Editor's Note: (H. Gilbert Welch is professor of medicine at the Dartmouth Institute for Health Policy and Clinical Practice and the author of "Less Medicine, More Health -- 7 Assumptions that Drive Too Much Medical Care." The opinions expressed in this commentary are his own. )
(CNN) A simple blood test to detect cancer early. How great is that?
There has been enthusiasm about the so-called "liquid biopsy" for years. In mid-January, however, doctors learned more -- both about this vision and its problems.
A widely reported study in the journal Science described a liquid biopsy test -- CancerSEEK -- which combined measuring eight tumor biomarkers with testing for pieces of DNA with cancer associated mutations in 16 genes.
It's not one test; it's a battery of tests. And while collecting the blood may be simple, the subsequent analysis is extraordinarily complex.
The task at hand is particularly challenging. We all have pieces of DNA in our blood. Distinguishing the tumor DNA from the background DNA requires finding the mutations specifically associated with cancer.
Adding to the complexity, healthy individuals can have mutations. To avoid labeling innocuous mutations as cancerous requires a bunch of statistical fine-tuning.
In other words, there are a lot of steps in a liquid biopsy and much potential for things to go awry.
To their credit, the CancerSEEK investigators were very forthright that the study conditions were ideal for the test to accurately detect cancer. The liquid biopsy simply had to discriminate between patients with known cancer (the majority of whom had symptoms) and healthy individuals. And the statistical fine-tuning was tailored to the study participants -- with the knowledge of who had, and who did not have, cancer.
Although the test was able to detect most of the late-stage cancers, it detected less than half of the stage 1 cancers.
But doctors don't screen to find advanced cancer, we screen to find early cancer. And we don't screen people with symptoms of cancer, we screen people who don't have symptoms of cancer.
There's no doubt that there would be more detection errors in the less controlled environment of the real world.
Just how often was made clear in a recent JAMA-Oncology study. Forty patients with metastatic prostate cancer received liquid biopsies to tailor therapy in real time to the genetics of their spreading tumors. That's the vision for precision medicine.
But the investigators added a little twist. They wanted to know whether it mattered which lab the liquid biopsies were sent to. So they sent each patient's blood for two different commercial liquid biopsies: Guardant360 and PlasmaSELECT. Both tests were designed to detect mutations in the same genes.
Yet in over half of the 40 patients, the tests gave different answers about which mutations were present. Different liquid biopsy tests give different answers in a majority of patients? That's not precision, that's awful.
Sure, the analyses of liquid biopsies will improve. But if this much confusion exists about what mutations are present in the blood of patients with metastatic cancer (who have a lot of tumor DNA), imagine the uncertainty that will exist for asymptomatic individuals not known to have cancer -- the very people who would be screened.
And then there is the question of what to do with a positive result. This is very different than detecting a concerning lung nodule on a screening chest CT scan or a concerning breast mass on a screening mammogram. In these cases, it's clear what to do to get a definitive answer: surgically biopsy the nodule or the mass. But with a liquid biopsy, the anatomic location of a cancer can be a mystery. It may not even be clear what organ the cancer is in.
Imagine what this might mean for a patient: A doctor says, "It looks like you have cancer, but we are not sure where."
Even if there is certainty that the cancer is in, say, the liver, doctors may not know where in the organ. What to do then? Randomly biopsy different parts of the liver?
This is doubly concerning when screening average-risk individuals, because most positive results are expected to be false alarms. We typically learn that a screening test is falsely positive because a surgical biopsy is normal. But absent the knowledge of where to biopsy, how can we ever be sure a positive liquid biopsy is wrong?
Doctors won't know where to look, but we will keep looking. Liquid biopsies are a recipe for more health anxiety, more procedures, more complications and more overdiagnoses. Not to mention, more out-of-pocket costs for our patients.
Of course, we should continue to study liquid biopsies. The detection of circulating tumor DNA may ultimately prove useful in selected settings, such as tailoring therapy for aggressive cancers that are rapidly mutating. But the real enthusiasm is for screening average-risk individuals.
One reason is obvious: there is a lot of money to be made. A Goldman Sachs video estimated the potential liquid biopsy market to be $14 billion annually, adding "and we're just at the beginning." That kind of money doesn't come from testing the few patients with aggressive cancer, that comes from screening millions of people.
And there is a less obvious reason: it is easier for a new test to pass regulatory muster than it is for a new drug. While the FDA has a longstanding mandate to protect us from snake oil treatments, this often doesn't extend to snake oil testing.
The enthusiasm for finding things that might benefit people in the future ignores the fact that doing so can cause people to have problems now. In short, a bad test can do as much damage as a bad drug. Worrisome liquid biopsies will start a cascade of subsequent, not-so-simple tests and procedures. People will be hurt in the process.