Worldwide, almost 16 million women are HIV-infected and most are of reproductive age and live in resource-limited settings.1 Access to effective contraception is a critical component of reproductive healthcare for HIV-infected women for a number of reasons. First, a high proportion of HIV-infected women do not want to become pregnant.2-4 Second, access to effective contraception for HIV-infected women not wanting to become pregnant has been identified by the World Health Organization as an important strategy for the prevention of mother-to-child HIV transmission5; this approach is as cost-effective as the provision of antiretroviral drugs to pregnant women.6,7 Use of effective contraception by HIV-infected women not wishing to become pregnant also reduces the number of HIV-related orphans. Third, pregnancy may increase the likelihood of HIV transmission from an infected woman to her uninfected male partner.8 Finally, use of highly effective contraception preserves the health of HIV-infected women by eliminating the mortality and morbidity associated with pregnancy and childbirth.
The choice of appropriate contraception can be difficult for HIV-infected women. Healthcare providers may put strong emphasis on condom use when counseling HIV-infected women and less emphasis on highly effective contraceptive methods. Providers may also be reluctant to prescribe hormonal contraceptives to women using highly active antiretroviral therapy (HAART) because of concerns about potential interactions between hormonal contraception with antiretroviral drugs that can theoretically result in lower contraceptive or antiretroviral efficacy.9
Several studies, including a trial in Zambia of women randomized to either hormonal contraception or intrauterine devices, have raised the concern that hormonal contraception (particularly depot-medroxyprogesterone acetate [DMPA] and oral contraceptives [OCs]) might accelerate HIV disease progression.10-12 However, other studies have found no increase in viral loads or decrease in CD4 levels13,14 and no increased risk in time to AIDS, death, or antiretroviral therapy (ART) initiation among OC or injectable users compared with women not using these methods.15-18
We evaluated whether women using DMPA and OCs had more rapid progression from time of HIV infection to AIDS than women not using hormonal contraception among women attending reproductive health clinics in Uganda and Zimbabwe. This cohort was uniquely suited to examine this question because of accurate measurement of HIV infection timing, short visit windows (every 12 weeks), long follow-up with high retention rates and accurate measurement of contraceptive use, CD4 counts, and clinical outcomes.
The research was approved by the Institutional Review Boards of collaborating institutions. All study participants provided written informed consent.
Study Population and Procedures
The study population were women who became HIV-infected while participating in the Hormonal Contraception and Risk of HIV Acquisition Study and a subsequent serosurveillance phase during 2001 to 2009. The 303 study participants were HIV-infected, ages 18 to 45 years, and used either DMPA (150 mg administered quarterly), OCs (low-dose pills containing 30 μg ethinyl estradiol and 150 μg of levonorgestrel), or no hormonal method.
Study procedures have been described in detail previously.19 Briefly, we notified Hormonal Contraception and Risk of HIV Acquisition Study participants who became HIV-infected about their infection status and scheduled interested women for enrollment into the Hormonal Contraception and HIV Genital Shedding and Disease Progression Study as soon as possible. After conducting informed consent procedures, we interviewed participants to collect sexual behavior, reproductive health, and contraceptive history data. We provided contraceptive, HIV risk reduction, and condom use counseling and free contraceptives and condoms. Study clinicians conducted a standardized physical (including pelvic) examination and collected specimens for reproductive tract infections (RTIs), pregnancy testing, Papanicolaou smears, lymphocyte phenotyping, and plasma and cervical viral loads. We tested for RTIs and pregnancy as previously described.20 We treated participants on-site for vaginal infections and recalled women diagnosed with asymptomatic chlamydia, gonorrhea, or syphilis infections for treatment. HIV subtype determination was performed as previously described using the C2-V3 region of the env gene.19
We conducted follow-up visits at 4, 8, and 12 weeks after enrollment and at 12-week intervals thereafter for up to 9.3 years. Follow-up procedures were similar to those at enrollment and included testing for RTIs and pregnancy.
Beginning in 2003, we offered HAART and trimethoprim-sulfamethoxazole to women who developed severe symptoms of HIV infection (World Health Organization clinical Stage 4 or severe Stage 3 disease) or who had successive CD4 lymphocyte counts of 200 cells/mm3 or less.
Analysis Population and Variable Definition
The analysis population included 303 Ugandan and Zimbabwean women contributing 5300 regular study visits and 1408 years of follow-up.
