"Liquid biopsies" for cancer: not ready for prime time

I've written many times about how the relationship between the early detection of cancer and decreased mortality from cancer is not nearly as straightforward as the average person—even the average doctor—thinks, the first time being in the very first year of this blog's existence. Since then, the complexities and overpromising of various screening modalities designed to detect disease at an early, asymptomatic phase have become a relatively frequent topic on this blog. Even more than ten years ago, I noted that screening MRI for breast cancer and whole body CT scans intended to detect other cancers early were not scientifically supported and thus were far more likely to cause harm than good. That was well over ten years ago. Now we have a company offering what it refers to as a "liquid biopsy" for the early detection of cancer. I fear that this is the recipe for the ultimate in overdiagnosis. I will explain.

The problem, of course, is that disease progression, including cancer progression, is not always a linear process, in which the disease progresses relentlessly through its preclinical, asymptomatic phase to symptoms to complications to (depending on the disease) death. There is such a thing as disease that remains asymptomatic and never progresses (at which point it's hard to justify actually calling it a disease). As I pointed out in my first SBM post on the topic, at least three-quarters of men over 80 have evidence of prostate cancer in autopsy series. Yet nowhere near three-quarters of men in their 80s die of prostate cancer—or ever manifest symptoms from it. This is what is meant by overdiagnosis, the diagnosis of disease that doesn't need to be treated, that would never cause a patient problems.

When teaching medical students and residents, I frequently emphasize that overdiagnosis is different from a false positive because overdiagnosis does diagnose an actual abnormality or disease. For example, ductal carcinoma in situ (DCIS) diagnosed by mammography leading to a biopsy is a real pathological abnormality; it is not a false positive. We just do not know which cases of DCIS will progress to cancer and which will not, leading to a question of how DCIS should be treated or at the very least whether we should treat it as aggressively as we do now, particularly given that the apparent incidence of DCIS has increased 16-fold since the 1970s, all of it due to mammographic screening programs and the increased diagnosis of DCIS and early stage breast cancer has not resulted in nearly as much of a decrease in the diagnosis of advanced stage breast cancer as one would expect if early diagnosis were having an impact in reducing the diagnosis of late stage disease.

Overdiagnosis would not be such an issue if it didn't inevitably lead to overtreatment. DCIS, for instance, is treated with surgery, radiation, and anti-estrogen drugs. Early stage prostate cancer used to be treated with radical prostatectomy, but now more frequently with radiation. Many of these men and women didn't actually need treatment. We just don't know which ones. This is why over the last six or seven years a significant rethinking of screening for breast and prostate cancer has occurred. There has been a backlash, of course, but the rethinking seems to have taken hold.

Not everywhere, of course.

A "liquid biopsy" for cancer

The concept of a blood test that can diagnose cancer is, of course, not a new idea. Indeed, prostate-specific antigen (PSA) screening for prostate cancer has been around for decades and is arguably just that: A diagnostic test used to diagnose or follow the progression or treatment of cancer. Ditto carcinoembryonic antigen (CEA) for colorectal cancer, CA19-9 for pancreatic cancer, and CA125 for ovarian cancer. Indeed, there are a wide variety of tumor markers, some useful, some not-so-useful, some worthless, that can be measured in the blood. Usually these are proteins or other chemicals produced more by specific cancers than by normal tissue that can be detected at elevated levels in patients with those cancers and often rise and fall with progression or response to therapy. For instance, for colon cancers associated with elevated CEA, CEA levels should fall to very low or undetectable levels after effective treatment and rising levels can indicate cancer recurrence.

These tests, as reliable or unreliable as they are, depending on the test, are associated with specific cancers. Even so, one question that always comes up whenever tumor marker levels rise is whether the recurrence—if recurrence there is—is a local recurrence where the tumor was resected or a distant (metastatic) recurrence. Local recurrences are often treatable for cure, albeit at much lower success rates than for a primary cancer, while metastatic recurrences usually indicate incurable disease. Another problem is that rising levels of tumor markers often precede detectable recurrence, resulting in a diagnostic challenge: Where is the recurrence? In the case of CEA, for instance, this problem has resulted in various clever strategies to label antibodies against CEA with radiotracers and then using intraoperative probes to seek where the antibodies bind. Even so, demonstrating improved survival targeting very early recurrences using such methods has been problematic. Indeed, it has been very difficult to demonstrate that finding very early asymptomatic cancer recurrences in general, regardless of the method used to detect them, improves survival. That is why we no longer do nearly as many routine scans after seemingly successful cancer treatment as we have in the past. In the vast majority of cases, it doesn't improve survival, but it does cause all sorts of problems.

