Results from the Caprisa 004 tenofovir microbicide trial

Simon Collins, HIV i-Base

In terms of conference headlines, the biggest news came from the results of a prevention study called Caprisa 004. This study reported that a microbicide gel containing 1% tenofovir reduced the risk of infection to women when used before sex to protect against HIV by 39%. [1, 2] Previous microbicides (not using HIV drugs) have not shown a benefit, so a positive result, no matter how slight, was likely to be greeted enthusiastically. When the results were presented, the audience gave the presenters a standing ovation.

Importantly, the presenters stressed that these preliminary results justified further research. This study was based on 98 endpoints for the primary analysis and the sample size ensured that they could be 90% confident of detecting a doubling/halving in the risk (ie an OR of 2 or 0.5). However, because the endpoints are by definition fewer in subgroup analyses, the study is not powered to analyse some of those interesting results. One of the most helpful aspects of the study is that the detailed results were published in a free-access article in Science Express. [3]

The theoretical benefit from an antiretroviral microbicide is similar to the use of pre- and post-exposure prophylaxis (PrEP and PEP) but instead of taking oral drugs, applying a gel enables the active drug to be absorbed in the tissues that are first exposed to the virus. If the cells in the genital tissues have antiretroviral activity, the hope is that this will reduce the risk of infection.

As with all studies, the complexity of the results is in the details, and the presenters themselves cautioned that their results primarily signaled the urgency of running additional studies.

Women were advised to use the microbicide ‘up to 12 hours before sex’ and ‘as soon after as possible’, using a maximum of two doses in any single 24 hour period. The gel was applied with a special pre-filled applicator, similar to a tampon container.

This phase 2b study was in around 900 women aged 18–40 years, living in two districts in South Africa where the risk of HIV for women reaches 50% by the age of 24. One trial site was in urban Durban (n=278) and the second was in a rural location 90 miles from Durban (n=611). This was a double-blind study with women randomised 1:1 to either the active gel or a placebo gel. Free condoms and counselling on the importance of safe sex were provided to all women, with monthly HIV testing and monitoring.

There were significant differences between the rural and urban women. Rural women were younger (mean 23.3 vs 25.1), poorer (86% vs 69% monthly income <R1000), less likely to have a stable partner (77% vs 93%), had fewer lifetime partners (mean 2 vs 6), used condoms less consistently (22% vs 42%) and had lower HSV-2 prevalence (48% vs 60%), see Table 1. However, randomisation ensured that there was no difference in these baseline characteristics between the active and placebo group.

Table 1: Demographic differences between rural and urban sites
Rural site n=611 Urban site n=278 p-value
Mean age (years) 23.3 25.1 <0.001
Monthly income <R1000 86.1% 69.1% <0.001
Married 6.5% 3.6% 0.085 *NS
Stable partner 77.0% 93.1% <0.001
Mean age sexual
17.3 17.7 0.014
Mean no. sexual partners (lifetime) 2.1 6.0 <0.001
Mean age of oldest partner (past 30 days) 26.4 29.6 <0.001
Sex in the past 7 days 58.9% 68.3% 0.007
Always use condom 22.9% 42.8% <0.001
partner (past 30 days)
0.5% 2.5% 0.014
Anal sex (past 30 days) 0.5% 0.4% 1.000
HSV-2 prevalence 47.6% 59.6% 0.001

* NS = non significant differences

The predetermined endpoint of 98 events was reached after a mean 18 months with 1341 person years (PY) of follow-up, with a low drop-out rate (~5%).

Of the 98 women who became HIV-positive over 12–30 months, 38 were in the active gel group and 60 were in the placebo group. The HIV incidence rate per 100 PY was 5.6 (CI: 4.0, 7.7) in the tenofovir gel arm compared to 9.1 (CI: 6.9, 11.7) in the placebo gel arm (incidence ratio rate [IRR]=0.61; CI: 0.40, 0.94; p=0.017). After adjusting for baseline covariates including, age, site, anal sex history, contraceptive method, HSV-2 antibody status, and condom use, the hazard ratio was 0.63 (CI: 0.42, 0.94; p=0.025). Sensitivity analysis produced similar results. Although this fell just short of the predetermine OR of 0.50, the results remained statistically significant.

