Development of a didanosine genotypic resistance interpretation system based on large derivation and validation datasets

To assess the genotypic determinants of the virological response to didanosine (ddI) in HIV-infected patients. The Forum database on ddI was randomly divided into a derivation set (n = 1000) and a validation set (n = 453). Linear regression models and bootstrap sampling were used to select resistance mutations and to estimate their resistance scores. Linear regression models, accounted for the censoring of viral load measurements due to assay lower limits, of the week 8 reduction in viral load from baseline were adjusted for baseline viral load, the exact number of weeks between baseline and the week 8 viral load measurements, and the Stanford genotypic sensitivity score. The ddI resistance mutations and their resistance scores based on the derivation set were as follows: M41L (score of 14), T69D (24), D123S (40), T139M (54), I180V (53), M184V (-12), V189I (55), Q207K (37), L210W (25), and T215Y (eight). The total score is obtained by adding the individual scores. Viruses with scores of 19 or less, 20-59, and 60 or more are considered sensitive, intermediate, and resistant, respectively. In the validation set, respectively, 58.7, 36.9, and 4.4% of viruses were predicted to be sensitive, intermediate, and resistant to ddI. The observed viral load reductions at week 8 were, respectively, 1.51 log10 copies/ml [interquartile range (IQR) 1.26-1.76] (P = 0.0001 versus resistant), 1.11 (0.94-1.30) (P = 0.0077 versus sensitive), and 0.46 (0.32-0.74) (P = 0.0079 versus intermediate). We developed a genotypic resistance score for didanosine including four mutations never previously used.