Enter CancerIntercept™ by Pathway Genomics

With that background in mind, how does a cancer clinician approach the recent press release by Pathway Genomics, which announced that "Pathway Genomics Launches First Liquid Biopsy Test To Detect Cancer-Associated Mutations In High-Risk Patients"? Here are some of the claims:

Pathway Genomics, a global precision medical diagnostics company, announced the launch of CancerIntercept™, its first liquid biopsy, a non-invasive screening test designed for early cancer detection and monitoring, for as low as $299.

The test detects mutations that are commonly associated with lung, breast, ovarian, colorectal cancers and melanoma, as well as mutations that occur less frequently in other cancer types (such as pancreatic, head and neck, thyroid, gastric and prostate cancers). View an animation of how CancerIntercept works here.

The test is offered for two general populations: CancerIntercept™ Detect is the first liquid biopsy designed to detect tumor DNA in high-risk but otherwise healthy patients; CancerIntercept™ Monitor monitors patients with active or previously diagnosed cancer. Both programs use advanced DNA analysis to identify small DNA fragments that are shed from cancer cells and released into the bloodstream. The tests analyze the presence of 96 frequently occurring DNA mutations in nine cancer genes.

"Early detection is the single most important factor in ensuring successful treatments and improved survival rates," said Jim Plante, CEO and founder of Pathway Genomics. "Cancer patients and those at risk for the disease can take proactive steps to safeguard their health and fight back against some of the most virulent forms of the disease."

In addition, with CancerIntercept Monitor, physicians are able to supplement more invasive tissue biopsies and scans with liquid biopsies to monitor cancer treatment efficacy, disease progression and recurrence. CancerIntercept Monitor can also be ordered with personalized Clinical Trial Matching for later stage cancer patients.

"Rising levels of tumor DNA may indicate progression of the cancer before there is clinical or imaging evidence of tumor growth" said Dr. Glenn Braunstein, MD and Chief Medical Officer of Pathway Genomics. "Our liquid biopsy tests may also detect new mutations that occur over time and signal that the patient is becoming resistant to current therapy."

Naturally, Pathway Genomics has produced a very sophisticated animated video to tout the benefits of its new product:

In the press release, Pathway Genomics states that the test can be either ordered through a patient's own physician or, helpfully, it has an "online physician referral network," presumably of physicians willing to order the test if the patient's physician is not. Of note, the $299 price is for patients receiving the test on a "subscription" service, with repeat tests done at quarterly intervals. For just one test, the price is $699; annual, $499, semi-annual, $399. In other words, the company is incentivizing patients to be tested more frequently by accepting a smaller profit margin in return for more frequent testing. As we will see, that there is a severe lack of compelling scientific evidence that quarterly testing with CancerIntercept™ Detect benefits patients more than semi-annual or yearly testing matters not at all. As we will also see, that there is an even more disturbing lack of compelling evidence presented that testing with CancerIntercept™ Detect decreases mortality rates from cancer even in high risk patients matters not at all.

A "cancer stethoscope," dissected

Featured prominently in the promotional video, press release, and company website is a reference to CancerIntercept™ Detect and CancerIntercept™ Monitor as a "stethoscope" for cancer. It sounds like an awesome marketing simile, at least if you don't think about it too closely. After all, CancerIntercept™ Detect is first a "yes-no" test: Is there DNA containing a cancer-specific mutation in the blood or is there not? Then it is—if you accept the company's advertising materials—a diagnostic test. Which cancer-associated mutation(s) is there in the blood and at what level? A stethoscope, on the other hand, simply amplifies what can be heard without it. Before stethoscopes existed, doctors could simply put their ear to a patient's chest or back. Moreover, interpreting heart sounds and breath sounds is a skill that requires a great deal of practice to master. It's an essential skill for primary care physicians and many other specialties. Of course, it's quite clever of Pathway Genomics to liken its test to this essential physician tool, but it's a very imperfect analogy.

