The HIV Resistance Response Database Initiative (RDI) developed these latest models specifically for use in the many settings where genotyping is unaffordable. They estimate the probability that any combination of HIV drugs will reduce the amount of virus to undetectable levels in patients whose current therapy is failing. They were trained with data from tens of thousands of patients in clinics all over the world, including for the first time, patients from Southern Africa. They were around 80% accurate, which is significantly better than the 57% accuracy achieved by genotyping.

"This study and these models are proof of principle that this could be a very helpful approach for selecting effective therapy in highly resource-constrained settings, such as Southern Africa," commented Professor Robin Wood, Head of the Desmond Tutu HIV Centre, University of Cape Town, South Africa and a co-author on the paper." As more of our patients fail first and even second line therapy, it is critical to optimise the selection from our limited range of drugs to achieve maximum suppression of the virus and this system could be very useful."

The new models are now available to be used by healthcare professionals as part of the RDI's HIV Treatment Response Prediction System (HIV-TRePS), which is freely available online at www.hivrdi.org/treps. The system also enables users to enter the costs of the drugs in their clinic and so model both the cost and the effectiveness of various treatment options. A previous pilot study using data from an HIV clinic in India indicated that the system could identify more effective and less costly combinations of drugs that those actually used in the clinic.

"Currently, most HIV patients in resource-limited settings are treated according to WHO public health guidelines that offer very limited treatment options," explained Dr Hugo Tempelman, Clinical Director of the Ndlovu Care Group, Elandsdoorn, South Africa and co-author on the paper. "The HIV-TRePS system, incorporating these models, enables doctors to tailor the HIV treatment based on the cost and predicted effectiveness of the treatment. What is wonderful is that HIV-TRePS provides us with high predictive value at no cost."

The data required by the system for its predictions includes a measure of the amount of virus in the patient's bloodstream (the viral load), a test that is not widely used in resource-limited settings. However, the potential cost savings offered by the system are likely to cover the costs of viral load testing, many times over. Moreover, viral load monitoring of HIV therapy is now recommended by the World Health Organisation in all settings and initiatives are underway to fund it, including the formation of the Load Zero Foundation, formed specifically with this goal.³

"We are very encouraged by the results with these models." Commented Dr Andrew Revell, Executive Director of the RDI. "They show that not having access to genotyping in resource-limited settings need not be barrier to providing individualised, optimised HIV drug therapy."