LONDON - The introduction of cocktails of antiretroviral drugs to treat human immunodeficiency virus infection has greatly improved the outlook for many people with HIV and AIDS. Treatment sometimes fails, however, because of the emergence of viral strains that are resistant to several of the drugs involved.

Clinicians and pharmacologists need to know whether such multidrug-resistant strains are present at the outset in patients who have never taken antiretroviral drugs, or whether they arise by mutation during replication while patients are taking the drugs. The answer to this question will help determine the best pharmacological strategies aimed at eliminating multidrug-resistant strains, or reducing the likelihood of such strains emerging.

Ruy Ribeiro, of the Wellcome Trust Centre for the Epidemiology of Infectious Diseases at the University of Oxford in Oxford, UK, and Sebastian Bonhoeffer, of the Friedrich Miescher Institut in Basel, Switzerland, have now devised models to allow them to quantify which of these scenarios is most likely to be true. They report their work in the July 5, 2000, Proceedings of the National Academy of Sciences, in a paper titled "Production of resistant HIV mutants during antiretroviral therapy."

Bonhoeffer, group leader in theoretical and computational biology at the institute, told BioWorld International, "Provided multidrug-resistant mutations are there when the patient is starting treatment, one should try to find combinations of drugs that make it less likely that the virus has already found a solution that will make it resistant before therapy starts. Conversely, if resistant strains arise during treatment, the strategy would be to use drugs which are more potent at inhibiting virus replication."

Ribeiro and Bonhoeffer decided to take a mathematical modeling approach to the problem because they were aware that current assays to detect multidrug-resistant viruses were inadequate to allow an experimental approach. Bonhoeffer explained, "The expected frequency of these mutants in a drug-naive patient is very low. It could be 1 in 10,000 of the viruses or less, and current assays cannot detect resistant viruses at such low frequencies."

They therefore set out to use known data, such as viral mutation rates, the size of the viral population in the body, rates of clearance of infected cells and so on, to calculate the expected frequencies of drug-resistant mutants.

Using both a deterministic model and a stochastic one, they calculated that, in most cases, treatment fails because resistant strains pre-exist in patients who have never taken antiretroviral therapy. Writing in PNAS, they conclude, "It is generally less likely that resistant mutants are generated for the first time during treatment. This suggests that efforts to reduce the risk of treatment failure need to concentrate on the combination of drugs with different resistance profiles in order to minimize the risk that multidrug-resistant strains pre-exist in a drug-naive viral population. Increasing the efficacy of replication inhibition is only of secondary concern."

Bonhoeffer added, "This is encouraging news because it means that provided a drug cocktail can be found that is effective against all pre-existing virus mutants, then if you adhere to the prescribed drugs, the chances are that you will not develop resistant virus during therapy."

He is setting up a collaboration with researchers at the Aaron Diamond Center in New York. "We are interested in testing experimentally with new assays for the frequency of drug-resistant mutants prior to therapy," he said, "to verify whether the mathematical models we have used have done a good job of predicting these frequencies."