Some history

The first product tested as a potential microbicide was nonoxynol-9 (N-9), a spermicidal surfactant that works by disrupting the sperm’s outer membrane. N-9 has been an active ingredient in over-the counter contraceptives for the last 50 years. In 1985, it was shown to destroy HIV by disrupting the viral coat. This evidence triggered 13 different trials examining various N-9 formulations for their potential applicability against HIV. Since the products tested varied widely in dosage, formulation and frequency of use, this research produced mixed results. Evidence of harm was noted in some trials in the form of vulvar ulcers, vulvitis and irritative symptoms.

Given the urgency of the microbicide demand, UNAIDS mounted a large, randomised, placebo-controlled trial of a low-dose N-9 vaginal gel in 1996 to obtain a definitive answer regarding its effectiveness. Advantage S, a spermicide produced by Columbia Laboratories, was tested by 892 female sex workers in four countries: Benin, Ivory Coast, South Africa and Thailand. All participants were supplied with condoms and encouraged to use them, as well as enhanced access to diagnosis and treatment of STIs (resources provided to all participants in subsequent microbicide trials as well).

The trial, called COL-1492, produced results in 2000. Among less frequent gel users, the HIV-seroconversion rates were similar between the N-9 and placebo arms of the trial. The number of seroconversions among the more frequent users, however, was twice the rate experienced by placebo users. N-9 appeared to have no effect on rates of gonorrhoea or chlamydia infection, although it exacerbated HIV risk when used frequently. Research into the effectiveness of nonoxynol-9 as a microbicide was abandoned immediately following the COL-1492 trial.

The second surfactant candidate to be tested was called C31G, or Savvy. Family Health International (FHI) mounted two large-scale efficacy trials of C31G between 2004 and 2006 - one in Ghana and a second in Nigeria. Both trials used a 1% C31G gel because of existing evidence that higher doses of C31G might cause vaginal irritation. CONRAD researchers had demonstrated that 0.5% and 1.0% C31G gels caused significantly less vaginal irritation than either 1.7% C31G gel or Extra Strength Gynol II (a nonoxynol-9 product).²

Both were stopped before completion due to futility. During interim analyses, their Data Safety and Monitoring Boards (DSMBs) observed that the rate of new infections was lower than anticipated. When used in connection with HIV-prevention trials, ‘futility’ means that HIV incidence in the study population is too low to demonstrate whether or not the candidate compound had an effect.

Several factors may have led to this low incidence rate among trial participants. One is the phenomenon (since observed in a number of other trials) that HIV incidence can fall dramatically amongst previously high-risk people when they join a clinical trial, and are then provided with standard-of-care prevention resources.

Another factor possibly contributing to futility in this case was the high rate of pregnancy (about four times as common as HIV seroconversion in this study). This decreased the likelihood of the study reaching a result, because many of the women most likely to become infected simply became pregnant first, and then they were ineligible to continue use of the test product. Most studies of microbicide-efficacy trials since then have included contraception within their design to reduce drop-out rates due to pregnancy.

Although the differences in seroconversion rates occurring in the test product versus the placebo arms of these trials versus control arms did not reach the level of statistical significance, they did raise some concerns. (For more information on what is meant by 'statistical significance' and how DSMBs work, please see the Global Campaign for Microbicides factsheets entitled Background Information for Advocates on the Premature Closure of Cellulose Sulfate Trials at www.global-campaign.org/clientfiles/GCM-Q&A-CSTrial-5Feb2007.pdf and Advocates’ Frequently Asked Questions About HPTN 035 at http://www.global-campaign.org/HPTN-035.htm.)

In the Nigeria trial, more seroconversions occurred among women assigned to use C31G than among those using the placebo (21 versus 12, out of a total of 2153 women enrolled), indicating a possible trend toward harm. In Ghana, more ‘adverse events’ (problems that may or may not have been related to the test product) were reported by the women assigned to use C31G. These, together with doubt that the trials could demonstrate anything conclusively, contributed to the decision for early closures.

  The main challenge in developing surfactants is to find products that disrupt the virus’s outer coat without also damaging healthy cells. While surfactants have the advantage of being contraceptive, and possibly having activity against other STIs, in addition to HIV, this approach has largely been abandoned because of the difficulty of striking an effective balance between safety and risk.

Vaginal-defence enhancers are designed to build upon the body’s naturally occurring defences against pathogens. These include optimal pH levels and the presence of naturally occurring microorganisms, such as lactobacilli, in the vagina.

