# Identifying In-Hospital Risk Factors for Post-liver Transplant Seizures

## TL;DR

This study identifies risk factors for seizures after liver transplants and creates a predictive model to help clinicians manage patient care.

## Contribution

The study introduces a new predictive risk model for post-liver transplant seizures using in-hospital clinical variables.

## Key findings

- 10.6% of liver transplant patients experienced seizures within 30 days post-surgery.
- A predictive model with an AUC of 0.90 was developed using seven perioperative variables.
- The model achieved an accuracy of 0.72 and could help improve neurological outcomes.

## Abstract

Background

Liver transplantation (LT) is a life-saving intervention for end-stage liver disease; however, postoperative complications, particularly neurological issues such as seizures, pose significant challenges. This study aims to identify perioperative factors associated with seizures following LT and develop a predictive risk model. By recognizing these in-hospital risk factors, clinicians may tailor perioperative management to mitigate seizure risk and improve neurological outcomes.

Methodology

We conducted a retrospective observational study of adult patients who underwent LT at a tertiary referral center between January 2009 and January 2019. Data were collected for the perioperative period, spanning seven days pre-LT to 30 days post-LT. Variables included demographic, clinical, neurological, and liver-related data. Statistical analyses compared patients with and without seizures using appropriate tests for categorical and continuous variables. A predictive model for seizures was developed using in-hospital factors. It was then internally validated and evaluated for accuracy, sensitivity, specificity, and receiver operating characteristic curve. From this model, a practical clinical risk score was created.

Results

Of 376 patients, 40 (10.6%) experienced seizures within 30 days post-LT. The median age was 54.8 years, and 41% were males. Chronic liver failure was the primary indication for LT, with alcohol abuse, hepatitis C, and hepatitis B being the most common etiologies. Patients’ preoperative conditions included hepatocellular carcinoma (23%), chronic hyponatremia (22.6%), and prior kidney failure (22.1%). Portosystemic encephalopathy (PSE) was present in 46.5% of patients. Patients’ preoperative blood tests revealed low hemoglobin (mean: 10.8 g/dL, SD: 2.2), hyponatremia (mean: 134 mEq/L, SD: 5), elevated international normalized ratio (mean: 2.21, SD: 1.56), high bilirubin (mean: 8.74 mg/dL, SD: 10.23), and creatinine (mean: 1.41 mg/dL, SD: 1.95). Combined liver and kidney transplantation was performed in 9.2% of cases, while 15.4% were emergency procedures. Graft complications occurred in 25% of patients, with functional delay being the most frequent (14.5%). Immunosuppressive regimens included prednisone with tacrolimus or mycophenolate. Seizures typically occurred on post-LT day seven (interquartile range: 5-13) and included uncertain onset (40%), non-convulsive status epilepticus (30%), generalized seizures (22.5%), convulsive status epilepticus (5%), and focal seizures (2.5%). Treatment was administered to 85% of seizure patients, primarily with levetiracetam (75%). Using seven acute perioperative variables, i.e., age at transplantation, history of PSE, pre-LT epilepsy diagnosis, pre-LT hemoglobin, procedure duration, graft cold ischemia time, and intraoperative blood transfusion, we developed a risk score for post-LT seizures. This score achieved an accuracy of 0.72 (95% CI: 0.63-0.80) and an area under the curve (AUC) of 0.90.

Conclusions

This study identifies key in-hospital factors associated with seizures following LT and presents a predictive risk model based on clinical preoperative and surgical variables. With an AUC of 0.90, the model demonstrates strong discriminative ability, suggesting it is a robust tool for predicting seizures in the immediate post-LT period. Further prospective multicenter studies are needed to externally validate the model and risk score, thereby enhancing its clinical applicability.