HIV polymerase chain reaction (PCR; Cobas AMPLICOR; Roche, Branchburg, New Jersey) was performed on samples from visits before HIV seroconversion to establish timing of initial infection.20 For women whose seroconversion visit was also their first PCR-positive visit, HIV-1 infection dates were estimated as the midpoint between this and the previous visit. Because HIV testing was conducted every 12 weeks in the Hormonal Contraception and Risk of HIV Acquisition Study, estimated infection date was usually within a 6-week window of the actual infection date. We estimated acute HIV infections (serologically negative but HIV PCR-positive) to have occurred 15 days before the first PCR-positive visit.
We defined contraceptive exposure for our primary analysis as the cumulative (time-varying) number of months of DMPA and OC exposure from the estimated date of HIV infection up to each study visit. We performed two sensitivity analyses using time-fixed contraceptive exposures. First, we defined contraceptive exposure as the total cumulative number of months of DMPA and OC exposure during the year before AIDS. Second, we used the contraceptive exposure at the estimated time of HIV infection. The contraceptive exposure definitions are similar to those proposed to assess the effect of contraceptive use on HIV-1 disease progression in the literature.21 We used a washout period of 120 days from last injection when women switched from DMPA to the nonhormonal (NH) group.
We defined two analysis end points. Our primary end point was an AIDS diagnosis, which we defined as two successive study visits (or two visits within 6 months) with a CD4 count 200 cells/mm3 or less or World Health Organization clinical Stage 4 disease or severe Stage 3 disease (three or more Stage 3 criteria) (whichever occurred first). Participants not reaching the AIDS end point were censored at the time of HAART initiation or their last follow-up visit. We conducted sensitivity analyses of different hormonal contraception (HC) exposure definitions and analyses of the time to AIDS, HAART initiation, or death (the secondary end point) using the same approaches as for the primary analysis.
Participant characteristics at the HIV seroconversion visit were summarized and compared by contraceptive group using Cochran-Mantel-Haenszel tests for categorical variables and the Kruskal-Wallis test for continuous variables.
We used Cox proportional hazard regressions and logistic regressions adjusted for repeated observations to evaluate bivariable associations between baseline and time-varying characteristics and AIDS diagnosis and to assess potential confounding factors. Time-independent variables were defined as confounders if the hazard ratio for the association between HC exposure and an AIDS diagnosis changed by at least 10% when the variable was added to the primary model and were retained in final multivariate models. This covariate selection process was applied to all data analyses. In addition, for time-varying cumulative contraceptive exposure, if a time-dependent covariate was associated (P < 0.05) with an AIDS diagnosis and predicted HC exposure and was predicted by past HC exposure, it was considered a time-dependent confounder.22 Because the effects of cumulative HC exposure on AIDS diagnosis could be biased by time-dependent confounding, we used a marginal structural Cox survival model with the stabilized inverse-probability treatment weighting approach as the primary analysis method to provide consistent estimates of the cumulative HC exposure effect on time to AIDS adjusted by baseline confounders and variables considered a priori to be important (site, age, HIV subtype).23 Because it was possible that contraceptive exposure at the time of HIV infection could influence the initial CD4 and set point viral load measurements (and thus be intermediary variables in a contraceptive exposure-HIV disease progression pathway), we did not control for these variables in our primary analysis but instead included these variables in sensitivity analyses. We calculated 95% confidence intervals for estimated hazard ratios using the robust sandwich estimate of the covariance matrix.24 We further used conventional Cox proportional hazards models to evaluate covariate association with time to AIDS.
Because all Zimbabwean participants with completed subtyping (n = 96) are Subtype C, we assumed that all the remaining 80 Zimbabwean participants also had Subtype C HIV infections. Additionally, three Ugandan participants with Subtype C infections and five Ugandan participants with missing subtype information were excluded from multivariate modeling.
Data analyses were conducted using SAS Version 9.2 (SAS Institute Inc, Cary, NC).
From a total of 333 seroconverters in the Hormonal Contraception and Risk of HIV Acquisition Study (213 seroconverters) and a follow-on serosurveillance phase (120 seroconverters), 306 women joined the Hormonal Contraception and HIV Genital Shedding and Disease Progression Study. Three study participants did not contribute sufficient follow-up time and were excluded from this analysis. Of the remaining 303 participants, 127 (42%) were Ugandan and 176 (58%) were from Zimbabwe. Women were retained for 95% of their expected follow-up time in the study (Uganda 94%, Zimbabwe 96%).