BufferGel was a candidate microbicide developed around the fact that vaginal fluids are naturally acid, seminal fluids are alkaline, and both HIV and sperm survive best in pH-neutral environments (around 4.5 to 5). Upon entering the vagina, semen raises the vaginal Ph temporarily to facilitate sperm survival. BufferGel was designed simply to keep the vaginal pH low, even in the presence of seminal fluid, to minimize HIV survival. While BufferGel was proven safe, the HPTN 035 trial results of 2009 indicated that it had no significant effectiveness. Research on its potential effectiveness as a contraceptive, however, is continuing.

Acidform (or Amphora) is another candidate using this mechanism of action. Acidform gel includes a bio-adhesive agent designed to enhance its sustained presence (and, hence, effect) in the vagina. While data indicate that it is safe for vaginal use and may have contraceptive and/or broad-spectrum activity against some STIs, it has not inspired sufficient support to move into largescale trials to test its effectiveness against HIV.

Lemon and lime juice (not used in an efficacy trial, but discussed here because of significant media interest) are also known to inactivate both sperm and HIV. Obviously, the idea that cheap and almost universally available options, such as fruit juices, could prevent HIV is very attractive. Professor Roger Short of the University of Melbourne presented a poster at the 2004 International AIDS Conference showing that a solution containing 20% lemon or lime juice was found to inactivate 90% of HIV reverse-transcriptase activity within two minutes. He and his colleagues surveyed 200 sex workers in Nigeria, 163 of whom said they rinsed with lemon or lime juice before or after sex to prevent pregnancy and infections.^{3 4}

Short and his colleagues mounted a strong advocacy campaign in support of this practice, urging researchers to test its impact. When conducted, the research produced the following:

• The Southern Research Institute (US) tested lemon juice, lime juice and nonoxynol-9 (N-9) on cervical explant tissue (human cells obtained from routine hysterectomies and kept alive in lab cultures). They found that the cell damage caused by lemon and lime juice was similar to that caused by N-9 (see above). In addition, this study showed that the juice concentration (50% or more) needed to stop HIV was also the concentration that caused cell damage.⁵

• Laboratory research at CONRAD and at St George’s, University of London, showed that a 25% concentration of lime juice could also cause cell damage to penile-tissue samples and that even a 25% juice concentration could damage vaginal cells. In order to inactivate HIV in semen, one had to expose it to a 50% concentration of lime juice for 30 minutes.⁶

• Other researchers atCONRAD conducted a clinical trial in which sexually abstinent women were randomised to four trial arms and inserted various concentrations (100% lime juice, 50% diluted with water, 25% dilution, or plain water) twice daily for six consecutive days during two menstrual cycles. Some of the women in the 100% or 50% lime-juice arms got small, but serious, abrasions, in the walls of their vaginas. None of the women in the 25% juice or plain-water arms experienced this. Among the women using 100% lime juice with no water, more than 65% experienced genital irritation, 50% experienced deep epithelial abrasions and more than 70% reported experiencing pain.⁷

• A researcher at University of California produced data demonstrating that the practice could be relatively safe - at lower concentrations. In her trial, women used 10% and 20% concentrations of lime juice for 14 consecutive days without significant problems.⁸

In summary, The Economist magazine wrote, on 29 April 2006, that: “as a microbicide, lime juice is safe when it is ineffective and effective when it is unsafe.”⁹

Candidates in this category are based on very large, stable polymers and three of them – cellulose sulphate (CS), Carraguard and PRO 2000 – showed sufficient promise to move into large-scale effectiveness trials. These so-called ‘big, sticky molecules’ coat cell surfaces, thus preventing the binding of viruses or the entry of microbes into tissue. PRO 2000 is also a sulphonated polymer, whose sulphur-containing components have a great affinity for the receptor molecules on the surface of the cells that HIV most readily infects within the genital tract. Thus, it is capable of latching onto these receptors and blocking them from HIV attachment.

The Microbicide Development Programme (see above) had planned to conduct an effectiveness trial of dextrin-2-sulfate (Emmelle) – a fourth candidate in this category. They cancelled the plan, however, because the other three products in this class, already entering phase III clinical trials, had demonstrated greater potential for effectiveness in preclinical testing.

None of these entry inhibitors appeared to generate inflammatory problems and Phase I and II clinical trials showed each to be very well tolerated. Their molecular size also made them unlikely to be absorbed into the bloodstream. They had the added advantage of being relatively easy and cheap to make. Analysis of the failure of cellulose sulphate as a candidate microbicide, however, revealed an unexpected form of tissue toxicity, as described below.

