What is the value of treatment options in precision oncology?
-
Are we ready to assess the value of treatment options in precision oncology?
Efforts to address the rising cost of cancer treatment drugs dominate the national conversation about the use of precision medicine in oncology, but there are other, equally-important issues to consider.
1. Certain pharmacogenomic tests are not routinely used in clinical practice.
While precision or “genomic” medicine has already been integrated into the practice of medicine in some medical subspecialties, it’s made the most progress in the field of oncology. Oncologists have been successful in using targeted drugs to treat specific genomic DNA alterations associated in certain cancers.
Patients with EGFR-mutated lung cancer, for example, generally benefit from erlotinib or gefitinib, while those with ALK mutations can benefit from crizotinib or ceritinib. Patients with colorectal cancer and the driver KRAS or NRAS mutation do not derive benefit from cetuximab or panitumumab, and may even be harmed by toxicities associated with the combination of these drugs with chemotherapy.
If those RAS genes are not mutated, the anti-EGFR targeted agents can work well – especially when combined with chemotherapy – and can prolong survival. Patients with breast cancer and ER/PR+ tumors benefit from hormonal therapy, while those whose tumors express Her2 benefit from Herceptin, unless their tumors have other mutations that cause resistance. There are genomic tests that predict the sensitivity of tumors, as well as the toxicity from classical drugs such as 5-FU, irinotecan, taxanes, and others. But, these tests aren’t consistently used in clinical practice
2. Outcomes are difficult to predict.
Genomics may offer a path to precision therapy by demonstrating the value of targeted agents for specific tumor signatures. The NCI MATCH trial, which opened in August, seeks to determine the effectiveness of precision medicine therapeutics in oncology, which represents a paradigm-shift in oncology. For the first time, patients will be treated based on what kind of driver mutation is found in their tumor, as opposed to the tumor’s histology, or tissue of origin.
But, it can get complicated.
For example, one predictable result may be that patients with melanoma and V600E BRAF mutations respond well to drugs such as vemurafenib that target BRAF, but patients with colorectal cancer and V600E BRAF mutations don’t respond to those same drugs.
Yet, oncologists did not predict that a combination of a mutant BRAF inhibitor, a MEK inhibitor, and an anti-EGFR inhibitor would come close in BRAF-mutated colorectal cancer to what a BRAF inhibitor as a single agent can do in melanoma.
3. Oncologists are making progress, but the issue of financial toxicity with triple combinations remains a problem.
Most clinicians are likely to recommend therapies that have high response rates for approved indications, and health insurers are likely to cover them. Therapies with documented short-term clinical benefits are clearly valuable, but the issue of rising costs is apparent.
In this scenario, a given mutation in different types of tumors, and clinical observations of a population with specific biological characteristics, are found to behave in a similar manner. The tumor heterogeneity within the BRAF-mutant melanomas did not have much impact on the high initial response rates to a mutant BRAF-specific inhibitor.
4. Genomic medicine in oncology is at a primitive stage.
Historically, scientists were unable to predict drug efficacy even with single gene defects, until breakthroughs such as the discovery of EGFR mutation’s usefulness in response to agents like erlotinib. With the dawn of precision medicine, oncologists will be able to predict how drugs affect hundreds, or even thousands, of genes. Our ability to use this information for the benefit of patients is lagging behind.
Currently, genomic medicine in oncology involves the use of information from one or two mutated genes in a tumor, as mentioned above. One gene may predict sensitivity, while a second gene predicts resistance. In oncology, we are challenged by the need to do more with tumor genomics in order to help our patients who otherwise will have limited survival.
Though technology has advanced, our ability to harness this information for the benefit of patients is still low. This is a pressing challenge that precision oncology promises to address in the coming decades.
In the future, extensive genomic testing may be affordable and routine in a general practitioner’s clinic. The ultimate impact on the population may be in the area of disease prevention.
President Obama recently launched the White House Precision Medicine Initiative to help figure out how the general population can benefit. For the population at large, the answer for most genomic mutations may not be a drug, but rather prevention and screening.
5. The human genome is vast and largely uncharted in terms of precision medicine applications.
To date, we do not have drugs for much of what can be uncovered by genomic analysis, i.e. for the few hundred genes we understand from decades of basic research. For most of the human genome, we still don’t know what it means beyond genes that encode proteins – most of which have not been studied – and a few non-coding sequences that are beginning to be better understood.
Precision medicine will be slow in adoption over the years with incremental progress as we learn more about the genome and as more drugs are developed. National precision medicine networks work to establish guidelines and demonstrate the value of genomic testing in oncology. Some efforts may solve problems and revolutionize medicine in general through precision medicine.
We will need to predict the benefit for patients based not only on their tumor genomes, but also on epigenome and immune responses. Again, predictions are easy when outcomes are impressive, such as with immune checkpoint therapy and high tumor mutation burden, i.e. mismatch repair deficient colorectal cancer.
6. We need to predict the benefits and the value of a particular treatment for a particular patient, and we need to accurately communicate it.
The need to predict benefit and value for individual patients is a critical component in the drug cost-value discussion. Predicting drug value inaccurately for a given patient’s tumor and its genome may deprive patients of exceptional responses if the value is deemed low and the drug isn’t covered by insurance based on a population and statistics argument.