Be that as it may, let's look at some of the specific claims made by Pathway Genomics. Which claims have merit? From my perspective, only one of the claims might have merit, while the rest are highly questionable at best. To understand which one, you must first know what, exactly, these tests do. Cancer cells release small fragments of DNA into the bloodstream, known as cell-free DNA or cfDNA. For CancerIntercept™ Detect, cfDNA in the blood is tested for the presence of circulating tumor DNA (ctDNA) by screening for specific cancer-associated mutations using polymerase chain reaction (PCR) to amplify both the mutant and wild type DNA, followed by a "specific enrichment of the mutant and simultaneous removal of the wild type DNA by using a proprietary technology," after which the "mutant DNA is sequenced on Illumina's next-generation sequencing platform." Pathway Genetics claims that the test can detect two mutant copies of DNA per 5 ml blood sample. I always get a bit worried when I hear of a "proprietary" method to enrich mutant DNA compared to wild type (i.e., "normal") DNA, because I can't evaluate whether it's an accurate and reliable method or not.

So what does this test detect? The test examines 96 mutation "hotspots" among nine cancer "driver" genes (i.e., genes that, when mutated, can cause cancer or contribute to cancer progression). The genes tested for are involved primarily in melanoma, breast, ovarian, lung, and colorectal cancers, and for those of you with the knowledge and inclination to evaluate them, these specific mutations in the following genes: BRAF, CTNNB1, EGFR, FOXL2, GNAS, KRAS, NRAS, PIK3CA and TP53. You don't have to know all these genes for purposes of this discussion, but I list them for those who might be interested. Basically, to me it sounds like a shotgun approach to a bunch of mutations without anything resembling adequate clinical validation to justify its recommended use.

For whom is CancerIntercept™ Detect intended?

Helpfully, Pathway Genetics includes a "white paper" that enumerates what its tests do and what, according to the company, the supporting evidence is for their indications. Unfortunately, it's mighty thin gruel. According to the white paper, CancerIntercept™ Detect can be used thusly:

For preventative surveillance of high-risk populations in order to screen for mutations that could indicate early disease. Examples of high risk populations include, but are not limited to:

• Significant family history of one of the cancers that are known to harbor the mutations in the genes within the screen.
• Known genetic predisposition for one of the screened cancer types.
• Personal history of smoking, or prolonged exposure to second-hand smoke.
• Exposure to known carcinogens, like radon.
• Prolonged radiation or UV light exposure.
• History of hormone use (fertility drugs, progestogen-containing hormone replacement therapy).
• Certain reproductive and dietary factors.
• History of certain infectious diseases or chronic health conditions.

(Cancer types included in the screen: breast, colorectal, lung, melanoma, and ovarian cancers).

In a FAQ, Pathway Genomics states that CancerIntercept™ Detect should be considered for basically the same sorts of patients and includes a link to its clinical questionnaire. As a questionnaire to determine the appropriateness of a screening test, it is quite risible. For instance, the very first question is "How would you rate your risk of cancer?" followed by a 1 to 10 scale, with 10 being the highest risk. If there's one thing that most people don't have a good handle on, it's their actual risk for cancer; those at high risk tend to underestimate their risk while those at low risk tend to overestimate their risk. For purposes of a screening test—and, let's face it, that's how Pathway Genomics is marketing this test, as a screening test—what the patient thinks his or her risk of cancer is really doesn't matter, except perhaps as a marketing tool for a test like this. What really matters are the parts of the clinical history that are predictive risk factors.

In fairness, the rest of the questionnaire does ask about family history, previous hereditary cancer tests, health conditions associated with a higher risk of cancer, and various environmental exposures, some common (alcohol, smoking), some less so (asbestos, nickel, chromium, silica dust, etc.). The problem is that nowhere on the form or in Pathway Genomics' online literature is there an indication of a risk level as assessed by the test where CancerIntercept™ Detect wouldn't be appropriate. If, for instance, a prospective patient were to check no relevant exposures, no family history of cancer, no diseases associated with increased risk, and that he or she was a teetotaler who never touched alcohol or tobacco, would the company refuse to run the test, even though that person has at most average risk for cancer but probably actually has lower than average risk? Somehow, I tend to doubt it. That is unlike, for instance, the OncoType DX test for breast cancer. If the patient is inappropriate for the test (her tumor is estrogen receptor negative or HER2-positive, for instance), the company will not run the test. Worse, in the white paper, there is no direct clinical trial evidence cited showing that this particular test decreases mortality for cancer or even results in earlier diagnosis. As is frequently the case with various alternative medicine modalities, there are lots of references cited; it's just that many of them are not directly relevant, nor do they support many of the claims being made, particularly for CancerIntercept™ Detect.