Cellulose sulfate – big sticky candidate #1

Cellulose sulfate (Ushercell) looked like a promising candidate when it entered large-scale efficacy trials in 2004 and 2005. Though named for its Canadian developer, George Usher, it was more commonly referred to by the name of its active ingredient, a cotton-based compound that had shown significant activity against HIV and other STIs in vitro. 

In ten earlier human-safety studies and two contraceptive trials, cellulose sulfate (or CS) showed no evidence of harm. Women in the placebo arms of the CS safety trials experienced symptoms, laboratory results and incidence of vaginal infections very similar to those experienced by women in CS gel arms.

On the basis of these data, two Phase III trials to assess the effectiveness of CS as a microbicide were initiated. One, conducted in Nigeria by FHI, started enrolment in December 2004. The other, conducted by CONRAD in Benin, South Africa, Uganda and India, started in July 2005.

In January 2007, CONRAD stopped its trial when the Data and Safety Monitoring Board’s interim analysis showed that more HIV infections were occurring among women using CS than among those using the placebo (25 versus 16, out of a total of 1425 women enrolled). FHI’s study in Nigeria was halted at the same time as a precautionary measure, even though interim data from that study did not suggest an increased risk of HIV infection.

Closure of the CS trials was a bigger shock to the microbicide community than the closure of the C31G trial. As a surfactant, C31G was automatically suspected of being cytotoxic (killing cells), so that the trial’s failure was a disappointment, but not a surprise. No such expectation, however, existed with regard to CS.

In its interim analysis, the CONRAD trial’s difference in seroconversion numbers fell below the level of statistical significance, although it suggested a trend toward harm. Stopping the trials was a wise move given that further analysis eventually showed a total of 60 women in the CS arm (8.4% of participants receiving CS) seroconverted, in contrast to 27 women receiving the placebo gel (3.8% of all participants in that arm). This difference still did not quite reach statistical significance, but certainly underscored the possibility of a trend towards increased harm.

Hypotheses to explain the disparity include the possibility that CS induced inflammatory reactions, localised immune dysfunction and/or that disruption of the normal vaginal flora could have occurred with frequent CS use. At the Microbicides 2008 Conference, investigators reported that, at the trial’s Durban site, they observed a threefold greater-than-average risk of infection among trial participants diagnosed with chlamydia, bacterial vaginosis or trichomoniasis, and a greater than fourfold increase in HIV-infection risk among participants who reported more than one new partner between the quarterly trial visits.¹⁰

The link between CS use and the apparently increased vulnerability to HIV was finally illuminated by Pedro Mesquita of Albert Einstein College of Medicine in New York. Novel imaging techniques¹¹ allowed Mesquita’s team to see that CS was not cytotoxic (did not kill cells), but did cause loss of tissue integrity by destroying proteins that bind cells together. This allowed HIV to leak into the underlying tissues. The study found that this unexpected toxicity was also observed to a greater degree with nonoxynol-9, and to a very limited degree with the microbicide PRO 2000, though in this case the tissue damage appeared reversible. This effect causes neither macroscopic tissue damage that can be seen by colposcopy, nor a significant inflammatory immune response (though repeated application can cause inflammation). Thus, it would have been unobservable by the standard assays for microbicide toxicity.

Carraguard - big sticky candidate #2

After the evidence that C31G and cellulose sulphate might not only be ineffective, but actually increase risk, the results from Carraguard, the next large Phase III study, came as a relative relief to researchers and advocates alike. Although Carraguard did not protect women against HIV infection, it at least appeared to be safe. There was even a hint of efficacy at some trial sites.

Carraguard is manufactured from carrageenan, a seaweed derivative that is well known, cheap to produce and already used for a number of purposes, including as a commercial food additive. The Population Council (see above) became interested in it as a potential microbicide in the mid-1990s when data began to emerge¹² indicating that it effectively prevented mice from catching the genital herpes virus (HSV-2), an infection that is both highly contagious and lethal for mice. In vitro, it also blocked cells from becoming HIV infected.¹³ Although no animal study was done specifically demonstrating that it prevented HIV-like infection (such as infection with SIV, SHIV or FIV, for example), it was shown to protect mice from some other STIs.¹⁴ Numerous safety studies had come up with a clean bill of health regarding its safety for vaginal use.