Participant Characteristics at Time of HIV Seroconversion
At the HIV seroconversion visit, the median age and education was 26 and 10 years, respectively (Table 1). Almost two thirds of women used hormonal contraception including DMPA (36%) and OCs (27%). Only 7% of women were currently pregnant and 11% were breast-feeding. Few women reported multiple partners (6%) or commercial sex (1%). Only 29% of women at HIV seroconversion reported either consistent condom use or not having sex in the previous 3 months. RTI prevalence was high, including 18 women (6%) with chlamydia, 30 women (10%) with gonorrhea, 16 women (6%) with trichomoniasis, 80 women (27%) with bacterial vaginosis, and 244 women (82%) infected with herpes simplex virus-2. Subtype C HIV infection was most common (59%), whereas 28% of women had Subtype A and 11% had Subtype D infections.
At HIV seroconversion, women in the NH group as compared with the hormonal groups were slightly older, more likely Ugandan and pregnant, more likely to have risky sexual behavior, more likely to use condoms but to have sex less often, and more likely to have HIV Subtypes A and D infections (Table 1). No difference existed between groups in cohabitation or educational status, primary partner risk, or in the prevalence of RTIs.
Analyses of AIDS Incidence
There were 255, 213, and 171 study participants who contributed at least 1, 3, and 5 years, respectively, of follow-up data. The median follow-up time of study participants was 58 months. During the study, 111 women received an AIDS diagnosis (median of 47.3 months), whereas 192 women did not (median follow-up, 63.1 months). The AIDS incidence rate (IR) was 7.9 per 100 women years overall. The AIDS IR was 6.6, 9.3, and 8.8 per 100 person-years for DMPA, OC, and NH users, respectively (Table 2). AIDS IR was higher for Uganda Subtype D than for Uganda Subtype A or Zimbabwe Subtype C (11.0, 6.9, and 7.9 per 100 person-years, respectively). AIDS incidence and was also higher for older (25 years or older) than for younger women (9.1 versus 6.1 per 100 person-years).
The cumulative probability of AIDS was 5.6% at 2 years (7.1% for Zimbabwe [ZM], 3.5% for Uganda [UG]), 29.4% at 5 years (29.7% for ZM, 29.1% for UG), and 49.5% at 7 years (49.0% for ZM, 50.7% for UG). There were no differences in the cumulative probability of AIDS by country (P = 0.87).
Cumulative Hormonal Contraceptive Exposure and Time to AIDS
Cumulative exposure to DMPA and OCs was not associated with time from HIV infection to AIDS in marginal structural modeling. The adjusted hazard ratio (AHR) for DMPA and OCs per year of use was 0.90 (95% confidence interval [CI], 0.76-1.08) and 1.07 (95% CI, 0.89-1.29), respectively (Table 3). When baseline CD4 and set point viral load were added to the marginal structural Cox model, the AHRs for DMPA and OC use (compared with no hormonal use) remained very similar.
We also calculated the association of covariates with HIV progression to AIDS from a conventional Cox proportional hazards model, including baseline CD4 and set point viral load. Younger age (18-24 years) (AHR, 0.60; 95% CI, 0.40-0.91) and higher baseline CD4 (per 100 cell increase) (AHR, 0.76; 95% CI, 0.67-0.86) were associated with slower progression to AIDS, whereas set point viral load (per log10 unit increase) (AHR, 1.40; 95% CI, 1.07-1.83) was associated with more rapid disease progression (Table 4). Uganda/Subtype D (AHR, 1.91; 95% CI, 1.00-3.68) infection was also marginally associated with more rapid disease progression.
