Kirsten Patrick: This old dog learns new tricks: genomics

kirsten2Although the term “genomics” was coined in 1986 by geneticist Tom Roderick, I didn’t learn about genomics at medical school in the early 1990s. We studied “genetics” and a bit of molecular biology, but the genetic medicine we were taught was, as far as I remember, mostly about patterns of heredity and the few diseases known to be underpinned by specific genetic abnormalities. Translational genomics, stratified medicine, personalised medicine…these are all concepts that I’ve had to try to get my head around since leaving medical school.

Yesterday I had a crash course in translational genomics at a one day meeting held at the Royal College of Physicians in London. “Translational genomics: the path from genomic insight to clinical applications, licensed drugs and treatment decisions” brought together leaders in genomic medicine from both academia and pharma, pharmacists, educators, regulators, and clinicians to discuss how genomic medicine is evolving, the state of research, and how the health service in the UK will be able to move towards offering personalised treatments for a wider range of diseases to patients in the future.

Munir Pirmohammed, the NHS Chair of Pharmacogenetics, began the day by presenting the four main ways that genomic medicine currently improves drug treatment for patients, often by making sure that they don’t get a drug that they are genetically predisposed to getting serious side effects with.

Even I was familiar with tests for particular phenotypes such as the one that measures thiopurine methyltransferase (TPMT), which is recommended prior to commencing patients on treatment with thiopurine drugs such as azathioprine and mercaptopurine. Apparently there’s variable use of this test; dermatologists diligently order them while gastroenterologists don’t tend to.

Then there are tests of genotype, for example the test for HLA-B*57:01. Patients who are positive for the gene are at high risk of hypersensitivity reactions (which can be severe in 4-8% of patients) if they take Abacavir, an antiretroviral commonly used in the treatment of HIV. A 2004 study that showed that doing this test was cost-effective transformed the way that patients with HIV were treated. One of the problems that comes with genotype testing is that it’s SLOW. Developing bedside genotyping is the next big challenge here.

Genome wide association studies (GWAS) have also contributed to better drug treatment, for example for patients with HCV. A series of genome-wide association studies independently identified a strong association between common IL28B polymorphisms and the outcome of PEG-IFN-α plus ribavirin combination therapy in patients chronically infected with HCV genotype 1. Since 2001, 22 new alleles have been associated with serious immune mediated adverse drug reactions. Many of these have been discovered through GWAS.

Lastly, gene sequencing technology has helped to identify mutations in the cells of certain tumors that can be targeted in treatment. An example here is the discovery of a mutation in the BRAF gene that is oncogenic. Vemurafenib is a B-Raf enzyme inhibitor that has been shown to cause programmed cell death in melanoma cell lines; following it’s release in 2011 it has begun to transform the outcome for many with late stage melanoma.

The existence and continued development of more personalised treatments through genomic medicine means nothing, however, if such tests and treatments are not approved, adopted, funded, and developed further.

Studies have shown that 2 gene mutations (CYP2C9 and VKORC1 genotypes) account for nearly 50% variation in dose response in patients treated with warfarin. However, guideline developers have said that the evidence is not strong enough for them to recommend routine genomic testing in patients being started on warfarin and that more RCT evidence is needed before genomic testing can be seen as better than usual care for patients treated with warfarin. Pirmohamed drew on Michael Rawlins’ 2008 Harveian Oration to question whether we should apply such strict rules regarding hierarchies of evidence to genomic studies, where quick, relatively cheap answers are available from large studies but randomised trials of adequate power would take many years and vast expense to undertake. (And for those of you who are about to cry out that genomic tests are expensive too consider that you can now get whole genome sequencing done for US$99.) A hospital in Nashville in the US operates an opt-out scheme whereby a patient’s clinical chemistry blood sample is used for DNA sequencing this information is then linked to central medication database. In 18 months they have amassed a huge amount of observational data cheaply that just wouldn’t be possible from an RCT of 2000 patients that would take ages to develop.

Interestingly, the EMA has called for incorporation of genomic susceptibility data in future submissions for drug licensing, particularly for drugs used to prevent cardiovascular disease.

And for the clinician…? There is a huge need to develop more decision support software and algorithms to help clinicians to interpret whole genome sequencing data.
Shortening turn around times for genomic testing depends a lot on the complexity of the test. For single gene tests we need to find ways to do these tests more locally and more rapidly.

