In 2008, the National AIDS Control Organization revised the 2006 India national HIV prevalence estimate for the adult population to 0.36% . This was a downward revision of the 0.8% estimate previously cited . The new estimate used multiple data sources including data from sentinel surveillance sites, the third National Family Health Survey (NFHS-3) 2005–2006, and other behavioural surveys .
This revision is important, as previous estimates had relied primarily on sentinel surveillance of women accessing antenatal clinics (ANC) and people attending clinics for the treatment of sexually transmitted infections (STI), both of which have been shown to have their own biases [3–5]. For example, women receiving private ANC care and women in rural areas are likely to be underrepresented in sentinel surveillance ANC populations, which are generally sampled from urban public sector hospitals . In some settings, women attending public facilities are also more likely to be of lower socioeconomic status and perhaps are more likely to be HIV positive . There has been an increase in the number of sentinel sites from 180 in 1998 to 1122 in 2006 ; however, the use of other data sources is particularly important in adjusting for bias.
India is experiencing a largely heterosexual but heterogeneous epidemic, concentrated in high-risk groups. ANC data have indicated declines in HIV prevalence in southern states of India  and in Pune, a city in the western state of Maharashtra , but whether these are true declines and whether this is occurring elsewhere in India is less clear [10–12]. HIV prevalence from sentinel surveillance varies substantially by state and risk group. Prevalence in the adult population ranges from less than 0.1% in some states to 1.67% in the state of Manipur . Eight out of 26 states have HIV prevalences exceeding 5% among female sex workers (FSW), whereas nine other states have prevalences in this group of less than 1% . These differences highlight the heterogeneity of the epidemic and the importance of measuring state and district-level HIV prevalence.
In a community-based study in the six high prevalence states in India, the NFHS-3 estimated the overall adult HIV prevalence in Karnataka state to be 0.69% . This was adjusted to 0.8% for the National AIDS Control Organization estimate . Mysore is one of the 187 districts in India identified as having an HIV prevalence of approximately 1% in ANC attenders and a relatively large, predominantly street-based, sex worker community. A study conducted in 2005 found 25% HIV prevalence among FSW in Mysore .
The current study is part of the monitoring and evaluation component of Avahan, the India AIDS Initiative of the Bill and Melinda Gates Foundation . This is a large scale HIV prevention programme primarily targeting high-risk populations, sex workers, high-risk men who have sex with men and injection drug users, in six high prevalence states in south and northeast India. Data from this study will also be used along with data from other cross-sectional surveys in mathematical modelling to predict the HIV epidemic trajectory and assess the impact of the intervention programme .
In this paper, we present HIV and STI prevalence estimates within the general population of Mysore district, and examine differences in the distribution of risk factors associated with HIV prevalence among men and women.
Sample determination and data collection
The target sample was 6000 randomly selected individuals, equally distributed between rural and urban areas (clusters) and between men and women (randomly selected from each cluster).
For rural areas, sampling proportional to population size, with a sampling frame based on a complete list of villages in the district with population sizes obtained from the most recent national census, was used for the selection of villages. For urban areas, the sampling frame maintained by the National Sample Survey Organization was used for the selection of urban blocks through systematic random sampling.
A complete household census was carried out in all of the selected 15 rural and 20 urban primary sampling units (PSU; clusters), and formed the basis of the sampling frame for the main survey. The number of respondents randomly selected varied according to the actual population size of the PSU in rural villages as determined during the census. For urban blocks, a fixed number of 150 respondents was randomly selected as blocks were of similar size. In total, 5991 individuals were sampled, respondents that were not at home were visited again by field workers up to five times to maximize the possibility of inclusion. Eligibility criteria for inclusion in the main survey were usual residents of the household (living in the household >6 months), aged between 15 and 49 years, slept in the household the night before the census. Those not mentally competent or who permanently moved out of the PSU between census and main survey were excluded. Informed written consent was obtained from every participant, separately for interview administration and biological testing. Names were needed for the identification of sample individuals obtained from the census but were never marked on the biological samples; results were linked to sociodemographic and behavioural information through a unique study code. The census sample file was destroyed when recruitment finished.
Questions were administered in an extensive face-to-face interview and included sociodemographic information, data on types of partnerships, sexual behaviour and STI history.
Urine and blood specimens were collected from all consenting participants. Dried blood spots were obtained when venous blood drawing was not possible or consented to. HIV testing was anonymous, but free referral to local HIV voluntary counselling and testing services was made, and participants who were symptomatic for an STI were referred to a specifically trained local physician for free assessment and treatment.