We also assessed the secondary (combined) end point of time to AIDS, death, or HAART initiation. Of the 303 women in this analysis, 116 women (38%) used ART at some point during the study. Mean CD4 cell count at ART initiation was 175 cells/mm3 and 16 participants initiated ART or died before an AIDS diagnosis. One hundred twenty-seven (49 UG, 78 ZM) women reached the combined end point (IR, 9.0 per 100 women-years). The IR for AIDS, HAART initiation, or death was 10.0, 7.3, and 10.8 per 100 person-years for OC, DMPA, and NH users, respectively. Cumulative exposure to DMPA and OCs was not associated with this end point; the AHR for DMPA and OCs per year of contraceptive use was 0.90 (95% CI, 0.77-1.06) and 1.02 (95% CI, 0.86-1.22), respectively (Table 3). Relationships of other covariates (age, sexually transmitted infection symptoms, unprotected sex acts) with the secondary end point were similar to those for the primary end point (Table 4).
The Effect of Hormonal Contraceptive Use During the Year Before AIDS
We hypothesized that cumulative HC use during the year before AIDS might influence whether a woman reached an AIDS end point. Therefore, we conducted analyses of HC use during the year before AIDS (or censoring). Neither cumulative DMPA (AHR, 0.82; 95% CI, 0.50-1.36) nor OC use (AHR, 1.29; 95% CI, 0.69-2.44) during the year before AIDS was associated with time to AIDS (Table 3). When we considered the effect of HC in the year before AIDS, HAART initiation, or death, we again found no association between either DMPA (AHR, 0.81; 95% CI, 0.51-1.31) or OC use (AHR, 1.26; 95% CI, 0.69-2.29) and this outcome (Table 3).
The Effect of Contraceptive Exposure at the Time of HIV Infection
Finally, we considered the effect of HC exposure at the time of HIV infection on time to AIDS. Neither DMPA (AHR, 1.14; 95% CI, 0.73-1.79) nor OC use (AHR, 0.93; 95% CI, 0.56-1.55) at the time of infection was associated with disease progression (Table 3). The results were largely the same for DMPA (AHR, 1.23; 95% CI, 0.81-1.87) and OC (AHR, 0.85; 95% CI, 0.52-1.39) use at the time of HIV infection when we considered the secondary end point of time to AIDS, death, or HAART initiation.
In this study of women with accurately timed HIV infection dates, we found that neither time-varying cumulative DMPA nor OC use was associated with time from HIV infection until AIDS. This result did not change when we examined the alternative study end point of time from HIV infection to AIDS, HAART initiation, or death. We also found no association of DMPA and OC use with time to AIDS when we considered definitions of HC use during the year before AIDS or at the time of HIV infection.
The cumulative probability of AIDS among women in our study population was 50% by 7 years. This estimate did not vary by country. Although adding baseline CD4 and set point viral load to the prediction model did not change the HC results, Uganda Subtype D infection became associated with faster HIV disease progression.
The results of our study agree with some but not all previous studies as well as a recent review25 of the literature examining the relationship between HC use and HIV disease progression. On the one hand, our results agree with the results of several prevalent HIV cohort studies15-17 and one incident HIV cohort study18 that found no increase in HIV disease progression among women using as compared with women not using HC. These results also confirm our earlier finding of no difference in viral set point levels (as an indicator of future disease progression) between women using DMPA and OCs compared with women not using HC at the time of HIV infection.19 On the other hand, our results conflict with the findings of a trial in which women with prevalent HIV infection randomized to DMPA and OC use experienced more rapid HIV disease progression compared with those randomized to a copper intrauterine device,11,12 Our results also conflict with the finding that sex workers in Mombasa Kenya who used DMPA at the time of HIV infection had higher HIV viral set points than women not using HC at the time of HIV infection.10 As emphasized by the study design, our study results were robust to different definitions of HC exposure before and throughout HIV disease. For example, we did not observe an association between HC use during the year before AIDS (when switches in coreceptor use can lead to accelerated CD4 decline26 and HIV disease progression). Likewise, we found no association between HC exposure at the time of HIV infection and subsequent disease progression. Past research had suggested that HC use might influence the diversity of the infecting virus and result in an increased viral set point.10,27
We found several variables that were associated with time from HIV infection to AIDS. In accordance with a previous report, we found that younger age was associated with less rapid HIV disease progression.28 As would be expected, we also found that higher baseline CD4 was also associated with slower disease progression and higher set point viral load was associated with faster disease progression. Controlling for the lower baseline CD4 levels we found among Zimbabwean than Ugandan women,29 Uganda Subtype D infection was associated with faster disease progression than Uganda Subtype A infection. This has also been reported previously.30-33
Our study has a number of important strengths. The study was prospective with both CD4 cell levels and physical examinations (including World Health Organization HIV clinical staging) conducted every 12 weeks. Clinical staging done by local clinicians was carefully reviewed by the study coinvestigator (R.S.). We had high retention levels in both countries and we followed women from time of HIV infection to AIDS (up to 9 years). We accurately timed HIV infection by conducting HIV PCR testing on serial samples before women HIV seroconverted. HC was supplied to women by the study and we measured exposure carefully using contraceptive calendars and checked data against clinic records. We also carefully measured a number of reproductive tract infections and pregnancy. We considered several different hormonal exposure definitions and thus assessed the robustness of our analyses. Finally, we enrolled women seeking family planning services in two sub-Saharan countries. This allows for greater generalizability of study results than a study population drawn from a selected high-risk group (eg, sex workers) in one locale.
