Share this article on:

Using Objective Markers to Assess Participant Behavior in HIV Prevention Trials of Vaginal Microbicides

Mauck, Christine K MD, MPH*; Straten, Ariane van der PhD, MPH†‡

JAIDS Journal of Acquired Immune Deficiency Syndromes: September 1st, 2008 - Volume 49 - Issue 1 - p 64-69
doi: 10.1097/QAI.0b013e318183a917
Critical Review: Clinical Science

The need to verify participant behavior exists in any study in which behavior may affect outcomes. In vaginal microbicide trials, the act of having sex and the use of study products and condoms all affect the risk of acquiring HIV/sexually transmitted infections (STIs). Until now, these behaviors have been assessed using self-reports. But self-reports are limited by participant cooperation in answering questions, imperfect recall, and social desirability biases. Biomarkers are increasingly being used in medicine to reduce the time and resources needed to bring a drug to market. The use of biomarkers in vaginal microbicide trials has been proposed as a means of assessing factors that affect the risk of sexual acquisition of HIV/STIs, namely, the presence of preexisting infection, cervicovaginal inflammation, and the presence of HIV/STIs. Biomarkers for some of these already exist. What are needed are validated markers of behaviors that might affect risk, namely, markers for sexual behavior and for the use of study products and condoms. Validating and working out the logistics of collecting such markers in large trials will be a challenge. But finding objective markers for behavior may help improve adherence measurement during a trial and is a rate-limiting step in the field of vaginal microbicides. Resources and funding should be mobilized to develop and validate markers of sexual behavior and product use as a high priority in vaginal microbicide research.

From the *CONRAD, Eastern Virginia Medical School, Arlington, VA 22209; †Center for AIDS Prevention Studies, Department of Medicine, University of California, San Francisco, CA; and ‡Women's Global Health Imperative, RTI International, San Francisco, CA

Received for publication November 2, 2007; accepted May 22, 2008.

Some of this material was presented at a conference entitled “Adherence and its Measurement in Microbicide Clinical Trials” sponsored jointly by the Alliance for Microbicide Development and Family Health International on December 18-19, 2007.

As a literature review, this article does not have a source of funding

Correspondence to: Christine K. Mauck, MD, CONRAD, Eastern Virginia Medical School, 1611 North Kent Street, Suite #806, Arlington, VA 22209 (e-mail:

Back to Top | Article Outline


The need to verify participant behavior exists in any study in which behavior may affect outcomes. However, certain behaviors, particularly those involving sex, are private and thus impossible to observe in a clinical trial setting. But knowing whether the participants were exposed to risk and whether they used the product designed to mitigate that risk is essential to correctly interpret study results.

This is the quandary in which the field of vaginal microbicide research finds itself. Vaginal microbicides are topical products designed to prevent male to female sexual acquisition of HIV and other sexually transmitted infections (STIs). Finding ways to measure sexual behavior and product use is a “priority action” in vaginal microbicide development.1 Assessing participant adherence to instructions regarding sexual behavior and product use, validating measures of self-reported behavior, and interpreting trial results when objective behavior measures are not available are all enormous challenges.2 Similar issues apply to studies of other female-initiated methods such as diaphragms that physically cover the cervix and may serve as microbicide delivery systems and to rectal microbicides used by men and women, although the latter are not specifically addressed here.

A recently completed trial illustrates the importance of being able to accurately assess study participant behavior. The Methods for Improving Reproductive Health in Africa trial did not show a difference in HIV acquisition between a group of women randomized to use condoms alone and a group randomized to use condoms plus a diaphragm with a lubricant gel.3 However, the rate of self-reported condom use during the last sex act was different in the 2 groups with an average of 85.1% in the condom-only group and 53.3% in the condom + diaphragm + gel group. It was speculated that “diaphragm use might have compensated for the difference in condom use,” but that “uncertainty about self-reported condom use makes other explanations equally plausible.” The study report went on to say that, “… our study emphasizes the challenges of relying on self-reported data for sexual behavior, and the need to improve measurements and develop objective measures of sexual activity, unprotected sex, and product use.”

