Prognostic Index for Localized Liver Radiation – Metastatic (PILLiR-M): Development and Analysis of a Clinical Prognostic Tool to Improve Patient Selection for Liver Directed Radiotherapy for Liver Metastases

Abstract

Liver metastases are common and associated with a high mortality rate. With improved treatment planning and delivery techniques, external beam radiotherapy (EBRT) is increasingly utilized for treating liver metastases. Many of these patients, however, are unlikely to live long enough to reap the benefits. The aim of this study was to develop a prognostication tool to aid in patient selection for liver directed EBRT in patients with liver metastases. A prospectively collected liver radiotherapy database was used to identify patients who had been treated with EBRT to the liver with local control intent. Exclusion criteria: single fraction EBRT, primary HCC or cholangiocarcinoma, benign pathology, lost to follow-up < 4 months. Pre-treatment patient and tumor characteristics associated with early death (<4 months) were identified using univariate analysis. Variables found to be significant were entered into a multiple logistic regression model using forward selection to identify factors independently predicting early death. Three fold cross validation and bootstrapping using 500 random samples was performed. The PILLiR-M was created based on this model. Two hundred twelve patients who had received liver EBRT were identified. One hundred nine patients were excluded (single fraction EBRT n=17, HCC n=63, cholangiocarcinoma n=22, benign pathology n=1, follow-up <4 months n=6). One hundred three patients remained for analysis with a 4 month mortality rate of 27.2%. Ascites, elevated bilirubin, low serum albumin, Child-Pugh score, ECOG performance status, non-colorectal primary, presence of extrahepatic disease and previous liver directed therapy were found to be significant predictors for early death on univariate analysis and were included in the multiple logistic regression model. Using forward selection to maximize the area under the curve, non-colorectal primary, ECOG, presence of extrahepatic disease and serum albumin were found to best predict patients unlikely to live longer than 4 months. A prognostic index (PILLiR-M) was created with 1 point for each of the following: non-colorectal primary, presence of extrahepatic disease, ECOG ≥2 and serum albumin<35g/L (AUC 0.852). Four month mortality was 0%, 3.1%, 43.6%, 66.7%, and 100% for patients with 0 (n=16), 1 (n=32), 2 (n=39), 3 (n=9), and 4 (n=3) points, respectively. Non-colorectal primary, presence of extrahepatic disease, poor performance status and decreased serum albumin were found to be risk factors for mortality less than 4months in patients being considered for liver directed EBRT for metastatic disease. We successfully created the PILLiR-M to aid clinicians in determining which patients are unlikely to live long enough to benefit from liver EBRT. Future work includes validating the PILLiR-M with an independent dataset.