Ethical approval for the study was granted by the institutional review boards of the Centre Hospitalier affilié Universitaire de Québec, Québec, Canada; University of Manitoba, Winnipeg, Canada and St John's Medical College, Bangalore.
Serum and dried-blood spot samples were tested for HIV using an enzyme-linked immunosorbent assay (ELISA) test (J. Mitra, J. Mitra and Company Private Ltd., New Delhi, India), with confirmation by a second ELISA test (Genedia, Greencross Lifescience Corp., Kyunggi-do, South Korea). A test was deemed HIV positive if both tests were found positive, only two specimens had discordant results and these were considered negative. Urine samples were stored at 4oC and then tested for HIV using an ELISA followed by Western blot confirmation (Calypte Biomedical Corporation, Berkeley, California, USA), for respondents who refused to provide blood. Sera were also tested for syphilis using a rapid plasma reagin test (Span Diagnostics, Surat, India), with a Treponema pallidum haemagluttination test confirmation (Glaxo-Omega, Alloa, Scotland, UK) and a respondent was considered as having active syphilis when both tests were positive. Herpes simplex virus type 2 (HSV-2)-specific antibodies were tested on a random subsample of 1/8 sera using an ELISA test (Kalon Biological, Surrey, UK).
Urine samples were tested using nucleic acid amplification tests (Aptima Combo-2 assay, Gen-Probe, San Diego, California, USA) for both Neisseria gonorrhoeae and Chlamydia trachomatis.
Prevalence estimates were weighted according to sampling and population distribution; separate weights were calculated for urine, blood, any biological specimen, and for the questionnaire items, in order to adjust for non-response in each sampling domain. These weights represented the inverse of the probability of selection. Logistic regression analyses were conducted using STATA 9.0 to calculate odds ratios (OR) to estimate the strength of associations between potential risk factors and HIV infection. Analyses were stratified by sex. Intraclass correlation was accounted for within PSU (urban/rural) using survey commands to correct the standard error and provide more accurate confidence intervals (CI) around the estimates. All models were constructed using a forward selection technique to allow for the rarity of the outcome, variables with a P value of less than 0.1 in univariate analysis were entered into the multivariate model. Variables with the strongest association were added first and treated as the main independent variables; those with the weakest association were added later as potential confounders of the main effect. Any variable that was not significantly associated in multivariate analysis and did alter the OR of other variables in the model by 10% or more was retained as a confounder.
Sample structure and demographics
Of the 5991 people from the census survey who were sampled for participation in the main study, 262 (4.4%) were not eligible at the moment of the main survey. Out of the 5729 eligible subjects, 81.2% (4653) were interviewed. Of those, 90% (4199 or 73.2% of all eligible subjects) also provided at least one biological specimen. Approximately half of all non-participants (i.e. those who were not even interviewed) were not at home. Specific refusal to participate accounted for the other half of the non-participants.
Using data from the household census questionnaire, respondents and non-respondents were compared according to the distribution of some main characteristics including age and marital status (Table 1). A difference in distribution between the two groups (P < 0.01) was found for all variables with the exception of age. This difference remained even when excluding those who refused to participate.
Differences between participants who underwent interview and then agreed or declined to give biological samples were examined (Table 1). Female and urban respondents were significantly less likely to provide a biological specimen, as were those with at least secondary level education. We also examined the proportion with potential risk factors for HIV. Those who reported ever having a medical injection (OR 1.7, 95% CI 1.29–2.69, P < 0.01), ever having genital discharge (OR 2.56, 95% CI 1.39–4.72, P < 0.01), travelling for work (OR 1.33, 95% CI 1.01–1.74, P = 0.04) and more than one lifetime partner (OR 2.76, 95% CI 1.50–5.08, P < 0.01) were more likely to provide a specimen. All respondents reporting a genital ulcer or new partner in the past year, and all women reporting having received money for sex, provided a biological specimen.
The mean age of respondents was 30 years and 67% of respondents were currently married; this proportion was significantly higher among women than men (74 and 59%, respectively), and slightly higher in rural than in urban areas (69 and 64%). Overall, 41% of the sample were illiterate, whereas 30% had completed secondary-level education or above. Literacy differences were most marked between rural and urban respondents, with illiteracy reported by 50% of rural and 28% of urban respondents compared with proportions reporting completed secondary-level education of 20 and 43%, respectively. The majority of respondents were Hindu (91%).
