Introduction

A number of key issues in microbicide trials require careful consideration at the planning stage. Before the recent success with antiretroviral-based microbicides, it was suggested that lack of attention to some of these important issues was a potential contributing factor to the relatively high proportion of trials delivering an indeterminate or negative result.

In prevention trials, unlike treatment trials, identifying those who will benefit from the intervention is difficult, except in well-defined, high-risk populations (e.g. sex workers). Prospective trial investigators must have prior knowledge of the potential target trial population, including the status of the human immunodeficiency virus (HIV) epidemic in that area, in particular reasonably robust estimates of HIV incidence. As with other HIV prevention trials, microbicide trials rely on adequate numbers of incident HIV infections from an HIV-negative population, and therefore targeting populations sufficiently at risk of infection is essential. During the trial, collecting and monitoring behavioural data, such as sexual frequency, condom use and adherence to the product, are essential to understanding the end results. In this chapter, we outline the main features of microbicide trial design and discuss the challenges specific to these types of clinical trials.

Design features of microbicide trials

Incidence of human immunodeficiency virus

The incidence of HIV is defined as the number of newly observed infections occurring within a specified time period, and is usually expressed as the number of infections per 100 person-years. The current estimated HIV incidence in a target trial population is a key parameter for determining the required sample size. The incidence determines the expected number of infections to be observed during the trial, which in turn drives the statistical power. The statistical power is the probability the trial will detect an effect of a microbicide (in reducing the incidence of HIV), if the microbicide is truly effective. An overestimate of incidence can result in an under-powered trial with an indeterminate result.

In recent years, HIV incidence has dropped by more than 25% between 2001 and 2009 in many countries within sub-Saharan Africa, where most new infections still occur. Although this overall decline will mask populations in which the incidence could still be increasing, the implication for microbicide trials conducted in the general population with declining HIV incidence, is that thousands of trial participants will be required in order to observe a sufficient number of HIV infections. The dramatic effect on the required sample size for varying levels of HIV incidence ranges from 2 to 5 per 100 person-years as shown in Figure 1 .

Statistical power and required sample size to detect a 50% reduction in human immunodeficiency virus incidence for varying levels of control arm human immunodeficiency virus incidence: 2.0, 3.0, 4.0 and 5.0 per 100 person years.

Previous studies have failed because of lack of power resulting from a lower than expected HIV incidence. For example, the SAVVY trial and phase 2 Tenofovir gel trial were both terminated prematurely for this reason; neither study specifically measured HIV incidence in the target population before starting the trial.

Incidence of HIV can be estimated from seroprevalence data using biomarkers to determine the timing of the infection. Recent advancements in improved assays could lead to improved incidence estimates. If no contemporary data exist, however, it is recommended that pre-trial feasibility studies in the target trial population are undertaken in order to obtain a reliable estimate of HIV incidence. Although pre-trial feasibility studies increase the overall costs of the trial, such studies also provide valuable insight into whether the product is acceptable, whether participants can be recruited and retained in follow up, and also help to assess whether any behavioural indicators could reduce the likelihood of demonstrating efficacy (i.e. increased condom uptake after counselling). Feasibility studies were conducted in each of the six study sites in the MDP301 trial of PRO2000 gel, in which the pooled estimated HIV incidence, weighted according to the expected number of women to be enrolled by site, was 6.2 per 100 person years. For this trial, a conservative estimate of 4 per 100 person years was used in the sample size. It was acknowledged that the incidence could decrease during the trial, as evidence showed a declining incidence in Uganda and Zambia at that time. This turned out to be the case, as the observed control arm incidence in this trial was 4.5 per 100 person-years.