Until now, sexual behavior and product use in clinical trials have been assessed by self-report. The studies most analogous to HIV prevention studies are those of barrier contraceptives. They yield pregnancy rates in “typical use” and “perfect use,” the distinction most commonly being based on behavior self-reported on paper diaries. But paper diaries are burdensome to participants, study staff, and data managers and are often filled out retrospectively. Electronic diaries may be easier to use but are expensive and still require daily input from participants. Diaries are sometimes replaced by face-to-face (FTF) questioning at follow-up visits or at the end of the study or by audio computer-assisted self-interviewing (ACASI), a computerized system of interviewing participants. Although the latter removes the human questioner and may thereby decrease self-presentation bias,4-6 any means of self-reporting behavior may be hampered by imperfect recall, misinterpretation of questions, and social desirability issues such as reluctance on the part of the participant to report lack of compliance and a desire to conform to group norms.7,8

Biomarkers are increasingly being used in medicine to indicate the presence of risk factors for a disease, of the disease itself, and of response to treatment. Biomarkers can give guidance early in drug development as to which therapies are likely to work, how, and in whom, thereby reducing the ever-escalating cost of bringing a new drug to market.

Vaginal microbicides are designed to prevent disease rather than treat it. The use of biomarkers in vaginal microbicide trials has thus been proposed as a means of assessing factors that affect the risk of sexual acquisition of HIV/STIs and determining the actual presence of HIV/STIs (but not assessing response to treatment). Specifically, biomarkers could be used to assess the presence of preexisting infection that increases risk (eg, genital herpes simplex), cervicovaginal inflammation that increases risk and may also indicate toxicity from vaginal microbicide use, and the actual final presence of HIV/STIs. Biomarkers for some of these already exist.

What are missing from the list above are validated markers of sexual behavior and product/condom use that might affect risk. Indicators of behavior are not, strictly speaking, biomarkers, but we will describe objective ways to measure some of them. Such objective markers could be used to reduce the reliance on self-reporting, to help choose the most accurate type of self-report among those we continue to use, and to increase participants' motivation to give accurate reports. This article is an attempt to provide a framework for thinking about markers for sexual behavior and product use in vaginal microbicide clinical trials, describing both the few we have and the many we need, along with the considerable challenges we will face in developing them.

Back to Top | Article Outline


The following markers might be useful in assessing behavior in vaginal microbicide trials: (1) markers of behavior that increase risk (exposure to semen) and (2) markers of behavior that reduce risk, which may be divided into study product use (vaginal microbicides and their delivery systems) and condom use.

The results that may be expected when using markers under various behavioral conditions in a vaginal microbicide trial are shown in Table 1. It must be recognized that only a few of these markers actually exist, and much work must be done to identify, develop, and validate most of them.



Back to Top | Article Outline

Markers of Risk: Semen Exposure

Markers of semen exposure have been extensively reviewed elsewhere.9 In brief, they involve taking a vaginal fluid sample and testing it for a marker of either seminal fluid or sperm cells. Prostate-specific antigen (PSA) is the best characterized marker of seminal fluid, having been studied for the past 30 years, and has a vaginal residence time of up to about 48 hours.10 Semenogelin is another seminal fluid marker, for which a new test has recently become available (RSID-Semen, Independent Forensics, Lombard, IL).11 It is detectable in vaginal swabs obtained up to 3 days after intercourse. The detection of sperm is used in certain types of trials, but it is cumbersome and subjective.9 Y-chromosome DNA is another marker of sperm cells that is detectable for weeks but is not specific for sperm, and concerns have been raised that results may be positive if any source of the male partner's DNA enters the vagina (eg, through oral or digital sex). The presence of a marker for semen exposure can indicate unprotected sex, incorrect condom use, or sex with an imperfect condom, but the absence of a semen marker cannot distinguish between the absence of sex and the occurrence of effective condom use (hence a reason to develop a marker for condom use). Recent work suggests that the presence of some vaginal products may interfere with the detection of semen markers (Johan Melendez, MS and Laurie Howard, MPH, personal communication 2007). This type of interference should be explored and quantified before these (and any other) markers are relied upon.