HIV and sexually transmitted infections
Thirty-two cases of HIV infection were found in the sample. Weighted HIV prevalence was 0.8% (95% CI 0.52–1.09) overall, and 0.7% (95% CI 0.35–1.08) and 0.9% (95% CI 0.51–1.37) in rural and urban populations, respectively (Table 2). Despite higher HIV prevalence in urban areas, the prevalence of syphilis, gonorrhoea and chlamydial infections were similar for urban and rural areas (Table 2). The prevalence of active syphilis among rural and urban populations was 1.6% (1.03–2.2) and 1.3% (0.78–1.9), respectively, and there was a statistically significant twofold difference in prevalence between men (2%, 1.3–2.7) and women (1%, 0.55–1.5) in the urban sample (OR 0.57, 95% CI 0.33–0.98, P = 0.04). The prevalence of gonorrhoea was extremely low, only five cases found in total. The prevalence of HSV-2 was 11.6% (95% CI 7.1–16.0), 12.7% (95% CI 6.2–19.1) among women and 10.6% (95% CI 6.2–14.9) among men.
Sexual partnerships and behaviour
Approximately three-quarters of the sample (72.6%) reported ever having had sex (Table 3), the difference between men and women reflecting the age difference at marriage; the mean age at marriage being 24.5 years for men compared with 17 years for women. Only 3.8% of respondents reported using a condom at first intercourse; this proportion was significantly higher in urban (5.6%) compared with rural (3.8%) areas.
The mean number of lifetime partners was two for men (range 1–50) and one for women (range 1–3), and 99.9% of women reported a spouse or cohabiting partner as their first sexual relationship type, compared with 76.4% for men. The proportion of respondents who reported having a non-regular partner in the past 12 months was 1.6%, and condom use at last sex was twice as high among this group as among those reporting sex with only a regular partner.
Forty-four male respondents reported sex with an FSW, 1.4% of rural and 2.5% of urban men. Married men were more likely to report ever having had sex with a sex worker than those who were unmarried (2.0% compared with 1.4%), as were men with more lifetime partners. This association between marital status and sex with a sex worker was not statistically significant, however, and when age-adjusted the association was reversed. Five out of the six women reporting having received money for sex were widowed, separated, deserted or divorced.
Having experienced heterosexual anal intercourse was reported by 2.6% (39) of men and 0.3% (10) of women. Interestingly, 68 (2.3%) of female respondents reported not knowing if they had ever experienced anal sex. Thirteen men (0.6%) reported ever having anal sex with another man and 17 (0.8%) reported not knowing if they had ever done so; none tested HIV positive.
Risk factors associated with HIV infection
For men, ever having used a condom, ever having paid for sex, and the number of lifetime sexual partners were all significantly associated with HIV infection in univariate analysis (Table 4). Active syphilis infection and ever having had a genital discharge or ulcer also showed some degree of association (although not statistically significant), and were included in the multivariate model. HIV prevalence in non-circumcised men was 1.1% (N = 1732) compared with 0% (N = 193) in circumcised men (Fisher's exact χ2 0.25). In the adjusted analysis, ever having used a condom (OR 2.75, 95% CI 1.01–7.5, P = 0.05) and the number of lifetime sexual partners (OR 6.9, 95% CI 2.2–21.9, P < 0.01) both remained strongly associated with HIV infection in men, with only a modest reduction in the adjusted OR for both.
Among women, variables associated with HIV infection in univariate analysis were anal sex, number of lifetime partners and whether a condom was used at last sex (Table 5). In multivariate analysis the number of lifetime sexual partners (OR 10.9, 95% CI 2.0–60.1, P < 0.01) and condom use at last sex (OR 10.5, 95% CI 2.1–53.8, P < 0.01) were strongly associated with HIV infection. Respondents who reported not knowing whether they had had anal sex were at higher risk of HIV than those reporting no anal sex (OR 9.1, 95% CI 1.1–72.3, P = 0.04) (Table 5). The inclusion of marital status in the model reduced the adjusted OR for the number of lifetime partners and condom use at last sex by 10%.
