Vaginal microbicides are one class of products being investigated for their potential to reduce the risk of penile-vaginal human immunodeficiency virus (HIV) transmission. Microbicide effectiveness studies typically enroll thousands of women in regions with high incidence of infection, particularly Sub-Saharan Africa.1,2 Participants are randomized to receive the candidate microbicide or placebo as well as a comprehensive package of counseling, diagnosis and treatment of sexually transmitted infections, and provision of condoms. The risk of infection is then compared between study arms based on monthly or quarterly HIV testing over a year or more of follow-up per participant.
A key challenge to assessing the performance of a vaginal microbicide is the fact that there may be little or no objective information regarding when and how exposure to HIV takes place, or whether the product was used as prescribed for acts of vaginal intercourse. As a consequence, clinical trials do not primarily estimate the reduction in risk of acquiring HIV due to an exposed act of vaginal intercourse when the product is used correctly (i.e., efficacy); they estimate the overall reduction in risk, regardless of the number and type of exposures that take place in each HIV test interval.
Achieving high levels of adherence is essential to the successful evaluation of a candidate vaginal microbicide. Protocol requirements to withdraw product due to adverse events or when a participant becomes pregnant diminish study power,3,4 but this can be largely accounted for when designing the trial (e.g., by requiring the use of effective contraception). Of greater concern is poor adherence due to personal preference; the decision to use (or not use) the product is largely controlled by participants and their partners, and the identification of a product efficacious against penile-vaginal transmission might be circumvented if adherence is poor.5 In this model, we assess the impact of a behavior which has the potential for an equal if not more devastating impact on the power of vaginal microbicide trials; the practice of receptive anal intercourse (RAI). Because we do not know whether vaginal or anal intercourse leads to infection, the true effect of a microbicide on the rate of vaginal HIV acquisition could easily be masked by rectally acquired infections.4
It is becoming increasingly clear that women in the developed6–8 and developing world9–12 engage in RAI. The frequency of this activity is highly variable but in one recent study of a US sexually transmitted disease clinic population, 16% to 21% of women reported RAI with their last partner.8 Unfortunately the rectal mucosa is highly vulnerable to HIV transmission, with a conservative estimate of a 10- to 20-fold increased risk of HIV transmission associated with anal13,14 compared to vaginal intercourse15,16; these data do not reflect the potential for even greater risk when sexually transmitted infection's coinfections are present.17 This differential in HIV transmission rates lies at the heart of our concern that RAI might have a serious impact on the ability to identify products that reduce the risk of penile-vaginal transmission of HIV infection.
For our concerns to have validity, we must have evidence that RAI is occurring in vaginal microbicide studies. Table 1 presents data on RAI in completed vaginal microbicide trials.18–24 These data are quite heterogeneous as various research groups capture data on RAI across a range of timeframes. Furthermore, stigma associated with RAI may lead to underreporting, especially when trial participants are counseled against practicing RAI. Nonetheless, examination of available data suggests that RAI is occurring among trial participants, albeit in levels that vary by study, site, and population. In the COL 1492 (a nonoxynol-9 based gel),19 prior history of RAI was reported to range from 41% among sex workers in Durban, South Africa to less than 5% in Thailand, with 75% of women at the Durban site reporting RAI at some point during the trial. In a recent multicountry study of cellulose sulfate,23 0% to 3.2% of participants reported RAI during the first month of follow-up (unpublished data). Given that participants in vaginal microbicide studies are practicing RAI, we sought to determine at what level it might impact the ability to detect the effect of a vaginal microbicide.
A mathematical model is proposed for expressing effectiveness and power as a function of microbicide efficacy, the probability that the microbicide is used for vaginal acts of intercourse with exposure to HIV, the probability that an act of intercourse with exposure to HIV is rectal, and the ratio of transmission probabilities for rectal versus vaginal intercourse.
The effectiveness of a microbicide being measured in a trial (denoted E) will be less than the typical use reduction in risk from penile-vaginal intercourse (EV) whenever nonvaginal exposure to HIV takes place. Although EV is of primary interest, it can at best be approximated using imprecise self-reported measures of adherence and exposure. Here we focus on penile-vaginal and penile-rectal exposures, but others routes (e.g., oral or intravenous) are possible.
