Learning curve and interobserver reproducibility evaluation of liver stiffness measurement by transient elastography.

Background/aims: Fibroscan allows liver stiffness examination (LSE) that is well correlated with fibrosis stages. Our main objective was to evaluate LSE learning curve.

Methods: LSE results of five novice observers with different medical status were compared with those of five expert observers (physicians with >100 examinations) in 250 patients with chronic liver disease. Each novice-expert pair had to blindly examine 50 consecutive patients divided into five consecutive subgroups of 10 patients.

Results: In each observer group, novice-expert agreement [intraclass correlation coefficient (Ric)] for LSE results was excellent from the first to the last subgroup. Novice-expert agreement for LSE results varied with liver stiffness level: <9 kPa: Ric=0.49; >or=9 kPa: Ric=0.87. Relative difference (%) between novice and expert LSE results was independently associated with the number of valid LSE measurements, and stabilizes around 20-30% after the fourth valid measurement. In each observer group, novice-expert agreement (Ric) for LSE success rate progressively increased as a function of time.

Conclusion: LSE requires no learning curve: a novice is able to obtain a reliable result after a single training session, whatever the professional status. However, success rate will progressively increase. An LSE with less than four valid measurements should not be considered as reliable.