Back to Top | Article Outline

Markers of Potential Protection: Study Products (Vaginal Microbicides and Their Delivery Systems)

Assays for study products in the female genital tract must take into account 3 factors: the route of application, the timing of application, and the anatomical location in which the product must end up to be effective. Although this article focuses on vaginally applied products, orally administered antiretroviral drugs are being studied for their ability to prevent vaginally acquired HIV (preexposure prophylaxis) and levels in vaginal fluids and tissue after oral administration are being studied. Preexposure prophylaxis is generally administered on a daily basis without regard to coitus. Most vaginal products to date have been administered within the hour before coitus because they have their activity in the vaginal fluids (eg, BufferGel). Others must undergo absorption and intracellular metabolism to become effective (eg, the antiretroviral drug tenofovir) but may remain effective long after application. This could allow application time more remote from coitus as in the Centre for the AIDS Programme of Research in South Africa trial, in which participants are instructed to insert one dose of tenofovir gel up to 12 hours before each act of vaginal intercourse and a second dose as soon as possible after coitus but within 12 hours. It could even allow daily vaginal application not linked to coitus.

Markers of microbicide in the vaginal fluid and tissues are not as well studied as markers of semen exposure but do exist. The Population Council assayed Carraguard levels in vaginal fluids 1, 2, 3, and 24 hours after a single dose. CONRAD is assaying tenofovir levels via vaginal aspirate, cervical brush samples, and vaginal biopsy after a single dose and after 14 days of once- and twice-daily dosing. Others are assaying UC-781 levels immediately after a single dose, and at 2, 3, or 8 hours, and 30 days later via cervicovaginal lavage, cells from vaginal swabs, cervical brush, and spatula sampling (S. Hillier, PhD, personal communication 2007). These tests provide essential pharmacokinetic data on these products and could, with certain caveats described below, serve as a marker of product use in an HIV prevention trial; however, some are relatively invasive and may not be practical to use in large-scale trials.

Counting used gel applicators is sometimes used as a proxy for assessing product use. In response to reports of microbicides being expelled from applicators (but not into the vagina) before those applicators being counted, tests have been developed to determine whether a microbicide applicator came into contact with the vaginal mucosa. The Population Council developed a test for mucin that was used in its phase 3 HIV prevention trial of Carraguard12,13 and in a UC-781 study and the VivaGel study carried out by the Microbicide Trials Network. The test involves spraying a dye on all used applicators and then rinsing. If mucin is present, the dye “sticks” and is visible as a color change. The test cannot indicate, however, which mucosal surface was touched by the applicator and whether the gel was actually expelled into the vagina. Furthermore, it may not work with all types of applicators.

XiGo Medical Devices (Morganville, NJ), in collaboration with the International Partnership for Microbicides, is developing a second-generation applicator designed to address these issues. This “smart applicator” will record the date and time that the piston in the device was depressed into the barrel and the presence of mucin. Temperature sensors and conductivity measurements between the plunger tip and the participant's body may make it possible to determine whether the dose was dispensed while the device was inside the body. Clinical testing of the applicator should begin in 2008 (Sean Race, MBA, personal communication 2007). A similar technology assessing temperature and pH variations could also be applied to the development of a “smart” vaginal ring and has been discussed for the creation of a “smart” diaphragm (Kelly Blanchard, MSc, personal communication 2007).14

Other new technologies are being developed to monitor drug treatment adherence that could potentially be adapted to assess use of vaginal products. For example, a novel, exhaled breath-based individual medication adherence monitoring system has been recently developed by Xhale Inc. (Gainesville, FL- It is conceivable that a vaginal product could be tagged with an innocuous marker which is then systemically absorbed and detected using this system. This approach would provide a noninvasive way to monitor microbicide use.

Back to Top | Article Outline

Markers of Protection Other than Study Products: Condoms

The importance of a condom marker is illustrated by the Methods for Improving Reproductive Health in Africa trial results. Active promotion and provision of male condoms is an ethical prerequisite of all prevention trials but may “mask” the effect of the test product, so it is critical to have an objective way to measure condom use. Counting used condoms is sometimes used for this. Besides being unappealing, the accuracy of this practice depends on the participant's cooperation in returning used condoms to the clinic. A marker for condom use does not currently exist, although there have been efforts to develop a marker for latex. It may be possible to test for the presence of a component of the condom or even to add a marker to the lubricant used on condoms in a clinical trial (Richard Melker, MD, PhD, personal communication 2007).

Back to Top | Article Outline


In a perfect world, the sequence of events (ie, application of microbicide, a coital act, and sample collection) would take place as shown in Figure 1. We could then look at Table 1 and determine exactly what took place during that act because the pattern of markers that would be expected under each behavioral condition is unique. However, we would be making a number of assumptions that would have to be recognized, with their associated caveats as described below.



