How to find what’s driving Africa’s HIV epidemics
HIV is not an easy virus to catch – it needs memorable skin-piercing events or intimate sexual contact. So it shouldn’t be too hard to find out why so many Africans are getting HIV. How? Find people with new infections, ask them about recent sex partners and skin-piercing events (injections, tattoos, accidents), and then go look for the source.
- Trace and test sex partners to see if any are HIV-positive.
- See if people with new infections got injections or other skin-piercing procedures from clinics that did sterile instruments; test others who visited the same clinic to see if they, too, were infected.
Then check to see if HIV someone with a new infection is a close match to HIV from a sex partner (if they are infected) or from one or more people who attended the same clinic (if they were infected). If you find a match, then you have very likely found the source.
Researchers didn’t trace the source of new infections
Over the past 30 years, international and foreign organizations have funded a lot of research on HIV in Africa. This research includes 44 randomized controlled trials during 18987-2011 that tested interventions to prevent HIV in adults (see appendix table in this link). In these trials, researchers recruited adults, then divided them into two or more groups: one group got nothing (no intervention) while another group or groups got something that — it was hoped — would protect them (for example, antibiotics to syphilis, circumcision). Studies then followed all adults to see who got new HIV infections.
How did these studies fail to explain what is different about HIV in Africa? As crazy at it seems – they didn’t ask, didn’t look, and didn’t report what they found. Collectively, these studies identified >4,000 new infections: >900 in men and > 3,100 in women. But studies didn’t look for the sources of these infections. (The details of this failure are described in a review available here or from the Social Science Research Network.)
Not looking for a sexual source
- Only 4 of 44 studies traced HIV-positive sexual partners — spouses only long-term partners only — and checked to see if HIV from both partners were similar, which was good evidence that one infected the other by sex. These 4 studies traced a total of only 186 infections (<5% infections in 44 studies) to long-term sexual partners (see table 5 in this link).
- No study traced and tested any spouse or other long-term partner that was not already enrolled in the study.
- No study traced and tested any non-spousal partner.
Not looking for a blood-borne source
- No study identified any facility that provided skin-piercing procedure to anyone with a new HIV infection.
- No study investigated sterilization practices in any facility that provided injections or other skin-piercing procedures, or tested any other patient or client attending such facilities.
Even so, other evidence from these studies suggests blood-borne and sex risks are equally responsible for Africa’s epidemics
Aside from tracing infections to their source, a less decisive method to identify risks responsible for new infections is to see people with new infections vs. people without new infections were more likely to report risks. Here’s what the 44 studies report:
- Only 5 of 44 studies reported recent sexual risks (having any vs. no sex partner, or less than 100% condom use) for people with or without new infections. According to the median (middle) results from these 5 studies, having any unprotected sex was responsible for less than 30% of new infections (see table 6 in this link).
- Only 5 of 44 studies reported skin-piercing procedures in people with new HIV infections (5 reported injections, two reported transfusions, two reported hospital or clinic visits without specifics about skin-piercing procedures). Looking for the median (middle) result from these 5 studies, reported skin-piercing procedures and hospital or clinic visits appeared to be responsible for roughly a quarter of all new HIV infections (see table 7 in this link).
As economist I strongly agree that with scarce resources, any decision on the allocation of funds should be based on economic principles. I therefore strongly welcome the rethinkHIV initiative to rank competing HIV/AIDS interventions based upon benefit-cost metrics aimed at helping policy-makers and donors prioritize investments.
One of the key problems with their approach is, however, that we have only very little empirical evidence on the benefits of these interventions. Only a small number of randomized field experiments were conducted. The ones on preventing sexual transmission often use self-reported sexual behavioral change as their key outcome variable but fail to show that these interventions result in a measureable decline in the number of new infections – which should be the prime outcome of the trials from my point of view. Trials on preventing non-sexual transmission exclusively focus on preventing mother-to-child transmission. Any empirical evidence on the effectiveness on preventing iatrogenic HIV transmissions is entirely missing.
So, on which basis are estimates for benefits made? The authors of the commissioned research papers use mathematical models similar to the ones provided by the UNAIDS “Know Your Epidemic & Modes of Transmission” initiative (http://www.unaidsrstesa.org/thematic-areas/hiv-prevention/know-your-epidemic-modes-transmission). Any estimates derived from these models, however, can be only as good as the underlying model.
One possibility to check reliability of the benefit estimates is to investigate if these type of models can predict the epidemic situation of many African countries in the long-run. Assuming that HIV is predominantly heterosexually transmitted so that other transmission mode can be ignored (at least for adults), it has been shown that simple models based on differential equations and sequential sexual partnerships (similar to the UNAIDS models) entirely fail to predict an HIV epidemic observed in Africa once they are parameterized with realistic model parameters. In fact these models would often not even predict an outbreak:
Click to access 16-5.pdf
Other models model concurrent partnerships. With these models, at least an outbreak can be predicted. But again, parameterizing them with realistic model parameters dramatically reduces the role of sexual transmission suggesting that concurrency cannot be an important driver of HIV epidemics in sub-Saharan Africa:
I strongly agree with UNAIDS: Knowing the epidemic and the importance of different transmission modes is an important pre-requisite for identifying gaps in current prevention response and resource allocation. However, I do not believe that using unrealistic mathematical models, ignoring the existing empirical evidence on the importance of non-sexual transmission modes, and downgrading statistical anomalies (such as the high share of HIV-positive children in Africa have HIV-negative mothers or the high prevalence among self-reported virgins) as data-errors will provide any new insights, neither on the importance of different transmission modes, nor on the cost-benefit ratio of different prevention methods.