Acurácia de métodos/modelos não invasivos para predição de varizes de esôfago em pacientes com doença hepática crônica avançada compensada secundária à doença hepática gordurosa não-alcoólica

Abstract

INTRODUCTION: Nonalcoholic fatty liver disease (NAFLD) is the most prevalent chronic liver disease worldwide and its aggressive form, nonalcoholic steatohepatitis, progresses to cirrhosis in about 20% of cases. In these there is an asymptomatic (compensated) stage in which esophageal varices are already present. However, no more than 20% of these patients have varices needing treatment (VNT). OBJECTIVE: to evaluate the accuracy of non-invasive methods/models for predicting esophageal varices in patients with compensated advanced chronic liver disease (cACLD) secondary to NAFLD. METHODS: this is a cross-sectional study that evaluated 21 patients with biopsy-proven cACLD secondary to NAFLD. Demographic, clinical and propaedeutic data were collected and it was evaluated the accuracy of liver stiffness measurement (LSM) by transient elastography (TE), platelet count/spleen diameter ratio (PSR), LSM-spleen diameter to platelet ratio score (LSPS), Varices risk score (VRS), spleen stiffness measurement (SSM) by TE, Baveno VI, Expanded Baveno VI and NAFLD cirrhosis criteria for predicting any varices and VNT. It was also evaluated the performances of these models for sparing EGD with risk of missing VNT less than 5%. RESULTS: mean age was 61 (± 6.6) years; 81% of the patients were female; 29% presented any varices and 14% presented VNT. Variables statistically associated with the presence of any varices were spleen diameter (p = 0.017), LSM (p = 0.011), LSPS (p = 0.003), VRS (p = 0.004) and Expanded Baveno VI criteria (p = 0.031). Variables associated with the presence of VNT were albumin (p = 0.012), LSM (p = 0.035), PSR (p = 0.033), LSPS (p = 0.007), and VRS (p = 0.003). For detection of any varices, LSPS and VRS presented AUROC above 0.9 (0.905 for both) and cutoff points were 1.18 and -4.35, respectively. The AUROC of LSM was 0.850 and cutoff point was 21.8 kPa. For detection of VNT, LSPS and VRS performed similarly (AUROC 0.961 for both) and cutoff points were 1.81 and -2.27. LSM presented AUROC of 0.889 and cutoff point of 21.8 kPa, too. The performance of all models was better to exclude than to diagnose any varices. The negative likelihood ratio (LR-) ranged from 0 to 0.71 and the positive likelihood ratio (LR+) from 2.08 to 6.99. For VNT, the LR- ranged from 0 to 0.75 and the LR+ from 3.0 to 16.9. LSPS and VRS made it possible to avoid 55% to 60% EGDs for any varices and 75% to 80% for VNT, with risk of missing no varices. Among the criteria with predetermined parameters, Expanded Baveno VI presented the best performance, allowing to spare 71% of EGDs with risk of 9.5% and 4.8% of missing any varices and VNT, respectively. CONCLUSION: LSPS and VRS presented excellent performance for both predicting varices and sparing EGD with no risk of missing any varices and VNT. Among the criteria with predetermined parameters, Expanded Baveno VI showed the best performance to avoid EGDs with minimal risk of missing any varices and VNT. LSM cutoff point was established and can be used for future comparisons.