Assumption #1: Only one sex act (or sex acts of only one kind) occurred during the inter-test interval.

Caveat: If this happened and we had valid markers, we could draw conclusions about what kind of behavior the participant engaged in during that interval. We would not know exactly when the behavior occurred or how many times. However, if more than one behavior occurred during the inter-test interval, more than one pattern of markers would be present and our ability to distinguish between behaviors would be lost. What could remain useful in this situation is not the presence of markers but rather their absence. For example, if there were no marker for microbicide, applicator, or diaphragm, we could conclude that the study product was never used during inter-test interval. If there were no condom marker but there was a semen marker, we could conclude that unprotected sex took place at least once.

Assumption #2: The residence times of the markers being studied are equal to or longer than the inter-test interval.

Caveats: In the real world of phase 3 vaginal microbicide trials, participants are seen monthly at most. Specimens could be obtained at these visits for markers that, ideally, would have residence times of 1 month. However, the residence time of PSA, for example, is about 48 hours. It is likely that the residence times of most markers will be relatively short. To obtain marker information about the whole interval, collections would have to be done at a frequency that would not be feasible for in-clinic collection. Participants could self-swab at home, either after each sex act or at other predetermined times, but self-swabbing is a behavior in itself and raises many of the same compliance issues that are associated with product use or diary use.

Assumption #3: The residence time is the same as the “effectiveness time” or “risk time.”

Caveat: Figure 1 shows a hypothetical pharmacokinetic curve of local drug concentration over time after application. It is possible that, by the time a sample is collected, the drug level has gone below whatever level is required for effectiveness but is still in the detectable range. A qualitative (yes/no) test may lead to the mistaken assumption that the entire period of time since drug application was protected. Similarly, it is possible that markers for semen (ie, risk) may persist long after the risk of infection has passed.

Assumption #4: The time of drug or semen application can be known.

Caveat: It may seem logical that if the quantitative decay curve of the marker is known, then the time of drug or semen application can be inferred from the level of marker that is seen at the time of specimen collection. However, drugs that accumulate with repeated application may give a higher level than that expected from a single application. Whether or not semen markers accumulate has not been studied. And both drug and semen markers are subject to all the factors that may speed up their removal from the vagina (see below).

Assumption #5: Once deposited in the vagina, a marker will follow the decay curve established in the validity trials.

Caveat: In real life and in effectiveness trials, many things happen that may accelerate removal of a marker from the vagina: menses, use of tampons or other intravaginal products, discharge, and sex. The collection of tissue samples or other nonvaginal samples, when relevant, instead of vaginal fluid samples gets around this but may be invasive and thus not feasible for large trials.

Assumption #6: Application of the microbicide occurs before sex.

Caveat: Application may, in fact, occur after sex. This may happen in an attempt to gain protection “postexposure” or may result from an unprotected act followed by drug application in anticipation of second act that does not materialize. In either case, it cannot be assumed, just because a marker for drug and a marker for semen are present, that the act was protected.

Assumption #7: The risk of HIV acquisition is present during the sex act in question and can be known.

Caveat: Adherence to study product use is only relevant when considering coital acts that pose a risk, that is, sex with an infected partner. But the HIV infection status of partners in HIV prevention trials is generally unknown. Data from someone who is adherent during all acts but never has sex with an infected partner are not informative when assessing the effectiveness of a product. Some participants may only use the product with partners who they perceive pose a risk. Such a participant may be misclassified as poorly adherent based on the percent of acts that were protected, when her adherence during acts that “matter” was actually high. Failure of the microbicide to protect in such a case may be mistakenly attributed to poor adherence rather than ineffectiveness of the drug. The “level of risk” of any given act is unknown, except perhaps in HIV-discordant couple studies. Such studies are not without precedent, but the risk per coital act among couples who have managed to remain discordant may be relatively low.15

Assumption #8: It is possible to determine which act resulted in infection.

Caveat: This would require that each participant engage in only a single coital act during the inter-test interval and that there is a test that would indicate HIV acquisition almost immediately after the fact. Obviously, this is not the case: relatively frequent sex is required to complete trials within a reasonable time frame and the width of the seroconversion window for the commonly used HIV tests is wide and varies among individuals.

