Syndicate content

What caused the HIV epidemic in Africa?

Shanta Devarajan's picture

Damien’s earlier post called into question one commonly-held view of the cause of the spread of HIV in Africa, namely male promiscuity. 

A paper by Pauline Leclerc and others (hat tip to Mark Gersovitz) seems to show that there is even greater uncertainty.  Leclerc and co-authors tried to simulate the dynamics of the epidemic in Zambia but found that the parameters needed to fit epidemiological models were beyond what the data would allow. 

In short, thirty years later, it appears as if we still don’t know what caused the disease to spread the way it did on the continent.  Perhaps there is no single set of causes, and that the evolution of the disease is different in different parts of Africa.  Perhaps we should move beyond epidemiological models and look to other disciplines for the answers. 

At any rate, to fight the epidemic effectively, we need to know how and why it became an epidemic.


Submitted by Lucy Hancock on
David Gisselquist thinks it is partly transmission in health care settings. This would account for anomalies such as why the rich are likelier to become infected, why rates rise when war is over, why is it that infection seems proportional to the number of health care "jabs." He and I were consultants next door to each other in the H building several years ago and I screamed at him until he wrote his ideas down in a book, now out, and online. I got him a domain name, - it's a terribly ugly web site (my fault) but it points to his book, You can buy his book at Amazon or read it online. - Lucy

Submitted by David Gisselquist on
Colleagues, AIDS was recognized in Africa by 1982. That recognition led to persistent indulgence of racial stereotpyes among many scientists and activists, blaming Africa's terrible AIDS epidemics almost exclusively on (fantasized) differential heterosexual behavior in Africa vs elsewhere. Much of this indulgence has been funded by the UN system. This indulgence continues despite repeated survey findings that heterosexual risk behavior in Africa is comparable to what is found in the EU/US. On the other hand, the risk you might receive health care with a reused and unsterilized skin-piercing instrument is greater in Africa than in the US/EU. WHO/UNAIDS warns UN employees that health care is a risk for HIV in Africa. The persistence of this risk in Africa (but not in the US/EU) almost 30 after AIDS was first recognized shows that health program managers in WB, WHO, and African ministries of health have not done what is required to find and stop HIV transmission in health care. Instead of doing their job to protect patients, health experts blame the victims. AHA -- you have too much sex! In addition to distractions and confusions stemming from racial stereotypes and emotive sexual issues, health professionals have a conflict of interest in addressing HIV transmission during health care. Left to their own devices, health professionals might well continue to deny and to ignore their contribution to Africa's AIDS disasters for another 30 years. Thus, this may be an area where the WB's Executive Board should step in to order some change -- to tell WB's health program managers (a) to do what is required to stop HIV transmission through health care in Africa, and (b) to stop indulging and publishing unsupported racial stereotypes that stigmatize Africa and African men and women living with HIV. Best regards, David

David, I think a healthy and dispassionate debate about the role of HIV transmission through health care vs. sexual transmission in Africa is useful, and I agree that the risk of HIV infection through unsafe injections and healthcare should not be ignored and should be reduced. However, I believe the role of health care transmission remains limited. My sense is that the studies, to which you have contributed greatly, that focus on the role of health care in HIV transmission, are often based on detecting anomalies in the data, for example girls who declared to be virgins but are tested positive for HIV. You would then suspect that this is a case of non sexual HIV transmission, for example in a health care setting. This is a possibility, but of course virginity is self-reported and there might be a bias, so another possibility is that the girl was not a virgin. One of the most talked about paper on the role of HIV transmission in health care setting is a paper by Deuchert and Brody in 2006 in the International Journal of STD & AIDS. They found a positive association between tetanus toxoid injections during pregancy and HIV infection in the Kenya Demographic and Health Survey (DHS). The paper by Deuchert and Brody can be found here: I was intrigued by this result and decided to look for this association in a larger set of DHS. My paper, published in 2008 in Sexually Transmitted Infections takes the Kenya DHS and six other African DHS (all the African DHS with HIV testing available at the time) and test the robustness of the correlation found in Kenya. Once I included standard controls for age group, ethnic, urban and regional indicators, none of the correlations, even the one in Kenya, was statistically significant. My paper can be found here: Best regards, Damien

