NEW YORK – A pan-cancer analysis involving tens of thousands of patients has spelled out new and known relationships between tumor mutations, survival outcomes, and responses to a range of cancer treatment strategies.
"This work demonstrates how computational analysis of large real-world data can facilitate precision oncology and generate insights and hypotheses in mutation-treatment interactions and mutation-mutation interactions in cancer," co-senior and corresponding author James Zou, a researcher at Stanford University, and his colleagues wrote in Nature Medicine on Thursday.
Using a gene-level analytical strategy, the researchers searched for ties between tumor mutation profiles, cancer treatment histories, and survival patterns. Their dataset included electronic health record (EHR) entries and targeted Foundation Medicine panel sequence profiles for hundreds of cancer-related genes in more than 40,900 de-identified cancer patients who are part of the Flatiron Health-Foundation Medicine clinicogenomic database.
The participants included more than 12,900 individuals with advanced non-small cell lung cancer, nearly 7,900 metastatic breast cancer patients, almost 3,900 individuals with ovarian cancer, some 3,500 patients with metastatic pancreatic cancer, and thousands more patients with advanced bladder cancer, renal cell carcinoma, or melanoma, the team explained, noting that the results were verified using data for nearly 3,900 additional advanced lung, breast, or colorectal cancer cases from an American Association for Cancer Research dataset.
"For each patient, we have data on tumor mutations, treatments received as first-line or further lines of therapy, real-world progression, survival outcomes, and detailed information on demographics, tumor stage, and laboratory values extracted from the EHRs," the authors explained.
The main analyses focused on overall survival and mutations occurring across specific genes, though follow-up work also considered progression-free survival and treatment discontinuation outcomes, along with certain mutation subtypes.
In the process, the team flagged 458 apparent mutation markers for survival in cancer patients receiving specific treatment protocols and uncovered specific mutations that typically co-occur with other tumor alterations.
The investigators found that mutations in 42 genes tracked with survival outcomes in at least one of the cancer types considered, for example. These genes, in turn, showed almost 100 significant interactions.
Consistent with past studies that showed ties between EGFR inhibitor resistance and KRAS mutations in advanced NSCLCs, they saw shorter-than-usual survival for KRAS-mutated cases treated with EGFR inhibitors and enhanced survival in EGFR inhibitor-treated advanced NSCLC patients with KRAS-wild type tumors.
Still other mutation-treatment ties turned up across cancer and treatment types, shoring up known relationships between tumor alterations, survival, and response to a range of chemotherapies or immune checkpoint immunotherapies and offering clues to new potential predictive biomarkers.
"Overall, our global analysis across multiple cancer types, treatment classes, and genes provides a more comprehensive picture of potential gene-treatment interactions," the authors reported, "and generates new hypotheses for future biological and clinical investigations."
Similarly, when the team looked at mutation-mutation interactions across genes such as ALK, BRAF, EGFR, MET, RET, ROS1, ERBB2, or PIK3CA that are currently targeted by US Food and Drug Administration-approved treatments, it found genes that were more or less likely to be mutated in conjunction with alterations affecting the targeted genes.
"Our findings demonstrate that high-quality, real-world clinicogenomic data from patients with cancer can be an important resource for investigating such mutation-treatment interactions by capturing outcome information of patients on diverse treatments," the authors concluded. "As tumor sequencing data become increasingly linked to the EHR, such data combined with careful computational analysis can greatly benefit precision medicine."