Assumption #9: Risk results only from vaginal sex with an infected partner.

Caveat: Risk can come also from anal sex and sharing needles, both of which may not be reliably reported in trials.

This list of caveats is long and somewhat discouraging. However, recognizing them ahead of time will allow us to move forward in an informed manner and to develop markers that give us as much relevant information as possible.

Back to Top | Article Outline


We have been considering the use of markers as an objective way to assess behavior that, up to now, has been self-reported. However, to consider a marker a gold standard, it must be validated or shown to accurately correspond with the behavior it is intended to indicate. An example of how this might be done comes from the field of markers for semen exposure. A few trials have been done in which a woman's vagina was inoculated with a known volume of her partner's semen in the clinic and then vaginal samples were taken at various time points afterward. In one study, all swabs taken immediately after inoculation were considered positive for PSA,16 and in another, 98% of samples were positive17: these values can be interpreted as sensitivity. By 24 hours in the second study, only 29% of samples were positive and by 48 hours, 3% were. If we assume that there should be no PSA present at 48 hours, this would imply a specificity of 97%.17 These values are well within the acceptable range for a biomarker. Additional studies of this type are being undertaken for PSA and should be carried out for any other marker of behavior.

Back to Top | Article Outline


As an alternative to replacing behavioral tools with markers, one could use markers to compare and select the most accurate among several self-reported behavioral measurement tools, should we continue to use them. During the Methods for Improving Reproductive Health in Africa trial, ACASI was used to collect data on sexual behavior and product use. After the trial, a subset of participants was randomized either to one additional ACASI session or to FTF interviewing. Discrepancies by mode of reporting (ACASI vs. FTF) will be monitored, using PSA as a “gold standard.” Analysis of these data is in progress.

Another example of using markers to compare behavioral assessments is a Population Council study that enrolled 800 women for 3 months each. ACASI and FTF will be compared with the applicator test and a semen marker (RSID-Semen). The goal is to see whether less reporting of consistent microbicide use and more accurate reporting of sensitive sexual behaviors (eg, unprotected sex) are observed with ACASI compared with FTF. The study is nearly complete with analysis expected in the fall of 2008.

Back to Top | Article Outline


When deciding whether to use markers in a trial, researchers must determine whether their goal is to understand participants' behavior, influence it, or both. The influence of knowing that behavior is being monitored on a participant's behavior is not at all clear. One study in which respondents were led to believe that inaccurate answers could be detected by a physiological recording device (eg, a polygraph) or a test performed on a body fluid resulted in more accurate reporting of sensitive behaviors.18 However, in a recent study conducted in Africa, more women who knew they would undergo vaginal PSA testing had discrepancies between their self-reported behavior and their PSA results than did women without such advance knowledge.19 In most studies, it would be considered unethical to use behavior markers without informing study participants. But it must be recognized that the influence of such knowledge on participant behavior is poorly understood. It is possible that using adherence data to identify nonadherent participants who should receive additional counseling may improve the quality of the trial.

Back to Top | Article Outline


If one has to choose, for practical reasons, among the different kinds of markers, a vaginal microbicide marker gives more directly valuable information than a marker for applicator use. Similarly, a marker for semen is more valuable than a marker for condom use because exposure to semen, whether in the context of no condom, incorrect use, or a broken condom, is what constitutes risk. If the semen marker is negative, we can assume there was no risk of HIV infection, at least during that act or during the marker's residence time. If it is positive and a microbicide marker is also positive, and the woman did not seroconvert, it cannot be determined with certainty that the vaginal product/microbicide was protective even if the male partner was HIV positive because of the low probability of HIV transmission per coital act. However, if the semen marker is positive and the study product markers are negative and the woman seroconverted, we may be able to avoid attributing “failure” to the study product.

Back to Top | Article Outline


In summary, using markers to assess behavior in HIV prevention trials is different from using them in trials of therapeutic modalities. In trials of vaginal microbicides, we are focused on risk reduction, not treatment, and risk is largely the result of unobservable behaviors. Being able to quantify those behaviors may allow us to more accurately interpret the results of prevention trials. Whereas some work has been done in the area of markers for semen exposure, product use, and delivery systems use, much work remains to refine these and to develop a marker for condom use. Understanding residence time, determining what factors may interfere with assays, validating markers against known standards, and working out the logistics of sample collection in the context of a prevention trial are huge challenges. But finding objective markers for behaviors may help improve adherence measurement during a trial and is a rate-limiting step in the field of vaginal microbicide research. Resources and funding for behavior marker development should be mobilized as a high priority in vaginal microbicide research.