Our study also had limitations. We do not have plasma viral load measurements from all visits and thus were unable to consider time-varying viral load as a potential mediating factor. The study was observational and it is possible that residual confounding remains despite using marginal structural models to properly adjust for time-dependent confounding. In addition, the issue of pregnancy in our analyses is complex and although we adjusted for time-varying pregnancy status, analytic issues remain. Finally, we only sequenced the C2-V3 region of the HIV env and thus cannot fully explore the issue of recombinant viruses on HIV disease progression.
In summary, we found that HC, including DMPA and OC use, was not associated with either time from HIV infection to AIDS or to AIDS, HAART initiation, or death. We also found that the cumulative probability of AIDS was 50% at 7 years from HIV infection and did not vary by country. These results suggest that HIV-infected women who do not want to become pregnant can safely use DMPA and OCs without putting themselves at greater risk of AIDS. Additionally, HIV prevention and family planning programs can make provision of these HCs to HIV-infected women not wanting to become pregnant an important strategy in the prevention of mother-to-child and heterosexual HIV transmission.
We thank Josaphat Byamugisha, MBChB, PhD, for his supervision of the Uganda clinical staff; Marshall Munjoma, MPH, for his supervision of the Zimbabwe laboratory; and Cynthia Kwok, MSPH, for helping in the analysis of study data. We also thank the Hormonal Contraception and HIV Genital Shedding and Disease Progression Study staffs in Uganda and Zimbabwe and especially the Hormonal Contraception and HIV Genital Shedding and Disease Progression Study participants for their long-standing participation in and loyalty to the study.
1. AIDS Epidemic Update, November 2009
. Geneva, Switzerland: Joint United Nations Programme on HIV
/AIDS (UNAIDS) and World Health Organization (WHO); 2009.
2. Hoffman IF, Martinson FE, Powers KA, et al. The year-long effect of HIVpositive test results on pregnancy intentions, contraceptive use, and pregnancy incidence among Malawian women
. J Acquir Immune Defic Syndr
3. Desgrees-Du-Lou A, Msellati P, Viho I, et al. Contraceptive use, protected sexual intercourse and incidence of pregnancies among African HIV
. DITRAME ANRS 049 Project, Abidjan 1995-2000. Int J STD AIDS
4. United Nations Family Planning
Association (UNFPA). UNFPA Annual Report 2000
. New York: United Nations Population Fund; 2001.
5. World Health Organization (WHO), United Nations Family Planning
Association (UNFPA). Glion Consultation on Strenghtening the Linkages Between Reproductive Health and HIV/AIDS: Family Planning and HIV/AIDS in Women and Children
. Geneva: WHO; 2006.
6. Reynolds HW, Janowitz B, Homan R, et al. The value of contraception to prevent perinatal HIV
transmission. Sex Transm Dis
7. Halperin DT, Stover J, Reynolds HW. Benefits and costs of expanding access to family planning
programs to women
living with HIV
. 2009;23(suppl 1):S123-S130.
8. Mugo N, Heffron R, Donnell D, et al. Pregnancy is associated with an increased risk for HIV
transmission among African HIV
-1 serodiscordant couples. Paper presented at: Microbicides 2010; May 23, 2010; Pittsburgh, PA.
9. World Health Organization (WHO). Medical Eligibility Criteria for Contraceptive Use,
4th ed. Geneva: WHO; 2010.