The combined rural/urban analysis produced a protection rate of 39% from using the active compared to placebo gel. However, the 95% confidence intervals are 6% and 60%. Further studies are likely to focus on dosing, adherence and other factors in order to see whether higher protection rates can be seen. Although the results were presented by site, showing effectiveness at the rural site of 43% (95%CI 5, 57; p=0.023) but not at the urban site (26%; 95%CI –59, 67; p=0.380), see Table 2. However, as the study was not designed to compare efficacy by site, while interesting, it was not powered for this comparison to be meaningful.

Table 2: Effectiveness results in Caprisa 004 study
HIV infections/PY HIV incidence
n Tenofovir Placebo Tenofovir Placebo IRR Efficacy 95%CI p-value
Overall effectiveness of tenofovir gel (n=889)
HIV endpoints 98 38/680.6 60/660.7 5.6 (4.0, 7.7) 9.1 (6.9, 11.7) 0.61 39% 6, 60 0.017
Site-specific effectiveness (n=889)
Rural 611 25/484.7 42/461.2 5.2 (3.3, 7.6) 9.1 (6.6, 12.3) 0.57 43% 5, 67 0.023
Urban 278 13/195.9 18/199.5 6.6 (3.5, 11.3) 9.0 (5.3, 14.3) 0.74 26% –59, 67 0.380 NS
HIV endpoints by levels of adherence (n=884) †
Adherence >80% 336 11/259.2 25/269.4 4.2 (2.1, 7.6) 9.3 (6.0, 13.7) 0.46 54% 4, 80 0.025
Adherence 50-80% 181 10/159.8 10/99.7 6.3(3.0,11.5) 10.0 (4.8, 18.4) 0.62 38% –67, 77 0.343 NS
Adherence <50% 367 16/258.5 25/290.6 6.2 (3.5, 10.1) 8.6 (5.6, 12.7) 0.72 28% –40, 64 0.303 NS

NOTE: Study was not powered for the subgroup analyses by site and adherence.

†Adherence could not be calculated for the 5 women who reported no sex during their follow-up in the study. NS=non significant

Adherence is essential to monitor in any intervention study, see Table 2. In Caprisa 004, the researchers determined that two applications of the gel were used for over 70% of occasions when participants had sex. While approximately 40% women reported >80% adherence, a similar proportion reported that they used the gel less than half the time. When adherence was 80% or higher (n=336), the protection increased from 39% to 54% (95%CI 4, 80; p=0.025). There appeared to be a trend between adherence and efficacy, and this is clearly plausible, though again the study was not powered for this comparison. The Science Express paper reported 38% protection (95%CI –67, 77; p=0.343) at 50–-80% adherence (n=181) dropping to 28% (95%CI –40, 64; p=0.303), when less than 50% (n=367).

The mean number of sex acts in the high, intermediate and low adherence groups was 3.2, 5.0 and 6.7 per month respectively. Median adherence in the women who become HIV-positive was similar throughout the study at approximately 60%, whereas in the HIV-negative women this started at 55% and increased to 75% in the first and last six months respectively. Even with an intensive education and support programme, only a minority of women achieved >80% adherence, and these were the women who had less sex (3 times a month). Condoms were reportedly used 80% of the time, though this may have been over-reported given the rough per-exposure risk this generates for the study, which is not uncommon in prevention studies.

No serious or significant safety issues (from the 4692 reported events) were associated to using the gel in terms of side effects, including renal toxicity or in the 54 unplanned pregnancies that occurred. Mild diarrhoea was reported in 16% people using the active gel compared to 11% of the placebo group. No safety concerns in terms of flares in liver enzymes were seen relating to the use of tenofovir in the small numbers of women who entered the study with active hepatitis B (19 in the active and 15 in the placebo group) or who acquired HBV during the study (22 women, 19 or who cleared the virus without additional treatment). The concern that continued exposure to tenofovir prior to HIV being diagnosed might exert sufficient pressure to generate drug resistance was not supported in genotypic results from 35 women (no K65R, K70E or RTI-associated mutations). Of interest, the use of the active gel had no impact of viral set point after infection (4.65 vs 4.30 log; p=0.15) and participation in the study did not lead to any increase in risk behaviour.

The study also reported an impact on transmission of HSV-2, the virus responsible for genital herpes. Of the 434 women who tested negative for HSV-2 at the beginning of the study, 29 became infected in the active gel group compared to 58 in the placebo group (IR/100PY 9.9 (6.6, 14.2) vs 20.2 (15.3, 26.1). This was reported as tenofovir providing 51% protection against HSV-2 (95%CI: 22%–70%; p = 0.003). Because genital herpes increases the risk of catching HIV, these results are complicated to understand. Although tenofovir has not shown protective effects against HSV-2 in mouse and test-tube studies, drugs with a similar structure to tenofovir such as cidofovir have activity against HSV-2.

