Randomized clinical trials have shown that medical male circumcision substantially reduces the risk of contracting HIV,1,2 leading many to espouse this practice as a means to combat HIV on a large scale.3–5 However, relatively little is known about the relationship between traditional male circumcision and HIV risk. Understanding this relationship is important for HIV prevention because many men in Africa continue to be circumcised by traditional providers, despite the scale-up of medical male circumcision.6
Traditional male circumcision (hereafter referred to as circumcision) has long been practiced in many parts of Africa as part of a broader initiation process marking the transition between boyhood and manhood, and early ecological research suggested that it may help reduce HIV prevalence.7 The relationship between traditional circumcision and HIV risk in Africa is, however, not obvious a priori. On the one hand, similar to medical circumcision, it may help reduce HIV infections. On the other, traditional circumcisions do not necessarily remove all the foreskin,8,9 thereby providing continued viral access to resident immune cells10,11 and potentially little or no HIV reduction benefit. And as traditional circumcisions in Africa typically occur after puberty,6 it is possible that this also raises the risk of HIV infection relative to earlier circumcision.12
Recent analysis of survey data from the African Demographic and Health Surveys (DHS) revealed an inconsistent relationship between HIV status and circumcision.13–16 These findings have prompted some analysts to argue that behavioral disinhibition may be at work and, hence, that population-wide circumcision programs may not be effective in reducing incident HIV infections as implied by the randomized clinical trials.14,17 However, an alternative explanation is that the aforementioned variations in age and extent of circumcision may lead to protective benefits for some individuals but not others, thus obscuring the relationship between circumcision and HIV at the population level.
Unlike the DHS, which typically only asks respondents if they have been circumcised, the data set used in this study—a survey of Xhosa-speaking Africans living in Cape Town—goes further by asking the men about the extent and age of circumcision and attitudes toward traditional initiations. Furthermore, it is the first data set with information on both the extent of foreskin removal during traditional circumcision and HIV status. This article uses these data to assess variations in circumcision practices among an ethnic group that has traditionally used circumcision as a rite of passage into manhood and to examine the consequences of this variation for the risk of contracting HIV.
MATERIALS AND METHODS
Data and Measures
The data for this study come from the Cape Area Panel Study (CAPS). The first wave of CAPS (in 2002) surveyed a representative sample of 4752 young adults aged 14–22 years living in Cape Town. For the first wave of the study, a 2-stage sample was used, stratified by the 3 main population groups (African, colored, and white). In the first stage, clusters were selected, categorized by predominant population group, and in the second stage, households were randomly selected from clusters to achieve a representative sample. Respondents were reinterviewed up to 4 more times, most recently in 2009 (wave 5), then with the cohort aged 20–30 years. The African male sample initially comprised 930 men, and 582 were reinterviewed in 2009. The final estimation sample for this article (n = 473) consists of African men interviewed in 2009 with complete data on all dependent and independent variables.
In all waves, study participants were asked detailed questions on a variety of demographic, socioeconomic, and behavioral topics. In wave 5, African respondents were asked to provide specimens for an HIV test and to fill in a self-administered module on circumcision. Ethical approval was granted by the University of Cape Town and Michigan University. Information regarding CAPS, including the initial sampling strategy, consent, ethical approval, and access to data and questionnaires can be found at http://www.caps.uct.ac.za.
In wave 5, African respondents were asked questions probing cultural preferences toward the practice of circumcision (Table 1). Men were also asked whether they had been circumcised and to what extent. We created a binary indicator for whether the individual reported being circumcised, equal to 1 for those who answered “yes” to the question “Are you circumcised? (that is, some or all of your foreskin has been removed)” and zero for those who answered “no.” Respondents were asked: “If you are circumcised: How much of your foreskin has been removed?” Response options were “the entire foreskin,” “only some of the foreskin,” “don't know,” “refused,” and “not applicable.” From these data, we constructed binary variables for (1) fully circumcised, (2) uncircumcised, and (3) circumcised but the extent of foreskin removal was not reported, with partial circumcision treated as the reference group in each case. Finally, we used responses to the question “About how old were you when you were circumcised?” to create a continuous measure of age of circumcision. In addition to an open response option, participants were also given the choice to report “Don't know but very young.” We coded these individuals as being circumcised at the age of 0; the results were robust to age assigned (0–12 years) and to the exclusion of these respondents from the analysis.