For whom is CancerIntercept™ Monitor intended?

CancerIntercept™ Monitor is a similar test but intended for patients who have already been diagnosed with cancer. With this test the company is on a bit more solid ground than it is with CancerIntercept™ Detect, but still, in my estimation, makes a number of claims that are not fully supported by the evidence. According to the same Pathway Genetics white paper, CancerIntercept™ Monitor can be used thusly:

• Monitor residual disease in patients with known mutations in the primary tumor.
• Monitor treatment efficacy in patients.
• Monitor disease progression and tumor evolution (i.e. development of tumor resistance).
• Help the physician explore other options of treatment when the patient is resistant to current therapies.
• Provide an alternative method for biopsy when tissue is difficult to obtain or not available, or when the primary site of metastatic disease is unknown.
• Provide an alternative method for biopsy when the quantity of tissue obtained in a biopsy sample is limited and traditional molecular genotyping is requested.
• Provide prognostic information for some patients.

I can see the potential utility of genomic tests applied to "liquid biopsies" of tumors. So can a lot of other investigators. It is true that, if a primary tumor has already been diagnosed, monitoring the mutation-containing DNA fragments in the blood might well be useful for identifying potential driver mutations and following response to therapy via a blood test rather than other more difficult tests. That's why liquid biopsies are being tested in clinical trials. Indeed, Pathway Genomics lists a number of such clinical trials examining the prognostic value of various mutations screened for by the CancerIntercept™ Monitor and Detect tests in terms of recurrence, survival, and response to treatment, as well as correlation with the actual presence or absence of the mutation in the tumor itself. These tests could potentially be used to support the utility of a test like CancerIntercept™ Monitor in identifying specific mutations in patients with an already diagnosed cancer of a type that often harbors mutations tested for by the test.

My concern, however, is the lack of specific evidence that this specific test does what the company claims it does and is useful for the indications in patients with cancer that the company lists. Again, given that the enrichment process is proprietary, it's impossible for me—and any other scientist with a molecular biology background—to satisfy ourselves that the methodology is sound, at least not without seeing some rather specific evidence from each step in the assay to show that it does what is claimed for a large number of samples and doesn't over- or under-enrich for ctDNA.

The FDA steps in

Not surprisingly, the Food and Drug Administration (FDA) took an interest in Pathway Genomics' claims after the splash it made with its announcement earlier this month, so much so that it sent a letter last week to Pathway Genomics Founder and CEO Jim Plante. Its letter doesn't mention CancerIntercept™ Monitor, likely for the reasons I mentioned above. After all, there is an argument to be made that, if the test works as described, it is only combining tests for known cancer-associated mutations in ctDNA, although it must be pointed out that such a test is still only useful in the context of a clinical trial. In any case, that's probably not a hill the FDA wanted to die on, at least not first, which is probably why instead the FDA does express a considerable amount of concern regarding CancerIntercept™ Detect. Specifically, in its letter the FDA expresses the very same concerns that I did:

We have conducted a review of our files and have been unable to identify any Food and Drug Administration (FDA) records reflecting the approval, clearance, or listing of these devices. We have also examined published literature and have not found any published evidence that this test or any similar test has been clinically validated as a screening tool for early detection of cancer in high risk individuals. We have reviewed the information presented on your website in the white paper, entitled "Liquid Biopsy for the Detection and Monitoring of Cancer: Analysis of 96 Hotspot Mutations via Plasma Derived Circulating Tumor DNA," dated September 2015. It is unclear how the literature that you cited, addressing the presence of circulating tumor DNA (ctDNA) in already-diagnosed patients, is adequate to support the expansive claims of screening for early cancer detection using ctDNA in undiagnosed patients for up to 10 different cancers with the CancerInterceptTM Detect.

