A second UK cost-effectiveness study of pre-exposure prophylaxis (PrEP) has concluded that a medium-sized PrEP rollout programme that started in 2016 and which immediately offered tenofovir/emtricitabine (Truvada) as daily PrEP to 5,000 gay men at risk of HIV infection, would not be cost-effective unless the price of Truvada was cut substantially.
The model is based on the observed HIV infection rates in sexual health clinic attendees, rather than the higher infection rates seen in those not taking PrEP in the groundbreaking PROUD study.
The base-case scenario it presents also assumes that, due to lower adherence and falls in condom use in those taking PrEP, its effectiveness will not be as great in a public provision programme as it has been in recent studies. And it assumes that the cost of HIV treatment will, in future, fall considerably, and at a faster rate than the cost of PrEP.
The model, from researchers at Public Health England (PHE), was presented at their recent conference on 15-16 September.
At present-day prices, it calculates, the cost per person of offering PrEP would exceed the cost of the HIV infections that would occur otherwise, if PrEP is not offered, by £34,000. This is called its incremental cost-effectiveness ratio or ICER. An ICER this size would exceed the usual criterion for cost-effectiveness within the NHS (£20,000-£30,000 per person).
The absolute cost of providing PrEP to 5,000 gay men would be £26.8 million per year. This money would eventually be recouped if cumulative reductions in HIV incidence were maintained – but not for 48 years.
The model and its assumptions
To produce its £34,000 ICER, the PHE model assumed several things, which they predict are reasonable assumptions about a 5000-per-year PrEP programme. These are:
- That daily PrEP is only offered for one year. After this another 5,000 gay men at high risk of HIV start using it and the original 5,000 stop. This is clearly not intended to reflect the likely reality, but is intended to be the simplest way to cope with the dilemma of falling risk discussed below.
- That background HIV incidence in those offered PrEP (i.e. the infection rate if they were not offered PrEP) is 3.3% a year, not the 9% seen in PROUD.
- That the ‘post-PrEP’ HIV infection rate in gay men taken off it is 0.48%, or just under one-seventh of the background incidence rate in year 1.
- That overall PrEP effectiveness is 64%, not the 86% seen in PROUD. This figure was chosen because 64% was the lower bound of the 90% confidence interval in PROUD, i.e. the lowest-likely effectiveness that would be seen were the study to be exactly repeated ten times.
- That behaviour change produces an overall 20% increase in HIV risk over and above this.
- That the programme starts in 2016 and that 5,000 men are immediately placed on PrEP.
- That the price of Truvada PrEP stays at its current level.
- That there are substantial falls in the cost of HIV treatment post-2020.
The model does not, unlike the other one devised by University College London, map out the cost of a developing PrEP programme over a number of years. This is partly because it incorporates one finding that the other model does not: that most gay men only spend a year at a time at the kind of level of risk of HIV infection that would justify offering them PrEP.
The single factor that has the biggest influence on PrEP’s cost-effectiveness is whether you offer it only to people who would otherwise have a high chance of catching HIV. If you offer it to people who are at less risk, you will be giving PrEP to a lot of people who would not have caught HIV anyway – clearly less cost-effective.
The PHE experts noticed two things when they looked at actual HIV incidence rates in gay men attending UK STI clinics, using incidence assays that calculate the time people were infected.
Firstly it was impossible to find any group that had an annual HIV infection rate (incidence) as high as the 9% seen in non-PrEP takers in PROUD. The nearest they got was 5.2%, seen in men who had turned up with a rectal STI infection in the last twelve months. This makes it very difficult to predict, in any public programme, who will and will not be a good candidate for PrEP.
Secondly, they found that gay men, on average, only stayed at the kind of risk that would justify daily PrEP for about a year. After a year, the HIV risk in 50% of them had fallen to the kind of rates that characterise the general gay male population in the UK. After a second year, this has happened to 50% of the remainder. In other words, being at very high risk had a ‘half-life’ of about a year.
This creates a dilemma. On the one hand, it implies that daily PrEP should only be offered for about a year and that clinics would need to have a very precise way of judging risk in order to decide who should be taken off PrEP. On the other hand, there remains a residual risk of HIV infection which, in gay men, is not negligible – it is at most 0.5% a year, but if gay men stay at that degree of risk for the following decade or so and are not on PrEP, then the number of HIV infections acquired in this ‘post-PrEP’ period starts to exceed the number of infections averted during the year or so they were on PrEP. This in turn clearly impacts on cost-effectiveness.
This need only be the case, however, if PrEP is daily and that men continue to take it daily even at times they are not at risk. If PrEP was prescribed to be used ‘on demand’ – as it was in the French Ipergay study – and if men were good at judging their risk and tailored their PrEP use to that risk – as did appear to happen in Ipergay and other studies – then continuing to offer PrEP to low-risk men would not be a problem.
The most contentious assumption the model makes is that risk behaviour would increase by 20% and that this would decrease the effectiveness of PrEP to a degree over and above what would happen if adherence was poor. This does not appear to be borne out by recent studies. In the recent study of PrEP provision via primary care in San Francisco, for instance, condom use fell in 41% of PrEP users but this had no impact on HIV infections – in fact none were seen.
The assumption that adherence needs to be very high for effectiveness to be high is also questionable. In Ipergay, the percentage of sex acts fully covered by the PrEP regime was quite low, at 43%, but this appeared not to have any impact on effectiveness, as the only two men infected had not been on PrEP for some time before their infection.
In short, over a certain adherence threshold, PrEP appears to be a sufficiently robust strategy that neither imperfect adherence nor behaviour change (that in someone not on PrEP would increase their risk) actually impact on the risk of infection.
The effect of alternative assumptions on cost-effectiveness
Clearly, a number of these inputs are not likely to be reflected in what actually happens, and the modellers then put their model through a number of so-called sensitivity analyses to find out what factors would most impact on cost-effectiveness.
The three most important were background incidence, the effectiveness of PrEP, and drug price.
In the model, background incidence has the most impact on cost-effectiveness. If background incidence was as high as it was in PROUD, PrEP would be cost-saving, not just cost-effective, at current prices even if effectiveness was only 44%, as it was in the iPrEx Study. If background incidence was 5%, it would be cost-saving at above approximately 60% effectiveness. And if PrEP effectiveness matched PROUD at 86%, then PrEP would be cost-effective at about 3.3% incidence, which is roughly the annual incidence seen in people who seek post-exposure prophylaxis (PEP) and then subsequently catch HIV. Conversely if background incidence was only 1% a year, PrEP would not be cost-effective even at 100% effectiveness.
If drug price was reduced to a quarter of what it is now, then PrEP would be cost-saving, and it would be marginally cost-saving if PrEP were offered to all as intermittent PrEP, using the same two-before-two-after protocol used in Ipergay.
Behaviour change would have some, but not a huge, influence on cost-effectiveness and if, as expected, the cost of HIV therapy falls substantially after 2020, that would make PrEP uneconomical – but only if the cost of PrEP did not fall at the same rate.
In short, whether PrEP in the UK will be cost-effective or not depends on what one thinks is likely to happen. Will it gradually be taken up, initially by the gay men at the highest risk of HIV infection and then in time, by lower-risk gay men and other people such as at-risk heterosexuals? Or will there be mass adoption of PrEP from the start? In the latter scenario PrEP would be unaffordable without immediate substantial reductions in drug cost. In the former it may even save money, as long as the expected future falls in drug prices keep pace with increased numbers on PrEP.
Ong K-J et al. Will HIV PrEP given to high-risk MSM in England be cost-effective? Preliminary results of a static decision analytical model. Public Health England Conference, Coventry. 2015.