Many of the presentations at the 3rd International Workshop on HIV Drug Resistance contained data on correlations between HIV-1 RT and protease mutations (genotype), drug susceptibility (phenotype), and response to anti-HIV drug therapy. However, despite the large number of presentations, no fundamentally new concepts emerged. The data were presented in a variety of different formats and many presentations were designed mainly with the intention of promoting individual drugs or diagnostic tests rather than furthering knowledge on drug resistance. The primary data were also often not provided, making the studies not very amenable to peer review and subsequent analysis.

The presentations described the following types of studies: (1) prospective clinical intervention studies in which patients were randomized to receive or not receive genotypic data on their virus prior to the selection of a new treatment (Viradapt and GART), (2) retrospective clinical studies in which genotypic testing (sequencing) or phenotypic testing was performed on samples collected prior to the start of a new drug regimen. (3) laboratory studies describing the phenotypic effect of specific mutational patterns, and (4) description of databases containing RT and protease sequences.

The prospective intervention studies suggested that genotypic tests are useful in allowing clinicians to choose treatment regimens that will lead to a greater short-term virologic response. The retrospective studies consistently showed that the number of mutations related to protease inhibitors (PIs) and RT inhibitors (RTIs) was predictive of a response to a new therapy. But there were too few studies correlating specific mutational patterns with in vitro drug susceptibility results, and no studies suggesting exactly how knowing the genotype of a patient’s virus could be used to help in the selection of a new treatment regimen.

Prospective clinical intervention studies: The Viradapt study has just been published in Lancet and has been reviewed on this site [1]. Viradapt enrolled 108 patients with plasma HIV RNA levels >10,000 copies/ml despite ^>/= 3 months treatment with a PI-containing regimen. 65 patients were randomized to have their next drug regimen selected on the basis of RT and protease mutations; 43 patients were randomized to have their next drug regimen selected on the basis of published guidelines alone. At month 6, plasma HIV RNA levels decreased 1.15 log copies/ml in the genotype group and 0.67 log copies/ml in the control group (p=0.05). Clevenburgh presented virologic data on study subjects during open-label follow-up [2]. At months 6 and 9, both treatment groups were offered open-label genotype testing. At month 12, the mean plasma HIV RNA reduction remained 1.15 log copies/ml in the genotype group, and increased to 0.98 log copies/ml in the control group patients who had open-label genotypic testing.

Preliminary results of the CPCRA 046 (GART) trial were first presented at the 6th Human Retrovirus Conference earlier this year. Baxter presented the final results at the Workshop [3]. GART randomized 153 patients with virologic rebound (3-fold rise in plasma HIV RNA) after 16 or more weeks on a PI-containing 3-drug regimen. 78 patients were randomized to received genotype interpretations and suggested drug regimens; 75 patients were managed without genotypic results. The primary endpoint was the mean of the 4 week and 8 week measurements and patients receiving the genotype had greater plasma HIV RNA decreases 1.19 vs 0.61 log copies/ml (p<0.001). At week 12, the difference between study arms decreased to 0.44 log copies/ml but was still statistically significant (p=0.003). Mayers [4] reported that in univariate analyses the RT mutation, T69D, and the protease mutation, L90M, were associated with poorer response to salvage therapy, whereas the protease mutation, D30N was associated with a better response to salvage therapy. RT mutations at positions 184 and 215 were present in most patients and were therefore less predictive of virologic response.

Retrospective clinical studies: Harrigan [5] presented retrospective data on 59 heavily treated patients (failed ^>/= 2 treatment regimens, received a median of 7 anti-HIV drugs) started on a new multidrug salvage regimen. Genotype and phenotype assays were performed using the Virco recombinant virus assay on stored samples and therefore the results were not used to influence treatment selection. The initial (presumably 4 or 8 weeks) plasma HIV RNA reduction was 0, 0.9, 1.6, and 1.3 log copies/ml for regimens containing 0, 1, 2, and 3 effective agents (as measured by Virco's genotype interpretation algorithm (Vircogen)) and 0.2, 0.05, 0.9, and 1.5 log copies/ml for regimens containing 0, 1, 2, and 3 effective agents (as measured by Virco's phenotypic antivirogram assay).

Moyle [6] described the 12 week response to salvage therapy in 52 heavily treated patients (100% had received nucleoside RTI (NRTIs), 81% had received PIs, and 48% had received nonnucleoside RT inhibitors (NNRTIs). Genotyping was performed in realtime and was used to help select the salvage regimen. After a median follow-up of 14 weeks following the start of salvage therapy, the median plasma HIV RNA decreased 1.6 log copies/ml. The number of key NNRTI mutations (K103N, Y181C) was predictive of poor virologic response in patients starting NNRTI-containing salvage regimens. The number of key PI mutations were predictive of poor response in patients starting PI-containing salvage regimens.

Lorenzi [7] correlated baseline genotype to clinical response in 62 patients starting a nelfinavir (NFV)-containing salvage regimen. Patients had a median CD4 count of 113, a median plasma HIV-1 RNA level of 5.2 log copies/ml, and had received RTIs for a median of 3 years and PIs for a median of 1 year. In a multivariate model, the number of RTI plus PI mutations was the only independent predictor of response to therapy (PI mutations were more predictive than RTI mutations). The overall response to NFV-containing salvage regimens was poor. The median plasma HIV RNA decrease was 0.38 logs and only 32% of patients had a ^>/= 1.0 log decrease at 4-12 weeks.

Condra [8] examined the use of indinavir (IDV) in 30 patients who had previous virologic failure while receiving NFV and reported that the subsequent response to an IDV-containing regimen was significantly worse for patients with isolates having the L90M protease mutation (11 patients) than for patients having the D30N mutation (17/30 patients).

Laboratory studies: Ziermann [11] examined patterns of cross-resistance to amprenavir in viruses from 200 patients failing therapy with NFV or IDV and reported that protease mutation N88S (which occurred with both drugs) caused 2-10-fold increased in vitro susceptibility to amprenavir.

Databases: Larder [12] described the proprietary Virco database linking genotypic and phenotypic data on >7000 isolates and the approach used to determine the "virtual phenotype" of a newly sequenced isolate. I described the publicly available, on-line, Stanford HIV RT and Protease Sequence Database (http://hivdb.stanford.edu; [13]). This database contains nearly all published RT and protease sequences (>5000) linked to patient treatment histories. Earlier this year, a program for analyzing user-submitted RT and protease sequences was added to the database-backed web site. In August, a new version of the database and web site will become available and users will have access to an expanded number of queries and to drug susceptibility data.