Non-invasive ménage à trois for the prediction of high-risk varices: stepwise algorithm using lok score, liver and spleen stiffness

Background & aims: Liver stiffness (LS), spleen stiffness (SS) and serum markers have been proposed to non-invasively assess portal hypertension or oesophageal varices (EV) in cirrhotic patients. We aimed to evaluate the performance of a stepwise algorithm that combines Lok score with LS and SS for diagnosing high-risk EV (HREV) and to compare it with other already-validated non-invasive methods.

Methods: We performed a cross-sectional study including 136 consecutive compensated cirrhotic patients with various aetiologies, divided into training (90) and validation (46) set. Endoscopy was performed within 6 months from inclusion for EV screening. Spleen diameter was assessed by ultrasonography. LS and SS were measured using Fibroscan. Lok score, platelet count/spleen diameter ratio, LSM-spleen diameter to platelet ratio score and oesophageal varices risk score (EVRS) were calculated and their diagnostic accuracy for HREV was assessed. The algorithm classified patients as having/not-having HREV. Its performance was tested and compared in both groups.

Results: In the training set, all variables could select patients with HREV with moderate accuracy, the best being LSPS (AUROC = 0.818; 0.93 sensitivity; 0.63 specificity). EVRS, however, was the only independent predictor of HREV (OR = 1.521; P = 0.032). The algorithm correctly classified 69 (76.66%) patients in the training set (P < 0.0001) and 36 (78.26%) in the validation one. In the validation group, the algorithm performed slightly better than LSPS and EVRS, showing 100% sensitivity and negative predicted value.

Conclusion: The stepwise algorithm combining Lok score, LS and SS could be used to select patients at low risk of having HREV and who may benefit from more distanced endoscopic evaluation.