Intention‐to‐treat survival benefit in liver transplantation: Comments on Lai et al.

model selections is no different, and effect estimates are largely unchanged. We also agree with Ridola and Riggio that the presence of minimal HE is predictive of incident overt HE. However, minimal HE is not routinely evaluated in clinical practice. Further, studies of minimal HE diagnostics, such as the animal naming test, uniformly excluded patients with psychiatric disorders, alcohol misuse in the previous 6 months, any psychoactive medication, and heart, respiratory, or renal failure. These comorbidities characterize roughly half of our cohort. The optimal cutoffs as well as their performance in real-world patients require future study. To effectively risk-stratify real-world patients in an intention-to-screen fashion, future studies of multimodal approaches, including our risk score and other modalities such as the EncephalApp and the animal naming test, are indicated. Ridola and Riggio raised three additional issues that deserve clarification. First, validation: we performed an internal validation using a bootstrapping method. Second, we excluded patients with a history of overt HE at baseline. Our goal was to predict incident overt HE and not recurrent HE as patients with a history of overt HE are known to be at high risk of recurrent HE. Third, albumin levels can vary and are subject to confounding by malnutrition and ascites. We agree. However, both malnutrition (associated with sarcopenia or zinc deficiency) and ascites (an indicator of severe portal hypertension) are also expected to be associated with the risk of HE.