Personalized Medicine Grows Up

Dr. Edwin Clark, Director of the Oncology Biomarker Team at Bristol-Myers Squibb, explores the increasingly important role of biomarkers in the pharma drug development toolbox and shares his hopes for the future of personalized medicine.

To use an analogy: personalized medicine is in late adolescence right now, with all the promises and issues that a teenager has. The first papers on personalized medicine surfaced in 1999, so that’s what I call birth. Early on, one of the issues that this child had to deal with was the Human Genome Mapping Project, completed in 2000. There were a lot of very high expectations placed on personalized medicine and how the results of the project would impact it. That’s a pretty heavy burden for a one-year old child – especially when we needed specific tools to be developed to help us extract the information necessary for personalized medicine to become a reality.

There’s also the need for what I call the parents of personalized medicine – in this case, the regulatory authorities on one side and the pharmaceutical and biotech industries on the other – to understand how best to realize the promise of this growing child. In April 2005, the FDA published a concept paper on drug diagnostic co-development, which is a blueprint for how to operationalize personalized medicine.

The next step was for the pharmaceutical and biotech industries to fully embrace the potential of personalized medicine, and then the diagnostic industry had to get on board. (In some sense, I think of them as a godparent). The maturation of this child has not happened as rapidly as some might have hoped, but there’s greater acceptance today of the concept of personalized medicine than five years ago. Considering the concept was only born seven or eight years ago, many of the pharmaceutical and biotech companies are now planning for the inevitability of personalized medicine.

At Bristol-Myers Squibb, every one of our oncology development programs has examined the potential value for patients of a personalized medicine plan. More than 90 percent of our programs have a development plan that includes the use of biomarkers that will enable a personalized medicine approach. So we’ve moved from an approach where biomarkers and personalized medicine is a ‘nice-to-have’ five years ago, to one in which it’s a must-have. The potential in other therapeutic areas outside of oncology is being addressed as well.

What caused pharma-biotech to see the light at this point in time? We realized it’s in the best interests of patients as well as pharmaceutical and biotech companies who want to develop drugs that reach the right patients sooner.

Finally, all of the work that we do to identify why certain patients do or do not respond to a given therapy gets fed back to our discovery colleagues to assist them in the selection of novel drug targets. For instance, when the expression of gene X predicts a lack of response to a therapy ‘number one’ this may suggest that gene X may be a good target for a novel therapy that gets combined with therapy ‘number one’. The results of such work are being utilized in our drug discovery efforts and are the results of a ‘holistic approach’ to drug discovery and development. This approach may take longer because the drugs themselves take longer to be developed, but I think we’ll have more success.

Biomarking for success
We discover, develop and use pharmacodynamic – or what we call PD – markers to help us in dose selection, predictive markers that we use in patient selection, and surrogate biomarkers that help to find activity in our compounds. Most of the work we do is in discovering and developing pharmacodynamic markers to help us get the right dose of our drug and predictive markers to get the right patients.

I have personal experience of how PD markers have been used in phase one development to support what we call a ‘proof of confidence’ decision, meaning that the biomarker result increased our confidence that the compound would be effective in phase two and phase three development because the biomarker told us that the compound was effective against the target.

Even more important are examples where the marker told us that the compound was not as effective in patients against the target as we expected based on pre-clinical models. Such results enable us to make a no-go decision sooner and hopefully reduce the number of compounds that fail later in development because they aren’t effective enough against the target. We believe that this will drive greater success in late-stage clinical trials where the cost of failure is significantly higher.

With regard to predictive markers, there are examples where a given therapy is most effective or safe in say 10 to 50 percent of patients and there are markers that can define this subpopulation, yet the drug has previously been developed against the entire population not against this select population. Such therapies typically tend to not be very successful – either regulatory agencies don’t approve them or, if they get approved, the payers become skeptical about therapies that are really marginally effective. Examples such as Herceptin, Avastin, and even our drugs ERBITUX and SPRYCEL, fit into this category where a drug will be developed in a specific subpopulation of a specific indication. This is likely to be the norm going forward rather than the exception.

We’re currently using array-based technologies such as the oligonucleotide micro-arrays that measure mRNA levels and SNP arrays, which measure SNPs but also can be used to measure NDA copy number changes to help discover and develop biomarkers. The advancement of these technologies in terms of their cost, speed and the quality of the data has enabled companies like Bristol-Myers Squibb to screen the entire genome to identify the best biomarkers that may someday be turned into diagnostics. We start well before a compound makes its way into the clinic by utilizing these array technologies in pre-clinical testing of candidate compounds to identify the candidate’s pharmacodynamic markers and predictive markers that we then test in early clinical studies. We use these technologies in our clinical discovery efforts when we utilize the specimens actually collected from patients in the clinical trials.

Progres
In my four years at Bristol-Myers Squibb, I have been involved in several projects around selecting, developing and validating biomarkers relevant to cancer, but I want to focus on two. The first is SPRYCEL, a drug approved for use in chronic myeolia leukemia (CML). It’s a drug I started with very early on so I saw the drug from the clinic through regulatory approval and beyond.

We developed an assay to measure the pharmacodynamic activity of SPRYCEL in the phase one studies, which helped us to define the dose we took forward into additional studies. We also generated biomarker data on BCR-ABL to better understand why patients who were resistant to Gleevec were sensitive to SPRYCEL treatment. We carefully sequenced the BCR-ABL gene in pre-clinical models and then in our patients in the CML trials and we were able to show that SPRYCEL was effective against the majority of mutations in BCR-ABL that conferred resistance to Gleevec.

The accelerated development of SPRYCEL and its rapid acceptance by regulatory authorities demonstrates the strategic importance of using biomarker information gathered from both the pre-clinical and the clinical studies. It enabled Bristol-Myers Squibb to be more confident in the results we were expecting in our registrational studies and to put the resources behind the development of this compound. In the end the result was one of the fastest development stories in the history of the pharmaceutical industry, and one that we like to talk about because it’s really a model for how to do it right.

Another example that we recently published on is the identification of candidate markers that predict survival in colorectal cancer patients who are treated with ERBITUX. We used array-based technologies in both pre-clinical and clinical samples and identified the EGFR ligands – epiregulin and amphiregulin – as markers that can double the response and time to progression in colorectal cancer patients. Now we’re doing further work with a diagnostic partner in this area to validate these markers in additional lines of therapy with a plan to utilize a diagnostic based on this work to get Erbitux to colorectal patients in those earlier lines of therapy. I’m excited about this because I think it will allow us to live up to the personalized medicine pledge of getting the right drug to the right patient at the right time. We believe that there are patients who are currently not receiving this drug who should be and we believe that these markers will help us identify those patients.

As Director of the Oncology Biomarker Group, Dr. Edwin Clark focuses on the identification and use of biomarkers in the development of the oncology portfolio. In this capacity, he has been involved in the early development of a number of cancer therapies at Bristol-Myers Squibb, including SPRYCEL (dasatinib), and the panHER and panHER/VEGFR compounds as co-chair of the exploratory development team. In addition, Dr. Clark was involved in the completion of pharmacogenomic-specific studies for ixabepilone in breast cancer and ERBITUX (cetuximab) in metastatic colorectal cancer.