HIV tests were conducted using a dry blood spot specimen (preferred) or saliva (using the OraSure Device; OraSure Technologies, Inc, Bethelhem, PA). Ninety-four percent of respondents consented to the test, and 86% provided a dry blood spot specimen. HIV tests were done using HIV antibody enzyme-linked immunosorbent assays. The screening assay was the Vironostika Uniform 11 Plus O (HIV-1p24; HIV-1gp160; HIV-1ANT70; HIV-2 env peptide amino acids 592-603; bioMérieux, Boxtel, the Netherlands). The second and third tests were the SD third generation Bioline (HIV-1 gp41 including Subtype O, p24, HIV-2 gp36; Standard Diagnostics, Inc, Suwon, Korea) and the HIV1/2 Bio-Rad Western blot. The testing algorithm is displayed in Figure 1.
CAPS collected a range of socioeconomic and behavioral information that we used as controls in our models. Measures of educational attainment (0–12 years of schooling, with any tertiary education recorded as 13) and socioeconomic status (monthly household income from wave 1) were included given their importance in predicting HIV status in other studies.18–21 In addition, we used a set of sexual behavior indicators obtained across all waves. First, given that sexually transmitted disease (STDs) have been shown to be associated with HIV serostatus,22 we created a binary indicator for respondents who reported ever having an STD or a history of dysuria, genital discharge, ulcers, or sores. Second, we included a binary measure reflecting the number of lifetime sexual partners (1 if the individual reported 5 or more lifetime partners and zero otherwise), as this has been found to be a statistically significant correlate of HIV status.22 We also constructed a continuous measure of years sexually active to control for increases in HIV risk that accrue from additional years of sexual exposure.23 Third, we included binary measures of whether the individual reported ever being in a concurrent sexual partnership [equal to 1 if the individual reported “ever (having) been in a sexual relationship with someone and had sex with somebody else”] and whether respondents used contraception the first time they had sex as additional controls for types of partnerships and attitudes/preferences toward unprotected sex and/or access to protective contraceptives, respectively.
We began with an analysis of cultural preferences for traditional circumcision among Xhosa men and then focused specifically on the link between circumcision and HIV status. We computed descriptive statistics for the entire sample and then stratified by circumcision practices. We then estimated 4 probit regression models on HIV status. The first assessed the relationship between HIV status and the binary indicator for whether the individual was circumcised. We started with this specification as it is the dominant one in the literature on male circumcision and the usual option for those using DHS data sets. In the second model, we replaced the general circumcision variable with the indicators for the extent of circumcision. The third examined the relationship between age of circumcision and HIV status. Because uncircumcised individuals cannot report an age of circumcision, the sample for this model comprised circumcised individuals only. Finally, our fourth model included both circumcision extent and age.
For each model, we controlled for age and age-squared, education, household income, and the sexual behavior variables. Age is included because HIV risk accumulates with time and age squared because this accumulation may not be linear. We also included the binary indicator for whether the individual had heard about the protective benefits of circumcision to control for possible behavioral disinhibition among circumcised individuals.24,25 The goal of these specifications was to assess the robustness of the results to the inclusion of factors that may jointly be correlated with the circumcision practice variables and HIV serostatus. In addition, the control variables may represent pathways between circumcision and HIV, that is, circumcision may lead to change in sexual behaviors (via change of social status or teachings about sexual conduct) that influence HIV risk.
Finally, for all models, we present marginal effects instead of the probit coefficients as these are more easily interpretable (for a continuous variable, the coefficient reflects the percentage point increase in the probability of observing the dependent variable for a 1 unit change in the independent variable; for binary variables, it reflects a similar change in the dependent variable from moving from 0 to 1 on the independent variable of interest). All standard errors were corrected for heteroskedasticity.
Table 1 presents our findings on attitudes toward traditional male circumcision among Xhosa men. Almost all men (99%) reported that it is important or very important that Xhosa men are circumcised. The data also show a strong cultural preference for traditional circumcision, with almost all men (89%) agreeing that a “man is not a man until he has been to the mountain/bush” (going to “the mountain” or “the bush” is the term used to describe participating in a traditional male circumcision ritual). A large majority of respondents (89%) reported the belief that men should always get their foreskin removed in the mountain/bush by a traditional surgeon only. Most of the other respondents (6%) indicated that circumcision should continue in the mountain/bush but with a male Xhosa nurse assisting the traditional surgeon. Only 3% of men thought it best for circumcision to be done in a clinic or hospital.
Interestingly, only a third of men reported having heard that circumcision reduces the risk of contracting HIV. This suggests that information about this form of HIV prevention is lacking in this group. However, it also indicates that young men are unlikely to be engaging in unsafe sex after circumcision on the assumption that having been circumcised confers a protective benefit.