In 1994, Carraguard became the first novel microbicide candidate to enter clinical trials (given that nonoxynol-9 was designed as a contraceptive, not a microbicide). A decade later, the Populations Council started enrolment in its Carraguard Phase III trial. Completed in 2008, this was the first large-scale efficacy trial of a new microbicide candidate to reach completion - a notable achievement given that 6202 women participated in three sites in the South Africa trial and each was followed for up to two years.

In total, 134 new infections occurred among women in the Carraguard arm of the trial and 151 in a placebo arm. The difference between the two groups was not statistically significant, nor was there any significant difference between the arms in the incidence of side-effects, such as vaginal irritation, loss to follow-up or withdrawal rates (participants choosing to leave the trial).

A crucial question in the Carraguard trial was the degree of adherence to gel use. Overall, participants claimed that they used the gel during 94% of their sex acts. But the researchers gathered additional adherence data by applying a dye to the returned applicators to detect the presence of vaginal mucous on them. This test showed that only 61% of the returned applicators had actually been used. This test was not included in the original study protocol but was added when it became apparent that adherence was a serious concern.

By combining such objective measurements with self-reported data, the researchers determined that the Carraguard gel had actually been used by women in the Carraguard arm of the trial during only 43% of sex acts. This effectively removed the statistical power of the trial to determine the low-to-moderate efficacy. With good adherence, the trial was powered to show effectiveness at the rate of 30% or greater. But only a product with substantially greater efficacy than that could have shown any measureable result when used less than half the time.

Low adherence was not the only reason the trial ended up having little power to generate a statistically meaningful result. Although the drop-out rate was only 13%, an additional 9% were lost to pregnancy, and the pregnant women did not return to the study after giving birth. Circumcision rates in men were also very different at the three trial sites, with 97% of men in the Cape Town area circumcised, 54% in Pretoria, but only 24% in Durban. At the Durban site, more seroconversions actually occurred amongst the participants using Carraguard than those using the placebo (48, or 3.3% of participants, versus 42, or 2.8%). In the other two sites, more seroconversions occurred among the placebo users. These data are not statistically significant, but may suggest a hint of efficacy for Carraguard when it comes to populations with lower HIV prevalence than the extremely high rate seen at Durban and/or higher rates of male circumcision.

Carraguard was popular: women said it improved sex and they liked the feel. Other gel studies have reported similar findings. In fact, participants at one site in the prematurely terminated cellulose sulphate trial refused to hand their gel back until they were given supplies of placebo.

Data on adherence in this trial must be considered in context. Reinforcement of the importance of regular microbicide use took place only at quarterly study visits and there was no supportive ongoing education in the community. One major challenge for the field now is devising innovative ways of addressing this issue. With closer adherence monitoring, it should be possible to do a ‘per protocol’ analysis that looks at product effectiveness only amongst actual product users, rather than the traditional ‘intention-to-treat’ analysis that factors in all trial participants equally, regardless of whether or not they completed the trial, were adherent to the protocol, etc. Researcher Elof Johansson commented that: “using an intention-to-treat analysis in people who do not feel sick is very difficult”, and noted that trials of blood-pressure drugs had run into the same problems in terms of inability to show statistically significant results.

PRO 2000 - big, sticky candidate #3

As noted above, PRO 2000 is a compound containing sulphonated polymers, large molecules whose sulphur-containing components are drawn to the receptor molecules on the cell surfaces that HIV tends to target. By latching onto these receptors before HIV exposure occurs, these molecules can block HIV attachment. Although similar to cellulose sulfate, PRO 2000 has been shown to have less toxicity.

PRO 2000 underwent two large-scale efficacy trials, the first of which provided a hopeful (although inconclusive) view of its potential and the second of which demonstrated that it was not effective against HIV.

The first of these, called HPTN 035, was started by the HIV Prevention Trials Network (HPTN) in January 2005 and completed in 2009 under the auspices of the Microbicide Trials Network (MTN, see above). The MTN was spun off from the HPTN in 2006 to serve as the NIH’s collaborative clinical-trials network, focused specifically on preventing the sexual transmission of HIV.

In HPTN 035, participants were randomly assigned to one of four study groups: BufferGel, PRO 2000 gel, placebo gel, and no gel. Women assigned to the three gel groups applied gel up to one hour before sexual intercourse using pre-filled applicators. Because all three gels and applicators look the same, neither researchers nor participants knew who has been assigned to use which gel for the duration of the study (although the no-gel, or ‘condom-only’ arm was necessarily obvious).