We make a number of simplifying assumptions to facilitate our analysis. Namely, we assume that exposure to HIV only occurs due to unprotected vaginal or rectal intercourse with an infected partner, that all unprotected acts of penile-vaginal intercourse with an infected partner have the same transmission probability (TV), and that all unprotected acts of RAI with an infected partner have the same transmission probability (TR). Thus we ignore intra- and intersubject variability in the probability of HIV acquisition, including variability due to incorrect or inconsistent condom use. Although inadequate for modeling infection dynamics in a population, these assumptions are reasonable when assessing study power.
In addition to the route of exposure, the measured effectiveness of a microbicide depends on the absolute number of exposures to HIV that take place over each HIV test interval. If the probability that a woman becomes infected in an interval is small, however, then E can be reasonably approximated as one minus the relative risk of infection for a single act with exposure to HIV. Let PV denote the probability that such an act is vaginal and UM equal the probability that a woman assigned to the active product uses microbicide for a vaginally exposed act. Then the effectiveness level measured in the study is given by,
If all exposures are vaginal (i.e., if PV = 1.0) then E = EV. If the probability of a rectal exposure is nonzero, however, then E will be less than EV; how much less depends on the ratio (TR/TV) of rectal to vaginal transmission probabilities.
In order to quantify the potential impact of RAI, we consider a hypothetical study that randomizes women in a 1:1 allocation ratio (active vs. placebo) and follows them through time until 160 incident infections have been observed (a trial of this size would not be atypical in the HIV prevention field). If the product is used correctly for UM = 80% of vaginal acts with exposure to HIV, then a 50% efficacious microbicide would result in a reduction in risk due to vaginal intercourse of EV = 40% (EV = efficacy × UM under the stated assumptions). In the absence of nonvaginal exposure to HIV, the study would have 90% power based on a standard formula25:
where L is the number of infections, Z1−α/2 is the (1−α/2)'th percentile of a standard normal distribution, α is the type 1 error of the test, ln(·) is the natural logarithm function, and φ(·) is the cumulative normal distribution function. Here we assume that L = 160; α, = 0.05; Z1−α/2 = 1.96; and EV = 0.4. By substituting E for EV in the above equation, we obtain power values that explicitly account for the practice of RAI.
We have generated data on measured effectiveness levels for the example study, assuming between 0% and 5% of all sex acts with exposure to HIV are rectal and that the rectal transmission probability is 1-, 10-, 20-, or 30-times that of vaginal exposure.13,14
Our model demonstrates that relatively modest levels of RAI occurring within the context of a vaginal microbicide effectiveness study can lead to a significant reduction in the power of the study to identify an effective vaginal microbicide for HIV prevention (Table 2 and Fig. 1).
If the rectal transmission probability is conservatively assumed to be 10 times that of vaginal intercourse and if 2% of all acts with exposure to HIV are rectal then measured effectiveness is reduced to 33.2% and power drops from 90% to 72%; if the rectal transmission probability is 20 times that of vaginal intercourse then power to detect an otherwise 40% effective product is only 56%. Higher rates of RAI clearly have more devastating impacts on study power.
In this paper, we demonstrate that heterosexual anal intercourse has the potential to significantly reduce the power of a vaginal microbicide trial, which could help to explain why no vaginal microbicide has been proven effective in preventing or reducing HIV acquisition secondary to penile-vaginal infection intercourse. To make a more precise assessment of the impact, it would be necessary to know the relative rate of HIV transmission for RAI versus vaginal intercourse in the Sub-Saharan settings where most microbicide trials are conducted. It seems unlikely, however, that this would differ substantially from the US and European estimates which formed the basis of our model.13,14,16 Greater detail is also needed regarding the absolute frequency of RAI in prevention trials. Despite these limitations, our results have clear implications for the design, conduct, analysis, and interpretation of vaginal microbicide effectiveness studies.
