If you have both in vitro and in vivo data pertaining to the resistance profile of a given drug, which is more important?
Moderator: Okay. Maybe we can move on to one last question here. I guess it's coming back to the clinical side again. And this actually was submitted by myself. And I'll address it to both Michelle Roland and to Doug Mayers. And it comes back to the issue of what information is the most important or the most dominant in interpreting these tests. For example, if you have both in vitro and in vivo data, which is more important? And I'll give you a specific example. The M184V mutation, which imparts high level resistance to 3TC, imparts relatively low level resistance in the laboratory to abacavir. At the same time mutations at positions 41 and 70 impart little, if any, resistance to abacavir. So if you had a combination of 41, 70, and 184 and you just looked at that set of genotypic mutations, it would say that the patient was in fact sensitive to abacavir because those low level resistance mutations were all that were found.
On the other hand, we know from Randall Lanier‘s data that the likelihood of responding to an abacavir-containing regimen with that group of mutations is actually quite low. So in this particular case I would argue that the clinical information always has to trump anything we see in vitro. So regardless of what we see in terms of genotypic information or phenotypic information, for that, if we have actual clinical trial results that say that that's not the correct interpretation, we have to go by the clinical trial or in vivo data? Doug?
Doug Mayers: I agree. I think that whenever possible you have to use the in vivo response data, and wherever we have it, that trumps any phenotype data type that we have.
I think we have two problems. One is that many of our drugs fail without a clean phenotype, and the other is that we haven't established phenotypic break points that relate to response at this point in time. This has only been done for a very limited number of drugs.
I think probably one of the best efforts we've seen is the recent study of ABT-378 by Abbott where they clearly defined their clinical response data in relation to a genotypic score and a phenotypic fold resistance and made clinically based cut points. And the hope is that we'll see that type of
effort for all new drugs and that we can try and obtain that type of data either genotypically or phenotypically for the drugs in current clinical practice.
Michelle Roland: Well, I also agree with you, and certainly in my clinical practice and in my teaching practice spend a lot of time thinking about what are the clinical outcomes data that we have available. And this is the perfect example, the abacavir example. It is an area of great concern to me when I watch clinicians try to interpret genotypes and try to interpret individual mutations in isolation, which I think is unfortunately still the predominant practice. So I think that there's a great risk that people are both using the genotypic information incorrectly and forgetting once again the limitations of these resistance assays and how we place these resistance assays into our larger knowledge base and decision algorithms in regard to an individual who has had virologic rebound on therapy. It's not as simple as what a resistance test tells you.
Mark Wainberg: As some of you know, I would argue that Randall is probably right and that there is a role for abacavir in this setting. But I think, as Doug was saying earlier, there needs to be clinical trials to address these and related issues. We have a lot of speculation right now. We have people who are interpreting a series of data mostly with some degree of bias I think one way or the other. And ultimately it's only going to be a randomized clinical trial that's probably performed that's going to give us the answers we need.
Jonathan Schapiro: Okay. Getting back, I think Steve, off the point is most important. I think it goes back to the previous question of a virtual clinical outcome. I think an attempt of that, as Doug mentioned, would be what was performed by Abbott. But I think we have to realize how difficult it will be to clean this data from trials, since the amount of variation in the response of patients is so great. To give maybe an example of the very nice study that was done by Abbott, those patients were actually in a very specific clinical setting in which a non-nucleoside reverse transcriptase inhibitor was given together with ABT378.
And all those cutoffs are only relative to a patient who is naive to NRTIs and receives an NRTI with ABT378. That is the caveat I think that we see that in each one of those individual situations will make accumulating this data so difficult. Therefore, if you have a patient who's the exact same patient but is experienced with an NRTI, so a very eloquent cutoff, lose their value.
Bob Shafer: I think one of the problems is that the primary data is really never made available. So with Randall Lanier’s data, we know that the more AZT mutations present, the less likely they'll be a response to an abacavir-containing regimen. To me it would be very interesting if all the data were present to see which of the three AZT mutations, which AZT mutations are more important than others. Are the AZT mutations perhaps a surrogate for something else?
It's the same thing with the ABT data. All they presented, they give a score of how many of these, I forgot if it was 9 or 11 mutations that were considered important, but the primary data on which mutations were present and how people responded and what the drug's susceptibility was is not available. So it makes it very hard to—I think Jonathan mentioned at one point that the clinical scenarios are very different. And that's the key thing with ABT-378, because they were receiving, as part of salvage, the non-nucleoside. But the other is without the actual data; it's very hard to really know what's going on. And you just have to take sort of peoples' word for it, but it's hard to extrapolate without the primary data.
