A PrEP programme targeting 5% of gay and bisexual men in the Netherlands would not only cause big reductions in HIV incidence, but would lead to even greater reductions in gonorrhoea.
In the ’base case’ of the model, gonorrhoea would almost be eliminated, with 2027 only having 3% of the infections seen in 2018, the year the model assumes the PrEP programme was started.
As with all modelling studies, its findings should not be regarded as predictions, but as a test of the assumptions fed into the model; the question it asks is: “If we were able to create these conditions, what might happen?”
Its most important finding is probably that, even if a PrEP programme results in a 75% rise in the likelihood of condomless anal sex in gay men, it would nonetheless lead to falls in gonorrhoea, due to more STI testing and treatment. The idea that PrEP might lead to a rise in STIs in itself is not upheld, unless condom use falls almost to zero.
A lot of the benefit of PrEP is due to PrEP users testing for HIV and STIs more regularly. This study suggests that a third to a half of the benefit of PrEP programmes is due to more testing, with the rest due to PrEP's direct effect against HIV.
The model
Researchers at the Dutch National Institute for Public Health and the Environment (RIVM) built a model based on their most recent (2017) data for HIV prevalence and diagnoses, gonorrhoea incidence, and current condom use within the Dutch gay community. They also factored in the 2017 costs for PrEP drugs and antiretroviral therapy (ART) as well as the three-monthly testing and monitoring a PrEP programme would entail.
In their ‘base case’ they assumed that 75% of Dutch gay and bisexual men eligible for PrEP would start taking daily PrEP on 1st January 2018. (This didn’t happen, of course; but the model follows what might have happened if it had.) Dutch eligibility criteria are as follows:
- At least one casual condomless sex partner in the last six months;
- A diagnosis of rectal or urethral gonorrhoea in the last six months; or
- A steady partner who has HIV and has a detectable viral load.
Seventy-five per cent of men fitting these criteria would total about 5.5% of the Dutch gay male population, or about 11,000 people on PrEP. For comparison, this is one person per 1556 of the entire Dutch population; when the English PrEP IMPACT study fully recruits, its 26,000 participants will represent one person per 2153 of the English population.
It was assumed that PrEP users stop taking it if an annual check-up finds they no longer meet any of the criteria, while gay men meeting the criteria steadily join the PrEP-taking population up till 31 December 2027. The model examined what would happen to HIV and gonorrhoea diagnoses over those ten years, and to the resultant cost and cost-effectiveness of the PrEP programme.
The model simulated four different scenarios. The first is as described above. In the second, a 75% increase in the likelihood of any one man having condomless anal sex on any one occasion of sex was assumed once people started PrEP. (This is not the same as a 75% rise in condomless sex across the board: in practice, once run through the model, it means that an average of eleven occasions of condomless anal sex every three months increases to 13 occasions.)
In the third, there was a cap of 5000 on the number of PrEP prescriptions available, or 2.5% of the entire Dutch gay male population. In this, it was assumed that eligible men start PrEP on a first-come first-served basis, rather than further criteria operating.
In the fourth scenario, the 75% increase in the likelihood of condomless sex was combined with a 5000-place cap.
“In the era of generic drugs, the cost of HIV testing and medical monitoring in PrEP is higher than the cost of PrEP drugs.”
As already mentioned, the model assumed daily PrEP. The cost of PrEP drugs in the model ranged from €30 to €50 a month. One interesting aspect of the model is that, based on the current costs of tests and staff time in STI clinics, the cost of testing and monitoring is actually higher than the cost of PrEP, at €150-€163 every three months. In other words, the total cost of PrEP ranges from €240 to €313 every three months. It was not assumed that STIs were all diagnosed or treated at sexual health clinics; the cost of diagnosing and treating them was shared between sexual health centres, primary care physicians, and other hospital settings, and varied accordingly.