The no-gel arm was designed to test whether the neutral placebo gel was actually neutral. In 2000, the results of the COL-1492 study on nonoxynol-9 caused speculation as to whether Replens (a vaginal moisturizer used as the comparator, or placebo, gel in that study) might have reduced HIV risk slightly among the women in the control arm of the study, thereby skewing the results. This led to the development of hydroxyethyl cellulose (HEC), a gel that is now used as the ‘universal placebo’ in microbicide trials. The absence of significant differences between the experiences of participants in the placebo arm of HPTN 035 and those in the no-gel arm confirmed that HEC was neutral and had no discernable effect on HIV risk.

HPTN 035 was not designed to compare the effectiveness of BufferGel to that of PRO 2000, but simply to see if their joint efficacy was more than 43.6%. Between 2005 and 2007, the trial enrolled a total of 3100 participants at seven sites in Malawi, South Africa, Zambia, Zimbabwe, and the US. Each study participant was followed for at least one year.

As noted above, this trial showed clearly that BufferGel was ineffective. Its findings regarding PRO 2000, however, were ambiguous. Women in the PRO 2000 arm had 30% fewer HIV infections than those in the placebo or no-gel arms. While very encouraging, this result did not reach the level of statistical significance. Given the size of the trial and the seroconversion rate, there was a one-in-ten probability that the 30% reduction in HIV infections among PRO 2000, versus those who used a placebo, was due to chance (p = 0.1). In the case of women who used no gel, this probability was one in 17 (p = 0.06). Scientifically, results are not regarded as ‘significant’ unless the chance that they are wrong is less than one in 20 (p = 0.05).

The second PRO 2000 effectiveness trial was MDP 301, a study undertaken by the Microbicides Development Programme (MDP) of the UK Medical Research Council to compare two doses of PRO 2000 (2% and 0.5%) to placebo. With three times the number of participants (9673 enrolled over six sites in South Africa, Zambia, Tanzania and Uganda), researchers and advocates fully expected that MDP 301 would be able to demonstrate PRO 2000’s efficacy conclusively if it detected a similar reduction in the HIV-infection rate.  

The hopefulness around MDP 301 was slightly dampened by the trial’s January 2008 closure of its 2% trial arm. When launched in 2005, MDP 301 set out to test two concentrations of PRO 2000, a 2% gel and a 0.5% gel. In January 2008, MDP announced that the 2% trial arm was being closed due to futility (meaning that the study would not be able to demonstrate whether or not the candidate compound had an effect). While PRO 2000 appeared less toxic than some other candidates, any protective effect that the 2% gel had may have been offset by a local irritant effect caused by the higher dose.

The 0.5% gel arm continued, however, and claimed adherence to gel use was high. The rate of reported condom use varied widely among sites, ranging from 80% in South Africa to 17% in Zambia. In-depth examination of condom-use rates in one South African site, however, revealed that true level of consistent condom use at that site was about 25%, half the rate indicated by self-reporting.

The results of MDP 301, announced in December 2009, showed no evidence that PRO 2000 reduced HIV risk in trial participants. The factors potentially contributing to this surprising result are still under examination. In addition to concern about adherence, anal sex may have been another confounding factor. At their monthly clinic visits, only 1% of women reported having had anal sex in the last month. But anal sex tends to be under-reported, and condom use tends to be less common among heterosexuals when they are having anal sex than it is during vaginal sex. Given that anal intercourse is a highly efficient method of transmitting HIV - and that participants were told that the gel was for vaginal use, not anal use - unprotected anal sex may have given rise to a number of infections.

Regardless of cause, the news elicited field-wide disappointment, as well as unanimous agreement that MDP 301 was a sound, well-conducted trial that produced valuable information. The trial also illustrates perfectly why multiple clinical trials are needed to evaluate novel prevention strategies. If MDP 301 had not already been underway when the HPTN 035 results came out, it would have been necessary to wait for such a confirmatory trial to be mounted and completed. Clearly, proceeding on the assumption (based on the HPTN 035 data) that PRO 2000 worked would have been dangerous, given PRO 2000’s now-documented lack of effectiveness. 

The MDP investment in behavioural and social-science research as a part of MDP 301 was also unprecedented in the field. Data from the trial’s broad and sophisticated array of behavioural sub-studies are expected to provide important insights into the ongoing challenges of acceptability and adherence that (as illustrated above) are key to researchers’ ability to access product effectiveness accurately.

MDP 301 is the last large efficacy study of non-specific microbicide products for the time being.  Although several non-ARV-based candidates are being investigated by preclinical researchers - and some may prove promising enough to advance to clinical trial eventually - virtually all of the candidates now moving though the clinical-trial pipeline are ARV-based.