Descriptive statistics for the estimation sample of African men are presented in Table 2. The average age was roughly 25 years, and the average respondent completed just over 10 years of schooling. Almost all respondents (99%) reported Xhosa as their preferred language, and most men either were Christian (51%) or indicated no religious affiliation (45%). With regard to the circumcision variables, a total of 92.5% reported having been circumcised. Most (66%) reported full removal of the foreskin, 9.7% reported partial removal of the foreskin, and 17% either refused to answer or said that they did not know how much of their foreskin was removed. The mean age of circumcision was roughly 19 years, with 90% of the sample circumcised between the ages of 17 and 22 years. Despite the near-universal normative endorsement of circumcision for men, a minority of men had not been circumcised at the time of the interviews. Uncircumcised men were, on average, younger (the majority being 20 or 21 years old) and from slightly poorer households than circumcised men, indicating that they had not been circumcised yet but might in the future. No significant difference was found between uncircumcised and circumcised men in terms of the sexual behavior variables (not displayed).
All men in the sample had had sex. The vast majority reported only having sex with women (96%), and less than 1% reported sharing needles while injecting drugs, indicating that the primary mode of HIV transmission in the sample was heterosexual intercourse. Significant variation in the other sexual behaviors was reported as follows: 55% reported 5 or more sexual partners, 23% a history of STDs, and 57% having had concurrent partnerships. Of note, 92% of circumcised participants reported sexual debut before circumcision.
In terms of HIV serostatus, 9% tested HIV positive. We computed descriptive statistics by HIV status (not shown here): We found that HIV-positive men were, on average, 1.3 years older (P < 0.01), 0.7 years less schooling (P = 0.02), sexually active for 1.2 years longer (P < 0.01), 15% points more likely to report an STD (P = 0.02), and 18% points more likely to have had a concurrent sexual partnership (P = 0.02). Differences in other characteristics by HIV status were not substantively or statistically significant. Results are available on request.
Table 3 presents descriptive statistics by extent (partial versus full, as this is the main comparison of interest) and age of circumcision. Regarding the former, there were no statistically significant differences between partially and fully circumcised respondents in terms of the demographic and sexual behavior variables. However, those who reported partial circumcision were generally circumcised at later ages (19.9 versus 19.2 years; P = 0.02). In addition, those reporting partial circumcision were nearly twice as likely to test HIV positive (P = 0.08).
With age of circumcision, we found that those who were circumcised earlier tended to be younger, more educated, and less likely to report a history of STDs. These differences were substantively small but statistically significant. We also found that early circumcisers were less likely to test HIV positive, although this difference was not statistically significant.
Table 4 presents the results of the multivariate regression analysis. Column 1 displays probit marginal effects for the circumcised or not variable. The estimate of −0.070 for the circumcised variable illustrates that being circumcised is associated with a 7.0% point reduction in the probability of testing HIV positive. This estimate does not reach statistical significance.
Column 2 examines the extent of circumcision variables, with the base group being those who reported partial removal of the foreskin. We find that, relative to partial circumcision, full circumcision is associated with a 7% point decrease in the probability of testing HIV positive (P < 0.10). Furthermore, the estimates suggest that partial circumcision provides no benefit over not being circumcised, as the estimate on the latter variable is close to zero and statistically insignificant. Interestingly, those who reported being circumcised but did not report an extent had a lower risk of testing positive than partially circumcised men, with the effect magnitude being similar to the estimate on full circumcision (P < 0.05). It is a priori unclear what is driving this result. It may be that the group not reporting extent comprises primarily those who had all their foreskin removed or it could be that those who felt uncomfortable reporting information about their penis were also more cautious sexually or that some other unobserved factor conferring reduced HIV risk was correlated with reluctance to report circumcision extent. Comparisons between nonreporters and fully circumcised individuals and nonreporters and partial circumcisers were unrevealing in this regard, as they show few significant differences in demographic, socioeconomic, and sexual behaviors (results not shown here), and removing these nonreporters from regression analysis did not change the substantive results.
Column 3 of Table 4 focuses on the age of circumcision variable. We find a robust association between timing of circumcision and HIV status, with a 1-year increase in the age of circumcision associated with roughly a 1.1% point increase in the risk of contracting HIV by the end of the survey period (P < 0.10). Given that most individuals were circumcised between 17 and 22 years of age, this amounts to a 5.5% point increase in HIV risk for those who were circumcised at the latest end of that interval vis-à-vis the earliest. Column 4, which presents results for the fully specified model, provides similar results: Full circumcision continues to show a 7% point reduction in HIV positivity vis-à-vis partial circumcision (P < 0.05), and a 1-year increase in age of circumcision is associated with a 1% point increase in HIV risk (P < 0.10).
