Treating everyone to halt the HIV epidemic in the worst-affected countries may not be the most cost-effective use of antiretroviral drugs as a prevention tool, according to mathematical modelling by London’s Imperial College Infectious Disease Epidemiology group. Nor would universal treatment and annual testing always be necessary to achieve profound reductions in new infections, they find.
Their modelling exercise, published in the March 13th edition of the journal AIDS, highlights the extent to which sexual behaviour in the population could influence the success of a strategy – universal testing and treatment – that has been proposed as a means of virtually eliminating the HIV epidemic in countries like South Africa within 20 years.
The Imperial College group also demonstrate that in some circumstances, expanding treatment coverage to 80% of those with CD4 counts below 350, and getting everyone to take an HIV test every four to five years, could be the most cost-efficient strategy for reducing new infections.
The model is a response to another widely discussed mathematical model, developed by a group within the World Health Organization, which concluded that new HIV infections could be eliminated by 2030 in high-burden countries like South Africa if universal annual testing was introduced, and immediate treatment provided for anyone found to be HIV-positive.
The WHO modelling exercise, led by Reuben Granich and Brian Williams, sparked huge interest and controversy.
At a consultation last November researchers and civil society representatives discussed the model, the implications of the approach and how its feasibility could be assessed in research studies.
One of the chief concerns raised was the need for further modelling work to test the robustness of the finding that more extensive treatment would drive down the scale of the epidemic in severely affected countries.
Imperial College is one of the institutions carrying out its own modelling exercise on 'test and treat' strategies.
Its model differed from the WHO model in several key respects. Firstly, it looked at the effects of treatment and testing in populations with different patterns of sexual behaviour. It also looked at the effects of different thresholds for initiating treatment, and different frequencies of testing, on long-term HIV prevalence and incidence.
The modelling exercise broadly confirmed the results of the WHO model, and of a similar model published by Professor Julio Montaner and colleagues in 2006. Scaling up treatment coverage in 'hyperendemic' countries would have a profound impact on new HIV infections.
However, the model also showed that impact would be highly dependent on the character of the local epidemic. In settings where people with large numbers of sexual partners frequently had sexual contacts with people who had a very small number of lifetime partners, new infections would be reduced by 85% but would not be eliminated.
The model also showed considerable variation in the necessary frequency of testing according to the sexual behaviour pattern in the population. In a setting where the level of sexual mixing between 'high-risk' and 'low-risk' was relatively low, testing people every two years would be necessary to bring down incidence by 90%. But where more sexual mixing took place, it might be necessary to diagnose every person with HIV within one month of infection to achieve a similar impact.
Depending on the epidemiologic context, say the authors, incidence might be reduced by 95% if 80% of the population were to be tested only every three to four years, and started treatment at a CD4 count around 400 (close to the threshold recently recommended by the World Health Organization).
Their analysis of the cost-efficiency of different frequencies of testing and treatment similarly found that in some epidemics the most cost-efficient strategy would be test everyone every four to five years and initiate treatment at a CD4 cell count of 350 to 400.
Only in the most 'robust' epidemics, where incidence has stubbornly failed to decline despite years of prevention efforts, would more frequent testing prove to be cost-efficient. In these settings, testing 80% of people every two to three years and initiating treatment at a CD4 count above 450 would be the optimal strategy.
“It is likely that the 'test and treat' approach is much better suited to some populations and poorly suited to others,” they conclude. “There are diminishing returns for increasing testing frequencies to once-per-year levels.
“Failing to achieve sufficiently high coverage levels or failing to test frequently enough could just lead to a dramatic spiralling of treatment costs.”
Reductions in incidence of 85 to 95% would take around 30 years to achieve, so – in the short term – treatment costs would rise.
They also speculate that targeting particular population groups or locations for testing, such as truck drivers or beer halls, might prove particularly effective. However, targetted approaches are likely to raise concerns over the human rights implications of 'test and treat' strategies, which have been strongly voiced at a number of international meetings.
But more information will be needed to refine the model, and the numbers produced by the modelling exercise shouldn’t be regarded as precise estimates, they say.
Dr Timothy Hallett, one of the epidemiologists who developed the model, told aidsmap that the group is now doing further work on the model, incorporating data on behaviour, HIV incidence and prevalence from a range of countries.
Dodd PJ, Garnett GP, Hallettt TB. Examining the promise of HIV elimination by 'test and treat' in hyperendemic settings. AIDS 24: 729-735, 2010