A machine learning-based model can help predict short-term outcomes in myasthenia gravis (MG), according to a new study published in the journal Therapeutic Advances in Neurological Disorders.
“Our predictive tool may help promote the clinical management of MG patients and build a follow-up surveillance system for professionals,” the authors wrote.
Read more about the prognosis of MG
The course of MG can vary with time making it challenging to manage.
In the present study, a team of researchers from China developed and validated a model based on machine learning to predict short-term clinical outcomes in patients with MG who have different antibodies.
The researchers assessed 890 patients who had regular follow-ups at 11 tertiary centers in China between January 2015 and July 2021. They used a 2-step variable screening to determine the factors for the construction of the model as well as 14 machine learning algorithms for its optimization.
The model identified patients whose condition improved, stayed the same, or worsened both in a derivation cohort consisting of 653 patients and in a validation cohort consisting of 237 patients. The areas under the receiver operating characteristic curve were 0.91, 0.89, and 0.89 for the improved, stable, and worsened patients, respectively, in the derivation cohort and 0.84, 0.74, and 0.79, respectively, in the validation cohort.
“Both datasets presented a good calibration ability by fitting the expectation slopes,” the researchers wrote.
The model will now be assessed for feasibility using a web tool.
MG is a rare autoimmune disease affecting the neuromuscular junction. It is characterized by the presence of various antibodies that attack and injure receptors found on the surface of muscle cells that receive nerve impulses. This leads to generalized fatigability as well as muscle weakness, which is the result of the reduced transmission of electrical impulses across the neuromuscular junction.
Zhong H, Ruan Z, Yan C, et al. Short-term outcome prediction for myasthenia gravis: an explainable machine learning model. Ther Adv Neurol Disord. Published online February 24, 2023. doi:10.1177/17562864231154976