ELM Score Could Shrink the “Grey Zone” in Non-Invasive CSPH Diagnosis: Journal of Hepatology | September 2026 Study

Introduction:

Baveno criteria have transformed non-invasive diagnosis of clinically significant portal hypertension (CSPH), but many patients remain in an indeterminate “gray zone.” The ELASTO-ML study developed a machine-learning pan-elastography ELM score to improve CSPH classification while remaining applicable across different elastography platforms.

Why was this study needed?

HVPG remains the reference standard for CSPH but is invasive and not widely available.

Baveno criteria leave a substantial proportion of patients unclassified.

Better non-invasive classification could improve selection for NSBB/carvedilol therapy.

A model applicable across different elastography technologies would improve real-world usability.

Key Findings:

The study included 1,435 patients with compensated advanced chronic liver disease and paired HVPG measurements.

ELM combines liver stiffness, spleen stiffness, platelet count, and clinical variables using machine learning.

In the 342-patient external-validation cohort, diagnostic performance was strong (AUC 0.91).

An ELM score ≤0.45 ruled out CSPH with an NPV of 90%.

An ELM score ≥0.60 ruled in CSPH with a PPV of 96%.

Only 12.3% of patients remained indeterminate, compared with 47.9% using Baveno VII criteria.

The model performed across VCTE, 2D-SWE, and point-SWE, supporting a pan-elastography approach.

ELM could potentially reduce unnecessary HVPG testing and improve non-invasive identification of patients who may benefit from NSBB therapy.

Prospective implementation and regulatory validation are required before replacing established Baveno algorithms.

Bottom Line:

ELM substantially reduced the CSPH diagnostic gray zone while maintaining strong rule-in and rule-out performance. It could make non-invasive portal-hypertension assessment more actionable, but remains a research tool pending prospective implementation.