First and foremost, the screening process for inclusion in trials should include a detailed assessment of whether women have a history of RAI. These data may be best collected using computer-assisted techniques that have been shown in some studies to encourage participants to report sensitive sexual behaviors.26 Women with a history of RAI could be excluded from entering the study, asked to avoid RAI, and/or provided with focused counseling on the need to encourage their partners to use condoms for RAI intercourse when it occurs. Similar counseling should be provided to all participants, irrespective of their sexual history. However, it is acknowledged that inclusion or exclusion of study participants based on self-reported behavior is difficult. As one example, vaginal microbicide studies routinely exclude women who desire to become pregnant during the projected time frame of the study. Nethertheless, pregnancy rates in women enrolled in vaginal microbicide studies are frequently greater than 10%.27 One concern about excluding women with a history of, or intent to practice RAI is that these women might have a higher risk of penile-vaginal HIV infection than women who do not practice RAI.28 Excluding these women from a vaginal microbicide trial might reduce the HIV seroincidence within the trial population and lead to longer or larger clinical trials.
Irrespective of decisions about inclusion or exclusion criteria concerning RAI, it will be critical to collect accurate sexual behavioral data on RAI throughout vaginal microbicide trials; without this information, a promising vaginal microbicide candidate could be improperly abandoned due to the impact of RAI on the study outcome.
From a statistical perspective, it has been argued that adherence-based analyses would allow for substantial gains in power, and perhaps unbiased estimates of efficacy, in randomized microbicide trials.29 A natural extension of this strategy would adjust for rates of nonvaginal intercourse as well. Implementing such strategies are fraught with difficulty, however, since the acts that lead to infection, and hence behaviors during those acts, are unknown. If risk-taking behaviors are not independent of the infection status of partners (as seems likely), and if the infection status of partners is unknown (as is common), causal modeling will fall victim to measurement error and confounding.
In summary, the described model and example data suggest that RAI has the potential to significantly reduce the power of a vaginal microbicide study to identify an effective product. Future effectiveness studies must include rigorous efforts to identify women who practice RAI and provide appropriate counseling to limit the impact of this behavior on study outcome. In addition, there is a clear need to accelerate development of rectal microbicides for men and women who are at risk of HIV acquisition through RAI.30
1. Braunstein SL, van de Wijgert JH, Nash D. HIV incidence in sub-Saharan Africa: A review of available data with implications for surveillance and prevention planning. AIDS Rev 2009; 11:140–156.
2. Abdool Karim SS, Churchyard GJ, Abdool KQ, et al. HIV infection and tuberculosis in South Africa: An urgent need to escalate the public health response. Lancet 2009; 374:921–933.
3. Raymond EG, Taylor D, Cates W Jr, et al. Pregnancy in effectiveness trials of HIV prevention agents. Sex Transm Dis 2007; 34:1035–1039.
4. Masse BR, Boily MC, Dimitrov D, et al. Efficacy dilution in randomized placebo-controlled vaginal microbicide trials. Emerg Themes Epidemiol 2009; 6:5.
5. Weiss HA, Wasserheit JN, Barnabas RV, et al. Persisting with prevention: The importance of adherence for HIV prevention. Emerg Themes Epidemiol 2008; 5:8.
6. Mosher WD, Chandra A, Jones J. Sexual behavior and selected health measures: Men and women 15–44 years of age, United States, 2002. Adv Data 2005: 1–55.
7. Misegades L, Page-Shafer K, Halperin D, et al. Anal intercourse among young low-income women in California: An overlooked risk factor for HIV? AIDS 2001; 15:534–535.
8. Gorbach PM, Manhart LE, Hess KL, et al. Anal intercourse among young heterosexuals in three sexually transmitted disease clinics in the United States. Sex Transm Dis 2009; 36:193–198.
9. Karim SS, Ramjee G. Anal sex and HIV transmission in women. Am J Public Health 1998; 88:1265–1266.
10. Lane T, Pettifor A, Pascoe S, et al. Heterosexual anal intercourse increases risk of HIV infection among young South African men. AIDS 2006; 20:123–125.
11. Schwandt M, Morris C, Ferguson A, et al. Anal and dry sex in commercial sex work, and relation to risk for sexually transmitted infections and HIV in Meru, Kenya. Sex Transm Infect 2006; 82:392–396.
