Revell AD, et al. - Researchers developed two sets of random forest models using 50,270 treatment change episodes from more than 20 countries in order to predict absolute viral load over time in HIV. Using variables including baseline viral load, CD4 count, and treatment history, the models estimated viral load at different time points following the introduction of a new regimen. In addition, genotypes were used in one set’s predictions. They noted the achievement of highly significant correlations between predicted and actual viral load changes with both models, indicating their utility for choosing the optimum combination treatment for patients needing a change in therapy in situations using any definition of virological response. Information regarding the likely response curve over time could also be gained using the models. For those in resource-limited settings, these models could be a useful addition to the HIV-TRePS system in view of no requirement for genotypes.