Although there is no framework for prediction models for postembolization fever, potential influencing factors include demographic, clinical, laboratory, and radiologic data. We aim to develop and validate machine learning-based prediction models for postembolization fever after transarterial chemoembolization. Data from 1495 patients who underwent transarterial chemoembolization at a tertiary hospital were reviewed retrospectively. Seven machine learning-based algorithms were used to develop prediction models of postembolization fever occurrence after transarterial chemoembolization using SPSS WIN 27.0 and Python. The proposed ensemble method was the best algorithm for predicting postembolization fever. Variables positively correlated with postembolization fever occurrence were posttransarterial chemoembolization aspartate aminotransferase, C-reactive protein, alanine aminotransferase, and bilirubin levels, and international normalized ratio and platelet counts; pretransarterial chemoembolization aspartate aminotransferase and alpha-fetoprotein levels and platelet count; lipiodol and doxorubicin amounts; and a >5 cm tumor. Conversely, variables negatively correlated with postembolization fever were posttransarterial chemoembolization lymphocyte and monocyte counts and albumin levels; pretransarterial chemoembolization albumin levels and lymphocyte count; and probably hepatocellular carcinoma. Oncology clinicians should monitor demographic, clinical, laboratory, and radiologic data before and after transarterial chemoembolization to assess postembolization fever. In addition, health care professionals should be aware of the potential side effects of transarterial chemoembolization and management strategies, such as medications.
Kwon et al. (Fri,) studied this question.