Accuracy of non-invasive methods/models for predicting esophageal varices in patients with compensated advanced chronic liver disease secondary to nonalcoholic fatty liver disease.

Introduction: Nonalcoholic fatty liver disease (NAFLD) patients can progress to cirrhosis. In these, there is a compensated stage in which esophageal varices can exist. However, no more than 20% of these patients have varices needing treatment (VNT).

Objective: Evaluate the accuracy of non-invasive models to predict esophageal varices, as well as their performance to avoid esophagogastroduodenoscopy (EGD) with a risk of missing VNT of less than 5%, in Brazilian patients with compensated advanced chronic liver disease (cACLD) secondary to NAFLD.

Methods: Twenty-one patients with biopsy-proven cACLD secondary to NAFLD were submitted to liver stiffness measurement (LSM) by transient elastography (TE), and data were collected to measure platelet count/spleen diameter ratio (PSR), LSM-spleen diameter to platelet ratio score (LSPS), varices risk score (VRS), Baveno VI, Expanded Baveno VI and NAFLD cirrhosis criteria.

Results: The mean age was 61 (±6.6) years, and 81% were female; 14% presented VNT. For detection of VNT, LSPS and VRS performed excellently, with an area under receiver operating characteristic (AUROC) of 0.961 for both. LSM presented an AUROC of 0.889 and a cutoff point of 21.8 kPa. LSPS and VRS enabled sparing 75-80% of EGDs for VNT, with no risk of missing varices. Expanded Baveno VI enabled sparing 71% of EGDs, with 4.8% risk of missing VNT.

Conclusion: LSPS and VRS performed excellently in both predicting VNT and sparing EGD, and Expanded Baveno VI showed good performance in sparing EGDs, with acceptable risk of missing VNT. An LSM cutoff point was established and had good performance.