OBJECTIVES:

Magnetic Resonance Imaging (MRI) has become the most useful para-clinical tool to detect (subclinical) disease activity in order to evaluate treatment response and inform clinical decisions about Disease Modifying Treatment (DMT) in people with Multiple Sclerosis (pwMS). However, current evaluation of brain MRI scans, in a clinical setting, requires visual inspection, the sensitivity of which is degraded by multiple human and technological factors. AI-derived software for medical image quantification offer gains in efficiency, accuracy, clinical relevance, reduced subjectivity, and have the potential to allow timely therapy decisions which can improve patient’s outcome. The aim of this research is to provide an initial evaluation of the cost-effectiveness of AI-assisted radiological assessment in the MS clinical decision-making pathway in the UK using a health-care perspective.

METHODS:

We developed a Markov model based on the Treatment Algorithm for Multiple Sclerosis Disease-modifying Therapies from the NHS. The model compares the current clinical decision-making pathway, which uses visual inspection for MRI assessment of disease activity, with a clinical decision-making using AI-assisted MRI assessment. The model provides estimates for the overall costs, Quality-Adjusted Life Years (QALYs), and incremental cost effectiveness ratio (ICER) per QALY gained.

RESULTS:

The AI-assisted MRI assessment was cost-effective at a willingness to pay threshold of £20,000 per QALY, with an ICER of £4,148. The incremental QALYs gain was 0,11 and the incremental costs £443 per patient over a time-horizon of 20 years. The probability of being cost-effective was 92.5% and 96.7% at a threshold of £20,000 and £30,000 per QALY respectively.

CONCLUSIONS:

These results suggest that AI-derived software can add value to neuroradiology assessment in pwMS, providing insights into clinical decision-making that can improve patient outcomes. This model is useful to identify key drivers of cost-effectiveness to inform research and decision about the use of AI-assisted MRI assessment in pwMS.