Development and validation of the AMMON-OHE model to predict risk of overt hepatic encephalopathy occurrence in outpatients with cirrhosis

•Neither PHES nor CFF are independent predictors of a first episode of OHE. •Ammonia levels above the upper limit of normal are independently associated with the risk of developing OHE. •The best model to predict a first episode of OHE (the AMMON-OHE model) includes sex, diabetes, albumin, creatinine and ammonia. •The AMMON-OHE model performs better than PHES and CFF to predict the first episode of OHE in outpatients with cirrhosis. •The AMMON-OHE model showed good predictive accuracy in two separate external validation cohorts. Background & Aims Neuropsychological and psychophysical tests are recommended to assess the risk of overt hepatic encephalopathy (OHE), but their accuracy is limited. Hyperammonaemia is central in the pathogenesis of OHE, but its predictive utility is unknown. In this study, we aimed to determine the role of neuropsychological or psychophysical tests and ammonia, and to develop a model (AMMON-OHE) to stratify the risk of subsequent OHE development in outpatients with cirrhosis. Methods This observational, prospective study included 426 outpatients without previous OHE from three liver units followed for a median of 2.5 years. Psychometric hepatic encephalopathy score (PHES) <-4 or critical flicker frequency (CFF) <39 was considered abnormal. Ammonia was normalized to upper limit of normal (AMM-ULN) at the respective reference laboratory. Multivariable frailty competing risk and random survival forest analyses were performed to predict future OHE and to develop the AMMON-OHE model. External validation was carried out using 267 and 381 patients from two independent units. Results Significant differences were found in time-to-OHE (log-rank p <0.001) according to PHES or CFF and ammonia, with the highest risk in patients with abnormal PHES plus high AMM-ULN (hazard ratio 4.4; 95% CI 2.4-8.1; p <0.001 compared with normal PHES and AMM-ULN). On multivariable analysis, AMM-ULN but not PHES or CFF was an independent predictor of the development of OHE (hazard ratio 1.4; 95% CI 1.1-1.9; p = 0.015). The AMMON-OHE model (sex, diabetes, albumin, creatinine and AMM-ULN) showed a C-index of 0.844 and 0.728 for the prediction of a first episode of OHE in two external validation cohorts. Conclusions In this study, we developed and validated the AMMON-OHE model, comprising readily available clinical and biochemical variables that can be used to identify outpatients at the highest risk of developing a first episode of OHE. Impact and implications In this study, we aimed to develop a model to predict which patients with cirrhosis are at risk of developing overt hepatic encephalopathy (OHE). Using data from three units and including 426 outpatients with cirrhosis, we developed the AMMON-OHE model – comprising sex, diabetes, albumin, creatinine and ammonia levels – which demonstrated good predictive ability. The AMMON-OHE model performs better than PHES and CFF to predict the first episode of OHE in outpatients with cirrhosis. This model was validated in 267 and 381 patients from two independent liver units. The AMMON-OHE model is available online for clinical use. Neuropsychological and psychophysical tests are recommended to assess the risk of overt hepatic encephalopathy (OHE), but their accuracy is limited. Hyperammonaemia is central in the pathogenesis of OHE, but its predictive utility is unknown. In this study, we aimed to determine the role of neuropsychological or psychophysical tests and ammonia, and to develop a model (AMMON-OHE) to stratify the risk of subsequent OHE development in outpatients with cirrhosis. This observational, prospective study included 426 outpatients without previous OHE from three liver units followed for a median of 2.5 years. Psychometric hepatic encephalopathy score (PHES) <-4 or critical flicker frequency (CFF) <39 was considered abnormal. Ammonia was normalized to upper limit of normal (AMM-ULN) at the respective reference laboratory. Multivariable frailty competing risk and random survival forest analyses were performed to predict future OHE and to develop the AMMON-OHE model. External validation was carried out using 267 and 381 patients from two independent units. Significant differences were found in time-to-OHE (log-rank p <0.001) according to PHES or CFF and ammonia, with the highest risk in patients with abnormal PHES plus high AMM-ULN (hazard ratio 4.4; 95% CI 2.4-8.1; p <0.001 compared with normal PHES and AMM-ULN). On multivariable analysis, AMM-ULN but not PHES or CFF was an independent predictor of the development of OHE (hazard ratio 1.4; 95% CI 1.1-1.9; p = 0.015). The AMMON-OHE model (sex, diabetes, albumin, creatinine and AMM-ULN) showed a C-index of 0.844 and 0.728 for the prediction of a first episode of OHE in two external validation cohorts. In this study, we developed and validated the AMMON-OHE model, comprising readily available clinical and biochemical variables that can be used to identify outpatients at the highest risk of developing a first episode of OHE.