Computational models that predict response to HIV therapy can reduce virological failure and therapy costs in resource‐limited settings

The results of genotypic HIV drug‐resistance testing are, typically, 60–65% predictive of response to combination antiretroviral therapy (ART) and have proven valuable for guiding treatment changes. However, genotyping is not available in many resource‐limited settings (RLS). The purpose of this study was to develop computational models that can predict response to ART without a genotype and evaluate their potential as a treatment support tool in RLS. Random forest models were trained to predict the probability of response to ART (<400 copies HIV RNA/ml) using the following data from 14,891 cases of ART change following virological failure in well‐resourced countries: viral load and CD4 count prior to treatment change, treatment history, drugs in the new regimen, time to follow‐up and follow‐up viral load. The models were assessed during cross‐validation, with an independent set of 800 cases, with 231 cases from RLS in Southern Africa, 206 from India and 375 from Romania. The area under the ROC curve (AUC) was the main outcome measure of the accuracy of the model's predictions. The models were used to identify alternative regimens for those cases where the salvage regimen initiated in the clinic failed. Finally, annual therapy costs were used to determine the potential cost effectiveness of this strategy for the Indian cases. The models achieved an AUC of 0.74–0.81 during cross validation and 0.76–0.77 with the 800 test TCEs. They achieved an AUC of 0.59–0.65 with cases from Southern Africa, 0.64 for India and 0.73 for Romania. The models identified alternative, locally available drug regimens that were predicted to result in virological response for 97% of cases where the salvage regimen failed in Southern Africa, 98% of those in Romania and 100% in India. Cost‐neutral or cost‐saving regimens that were predicted to be effective were identified for 88% of the Indian salvage failures with a mean saving of $638 per year. We developed computational models that predict virological response to ART without a genotype with comparable accuracy to genotyping with rules‐based interpretation. The models were able to identify alternative regimens that were predicted to be effective for the great majority of cases where the new regimen prescribed in the clinic failed. The models were also able to identify cost‐saving alternatives for most cases of failure in India. These models are now freely available over the internet as part of the HIV Treatment Response Predictions System (HIV‐TRePS), which has the potential to help optimise antiretroviral therapy in countries with limited resources where genotyping is not generally available.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.