BETA
This is a BETA experience. You may opt-out by clicking here

More From Forbes

Edit Story

Companion Diagnostics? For Cancer Care, We Need Better Ones

This article is more than 8 years old.

Maybe it takes a certain set of experiences like going to medical school and then having cancer, or geekiness, to get excited by spending two days at a biotech conference on companion diagnostics, CDx. But I came away from a recent, sold-out World CDx meeting in Boston eagerly reading. I couldn't learn enough about these medical match-makers.

Companion diagnostics are tests designed to predict if a person is likely to respond favorably to a specific, "companion" treatment, or not. They include a wide – and expanding – range of genetic, RNA and protein measurements, along with novel ways to evaluate cancer cell components and activity, like stickiness or the capacity to repair DNA. These biomarkers should link patients to optimal therapy (when such exists), and lead some to decline therapy that's unlikely to help.

The field is moving forward rapidly in oncology, where there's a crucial need. Ideally doctors and patients would know – before prescribing or taking treatment, or choosing another cycle of a costly remedy – who is likely to respond favorably, to which drugs, and at what doses. These tests have the potential to make precision cancer care both more effective and more affordable.

Some of the science behind these diagnostic tools is cutting edge, and some is long overdue. Unfortunately, many of the tests aren't yet sufficiently accurate. Even seemingly straightforward DNA tests, for specific mutations, may not predict an individual patient's responsiveness to a matched treatment.

A Perspective from Pharma

"If I were a patient, I would want a test that will predict if I will be helped by a treatment," said Cecilia Schott, of AstraZeneca, during a panel discussion of companion diagnostics at the Boston meeting. She heads that company's program in personalized healthcare.

But not all pharmaceutical executives are, or were, gung-ho about promoting tests that would be gatekeepers for prescription and insurance coverage of medications. Accurate tools are costly to develop and evaluate, on top of expensive research and trials of the relevant drug. And if a diagnostic test proves effective, it would direct therapy to a more limited patient group than, potentially, everyone who might want to try the companion treatment.

Industry's approach to companion diagnostics has shifted dramatically over the past five years of so, Schott confirmed. "There was an assumption that personalized medicine is costly," she told me in a subsequent phone interview.

"Now at AstraZeneca, about 85% of our pipeline has personalized care embedded in our development programs," Schott said. "We follow the science to identify if a patient population will respond better. The reason is very simple," she continued. "Having good biomarkers decreases the chances of a drug's failure. It means that the pivotal trials have a greater chance of succeeding, which is good for everyone."

The FDA defines a companion diagnostic as a tool or imaging device that provides information essential for the safe and effective use of a corresponding treatment. As of today, the agency lists two dozen or so approved tests, most having to do with cancer.

Companion Diagnostics and the FDA

Just last week, the FDA granted rapid approval to a new anti-cancer drug, Tagrisso (osimertinib or AZD9291, AstraZeneca), for treatment of patients with progressive lung cancer harboring a very specific mutation. This new drug comes in pill form, much like old-fashioned antibiotics that can be taken by mouth. Tagrisso belongs to an expanding category of drugs called tyrosine kinase inhibitors (TKIs). It's thought to inhibit cancerous cell growth by blocking signals through the epidermal growth factor receptor (EGFR), which is inappropriately active due to mutations of that gene in some lung cancers.

With treatment, malignant cells tend to become resistant to medications. So tests have evolved – and this is an active area in companion diagnostics – to sequentially check genetic material from cancer cells (such as by liquid biopsies) for additional DNA mutations, changes that cause or permit the malignancy to progress after a period of responsiveness to a drug that had been working for a while, such as months or years.

The new drug can overcome resistance and help lung cancer patients who have acquired the T790M mutation. When the FDA approved Tagrisso last week, it also approved an updated test, Cobas EGFR Mutation Test v2, manufactured by Roche. This companion diagnostic checks for the particular EGFR mutation, T790M, which may cause or mark resistance to other lung cancer medications. As considered recently at Medscape, the T790M change appears to confer greater responsiveness to Tagrisso. (As reported, some patients who lacked this mutation also responded to the drug.)

The Wall Street Journal reported that the FDA's approval of Tagrisso came in record speed, after a "development process that took just 2½ years." There was no waiting for phase 3 data.

As outlined by the FDA, the agency reviewed data for Tagrisso from two early-phase studies involving 411 patients. All had lung cancers with the EGFR mutation of interest, T790M, with disease that had progressed after treatment with other first or second generation TKIs that target abnormal EGFR activity. In the two studies, 57% and 61%, of the mutation-enriched patients enrolled, respectively, had either a complete or partial lessening of the tumor size.

