Decoding Health Care

Part 17: Precision Medicine

I am extremely proud to be an alumnus of the Washington University in St. Louis School of Medicine. In fact, whenever I attend a medical school reunion and the Dean of Admissions reviews the qualifications of the incoming class, I am amazed they ever let me in. My fellow students, the medical residents, and the faculty were the smartest people I ever met. As an example, one of my internal medicine residents, Dr. Brian Kobilka (who was the nicest, most soft spoken guy you could ever meet), won the Nobel Prize in Chemistry in 2012 for his work with G proteins, a critical cellular signaling pathway. Another one of my residents was Dr. Brian Druker. In 1993 he identified imatinib as the “magic bullet” for Chronic Myelogenous Leukemia (CML), and thus arguably became the father of precision medicine.

CML is an excellent starting point for our discussion of precision medicine. It is a  fascinating disease (Milestones in Chronic Myeloid Leukemia - Hematology.org). Initially described in the 1840’s, the first clue to its etiology was discovered by Dr. Peter Nowell at the University of Pennsylvania and Dr. David Hungerford at Fox Chase Cancer Center in 1960. While studying the white blood cells of patients with CML, they noted an abnormality of chromosome 22. They named this the Philadelphia chromosome. In 1973, Dr. Janet Rowley at the University of Chicago discovered this chromosomal abnormality was actually a DNA translocation involving chromosomes 9 and 22. This was the first time DNA translocation was associated with human malignancy. Ten years later, a number of scientists found that this translocation resulted in the creation of a fusion gene/protein called BCR-ABL which turned normal blood cells into cells that grew without any response to normal control mechanisms, ie cancer cells. BCR-ABL was the single explanation for the cancer called CML. BCR-ABL was an on switch stuck in the on position. The stage was set for Druker, who subsequently found the drug that turned that signaling mechanism off. BCR-ABL was the lock and imatinib was the key.

Imatinib was an amazing breakthrough; just ask any hematologist-oncologist who cared for CML before imatinib. Prior to imatinib, average survival of a CML patient was about 3 years. Medications were available to reduce the symptoms of the disease, but the only treatment that occasionally changed the trajectory was interferon which was very toxic. Some patients underwent a bone marrow transplant which could cure the disease but you needed a bone marrow match (usually a sibling) and the risks of the BMT were very high. Imatinib worked in 95% of patients and gave patients with CML a normal life (in both quantity and quality).

CML is a rare disease. But it became the paradigm for cancer causation and treatment. The hypothesis was that each cancer must have a mutation explaining it and therefore must have a precise, directed therapy. And each patient’s tumor, having undergone mutation testing, would yield a personalized cancer treatment plan.

Over the years, many mutations in a wide variety of cancers have been discovered. And these mutations cause cancer by many diverse mechanisms that include dysregulation in growth and metabolism, life cycle/cell death signaling, and other genomically controlled functions. But the absolutism of BCR-ABL in CML never seemed to be present, and there were rarely specific therapies that worked as well as imatinib. In some ways, CML and imatinib ruined everything because they created wildly unrealistic expectations for precision medicine.

Mutations are common in cancer. A few are inherited, like BRCA in breast and ovarian cancer. Many more are acquired, so called somatic mutations. In most common cancers, each tumor has several of these mutations. Exactly which mutations are responsible for what is largely unknown. Is there one major mutation that explains everything, or do they work together to determine what happens to the patient? In some cancers, like colon cancer, the sequence of mutation acquisition explains progression of the cancer from a polyp to a life-threatening tumor.

As knowledge of the number of mutations associated with various cancers grew, so did the technology to identify them. If you are looking for a single mutation in a gene you know you are interested in, the test can be very straightforward. You can sometimes actually look at the chromosomes. Or you can look for the RNA transcript. Or you can sequence the DNA of that particular gene. Or sometimes you can even look for the protein product. But when you want to look at a lot of genes at the same time (especially if more than one can be implicated), you need a different approach.

