Binomial Response Model Used to Evaluate Subtype and Treatment Effects on Genotypic Evolution for Each Protease and RT Position

<p>A separate model was created for each non-B subtype. The frequencies of mutations at each position in four patient groups (untreated subtype B, treated subtype B, untreated non-B, and treated non-B) were converted to <i>Y</i> scores using a cube root transformation (similar to a logistic transform). Subtype effect was evaluated by calculating θ_1, the score differences between non-B and B subtypes in untreated persons_. The treatment effect was evaluated by calculating θ_2, the score differences between treated and untreated persons within the same subtype. The subtype–treatment interaction was evaluated by calculating θ₃, the difference of differences in the 2 × 2 table, or the difference in treatment effects between non-B and B subtypes.