Exploring algorithms to select candidates for non-selective beta-blockers in cirrhosis: A post hoc analysis of the PREDESCI trial

Background & Aims: Whether non-invasive tests (NITs) can accurately select patients with cirrhosis requiring non-selective beta-blockers (NSBBs) for clinically significant portal hypertension (CSPH) and prevention of decompensation is unclear. Our aim was to test the performance of NIT-based algorithms for CSPH diagnosis using the prospective PREDESCI cohort. We investigated whether a new algorithm combining NITs with endoscopy could improve performance. Methods: We included patients with compensated cirrhosis and available liver elastography who were screened during the trial. The performance of models based on liver stiffness measurement (LSM) and platelet count was evaluated. An algorithm considering endoscopy for patients with inconclusive results (the “grey zone”) was then developed and validated in an independent cohort of 195 patients in whom spleen stiffness was also available. Results: We included 170 patients from the PREDESCI cohort. An LSM ≥25 kPa alone (Baveno VII criteria) or combined with an LSM >20 kPa plus thrombocytopenia (AASLD criteria) ruled-in CSPH with positive predictive values of 88% and 89%, respectively. However, 37%-47% patients fell into the grey zone while at high risk of decompensation or death. Performing endoscopy in inconclusive cases identified patients with varices that, when reclassified as high-risk for CSPH, significantly reduced the grey zone to 22%. In this algorithm, 86% of patients with CSPH were correctly classified as high risk. The diagnostic performance was confirmed in the external validation cohort, where combining Baveno VII criteria with spleen stiffness showed similar accuracy to the model using endoscopy. Conclusions: Algorithms based only on LSM and platelet count are suboptimal to identify NSBB treatment candidates. Performing endoscopy in patients with indeterminate findings from NITs improved diagnostic performance and risk stratification. Endoscopy may be substituted by spleen stiffness for stratifying risk in the grey zone. Impact and implications: The PREDESCI trial demonstrated that non-selective beta-blockers prevent decompensation in patients with clinically significant portal hypertension (CSPH). Still, it is unclear whether we can select treatment candidates using non-invasive tests to assess the presence of CSPH without measuring HVPG (hepatic venous pressure gradient). In the prospective cohort of patients screened during the PREDESCI trial, we showed that algorithms based on liver stiffness and platelet count had suboptimal performance, mainly due to a high rate of indeterminate results. Performing endoscopy on patients in the grey zone significantly increased the number correctly characterized as having CSPH and improved the risk stratification for decompensation or death on long-term follow-up. These findings were validated in an independent cohort. In addition, a model using spleen stiffness instead of endoscopy showed similar diagnostic performance in the external validation cohort, suggesting that adequate risk stratification to select treatment candidates can be achieved with a fully non-invasive algorithm.