Interview with Don Smith – Community HIV Research Network, Sydney, Australia
Dr Don Smith is an HIV clinician and the coordinator of the Australian Community HIV Research network – the organisation which overseas HIV/AIDS treatment trials in clinics and hospitals around Australia. He spoke to Megan Nicholson in Sydney about the recent Resistance and Treatment Strategies Workshop.
ABT 378
Q: ABT is one drug that is currently available as a salvage therapy option. Was there any new information about ABT-378’s resistance profile and its effectiveness as a salvage therapy?
There was limited data presented on ABT-378 which is supposed to work against many PI resistant strains. Abbott pointed out that it was difficult to get good quality data because so few people failed this drug.
In the study they looked at people who were PI experienced, viral loads above 1000 but non-nuke naive. They put them on an open label study of ABT 378 plus efavirenz, plus recycled and new nukes. They had 57 patients who started off with quite high viral loads of about 80,000 (abstract 89). They found that if you had more than six PI mutations then you had a poor response to new treatment – only one in three patients responded. If you looked at the phenotype, 93% responded when they had less than ten fold change in sensitivity. For a 10-20-fold change, 78% responded; 20-40-fold change, 67% responded and if it was more than a 40-fold change, then 50% responded.
Q: Did the meeting think that was a good response?
A: The Abbott presentation was challenged immediately concerning the relative role of ABT 378 in the response rates. All the participants were getting efavirenz and studies of efavirenz indicate that you’d expect that at least half of this effect was caused by efavirenz. I was sitting beside some Canadian physicians who were all saying "buy, buy, buy" and then after the questions they were saying "sell, sell, sell".
Q: Good to hear that they were thinking of their patients.
A: They were, but what they were intimating was that a lot of the presentations were quite commercial – presentations with a lot of spin on them. But you didn’t know what effect efavirenz and the nucleoside had. Abbott was trying to say that all this benefit was ABT 378 but others were saying ‘No, it’s not. You can’t say how much is (ABT 378), you haven’t done a randomised trial. It’s an open label study, everyone is on the same treatment and you are really fudging it by saying that these changes in sensitivity have accounted for your clinical responses.’
The meeting was really left up on the air as to the effectiveness of ABT against virus with resistance mutations. No one really knew. It wasn’t clear from that presentation but that was about all that Abbott was presenting on ABT 378. So while it looked like a really good presentation, like a number of presentations, there was a very overtly commercial slant.
I think that was probably more frustrating at this meeting than some of the other meetings in the past, where there was a lot more of honesty about the data. A number of presentations were from the companies and they were much more competitive about the way they presented the data than previously.
Resistance testing
Q: Was there any data to counteract the recent NARVAL findings which found very little benefit in resistance testing?
A: Cal Cohen of New York presented the final results of the VIRA 3001 study - phenotype versus standard of care, in people who were PI experienced and had more than 200 copies. They showed that by 16 weeks the change in viral load in those that had a phenotypic test was 1.23 logs compared to 0.87 log in those on standard of care treatment – which was a significant result. The number who were undetectable was 59% versus 42%.
Brendan Larder put it best – the value of resistance testing is equivalent to adding one new agent into our combination. You are getting an additional half log difference with resistance testing.
VIRA 3001 was presented at the same time that NARVAL was updated. NARVAL showed at 16 weeks no difference between genotype or phenotype or standard of care. At 24 weeks, genotype looked a bit better.
Interestingly enough, in the VIRA study, there was no difference in CD4 count. So that small difference in viral load didn’t correlate to a meaningful difference in CD4 count. The people who didn’t get the resistance test weren’t any worse off immunologically in the short term.
The key difference in these studies was the patient group. The VIRA 3001 patients had all failed their first PI combination whereas the NARVAL patients had been exposed to an average of 7 drugs. NARVAL was suggesting that it doesn’t really matter what patients’ resistance profiles are if you can’t find new drugs that are still active.
Resistance testing might have a role in people who can come up with good new drug combinations. Unfortunately, most people will want resistance testing for those who have undergone multiple failures because they are the group least likely to benefit.
Q: International guidelines generally recommend use of resistance testing or consideration of use of resistance testing. Was there any shift in people’s view of those recommendations?
A: Not in terms of the whole meeting. There’s certainly some role for resistance testing but there are still a lot of grey areas.
There was one good presentation that compared the VIROLOGIC with the VIRCO phenotype - the two main companies involved with resistance testing. VIROLOGIC are trying to capture the US market and VIRCO are trying to capture the European market. That presentation showed very good correlation on matched samples. That indicates that both the assays are probably reasonably accurate. Limitations apply, but one is not outstanding and you could use either.
Q: How does the interpretation of those results play a role if the assays are fairly comparable?
A: The interpretation is more critical than the assay. With some of the assays you just get a print out from the company which looks at the dominant virus. The companies have different cutoffs levels for resistance. VIROLOGIC say that anything under a 2.5 fold change in sensitivity is meaningless, anything above a 2.5 change and you are starting to call that resistance. VIRCO use a 4-fold change as their cut off level with anything above 10-fold being highly resistant. However, research indicated that the ‘one-size fits all’ approach doesn’t work for all drugs.
You can get a large variation in sensitivity to the non-nukes. The drug levels you can achieve with the NNRTIs are often greater than 100 times the IC50, so a 10-fold change is clinically meaningless for NNRTIs but a 10-fold change for some other drugs is very meaningful. So the cut off levels for the NNRTIs should be greater than 10 before it is called resistance. Resistance to PIs is a much more variable thing. A 2-3 fold change is probably significant there.