There is growing interest in using medical male circumcision as a tool for HIV prevention.4,26 But where there is strong cultural support for traditional circumcision, this can stymie efforts to promote early and medical circumcisions.27 Under these circumstances, it is important to gain more understanding of the nature of traditional circumcision and its implications for HIV risk. To the best of our knowledge, ours is the first study to examine the relationship between variations in traditional circumcision practices and HIV status.
We found significant variation in circumcision practices in a population that expressed strong preferences for traditional circumcision and very little support for medical circumcision. Nearly 10% of respondents reported only partial removal of foreskin during circumcision, a figure that concords with previous research using medical examinations among a Xhosa sample.9 We found that full circumcision and earlier age of circumcision were protective against contracting HIV: Circumcising an individual at the age of 17 years as opposed to 22 years was associated with a 5%–6% point reduction in the likelihood of contracting HIV, and full versus partial circumcision conferred a 7% point reduction. Partial foreskin removal conferred no benefit toward reducing HIV risk vis-à-vis no circumcision at all. These substantive results were robust to the inclusion of a rich set of covariates.
There are several limitations to this study, many of which motivate further research. First, extent and age of circumcision were self-reported, and a fifth of the men declined to report their extent of circumcision. Second, omitted variables may confound the relationship between circumcision practices and HIV status. Although we attempted to address this by demonstrating the robustness of the results to the inclusion of a rich set of controls, we still cannot be sure that our estimates reflect causal relationships. Also, complicating causal inference is the fact that some men may have seroconverted before circumcision. However, this is unlikely to be a major source of bias because the HIV prevalence of men 20 years and younger in South Africa is very low.28 Third, sampling error could potentially affect the relationship between HIV status and circumcision extent due to the relatively small percentage of men who were both HIV positive and partially circumcised. Fourth, although the sample was initially representative in 2002, we cannot be certain that the sample was representative in 2009 due to survey attrition and migration into Cape Town. Fifth, our results are for a specific population group, and it is unclear whether they generalize to other populations.
Our findings have several policy implications. Our key finding is that the extent and age of traditional circumcision matters for HIV risk. Given that there is a strong preference for traditional circumcision among young Xhosa men, and medical circumcision is unlikely to be adopted in the near future, working with traditional surgeons to ensure that as much of the foreskin is removed as possible during initiation rites and encouraging earlier circumcision may play an important role in reducing HIV infections. Further research is required to improve our understanding of the relationship between partial removal of the foreskin during traditional circumcision and HIV infection and its implications for male circumcision provision.
Our findings also imply that analyses of DHS data showing no link between male circumcision and HIV status are likely hampered by incomplete measurements. It is not enough to report that the respondent is circumcised or not (as is often the case with the DHS): data must be collected on the extent and age of circumcision if meaningful conclusions are to be drawn about the relationship between HIV status and circumcision. Future studies of HIV determinants—especially in countries where a large percentage of men are circumcised in traditional settings—should measure and control for variations in traditional circumcision practices.
The authors would like to thank Rebecca Thornton, seminar participants at the University of Cape Town Centre for Social Science Research, and 2 anonymous referees for helpful comments and suggestions. Dr. Venkataramani acknowledges the Washington University School of Medicine Forum for International Health and Tropical Medicine for funding travel to South Africa.
