Research News

Genomic sequencing of only tumor tissue could be misleading in nearly half of patients, study shows

BMJ 2015; 350 doi: (Published 16 April 2015) Cite this as: BMJ 2015;350:h2036
  1. Michael McCarthy
  1. 1Seattle

Clinicians who rely on genomic analysis of only a patient’s tumor tissue to guide cancer therapy could be misled by the presence of harmless mutations that are also present in the patient’s normal germline cells and are unrelated to the cancer, a study published in Science Translational Medicine has found.1

Increasingly, it has become standard practice to perform genomic sequencing of tumor tissue to identify cancer-causing mutations responsible for tumor development and growth. Of particular interest to clinicians are “actionable genes” whose presence can help identify the most promising treatment approach.

In the new study Victor Velculescu, professor of oncology and pathology and codirector of the Cancer Biology Program at the Johns Hopkins University School of Medicine in Baltimore, Maryland, USA, and colleagues compared the genomes of tumor and normal tissue from 815 patients who had 15 types of cancer, including breast, brain, renal, gastric, lung, pancreatic, blood, and melanoma.

When they looked at only the tumor genomes and filtered out well known germline changes they identified 382 possible genetic variations that seemed to be tumor related. However, when they then compared patients’ tumor genomes with the genomes of their normal tissues, 249 of the 382 possible tumor related variations turned out to be harmless genetic variations also found in the patients’ normal germline cells.

In the case of mutations in actionable genes the researchers found that, when they looked at tumor genomes alone, they identified on average 2.4 changes in each patient. However, when they compared these changes with those found in a patient’s normal tissue, 33% seemed to be false positives—that is, germline mutations unrelated to the cancer. In all, such false positives were present in 48% of the patients analyzed.

In a telephone briefing with reporters Velculescu said that, because each patient on average had multiple alterations of this type, false positive “actionable” variants appeared in about half of the patients analyzed. Therefore, the odds of correctly identifying actionable genes on the basis of a tumor only genomic analysis was “the equivalent of flipping a coin,” he said.

Velculescu noted a number of challenges to overcome, including cost, before clinics can routinely use analyses that include both a patient’s tumor and normal genomes. However, therapy based on inaccurate genetic information “can have substantial consequences ranging from serious side effects from inappropriate therapies, to lack of useful targeted therapies, and to increased costs of patient care from misguided medicines,” he added.

The study was done in collaboration with Personal Genome Diagnostics, a US company of which Velculescu and Luis Diaz, coauthor and associate professor of oncology at Johns Hopkins, are cofounders.


Cite this as: BMJ 2015;350:h2036


View Abstract

Log in

Log in through your institution


* For online subscription