12. Kalichman SC, Simbayi LC, Cain D, et al. Heterosexual anal intercourse among community and clinical settings in Cape Town, South Africa. Sex Transm Infect 2009; 85:411–415.
13. Vittinghoff E, Douglas J, Judson F, et al. Per-contact risk of human immunodeficiency virus transmission between male sexual partners. Am J Epidemiol 1999; 150:306–311.
14. Leynaert B, Downs AM, De VI; European Study Group on Heterosexual Transmission of HIV. Heterosexual transmission of human immunodeficiency virus: Variability of infectivity throughout the course of infection. Am J Epidemiol 1998; 148:88–96.
15. Gray RH, Wawer MJ, Brookmeyer R, et al. Probability of HIV-1 transmission per coital act in monogamous, heterosexual, HIV-1-discordant couples in Rakai, Uganda. Lancet 2001; 357:1149–1153.
16. Padian NS, Shiboski SC, Glass SO, et al. Heterosexual transmission of human immunodeficiency virus (HIV) in northern California: Results from a ten-year study. Am J Epidemiol 1997; 146:350–357.
17. Renzi C, Douglas JM Jr, Foster M, et al. Herpes simplex virus type 2 infection as a risk factor for human immunodeficiency virus acquisition in men who have sex with men. J Infect Dis 2003; 187:19–25.
18. Roddy RE, Zekeng L, Ryan KA, et al. A controlled trial of nonoxynol 9 film to reduce male-to-female transmission of sexually transmitted diseases. N Engl J Med 1998; 339:504–510.
19. Van Damme L, Ramjee G, Alary M, et al. Effectiveness of COL-1492, a nonoxynol-9 vaginal gel, on HIV-1 transmission in female sex workers: A randomised controlled trial. Lancet 2002; 360:971–977.
20. Peterson L, Nanda K, Opoku BK, et al. SAVVY (C31G) gel for prevention of HIV infection in women: A phase 3, double-blind, randomized, placebo-controlled trial in Ghana. PLoS One 2007; 2:e1312.
21. Feldblum PJ, Adeiga A, Bakare R, et al. SAVVY vaginal gel (C31G) for prevention of HIV infection: A randomized controlled trial in Nigeria. PLoS One 2008; 3:e1474.
22. Halpern V, Ogunsola F, Obunge O, et al. Effectiveness of cellulose sulfate vaginal gel for the prevention of HIV infection: Results of a phase III trial in Nigeria. PLoS One 2008; 3:e3784.
23. Van Damme L, Govinden R, Mirembe FM, et al. Lack of effectiveness of cellulose sulfate gel for the prevention of vaginal HIV transmission. N Engl J Med 2008; 359:463–472.
24. Skoler-Karpoff S, Ramjee G, Ahmed K, et al. Efficacy of Carraguard for prevention of HIV infection in women in South Africa: A randomised, double-blind, placebo-controlled trial. Lancet 2008; 372:1977–1987.
25. Fleming TR. Evaluation of active control trials in AIDS. J Acquir Immune Defic Syndr 1990; 3(suppl 2):S82–S87.
26. Kurth AE, Martin DP, Golden MR, et al. A comparison between audio computer-assisted self-interviews and clinician interviews for obtaining the sexual history. Sex Transm Dis 2004; 31:719–726.
27. Raymond EG, Taylor D, Cates W Jr, et al. Pregnancy in effectiveness trials of HIV prevention agents. Sex Transm Dis 2007; 34:1035–1039.
28. Gross M, Holte SE, Marmor M, et al; The HIVNET Vaccine Preparedness Study 2 Protocol Team. Anal sex among HIV-seronegative women at high risk of HIV exposure. J Acquir Immune Defic Syndr 2000; 24:393–398.
29. Institute of Medicine. Methodological Challenges in Biomedical HIV Prevention Trials. Washington, DC: National Academies Press, 2009.
© Copyright 2010 American Sexually Transmitted Diseases Association
30. McGowan I. Rectal microbicides: A new focus for HIV prevention. Sex Transm Infect 2008; 84:413–417.