Hippocrates would approve. The early days of personalized medicine are as much about "doing no harm" as about matching patients with the most appropriate therapies.

In cancer, a handful of molecular markers can help guide clinical decisions. For instance, tests can identify candidates for drugs such as imatinib (Gleevec). At the same time, markers are emerging that can identify patients who should not receive a treatment, and as new research suggests, this is valuable information.

"Patients are well served when physicians can forego a particular therapy and avoid the treatment's toxic effects," said Dr. Richard Schilsky of the University of Chicago, who writes about predictive biomarkers in the June Lancet Oncology.

"There really are no highly reliable biomarkers that predict a positive response to a cancer therapy," he continued. "The estrogen receptor for breast cancer is probably the best one going."

The biology of cancer is so complex that biomarker studies are more likely to find, say, a genetic change that causes resistance to a drug than a single biomarker that predicts a positive response.

One such marker of resistance is the gene KRAS in colorectal tumors. Patients with advanced tumors that carry KRAS mutations appear not to benefit from the addition of cetuximab to chemotherapy, as researchers reported this month at the American Society of Clinical Oncology (ASCO) annual meeting.

Patients with these mutations (one-third of those with the disease) should be spared the side effects and cost of cetuximab and a related drug, panitumumab, researchers said at the meeting. In Europe, cetuximab and panitumumab are approved for colorectal cancer only in patients with normal KRAS genes.

The new results, supported by other studies, were a tipping point. Within days, NCI and its Cooperative Group partners stopped enrolling patients on colorectal cancer clinical studies using cetuximab until this new information on KRAS can be appropriately incorporated into the trials.

As new drugs continue to expand treatment options, it will be even more important to know which ones to withhold.

"We are in an exciting era of targeted agents, and there are hundreds more in the pipeline," said Dr. Julie Gralow of the Fred Hutchinson Cancer Research Center at the ASCO meeting. "How are we going to afford these? One answer is that we need to identify patients who are likely to respond and [then] hold the money and hold the toxicity."

A goal for the field is to develop biomarkers in parallel with new drugs, so that the biomarkers can be validated in final-stage clinical trials. Validation and reliable testing is critical, because errors could deprive patients of a valuable treatment or expose them to unnecessary risk.

A decade after the approval of the breast cancer drug trastuzumab (Herceptin), researchers are still exploring the best way to identify candidates for the drug. Dr. Schilsky noted that while the current test is considered a positive predictor, only about 25 percent of advanced HER2-postive tumors respond to the drug. "So, the test may be better at identifying non-responders," he said.

Another biomarker test designed to help patients avoid an unnecessary treatment is Oncotype DX. A clinical trial (TAILORx) is asking whether the genomic test can identify women with ER-positive, early stage breast cancer whose risk of recurrence is sufficiently low that they can safely avoid chemotherapy after surgery.

Beyond what biomarkers can do for individual patients, many researchers believe they could speed the development of new drugs by enrolling patients who are "likely responders" in clinical trials. Fewer patients might then be needed to meet the goal of the study, and this could save money, as well.

Some companies are searching for biomarkers in preclinical studies, but large trials may be needed to find useful markers. The evidence on KRAS and cetuximab, for instance, came from 1,200 patients in many different trials.

Preclinical studies often don't produce biomarker candidates, and "you may only begin to relate a biomarker to efficacy after treating hundreds or thousands of patients," said Dr. Eric K. Rowinsky, the chief medical officer at ImClone Systems, which initially developed cetuximab. The company starts biomarker studies for all of its experimental drugs early in their development, he said.

Many experts believe it is only a matter of time before all patients with colorectal cancer have their tumors tested for KRAS mutations following surgery. It also seems likely that physicians will use combinations of positive and negative predictive markers in the future.

"The KRAS story is tremendous," said Dr. Rowinsky. "We've always known in oncology that most patients won't respond, and we have become accustomed to that. We have kind of shut our eyes to the reasons. This story tells us we should always look for the reasons patients do or do not respond to any therapy."

—Edward R. Winstead