Using a dynamic, compartmental model our analysis indicated that a five-component test-and-treat strategy such as we have proposed might dramatically reduce new HIV infections over the next 20 years within a heavily affected U.S. population, namely urban MSM. Our strategy required implementing multiple interventions to ensure widespread and frequent testing of the at-risk populations and greater and more comprehensive provision of treatment. Our analysis explored the effect of implementing each intervention individually and combined to both intermediate and best-case levels of improvement.
While improving each intervention individually had some effect, the most significant impact resulted from improving all simultaneously. Even with intermediate levels of improvement in the implementation of these interventions, their combination reduced the cumulative number of new HIV infections over 20 years by 39.3% and reduced HIV prevalence from a projected 18.3% to 13.3%.
Among individual interventions, the most effective was increasing the annual testing rate. Diagnosing previously undiagnosed HIV-infected individuals was associated with a substantial reduction in risk behavior and enabled infected individuals to enter care with the possibility of achieving viral load suppression. However, attaining a 48% annual testing rate will likely require considerably expanded HIV screening and potentially costly outreach to the infected, undiagnosed.
Our findings suggest that initiating ART at the intermediate CD4 count of 500 cells/mm3, according to current guidelines, could by itself have a substantial impact on HIV incidence, reducing the expected number of new HIV infections by 8.5% over 20 years. In comparison, initiation of treatment at diagnosis, the best-case scenario, resulted in a reduction of 11.7% over 20 years. The fact that the bulk of the benefit was associated with the intermediate case is not surprising considering that diagnosis, in the base case, occurs on average at CD4 300 cells/mm3, meaning that few individuals would be diagnosed in time to start ART at counts higher than 500 cells/mm3.
Viral load suppression on ART, another important variable in our test-and-treat strategy, required not only linkage, but also retention in care, as well as adherence to treatment. Although there is a fairly extensive literature on strategies to promote adherence to antiretroviral therapy 
, less is known about how to promote retention in care 
. Meeting the intermediate targets for earlier initiation of treatment and viral load suppression may require an increased commitment to the provision of treatment and to programs that maintain individuals on treatment once they begin. Nearly 9,000 low-income individuals with HIV were on waiting lists to receive treatment in September 2011 
. A longitudinal study in Nigeria showed that patients who started treatment at higher CD4 counts were at greater risk for dropping out of care and for non-adherence 
Implementing notification of test results and linkage to care to intermediate levels of improvement led to a more modest reduction in number of new HIV infections over 20 years (0.9% and 2.1% respectively). The reduced impact of these interventions seems to be due to already relatively high estimated notification and linkage rates and thus more limited room for improvement. If the proportion of persons tested with conventional testing is more than our assumed 50%, then improved notification may be more important. However, if the proportion of persons tested with rapid testing is more than our assumed 50%, then notification will be higher than our estimates; this may occur as rapid testing becomes more widely used.
The results of our sensitivity analyses indicated that a reduction in condom use to 50% from 75% among all urban infected MSM negated the benefits of the intervention. This finding underscores that safer sex practices such as condom use must be maintained in the MSM community.
Published reports vary regarding reductions in HIV transmission associated with suppression of plasma HIV RNA viral loads. By our estimate, even with an 80% reduction, a value that is at the lower bound of reported estimates, implementation of a test-and-treat strategy to intermediate levels of improvement could reduce the number of new infections by 26.4% over 20 years. If we assumed viral load suppression achieved a 99% reduction in the risk of HIV transmission, then the number of new infections over the same period was reduced by 51.6%. Our analyses conservatively assume that ART confers a reduction in transmission only when viral load suppression is achieved.
We also explored the ability of a test-and-treat strategy to achieve the shorter-term goals laid out by the National HIV/AIDS Strategy. We found that a combined test-and-treat strategy such as the one we have proposed that achieves intermediate levels of improvement in each intervention can meet four of the five goals of the National HIV/AIDS Strategy. Thus, a multi-component test-and-treat strategy could be a valuable part of the National HIV/AIDS Strategy.
Other models forecasting the effect of test-and-treat strategies on HIV epidemics in North America have been constructed and parameterized somewhat differently than ours. Our results are similar to those that assessed the impact of test-and-treat strategies on new HIV infections in British Columbia, where a 37% to 62% reduction in new cases over 25 years was estimated if the proportion of eligible individuals who received ART increased from 50% to 100% 
. Another modeling exercise of the epidemic in the United States indicated that a strategy of test-and-treat could achieve an 18% reduction in new HIV infections over 20 years 
. That model focused on low- and high-risk populations and assumed more modest reductions in risky sexual behaviors than we did (i.e., 20% vs. 50%) associated with the diagnosis of HIV infection.
Our model extrapolated past trends and current conditions into the future. If future conditions change with the emergence of more effective behavioral or biomedical prevention strategies, then the prevention benefit from the test-and-treat strategy we have proposed would likely change as well. Further, the benefits of a test-and-treat strategy as forecast by our model depend on correct estimates of levels of current implementation. For example, the timing of initiation of ART is an important parameter that we estimated using data from a cohort of HIV-infected MSM receiving care. However, this cohort may not be representative of all HIV-infected MSM, particularly those with limited access to care and persons who do not reside in metropolitan areas of the U.S. Similarly, our estimate that 80% of persons who initiate ART are able to achieve and maintain viral load suppression, likewise taken from HIV-infected MSM in care, may be optimistic. However, in both cases, our model likely underestimates the benefits of a test-and-treat strategy.
We did not assess the potential effect of antiretroviral resistance. Increases in the prevalence and transmission of antiretroviral resistant HIV could reduce population-based responsiveness to treatment that in turn reduces the effectiveness of a test-and-treat strategy. Findings from a previously published modeling exercise using population-level resistance data indicated that the effect of antiretroviral resistance on the projected number of new HIV cases was minimal 
Our analysis was based on New York City data. We assumed that the analysis may extend to other urban areas in the United States. However, for all estimates of current implementation, there is likely to be variation by race, ethnicity, age and geography.
The findings from our model provide decision makers with more information on how best to implement a test-and-treat strategy among MSM, highlighting the importance of a multipronged approach and allowing an assessment of which individual interventions might be most important to the success of such a strategy. It offers guidance based on reasonable implementation targets. It indicates how implementation of our multi-component test-and-treat strategy can attain the National HIV/AIDS Strategy goals. Ideally, findings from this model will spur more research on how best to improve implementation of the individual interventions; particularly interventions that increase the annual HIV testing rate and that improve adherence and retention in care. Our model suggests that a test-and-treat strategy could have a substantial impact on the urban MSM HIV epidemic, but would not replace the need for consistent condom by MSM.