August 2026 Research Update

This month added evidence that AI may help score motor function in MLD from clinical-trial assessment videos. In one study using clinical trial videos, an AI system estimated visit-level GMFC-MLD (a motor function score for MLD) with good agreement against adjudicated labels on visits where human raters were fully aligned. This matters because motor function is a central part of how MLD progression is followed over time, and better consistency could make comparisons across visits easier. [1]

The model was built to protect privacy by converting regular video into a simple 2D skeleton (body-joint positions) before scoring, instead of keeping identifiable images. In 44 children with 303 visits from the EMBOLDEN phase-2 trial, the system reached 82.1% accuracy and a 0.806 quadratic weighted kappa on visits where raters fully agreed on the underlying score. The authors also reported that the model runs in real time on 2D skeleton data and supports streaming inference, which could allow future on-device use; they also released de-identified data and code. [1]

For families, this is mainly a research step toward more standardized motor tracking, potentially using videos already collected during care or trials. It does not show that care changes now, but it helps clarify that objective, privacy-preserving video-based scoring is technically feasible and can match expert scoring reasonably well in a trial dataset. What remains uncertain is how well it performs in everyday clinic videos, in different ages and disease stages, and in visits where human scoring is less clear-cut. [1]

This month included a study asking a practical question: if MLD were added to newborn screening, how much earlier might children be identified, and how might that change outcomes at the population level. Using a decision-analytic model (a structured way to project outcomes using available evidence), the authors projected that screening would find more early-onset cases than clinical identification and would increase the number of children projected to survive without motor impairment around 5 years after symptoms start. The main message is that earlier identification is projected to translate into better short-term functional outcomes in the model. [2]

In a modeled US birth cohort of 3.6 million newborns, the study projected that newborn screening would identify 36 infants per year with early-onset MLD, compared with 11.5 identified without screening. The model also projected 32.8 children surviving without motor impairment at about 5 years after symptom onset with screening, compared with 9.1 without screening. These numbers are model outputs rather than real-world results, but they provide a concrete estimate of how much earlier case-finding might matter. [2]

For patients and caregivers, this is forward-looking evidence that helps explain why newborn screening is being considered for MLD: earlier diagnosis is projected to increase the chance of reaching treatment eligibility and of better near-term motor outcomes in the model. It does not prove what will happen in every state or health system, because it depends on how screening, follow-up testing, and access to treatment work in practice. What remains uncertain is how closely these projections match real-world outcomes once screening is implemented more widely. [2]