Thanks to new biomarker data, Scott Patterson signals success in cancer diagnostics.
April 1, 2008 | Four years ago, Amgen created a division of medical sciences. One of the founders was Australian Scott Patterson, an executive director responsible for many of the division's biomarker activities. The group of more than 200 physicians and scientists runs the company's early stage development studies, including the biomarker program, and has produced what Patterson hails as "the new story" in personalized medicine.
It's the story of the oncogene KRAS, its involvement in the epidermal growth factor (EGF) signaling pathway, and the relevance of this biomarker on Vectibix, Amgen's colon cancer drug. Amgen acquired the rights to Vectibix as part of the Immunex acquisition in 2002. The FDA approved the drug in 2006, but its European counterpart initially rejected the drug in early 2007. However by that time, Amgen was discovering compelling new biomarker data: patients with 'normal' KRAS genes tended to respond well to Vectibix, whereas those harboring KRAS mutations derived no benefit from the drug.
Here, Patterson discusses the results, which were first presented at the European Cancer Organization (ECCO) conference in September 2007 and published this month in the Journal of Clinical Oncology, and their implications with Bio-IT World's Kevin Davies.
Bio•IT World: Scott, what is the role of medical sciences within Amgen's drug development program?
Patterson: Four years ago the FDA critical path initiative described the phases of drug development, including a phase called "exploratory." Well, we established medical sciences three months before it came out, and that 'exploratory' phase described what Medical Sciences' remit was - the transition from "R" to "D." We thought we'd do something a little differently at Amgen. Normally you matrix in different components with the early development group - which consists of physician scientists, clinical pharmacologists, and clinical research study managers, the people who design and execute the first-in-human studies. We said, "we want biomarkers to be a component of our early clinical development program, so let's bring in bench lab departments, so that they're working together." Not just early development with matrixed biomarker resources, but actually one function. This function is now led by Brian Kotzin, and I have responsibility for about a third of that function. My department was formed from former functional genomics groups, so genomics and proteomics technologies. Roger Perlmutter, the head of R&D, had said he wasn't really interested in investing in technology for new targets four year ago, but would invest in technology to help develop drugs. And so, we wanted to bring those people in and apply their expertise into drug development.
You also brought in a computational biology group. What does it do?
One way they're involved is in helping us establish the appropriate way to get the assays run and the data analyzed. As an example, they introduced "design of experiments," which is a paradigm that's applied to understanding assay performance characteristics. If you want to run the complete matrix of the different parameters that you'd want to explore to understand your assay, it takes a long time. A smarter way to do it is called "design of experiments." This computational approach picks out different components you want to explore, making for an efficient assay qualification process... We're making [clinical] decisions so they have to be very robust assays. They help us with the analysis of the data as well. They're also involved in experimental design and analysis when we're doing discovery for biomarkers because we're still using, in some cases, 'omics technologies. So we need that expertise for mass spec data and microarray experiments.
What's so special about the Vectibix biomarker story?
Vectibix was part of the Immunex acquisition, so it was already in development before we even set up medical sciences. Vectibix received accelerated approval in the U.S. for refractory metastatic colorectal cancer patients in third-line therapy as a monotherapy with best supportive care in September 2006. But with the KRAS biomarker data, it's an exemplar for oncology development now. It is the example of personalized medicine. We've known about Herceptin for ten years, but this is new. The KRAS biomarker data is what's really showing how you can now look at the effect of a mutation downstream of a targeted receptor to define a patient population.
Are you saying this is the gold standard now for personalized medicine?
Yes, this is a prime case study for personalized medicine in oncology. This is the new story on the block... So, Vectibix (or Panitumumab) is a fully human monoclonal antibody that blocks binding of the ligand to the EGF receptor (EGFR). In our clinical trials, approximately 10% of patients responded to the therapy - they had tumors that were primarily driven through the EGFR pathway...
Thirty years ago, oncogenic mutations were discovered in RAS. It's known as a poor prognostic factor. So, if you have KRAS mutations, that's not good for disease progression. The theory was, if we have activating mutations of KRAS, then blockade of the receptor upstream would be irrelevant. Therefore, if you have a KRAS mutant population, then they won't respond to the therapeutic, so that'll be a negative predictor...
We knew it was a smaller percentage of patients who were responding, we had to try to understand the strata within the population. So, we and others began to assay samples from Phase 2 studies, and it looked like there were no KRAS mutant samples from patients in the population responding to anti-EGFR antibodies, which is what we were expecting. We then confirmed these findings in the pivotal Phase 3 randomized trial. KRAS mutation frequency in colorectal cancer is around 50%, reported ranges are 30 to 57%.
How do you do test for the KRAS biomarker?