The key difference in prevention trials compared with treatment trials is the difficulty of defining a target population at risk of HIV, unlike treatment trials in which potential participants are usually identified in hospital after a confirmed diagnosis of the disease under investigation. To address this in HIV prevention, some studies were set up to recruit serodiscordant couples (with an HIV positive male partner). For example, HPTN 039, enrolled over 1300 discordant couples in 14 sub-Saharan African sites to assess whether acyclovir suppression in people who were HSV-2 positive reduced the chance of HIV transmission. The MDP301 trial of PRO2000 gel enrolled 840 seronegative women with HIV-positive male partners in Masaka, Uganda. The specific target population will also depend on the main objective of the trial. If it is to demonstrate ‘proof of concept’ of the product, then a high-risk, high-adherent population would be most desirable. Study participants in sero-discordant relationships who know the status of their partners would arguably be more motivated to adhere to a clinical trial of a microbicide. Recent results have shown that treating the HIV-positive partner protects the negative partner from HIV infection by up to 96%. The recruitment of serodiscordant couples in prevention trials needs careful consideration, as partners being treated would be providing a level of protection far over and above anything potentially being provided by a microbicide.

Design features of microbicide trials

Incidence of human immunodeficiency virus

The incidence of HIV is defined as the number of newly observed infections occurring within a specified time period, and is usually expressed as the number of infections per 100 person-years. The current estimated HIV incidence in a target trial population is a key parameter for determining the required sample size. The incidence determines the expected number of infections to be observed during the trial, which in turn drives the statistical power. The statistical power is the probability the trial will detect an effect of a microbicide (in reducing the incidence of HIV), if the microbicide is truly effective. An overestimate of incidence can result in an under-powered trial with an indeterminate result.

In recent years, HIV incidence has dropped by more than 25% between 2001 and 2009 in many countries within sub-Saharan Africa, where most new infections still occur. Although this overall decline will mask populations in which the incidence could still be increasing, the implication for microbicide trials conducted in the general population with declining HIV incidence, is that thousands of trial participants will be required in order to observe a sufficient number of HIV infections. The dramatic effect on the required sample size for varying levels of HIV incidence ranges from 2 to 5 per 100 person-years as shown in Figure 1 .

Statistical power and required sample size to detect a 50% reduction in human immunodeficiency virus incidence for varying levels of control arm human immunodeficiency virus incidence: 2.0, 3.0, 4.0 and 5.0 per 100 person years.

Previous studies have failed because of lack of power resulting from a lower than expected HIV incidence. For example, the SAVVY trial and phase 2 Tenofovir gel trial were both terminated prematurely for this reason; neither study specifically measured HIV incidence in the target population before starting the trial.

Incidence of HIV can be estimated from seroprevalence data using biomarkers to determine the timing of the infection. Recent advancements in improved assays could lead to improved incidence estimates. If no contemporary data exist, however, it is recommended that pre-trial feasibility studies in the target trial population are undertaken in order to obtain a reliable estimate of HIV incidence. Although pre-trial feasibility studies increase the overall costs of the trial, such studies also provide valuable insight into whether the product is acceptable, whether participants can be recruited and retained in follow up, and also help to assess whether any behavioural indicators could reduce the likelihood of demonstrating efficacy (i.e. increased condom uptake after counselling). Feasibility studies were conducted in each of the six study sites in the MDP301 trial of PRO2000 gel, in which the pooled estimated HIV incidence, weighted according to the expected number of women to be enrolled by site, was 6.2 per 100 person years. For this trial, a conservative estimate of 4 per 100 person years was used in the sample size. It was acknowledged that the incidence could decrease during the trial, as evidence showed a declining incidence in Uganda and Zambia at that time. This turned out to be the case, as the observed control arm incidence in this trial was 4.5 per 100 person-years.