I am by no means an expert at DNA sequencing. When I was a fellow working in a hematology research lab, I learned how to do Sanger sequencing. It was a slow, laborious, manual process. It involved using radioactive nucleotides, huge glass plates, and a toxic polyacrylamide brew that you needed to prepare each time you wanted to do a sequence. But this entire methodology was replaced by sequencing machines and the development of massive parallel sequencing, also known as NGS or next generation sequencing. In this process, a machine carries out the sequencing process on many different genes at the same time. It does not involve radioactivity or polyacrylamide. You just put in the pieces of DNA that flank your genes of interest (the primers), an enzyme to accomplish the sequencing (a polymerase), and (usually) color coded nucleotides. Voila: many genes sequenced all at once. This revolutionized the ability to analyze cancers for all the genes you were interested in.

For oncologists, this became a reality with Foundation Medicine. Foundation developed a panel of genes that could be sequenced on an individual’s tumor specimen. The DNA could be extracted from the pathology slides (formalin fixed, paraffin embedded) and did not require fresh tumor tissue. All you needed to do was send the slides to Cambridge Massachusetts and they would do the rest. Amazing. Foundation’s panel included virtually every gene associated with human cancer, about 300 or so initially. They issued a report with all the mutations that had been identified. It was up to the oncologist to decide which ones mattered.

When this test became available, there weren’t that many gene mutations that you could do much about. And the mutations were largely context specific. A mutation in the epidermal growth factor receptor (EGFR) meant a lot in lung cancer but not in lots of other cancers. As time has gone on the number of gene mutations you could target has increased (“actionable” mutations) and at least some of them have been recognized as being not quite so context dependent (“pan tumor” mutations). In some cancers, like lung cancer, there are a lot of potential gene mutations and so a lot of information that informs treatment. In others, like breast cancer, that is not so true. But you still get a report on 300+ genes, and that test costs several thousand dollars.

This is where conflict arose. Commercial health plans couldn’t see spending thousands of dollars on every cancer patient to get a lot of information they didn’t need. They were perfectly fine with paying for an EGFR mutation test in lung cancer, but they saw no reason to pay for it in breast cancer. But the panel was the panel. Medicare and the FDA collaborated to solve this quandary for Medicare beneficiaries.

I have previously written about how Medicare establishes coverage policy. In general Medicare leaves policy to their MACs (Medicare Administrative Contractors). But for topics of substantial importance they will issue an NCD (National Coverage Determination) which establishes coverage rules for all Medicare beneficiaries including Medicare Advantage. In the case of NGS in solid tumors, they decided to issue such a coverage rule. But they went one step further. There is an infrequently used process called parallel review whereby the FDA and CMS work together to document that a particular product (in this case, the Foundation NGS test) meets both FDA and CMS standards for approval and coverage. And that is precisely what happened here.

Foundation won an incredible victory by succeeding in this parallel review process. The initial proposal by Medicare was a little bit restrictive in whom the test would be covered. The final NCD was not. Medicare pays for NGS testing (by a company that has an FDA approved test including Foundation Medicine and a few others) for essentially all Medicare beneficiaries with a diagnosis of an “advanced” cancer in circumstances where the test might be used to make a treatment decision. Medicare acknowledged that we didn’t know how much the test added in some clinical scenarios but that we knew enough to justify coverage broadly.

This was a huge win for NGS. But commercial health plans are not required to mirror Medicare coverage policy and outside of their Medicare Advantage members they did not. They continued to be hung up on all of the unusable information that the huge panel produced. Included in this “unusable” data were mutations that were “actionable” in certain contexts but not necessarily in others. A good example are mutations in B-RAF. In melanoma, B-RAF mutations point the doctor in a specific direction, as there are a number of drugs that are FDA approved and are quite effective. But B-RAF mutations are seen in colon cancer, and the melanoma drugs are mostly ineffective. Not only do they not work, they are expensive and have a lot of side effects. B-RAF status DOES tell you something about colon cancer, specifically that the colon cancer is particularly nasty. But it does not tell you to use the melanoma drugs. And I assure you that if you are in desperate circumstances, it is very tempting to give them a try.

