Prioritisation is crucial for the success of PrEP, model confirms

A mathematical model developed by two researchers at Imperial College in London, and based on what would happen if pre-exposure prophylaxis (PrEP) was introduced to a high-prevalence region in Kenya, shows that PrEP could be a ‘runaway success’ or a ‘runaway failure’, depending on a number of factors.

These include adherence, whether new longer-lasting drugs are used, the cost of drugs, and the overall efficiency of distribution. But the model shows – as other cost-effectiveness models have done – that by far the most influential determinant of PrEP’s effectiveness is whether it is targeted accurately at those at highest risk of HIV. A programme with a fixed budget that was targeted poorly would be prohibitively expensive in terms of the money spent to prevent one infection, and would actually prevent few infections, because a high proportion of people would be taking PrEP who would not have acquired HIV anyway.

The model

The model assumes that a PrEP programme is introduced into the Nyanza province of Kenya, which has an HIV prevalence (as of 2009) of nearly 14%, or about 370,000 people in an adult population of 2.65 million.

The model sees what would happen if a PrEP programme with a fixed annual budget of US$20 million was introduced in 2015 and scaled up to maximum size by 2020. Its base case assumes a fixed cost per person per year for PrEP of $250 (with HIV treatment for those acquiring HIV costing more or less the same) and that PrEP users will spend an average of five years on PrEP. It stratifies the population by gender, male circumcision status (Nyanza hosts a large voluntary medical male circumcision programme), and low-, medium- or high-risk behaviour.

It assumes PrEP with full adherence is 90% efficacious but also assumes that adherence is imperfect; it stratifies the population into half with adequate adherence (60% of doses taken) and half with poor adherence (only 20%).

This base-case model finds that 24,603 HIV infections would be prevented between 2015 and 2025, with a steady 3400 prevented per year by 2020. This would cost roughly $6000 per HIV infection averted. Although this cost is not calculated, this represents maybe 15-20 years’ worth of HIV treatment, had the person acquired HIV.

What if things change?

The researchers then posit what would happen if certain positive or negative developments happen over time as the PrEP project matures:

• What if the programme becomes more efficient at targeting those at highest risk, so the proportion of people taking PrEP who are at highest risk rises over five years from 50% to 100%? Conversely, what if it shrinks to zero (i.e. no prioritisation)?
• What if adherence improves, to 90% in people with ‘good’ adherence, and 70% in people with ‘poor’ adherence? Conversely, what would happen if it declines, to 30% and zero respectively?
• What if the cost per unit of PrEP declines to $125 or, conversely, increases to $500?
• What if the use of longer-lasting drugs improves efficacy in people with good adherence by 64% more than baseline efficacy, and even by 44% in people with poor adherence?
• What if economies of scale increase the number of people the programme can put on PrEP per year over time? Or what if, conversely, decreased efficiency (maybe due to the cost of reaching ever harder-to-reach populations) reduces that figure?

The model finds that good prioritisation would be by far the biggest influence on the success or failure of PrEP. Good prioritisation would reduce the cost per infection averted from $6000 to $2060 and increase the annual number of infections averted to over 9000. Conversely, lack of prioritisation would increase the cost per infection averted to $36,360 and reduce the annual number of infections to a couple of hundred at best.

The other changes posited would have less of an impact. Improved adherence as above would decrease the cost per infection averted to $4000 and increase infections averted to 7300 a year, while poor adherence would increase the unit cost to $9000 and decrease infections prevented to 1400 a year. Economies of scale could decrease the cost to $4700 and increase infections averted to 7300, while increased inefficiency would increase the cost to $7200 and decrease infections averted to 2150.

The model finds that the introduction of longer-lasting drugs/formulations would not, in this model, make much of a difference: perhaps only 400 or so more infections prevented a year and $200 less in unit cost.

Comments and conclusions

The researchers comment on the importance of prioritisation and suggest the introduction of a ‘risk score’ as was used in the recent Partners PrEP Demonstration Project, but also praise the success of community-based projects in India, such as Avahan, where peer-led outreach, community mobilisation, concentration on HIV ‘hotspots’ and good data collection have optimised the prioritisation of those at most need of HIV prevention measures. However, they also point out the difficulty of prioritising “highly mobile and marginalised” populations and say that if stigma were to arise against people taking PrEP, this would be an additional barrier. A recent report from Avahan shows that the programme’s efficiency may have started to decline after its funding was handed over to the Indian government, showing the importance of skilled management and ‘bottom-up’ priorities informed by community members.

Implementation science, they say, will have a crucial role in ensuring that PrEP becomes a “runaway success” rather than a “runaway failure”.