Machine Learning Clustering Identifies Achalasia-Spectrum Phenotypes in Distal Esophageal Spasm

Area of Science:

• Gastroenterology
• Esophageal Motility Disorders
• Computational Biology

Background:

• Distal esophageal spasm (DES) is a rare esophageal motility disorder with varied treatment responses.
• Machine learning (ML) can identify distinct phenotypes within heterogeneous conditions like DES.
• This study aimed to use ML to define and characterize DES phenotypes for better therapeutic guidance.

Purpose of the Study:

• To identify and characterize unique distal esophageal spasm (DES) phenotypes using unsupervised machine learning clustering.
• To explore potential correlations between identified DES phenotypes and clinical characteristics, manometric findings, and treatment outcomes.

Main Methods:

• Adult patients undergoing high-resolution manometry (HRM) for esophageal symptoms were analyzed.
• Uniform Manifold Approximation and Projection (UMAP) was employed for dimensionality reduction of HRM data.
• DES patients were clustered, and characteristics, manometric parameters, and treatment outcomes were compared between clusters.

Main Results:

• Of 5360 patients, 42 had DES. UMAP identified 33 DES patients clustering with type III achalasia, while 9 remained separate.
• The key differentiator was premature supine contraction frequency (median 8 vs 0, P < 0.001).
• Within-cluster patients showed higher anxiety scores and better symptom resolution post-myotomy.

Conclusions:

• Machine learning successfully identified a DES phenotype with high premature contraction frequency, clustering with type III achalasia.
• Premature contraction frequency may guide selection of patients for lower esophageal sphincter-directed therapies.
• Larger validation studies are necessary due to the rarity of DES.