Poster presentation at:
XIX International AIDS Conference
22nd July 2012 - 27th July 2012
Washington DC, USA
Revell AD1, Wang D1, DeWolf F2, Gatell J3, Ruiz L4, Pozniak A5, Perez-Elias MJ6, Lane HC7, Montaner JSG8, Larder, BA1 on behalf of the global RDI study group
1 RDI, London UK
2 Netherlands HIV Monitoring Foundation, Amsterdam, The Netherlands
3 Hospital Clinic of Barcelona, Barcelona, Spain
4 FundaciA3 irsiCaixa, Badalona, Spain
5 Chelsea and Westminster Hospital, London, UK
6 RamA3n y Cajal Hospital, Madrid, Spain
7 NIAID, Bethesda, USA
8 BC Centre for Excellence in HIV/AIDS, Vancouver, Canada
Background Computational models that predict virological response to combination antiretroviral therapy (ART) can help treatment decisions. Accurate models have been developed using a range of input variables, including treatment history, and considerable data collected from clinical practice over several years. It is critical to the utility of such models that they include the latest drugs in clinical practice. Here we developed random forest (RF) models, using limited or comprehensive treatment history information, that for the first time predict responses to raltegravir.
Methods 7,263 treatment change episodes (TCEs) were used to train two committees of 5 RF models to predict the probability of virological response to a new regimen. The input variables were baseline CD4 cell count, viral load and genotype (protease and RT), drugs in the new regimen and time to follow-up. Committee 1 also included six treatment history variables (zidovudine, lamivudine/emtracitabine, any PI, any NNRTI, enfuvirtide, raltegravir), whereas Committee 2 included separate variables for previous exposure to each of 18 individual drugs. Accuracy of the models' predictions was assessed during cross-validation and with 375 independent TCEs (obtained from: http://hivdb.stanford.edu) by plotting receiveraEUR"operator characteristic (ROC) curves.
Results The AUC for Committee 1 during cross-validation was 0.78aEUR"0.83 (mean=0.815) and for Committee 2, 0.80aEUR"0.84 (mean=0.820). The AUC for the 243 TCEs containing raltegravir ranged from 0.66-0.76 (mean=0.71) and 0.63-0.78 (mean=0.71). When tested with the 375 independent TCEs, Committee 1 achieved an AUC of 0.87 and Committee 2 0.86).
Conclusions The models achieved a consistent, high level of accuracy in predicting treatment responses. The difference between models with restricted or comprehensive treatment history variables
were minimal and not statistically significant. The accuracy of the models for raltegravir TCEs was somewhat reduced, reflecting perhaps the relatively small number of raltegravir TCEs. The models are being used to power HIV-TRePS, the free online treatment selection tool.
New predictive models accurately predict HIV treatment response despite missing data
Predicting HIV therapy response despite missing data - new models published
2021 update to HIV-TRePS: a highly flexible and accurate system for the prediction of treatment response from incomplete baseline information in different healthcare settings
New models that predict treatment outcome without the need for a baseline CD4 count or genotype added to HIV-TRePS
The most accurate models to date for predicting response to HIV therapy are released for clinical use
RDI expands the range of models available through HIV-TRePS to include their most accurate models to date
AI models predict absolute virological response to HIV therapy over time