I agree. The literature cited by Pathways Genomics is in no way adequate to support the claims made for CancerIntercept™ Detect. I'd also add that the literature cited would not be adequate (to me, at least) to support the claims made for CancerIntercept™ Monitor, either, although for CancerIntercept™ Monitor the disconnect is much less egregious. Be that as it may, screening an asymptomatic population is a very different thing from monitoring a tumor marker in an already-diagnosed patient with cancer.

There are a number of perils and pitfalls in using tumor markers to follow cancer patients for response to treatment and recurrence. It's not as though we don't have considerable experience with them. Indeed, the scientific literature surrounding CEA and PSA, which have often been used for this purpose, is replete with the difficulties and complexities of following just a single molecule for a single cancer known to produce it. What does a rising level of a ctDNA with one or more of the mutations tracked by CancerIntercept™ Monitor mean? What should a clinician do? How sensitive is the test for detecting recurrences of the cancers covered by the test? How specific? How do you use the information to guide treatment, or do you use it at all? After all the hype, precision medicine is in its infancy and a recent clinical study (SHIVA) published in The Lancet Oncology failed to find a survival benefit using genomic profiling and targeted treatment in patients with advanced cancer, leading the authors to specifically recommend against using targeted therapies off-label for cancers for which they are not indicated based on the presence of mutations targeted by the drugs (perhaps a topic for another post).

As for monitoring for recurrence in a patient successfully treated for cancer, perhaps Pathway Genetics should consider that, of the original "Choosing Wisely" guidelines from the American Society of Clinical Oncology, one of the guidelines against scientifically unsupported treatment stated, "For individuals who have completed curative breast cancer treatment and have no physical symptoms of cancer recurrence, routine blood tests for biomarkers and advanced imaging tests should not be used to screen for cancer recurrences"—exactly the indication Pathway Genomics proposes for CancerIntercept™ Monitor, particularly given that in 2013 ASCO widened the recommendation to avoid advanced imaging as routine follow-up for pretty much all cancers unless there is high level evidence supporting it.

These issues are complicated enough when monitoring a patient known to have cancer and whose ctDNA mutation profile might agree with the mutation profile of his or her cancer, but start screening an asymptomatic population, even a "high risk" population, for ten cancers using a test like CancerIntercept™ Detect, and the issues and complexity skyrocket.

We know that our bodies are always producing cells with various mutations that could grow into cancer. Usually, these cells either die or the body eliminates them. Often, these tiny cancers never progress beyond a tiny cancer because they don't acquire the ability to induce the body to provide them with a blood supply. Without a blood supply, they remain limited by how far oxygen will diffuse in plasma (less than 1 mm) and never grow larger. As I described a long time ago, most men over 80 have tiny prostate cancers. Thyroid cancer is an uncommon cancer, but a Finnish study found evidence of papillary thyroid cancer in 36% of adults in an autopsy series. We're constantly producing tiny cancers that either don't progress beyond microscopic size or regress. How often does CancerIntercept™ Detect identify these tiny cancers in normal people?

More importantly, what do we do if these ctDNA fragments with cancer-associated mutations are found in an asymptomatic patient who just got the test because the company's marketing made it sound as though he should? These mutations are not cancer type-specific. They're found in many kinds of cancer. A clinician would have little idea where to look, particularly for the more "promiscuous" mutations found in many types of cancer. Will clinicians do PET scans, whole body CT scans, and/or MRIs in a frantic search for a tumor that might never harm the patient? Would finding tumors, be they from breast, prostate, the GI tract, ovaries, or elsewhere result in lower mortality from those cancers, or would we see a massive wave of overdiagnosis, with subsequent overtreatment, based on these findings? I don't know for sure, but my prediction is the latter. Think of it this way. We have significant overdiagnosis and overtreatment from far less sophisticated and sensitive tests, like mammography and PSA. Imagine how much more we would be likely to see from a test like CancerIntercept™ Detect.

Think of it this way: If a patient had an abnormal CancerIntercept™ Detect test, he would almost certainly be imaged up, down, left, and right, undergo blood tests and possibly invasive tests like colonoscopy (for GI cancer-associated mutations). Then, if all of those were negative, there would be pressure to "do something," possibly even administer targeted agents or chemotherapy, because watchful waiting is not something most physicians and patients are comfortable with for an unknown cancer.

