a Motor Neurone Disease Diagnostic Study

Abstract

Abstract Background Motor neurone disease (MND) is a rapidly progressing and rare neurodegenerative disorder characterized by progressive weakness, muscle wasting, and death from respiratory failure within 36 months of symptom onset. To date, clinical trials in MND have failed to identify therapeutic interventions that halt disease progression, possibly because the majority of patients are recruited to trials too late in the disease course. To recruit patients earlier, diagnostic criteria for MND now include evidence of subclinical disease in unaffected muscles, as assessed by needle electromyography (EMG). Whilst other electrodiagnostic tests of subclinical disease could be incorporated into these criteria alongside needle EMG, it is unclear whether this would provide additional diagnostic accuracy/certainty. Here we use beta-band intermuscular (EMG-EMG) coherence (BIMC) as an example of how this issue can be addressed with statistical confidence in future studies. Methods Using the BIMC test as a case study, we provide a statistical framework for the incorporation of expert knowledge into the choice of sample size using expert elicitation and Bayesian assurance calculations. Probability distributions were elicited from seven clinical experts and aggregated to form group consensus distributions. Results The Bayesian assurance calculations led to a smaller required sample size than traditional statistical power calculations. The quantification and incorporation of clinical expert knowledge and uncertainty in sample size calculations can provide better calibrated predictions of study outcomes and ensure the most appropriate sample size is chosen. Clinical experts reported the sensitivity of the Awaji criteria in line with previous studies, providing evidence of the validity of the results. We note that multiple experts understated estimates of specificity compared to the literature, though this may be due to the format of the questions or the design of the case study. Conclusions Bayesian assurance can be used alongside expert elicitation to design diagnostic accuracy studies. While we focus on the BIMC test case study, the methods presented are relevant and can be applied to other emerging tests relevant to MND.