Home Search Abstracts Browse Sessions Program Committee View Session E-mail Abstract Author

 

 

Session 121 Poster Abstracts
Impact of Drug Resistance on Virologic Response and Clinical Outcomes
Friday, 1:30 - 3:30 pm
Hall A


705
Validation of Clinically Relevant Breakpoints for HIV-1 Phenotypic Resistance Data
Bart Winters*1, A Rinehart2, J Montaner3, P Harrigan3, D Castor4, S Hammer5, B Wasikowski6, M Miller7, S Emery8, F van Leth9, P Robinson10, J Baxter11, B Gazzard12, A Pozniak13, and L Bacheler2
1Virco BVBA, Mechelen, Belgium; 2Virco Lab, Inc, Durham, NC, USA; 3British Columbia Ctr for Excellence in HIV/AIDS, Vancouver, Canada; 4New York Academy of Med, NY, USA; 5Columbia Univ, New York, NY, USA; 6xLeo, Inc, Durham, NC, USA; 7Gilead Sci, Foster City, CA, USA; 8Natl Ctr in HIV Epidemiology and Clin Res, Univ of New South Wales, Sydney, Australia; 9Intl Antiviral Therapy Evaluation Ctr, Amsterdam, The Netherlands; 10Boehringer Ingelheim Pharma Inc, Ridgefield, CT, USA; 11Robert Wood Johnson Med Sch, Camden, NJ, USA; 12Chelsea and Westminster Hlthcare NHS Trust, London, UK; and 13Chelsea & Westminster Hlthcare NHS Trust, London, UK

Background:  Cut-off values based on clinical outcomes data (CCO) can enhance the clinical utility of HIV-1-resistance tests by optimizing the selection of treatment regimens. We defined CCO for the vircoŽTYPE resistance analysis, based on treatment response to > 3150 HAART regimens in either clinical trial or cohort settings.

Methods:  Quantitative phenotypic resistance levels were predicted by VircoŽTYPE analysis of viral genotypes. For each drug, linear regression models of virologic response were derived as a function of baseline phenotype (fold change in IC50) and viral load, activity of the background regimen (cPSS) and treatment history parameters. Cutoffs were defined as VircoŽTYPE baseline fold change values associated with 20 and 80% of reference response observed for susceptible viruses. The variability of proposed CCO was assessed by bootstrap testing. The ability of the models to discriminate between different responses (c- index) and the odds ratios for treatment response (> 1 log drop in viral load or below quantification limit at week 8) per resistance class, defined by CCO, were assessed on the training data and an independent test dataset of 352 regimens. The test dataset was also a mix of trial and cohort data. 

Results:  Response models were developed for 16 drugs (boosted and un-boosted protease inhibitors modeled separately), 11 of which demonstrated non-overlapping 90% CCO confidence intervals. Median c-index of these 11 models on the training dataset was 0.76 (range 0.70 to 0.86). The test data set contained regimens useful for evaluation of 9 models resulting in a median c index of 0.77 (0.68 to 0.79). Median odds ratios for response per resistance class were 0.10 (0.01 to 0.23, 11 drugs) on the training data, and 0.25 (0.002 to 1.07, 9 drugs) on the test dataset. The odds ratios for response per unit increase in the cPSS for the entire treatment regimen, calculated using CCO, were 0.33 (training data) and 0.40 (test data).

Conclusions:  Clinically relevant breakpoints for vircoŽTYPE phenotypic resistance data have been defined using a consistent approach and definition across multiple drugs based on virologic response. The CCO and the models used to derive them performed well on an independent dataset. These CCO should enhance the clinical utility of vircoŽTYPE in the selection of optimal antiretroviral treatment regimens for HIV-1+ subjects.

 

 

Keywords: Clinical cutoff; HIV Drug Resistance; Phenotype