Dear Dr. de Walque, You have mischaracterized the evidence for blood-borne HIV transmission in sub-Saharan Africa. David Gisselquist's book, cited in Lucy Hancock's post, covers much of the evidence through 2005. Since his book was written, the evidence continues to accumulate and suggests that blood-borne HIV transmission in healthcare, cosmetic care, and ritual settings is widespread in sub-Saharan Africa. For example, see: St. Lawrence JS, Klaskala W, Kankasa C, et al. (2006) Factors associated with HIV prevalence in a pre-partum cohort of Zambian women. Int J STD AIDS 17: 607-613. Brewer DD, Potterat JJ, Roberts JM, et al. (2007) Male and female circumcision associated with prevalent HIV infection in adolescents in Kenya, Lesotho, and Tanzania. Ann Epidemiol 17: 217-226. Peters EJ, Brewer DD, Udonwa NE, et al. (2009) Diverse blood exposures associated with incident HIV infection in Calabar, Nigeria. Int J STD AIDS, 20, 846-851. Okinyi M, Brewer DD, Potterat JJ (2009) Horizontally acquired HIV infection in Kenyan and Swazi children. Int J STD AIDS 20: 852-857. Reid S. Non-vertical HIV transmission to children in sub-Saharan Africa. Int J STD AIDS. 2009;20:820-7. Vaz P, Pedro A, Le Bozec S, et al. (2010) Nonvertical, nonsexual transmission of human immunodeficiency virus in children. Ped Infect Dis J 29: 271-274. Instituto Nacional de Saúde (INS), Instituto Nacional de Estatística (INE), e ICF Macro. 2010. Inquérito Nacional de Prevalência, Riscos Comportamentais e Informação sobre o HIV e SIDA em Moçambique 2009. Calverton, Maryland, EUA: INS, INE e ICF Macro. (see pp. 177-181) Eva Deuchert replicated and extended the analyses of tetanus toxoid vaccination and HIV in women to Burkina Faso and Cameroon (Deuchert E (2007) Maternal health care and the spread of AIDS in Burkina Faso and Cameroon. World Health Pop 9: 55-72. Both her analyses and Deuchert and Brody's (2006) analyses were appropriately based on the unweighted DHS data. Your analyses, as reported in your 2006 article in Sexually Transmitted Infections, were based on the weighted data – a procedure that rarely, if ever, is done for epidemiologic analyses of correlates of infection. In your article, you state “The data have been weighted as recommended by the data provider” (p. 123). In fact, ICF Macro–the organization that oversees the DHS–advises exactly the opposite: “Use of sample weights is inappropriate for estimating relationships, such as regression and correlation coefficients” ( This point was also made in a prior exchange of letters about the Deuchert and Brody (2006) article (see Brody S. The unhealthy attempts by CDC and WHO to deny the importance of HIV transmission through unsafe health care. Int J STD AIDS. 2009;20:70-2. and In addition, I have conducted analyses of the association between tetanus toxoid vaccination and phlebotomy during antenatal care and prevalent HIV infection in women with DHS data from 11 countries in sub-Saharan Africa (the 7 you analyzed plus 4 others). These analyses were similar to those performed by Deuchert and Brody (including several demographic and sexual behavior control variables). The results show consistent independent associations between these blood exposures and prevalent HIV infection in women who had given birth in the prior 5 years and had not tested for HIV previously. That is, my results replicate and extend those of Deuchert and Brody. I presented these results at the 2007 meeting of the International Society for Sexually Transmitted Diseases Research in Seattle. I am in the process of updating these analyses to include all of the currently available DHS data. As my colleagues and I noted in our Peters et al. (2009) article, “To identify HIV transmission modes with confidence, three study design elements are required: (1) comprehensive assessment of blood and sexual exposures in uninfected persons and persons with incident infection; (2) tracing, testing and assessment of infected and uninfected persons’ contacts to those exposures; and (3) sequencing of infected persons’ HIV DNA to detect genetically related infections.” We have called for this kind of investigation many times in print since 2003. We have repeatedly sought funding to carry out such investigations but have been rejected by reviewers extremely hostile to the idea. Apparently, no one else has attempted such investigations. Consequently, Dr. Devarajan's summary of HIV epidemiology couldn't be more accurate:“In short, thirty years later, it appears as if we still don’t know what caused the disease to spread the way it did on the continent.” It would be wonderful if the World Bank and other funders prioritized such investigations. Devon Brewer,

