Using tests to sequence the genome of a cancer patient's tumor without also testing their normal tissue could be leading doctors to prescribe tumor-attacking drugs that may be ineffective, a study suggests.

Looking at genetic changes in a tumor could misdirect treatment decisions for almost half of patients unless they are first compared with similar genetic examinations of their noncancerous tissues, researchers at the Johns Hopkins Kimmel Cancer Center say.

The comparison could eliminate false positives that can have doctors prescribing treatments that turn out to be ineffective, they say in a published report on their study in Science Translational Medicine.

Genomic testing of tumors is increasingly utilized to help doctors choose cancer therapies designed to target DNA changes in tumor cells that allow them to survive and spread.

The problem, the researchers say, is that the tests don't often sequence the DNA of a patient's healthy cells to determine what mutations a patient was born with as opposed to those that are unique to the patient's particular cancer.

In a study led by Dr. Victor Velculescu of Johns Hopkins, who is also a co-founder of Personal Genome Diagnostics, both tumor tissue and healthy tissue from 815 patients with various kinds of cancer were sequenced.

Two-thirds of the mutations detected in the tumors were also present in patient's healthy tissues, they found, suggesting they weren't driving the cancer.

"They were false positives," Velculescu says, explaining that not all genetic changes seen in a cancer are directly linked to that cancer.

Some of them, he says, are what are known as germline changes, inherited mutations in genes found in a person's normal tissues that differ from individual to individual.

"We all carry variations in our germline genome. They're part of what makes us individually unique," says study co-author Dr. Valsamo Anagnostou of the Johns Hopkins University School of Medicine.

"Only by comparing the genetic sequence of an individual's tumor and his or her own normal cells can clinicians know which changes are more likely to be cancer-related and which treatments are likely to work," he says.

However, the chief scientist at Foundation Medicine, a pioneer in tumor profiling, says he refutes the suggestion that sequencing of both tumors and normal tissue can improve results.

Most of the false positives in the new study are of unknown significance, Foundation's Phil Stephens says, and the genes his company's tests focus on are "unambiguous cancer drivers."

The Johns Hopkins researchers acknowledge there would be hurdles to implementing genetic analyses of both tumors and normal tissue in a clinical setting.

One is cost; gene sequencing of tumors can begin at several thousand dollars and more, a price that would increase if DNA of normal tissues was also sequenced.