1. Auvert B, Taljaard D, Lagarde E, et al. Randomized, controlled intervention trial of male circumcision for reduction of HIV infection risk: the ANRS 1265 Trial. PLoS Med. 2005;2:1112–1122
2. Siegfried N, Muller M, Deeks JJ, et al. Male circumcision for prevention of heterosexual acquisition of HIV in men. Cochrane Database Syst Rev. 2009;2:CD003362 doi:003310.001002/14651858.CD14003362.pub14651852
3. Binagwaho A, Pegurri E, Muita J, et al. Male circumcision at different ages in Rwanda: a cost effectiveness study. PLoS Med. 2010;7:1–10
4. Kahn JG, Marseille E, Auvert B. Cost-effectiveness of male circumcision for HIV prevention in a South African setting. PLoS Med. 2006;3:2349–2358
5. WHO/UNAIDS. New Data on Male Circumcision and HIV Prevention: Policy and Programme Implications. 2007 Montreux, Switzerland WHO/UNAIDS
6. Wilcken A, Keil T, Dick B. Traditional male circumcision in eastern and South Africa: a systematic review of prevalence and complications. Bull World Health Organ. 2010;88:907–914
7. Caldwell J, Orubuloye I, Caldwell P. Male and female circumcision in Africa from a regional to a specific Nigerian examination. Soc Sci Med. 1997;44:1181–1193
8. Bailey R, Egesah O, Rosenberg S. Male circumcision for HIV prevention: a prospective study of complications in clinical and traditional settings in Bongoma, Kenya. Bull World Health Organ. 2008;86:669–677
9. Peltzer K, Nqeketo A, Petros G, et al. Traditional circumcision during manhood initiation rituals in the Eastern Cape, South Africa: a pre-post intervention evaluation. BMC Public Health. 2008;8:64
10. McCoombe SG, Short RV. Potential HIV-1 target cells in the human penis. AIDS. 2006;20:1491–1495
11. Patterson BK, Landay A, Siegel JN, et al. Susceptibility to human immunodeficiency virus-1 infection of human foreskin and cervical tissue grown in explant culture. Am J Pathol. 2002;161:867–873
12. Kelly R, Kiwankuka N, Wawer MJ, et al. Age of male circumcision and risk of prevalent HIV infection in rural Uganda. AIDS. 1999;13:3
13. Akwara P, Fosu G, Govindasamy P, et al. An In-Depth Analysis of HIV Prevalence in Ghana: Further Analysis of Demographic and Health Surveys Data. 2005 Calverton, MD USAID
14. Garenne M. Long-term population effect of male circumcision in generalised HIV epidemics in sub-Saharan Africa. Afr J AIDS Res. 2008;7:1–8
15. Siegfried N, Muller M, Deeks J, et al. HIV and male circumcision—a systematic review with assessment of the quality of studies. Lancet Infect Dis. 2005;5:165–173
16. Warren K HIV and Male Circumcision in Swaziland, Botswana and Lesotho: An Econometric Analysis. 2010 Cape Town, South Africa Centre for Social Science Research, University of Cape Town:1–54 CSSR Working Paper 273
17. Green LW, McAllister RG, Peterson KW, et al. Male circumcision is not
the HIV “vaccine” we have been waiting for! Future HIV Ther. 2008;2:193–199
18. Bärnighausen T, Hosegood V, Timaeus I, et al. The socioeconomic determinants of HIV incidence: evidence from a longitudinal, population-based study in rural South Africa. AIDS. 2007;21:S29–S38
19. Corno L, de Walque D The Determinants of HIV Infection and Related Sexual Behaviors: Evidence From Lesotho Policy Research Working Paper 4421. 2007 World Bank Development Research Group
20. De Walque D. How does the impact of an HIV/AIDS information campaign vary with educational attainment? Evidence from rural Uganda. J Dev Econ. 2007;84:686–714
21. Landman K, Ostermann J, Crump J, et al. Gender differences in the risk of HIV infection among persons reporting abstinence, monogamy, and multiple sexual partners in northern Tanzania. PLoS One. 2008;3:1–6
22. Auvert B, Ballard R, Campbell C, et al. HIV infection among youth in a South African mining town is associated with herpes simplex virus-2 seropositivity and sexual behaviour. AIDS. 2001;15:885–898
23. Gregson S, Nyamukapa C, Garnett G, et al. Sexual mixing patterns and sex-differentials in teenage exposure to HIV infection in rural Zimbabwe. Lancet. 2002;359:1896–1903
24. Cassell M, Halperin D, Shelton J. Risk compensation: the Achilles' heel of innovations in HIV prevention? BMJ. 2006;332:605–607
25. Vincent L. “Boys will be boys”: traditional Xhosa male circumcision, HIV and sexual socialization in contemporary South Africa. Cult Health Sex. 2008;10:431–446
26. Brooks RA, Etzel M, Klosinski LE, et al. Male circumcision and HIV prevention: looking to the future. AIDS Behav. 2010;14:1203–1206
27. Kalichman SC. Neonatal circumcision for HIV prevention: cost, culture and behavioral considerations. PLoS Med. 2010;7:1–2
28. Shisana O, Rehle T, Simbayi L, et al. South African National HIV Prevalence, Incidence, Behavior and Communication Survey 2008: A Turning Tide Among Teenagers?. 2009 Cape Town, South Africa HSRC Press