MARTIN GLENNIE, director of Cancer Sciences Division at Southampton University in the UK, melted my brain a little with a talk on the development of monoclonal antibodies to treat cancer. Immune response is related both to the generation of cancer and its treatment. Tumours control the type of T cells that infiltrate them (via T cell receptors); mostly they just kill them or turn them off. But some synthetic monoclonal antibodies designed to work with the right co-receptor that determines a T cell response can turn this around and get T cells killing tumor cells. An example is ipilimumab—which blocks cytotoxic T-lymphocyte-associated antigen 4 to potentiate an antitumor T-cell response—which has reduced tumor burden and prolonged survival in a proportion of patients with metastatic melanoma (albeit with some very troubling side effects for some). Another example is rituximab for non-Hodgekin’s lymphoma.

Such drugs are exceedingly costly and don’t work for everybody. It’s still not known how to tell in advance who they will work for and who they won’t, although when they work the effects are dramatic. How to fund their refinement and development for mainstream use, and how to work out how payers will pay for them is a question that looms large.

LON CARDON of GSK pointed out that personalised oncology medication isn’t “the future,” it is happening today. Pharma is much less interested now in the development of big blockbuster drugs and turning towards development of stratified medicine. Although the number of patients that might benefit from a new therapy is relatively small the hope is that research will bear fruit when it comes to discovering more new treatments for other (less rare) diseases. The trick is to accurately choose gene/mutation targets at an early stage before going into clinical trials to avoid costly late failures. Orphan drugs made up a third of new medicines approved in the EU and USA in 2009.

He emphasised again the incredible usefulness of GWAS signals for the detection of serious adverse events. Drug trials will often terminate early if high levels of AEs are shown—at great expense—or they won’t be picked up at all in early trials if they are rare. However, genomic studies of very small samples of patients who display a particular AE are good value—when gene-related AEs are detectable they are “loudly” detectable (such as in the case of Abacavir—see above).
Some AE candidate genes are not quite so perfectly predictive and it’s more of a challenge when the CIs around estimates of risk are wide.

Cardon taught me something I had never considered before: the concept of “human knockouts”
Some naturally occurring mutations mimic the actions of drugs, and genomic studies of populations in which the rate of the mutation is high can give proxy information regarding the likely effect of the drug that has the same effect as that mutation.

Fred  van Goor of Vertex Pharmaceuticals gave a stark example of how much things have changed since I was a medical student. I was taught about cystic fibrosis as a symptom-complex disease where patients’ pathology fell on a spectrum of severity. Who knew that today we would know that any of a number of different mutations might underpin the clinical disease entity we call CF?

Many speakers argued for the “genetically enabled electronic medical record” to be made available to researchers for the purposes of discovering new drug targets. I wondered who is thinking about how to get patients’ buy-in for this. We can’t just assume that if we tell them it’s a good thing they’ll all agree. I can see the tabloid headlines: “NHS gives our DNA information to private companies!”

But moving on…what about regulation of the genomic diagnostics that inform the prescribing of drugs?

ELISABETH GEORGE, associate director of appraisals at NICE, explained that the FDA currently has a fixed process for the approval/clearance of companion diagnostics. Drugs must be considered together with any genomic test that is designed to inform dosing or patient suitability. This requires submission of a double RCT. The European Medicines Agency does not have an evaluation plan like this for companion products. The EMA may stipulate a patient sub-group for a particular drug without stipulating a specific test to work out who gets it, or it may recommend a companion diagnostic test but not stipulate that it must be done for a particular drug. Peter Johnson of Cancer Research UK argued that the FDA’s approach is too rigid (and unsustainable) as there is often more than one diagnostic test or biomarker available to test for drug susceptibility, and also the complexity f vertain diseases/cancers means that multiple different susceptibility tests are needed for one patient. The EMA approach is therefore more flexible.

NICE just looks separately at the clinical and cost effectiveness of the drug and the biomarker test.
NICE cannot recommend any particular diagnostic test because the DOH does not refer biomarker tests to NICE as technology assessments. Therefore, NICE can only implicitly recommend a diagnostic test to accompany prescription of a drug and can’t overtly recommend its use. Better uptake of testing has been seen in the UK when the manufacturer of the diagnostic test funds its use, for example with Vemurafinib (Roche produces both drug and the BRAF test, and makes BRAF testing free of charge to the NHS in three centres in the UK.

Only a handful of drugs that have been assessed by NICE in recent years have been assessed with a companion diagnostic test:

*Cetuximab for colorectal cancer 2009
*Gefitinib for lung cancer 2010
*Vemurafinib for metastatic melanoma (BRAF mutation)

Genomic testing before treating increases the cost of technology for the NHS as all potentially eligible patients must be tested, but only a subset who will benefit get treated. Yet if side effects are avoided testing may be cost effective, as was convincingly shown for Abacavir. The lower the prevalence of a mutation in the population the higher the impact on technology costs, and sometimes prevalence of a mutation can only be guessed at so making decisions about whether or not it would be good practice to do gene testing can be difficult.