Pathway Genomics just doesn't "get it"

Does Pathway Genomics "get" these issues? If its response to the FDA is any indication, the answer is a resounding no:

We are carefully considering the concerns of the FDA as stated in their letter, and we will be responding to that letter. We assure that there is physician involvement in the ordering, review and follow-up of CancerIntercept™ testing. We believe that CancerIntercept™ Detect is a laboratory developed test and, as a CLIA and CAP certified clinical laboratory, we are offering it as such. While Pathway Genomics is involved in educating and marketing the tests to physicians and consumers, we do not believe this is a direct-to-consumer model. We believe we have performed appropriate validation of the test as a laboratory developed test, and we are in the process of performing additional studies.

Now you know why Pathway Genetics has its own network of physicians for this test, to avoid the "direct-to-consumer" model while still being able to do the test for pretty much any consumer who wants it. Moreover, while Pathway Genomics might believe that it has "validated" its test sufficiently as a "laboratory-developed test" (LDT), in my opinion it hasn't validated its test anywhere near adequately as a screening test for the general population, "high risk" or not.

Why is Pathway Genetics emphasizing that it considers its test a "laboratory-developed test" in a CLIA- and CAP-certified clinical laboratory? Dr. J. Leonard Lichtenfeld, Deputy Chief Medical Officer for the American Cancer Society, describes in a blog post that echoes many of the concerns I have about these tests:

The test in question is only done by this one company, measuring over 90 genetic abnormalities in the blood sample. And, since this is a proprietary laboratory test it can be offered directly to the public and the company can make whatever claims it chooses to consumers and health professionals–who may not even be aware there is no oversight by the Food and Drug Administration to verify the claims made for the test. (The Centers for Medicare and Medicaid services oversees the general quality of the laboratories offering these tests, but does not do specific review of a test in these situations).

This is basically the situation right now, but the FDA's policy towards these tests could well be changing, as its regulations are from an era before tests like CancerIntercept™ Detect and CancerIntercept™ Monitor (Jann Bellamy has discussed the changing role of the FDA in lab-developed tests before). As the FDA described on its website, the FDA "does not consider diagnostic devices to be LDTs if they are designed or manufactured completely, or partly, outside of the laboratory that offers and uses them," explaining, "While the uses of an LDT [laboratory developed test] are often the same as the uses of FDA-cleared or approved in vitro-diagnostic tests, some labs may choose to offer their own test. For example, a hospital lab may run its own vitamin D assay, even though there is an FDA-cleared test for vitamin D currently on the market."

In other words, back in the day, LDTs were generally simple tests, such as for vitamin D levels, electrolyte levels, or other commonly measured parameters. They were not complex genomic tests involving enrichment of plasma for specific types of DNA and then next generation sequencing of the resulting fraction. As a result, the FDA has provided draft guidance for the development of a regulatory framework for LDTs, invited public comment, and held a public workshop earlier this year for public input.

Tests used to diagnose or guide treatment of a disease, particularly cancer, should have strong scientific evidence supporting their indications, accuracy, and efficacy. Barring that, results from tests like the ones marketed by Pathway Genomics can confuse far more than illuminate our decisions on how to treat patients. As Dr. Lichtenstein puts it, "Take an abnormal result from this test to your doctor and the real possibility exists they will have no clue what to do with it." Exactly so.

He also notes, as I've noted before, that there will be testimonials that this test saved people's lives. Of course, as I've described so many times before, just because a patient believes a test (such as thermography for breast cancer) or a treatment (such as Stanislaw Burzynski's antineoplastons) saved her life, that does not make it so.

Tests like CancerIntercept™ Detect demonstrate just how urgent the need is for the FDA to modernize its regulation of laboratory-developed tests. Maybe Pathway Genetics is the perfect test case. Liquid biopsies will very likely be validated as a means of guiding therapy in cancer patients, but it's way, way too early to consider using them as screening tests for cancer in asymptomatic patients. Our experience with much simpler screening tests like mammography and PSA screening should tell us that much. The potential of genomic tests is huge, but such tests need to be validated by science before being offered to patients on a large scale.

And Pathway Genomics is far from alone in this field.