This document summarizes a study that evaluated methods for handling missing data when developing a predictive model for liver cancer recurrence. It assessed complete case analysis, complete variable analysis, and multiple imputation methods. The study found that using imputation methods could achieve better predictive accuracy than complete variable analysis, with one method called BPCA using just four features, including ones with missing values, to predict recurrence with 86% sensitivity and 88% specificity. Imputation methods were shown to still provide benefit for the predictive model even when handling data with missing values.