Location matters, especially when it comes to tumors in colon cancer. The issue of “sidedness” – whether tumors are located on the right or left side of the intestine – and what impact that has on the progression of metastatic colorectal cancer (CRC) has been a topic of discussion in the clinical community for several years. A complete understanding of sidedness has not been well understood by researchers and clinicians. Is sidedness an independent driver, confounded by other factors, or a simply a surrogate for other determinants? By leveraging artificial intelligence (AI) and causal modeling, researchers have now identified a comprehensive set of clinical and molecular drivers of survival for patients with metastatic colorectal cancer, leading to a better understanding of the role of sidedness in CRC.
The discovery is a step towards better understanding the drivers of prognosis and progression that can enable better patient care. Being able to measure these drivers at baseline will allow for better risk stratification at diagnosis.
Colorectal cancer is the third most common cancer diagnosed in the U.S. and it is estimated that 140,000 new cases will be diagnosed in 2018. The American Cancer Society estimates a lifetime risk of developing the disease at 1 in 22 for men and 1 in 24 for women. Metastatic, or stage IV colon cancers, have a five-year relative survival rate of about 11 percent.
Discovering key insights with causal modeling
Researchers manually reviewed the records of 3,000 patients to try to determine the validity of the “sidedness” theory in CRC. They quickly realized that analyzing the millions of possible models for CRC outcomes was beyond the capacity of basic statistics. They needed a much more powerful engine to speed up the process and gain true insight into the issue. AI supplied that firepower.
GNS, working with the Alliance for Clinical Trials in Oncology (Alliance), applied its Causal Machine Learning (CML) platform REFS to the clinical data collected during the CALGB 80405 (Alliance) phase III clinical trial. Within three months, REFS was able to provide hypothesis-free evidence to support the results that the location of the primary tumor within the colon plays a central role as an independent driver of overall survival in patients with metastatic cancer.
The Alliance and GNS presented a poster of the analysis at the ASCO 2018 annual meeting on June 4-5. The results revealed that primary tumor side is an independent driver of survival outcomes and is not confounded by any other clinical variable. The analysis also showed that aspartate aminotransferase (AST), hemoglobin concentrations, metastatic sites, BRAF mutation status, RAS mutation status, and expression of an angiogenesis-related gene signature are independent drivers of survival outcomes and can serve as potential biomarkers.
The work continues
The GNS/Alliance collaboration is ongoing and includes causal modeling of the rich molecular data from the CALGB 80405 (Alliance) study including gene expression, somatic mutation, microsatellite instability and blood-based biomarker data. The collaboration is focusing on uncovering the specific biological mechanisms that explain why patients with right-side tumors have a worse prognosis than those with tumors on the left side, and on identifying biomarkers that can predict the optimal drug for each patient.
The results from this collaboration showcase the power of causal AI to discover disease and drug mechanisms directly from multi-modal patient clinical trial data. Using this innovative technology, researchers are able to cut through noise to gain an understanding of underlying causes of CRC.
Additional validation and clinical follow up is needed, but the initial results of the collaboration offer a promising first step toward improving treatment effectiveness for CRC patients and will hopefully help find a cure for the disease.
For further information, check out the ASCO/GNS research poster here.
Reprint: GNS Healthcare Blog
For other articles in this issue of the Alliance E-News newsletter, see below.
