BACKGROUND: Phenotypic assays are considered the gold standard for HIV-1 tropism assessment. However, they are expensive and not widely available. Genotypic assays may provide an easier alternative, but their sensitivity remains low. We hypothesize that combining clinical data with V3 sequences may improve the diagnostic accuracy of genotypic tools. METHODS: We analyzed clinical and biological data from 159 HIV-1-infected adults, 88 (56%) of whom were treatment experienced. Coreceptor phenotype was performed with Trofile and ES Trofile assay. V3 loop sequences were interpreted according to genotypic algorithms available at website. Multivariate logistic regression analyses were used to identify variables predicting HIV-1 tropism. Cut-off values for the prediction of CXCR4-using virus were defined. RESULTS: A total of 170 samples with phenotypic and genotypic determination of HIV-1 tropism were included. When only treatment-experienced patients were selected, a predictive model of HIV-1 tropism had an area under the receiver operating characteristic curve of 0.966 (95% confidence interval: 0.930 to 1.000, P < 0.001). The equation of the model included 2 bioinformatic tools (Geno2pheno-clinical model and net charge rule), the false positive rate score of Geno2pheno, and the following clinical data: exposure to more than 3 antiretroviral classes, years since HIV infection diagnosis and log10 HIV-1 RNA. A cut-off value ≥ 5.75 showed the highest accuracy to predict CXCR4 usage (96.6% sensitivity and 92.3% specificity). CONCLUSIONS: A genotypic-clinical model is highly accurate in predicting phenotypic tropism of HIV-1 in treatment-experienced patients. This may provide a cheap and rapid tool to select candidates for treatment with CCR5 antagonists in a routine clinical setting.
BACKGROUND: Phenotypic assays are considered the gold standard for HIV-1 tropism assessment. However, they are expensive and not widely available. Genotypic assays may provide an easier alternative, but their sensitivity remains low. We hypothesize that combining clinical data with V3 sequences may improve the diagnostic accuracy of genotypic tools. METHODS: We analyzed clinical and biological data from 159 HIV-1-infected adults, 88 (56%) of whom were treatment experienced. Coreceptor phenotype was performed with Trofile and ES Trofile assay. V3 loop sequences were interpreted according to genotypic algorithms available at website. Multivariate logistic regression analyses were used to identify variables predicting HIV-1 tropism. Cut-off values for the prediction of CXCR4-using virus were defined. RESULTS: A total of 170 samples with phenotypic and genotypic determination of HIV-1 tropism were included. When only treatment-experienced patients were selected, a predictive model of HIV-1 tropism had an area under the receiver operating characteristic curve of 0.966 (95% confidence interval: 0.930 to 1.000, P < 0.001). The equation of the model included 2 bioinformatic tools (Geno2pheno-clinical model and net charge rule), the false positive rate score of Geno2pheno, and the following clinical data: exposure to more than 3 antiretroviral classes, years since HIV infection diagnosis and log10 HIV-1 RNA. A cut-off value ≥ 5.75 showed the highest accuracy to predict CXCR4 usage (96.6% sensitivity and 92.3% specificity). CONCLUSIONS: A genotypic-clinical model is highly accurate in predicting phenotypic tropism of HIV-1 in treatment-experienced patients. This may provide a cheap and rapid tool to select candidates for treatment with CCR5 antagonists in a routine clinical setting.
Authors: A Gonzalez-Serna; R A McGovern; P R Harrigan; F Vidal; A F Y Poon; S Ferrando-Martinez; M A Abad; M Genebat; M Leal; E Ruiz-Mateos Journal: Antimicrob Agents Chemother Date: 2011-12-05 Impact factor: 5.191