10. Lavreys L, Baeten JM, Kreiss JK, et al. Injectable contraceptive use and genital ulcer disease during the early phase of HIV
-1 infection increase plasma virus load in women
. J Infect Dis
11. Stringer EM, Kaseba C, Levy J, et al. A randomized trial of the intrauterine contraceptive device vs hormonal contraception
who are infected with the human immunodeficiency virus. Am J Obstet Gynecol
12. Stringer EM, Levy J, Sinkala M, et al. HIV disease progression
by hormonal contraceptive method: secondary analysis of a randomized trial. AIDS
13. Richardson BA, Otieno PA, Mbori-Ngacha D, et al. Hormonal contraception
-1 disease progression
among postpartum Kenyan women
14. Cejtin HE, Jacobson L, Springer G, et al. Effect of hormonal contraceptive use on plasma HIV
-1-RNA levels among HIV
15. Stringer EM, Giganti M, Carter RJ, et al. Hormonal contraception
and HIV disease progression
: a multicountry cohort analysis of the MTCT-Plus Initiative. AIDS
. 2009;23(suppl 1):S69-S77.
16. Allen S, Stephenson R, Weiss H, et al. Pregnancy, hormonal contraceptive use, and HIV
-related death in Rwanda. J Womens Health
. Sep 2007;16:1017-1027.
17. Kilmarx PH, Limpakarnjanarat K, Kaewkungwal J, et al. Disease progression
and survival with human immunodeficiency virus type 1 subtype E infection among female sex workers in Thailand. J Infect Dis
18. Polis CB, Wawer MJ, Kiwanuka N, et al. Effect of hormonal contraceptive use on HIV
progression in female HIV
seroconverters in Rakai, Uganda. AIDS
19. Morrison CS, Demers K, Kwok C, et al. Plasma and cervical viral loads among Ugandan and Zimbabwean women
during acute and early HIV
-1 infection. AIDS
20. Morrison CS, Richardson BA, Mmiro F, et al. Hormonal contraception
and the risk of HIV
21. Baeten JM, Lavreys L, Overbaugh J. The influence of hormonal contraceptive use on HIV
-1 transmission and disease progression
. Clin Infect Dis
22. Robins JM, Hernan MA, Brumback B. Marginal structural models and causal inference in epidemiology. Epidemiology
23. Hernan MA, Brumback B, Robins JM. Marginal structural models to estimate the joint causal effect of nonrandomized treatments. J Am Stat Assoc
24. Lin DY, Wei LJ. The robust inference for the Cox proportional hazards model. J Am Stat Assoc
25. Curtis KM, Nanda K, Kapp N. Safety of hormonal and intrauterine methods of contraception for women
/AIDS: a systematic review. AIDS
. 2009;23(suppl 1):S55-S67.
26. Esbjornsson J, Mansson F, Martinez-Arias W, et al. Frequent CXCR4 tropism of HIV
-1 subtype A and CRF02_AG during late-stage disease—indication of an evolving epidemic in West Africa. Retrovirology
27. Sagar M, Lavreys L, Baeten JM, et al. Identification of modifiable factors that affect the genetic diversity of the transmitted HIV
-1 population. AIDS
28. Bonnet F, Thiebaut R, Chene G, et al. Determinants of clinical progression in antiretroviral-naive HIV
-infected patients starting highly active antiretroviral therapy. Aquitaine Cohort, France, 1996-2002. HIV Med
29. Lovvorn AE, Patnaik P, Walker CJ, et al. Variations in CD4 cell counts among HIV
-uninfected and infected women
in Uganda and Zimbabwe. Int J STD AIDS
30. Kiwanuka N, Laeyendecker O, Robb M, et al. Effect of human immunodeficiency virus Type 1 (HIV
-1) subtype on disease progression
in persons from Rakai, Uganda, with incident HIV
-1 infection. J Infect Dis
31. Baeten JM, Chohan B, Lavreys L, et al. HIV
-1 subtype D infection is associated with faster disease progression
than subtype A in spite of similar plasma HIV
-1 loads. J Infect Dis
32. Kaleebu P, French N, Mahe C, et al. Effect of human immunodeficiency virus (HIV
) type 1 envelope subtypes A and D on disease progression
in a large cohort of HIV
-1-positive persons in Uganda. J Infect Dis
33. Senkaali D, Muwonge R, Morgan D, et al. The relationship between HIV
type 1 disease progression
and V3 serotype in a rural Ugandan cohort. AIDS Res Hum Retroviruses