While providing important information that helps in discussion, Median OS or PFS aren’t relevant to the exceptional responders, or the outliers, as those are not predicted.
Without a doubt, the cost of drugs is a major societal issue facing our time. It’s, in part, a product of the many impressive advances of biomedical research and the many investments made by the pharmaceutical sector. And, some important efforts are addressing the ‘value’ of drugs.
I have previously weighed in on the appropriateness of bringing in the cost-value discussion at a stage in drug development prior to drug approval, or in a situation of off-label drug use when there may be no better therapy options or clinical trials.
In thinking about the most recent efforts and what ultimately will be communicated with patients, the issue of whether we can actually predict benefit for a given patient and their tumor, along with its genomic signatures comes to mind.
There’s a bit of a catch-22 in predicting outcomes and demonstrating value from currently available expensive drugs, in the era of precision medicine and N-of-1 experiences. Unless it is shown that patient outcomes are impacted – which can take years – clinicians are unlikely to order expensive precision medicine tests, especially if they are not reimbursed by health insurance companies.
There is little chance of making accurate predictions of expected benefits, which are fundamental to determining value, unless test results are available and can be associated with clinical outcomes.
The exception, again, is in situations where response rates are known to be very high. In the case of very active regimens, there is clearly value for patients while the cost issue is a huge and worsening problem.
Rare tumors and orphan diseases remain a problem in the era of precision medicine with the vast molecular heterogeneity of cancer (no two tumors are alike). Added to this catch-22 is the lack of universal genomic testing platforms and widespread sharing of clinical outcomes associated with genomic signatures for specific individual tumors in real time.
7. We need more data-sharing.
We need to fix the EMR problem to facilitate not only the association of clinical outcomes with molecular profiles, but also their sharing across platforms. This may help us get to a point where we can move towards making predictions regarding the actual value of a particular treatment for a particular patient.
While in general the lack of evidence for benefit may associate with actual lack of benefit, this may not be the case in the era of precision medicine where obstacles and challenges to demonstrating impact of various therapies are present.
That, along with other complexities discussed above, slows down the availability of the best quality of evidence. In fact, the promise of precision medicine is to deliver the right treatments to the right patients which should ultimately lower costs.
8. There’s skepticism and push-back on the intensity of focus on precision medicine.
Precision medicine is not a distraction. In fact, it should impact population health in the future.
Evolving technology and may help over time as less expensive tests help to provide information about the genomics and resistance mechanisms. Liquid biopsy has a future in the approach to deliver precision medicine over time to a patient. And we’re quickly learning from the N-of-1 experiences with targeted therapies and genomic information that are found in specific patients.
Other progress is being made to improve or standardize approaches – an area of great need. The FDA, for example, is working to develop a crowd-sourced platform for standards in precision medicine testing technology.
9.Within the same PFS or OS curve, value can be huge for those who have durable responses and non-existent for those who don’t.
In the end, associating drug cost with value is rational if, and only if, value is predictable for a given patient. There is no generic answer, although I will concede that there is little argument about value in situations where drugs work very well and most patients are known to benefit.
Obviously, even where drugs work well, the drug costs remain a serious problem. The value issue is even more problematic if we can’t predict the 10 or 20 percent of patients who derive benefit from a given therapy.
For those who have exceptional responses, the ‘value’ is huge, while for those who have no response, there is no value and possibly harm. It is the problem of statistics and why most clinicians often have to explain to patients that statistics are just that; they do not predict what will actually happen in the course of treatment of a given patient.
With better tools, data, modeling and experience, we can move towards making better predictions for given patients. But we’re not there yet.
10. If we can’t accurately predict the value of a specific drug for a given patient’s tumor, are we ready to assess the value of treatment options in oncology?
We are generally good at predicting benefit and assessing value of particular treatments when response rates are very high or when response rates approach zero. However, we need to improve our ability to predict who will get great benefit versus little or no benefit for much of what is between the extremes.
For efficacious drugs, we don’t know how to predict who may have long-term benefit when there is a tail of long-term survival; thus, for the same PFS (Progression Free Survival) or OS (Overall Survival) curves, some patients hardly benefit while others can have long term benefit.
So, if insurance companies stop paying for drugs because of flawed analysis of value, who does this help? Will value assessments for cancer therapeutics impact on health disparities? Can we do better with accurately predicting survival advantage from a given therapy through molecular and genomic analysis, so we only treat those patients who very likely will benefit?
Isn’t this the promise of precision medicine? How will we get there before the system goes bankrupt? If we only treat patients who are accurately predicted to benefit, then we maximize value, such as lowering overall costs.
Perhaps in the future, genomics may help predict accurately which specific patients will benefit and for how long, for their specific tumor signatures, at the time of drug approval. But how do we get there with the hundreds of already approved drugs?
The issue of rising drug costs over time for the same as well as new drugs remains a major problem that threatens to deprive patients of potentially effective therapies, despite all of the exciting progress that is being made.
Precision medicine and genomics may impact the cost of health care and drugs by improving the science of drug selection, and by allowing more accurate prediction of expected benefit from alternative medical interventions.