What is probably the most detailed and realistic modelling study so far done of the impact and cost-effectiveness of pre-exposure prophylaxis (PrEP) finds that it could be very cost-effective in terms of the price of an extra year of healthy life for an individual taking it.
However, its absolute cost is likely to restrict its use to only the highest-risk individuals in a population, and this may mean it may only make a minor impact on HIV incidence.
These, at least, are the effects forecast amongst men who have sex with men (MSM) and transgender women (TGs) in the capital of a middle-income country. The country in the model is Lima, Peru, and the reason it was chosen as the base for the model is because it featured the highest number of enrolments to the seminal iPrEx study of PrEP in MSM/TGs, the first of several studies to report significant efficacy (in the case of iPrEx it cut HIV infections by 44%, even though only 50% of trial participants took enough doses of PrEP for it to work).
The inputs for the model were derived both from the data produced by iPrEx and from the anthropological and social research into the MSM/TG communities of Lima and Peru conducted in advance of the actual study; the latter was crucial to iPrEx's success because it enabled appropriate targeting and recruitment.
The model divides the MSM/TG population into four categories and assumes different levels of HIV incidence in them, based on projections derived from the original background research. These are:
MSM who mainly have sex with men (MMSM) – these form 70% of the target population and have an annual HIV incidence of 2.5%.
MSM who mainly have sex with women (MMSW) – 15% of the target population, annual incidence 1.0%.
Male sex workers (MSW) – 10% of the target population, annual incidence 3.1%.
Higher risk transgender women (TG) – 5% of the target population, but the highest annual incidence at 7.3%.
These are subdivided into nine risk types, according to whether they are exclusive ‘tops’ (insertive), ‘bottoms’ (receptive) or ‘versatile’ (both).
The model assumes a basic biological efficacy of 92% for PrEP if there is perfect adherence – so called ‘conditional efficacy’ – and then modifies this by inputting three different levels of adherence (95, 45 and 15%) and varying proportions of people who achieve these different adherence levels. These lead to a ‘functional effectiveness’ of 52, 62 or 35% for the level of adherence typical of iPrEx, a higher level or a lower level.
It also inputs two, fairly modest, levels of coverage – 5% of the target population and 20% of the target population.
Some other PrEP modelling studies have assumed higher coverage levels but this model chose an upper cost limit of 10% of Peru’s current spending on HIV prevention. Depending on how groups are prioritised, this would provide from 2500 to 18,500 men with PrEP for every year for the next ten years (if each person took it for the whole decade), and was used as a cap for the likely amount of PrEP possible. The model assumes that PrEP rollout is started today, reaches a maximum level in five years and is then sustained for five years more.
One important aspect of the model is that it assumed three different levels of targeting, with PrEP uptake either distributed evenly amongst the MSM/TG population, or moderately or strongly targeted to the highest-risk people, meaning that coverage is high or very high in TGs and MSWs and not as high in the rest of the MSM population. Because there was less PrEP to go around, in the case of 5% coverage it is assumed ‘high targeting’ means that it is only TGs and MMSWs who receive PrEP. (In a city with a different gay scene, it could be other high-risk groups who formed these target populations.)
The costs assumed for PrEP range from $525 to $830 per person per year of which the majority of the cost is for the drug – $420 for tenofovir and $600 for tenofovir/FTC (Truvada). The cost of antiretroviral therapy for people who are infected within the community during the model’s ten years is assumed to range from $1000 to $3000 a year. The PrEP programme’s average annual cost would range from $24 million to $152 million, depending on coverage and programme costs.
These different inputs were then used to calculate a number of different outputs:
The percentage drop in new HIV infections.
The total number of new infections.
The cost per disability-adjusted life-year with HIV (HIV DALY) averted.
How this cost measured up against two different definitions of cost-effectiveness.
The effect of behaviour change, measured as a change in condom use in PrEP users from a 30% increase to a 100% decrease.
The cost-effectiveness of PrEP compared with other interventions.
There is no single definition of ‘cost effectiveness’ and the model used two very different ones. The World Health Organization sees an intervention that costs less per person per year than the annual per capita gross domestic product (GDP) as ‘very cost effective’ and less than three times the annual GDP as ‘cost effective’ – in Peru these figures are $5401 and $16,203 respectively.
The World Bank uses a far stricter definition of cost effectiveness, which uses a figure (adjusted for inflation) of $149 a year for very cost-effective measures and $745 for cost-effective ones.
Results - effectiveness
The most striking finding of the model is that while, as might be expected, the number of infections prevented goes up if coverage is increased, the cost-effectiveness – the efficiency of the PrEP programme in cost per infection averted – decreases. This is because at low coverage levels, the fact that only people at extremely high risk of HIV are taking PrEP means that more infections are in fact averted per PrEP recipient – especially as the model includes the contacts of people receiving PrEP as well as the people themselves. For instance, giving PrEP only to MMSWs would reduce HIV incidence in the whole MSM population over ten years by just 0.9%, and only to MMSM by 1.2%. But if it was targeted solely at MMSWs annual incidence would drop by 3.4% and if solely to TGs then by 4.7%.