While far from perfect, that is a remarkably high objective response rate – on the order of 60% – for a single drug, a pill no less, in patients with metastatic, heavily pre-treated cancer that's progressed through other medications. As I've considered previously, investigators in precision oncology (and this one writer) envision giving multiple medications, in cocktails and other sorts of combinations, in an effort to squelch a tumor rather than just temper its growth.

But taking this drug is not easy. Apart from cost issues, the FDA noted considerable toxicity among the 411 patients in the two studies who received Tagrisso at the recommended dose. Diarrhea and rash were both common, each affecting 42% of the research subjects. Nausea, dry skin, nail changes, eye problems and loss of appetite were also fairly frequent. The most dangerous reactions included heart rhythm changes, inflammation of the lungs, and blood clots that formed or spread in the lung. Fatal adverse events were attributed to Tagrisso in 3.2% of patients who received it.

So it's fairly toxic stuff; most people wouldn't want to take this new medication without knowing there's a reasonable probability of its effectiveness against their tumor.

From this example, it's evident that a rapid FDA approval process has the potential to lower the costs of a drug's development and, therefore, possibly, to reduce its justifiable price. It also serves as an instance of a companion diagnostic enriching the fraction of patients in a trial who are likely to respond to a medication, with favorable results.

Not Just About DNA

Earlier this month, the FDA held back-to-back public workshops on Performance Evaluation of Next Generation Sequencing In Vitro Diagnostic Tests and Databases for Establishing the Clinical Relevance of Human Genetic Variants. These agenda reflect the challenge for regulators in evaluating genetic tests by methods such as next-generation sequencing (NGS), confirming the accuracy and interpretation of results in context of new, gigantic and proprietary cancer IT systems.

You might think, naively, that in the future world of cancer care, most smart treatment decisions will depend on genetics. But that is not the case.

Yet there's loads of genetic information to be gathered: a sequence for every gene-encoding part (exome) in cancer patients' inherited genetic material, and possibly the rest of it, along with the full cancer genome. When cancer persists during treatment, its genome may acquire mutations over time, and variably among cells within the tumor, so it might need  be checked repeatedly, and at multiple spots.

Investigators are interested in epigenetic changes, too: the smaller molecules that bind to genetic segments and influence whether they're active, or shut down. There's also, still, value in traditional karyotypes, FISH (Fluorescence) and newer CISH (Chromogenic) In Situ Hybridization, and other ways of examining cancer cell chromosomes. My point here is not to detail each method of evaluating DNA, but to give a sense of the scope of what's feasible, with current technology. 

Genetic checks are straightforward, as results go. When provided by a competent lab, the readout is clear: a sequence, or a mutation, or many mutations. (How to interpret that information gets harder.) And they're, perhaps, the tip of the iceberg for companion diagnostics.

Assays of RNA for instance, can be quite valuable in oncology decision-making, reducing treatment needs and costs. Already tests like MammaPrint and OncotypeDx enable some patients to wisely choose less treatment. However, these might not be considered true "companion" diagnostics, as they're not designed to check a cancer or patient's sensitivity to a particular drug. They're more like prognostic indicators, with value that's Rx-independent.

But let's say you're testing an RNA inhibitor in an early-phase or "matched" clinical trial, an RNA assay might be just the thing.

Meanwhile protein assays, including immunohistochemistry (IHC) – a very old method – remain very useful (considered below, for Her2 measurement). These may be jazzed up or updated versions of IHC with computer-assisted scanning devices capable of quantifying protein levels in thousands or even millions of cells. It may not matter if a gene is mutated if it's not turned on, or causing resistance; protein, including enzymes and receptors that signal for movement, stickiness, survival or death, may represent the cancer cell's "bottom line."

So one issue for companion diagnostics inventors – besides financing and FDA regulation and all of that – is what's useful to measure, when, and how often.

Tests like liquid biopsies – to evaluate circulating tumor cells (CTCs), tumorous DNA in blood, or cell elements like exosomes, about which I wrote recently – are in-the-works. At the Boston meeting on Companion Dx, quite a few presentations had to do with liquid biopsies. The basic idea is to test cancers, and monitor responsiveness to treatment, with serial blood samples rather than invasive biopsies. This is a hot topic in oncology now, and I expect to cover it separately.

Consider also the value of functional assays. Just to offer one example: a few months back, I wrote about PARP inhibitors with unexpected activity in tumors having DNA repair defects, but without known BRCA mutations. Meanwhile companies are developing novel and gene-agnostic ways to see if patients' cells don't fix damaged genetic material normally, and so might be exquisitely sensitive to a combination PARP inhibitor and DNA-damaging chemotherapy. It appears that a "scarred" genome reflects this sort of cellular incompetence, and portends responsiveness to treatment. At the meeting in Boston, Keith Wilcoxen of Tesaro, Inc., gave an update on quantifying these sorts of chromosomal changes, as a potential companion diagnostic, in the not-so-distant-future.

