Table 3.

Multivariate Logistic Regression Analysis Evaluating the Vi-rologic Function of 3 Weighting Schemes

Algorithma	Mean rGSS ± SD	Adjusted OR of rGSSb	Mean AUC ± SDc
Unweighted
ANRS	2.26 ± 0.94	1.70	0.76 ± 0.08
HIVdb	1.91 ± 0.91	1.71	0.76 ± 0.08
Rega	2.22 ± 0.91	1.63	0.76 ± 0.08
ViroSeq	2.14 ± 0.96	1.66	0.76 ± 0.08
Boosted PI weighting
ANRS	2.51 ± 1.03	1.89	0.77 ± 0.05
HIVdb	2.12 ± 0.99	1.88	0.77 ± 0.05
Rega	2.49 ± 0.91	1.83	0.77 ± 0.06
ViroSeq	2.37 ± 1.05	1.81	0.77 ± 0.05
Comprehensive weighting
ANRS	2.14 ± 0.97	2.14	0.79 ± 0.03
HIVdb	1.82 ± 0.88	2.22	0.80 ± 0.03
Rega	2.18 ± 0.91	2.15	0.80 ± 0.04
ViroSeq	2.02 ± 0.95	2.08	0.80 ± 0.03

NOTE. For boosted protease inhibitor (PI) weighting, the drug-specific genotypic susceptibility score (GSS) was multiplied by 1.5 before the regimen-specific GSS (rGSS) was calculated. This is the weighting used by the Rega algorithm. For comprehensive weighting, the drug-specific GSS was multiplied by 2.0 for boosted PIs and by 0.5 for nucleoside reverse-transcriptase inhibitors before the rGSS was calculated. ANRS, Agence National de Recerche sur le SIDA; AUC, area under the receiver operating characteristic curve; HIVdb, HIV Drug Resistance Database; OR, odds ratio; SD, standard deviation.

^aAlgorithm versions were as follows: ANRS 2007.10, HIVdb 4.3.1, Rega 7.1.1, and ViroSeq 2.8.

^bAdjusted ORs for attaining virologic suppression (plasma human immunodeficiency virus type 1 RNA level below the limit of quantification) for each unit increase in rGSS. For all ORs in this column, P values were <.001.

^cMean AUCs were calculated from the prediction results of multivariate logistic regression models tested in 10-fold cross-validation.