Back to Top | Article Outline


1. MDS Working Groups. The Microbicide Development Strategy. Silver Spring, MD: Alliance for Microbicide Development; 2006.
2. Boyce L, Plescia C, Finley B, et al. The microbicide field response to clinical trial challenges. Poster presented at: International AIDS Society meeting, Sydney, Australia, 2007. Available at: Accessed May 16, 2008.
3. Padian NS, van der Straten A, Ramjee G, et al. Diaphragm and lubricant gel for prevention of HIV acquisition in southern African women: A randomized controlled trial. Lancet. 2007. July 21, 2007;370:251-261.
4. Mensch B, Hewett P, Gregory R. Sexual Behavior and STI/HIV Status Among Adolescents in Rural Malawi: An Evaluation of the Effect of Interview Mode on Reporting. Los Angeles, CA: Population Association of America Annual Meeting; 2006.
5. Minnis AM, Muchini A, Shiboski S, et al. Audio computer-assisted self-interviewing in reproductive health research: reliability assessment among women in Harare, Zimbabwe. Contraception. 2007;75:59-65.
6. Turner C, Ku L, Rogers S. Adolescent sexual behavior, drug use, and violence: increased reporting with computer survey technology. Science. 1998;280:867-873.
7. Clark J, Tifft L. Polygraph and interview validation on self-reported deviant behavior. Am Sociol Rev. 1966;31:516-523.
8. Catania JA, Binson D, van der Straten A, et al. Methodological research on sexual behavior in the AIDS era. In: Rosen RC, Davis CM, Ruppel HJ Jr, eds. Annual Review of Sex Research: An Integrative and Interdisciplinary Review. A Publication of the Society for the Scientific Study of Sexuality; 1995;VI:77-125.
9. Mauck CK, Doncel GF; Biomarkers of Semen Exposure Clinical Working Group. Biomarkers of semen in the vagina: applications in clinical trials of contraception and prevention of sexually transmitted pathogens including HIV. Contraception. 2007;75:407-419. Epub April 30, 2007.
10. Macaluso M, Lawson L, Akers R, et al. Prostate-specific antigen in vaginal fluid as a biologic biomarker of condom failure. Contraception. 1999;59:195-201.
11. Pang BCM, Cheung BKK. Identification of human semenogelin in membrane strip test as an alternative method for the detection of semen. Forensic Sci Int. 2006, doi:10.1016/j.forsciint.2006.07.021.
12. Wallace A, Thorn M, Maguire RA, et al. Assay for establishing whether microbicide applicators have been exposed to the vagina. Sex Transm Dis. 2004;31:465-468.
13. Wallace AR, Teitelbaum A, Wan L, et al. Determining the feasibility of utilizing the microbicide applicator compliance assay for use in clinical trials. Contraception. 2007;76:53-56.
14. Ibis Reproductive Health Organization Report 2002-2005, Page 6. Available at: Accessed May 16, 2008.
15. Wawer MJ, Gray RH, Sewankambo NK, et al. Rates of HIV-1 transmission per coital act, by stage of HIV-1 infection, in Rakai, Uganda. J Infect Dis. 2005;191:1403-1409. Epub March 30, 2005.
16. Lawson ML, Macaluso M, Bloom A, et al. Objective biomarkers of condom failure. Sex Transm Dis. 1998;25:427-432.
17. Macaluso M, Lawson L, Akers R, et al. Prostate-specific antigen in vaginal fluid as a biologic biomarker of condom failure. Contraception. 1999;59:195-201.
18. Tourangeau R, Smith TW, Rasinski KA. Motivation to report sensitive behaviors on surveys: evidence from a bogus pipeline experiment. J Appl Soc Psychol. 1997;27:209-222.
19. Gallo MF, Behets FM, Steiner MJ, et al. Validity of self-reported ‘safe sex’ among female sex workers in Mombasa, Kenya-PSA analysis. Int J STD AIDS. 2007;18:33-38.

biomarkers; microbicides; adherence; behavior; condoms; clinical trials

© 2008 Lippincott Williams & Wilkins, Inc.