A year ago I interviewed Deborah Fletcher, a 54-year-old manager at Deloitte who was fighting inflammatory breast cancer with all her professional skills. She carried a briefcase of spreadsheets, documenting treatments, bills, research, notes about who’d said what and what her plans were.

Like many patients with advanced cancer, Fletcher eventually put her hopes in an experiment. In her case it wasn’t a new drug but a new kind of test, called a liquid biopsy.

The test scans a person’s blood for small fragments of DNA released by tumors. The one she got, from Guardant Health, a California startup heavily financed by venture capitalists, looks for the DNA sequences of 68 well-known cancer genes and can reveal the mutations driving a person’s cancer.

Fletcher hoped these clues would lead her doctors to pick the right drugs and cure her. “She was extremely smart, and she had the hope that the tests would help,” says her mother, Patricia Cain. Fletcher had her tumor’s DNA profile taken six times and switched drugs twice. It wasn’t enough: she died in May.

Fletcher’s case is important because she was at the vanguard of what amounts to a global field study of the new DNA tests. Since laboratory diagnostics are regulated lightly in the United States, these tests have gone to market without consensus about their accuracy or data showing that they really help patients live longer.

“You have tests coming to market that are sometimes proven and sometimes unproven,” says Tycho Peterson, an analyst at J.P. Morgan who tracks the industry. “Commercial activity is increasing very quickly.” He estimates $20 billion a year in tests globally by 2020, up from about $100 million today.

So far, most insurers, including Medicare, don’t pay for these kinds of tests. They don’t think it’s their role to underwrite what looks like a research experiment. Health insurer Anthem labels the tests “investigational and not medically necessary.” Cigna calls them “unproven.”

Eventually, the most important use of liquid biopsies should be to catch signs of cancer early, before symptoms arise—when a surgeon can cure it by cutting it out (see “Spotting Cancer in a Vial of Blood”). Such screening could profoundly reshape cancer medicine.

For now, though, they are being used as “theragnostics”—that is, tests that guide decisions about treatment. If doctors can spot a patient’s key mutation, they can pick one of about 50 drugs designed to work directly against these DNA defects. The number of such “targeted” drugs is up sharply, from only around 15 in 2008.

Guardant’s test, launched a year ago, was the first liquid biopsy to be offered to doctors in the United States. It costs around $5,400. Since then at least eight other tests have been commercialized, including a liquid biopsy for lung cancer patients launched last week by Personal Genome Diagnostics in Baltimore. Other companies, including Illumina and Sequenom, have plans for their own tests.

So far the accuracy of the blood-based tests is in dispute. Foundation Medicine, a company in Cambridge, Massachusetts, that offers a DNA test for biopsied tissue samples, says they may yield false results and steer patients to the wrong drugs (see “Foundation Medicine: Personalizing Cancer Drugs”). “We get many calls from doctors with results from [a tissue biopsy] and then totally different results from a liquid biopsy,” says the company’s chief scientific officer, Phil Stephens. “The results are completely discordant and all over the map. That is bad for physicians and bad for patients.” Guardant CEO Helmy Eltoukhy, however, says his company has data from over 3,000 patients indicating that the two test types agree more than 90 percent of the time. “We don’t look at it as experimental,” he says.

Tissue tests haven’t escaped criticism, either. Johns Hopkins doctors connected to the Baltimore company have said tests like Foundation’s could potentially be wrong much of the time, leading patients to get a drug that won’t help them. Last year Illumina, the company that makes the sequencing machines used to perform the tests, formed a consortium to try to sort out which DNA mutations are really important to study and how to standardize results.

Whether DNA tests really help patients when they’re used to guide treatment is also a matter of dispute. Foundation Medicine issued a press release in May describing work by MD Anderson Cancer Center that compared 175 people who got its test, called FoundationOne. It said those who received a therapy matched to one of the mutations it uncovered lived three months longer on average than those who didn’t. At Guardant, Eltoukhy promises “a deluge of data” in coming months. “We have plans for large clinical trials, essentially to monitor patients in terms of outcomes and response. This is where the technology needs to go,” he says. “We think we are beyond accuracy—it’s about showing clinical utility.”

Experts say large, controlled studies will be needed, though. This month, the American Society of Clinical Oncology said it would launch a clinical trial of a dozen targeted drugs to see how they work when their use is guided by DNA tests. It’s also established a database to track treatment outcomes. The same day, the National Cancer Institute said it would launch a similar study of 20 drugs.

In the meantime, pressure to use the tests is growing, even if insurance won’t pay. Last year Google (which is an investor in Foundation Medicine) said all its employees would have the DNA test paid for if they needed it. And some patients have started directly seeking out doctors using the newest tests.

One of those people was Fletcher. She began her treatment near her home in Fairfax, Virginia. But gradually, she told me, she realized her doctors were picking her drugs on the basis of how they had performed, on average, in many women. “To say that you are going to fall into a statistic and that things are going to work out—I felt very uncomfortable with that,” she told me.

She decided she wanted the blood tests after hearing a radio program about them. She sought out Massimo Cristofanilli, an oncologist at Thomas Jefferson University, whose work had been featured on the show. Cristofanilli asked Fletcher when her breast had last been biopsied. It had been a year before. He looked surprised. The reason, he explained to her, is that cancer keeps changing.

Cristofanilli ordered both the Foundation Medicine test and the Guardant blood test. In this case, the results were in agreement: there was a new mutation in a gene called PIK3CA that he believed was making Fletcher’s cancer cells resistant to Herceptin, the drug she had been taking. So he switched her to a different drug, Afinitor. After a second blood test showed that the new mutation was gone and a different one had cropped up, she went back on Herceptin. At first, the angry red lumps on her left breast shrank so much that her case was featured in a medical report.

Cristofanilli says doctors in his clinic order more than 100 DNA tests every month; that’s a bill of half a million dollars. But the cost seems acceptable to him. A bottle of 30 Afinitor pills costs about $7,500. Herceptin costs $70,000 a year. That’s a lot to pay for a drug if it isn’t helping. “We want to interrupt some treatments, start others, and keep up with the ever-changing biology of the disease. I think that is the future,” he says.

I asked Cristofanilli if he thought the drug switches had extended Fletcher’s life. “I think in a way that she survived a year more,” he says. She’d come to him with a very serious cancer. But he says he can’t be sure. “I do not think we have the data yet to show an extension of life,” he says.

By last October, Fletcher’s cancer was getting worse again. Cristofanilli was out of new drugs to try. Fletcher decided to move to Florida to be near her parents. Her mother says she looked around for trials of experimental drugs but never found one to join. She tried acupuncture and massage.

The DNA tests had been her big chance. “If her condition had been less serious, I think she would have found the answer, with all the research that she did,” says her mother. “But she could not find the answer.”