Mysore has been identified as a high prevalence area for HIV, but this study found a prevalence of approximately 0.8%, which is only slightly higher than the NFHS-3 findings for Karnataka state overall (0.69%), and approximately the same as the 2006 ANC attender estimate of 1% for Mysore . The prevalence was approximately twofold higher than the overall India national adult HIV prevalence estimate for 2006 (0.36%), but the latter includes large states in northern India where prevalence is very low.
HIV prevalence was higher in urban than rural respondents, as has been found elsewhere in India , but this difference was not statistically significant. The reverse was seen in a similar community-based study in Bagalkot district, northern Karnataka, where HIV prevalence was higher in rural respondents (3.6% compared with 2.4%). The overall prevalence was also much higher in Bagalkot; 2.9% . HIV transmission dynamics may be different between southern and northern Karnataka, and this merits further study.
The prevalence of HSV-2 in the random subset of those providing sera was approximately 10% overall, similar to the prevalence found in a community-based survey of STI prevalence in Chennai, Tamil Nadu , but much lower than in the Bagalkot survey . In that study, self-reporting of ever having had symptoms of a genital ulcer or discharge was 1% for men and 12% for women, and 0.3 and 5.3%, respectively, for symptoms of an STI in the past 12 months. This compares with 1.1 and 4% who reported such symptoms in the previous 12 months during the NFHS-3 . Comparison of self-reported symptoms and laboratory data suggest underreporting of symptoms among men. The prevalence of any bacterial STI among men was 2.8%, nine times the prevalence of self-reported symptoms and almost twofold higher than the prevalence of bacterial STI among women. This could, however, also be a result of syphilis comprising the majority of bacterial STI and those infected being in the latent stage of infection and truly asymptomatic.
The response rate was relatively high; 80% of those eligible were interviewed. High-risk behaviour seemed to be associated with providing a sample and could mean an overestimate in HIV prevalence. It is likely, however, that those who did not participate at all were also high risk, thus reducing the aforementioned bias. For example, underrepresentation of those travelling for work is a possible source of bias, as these individuals were absent from home at the time of interview, and this has been linked to having a higher risk of HIV infection .
Greater numbers of lifetime partners were associated with HIV infection in the adjusted analysis for men and women, and this association has been found in another India-based study . Despite low levels of condom use generally, condom use was also associated with higher HIV prevalence, ever having used a condom for men, and having used one at last sexual intercourse for women. This is in opposition to the trend found among STD clinic attendees in Pune, where condom use was found to be associated with a reduced risk of HIV , but concurs with the population-based findings of NFHS-3 . An explanation for this reverse association is that people with high-risk behaviour first get infected by HIV and, after learning about HIV and becoming aware of their high-risk behaviour, begin to use condoms.
Among women, the odds of HIV infection were increased 2.5 times for those reporting anal sex compared with those not reporting it; however, the strength of the association was weaker in the adjusted analysis. Ten female respondents reported anal sex compared with 60 who reported not knowing whether they had had anal sex. This latter category was significantly associated with HIV infection, compared with the group reporting no anal sex. This could be a result of the misreporting of anal sex because of cultural taboos, although this is complicated by the possibility that this population may have been less aware of the term or practice.
There is evidence that social desirability bias has had an impact on the findings of this study. Comparative data on sensitive behavioural questions, such as having paid/received money for sex, and having had anal sex, were collected via polling booth survey methodology at the same time as the current study. This methodology involves giving a simple yes/no answer using a voting card placed anonymously, with no respondent identifiers, and with no individual interaction with an interviewer, into a sealed box [23,24]. Results were compared with those using the face-to-face interview techniques described in this paper, and higher rates of a number of HIV-related risky sexual behaviours were reported using the polling booth survey methodology. For example, a higher proportion of married and unmarried men reported having paid for sex, and a higher proportion of married women reported having received money for sex . Such differences suggest significant underreporting of risk behaviour. Furthermore, there are discrepancies between the biological and interview data from the present study; six respondents (all male) who reported never having had sexual intercourse were found to be HIV positive. These problems with validity have been identified by others  and underline the difficulty of obtaining accurate information in the context of strong societal mores around sexual behaviour.
As a consequence of the likely underreporting of risk behaviour, odds ratios are likely to have been biased towards null. Also, because of the lower than expected HIV prevalence, there were wide CI around some of the adjusted OR. Other potential risk factors for HIV infection such as injecting drug use and lack of circumcision were difficult to assess in this sample. Injecting drug use was rare, and the majority of the sample comprised uncircumcised Hindus (93% of those who were circumcised were Muslim), so despite evidence that male circumcision has a protective effect against HIV infection [26–28], absolute numbers were too small to detect an association.