We ended up running the [Manchester-based] DxS assay - an allele-specific PCR-based assay which uses their amplification refractory mutation system (ARMS), that essentially amplifies only mutant alleles, so very high sensitivity [is achieved] in a background of wild-type DNA... Whenever you have the tumor that's excised, and it's put into a [paraffin] block format and then you take sections, not all the tissue is tumor tissue. If you want to do direct sequencing - PCR-amplify a region and do population-based sequencing - you need to remove all of the normal tissue because of sequencing sensitivity... We don't need to that with the DxS assay. There has to be a minimum amount of tumor, but we don't have to go in there and remove normal tissue because of the 1% sensitivity, whereas with direct sequencing, it's got to be more than 20%.
[By then,] we'd also submitted Vectibix for approval in the E.U. In March 2007, the EMEA came back with a negative opinion, said it doesn't really look good enough guys. We were just poised to run our samples... We established a pre-specified data analysis plan, ran the samples at a Belgian CRO (HistoGeneX), data came in and were analyzed. When we submitted the data to the CHMP [Committee for Medicinal Products for Human Use] at EMEA to review, they gave us a positive opinion... So we got conditional approval in metastatic colorectal cancer for Vectibix in wild-type KRAS patients.
Was the EMEA's earlier rejection because your positive responders were masked by the non-responders?
Yeah, I presume that was their basis. Importantly, the [KRAS] mutation frequency in our pivotal trial was 43%, pretty much what everybody sees. So it's pretty simple: half of them are KRAS mutant, half are wild types - you essentially double your response rate, thereby increasing the benefit to risk ratio through exclusion of non-responders.
All of the benefit resides in wild-type Panitumumab-treated [patients]. You couldn't dream of a better result! But statistically speaking, it's a negative predictor. We're predicting non-response... The responders are in the wild-type population, but there are also some non-responders there. So, we're continuing to try to see if we can chip away at this. What other pathways are affected? Is it loss of PTEN? Activating PI3 kinase mutations? Is it over expression of the ligand? ...
For a company whose mission statement is to serve patients, [we're] really living up to the mission statement because we don't want to have the drug given to patients, exposed to the drug, who can't respond, who have no likelihood of response. And this essentially takes out 50% of the population.
One of the conditions of approval in the E.U. was that there was a CE-Marked test kit available. The EMEA doesn't specify a test kit, but we wanted us to make sure there was one available, so we worked with DxS to ensure that they had their test CE-Marked in 22 countries. Obviously we want this in the U.S. So, we've presented this data to the FDA and we're working with DxS and the FDA so that this test can eventually be approved [in the U.S.]
The FDA originally approved the drug without any patient stratification data. Don't your corporate colleagues feel a little ambivalent about reducing the target population?
Nobody feels good about the drug being delivered to people who haven't got a chance of responding. So, we're just in the stages of discussing with the FDA and DxS to determine the best approach to share this information with prescribers.
For the wild type patients, what's the average clinical benefit to the drug?
One of the things that you'd always like to see is overall survival, but in this trial, because the patients who were on the best supportive care alone were allowed to crossover to the Panitumumab arm once they had progressed, it's very difficult to get overall survival data... So it's not really ethical to leave patients on the control arm when they've progressed.
What has been the early response of the cancer community?
They were really pleased to hear about it. The drug is now launched in the E.U. Tests are available. Interestingly, a lot of hospitals in the E.U. have been sequencing KRAS anyway. In Austria, they've been doing it for ten years. So, they didn't need a kit sort of forced upon them to do this analysis, but the EMEA wanted to make sure there was a kit available... M.D. Anderson, Harvard, Cleveland Clinic all have KRAS testing capability. Genzyme does, for example, but it was for lung cancer. But I'm sure they'll have a colorectal cancer test...
It sounds similar to the EGFR/Iressa story in lung cancer (See, "Minority Report," Bio•IT World, June 2004), in a way the answer was right there in front of people?
Well, people have said, "You knew it for 20 years, why wasn't this a stratification factor?" Pre-clinically it didn't really work out that way, which was surprising. And we knew it was a poor prognostic factor, so it's difficult to tease apart prognostic implications from the predictive implications. And this is the first randomized controlled trial showing this data and it's so clear...
This is one of the reasons we set up medical sciences. We're now working on [oncology, inflammation, and metabolic] programs to get ahead of the game. We would much prefer to define the potential diagnostic before we go into Phase 3, so that we would have the initial stages of validation of the kit accepted by the FDA. We'd be randomizing patients on that basis. We may still be putting positive and negative biomarker patients into the trial, but the hope is you'd complete your trial, your response data would look good, [and] your companion diagnostic would get approved at the same time. That's what the FDA would love. That's what we'd love. Unfortunately, you don't always know enough physiology going in.
Further Reading: Amado, R.G. et al. J. Clin. Onc. 2008: April 1
The EGFR signaling pathway. Inhibition of EGFR can halt tumor growth, but not in the presence of activating KRAS mutations.
This article appeared in Bio-IT World Magazine.
Subscriptions are free for qualifying individuals. Apply Today.