Abstract

Background: Hepatic encephalopathy (HE) is the most frequent cirrhosis complication leading to hospital admissions and is associated with significant mortality. The aim of this study was to determine the ability of the CyberLiver-Animal Recognition Test (CL-ART) to predict future hospitalisation due to decompensation, especially through HE, comparing its performance to established HE tests.

Methods: A prospective study of cirrhosis patients applying three different cognitive tests at two tertiary hepatology centres was performed. The CL-ART involved a timed (usually <30 seconds) recognition of animals using a smartphone app (Figure 1). EncephalApp Stroop Test and Psychometric Hepatic Encephalopathy Score (PHES) were chosen as test comparisons. Follow-up clinical data was collected for a 6-month period.

Results: 43 healthy controls and 103 cirrhosis patients were included (median CL-ART time 15.7s vs 24.0s). The baseline characteristics of the cirrhosis patients were 65% male, median age 58, Child-Pugh Score 8 [IQR 7-10], MELDNa 15 [IQR 11-19], CLIF-C AD 48 [IQR 44-52]. CL-ART demonstrated a good correlation with EncephalApp (r=0.81, p<0.001) and PHES (r= -0.63, p<0.001). When analysing patients admitted due to HE during their follow-up, baseline CL-ART was significantly higher compared to participants who were not hospitalised (31.5 vs 22.6s, p<0.001) with an AUROC of 0.85 (95% CI 0.77-0.93) for predicting future HE admissions. This was comparable to EncephalApp (AUROC 0.83, 95% CI 0.74-0.92) and ammonia (AUROC 0.81, 95% CI 0.71-0.91). In multiple logistic regression analysis, CL-ART remained an independent predictor of future HE admissions (OR 1.15, p=0.049). Using the Youden index, the optimal CL-ART cut-off to predict HE-related admissions is 26s (sensitivity 91.7%, specificity 71.4%). When analysing all subsequent admissions due to any decompensation event, baseline CL-ART scores were significantly higher in those subsequently hospitalised (27.0 vs 21.3s, p<0.001) with an AUROC of 0.76 (95% CI 0.66-0.85). Finally, the CL-ART also demonstrated superior participant useability (Figure 1).

Conclusion: This study demonstrates that CL-ART can help predict hospitalisation due to all decompensation, with highest sensitivity and specificity for HE-related admissions. Its rapid testing, smartphone application and high useability mean it can be used remotely, and therefore, play a crucial role in predicting decompensation, enabling early community intervention.