Ultrasonic spleen thickness-based indexes surpass Baveno VI criteria in high-risk gastroesophageal varices detection

Aims To evaluate the efficiency of ultrasonic spleen thickness (UST), routine variables and (expanded) Baveno VI criteria for high-risk gastroesophageal varices (HRGOV) detection in cirrhotic patients. Methods In total, 305 cirrhotic patients were retrospectively enrolled in the deriving cohort and 328 cirrhotic patients with hepatitis B sustained viral response were prospectively enrolled in the validation cohort. HRGOV was defined as medium and severe gastroesophageal varices (GOV), mild GOV with red signs or Child–Pugh C. The cut-offs for HRGOV were determined by likelihood ratio indicating strong evidences. Algorithms of Spleen thickness-Age-Liver stiffness measurement (LSM, by Fibroscan ® )-Albumin (SALA) and Spleen thickness-Platelet-Albumin (SPA) were derived by multivariate analyses. Results The area under receiver operating characteristics curve of SALA, SPA, UST, platelet, and LSM were 0.849, 0.835, 0.808, 0.746, and 0.655 in the deriving cohort, and improved to 0.901, 0.904, 0.858, 0.876, and 0.811 in the validation cohort, respectively. While SALA, SPA, UST, platelet, Baveno VI criteria (BVI), and expanded BVI spared 46.6%, 38.0%, 29.2%, 21.0%, 12.1%, and 23.6% esophagogastroduodenoscopy in the deriving cohort, these numbers were improved to 68.1%, 66.8%, 27.1%, 37.8%, 36.0%, and 61.0% in the validating cohort, respectively; however, the negative likelihood ratio of expanded BVI was up to 0.16. SPA spared less esophagogastroduodenoscopy than SALA, which can be supplemented by stepwise applying UST and SPA. Sequentially combining UST and SALA, BVI and SALA exempted additional 10–5% endoscopies. Conclusions SPA, without LSM, improves HRGOV detection comparing with BVI. UST based algorithms combination can achieve the best efficiency especially in sustained virus response hepatitis B.