This difference in coverage drives NGS companies crazy. And there are now a lot of NGS companies. Many health plans have agreed to cover NGS in certain tumors (non-small cell lung cancer being the predominant one) but not broadly. Obviously as evidence accumulates for the clinical usefulness of testing in other tumors this will improve access. But at this point there is a stalemate. In fact, things are getting worse rather than better. Testing companies are adding more genes without evidence of clinical benefit to their panels and payers are digging in their heels. So what happens next?

The answer is that we can start by learning from the thousands of patients who have already had their tumors sequenced. Medicare actually wanted to do that when they issued their NCD. The initial proposal was to establish a registry collecting the sequencing results, treatments, and outcomes for patients who had sequencing done and there wasn’t a clear FDA approved treatment. But this proposal was shot down: too expensive, too complicated, etc. Nobody wanted to own this registry. Opportunity missed.

 As we reviewed in the real world evidence post, everybody is holding on to their own data. That includes the sequencing companies, academic medical centers, and whoever else thinks they can make some money on the data they possess. If we can liberate that data, and in a perfect world marry it to the clinical and payer data currently also being held hostage by those who control that data, we might actually learn something useful. Again, this would require a role for the government in establishing and maintaining this data commons and a mandate for all parties to participate. Not easy, but possible.

And there are a lot of questions that still need to be answered about this precision medicine approach. The question about what markers are truly pan-tumor, which are poly-tumor and which are context dependent is an obvious one. That alone would help us understand how best to treat our patients immediately. NCI has started down this path with NCI MATCH. NCI MATCH was a clinical trial, launched in 2015, that enrolled patients with a host of mutations for which there was no FDA approved precision therapy but where there was a potential targeted therapy (https://www.cancer.gov/research/infrastructure/clinical-trials/nci-supported/nci-match). This trial was closed in 2022 having screened over 6000 patients and enrolled over 1200. There were ultimately 38 drugs tested and about 25% of these drugs had “some activity” (defined as 15-20% response rate). Only one new FDA approval came out of this trial. These results are modest at best and the progress has been painfully slow. It was an ambitious undertaking, and the NCI does deserve credit for trying. But this snail’s pace is just inadequate.

We now live in an age where the mutation analysis can be done on blood specimens, so called liquid biopsy. Many tumors shed DNA into the blood and this cell free DNA can be sequenced easily. It is still not completely clear if this is just as good as tissue based testing. If I were to venture a guess, sometimes it is and sometimes it isn’t. We can answer important questions about how to manage conflicting results and in what scenarios blood based testing is as good as (or even better than) tissue based testing.

Perhaps the most important question is why precision medicines are almost never curative. Why do we rarely if ever get the same kind of results we get with imatinib? Are we underplaying the role of context, ie the tumor type and clinical patient characteristics? Are we ignoring the importance of the mutations in genes that we cannot do anything about (we call them bystander genes)? What drives resistance? As our basic knowledge advances and our therapeutic armamentarium grows we should be poised to use this knowledge to the benefit of our patients. Artificial intelligence should facilitate answering these questions. But none of this will happen as long as parties hoard their data because of its potential monetary value.

Some of these questions are certainly going to require clinical trials. Ever since the NCD and subsequent commercial payer intransigence, testing companies have argued that one of the major benefits of testing these non-actionable genes is to promote clinical trial enrollment. As it stands, many trials require the presence of a particular mutation as an eligibility criterion, and if you cannot get NGS done you cannot be eligible. That’s a good story but I don’t know of any convincing data that trial enrollment has actually improved as a consequence of the NCD.

As data becomes available and we learn about potential gene interactions, mechanisms of resistance, and the importance of clinical variables, there should be an explosion of clinical trials. This should be further fueled by AI driven drug discovery. The NCI has already recognized this by launching NCI ComboMATCH which studies combination precision therapies. If we can open these trials to more people, the argument that NGS reaps benefits by facilitating clinical trial enrollment will be irrefutable. And payers will have no choice but to lift the restrictions on NGS coverage that are currently in place.

It is frustrating that we have not yet achieved the potential promised by imatinib, and we have a long way to go before we have fully investigated the power of precision medicine. I am bullish on this potential.

NGS has another important application, in cancer screening using multi-cancer early detection tests (MCEDs). To understand this, we should start with the topic of cancer screening, which we will do in our next post.