Challenging research on drug concentrations
Q: Were there any key controversies at the Resistance Workshop?
Andrew Hill was the man who got up at nearly every clinical presentation at the Geneva World AIDS Conference in 1998 and asked presenters if they used an intention-to-treat analysis. Out of that meeting everyone was forced to standardise the way they report.
Andrew got up again at this meeting and said that a number of companies are actually looking at two different pieces of data: the IC50, IC90 or IC95 for resistant viruses, and the minimum plasma concentration required to inhibit that virus. They were coming up with were little schemas that examine the Cmin of a drug (usually a boosted PI) and compare it to the IC50 of resistant virus. The implication is that you boost the PI levels well above levels that would work against resistant strains.
Andrew Hill was saying, ‘A load of bollocks. You can fudge the data in many different ways’. He was saying that Cmin concentrations don’t necessary reflect the amount of active drug around. More importantly, the way you measure the IC50, IC90 or IC95 should be standardised. He also went on to say that there was no such thing as a single IC50 for a drug; whenever you test it in the lab you end up with a range of IC50 because you’ve got multiple strains of virus all with slightly different sensitivities. So your IC50 can vary a lot and if you just pick out one then you can move the data set anyway you want to.
The other significant factor is the amount of protein binding that you account for. Usually measuring your Cmin is modelled, ignoring the huge interpatient variability that exists. When you calculate the IC50, you’re often doing that in a tissue culture where you grow the virus and add drug to see how much it takes to inhibit 50% of the virus – that’s not 100% human serum, yet you are saying that the IC50 generated is suddenly equivalent the IC50 you achieve in human serum. That’s not to negate the whole principle but to say that we all have to be a bit more careful because the companies can present this data in ways that are going to look good for them.
Viral fitness
Q: Turning to viral fitness, were there studies presented addressing viral fitness?
A: More discussion than a lot of data presentation. There were previous data from the Frankfurt cohort saying that for people failing therapy (more than 100 000 copies of virus and less than 50 T cells), those on treatment survived longer. There was discussion about the relative fitness of different strains of virus. It’s made complicated because you don’t get uniform changes or a single mutation that occurs all the time.
There‘s a little bit of preliminary data suggesting that it was the PIs that were more important than any of the other drugs in maintaining the less fit virus. The 184 mutation is only 75% as fit as wild type and this has led people to try to maintain that in any failing therapy. But there is increasing speculation that PI resistance may be worth maintaining.
Q: Was there any specific data?
A: There were a couple of posters about viral blips – people responding to therapy who suddenly got a low level detectable result. Dianne Havlir looked at whether people were more likely to fail therapy if they had these blips.
Q: Were they?
A: No. Maybe because they are priming their immune system with low levels of virus. But it was really hard to say because she filtered the group so much. She only looked at those people who went undetectable, who had the occasional single blip and not a sustained rise. Tests were conducted on single stored blood samples that weren’t repeated to look for lab variations. If anything, the study suggested that these transient blips shouldn’t be cause for panic attacks.
Dianne used an assay that went down to 3 copies. She found that people who were below that point were less likely to blip, and the blippers tended to be between 3-50 copies. However, in terms of their long-term failure rates the blippers seem to do slightly better. Having a little bit virus there might stimulate your T cells to be more reactive to the virus. It’s always a balance; you need to have an immune response against the virus but if you have too much virus that becomes destructive.
All these studies were heavily criticised as well because they completely ignored other factors such as compliance.
Q: What does viral fitness mean for treatment strategies then?
A: It means we are still unsure if you need to be much more aggressive with people on treatment, in changing their treatment around, or if you should be happy maintaining low level replication and potentially unfit virus.
What should you do with people with low level replication on treatment? I think generally it would still be acknowledged that, if you have good opportunity to suppress that virus because you’ve got a number of new drugs to use, you should do that. But for the increasing number of patients where that is not an option, they are probably better to stay on treatment and try to maintain a wimpy virus. You could call it the armour-plated virus - to survive in the war zone it‘s got so much padding on it, but it is slow and ponderous.
STI
Q: The notion of structured treatment interruption has taken a battering recently. Was there any new evidence to support STI?
A: The Spanish group presented some early data on their patients who had undergone three treatment interruptions. Five of the eight patients were below less than 10,000 copies after their third interruption, indicating that in some people you can re-establish some sort of setpoint. How long that endures is not clear and it obviously doesn’t apply for everyone.
The CD4 counts all dropped to their baseline levels during the interruptions. This indicates that it would be a risky strategy in advanced patients. For the people who were unfortunately treated a few years ago very early, who are now worried about toxicities, it might have a role.
Summary
Q: Do you have any final comments or observations?
A: I suppose there’s the increasing complexity of resistance. There are lots of mutations being described, some that will confer good resistance to some compounds, but not to all compounds in the same class. More variation, decreased susceptibility or hypersensitivity is being found. Some of the people failing PIs still respond nicely to other PIs. But given the huge number of mutations that could arrive within the virus, and given that you don’t see a single common pattern arising, interpreting resistance is really quite complicated.
I suppose that’s where VIRCO’s virtual phenotype might provide some salvation. They have genotyped and phenotyped 40 000 samples. You can email results of a genotypic test to VIRCO. They are using neural networks – almost semi intelligent computer networks – to search through databases, comparing your mutations with previously noted mutations and correlates with phenotype.