Dear Devon, Thank you for your response to my earlier comment on HIV transmission in health care settings. You pointed to the fact that my analysis of the association between tetanus toxoid injections during pregancy and HIV infection published in Sexually Transmitted Infections ( weighted data. This is true and the use of sample weights in regressions is debated, but generally considered as a minor issue. I took your suggestion seriously and re-run my analysis without weighting the data for the regressions. The results are similar to the ones in the published paper. Importantly, as in the published paper, none of the adjusted odds ratios are statistically different than 1. I therefore stand by my conclusion that once we include standard controls for age group, ethnic, urban and regional indicators, none of the associations between tetanus toxoid injections during pregancy and HIV infection, even the one in Kenya, was statistically significant. I am looking forward to reading your analysis extended to 11 Demographic and Health Surveys. Best regards, Damien

Submitted by Eva Deuchert on
Dear Damien I found your blog by coincidence. From my understanding of our papers, the main is the number of confounding factors. Your analysis controls for 5-year age groups and for ethnic, urban and regional indicators. Stuart and I mainly control for sexual behavior (and some socio-economic factors in the paper using data from Burkina Faso and Cameroon). Could it be case that the number of predictors is a little bit too high given the number of positive cases in your sample? Could overfitting be a problem? At the end of the day, any kind of “selection on observable” strategy would not lead to a causal effect anyway. All we provide are partial correlations that may or may not indicate the existence of a problem. I think we both agree that a final answer on the relative importance of non-sexual HIV transmission in Sub Saharan Africa cannot not be provided with our current state of knowledge. More precise (ideally longitudinal) data focusing on all possible transmission mechanism would be needed. Best regards, Eva PS: There is quite interesting policy brief which may be interesting:

Submitted by David Gisselquist on
Colleagues, In two comments to this blog, Damien maintains that evidence is lacking to show HIV transmission through health care in Africa. His main argument is this: The association between HIV in women and tetanus vaccinations during antenatal care is not significant – ie, there is at least a 5% chance that injections did not infect women. He also speculates that HIV-positive women who claim to be virgins might have lied. Those arguments do not shake the abundant evidence, which Lucy and Devon introduce and reference, that blood-borne transmission makes a substantial contribution to Africa’s epidemics. For example, national surveys in Swaziland and Mozambique found 20%-30% of HIV-positive children to have HIV-negative mothers (among children with tested mothers; see references in Devon’s comment). And a recent national survey in Congo (Brazzaville) reports higher HIV prevalence in virgin vs. non-virgin women aged 15-49 years (4.2% vs. 4.1%; see Table 7.5 in: In initiating this dialogue, Shanta suggested: “Perhaps we should…look to other disciplines” to explain Africa’s HIV epidemics. But solutions are not likely to come from other disciplines. If your mechanic has not fixed your car’s brakes, it won’t help to bring in a baker or painter. What you need is a mechanic who will do the job. To stop AIDS in Africa, we need to know how HIV goes from one person to another so much more often in Africa than elsewhere. Tracking HIV transmission is a job for infection control experts (to find and stop transmission in health care) and epidemiologists. Not investigating unexplained infections in Swazi and Mozambican children or in Congolese virgins exposes a double-standard on the part of AIDS experts. Such infections would be investigated in the EU or US. This double-standard surfaces in other ways. For example, studies in Africa have tested and followed HIV-positive women and their children, without telling the women they are infected, but watching them infect their children (see chapter 7 in: Such studies could not be done outside Africa. In a previous comment, I urge the Bank’s Executive Board to tell the Bank’s public health staff to do what they know how to do to protect African patients, and to solve and stop Africa’s epidemics. I don’t expect that will happen. My scenario for ending Africa’s HIV epidemics is this: Somewhere in Africa, sometime over the next several years, people will get fed up with unexplained infections. Public pressure will force Ministries of Health to trace the source of HIV infections in people with no and modest sexual risks, and to find and stop transmission in health care. In other words, I expect the people who are standing in front of the run-away car will get out of the way. And they will also take care of the brakes. David

It is an eye-opener to Africa that they must watch after for their sexual health. There is no assurance of their health, especially when they don't have enough facilities to relay the information regarding such disease.

Add new comment