There's no reason, in principle, why all of this information can't be recorded, and uploaded to the cloud or Watson or whatever oncology informatics system is in use. Unfortunately the tests, available now, are for the most part industry-driven and limited in their reporting of results, based on what's useful to the providing company or, possibly, demanded by the FDA.

Questionable Measures

Some companion diagnostics simply are not ready for prime time, to borrow language from a recent tweet by Dr. Eric Topol, a cardiologist and technophile. He pointed to a critical article in JAMA Oncology concerning tools for assessing a protein, PD-L1, and related biomarkers.

The current paper highlights some limits to companion diagnostics – and anticipates future problems with these, affecting patients – when tests are developed by competing pharmaceutical or lab companies with the possible aim to direct therapy toward one drug over another. The authors, both of Princess Margaret Cancer Centre in Ontario Canada, review several ways of assessing PD-L1 in clinical cancer specimens that have been approved by the FDA, or are in-the-works.

Dr. Lillian Siu is an author on the JAMA Oncology paper and a professor of medicine at the University of Toronto whose research career has focused on trials of experimental cancer drugs. We spoke by phone about the PD-L1 assay and related issues.

"The siloed development of these assays is detrimental to patients, and is also bad for clinical research," she told me. "When each company has its own antibody, and its own companion diagnostic, it's harder to predict how patients might respond to each drug."

The report highlights differences among the assays for these and related molecules – which might predict the effectiveness of Keytruda and Opdivo and newer, pipeline drugs – based on each company's use of particular lab reagents, like distinct antibodies, and what might be arbitrary, or result-enhancing, thresholds for calling a test "positive."

Inhibitors of PD-1 and PD-L1 offer a new axis for treating cancer, Siu said. These immune-harnessing drugs will likely serve as the backbone of many treatment combinations. "But the assays we have now are problematic." The ideal companion diagnostic would provide a clear and binary result, Siu considered. "But the reality is not like that."

The method most used to evaluate the presence of the protein, PD-L1, in specimens is called immunohistochemistry, IHC. Last month, the FDA approved one such companion test, PD-L1 IHC 22C3 pharmDx, manufactured by Dako, for patients with refractory lung cancer who might benefit from Keytruda. The test will be provided through Quest Diagnostics, in partnership with Merck's CDx Division.

There's a lot of enthusiasm in the oncology community, including patients, for trying immune checkpoint inhibitors, agents like Keytruda and Opdivo. These drugs both are monoclonal antibodies, produced by separate companies (Merck and Bristol-Myers Squibb, respectively). Each is designed to release the brakes that otherwise prevent a patient’s immune system from destroying cancer cells. Both Keytruda and Opdivo are thought to work by blocking an immune cell death-inducing molecule, called PD-1 (programmed death-1), from binding to its normal signaling partners, called ligands, with names like PD-L1 (programmed death ligand-1), PD-L2, PD-L3, etc.

Each of these drugs, Keytruda and Opdivo, has shown promising results in some – but not all – patients with lung cancer and skin cancer (melanoma). I wrote on Keytruda, as is being tried experimentally in patients with metastatic, triple negative breast cancer. Meanwhile Nivolumab has shown significant activity in refractory Hodgkin lymphoma. Most recently, a study found Opdivo shrinks tumors in 25 percent of patients with advanced kidney cancer, improves patients' quality of life, and extends overall survival. Clinical researchers are testing these drugs in patients with many solid tumor types, including head and neck cancer, and bladder cancer.

These details matter more than you might think. For PD-L1 testing, these kinds of distinctions may affect patients' choices of drug, insurance coverage, and treatment decisions including eligibility for clinical trials.

So there's a lot of interest, money, and clinical outcomes at stake, just for this one group of molecules, PD-1, and its binding partners. Competition is real. Recently, Medscape reported about a TV ad for Opdivo, evidence that pharmaceutical companies are appealing to patients, asking them to ask their doctors about these kinds of drugs.

"To top it all out, PD-L1 is probably not a perfect predictive marker," Siu said. "We know that some PD-L1 negative tumors do respond, and that some PD-L1 positive tumors don't respond." One explanation for this apparent discordance is that a PD-L1 drug could work by binding PDL-2, she suggested (among other possibilities). As things stand, PD-L1 serves "is at best an enrichment biomarker" Siu added. In other words, this kind of marker enables and facilitates prescription of the drug to specific patient populations who may be more likely to respond.