Other inputs included the cost of ART in those who acquire HIV if they do not take PrEP. This is €2641 every three months for the most frequently used regimen in the Netherlands. The cost of HIV care varies: it is higher immediately after infection and then falls again until people become ill, the likelihood of which is derived from current HIV statistics.
The impact of the PrEP programme in terms of quality-adjusted life-years (QALYs) saved was measured over a ten-year time horizon, from 2018 to 2027. As is usual in cost-effectiveness studied, the crucial figure is the ICER or incremental cost-effectiveness ratio, which is the amount of money the intervention (i.e. providing PrEP) costs per person, as compared with not making the intervention (i.e. not providing PrEP). Because it's acknowledged that health improvements are worth spending money on, PrEP is not necessarily expected to save money: for the purposes of this study, an ICER of less than €20,000 per person was deemed to be cost-effective.
The effectiveness of PrEP, for the purposes of this model, was 86%, in line with its effectiveness in the PROUD study.
The four models were run many times to produce various different permutations of scenarios. This resulted in a range of costs, and different degrees to which the models were either cost-saving or cost-effective.
In sensitivity analyses, the parameters were stretched further, to produce ‘best and worst case’ scenarios. Effectiveness was varied from 97% to 64% (the upper and lower bounds of the 10% confidence interval in the PROUD study). A further analysis was done assuming an effectiveness of zero: the reason for doing this was to measure the impact solely due to the increase in HIV and STI testing caused by PrEP, independent of its intrinsic efficacy. Other assumptions that were varied were that the probability of condomless anal sex tripled (up 200%); that there was an 80% reduction over time in the cost of HIV treatment; and that only 50% of qualifying men offered PrEP started it rather than 75%.
The results
When all these inputs were fed into the cost-effectiveness model, the base case scenario produced a 61% reduction in the overall number of HIV infections over the ten years, and a 69% decline in the incidence rate by 2027 relative to 2018. HIV incidence would decline from one infection per 342 per year in 2018 to one per 1075 by 2027, and the total number of new infections averted would be 3486.
Importantly, due to the prevention of secondary infections, HIV incidence would fall by nearly as much in gay men in general as in men taking PrEP. In fact by 2027 HIV incidence would be higher in men taking PrEP than in men not taking it, because they would be the more at-risk group. There would remain a very high-incidence subgroup of men who fitted the criteria for taking PrEP but chose not to take it; but even among them, HIV incidence would fall from 2.2% a year (one infection in 45 men a year) to 1.2% a year (one in 83).
“A lot of the benefit of PrEP is due to PrEP users testing for HIV and STIs more regularly.”
The most striking result, however, and the one that is novel to this study, is that gonorrhoea rates would plummet, by a lot more than HIV. In the base case scenario annual gonorrhoea incidence in men starting PrEP would be 0.782% a year, or one case per 128 a year. But by 2027 it would be 0.023%, or one case per 4348 men a year – a 97% decrease in annual incidence. The incidence of gonorrhoea would drop so much it would mean a considerable drop in the number of men on PrEP, because STIs would nearly disappear as a reason to take it, leaving condomless sex as the main criterion. As a result the number on PrEP would nearly halve, from 11,000 to 6000.
On balance, PrEP would save money; the total cost of the PrEP programme would be €3.7m lower than not doing it. However, it was only cost-saving in 52% of the times the model was run (it was cost-effective 97% of the time). As mentioned above, testing and monitoring, at a cost of €39m over the decade, ended up being considerably more expensive than the cost of PrEP drugs (€22m).
In the risk compensation scenario, interestingly, HIV incidence fell even further than in the standard scenario, down to just one infection per 1191 people a year, or a 70.5% reduction between 2018 and 2027, and a total reduction of 63% in the number of infections over ten years.
Why, seemingly counter-intuitively, did HIV fall slightly more even if condom use also fell? The reason is that more people would both start and stay on PrEP in this scenario (there would be 11,600 PrEP users in 2018 and 7300 in 2027) and the effect of more people staying on PrEP would outweigh the effect of fewer people using condoms.
