Quantitative CT Biomarkers to Predict Metastatic RCC Response to Antiangiogenic Therapy
Abstract
Background: Nearly 30% of patients present with metastatic renal cell carcinoma (RCC) at diagnosis, and an additional 7% to 32% develop metastatic disease after diagnosis. During the last decade, treatment of advanced and metastatic renal cell carcinoma (RCC) was revolutionized with the advent of targeted (molecular) therapies, but not all patients respond favorably. Some patients fail targeted therapy and are subjected to increases in tumor burden, drug toxicities, and costs from an ineffective treatment. It is currently not possible to identify which patients will respond to or fail therapy. It would be ideal to develop a biomarker that would confidently identify responders and nonresponders, so that nonresponders could halt an ineffective expensive and toxic therapy prior to disease expansion. Nonresponders could then be directed to a more effective therapy. Computed tomography (CT) imaging is widely available and used to evaluate treatment response in almost all patients with metastatic RCC. Evaluation of treatment response is complex, but we developed eMASS software, an image viewer that performs computer-assisted response evaluation, which is a method to standardize image evaluation, reduce errors and variability, capture quantitative information from CT images, and improve display and archiving of tumor response information. eMASS is also used to measure the Vascular Tumor Burden (VTB), or the amount of vascularized tumor. It turns out that metastatic RCC is highly vascular, and anti-vascular targeted agents are highly successful at prolonging survival and are designed to devascularize the tumors. We have shown that the changes in the VTB after only one cycle of anti-vascular targeted therapy are predictive of survival in patients with metastatic RCC treated with sunitinib (an anti-vascular agent), but it is unclear if the VTB will predict response to other anti-vascular targeted agents. Objective and Hypothesis: The first objective is to use eMASS to validate the VTB as an accurate and reproducible CT biomarker for predicting metastatic RCC response to three different anti-vascular agents. We hypothesize that changes in the VTB on initial post-therapy CT images are highly reproducible and accurately predict survival in patients with metastatic RCC treated with different anti-vascular agents. A second objective is to develop and validate a machine-learning algorithm to accurately predict survival on an individual basis (aka precision medicine). We hypothesize that a machine-learning algorithm that utilizes baseline clinical data and annotated data and images from eMASS will be highly reproducible and accurately predict survival in patients with metastatic RCC treated with different anti-vascular agents. Study Design: Our team has partnered with Pfizer, who is sharing the data and CT images from two previously completed landmark multi-institutional clinical trials (phase III studies) of metastatic RCC treated with different anti-vascular agents. The CT images from 818 patients were archived, and we know when all of the patients failed therapy and died. We will use eMASS to measure the VTB in order to see how well it predicts time to drug failure (progression) and overall survival. When we evaluate the CT images, eMASS is used to label (annotate) all the measured metastases on the CT images. We draw around the periphery of each lesion, and eMASS keeps track of the metastasis type, location, shape, and other features. We will combine this annotated data from eMASS and the laboratory and other clinical data available at baseline (prior to starting therapy) to develop a machine-learning (artificial intelligence) algorithm to predict response to targeted therapy and overall survival. We have partnered with Innolitics, commercial software engineers who built eMASS in collaboration with PI Andrew Smith MD, PhD, and who have extensive experience developing and validating image-based machine-learning algorithms.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Nov 19, 2019
- Source ID
- W81XWH1910764
Entities
People
- Andrew N. Smith
Organizations
- United States Army
- University of Alabama at Birmingham