The use of computational models to predict response to HIV therapy and support optimal treatment selection

Slides presentation at:

Scientific Days of the National Institute of Infectious Diseases

10th November 2011 - 30th November -0001

Bucharest, Romania

Revell AD1

1 HIV Resistance Response Database Initiative (RDI), London, UK

Combination antiretroviral therapy (ART) has made the long-term suppression of HIV and prevention of HIV disease progression a reality in well-resourced health care settings. Nevertheless, despite the availability of approximately 25 antiretroviral drugs from six classes, viral breakthrough, often with the emergence of drug-resistant virus, remains a significant challenge. Sustained re-suppression of drug resistant virus requires optimal selection of the next regimen. The complexities of resistance and the number of potential drug combinations make optimal individualised sequencing of therapy highly challenging. For physicians with limited experience or resources, treatment decisions can be even more problematic.

In well-resourced settings, a genotypic resistance test is usually performed following treatment failure. Interpretation of this genotype can be complex and is usually performed using rules-based software that relates point mutations to viral susceptibility to single drugs. However, different systems provide different interpretations and it is difficult to relate the results for individual drugs to the likely responses to potential drug combinations. Indeed, raw genotype sensitivity scores have been shown to be relatively weak predictors of virological response.

Computational models that provide a quantitative prediction of virological response to combination therapy, rather than to individual drugs, directly from the genotype and other clinical information may offer a potential clinical advantage. However, this can be challenging given the quantity of data required to incorporate a range of prognostic variables, multiple possible drug-genotype permutations and their respective drug response data. The HIV Resistance Response Database Initiative (RDI) was established in 2002 explicitly to take on this challenge to collect data required to develop such models.

Currently we have data from approximately 85,000 patients, predominantly from Western Europe and North America but also from Africa, Australia, Japan and most recently Romania. We have trained computational models to predict virological response to treatment from genotype, viral load, CD4 cell count, and treatment history with 80% accuracy. This compares favourably with a 50-70% predictive accuracy for genotypic sensitivity scores.

In order to assess the clinical utility of the RDI tool, a web-based user interface was developed through which clinical investigators could gain access to the models and the system was evaluated by HIV physicians in two clinical pilot studies. One third of treatment decisions were revised on the basis of the systemaEUR(TM)s predictions and the revised regimens were predicted to produce significantly greater virological responses and involve fewer drugs. The system was rated as being easy to use and a useful aid to clinical practice.

This approach could be of particular clinical utility in resource-limited settings (RLS). However, the models described above require genotyping, which is not generally available in RLS. We therefore decided to explore modelling response without a genotype. In the absence of substantial data from RLS, prototype models were developed with cases from well-resourced settings. The accuracy of the no-genotype models was only diminished by 4-5% compared to those using a genotype.

Previous studies have shown that models are most accurate for patients from aEUR~familiaraEUR(TM) settings from where the training data were collected, a concern was that this encouraging performance would not be replicated when the models were applied to cases in RLS. New models were trained using data from almost 15,000 treatment change episodes (without genotypes) from well-resourced countries. The models achieved accuracy of 72% with cases from the same settings and 60-67% for cases from RLS and Romania. The models identified alternative, available drug regimens that were predicted to result in virological response for 63-100% of cases that had resulted in failure in the clinic.

In conclusion, computational models can predict virological response to ART with a high degree of accuracy. They can identify potentially effective alternative regimens for the majority of cases of treatment failure and have been shown to be useful as a support tool for treatment decision-making. The RDIaEUR(TM)s HIV Treatment Response Prediction System (HIV-TRePS) is now available as a free experimental treatment support tool at

This approach has the potential to improve patient outcomes in RLS and to reduce treatment costs through the avoidance of unnecessary or premature switching to expensive, newer drugs, even in settings where genotyping is not generally available. The RDI is working to further improve the accuracy and utility of the system through the collection of sufficient data from specific regions, such as Eastern Europe and sub-Saharan Africa for the development of models specifically for those regions.

Slides in PDF View Slides

Join the RDI mailing list

Enter your email address below to start receiving the RDI newsletter.

You can unsubscribe at any time. View the RDI privacy policy.