From the Department of Infectious Disease Epidemiology, Imperial College London, St. Mary's Hospital, Norfolk Place, London, United Kingdom
Correspondence: Geoff P. Garnett, Imperial College London, St. Mary’s Hospital, Norfolk Place, London W2 1PG, UK. E-mail: firstname.lastname@example.org.
Received for publication June 7, 2007, and accepted July 12, 2007.
Given the salience of HIV as a health problem and the associated resources invested in research, it is remarkable how few good quality epidemiologic studies have investigated the risks of acquiring infection. Perhaps, we assume we know what the risks are already and now need to reduce them. However, patterns of risk might well change over time and across different populations. As illustrated in the study by Mattson and colleagues,1 new insights and risks can emerge. This study, from Kisumu in Kenya, used the opportunity provided by recruitment into a randomized controlled trial of circumcision to explore risks of HIV infection. The study used a case control design, with the risks reported by those with prevalent HIV infections compared with the risks of those uninfected. The participants, the 65% of uncircumcised men aged 18 to 24 years who agreed to enroll, were not representative of the wider population, but differences between the cases and controls reveal the determinants of HIV acquisition in this subsection of the population. What makes the study interesting is the great care taken in collecting detailed information on a wide range of potential risks. By going beyond the usual risk factors, the study managed to explain most of the individual variation in risk of acquiring HIV in this population of men. Nonetheless, questions remain concerning the interrelationship between the different variables and how they operate to influence HIV acquisition. To understand the role of different risk factors a theoretical framework, such as the proximate determinants2 or social epidemiology3 framework, is required. In designing appropriate responses, it is important that the relationship between social and cultural determinants, such as membership in a church, can be understood in terms of behaviors that directly determine risk.
As HIV spreads from its initial focus, we expect that the risk behaviors of the average infected individual will decline.4 With the epidemic becoming more generalized, the difference in behaviors between those infected and uninfected will likely become less, so much so that even if we measure the correct variables we might not be able to explain much of the risk.5 Since, over time, risk factors will change, those associated with prevalent infections like lifetime numbers of partners will be less useful than those associated with incident infections. Although studies of prevalent infections are informative, studies of recent infections would provide better information with which to target prevention interventions. This is a challenge because cohort studies to directly measure incidence are logistically difficult and expensive, and the sample sizes required to detect recent infections in cross-sectional studies would be large even if reliable tests became available. An alternative approach might be provided by repeated cross-sectional surveys such as is planned with demographic and health surveys. It is important that such surveys pay attention to generating valid reports of risk behaviors.
The interrelationship between risk factors, such as those between lifetime sexual partner numbers and concurrent partnerships and between being a Catholic and concurrent partnerships, means that some can fall out of multivariate analyses, as seems to be the case here for concurrent partnerships. Here a theoretical framework can help interpret the results of an analysis. When considering transmission dynamics, it is likely that concurrent partnerships would be more important to ongoing spread than lifetime partnerships,6 although in a proximate determinants framework, being Catholic would seem to be a distal risk acting through its association with the likelihood of concurrent partners. Understanding the attitudes and perceptions that lead to this association would be helpful. The interpretation of statistical analyses, such as multivariate logistic regression, needs to go beyond whether associations exist and how strong they are. This is illustrated by what will perhaps be the most controversial aspect of Mattson and colleagues’ results.
The study found having received an injection in the last 6 months to be significant in predicting HIV infection. Debate has ranged around the role of iatrogenic transmission.7 This is now being informed by epidemiologic studies, which have a diverse set of findings. In some cases, many infections can be associated with injections,8 in others very few.9 In some cases, the distribution of infections in specific groups like children suggests the role of injections is small;10 in others, there is great concern over the high prevalence of infection found across all ages of children.11 How should we interpret this diversity? First it should be noted that different locations might have widely divergent patterns. The hygiene practices of a particular local hospital might have a profound influence. Beyond this, there has been debate over the fraction of infections in generalized heterosexual HIV epidemics that can be attributed to sexual transmission or iatrogenic transmission.12 Two fundamental questions have emerged: whether we got the “cause” of the major HIV epidemics wrong and whether we need to invest in preventing iatrogenic transmission. I believe the answer to the former is no but that to the latter is yes. To understand the role of medical injections, we need to consider how we attribute infections to risk factors.