As these groups and their contacts are not isolated from each other and infections averted in one population will mean fewer transmissions to another, in practice giving PrEP would reduce annual incidence by more than this. For instance: if coverage was a modest 5% and adherence only at the level seen in the iPrEx trial, then the reduction in annual incidence would be 8%.
Although these declines in incidence sound modest, over ten years they would amount to several thousand infections averted. At the adherence level seen in iPrEx, 1000 infections would be averted with 5% coverage, 50% adherence and no targeting; with targeting, it would be about 2500 infections. If coverage was 25%, 3800 infections would be averted if there was no targeting and 6300 if PrEP was highly targeted; if adherence increased to high levels, it would increase to 7200 infections averted.
Results - cost effectiveness
These figures are there simply as hypothetical case scenarios: a more important consideration is whether PrEP would be cost effective.
Here, the modellers found that cost effectiveness depended crucially on whether the likely future costs of treatment for those who would be infected without PrEP were included. If these were not included, and the cost-effectiveness scenario simply calculated the cost of PrEP per extra DALY with HIV averted, then although PrEP would always be cost effective by the WHO definition, by the World Bank definition it would only be cost effective at 5% coverage if it was highly targeted. In this scenario, it would cost about $500 per DALY. At 20% coverage, targeted PrEP would cost $800 per HIV DALY averted. At both levels of coverage, non-targeted PrEP would cost about $1400 per DALY.
This picture changes if the costs of ART for people who would otherwise be infected is also included. If ART were to cost $1000 a year, then while the cost per extra HIV DALY averted for non-targeted PrEP would only go down to about $900, if targeted it would go down to $140 or even lower per HIV DALY at 5% coverage – in other words ‘very cost effective’, even by World Bank standards – and to about $380 a year at 20% coverage.
If ART were to cost $3500 a year, then PrEP would actually save money in both coverage scenarios – but it is unlikely that there would be such a wide gap between $600 for Truvada and $3500 as the average cost of ART.
To achieve larger cuts in HIV incidence would require much larger resources. For instance, to cut annual incidence by one-third within a decade using PrEP alone, the model finds, it would be necessary to spend half the country’s entire current HIV treatment, prevention and care budget every year for each of those ten years.
The model finds, nontheless, that PrEP is as cost effective as some other strategies. The paper compares the relative cost effectiveness of maintaining current condom and HIV therapy coverage, but putting all other HIV prevention money into one single HIV prevention strategy. If PrEP were that sole strategy, it would be more cost effective in prevention terms than (in terms of HIV DALYs averted) spending the extra money on ART treatment and as cost effective as spending it on MSM outreach programmes or presumptive STI treatment. It would not be as cost effective as spending the extra money on more condoms (except, just possibly, in a low-coverage but very highly targeted scenario) and would be much less cost effective, at least in Peru, than spending it on sex worker outreach or on expanding voluntary testing and counselling.
Impact of behaviour change and resistance
One interesting finding from the model is that behaviour change would have a relatively small effect on PrEP effectiveness and cost effectiveness. If PrEP was highly targeted, it would still be cost effective if there were a 30% drop in condom use in PrEP users; even a 50% drop would probably remain cost-effective. PrEP would only lead to an increase in HIV incidence if condom use in all MSM dropped by at least 50% and that the actual biological efficacy of PrEP was below 40% (which it clearly isn’t).
Drug resistance would also probably not be a big problem, if recipients took HIV tests regularly (which would presumably be a condition of their getting PrEP). If people got tested every three months, the proportion of PrEP doses taken by people who were in fact recently infected would only be about 0.4%.
It is important to remember that this model is not a demonstration of the potential of PrEP in all scenarios. Its strength, and its weakness, is that it is specifically tied to the current resources likely to be available in a country like Peru. In particular, it does not factor in any change in the price of PrEP.
Nonetheless, it finds that PrEP could be a cost-effective intervention, even using quite strict criteria, in a middle-income country if targeted at the gay male population and especially at its highest-risk members. The model also finds that some concerns about PrEP, such as behaviour change and drug resistance, are not crucial to is effectiveness.
Clearly, then, the biggest limiting factor to PrEP will be its sheer cost relative to other prevention methods. Where will the money for it come from?
The researchers comment that in a country like Peru where 100% HIV treatment coverage for those who need it has not yet been achieved, it would not be ethical to finance a national rollout of PrEP.
They add: "Theoretically, PrEP could compete with treatment scale-up by using limited funding or limited clinic capacity, and this should be avoided.
“Alternatively, PrEP could enable treatment programmes, by allowing greater volume discounts on drug prices and costs, by increasing testing coverage, by fostering retention in treatment by destigmatising ARV drugs and the people who use them, or by fostering popular and political support for attracting more funding to HIV/AIDS initiatives.”
Gomez GB et al. The potential impact of pre-exposure prophylaxis for HIV prevention among men who have sex with men and transwomen in Lima, Peru: a mathematical modelling study. PLoS Medicine 9(10): doi:10.1371/journal.pmed.1001323. 2012. See open access paper here.