BACKGROUND:
Antiretroviral therapy response has been shown to be dependant of HIV drug resistance and sufficient drug plasma levels. Since a various number of interpretation systems (IS) are available for years, their predictive power on therapy response could be shown already. However, despite these findings comparative analyses including additional essential factors like therapeutic drug monitoring (TDM) and quantifying the effect of each IS are hard to perform. In this study we re-analysed a dataset of 131 clinical isolates from pretreated patients combining drug resistance interpretation of nine IS and TDM to evaluate the predictive power of the IS in a clinical setting.

METHODS:
For 131 patients genotypic drug resistance testing was performed before the antiretroviral treatment was changed. All analyses has been interpreted retrospectively by the following nine IS:Retrogram_v1.4 (RG), Rega_v5.5, ANRS_AC11, CHL_v3.2, Grupo de Aconselhamento Virologico (GAV), Detroit Medical Center 2000 (DMC), VGI_5.0, Beta Test of Stanford database (SDB-β) and geno2pheno (g2p). According to the IS an active drug score (ADS) was given for each drug [from inactive (0) to fully active (1)]. Linear regression was performed to analyse the correlation between ADS and viral load. For a subset of 66 samples TDM could be performed. The ADS for each drug were corrected by the results from the TDM analyses and linear regression analyses were performed a second time resulting in a TDMADS representing the remaining activity of the drug.

RESULTS:
Correlation coefficients (R) varied from 0.44–0.61 for ADS and from 0.47–0.59 for TDMADS. The best correlation was found for ANRS_AC11, respectively, although it was lower for the TDM-ADS analysis. For all other IS the TDM-ADS analyses correlated better than with ADS alone. Interestingly, it has been found that 1.3–1.9 active drugs would be necessary according to the IS to avoid viral load increase. These values were slightly lower for TDM-ADS analyses (1.2–1.7). To induce viral load decrease additional 0.5–0.8 active drugs/delta log (TDM-ADS: 0.5–0.7) would be indicated. This value increased for IS with higher R, but decreased in allTDM-ADS analyses.

CONCLUSION:
1) All IS were predictive for therapy response. However, differences could be found in particular for ANRS_AC11, that showed to be most predictive. 2) Including of TDM was more predictive than resistance interpreted by IS alone, indicating that there is additional benefit by performing TDM, although the differences were not high. 3) The finding that a substantial part of the new therapy would be necessary to avoid an viral load increase could represent the remaining activity of the pretreatment. Therefore, the influence of the actual pretreatment needs further to be evaluated.