Standard epidemiologic measures to calculate the proportion of infections attributable to a specific cause fail us in infectious disease epidemiology, where the risks of infection depend on how well the infection has penetrated the population. Some risk factors that currently expose individuals to HIV would have no cases attributed to them were the infection not widespread, whereas other risk factors combine to create the conditions that allow the infection to initially spread, which means all the infections are “attributable” to them. These latter are the risks that allow each new infection to cause more than one other new infection when the infection first enters a population (i.e., the basic reproductive number). HIV globally is not found predominantly where unclean injections predominate, but it has spread where sex work is common or where many in the population have multiple overlapping sex partnerships. This shows that contacts through contaminated medical injections have not regularly generated basic reproductive numbers of greater than one, whereas sexual contact patterns have. In such circumstances, we would expect risk factors associated with sexual behavior to be more important than those associated with injection. Nonetheless, in some locations, medical injections may play an important role spreading infection from those infected sexually to others. It is doubtful given that iatrogenic transmission did not initial generate epidemics whether infections acquired through this route would be self-sustaining. Such a pattern would explain the diverse results of observational studies and would at the same time indicate that safe injection practices are an important part of HIV prevention where the virus is common but would not be sufficient on its own to control the spread of infection.
1. Mattson CL, Bailey RC, Agot K, Ndinya-Achola JO, Moses S. A nested case-control study of sexual practices and risk factors for prevalent HIV-1 infection among young men in Kisumu, Kenya. Sex Transm Dis 2007; 34:731–736.
2. Boerma JT, Weir SS. Integrating demographic and epidemiological approaches to research on HIV/AIDS: the proximate-determinants framework. J Infect Dis 2005; 191(suppl 1):S61–S67.
3. Poundstone KE, Strathdee SA, Celentano DD. The social epidemiology of human immunodeficiency virus/acquired immunodeficiency virus. Epidemiol Rev 2004; 26:22–35.
4. Wasserheit JN, Aral SO. The dynamic topology of sexually transmitted diseases: Implications for prevention strategies. J Infect Dis 1996; 174(suppl 2):S201–S213.
5. Lewis JJC, Donnelly CA, Mare P, Mupembireyi Z, Garnett GP, Gregson S. Evaluating the proximate determinants framework for HIV infection in rural Zimbabwe. STI 2007 [epub ahead of print].
6. Morris M, Kretzschmar M. Concurrent partnerships and the spread of HIV. AIDS 1997; 11:641–648.
7. Schmid GP, Buve A, Mugyenyi P, Garnett GP, Hayes R, Williams BG, Calleja G, De Cock KM, Whitworth J, Kapiga SH, Ghys P, Hankins C, Zaba B, Heimer R, Boerma JT. Eliminating unsafe injections is important, but will have little impact on HIV transmission in sub-Saharan Africa. Lancet 2004; 363:482–488.
8. Lopman BA, Garnett GP, Mason P, Gregson SA Individual Level Injection History: A Lack of Association with HIV Incidence in Rural Zimbabwe PLoS Medicine 2005; 2:142–146. e37.
9. St. Lawrence JS, Klaskala W, Kankasa C, West JT, Mitchell CD, Wood C. Factors associated with HIV prevalence in a pre-partum cohort of Zambian women. Int J STD AIDS 2006; 17:607– 613.
10. White RG, Cooper BS, Kedhar A, Orroth KK, Biraro S, Baggaley RF, Whitworth J, Korenromp EL, Ghani A, Boily M-C, and Hayes RJ. Quantifying HIV-1 transmission due to contaminated injections. PNAS [published online May 30, 2007].
11. Rehle T, Shisana O, Pillay V, Zuma K, Puren A, Parker W. National HIV incidence measures – new insights into the South Africa epidemic. S Afr Med J 2007; 97:194–199.
12. Gisslequist D, Potterat JJ. Heterosexual transmission of HIV in Africa: an empiric estimate. Int J STD AIDS 2003; 14:162–173