How To Pick A Winner In Oncology: An Interview With AstraZeneca's Andrew Hughes (Part 1 of 2)

In the contest to predict early winners in the pipeline game, the challenge of this decade is predictive efficacy, according to Hughes.

Big pharma's losing ways when it comes to picking clinical candidates that can go the distance are legendary. Almost every drug maker can point to changes in its pipeline strategy designed not only to avoid the threat of a financially devastating Phase III blow-up but also to guard against a newer peril: the danger of investing millions in a drug that isn't differentiated enough to earn reimbursement.

AstraZeneca is no different. But, not only is the U.K.-based drug maker focused on making go/no-go decisions early, it also aims to shorten the period between development phases, with the long-term goal of achieving an eight-year development cycle. So far the firm has seen the most impact in Phase I, where company-wide the time has been nipped from 20 months to near 10 (¹ (Also see "AstraZeneca R&D: Aim Early To Cut Cycle Time, Build Phase II Pipeline" - Pink Sheet, 18 Sep, 2008.)).

Having sorted out pharmacokinetics in the 1980s and predictive toxicology in the 1990s, the drug maker now is tackling the third challenge to identifying failures early: the headache of predictive efficacy, Andrew Hughes, global early development director for oncology, explained.

During a recent interview aimed at discovering what exactly pharma means by "translation," Hughes said that in AstraZeneca's transitional oncology program, a molecule has to pass six tests to be considered worth moving into the clinic. And even then, if it doesn't move the needle on its predictive biomarker(s) it will be abandoned.

"The Pink Sheet" DAILY : At AstraZeneca, what do you mean when you say "translational science"?

Andrew Hughes: The definition that we use very much within AstraZeneca is translating late-phase discovery into early clinical, so that we can make the decision on whether the drug is worthwhile investing in pivotal registration studies, and I guess the subtext that most people use is understanding how our drugs work in man.

When you break that down, certainly in oncology, we tend to really try and answer six questions before a compound has rights of passage into pivotal registration studies. The way we run that within AstraZeneca is by having about 100 people within our discovery department and 40 people in our clinical department whose day job really is to try and tease out the scientific basis to those questions.

"The Pink Sheet" DAILY: What moves are you making to identify failures earlier in development?

Hughes: There are three key things that we tend to look for here. Going back to why drugs failed in the '80s, far and away the majority at that point tended to fail because of poor pharmacokinetics, so the industry as a whole, including AstraZeneca, invested an awful lot in predictive pharmacokinetics. Now, we're surprised by the pharmacokinetic profile, and therefore terminate the drug, in only about 5 percent of drugs that we take into man. We have very good models and very good computer simulations.

[The next] issue became predictive toxicology. We'd sorted the pharmacokinetics, but then the tolerability in man didn't come up to spec. Again in the '90s and the first part of this decade we invested, a lot of us did, our competitors, in terms of predicting toxicity early.

So now again we have computer models that can demonstrate whether a drug is likely to produce effects on the heart or damage chromosomes and be toxic in man. So we can again screen out some of the encouraging leads from our compound screening library at a very early stage.

Having sorted out those two problems we are now left with this decade's challenge, which is the headache of predictive efficacy. Can we do preclinical models, or probably more beneficially, early efficacy predictors in man?

The Critical Path Initiative led by FDA has really spoken to the heart of that. Can we, by looking at biomarkers in man, pick up early hints of activity that we believe will translate to some clinical benefits?

If a drug has gone into Phase I, might have great pharmacokinetics, might have great tolerability, but if it hasn't changed these biomarkers, with a moral and ethical obligation to only try agents which we believe are going to be beneficial in man, we're confident to kill those molecules.

"The Pink Sheet" DAILY: If you're screening with biomarkers and you have something that looks perfectly safe but it's not going to move the ball, you will abandon it?

Hughes: Yes. As a matter of fact in cancer - I don't have the data at hand in the other therapy areas - but since the turn of the century we've killed two molecules in cancer for that very reason. They have ok pharmacokinetics, they have reasonable tolerability, but because they didn't hit the bar that we expect in terms of the biomarker activity, we reprioritized them and moved on to other molecules that truly did hit those biomarker hurdles.

"The Pink Sheet" DAILY: Is the availability of biomarkers a limiting factor?

Hughes: Finding biomarkers that are clinician-proof is not a small challenge. Sometimes things can work great on a bench, but when you move them into a hospital environment where the sample might sit on a laboratory bench for a couple of hours and go off, or it might preserve in the wrong way, or it might stand at Philadelphia airport for six hours on the tarmac in 40 degree heat. You don't know what happens to the sample in those conditions, so what the FDA calls qualifying a biomarker - making sure it's robust enough to be used in a real clinical trial - is in fact a science in and of itself with quality control measures that have to be gone through before you can deploy a biomarker in a clinical trial.

"The Pink Sheet" DAILY: Let's get back to those six questions.

Hughes: (1) Which tumor type do you want to position this molecule in? At times we can have a great drug, but try it in the wrong tumor type and we could be overlooking a major blockbuster.

2) What schedule do you try this drug at, because there are huge ranges of oncologic cancer schedules, from administering a drug once every 21 days to giving it four times every day of your life. And again we have great examples of drugs that were tried at the wrong schedule and were found to be too toxic, and so the schedule had to be amended until finally you could find a schedule which patients could tolerate and the efficacy of the drug could reveal itself. The taxanes are a good example of this.

3) Looking for things like survival in early phase trials is a tough task for a lot of our agents, so can we find pharmacogenomic biomarkers that give us a very early hint of predictive efficacy? We've invested a lot of time as a company and also working with the FDA biomarkers consortia to try and qualify biomarkers so we can at least demonstrate that the drug is having activity, biological activity, so that we can have the confidence in investing in a trial to demonstrate it has clinical activity as well.

And again, the translational group I head up spends a lot of time developing biomarkers that are relevant to both the drug and the disease in order to give us an early hint in Phase I and Phase II that the drug is actually hitting target and modulating the biology of the disease so we have confidence to do the more expensive Phase III investment.

4) How do we pull out the patients that are more likely to respond from those that aren't, so, personalized health care. We're looking here for predictive biomarkers rather than pharmacogenomic biomarkers. A great example of this is in Herceptin development, which, had it been tried in unselected populations, the response rate would have been in single-digit figures and the drug would have been abandoned.

But because it's targeted to those patients who just express Herceptin receptors on their cell surface, about a third of patients are getting benefit from this agent. Our own agent Iressa , which has just received European license for treatment of EGF positive non-small cell lung cancer, is our own example of that. In an unselected population, about 10 to 13 percent of patients typically respond, whereas in EGF mutation positive patients, the benefit rate is around 90 percent. So [there is] a really dramatic difference between unselected and selected populations.

5) The drug might be great on its own, but it might be even better in combination with another agent, and in oncology there are about 20 or 30 different standard of care agents that you could combine it with. Translational science is there to prioritize which of those is better tested with the agent because it might help in preventing either innate or acquired resistance.

6) The last question, which is really a question that patients themselves are asking, is 'tell me why this drug is better than anything else out there?' In other words, differentiation. And that differentiation might be it has a broader impact or has less toxicity or a more convenient formulation, but it has to be clearly a plausible differentiation from choices that patients and payers already have.

-Shirley Haley ([email protected])