Discrepant results in the interpretation of HIV-1 drug-resistance genotypic data among widely used algorithms

The aim of this study was to assess the concordance on the interpretation of HIV-1 drug-resistance genotypic data by three widely used algorithms: Stanford University Database (SU), TruGene (Visible Genetics, Canada) (VG) and VirtualPhenotype (Virco, Belgium) (VP).

Genotypic data from 293 HIV-1-infected individuals with treatment failure was interpreted for 14 antiretroviral drugs by the three algorithms.

Complete concordant results among the three systems for all the drugs studied were found in 40/293 (13.7%) samples. Low concordance in the interpretation was observed for most nucleoside reverse transcriptase inhibitors (NRTIs), while results agreed highly for all nonnucleoside reverse transcriptase inhibitors (NNRTIs) and most protease inhibitors (PIs). In pair-wise comparisons, discordant interpretations between SU and VP were found in over 50% of the samples for didanosine, zalcitabine, stavudine and abacavir, and the level of disagreement between VG and VP exceeded 40% for the same drugs. Major discrepancies (high-level resistance interpretation by one algorithm with sensitive interpretation by another) were observed between VG and VP in over 10% of the cases for didanosine, zalcitabine, stavudine and abacavir. On the other hand, the three algorithms had concordant results for lamivudine in over 90% of the cases.

This work demonstrates the great level of discordance in the interpretation of genotyping results among algorithms, clearly showing the necessity for clinical validation. Moreover, these results suggest that a joint effort from the scientific community as well as national and international HIV societies is needed to achieve a consensus for the interpretation of genotypic data.