Microbicides 2008: Accurate adherence reporting essential for microbicide trials

Researchers need much better ways of determining adherence to candidate microbicides in trials, and also of ways of determining sexual behaviour, the 2008 Microbicides Conference heard today in Delhi, India.

Barbara Mensch of the Population Council told the Conference: “The inability to prove effectiveness leaves the question of product efficacy unanswered.”

Several trials of new prevention technologies have failed recently despite excellent adherence to the intervention reported by participants. In the recent Carraguard trial (run by the Population Council), 96 per cent of women claimed to be using gel correctly. However when applicators were tested to see if they had actually been used, only 44 per cent had. Similarly the figures for women who claimed always to use condoms only, condoms plus gel, or gel only, added up to more than 100 per cent.

Similar problems may have plagued the recent HPTN 039 trial of aciclovir for herpes prophylaxis, in which the reduction in genital ulcers was suspiciously low despite reported 94% adherence amongst participants and pill counts.

Low adherence in itself may not jeopardise a trial as long as sufficient participants are high adhererers and as long as we know which is which, Mensch said.

“Poor adherence means that results of an intention-to-treat analysis will be very different from a per-protocol one,” she said, “but inability to measure poor adherence biases effects towards the null.” In other words, the inability to even compare intention-to-treat analysis, in which results are viewed as if all participants adhered to the trial medication, with per-protocol analysis, in which only participants that took it as prescribed are counted, is likely to lead to negative results even when the microbicide is highly effective in those who use it.

What are researchers doing to try and establish true adherence and behavioural data? One way is to use Automated Computer-Assisted Self-Interviewing (ACASI) with participants instead of face-to-face interviews live researchers.

This technique worked well in a substudy of the Carraguard study. This was a simulated three-month microbicide study in which 848 women were given placebo gel but assessed as if they were taking a trial compound.

As well as using ACASI to assess adherence, researchers visited the women to count off used applicators. The trial also used a PCR test to detect sperm in vaginal fluids to assess the true levels of sexual activity.

Establishing the latter is also extremely important; microbicide researchers are just beginning to understand the degree to which trial participants give unreliable reports of their own sexual activity.

For instance, in the MIRA trial, which looked at the possibility of using the female diaphragm in HIV prevention, the pregnancy rate was actually higher in women who reported contraceptive use than in women who did not. In the MTN 035 phase IIb trial of the microbicides Buffergel or 0.5% PRO-2000, at one site participants reported more sex acts protected by a condom than the total number of sex acts reported to invvestigators at that site.

In the simulated trial, 19% of women who reported no sex in the last 48 hours tested positive for semen in the vagina.

Sex may be misreported for a variety of reasons, not all of them due to deception; women may differ in their definitions of what ‘sex’ is, may fail to report non-consensual experiences, or may simply tell the researcher what they want them to hear: very few of us accurately report our sex lives, even to ourselves.

The ACASI technique proved particularly useful in assessing non-adherence to the product. Fact-to-face interviewing (FTFI) consistently produced lower figures for non-adherence than the ACASI technique, as in the following table:

Researchers also had to take lessons from demographers and opinion pollsters in how they phrased questions, Mensch added, for instance, asking “Were you able to use the gel last time you had sex?” instead of the more direct “Did you use the gel last time you had sex?”

Better still might be to ask about non-use rather than use: “In how many sex acts were you not able to use the gel?”

Interviewers should be trained to use supplementary questions to establish how participants arrived at answers (“When I just asked you how many sex partners you’d had, how did you count them up?”) and what their understanding of terms was (“What counts as a sex partner for you?”).

Mensch asked a list of questions microbicide trials had failed so far to ask, or have answered. Are the highest-risk participants the most or least adherent? Does adherence and sexual behaviour change over the trial? Under what circumstances are microbicides most and least likely to be used? What influence do partners and family have on women? Are trial populations really representative of the target population and if not, does it matter?

Future trials will try and answer some of these questions with more sophisticated monitoring. The MTN 035 trial site in Malawi, for instance, has started using an ACASI system asking questions of participants in writing, in voice and with graphics.

The CAPRISA 004 tenofovir-gel study, which has recently started in South Africa, is considering issuing participants with modified Blackberry terminals in which to record their gel use and behavioural data and transmit them to the centre. And the International Partnership for Microbicides is considering how to do the most direct-possible observation of women’s gel use in its series of studies of dapivirine (TMC120) without endangering or stigmatising participants.

In conclusion: the recent failure of the Carraguard trial to produce a positive result has not dampened researchers’ and advocates’ enthusiasm or belief in the concept of microbicides. It has instead highlighted the extreme complexity of these large trials and the numerous social and behavioural influences that need to be considered if researchers are to generate a meaningful result for the millions of dollars spent on them.