"The problem with PD-L1 is it's not black and white, even with one antibody. Let alone that there are five or six antibodies, and the cut-off thresholds for positive or negative results used for each antibody differ," she said. "It's going to be very challenging, to sort through."

Evaluation of PD-L1 is just one of many examples of these kinds of tests that will guide current and future cancer drug prescriptions. More generally, a problem in protein tests is that there's subjectivity in the measurement, including the threshold by which pathologists – or researchers employed by a company – decide to call a tumor sample "positive" or "negative" in a cancer specimen.  

Siu contrasted the PD-L1 assays with tests for k-ras mutations in colon cancer. Those particular checks for DNA mutations are typically more clear-cut biomarkers, she said.

"We use the presence or absence of k-ras mutations to determine if colon cancer patients should not get certain drugs," she said. "That's a clear 'yes/no,' because if you have the mutation, you shouldn't respond," she said. "That's a dichotomous result, which is very helpful."

Dr. Siu mentioned non-profit lung cancer groups that are pushing for greater "harmony," cooperation and consensus on PD-L1 tests and other diagnostic tools. The JAMA Oncology paper refers to efforts by the International Association for the Study of Lung Cancer to describe and compare the various tests, including feasibility and reproducibility. She and her coauthor, Dr. Aaron Hansen, call on the College of American Pathologists or European Society of Pathology, among others, to address these issues, facilitate post-marketing standardization of tests, and develop guidelines in this and other cancer types, such as head and neck cancer, for which the PD-1 and PD-L1 inhibitors are being tried in clinics.

Dr. Siu reports she has been an uncompensated consultant to Boehringer-Ingelheim, Regeneron, Merck and Novartis. For her work in clinical trials, the University Health Network has received grants and research support from Novartis, Bristol-Myers Squibb, Pfizer, Boehringer-Ingelheim, Regeneron, GlaxoSmithKline, Roche, Karyopharm, AstraZeneca, Merck, and Celgene.

HER2 Testing in Breast Cancer:  An Example

When I considered the recent paper on the PD-L1 tests, including one approved by the FDA as a companion diagnostic, I was reminded of a controversy that lasted years (and may still simmer), over how to accurately and best measure a complex protein in some cancer cells, Her2, or the gene encoding that molecule, Her-2 in breast cancer.

Over the past 15 years, routine testing for Her2 in breast cancer and availability of companion drugs, have changed the standard of care, and outlook, for those affected. Approximately 20% of breast cancers, the Her2 gene is turned on, too much, so the protein appears at high levels. This also happens in some stomach cancers and other malignancies.

Knowing if a tumor is Her2 positive, or negative, can make all the difference in a breast cancer patient's course, including survival. Which is why these companion diagnostics, and the accuracy of those, matter so much.

In 2006 the FDA approved Herceptin as extra treatment after surgery for early-stage, Her2 positive tumors. It approved Herceptin so based on multiple studies demonstrating that the antibody reduces recurrence in women with Her2 positive tumors. In patients with metastatic disease, this agent is effective in Her2 positive cases, usually in conjunction with other drugs. Herceptin and newer anti-HER2 agents like Perjeta, have expanded choices and dramatically improved the outlook for people with Her2-positive breast cancer, with early or metastatic disease.

As recently as 2010, the New York Times ran a front-page story about the lack of clarity in Her2 test results in breast cancer. The paper detailed how some women with early-stage tumors weren't sure if they should take Herceptin, and elected not to do so because their Her 2 results varied between labs. Since then, the American Society of Clinical Oncology and the College of American Pathologists have both reviewed and updated guidelines for Her2 testing.

It's for this reason – a need for clear and valid results – primarily, that the FDA has taken pains to review and approve some, but not all, tests for Her2 as companion diagnostic tools for prescription of Herceptin and PerjetaAs of last week, a good fraction (10 of 24) of the FDA's approved companion diagnostics represent distinct ways of gauging if that molecule is normal or excessively present in cancer cells.

And this is all for just one molecule! Her2 is just one of so many examples, of molecules and cancer cell parts that can be measured with current technology, and which does, or will, inform treatment choices and, potentially, be used and combined to achieve cures.

A Personal Thought

When I prescribed chemo years ago, we had few tools (and almost none standardized) to assess genetic changes and proteins in each case. Responses were hard, if not impossible, to predict. That's part of why I did research, and didn't prescribe so much treatment. I didn't want to cause harm. And so I'm encouraged, thrilled by this emerging technology – the capacity to evaluate patients' cancer cells at the molecular level – and, based on science, having insights about corresponding drugs' likely effectiveness in each case, knowing when to give or change or stop giving treatment. I find it promising, for real.

Follow me on LinkedInCheck out my website