Because we understand the molecular mechanism of action in our drugs, the very first time we go into human beings, we try
to see if the scientific hypothesis that we have made in preclinical research translates to the human situation. An example
is the Alzheimer's vaccine. We try to make the human body make antibodies against beta amyloids. Well, at the same time we
are now testing for safety, we are also testing that proof of concept. Previously, that would have been done much later in
For several years, we have been systematically doing that for all of our drugs going into the clinic. Actually, even during
the selection criteria—if the development and research teams cannot come up with a scientific readout that indicates that
the working hypothesis translates to the human situation, these drugs now have a lower priority than the ones that do have
such a concept.
This is a very important go/no-go criteria for full development before it costs the hundreds of millions that it does in Phase
IIs and IIIs. If I can put the drug in people, I should see this or that enzyme go up or down, and if it doesn't happen, then
the drug is dead then and there.
In terms of toxicology and safety studies, are there any new ways of testing early on that are going to be particularly applicable?
Yes. First, a lot of things that used to be done only in animals can now be done earlier in vitro. So there are quite a number
of in vitro assays developed that can predict prohibitive side effects in people.
Second, just like we do expression patterns of drugs with the clinical studies, the toxicologists are generating databases
of drugs that fail for toxic reasons and seeing if there's a typical genetic fingerprint and other commonalities. For example,
the family of genes that would lead to liver failure.
Now, if you could do that earlier, possibly in animals or even in in vitro situations, rather than waiting until you are in
expensive human studies, you could eliminate drugs that have a lower probability of survival because of these side effects.
There is a big European program that we are discussing to share those databases—they are proprietary and they are the historical
experience of that company with particular drugs. The question arises, if these drugs are not going to be competitive because
they cannot be translated into real medicine, could companies be convinced to pool this data and thereby find common genetic
patterns that are predictive of toxicology much faster than if each company does it on their own? That is one of these things
that people are thinking about.
Novartis is already starting to contribute some drugs with expression patterns, and several other companies are also signing
So it's moving forward.
Yes, hopefully. Of course, the scientific question that's still open is, Will we find sufficiently predictable patterns to
really be able to say with some level of confidence that one drug should not be further developed, while that one is OK? But
these methods all need to be validated and confirmed. And what will be great is if we can get a sufficient level of reliability
that the regulatory authority accepts these methods. That might help to reduce the number of animals used in safety studies.
It's still too early to call. As usual, we will learn by experience that there are some classes of drugs for which it works
very well and others much less. Especially for completely new mechanisms, it will be more difficult.
How is this similar to FDA's Critical Path Initiative that's encouraging companies to share information?
Oh, this is very similar, but it is under the auspices of the European Union.