The key difference in prevention trials compared with treatment trials is the difficulty of defining a target population at risk of HIV, unlike treatment trials in which potential participants are usually identified in hospital after a confirmed diagnosis of the disease under investigation. To address this in HIV prevention, some studies were set up to recruit serodiscordant couples (with an HIV positive male partner). For example, HPTN 039, enrolled over 1300 discordant couples in 14 sub-Saharan African sites to assess whether acyclovir suppression in people who were HSV-2 positive reduced the chance of HIV transmission. The MDP301 trial of PRO2000 gel enrolled 840 seronegative women with HIV-positive male partners in Masaka, Uganda. The specific target population will also depend on the main objective of the trial. If it is to demonstrate ‘proof of concept’ of the product, then a high-risk, high-adherent population would be most desirable. Study participants in sero-discordant relationships who know the status of their partners would arguably be more motivated to adhere to a clinical trial of a microbicide. Recent results have shown that treating the HIV-positive partner protects the negative partner from HIV infection by up to 96%. The recruitment of serodiscordant couples in prevention trials needs careful consideration, as partners being treated would be providing a level of protection far over and above anything potentially being provided by a microbicide.

Randomisation, blinding and choice of control arm

Randomised double-blind, placebo-controlled trials are considered to be the gold standard in clinical trial design. Randomisation ensures estimates of effect sizes are unbiased, and helps to ensure groups are balanced for known and unknown risk factors. Double-blinded trials, in which both participants and investigators are blind to which individuals are taking the active product and which individuals are taking placebo, ensure that no conscious or subconscious influence from this knowledge affects the results. Matching products for outward appearance is particularly important for maintaining the blind, and it is recommended that blinded assessments are made of active and placebo products.

Microbicide trials typically use an inert placebo gel as the control arm in a double-blinded design. As an effective microbicide moves closer to implementation, however, using a placebo-control will become unethical and unacceptable to participants, so that active control arms will need to be considered. One of the early microbicide trials, HPTN 035 of PRO 2000 gel, used two control arms: a placebo gel arm and an unblinded ‘no gel’ (or ‘condom only’) arm. The trial investigators argued that, given the objective nature of the outcome (HIV infection), the risk of introducing bias from the inclusion of an unblinded control arm was minimised. The advantage of including these two control arms, however, would mean that the possible protective effects of the placebo gel could be assessed (i.e. from the resulting lubrication or the gel acting as some sort of barrier to the virus). It is also well known that behaviour of participants within a trial is often different to those not enrolled. This is because, within a trial, the level of supervision and contact with medical and paramedical staff is likely to be much greater than in real life (i.e. the so-called ‘Hawthorne effect’). Differences from real life do not matter so much in a placebo-controlled investigation, as they will be independent of the treatment to which a participant has been allocated. A no-gel arm probably results in behaviour change, such as more diligent condom use, increased risk of gel sharing, and poorer retention in the no-gel arm. Results from this trial showed that, although condom use was significantly higher in the no-gel arm, no measurable difference in HIV incidence between the two control arms was observed.

Duration of follow up

It is important to recognise that poor retention can undermine any clinical trial, as the reasons for dropping out could be related to the intervention, and clinical trials with poor follow up are difficult to interpret. Both adherence to gel use, and frequency of attendance at follow-up visits, are likely to be influenced by the duration of the trial. Maintaining good adherence and retention with preventive therapy becomes increasingly difficult with time. Even in people who are being treated for active disease, maintaining high adherence and attendance is difficult (e.g. in the treatment of tuberculosis). In determining the length of follow up, a trade-off can exist in selecting a shorter duration for maintaining full adherence and a longer duration for generalisability. Short study designs of say, 9, 6 or even 3 months, are more likely to be able to demonstrate proof of concept than studies requiring participants to adhere for 24 months or more. Long-term safety data necessary for regulatory submission could be obtained from such studies by following a subgroup for longer. Typically, microbicide trials are powered to require a pre-determined number of HIV infections and are based on a minimum follow up of at least 1 year for each participant.

Effect size

The postulated size of the effect (or effectiveness) of the microbicide product under investigation is an essential parameter for determining the size of the trial.

Product effectiveness is usually expressed as a percentage and can be estimated from

Stay updated, free articles. Join our Telegram channel

Join