Results from the pharmacology substudy of CAPRISA 004 were presented in the same session by Angela Kashuba from the University of North Carolina. [5]

For the HIV analysis, 90 samples were available (37 active and 13 placebo in the HIV-positive women plus 24 active and 16 placebo from women who remained HIV-negative. Tenofovir levels were measured in blood plasma (BP), and cervicovaginal fluid (CVF) for all samples and additionally in vaginal and cervical tissue biopsy samples in the HIV-positive women. Plasma concentrations were minimal (<1 ng/mL), with detectable levels in only 12% of the HIV-positive women (median 0, range 0–0.1 ng/mL) a median 6 days (range 1-25) after application vs 50% of the HIV-negative women (median 0.1, range 0–0.8 ng/mL) after a median of 4.5 days (range 2–28), indicating very low systemic uptake even given the delay in sampling.

Tenofovir was more frequently detected and at higher CVF levels in the HIV-negative compared to HIV-positive women at 45% (median 1 ng/mL range 0–300,000) vs 96% (median 520 ng/mL (range 0–1,340,000), both at 4.5 days. CVF concentrations correlated well with infections and also importantly with intracellular levels of tenofovir diphosphate. This will help establish the target dose in future formulations. A separate PK study of 250 samples from 172 highly adherent HIV-negative women showed a mean half-life of about two days with most concentrations over the first few days of ~1000 ng/mL. It is important to note that there are currently no data on appropriate target levels of either tenofovir or tenofovir diphosphate and that data, as for early absorption (ie how soon before sex would you get protection?) will be the focus of the next studies. These results suggest that drug levels are a marker for adherence
rather than poor absorption potentially due to interpatient variability of cellular transporters such as MRPs.

A similar relationship was observed between drug levels and acquisition of HSV-2. While oral tenofovir is not able to achieve sufficient drugs levels to suppress HSV-2 (EC50 ~10,000 ng/mL), this is possible with a topical gel. 24% of the women with levels below this became HSV-2 positive compared to only 6% of women who had levels above.

Very low levels of tenofovir found in two women in the placebo arm was explained by possible shared sexual partners.


The proof of principal that an antiretroviral microbicide can protect against HIV and HSV infection is clearly important news.

The discussion in the published paper suggests that many of the infections may be due to infrequent but very high risk exposures with migrant workers and the investigators noted that the HIV incidence rate was similar in the low frequency placebo group to women who reported much more frequent sex.

In this high-risk setting, infection rates remained high in the women using the active gel (>5/100PY) and protection dropped significantly after 18 months for reasons that are unclear.

The differences in the urban/rural results may just be an issue of overall sample size (as opposed to something connected to the difference in lifetime sex partners or other factors). A good precedent for caution over the adherence analysis however comes from an earlier microbicide study. A similar adherence analysis in the phase 2b PRO2000 HPTN 035 study showed protection rates of 9%, 44% and 78% in low gel users, high gel users, and low condom/high gel users, respectively. Yet this microbicide was subsequently shown not to work.

Of note, the findings on prevention of HSV-2 transmission were more significant and robust than protection against HIV, and this will clearly be the focus for further research study.


Unless otherwise stated, all references are to the Programme and Abstracts of the 17th International AIDS Conference, 18-23 July 2010, Vienna.

  1. Karim QA et al. Effectiveness of 1% tenofovir vaginal microbicide gel in South African women: results of the CAPRISA 004 trial. Oral abstract
  2. Karim SA et al. Safety of 1% tenofovir vaginal microbicide gel in South African women: results of the CAPRISA 004 trial. Oral abstract TUSS0502.
  3. Karim QA et al. Effectiveness and safety of tenofovir gel, an antiretroviral microbicide, for the prevention of HIV infection in women. Science Express. Published Online 19 July 2010. Science DOI: 10.1126/science.1193748.
  4. Kashuba
    ADM et al. Do systemic and genital tract tenofovir concentrations predict HIV seroconversion in the CAPRISA 004 tenofovir gel trial? Oral abstract TUSS0503.

Further information:

Links to other websites are current at date of posting but not maintained.