Platelet count, liver and spleen stiffness-based model reliably identifies esophageal varices needing treatment in acute-on-chronic liver failure.

BACKGROUND AND

OBJECTIVES

Transient elastography of liver and spleen, along with platelet count, identify cirrhosis patients with esophageal varices needing treatment (VNT). Inflammation can drive rapid rise in portal pressure and development and progression of varices in acute-on-chronic liver failure (ACLF) patients. We aimed at developing a non-invasive model to predict VNT in ACLF patients.

METHODS

Baseline liver stiffness measurement (LSM) and spleen stiffness measurement (SSM) were done using Fibroscan 630 Expert and esophagogastroduodenoscopy was performed to detect VNT. Varices with red color signs or large size were considered VNT. A model was developed and validated. Decision curve analysis (DCA) was used to assess net benefit (NB) of model at different threshold probabilities.

RESULTS

Of the 216 patients analyzed (males 95.4%, mean age 42.2 yr, model for end-stage liver disease [MELD] score 25.5, alcohol etiology 83.4%), 104 (48.14%) had VNT. Patients with VNT had higher SSM (kPa) (76.71 ± 17.54 vs. 60.56 ± 20.21, p = 0.006), LSM (kPa) (71.96 ± 6.87 vs. 61.93 ± 17.76, p = 0.001) and lower platelet count (103 /µl) (91.73 ± 17.71 vs. 126.94 ± 42.97, p < 0.001). A model incorporating these parameters had an area under the receiver operating characteristic (AUROC) of 0.842 and could spare esophagogastroduodenoscopy in 34.3% of patients with missing VNT rate of 4.8%. The results were internally validated by bootstrap analysis. On DCA, NB of model was higher for threshold probabilities range 0.07 to 0.8. The model had good calibration with Brier score of 0.15. It performed well in alcohol and non-alcohol related causes with AUROCs of 0.830 and 0.884, respectively.

CONCLUSION

A model comprising SSM, LSM and platelet count identifies VNT non-invasively in ACLF patients.