Despite the difficulties with underreporting of sexual behaviour, population-based surveys are valuable for validating and improving HIV infection estimates, as well as examining potential risk factors for transmission, particularly if conducted repeatedly over time. Important differences in HIV prevalence and risk factors have been found between states, but also between districts within states; monitoring the epidemic therefore requires a multidistrict, multistate approach. The relatively low HIV and STI prevalences in the sample suggest that levels of risky behaviour are indeed relatively low in the general population. HIV prevalence among high-risk groups in Mysore has, however, been shown to be high [6,14], so HIV prevention efforts should continue to focus on high-risk groups such as FSW, clients of FSW, men who have sex with men and transgendered populations.
The authors would like to thank all the field and data entry staff for collecting and compiling the data, as well as the laboratory that tested the specimens, and of course the study participants in Mysore district. Thanks also to Jan Bradley for revisions made to the manuscript. Michel Alary is a national researcher of the Fonds de la Recherché en Santé du Québec, Canada (grant no. 8722). Regulatory approval for the study was obtained from the Health Ministry Screening Committee (HMSC), India.
HLM was responsible for data analysis and writing the manuscript. PB and AAJ were responsible for coordinating the data collection and contributed to the interpretation of the results. BM contributed to the data analysis and to the interpretation of results. CML participated in the study design, the development of the questionnaires, the overall supervision of the supervision of the study and the interpretation of the data. BMR contributed to the study design and supervised the data analysis. RW was the principal investigator in India and was responsible for the daily supervision of the coordination work. LJ and KM were responsible for the design of the laboratory aspects of the study and carried out the laboratory testing. SM and JB participated in the design of the study and contributed to the interpretation of the results. MA was responsible for the study design and the overall interpretation of the data and participated in the development of the questionnaires and the overall supervision of the study. All of the authors participated in revising the manuscript critically for important intellectual content and approved the final version of the paper.
Sponsorship: This research was funded by the Bill and Melinda Gates Foundation.
The views expressed herein are those of the authors and do not necessarily reflect the official policy or position of the Bill and Melinda Gates Foundation.
Conflicts of interest: James F. Blanchard receives funding from Canada Research Chairs, Health Canada.
All other authors declare no conflict of interest.
2. UNAIDS/WHO. Epidemiological fact sheets on HIV/AIDS and sexually transmitted infections
. 2006 Update. Available at: http://www.who.int/GlobalAtlas/predefinedReports/EFS2006/EFS_PDFs/EFS2006_IN.pdf
. Accessed: 29 March 2008.
3. Dandona L, Lakshmi V, Kumar GA, Dandona R. Is the HIV burden in India being overestimated?. BMC Public Health 2006; 6:308.
4. Correa M, Gisselquist D. Routes of HIV transmission in India: assessing the reliability of information from AIDS case surveillance. Int J STD AIDS 2006; 17:731–735.
5. Ghys PD, Kufa E, George MV. Measuring trends in prevalence and incidence of HIV infection in countries with generalised epidemics. Sex Transm Infect 2006; 82:52–56.
6. National AIDS Control Organisation (NACO), Government of India. Annual sentinel surveillance country report, 2006
. Available at: http://www.nacoonline.org/Search/?quicksearchquery=Annual+Sentinel+Surveillance+Country+Report+2006
Accessed: 2 March 2008.
7. Dandona L, Lakshmi V, Sudha T, Kumar GA, Dandona R. A population-based study of human immunodeficiency virus in South India reveals major differences from sentinel surveillance-based estimates. BMC Public Health 2006; 4:31.
8. Kumar R, Jha P, Arora P, Mony P, Bhatia P, Millson P, et al
. Trends in HIV-1 in young adults in south India from 2000 to 2004: a prevalence study. Lancet 2006; 367:1164–1172.
9. Mehendale SM, Shepherd ME, Divekar AD, Gangakhedkar RR, Kamble SS, Menon PA, et al
. Evidence for high prevalence & rapid transmission of HIV among individuals attending STD clinics in Pune, India. Indian J Med Res 1996; 104:327–335.
10. Gupte MD, Mehendale SM, Pandav CS, Paranjape RS, Kumar MS. HIV-1 prevalence in young adults in south India. Lancet 2006; 368:114.
