Although the overall number of new HIV infections has declined worldwide, it is estimated that 5000 new infections occur each day.1 With continuing progress in the implementation of treatment and prevention programs, it is becoming increasingly important to define metrics by which HIV incidence may be considered controlled to gauge success and guide resource allocation.
In a paper published in the Lancet HIV2 in October 2018, researchers from the Yale School of Public Health and Oregon State University examined the 4 potential types of control criteria and their shortcomings, which were discussed in 2017 at a United Nations Programme on AIDS (UNAIDS) meeting. This meeting used a transmission model to simulate trajectories of prevalence, incidence, and mortality through 2030 in 38 sub-Saharan African countries.
- A percentage reduction in incidence over time, although useful for assessing progress, is not an adequate measure of control. However, even in the case of achieving the UNAIDS goal of 90% incidence reduction, the incidence of HIV would remain unacceptably high in some countries compared with others; for example, 21 per 10,000 in Lesotho vs 0.14 per 10,000 in Niger.
- The incidence-to-mortality ratio, in which a higher rate of mortality vs incidence is considered indicative of control, is problematic for several reasons. However, given that more than 70% treatment coverage is required for incidence-to-mortality ratio-based definitions of control, this would be inapplicable to 35 of the 38 countries assessed.
- The value of using an incidence-to-prevalence ratio is also limited, partly because of variations in life expectancy. “For a lifelong infection such as HIV, prevalence rises as life expectancy increases,” as stated in the paper.2 “As the life expectancy for an uninfected person can vary substantially across nations, an [incidence-to-prevalence ratio] indicative of low incidence in one context may translate to an unacceptably high incidence in another setting.”
- The authors supported the use of incidence-based criteria; specifically, fewer than 1 in 10,000 new cases per year, to define HIV control. This metric is unaffected by fluctuating HIV mortality rates, and it reduces empirical uncertainty while more clearly conveying the magnitude of the problem compared with the other metrics.
“Efficiently achieving a low national incidence will undoubtedly involve targeted interventions within high-risk groups who might also be less likely to access services,” the researchers wrote.2“Outreach to high-risk groups will be instrumental in effective treatment-as-prevention approaches and will ensure that control is steadily achieved at subpopulation levels and for the country overall.”
Infectious Disease Advisor spoke with the following experts to discuss their thoughts about HIV control criteria: Willi McFarland, MD, PhD, MPH, professor of epidemiology and biostatistics at the University of California, San Francisco, School of Medicine, principal investigator of the Trans*National Cohort Study, and director of the Center for Public Health Research at the San Francisco Department of Public Health in California; Henry F. Raymond, MPH, DrPH,associate professor of biostatistics and epidemiology at both Rutgers School of Public Health, Piscataway, New Jersey, and the University of California, San Francisco’s Global Health Group; and Ken Ho, MD, MPH, assistant professor of medicine in the University of Pittsburgh’s Division of Infectious Diseases, and medical director of the National Institutes of Health-funded Pitt Men’s Study (part of the Multicenter AIDS Cohort Study).
Infectious Disease Advisor: Galvani et al concluded that incidence-based criteria, and specifically incidence less than 1 in 10,000 people per year, would be best for defining control of HIV epidemics. What are your thoughts on this, and what are potential issues related to this approach?
Dr McFarland: First, I am not a modeler or a fan of modeling, having seen how appalling the predictions of HIV models have been during the last 30 years. Therefore, I don’t feel I have the expertise to evaluate the specifics of the paper’s models, but I am not impressed by such models.
Second, I think the notion of “epidemic control” is more aspirational, political, and geared to advocacy than concretely or epidemiologically defined by less than 1 in 10,000 or a ratio of new infections to deaths (2 examples from the paper). The UNAIDS is pushing for the goals of zero new infections, zero AIDS-related deaths, and zero HIV stigma, which of course is aspirational, political, and advocacy-oriented. San Francisco and other “Fast Track” cities have taken these targets up.
These definitions of control/success would not accept 1 in 10,000 or a ratio of new infections to deaths as measures of success. Officially, we and others are embracing zero as the sign of epidemic control. I particularly don’t like the control definition of more deaths than new cases. Although this would make prevalence decrease, you can cynically achieve that by doing a poor job of treating people (ie, more death).
Third, the 1 in 10,000 metric would not be acceptable in San Francisco as defining HIV control, [where the approximate rate of new infections] are at approximately 2 per 10,000 per year (and 100 deaths). Getting to 1 in 10,000 would not be good enough to declare success, as this is only a halfway point and would equal about 80 new infections per year in San Francisco. That number would likely maintain HIV in the city. In addition, we know that those 80 individuals would be concentrated in the most vulnerable populations, and this would mean a relatively greater disparity of HIV for some populations. This would also be unacceptable.
As for a ratio with deaths, with good care, it would never be plausible for there to be more deaths than new cases as we get lower incidence. With good care, people with HIV live a full life span and are increasingly dying not of AIDS, but other causes.
Dr Raymond: I think these types of indicators are helpful, and the study authors made good arguments for their recommended indicator; particularly, their explanation of how demographic changes and mortality are important considerations in choosing an indicator of control of HIV incidence. They astutely noted that a country, for instance, could achieve control but not uniformly across all areas of a country. I would extend this to include that a geography could reach HIV incidence control in some populations (such as heterosexuals), but not in others (such as transgender women) who already bear a disparity in burden of HIV, stigma, and discrimination.
However, these types of indicators are also overly simplistic (and perhaps that’s the point), and don’t adequately address all the important aspects of epidemic response and control that should be important to clinicians. For example:
- clinical care indicators
- ensuring that all infected individuals are receiving care and treatment
- reducing AIDS-related opportunistic infections
- reducing mortality resulting from AIDS among infected persons
- societal indicators
- eliminating HIV stigma
Dr Ho: I think the study authors presented a compelling argument that identifies flaws in using certain statistics as evidence of progress. In this case, they described situations in which the statistics used to gauge the epidemic may suggest improvement and that the work is done, when in fact, the epidemic is worsening. The study authors proposed using incidence to define control, and in light of what they have presented, I think this is sensible. The number does seem fairly aspirational, although it is based on the standards of the World Health Organization. One issue with setting such ambitious goals might be consideration of the resources needed to achieve very low incidence rates in sub-Saharan Africa, where HIV incidence and prevalence are quite high at baseline.
Infectious Disease Advisor: What should the focus of further research pertaining to HIV control criteria be?
Dr Raymond: Perhaps, a more comprehensive set of indicators that cover incidence, care, and society should be developed.
Dr Ho: I believe a focus on cost/benefit analyses of this approach is warranted. As we make strides in fighting the epidemic, it may take more resources to accomplish incrementally smaller gains. I think further research is needed to inform logical cost/benefit standpoints, in terms of how resources get distributed as we get closer and closer to our goals.
Infectious Disease Advisor: Are there any additional points you would like to note about the topic?
Dr Ho: I think that the importance of U=U (Undetectable Equals Untransmittable) remains ever-present when we talk about HIV globally. Increasing access to antiretrovirals is simultaneously helping to improve the lives of people living with HIV and to prevent the spread of the epidemic. Moving forward, I think that maintaining support for such programs is crucial.
By Tori Rodriguez
1. HIV.gov. Global statistics. https://www.hiv.gov/hiv-basics/overview/data-and-trends/global-statistics Updated November 20, 2018. Accessed November 26, 2018.
2. Galvani AP, Pandey A, Fitzpatrick MC, Medlock J, Gray GE. Defining control of HIV epidemics [published online October 9, 2018]. Lancet HIV. doi: 10.1016/S2352-3018(18)30178-4