One of the most highly infectious time periods in the life of a person with HIV is also when that person is most likely unaware of their HIV status. This is the acute stage.
Shortly after infection, the virus explodes with a measurable peak in the number of copies floating around. This feature of the sneaky virus keeps it entrenched in communities, spreading like wildfire among people with concurrent partnerships and frequent sex acts. By the time a test catches it at a later stage of disease, HIV has already been likely and unknowingly been transmitted to partners.
Here are some things that might help from a public health perspective:
- Change national guidelines to test everyone more often (politics)
- Prescribe PrEP to people at high risk (support adherence)
- Start diagnosed cases on treatment sooner (adopt in regions now waiting until white blood cells dip below a level)
- Find an HIV vaccine and protect everyone (invest)
All this stuff costs a lot. Should we do a lot of one thing or a little of all? Advocates for efficiency may want to understand how to get the greatest bang for the buck – promoting stewardship of limited healthcare resources. Spending money to end HIV may mean not paying for services or medicines to help other diseases. Economists call this opportunity cost, and you can’t escape it.
Mathematical modeling can help. It’s useful in these situations where we can’t directly measure the outcome of interest, such as changes in incidence. This is common in infectious diseases, where prevention and treatment for an individual likely creates spillover benefit to others in the community. If your aim is to maximize health gains for a population, models are a practical tool to estimate the magnitude of impact.
I’m going to tell you about one approach to break the curse of acute HIV infection in a hot epidemic and show you what we learned from a math model of the expected impact.
The Sabes Study in Peru
The Sabes Study was designed to evaluate a treatment-as-prevention intervention among cisgender men who have sex with men and transgender women in Lima, Peru. This is where we see a hot epidemic and a group of people disproportionately affected by the HIV.
Peer educators visited saunas, adult movie theaters, sexwork areas, discos, bars, beauty parlors, sporting events, and internet cafes to find potential participants. The research sites in Lima were based at the Health and Education Civil Association (IMPACTA), Asociación Vía Libre, and a community-based organization called Epicentro.
From July 2013 to September 2015, a total of 3,337 subjects were screened for HIV; 2,685 (80.5%) were negative, and 2,109 began monthly testing. All participants were followed for 48 weeks and were then referred to the Peruvian Ministry of Health to continue receiving free HIV care and treatment.
One interesting side finding from Sabes showed that dispositional coping strategies of men who have sex with men and transgender women was related to the probability of linkage to care after HIV diagnosis.
Our Question: what would be the impact on the HIV epidemic in Peru if targeted acute HIV testing and linkage to care, as offered in the Sabes study, was extended to all high risk men who have sex with men and transgender women?
What we learned
Our new research published in the journal Infectious Disease Modelling used findings from the Sabes study in a math model that showed that reaching 50% of the high-risk men who have sex with men and transgender women in the acute phase would reduce HIV incidence in 2038 by 60% and prevent 36% of new infections in Peru between 2018 and 2038. Notice in the figure below how the benefits grow over time.
A) Effectiveness in terms of cumulative fraction of infections prevented; B) Effectiveness in terms of reduction in HIV incidence; C) Effectiveness in terms of reduction in HIV prevalence; D) Estimated contribution of the acute infections to the HIV epidemic (Dimitrov 2019).
How we did it
When it was time for us to do the math, we needed a model that captured both the course of HIV disease progression in an individual and the complex dynamics of HIV transmission and spread within a community.
If you were fortunate enough to take a high school calculus class, you may faintly recall equation derivatives being used for rates or slope and how you solved integrals for the area of a shape. As a teenager, I thought calculus was like Latin: an ancient language that no one speaks anymore but we are still forced to learn because it is at the root of so many things.
Now I’m a grown-up and I use calculus every day. It baffles me to reflect on how the computational power that I use to solve equations today did not exist when I was in high school. It would have been difficult for me then to imagine my future self anywhere near the role of scientist that I am today (I would have accurately predicted a life dedicated to ending HIV).
Calculus is a useful way to describe things. In this case, we needed to define a series of important health states that people live in and the movement of people through these states. We create a series of differential equations to describe the transitions between states over time. The magic of dynamic models is that the movement of people from one health state to the next is not only dependent on the population size in the prior state just a moment before, but can also be dependent on the number of people in a totally different state not directly connected by arrows. Different from static, deterministic, Markov models, dynamic models like incorporate concepts like herd immunity. The HIV incidence rate (S -> I) at any moment in time is dependent on how many people in the community are infected and what fraction of those are virally suppressed.
Simulated population is stratified in compartments by HIV infection status and progression towards AIDS during five disease phases (Dimitrov 2019).
To really solve the mathematical model diagramed above, we needed to define a pretty complicated series of differential equations. The susceptible population (S), undiagnosed infected (I), diagnosed infected (D), infected and engaged in care but not on ART (E), infected on ART but unsuppressed viral load (U), and infected on ART with suppressed viral load (T) populations each have sub-groups with risk status i, anal sex positioning preference j, and age group k. Combined together, you can find a solution that tells you the number of people in each health state (boxes in diagram) any any time t.
tl;dr: Each equation basically means: at any point in time you have the people who were here just a moment before, minus the people who died or left to go somewhere else, plus the people who just arrived. The dt in each equation is your the high school calc clue that these equations are describing rates.
To learn more about general dynamic transmission modeling of infectious diseases, there are several good books recommended on my Methods page. If you want to go crazy and try this at home, you can find the complete list of parameter values in our Supplementary Appendix.
Why it matters
The hands-on human study Sabes showed that it’s possible to find high-risk men who have sex with men and transgender women, test them frequently for HIV, identify acute infections, and start them on treatment. This math modeling study matters because it generated the evidence about country-level impact that was needed by the real decision-maker for action. Government officials must to weigh the tradeoffs and assess the population health benefits with the resources required to update clinical guidelines and scale this program locally in Lima or nationally in Peru.
Middle-income countries do not have infinite resources to implement every single opportunity to improve health. Careful stewardship and difficult decisions are required by governments trying to maximize health with available funds.
Acknowledgements
Dobromir ran the show on this one. Daniel did most of the coding (C++). Angie assembled inputs from Sabes. Javier an Jorge defined the epidemic in Peru. Ann found the money. Tanya Elshahawi deserves credit for the header photo of equation writing. Enrique Saldarriaga and Josesph Babigumira are helping with the cost-effectiveness analysis, so stay tuned for those results.
References
- Lama JR, Brezak A, Dobbins JG, Sanchez H, Cabello R, Rios J, Bain C, Ulrich A, De la Grecca R, Sanchez J, Duerr A. Design strategy of the Sabes study: Diagnosis and treatment of early HIV infection among men who have sex with men and transgender women in Lima, Peru, 2013-2017. American Journal of Epidemiology, 187 (8) (2018), pp. 1577-1585
- Dimitrov D, Wood D, Ulrich A, Swan DA, Adamson BJA, Lama JR, Sanchez J, Duerr A. Projected Effectiveness of HIV Detection during Early Infection and Rapid ART Initiation among MSM and Transgender Women in Peru: A Modeling Study. Infectious Disease Modelling, 4: 73-82 (2019). doi: 10.1016/j.idm.2019.04.001.