“Due to the prevention of secondary infections, HIV incidence would fall by nearly as much in gay men in general as in men taking PrEP.”
In the risk-compensation scenario the fall in gonorrhoea cases would still be considerable, but nothing like the 97% in the base case; it would be about 73%, down to one case per 472 men per year in 2027 from one per 128 in 2018.
Because there would be more people on PrEP, the programme would cost more, at €45.5m for monitoring and €25.5m for the medication; because of this, the PrEP programme would cost money, at €3.8m more than not doing it. It would be cost-effective, on average, only 73% of the time, and cost-saving 23% of the time.
Capping the programme to 5000 places meant it cost less, but also produced fewer savings in HIV treatment and monitoring. Annual HIV incidence only fell by 56% between 2018 and 2027 (with 48% fewer cumulative infections over ten years) and gonorrhoea by 75%. The programme would on average save money, in the order of €10.7m over ten years, but largely because it would spend less money on testing and therefore on ART.
Finally, capping the programme to 5000 and assuming a 75% increase in the likelihood of condomless sex would result in the lowest fall in annual HIV incidence (52%), the lowest fall in the cumulative number of HIV infections (46%) and a much smaller fall in annual gonorrhoea cases (28% reduction). The programme would on average save €5.2m but almost as many people would be on PrEP at the end as at the beginning.
As mentioned earlier, one sensitivity analysis assumed that PrEP had zero effectiveness, so that the model isolated the effectiveness of more HIV and STI testing. Under this scenario 1219 HIV infections were still averted over ten years, compared to not offering even ‘placebo PrEP’, so to speak. However this is less than half of the 2530 infections averted in even the capped-and-risk-compensation programme, and one-third as many as in the largest (11,000 plus risk compensation) programme.
If PrEP was 96% effective, the programme would be cost-effective in 97% of the model runs and cost-saving in 63% of runs.
If the likelihood of condomless sex per act increased by 200% (i.e. threefold), this would work out as an increase from 11 to 15 occasions of condomless sex a month per person. Under this scenario, even though more HIV infections would be prevented, more people would remain on PrEP by 2027 and the total costs of the PrEP programme would exceed the costs of HIV treatment and monitoring so much that it would, on average, no longer be cost-effective, with an average ICER of €58,558 per QALY gained. The same would happen if the cost of HIV drugs fell by 80% (though this is a counterfactual scenario in that it would had to have happened from the beginning of 2018 onwards).
Conclusions
This study shows that the cost of PrEP programmes is quite sensitive to other changes. In particular, a fall in HIV treatment costs, by the same extent as PrEP drug costs, could make PrEP uneconomical, as could a large increase in condomless sex.
It is arguable, however, that with condom use already quite low in people eligible for PrEP, the impact of further falls may be limited; small further falls in condom use may actually have the paradoxical effect of increasing the overall impact of PrEP, by making more men eligible for PrEP and therefore preventing more HIV infections than otherwise would have happened.
In terms of the cost of ART, it could be argued that they are unlikely to decrease by as much as PrEP costs in that patented, full-cost combination-therapy drugs are likely to form at least part of ART for the foreseeable future.
The study also shows that in the era of generic drugs, the cost of HIV testing and medical monitoring in PrEP is higher than the cost of PrEP drugs. This suggests that savings should be sought by means such as screening for HIV, STIs and kidney function being done at home, or perhaps less frequently.
The study’s most striking findings are twofold. Firstly, even if PrEP had no biological efficacy, the increase in HIV testing that a PrEP programme entails would reduce HIV infections by a third to a half by itself.
Secondly, under nearly all scenarios in this model and especially in scenarios where further falls in condom use are contained, the increase in STI screening due to the PrEP programme results in falls in gonorrhoea (and presumably other bacterial STIs) that are quite dramatic.