11. Soman CR, Mathew A. HIV-1 prevalence in young adults in south India. Lancet 2006; 368:113.
12. John T. HIV-1 prevalence in young adults in south India. Lancet 2006; 368:113–114.
13. International Institute for Population Sciences (IIPS) and Macro International. National Family Health Survey (NFHS-3), 2005–6: India
,Vol. I. Mumbai: IIPS; 2007.
14. [14.] Reza Paul S, Moses S, Ramesh BM, Halli S, Washington R, Modal S, et al
. An integrated behavioral and biological study among female sex workers in Mysore, India.
In: 15th Biennial Congress of the International Society for Sexually Transmitted Disease Research (ISSTDR)
. Amsterdam, the Netherlands, 10–13 July 2005 [abstract MP-106].
15. Avahan – The India AIDS Initiative. The business of HIV prevention at scale
. New Delhi, India: Bill and Melinda Gates Foundation; 2008. Available at: http://www.gatesfoundation.org/nr/Downloads/globalhealth/aids/Avahan_HIVPrevention.pdf
Accessed: September 2008.
16. Boily M-C, Lowndes CM, Vickerman P, Kumaranayake L, Blanchard J, Moses S. Evaluating large-scale HIV prevention interventions: study design for an integrated mathematical modelling approach. Sex Transm Infect 2007; 83:582–589.
17. National Institute of Health and Family Welfare (NIHFW) and National AIDS Control Organisation (NACO). Annual HIV sentinel surveillance country report 2006. New Delhi: NIHFW; 2007.
18. Kang G, Samuel R, Vijayakumar TS, Selvi S, Sridharan G, Brown D, et al
. Community prevalence of antibodies to human immunodeficiency virus in rural and urban Vellore, Tamil Nadu. Natl Med J India 2005; 18:15–17.
19. Becker ML, Ramesh BM, Washington RG, Halli S, Blanchard JF, Moses S. Prevalence and determinants of HIV infection in South India: a heterogeneous, rural epidemic. AIDS 2007; 21:739–747.
20. Panchanadeswaran S, Johnson SC, Mayer KH, Srikrishnan AK, Sivaran S, Zelaya CE, et al
. Gender differences in the prevalence of sexually transmitted infections and genital symptoms in an urban setting in southern India. Sex Transm Infect 2006; 82:491–495.
21. Solomon S, Kumarasamy N, Ganesh AK, Amalraj RE. Prevalence and risk factors of HIV-1 and HIV-2 infection in urban and rural areas in Tamil Nadu, India. Int J STD AIDS 1998; 9:98–103.
22. Rodrigues JJ, Mehendale SM, Shepherd ME, Divekar AD, Gangakhedkar RR, Quinn TC. Risk factors for HIV infection in people attending clinics for sexually transmitted diseases in India. BMJ 1995; 311:283–286.
23. Lowndes CM, Jayachandran AA, Pradeep BS, Ramesh BM, Washington R, Mahapatra B, et al
. Higher levels of HIV risky behaviour reported in polling booth surveys compared to face-to-face interviews in a general population survey in Mysore district, Karnataka state, southern India
. In: 17th International Society for Sexually Transmitted Diseases Research (ISSTDR)
. Seattle, USA, 21 July–1 August 2007 [abstract 548].
24. Blanchard JF, Halli S, Ramesh BM, Bhattacharjee P, Washington RG, O'Neil J, et al
. Variability in the sexual structure in a rural Indian setting: implications for HIV prevention strategies. Sex Transm Infect 2007; 83(Suppl. 1):i30–i36.
25. Cleland J, Boerma JT, Carael M, Weir SS. Monitoring sexual behaviour in general populations: a synthesis of lessons of the past decade. Sex Transm Infect 2004; 80:1–7.
26. Auvert B, Taljaard D, Lagarde E, Sobngwi-Tambekou J, Sitta R, Puren A. Randomized, controlled intervention trial of male circumcision for reduction of HIV infection risks: the ANRS 1265 trial. PLoS Med 2005; 2:e298.
27. Bailey RC, Moses S, Parker CB, Agot K, Maclean I, Krieger JN, et al
. Male circumcision for HIV prevention in young men in Kisumu, Kenya: a randomised controlled trial. Lancet 2007; 369:643–656.
28. Gray RH, Kigozi G, Serwadda D, Makumbi F, Watya S, Nalugoda F, et al
. Male circumcision for HIV prevention in men Rakai, Uganda: a randomised control trial. Lancet 2007; 369:657–666.