Computer Models Are Changing How We Craft HIV-AIDS Policies

A computer modeling of AIDS-afflicted populations of sub-Saharan Africa has recently enabled the World Health Organization (WHO) to identify a cost-effective measure to address the increasing prevalence of drug-resistant HIV in the region.

The HIV Synthesis Model, developed by Andrew Phillips, PhD, and colleagues at the Institute for Global Health, University College London, in the United Kingdom, is an individual-based simulation model of HIV transmission, progression and treatment response. The computer model incorporates extensive data from published sources, accounting for demographics and behaviors, as well as specific drug effects and HIV resistance mutations.

The program generates simulated populations of individual adults along with quarterly interval status throughout a lifespan, or a designated period such as 20 years, for a range of factors, including HIV testing, condomless sex, and risk of HIV acquisition

>> READ: Big Data and Machine Learning Take on HIV

Although the model was initially created to analyze patterns in the UK, Phillips and colleagues were tasked with applying the HIV Synthesis Model to sub-Saharan Africa, which is struggling with insufficient resources to provide antiretroviral treatment (ART) to more than 18 million HIV-infected patients.

The daunting challenges of high affliction rates across large geographic regions that are under-resourced have been compounded by emergence of treatment-resistant HIV. Public health workers have learned that the virus infecting approximately 10% of patients is resistant to the nonnucleoside reverse-transcriptase inhibitor (NNRTI) component of the WHO recommended first-line drug regimen.

In their report on the application of the modeling program to the growing problem of ART resistant HIV in sub-Saharan Africa, Phillips and colleagues note that approaches were analyzed for both affordability and effectiveness.

The HIV Synthesis Model “is probably one of the models that contains most detail on drug resistance to specific drugs, so it is quite well suited to study questions relating to choice of drug regimen,” Phillips told Healthcare Analytics News^TM.

Simulating Populations of Infected Individuals

The model analyzed 2 options. One is to test for NNRTI resistance at time of treatment initiation, with the WHO alternative dolutegravir (Tivicay, ViiV Healthcare)-based regimen provided in lieu of the first-line efavirenz (Sustiva, Bristol-Myers)-based regimen for those with treatment resistance. The other option is to designate the alternative as the new first-line regimen for all patients and obviate the need for, and cost of, testing and the logistics of reestablishing contact and treatment after obtaining results.

Parameters tracked in the individuals modeled for acquiring HIV include viral load, CD4 cell count, occurrence of WHO stage 3 and 4 HIV disease condition, clinic attendance and drop-out, use of specific antiretroviral drugs, presence of specific resistance mutations, adherence to ART, and toxic effects of particular regimens of ART.

Among the assumptions incorporated into the model for this application is that dolutegravir would exhibit a similar rate of resistance to that of the protease inhibitor, atazanavir (Reyataz, Bristol-Myers Squibb) boosted with ritonavir (Norvir, Abbvie). The investigators inferred this would be approximately 27 times lower than the resistance rate for efavirenz.

>> READ: Hunting for the Heart of a Changing Community

Phillips and colleagues report the modeling predicted that a transition to a dolutegravir-based first-line regimen in regions with more than 10% NNRTI drug resistance would be the most cost-effective approach and achieve the most health benefits, with a reduction of about 1 death per year per 100 people on ART over the next 20 years.

The results were considered by the WHO Guidelines Group, which subsequently issued a consensus statement recommending that regions with that prevalence of NNRTI resistance “urgently consider” a first-time ART regimen that does not contain an NNRTI.

“The urgency of the transition will depend largely on the country-specific prevalence of NNRTI resistance,” Phillips and colleagues noted.

Other Purpose-Built Models

The HIV Synthesis Model is one of several that have been brought into the HIV Modelling Consortium in the Department of Infectious Disease Epidemiology at the Imperial College London, which is funded by a grant from the Bill & Melinda Gates Foundation. The Consortium leads coordinated interaction between modeling groups and policy makers, in conjunction with targeted and responsive commissioning of new work.

“Its central objective is to help improve scientific support for decision making by coordinating a wide range of research activities in the mathematical modeling of the HIV epidemic,” according to a statement.

In addition to the HIV Synthesis Model, examples of others in the Consortium and their uses include:

• ASSA: Assess effects of prevention and treatment programs, demographic impact of HIV, and understanding of epidemic drivers.
• BBH: Investigate resource allocation issues, including cost-effectiveness of combination interventions and optimal allocation of a fixed budget across different interventions.
• Birger Saigon IDU Model: Consider HIV-HCV co-infection, prevention, mortality.
• Eaton, Hallett, Garnett 2011: Explore interaction between concurrent sexual partnerships and elevated infectiousness during primary HIV infection.
• ICRC HIV Transmission Model: Evaluate ART as treatment and as PrEP (pre-exposure prophylaxis).
• Menzies TB-HIV: Weigh burden of this comorbidity and evaluate resource utilization of interventions for both diseases
• PopART: Examine effects of universal testing and universal “test and treat.”

The HIV Synthesis Model, described as an individual-based stochastic simulation model, is calibrated with Approximate Bayesian Computation to input parameter values that enable modeled outcomes that are similar to what would be observed if there were sufficient surveillance data available in the real-world setting.

“However, it is nowhere close to being able to replace the need for actual clinical studies, of course, in discerning differences in outcomes between where such studies are feasible,” Phillips told HCA.

“We have used the model now in around 20 papers on different questions, and I have been quite surprised that new questions continue to emerge for which our model can provide some useful insights in relation to potential effectiveness and cost-effectiveness,” Phillips said.

Get the best insights in healthcare analytics directly to your inbox.

Related

Aetna and KCC Spar in 2 Lawsuits Over HIV Data Breach

How EHR Data Revealed Disparities in the Fight Against HIV

Data Suggest Demise of Tuberculosis Is Slowing. But Tech Offers Hope.