An Australian group recently developed a framework for reviewing codependent technologies with personalized medicines, which is useful in decision making.

And then there’s quality assurance. If a test is recommended it needs to be performed reliably if we want it to properly inform personalised treatment decisions. Investment in quality assurance systems will be needed as more tests are developed, approved, and recommended

ROSE McCORMACK of AstraZeneca who spoke about the development of Gefitinib—which is effective for some individuals with advanced non-small cell lung cancer—mentioned that there is a huge need for educating patients (patient-information websites and more general information) about personalised treatments. She also said that, given lessons learned with the early personalised cancer treatments, in future co-development of drug and diagnostic test will be the norm (in the past one has followed the other).

PETER JOHNSON talked about the stratified medicine pilot study that Cancer Research UK has set up at eight clinical hubs in the UK in collaboration with with AstraZeneca, Pfizer, and the UK government’s technology strategy board. Their aim is to recruit 9000 patients with the following types of cancer: breast, bowel, lung, prostate, ovary, and melanoma. They will test the feasibility of offering detailed gene sequencing for all these patients and then acting on the results to try to ensure personalised cancer treatment.

BOBBY GASPAR a paediatric oncologist from UCL’s Institute of Child Health spoke about his work in the development of viral vectors to deliver modified genes in rare genetic diseases. The technology is in its infancy, although it appears to work well and the hope is to develop it further for the treatment of many more diseases. He said that the really interesting part would be seeing what happened when the academic work was finished. Would the therapy be picked up by a product developer, funded, and marketed?

JOERN ALDAQ of the Dutch company UniQure that develops genomic medicines told a harrowing tale that clearly showed how difficult it can be for a smaller company to get a highly novel, albeit effective, treatment for a disease that is very rare through regulators and to market.

He spoke about his journey through market authorisation for the gene therapy Glybera (for LPL deficiency). Four iterations of the proposal were submitted to the EMA. Although two expert committees recommended approval of the drug it fell twice before another committee (of the 27 member states’ representatives) because they hadn’t proven efficacy in enough patients. (An RCT would have taken 350 patients to achieve statistical significance and they estimated that there were probably only 250 patients in the whole of Europe with the disease. They had shown efficacy in 27 patients.) It cost the company about 70 million euros and it had to go into liquidation. Fortunately a venture capital company acquired it and they went ahead with developing the drug on a smaller scale. Eventually after the fourth submission the drug was approved for a licence.

June Raine of the MHRA assured us that regulation is changing, and that the UK government is playing a key part in this through the government’s health sciences policy. The idea of relying more on post-authorisation efficacy studies (that should be properly designed and planned at the time of drug regulation) to see if a drug’s efficacy is sustained in the long term is one that is now being thrashed out. This might allow novel drugs to gain a conditional licence earlier in the process of development.

Professor Adrian Towse, director of the UK’s Office of Health Economics attempted to answer the question “How should we value genomic medicine?” What if a therapy works to ensure a lifetime cure? How should we value that? How much should the patient/funder be expected to pay up front? He didn’t really answer it. I guess it’s something for which an answer is found when a potential-life-time-cure gene-therapy appears on the market. NICE does not, apparently, plan to use a traditional cost per QALY approach to estimating the cost effectiveness of orphan drugs.

What’s abundantly clear is that drug development in stratified medicine can not be done entirely within one pharma “house.” Mike HARDMAN, vice president of R&D Science relations at AstraZeneca
discussed how academics, pharma, regulators, and clinicians will have to work together to optimise development of personalised medicines. He said, “Industry has moved away from thinking that we can do it all alone towards sharing the research, as well as the risks and the rewards of new drug development.” He drew attention to the European alliance for personalised medicine’s [] “Call to Action” in 2012 set out 5 key calls to policy makers, politicians, and regulators in the EU. These include:

1.    Ensuring a regulatory environment which allows early patient access to novel and efficacious personalised medicine.
2.    Increasing research and development for personalised medicine.
3.    Improving the education and training of healthcare professionals.
4.    Acknowledging new approaches to reimbursement and HTA assessment, which are required for patient access to personalised medicine and its value to be recognised.
5.    Increasing awareness and understanding of personalised medicine.

And the bottom line? We are ALL going to have to learn about genomic medicine, whether we’re a clinician in a specialty where no gene-specific therapies seem to be on the horizon or at the forefront of delivering personalised medicine, and whether we’re medically trained or lay. Personalised medicine is the future. Let’s get our heads around it and help others to do the same so that we can avoid having a public backlash against each new treatment that is approved for only a subset of individuals with a given disease (remember Herceptin…?)

Kirsten Patrick is the editorials editor, BMJ.