A study from the Barcelona Checkpoint gay men’s HIV clinic, presented at the recent 15th European AIDS Conference has found a strong and consistent relationship between the annual HIV infection rate (incidence) in different categories of clinic users and the particular combination of risk factors they self-report. The study found that having casual partners, multiple partners, condomless sex, taking the receptive role in anal sex or catching a bacterial STI all had strong and independent associations with subsequent HIV infection and that the more of these largely self-reported risk factors a clinic user had, the more likely they were to become HIV positive.
The primary intention behind this study was to estimate what proportion of clinic users might benefit significantly from PrEP. In doing the exercise, however, the clinic also managed to establish interesting new estimates for the relative contribution to HIV risk of condomless sex, STI infection, high partner numbers, and sexual role.
Cost-effectiveness studies of PrEP unanimously find that the biggest single influence on whether PrEP will be cost-effective or even save money over time is the HIV infection rate (incidence) in the people offered PrEP – the proportion of people who will be infected with HIV over the next year unless they are offered PrEP. If PrEP is taken by groups at low risk of HIV it is not cost-effective: as the vast majority would not have caught HIV anyway, the money saved by avoiding the cost of future HIV treatments does not match the money spent on PrEP.
Studies of PrEP in gay men including recent ones from the US and Brazil suggest that PrEP, initially at least, may be cost saving, because the people most likely to come forward for it are those at the highest risk of HIV infection. But it is very difficult to prove a person's future risk of HIV on the basis of their past risk.
Nonetheless, it is important to make the best estimates possible of this so-called ‘background incidence’ as it helps establish criteria for who should and should not be offered PrEP, and give guidance to national health systems about how many people are likely to need PrEP, how many HIV cases would be prevented by it, and what it would cost. The World Health Organization (WHO) in its latest guidelines estimated that PrEP would be cost-effective in populations with a background HIV incidence of over 3% a year, and UK studies suggest cost-effectiveness at an incidence of about 5% due to the higher cost of drugs in a high-income country.
In randomised studies like the PROUD trial, the incidence seen in the control arms not taking PrEP is assumed to be equivalent to the pre-study background incidence in participants. In PROUD this was 9% and in the French Ipergay study 6.75%. One HIV infection was prevented for every 13 people given PrEP in PROUD and every 18 in Ipergay, and this means that if background incidence is the same as that seen in these studies PrEP should save money over a relatively short period of time: more so if some are taking PrEP intermittently, as in Ipergay.
The problem is that no-one knows how to predict whether people coming forward for PrEP will have this level of HIV risk. The database of UK sexual health clinic patients (GUMCAD) cannot account for the kind of HIV incidence seen in PROUD. Even if the relatively small group of gay men who have high numbers of condomless sex partners and acquire rectal STIs is selected, incidence is no more than 5.2%. This may be because the GUMCAD data is national and high-incidence groups for HIV have something very specific about them. In particular, they need to be in touch with each other – digitally, or geographically.
This is why single-centre studies conducted in large clinics like Checkpoint that serve city-centre gay scenes may uncover more detail and isolate higher-incidence groups that should be first in line for PrEP.
The Checkpoint study
This study found that, depending on five different variables – relationship status, number of partners, condom use, sexual role, and STI status – the observed annual HIV incidence in different groups of clinic users varied from below 1% a year to an extremely high 25% a year.
It found that HIV incidence was higher than the 3% WHO threshold in all patients who were not in monogamous relationships unless (in the previous six months) they were exclusively the insertive partner, always or usually used condoms, and had not caught an STI.
It found that HIV incidence was over 5%, which in one cost-effectiveness study based on PROUD found cost savings even if PrEP was only 64% effective, in all Checkpoint users who said they ‘only sometimes or never' used condoms, and even in those who said they 'usually' used condoms if they had had more than 20 partners in the last six months.
By establishing these different incidence rates, the study is also able to give new estimates for the relative contributions of multiple partners, condomless sex, STIs and sex role to HIV risk.
The total number of patients seen at Barcelona Checkpoint from 2009 to 2014 was 9780 of whom 5430 had at least two visits; this latter group contributed 12,192 patient-years of follow-up data. During this time 306 of the 5430 became HIV positive, an overall annual incidence rate of 2.5%.
The researchers first looked at the relationship status of those diagnosed with HIV. In the entire clinic user population, 17.5% of men said they were in a steady, monogamous or committed relationship and in these men HIV incidence was 1.3% a year or one infection per 77 men per year. The rest were either in various forms of regular open relationships (36%) or only had casual/occasional sex (46%). HIV incidence in these groups was respectively 2.4% and 2.5% – about one infection respectively per 42 and 40 men per year or 83% and 87% higher than HIV incidence in committed couples.
The investigators then concentrated on a specific subset of clinic users. Excluding those in a steady relationship, they looked at HIV incidence only in clinic users for whom they had complete data on condom use, STI diagnosis, partner numbers and sex role in the previous six months. They furthermore excluded men who had fewer than ten partners in the previous six months and also ones who said they 'always' used condoms. This left a total of 739 men.
The researchers combined the men who said they sometimes, almost never or never used condoms in one 'sometimes' group and compared incidence with those who said they almost always or usually did in a 'usually' group. They also compared incidence in those who said they always took the insertive role in sex with those who sometimes or always took the receptive role. They compared incidence in all men with those who reported more than 20 partners in the previous six months. And they compared incidence in those diagnosed with STIs in the previous six months with those who were not.
This meant there were 16 different groups of men in whom HIV incidence was compared, though some were relatively small groups numerically, as in the table below (click on the image to see it full size).
Of the 739 men 84% said they 'usually' used condoms and the others (121 men) said they 'only sometimes or never' did (there were very few men in the 'never' and 'almost never' categories). Forty-one per cent of the men said they only ever took the insertive role in sex while the other 59% were receptive or versatile. All men had more than ten partners in the last six months, but 52% had had more than 20. And 7% (54 men) had been diagnosed with a bacterial STI in the last six months.
Both the proportion ‘usually’ using condoms and the proportion saying they were always ‘top’ are a lot higher than those found in some other studies of gay men. This may be due to ‘social desirability’ bias, but in a sense it does not matter if men are under-reporting their condomless and receptive sex: the study nonetheless found a strong relationship between self-reported risk and HIV incidence, showing it could be used as a predictor.
The ‘Incidence rate’ figures in the fourth column vary tremendously, as can be seen, and each risk factor raised incidence. Firstly, having an STI more than doubled the risk of subsequent HIV infection: across the whole group it raised the risk by 122%, and in men who were sometimes receptive and only sometimes used condoms, it raised an annual incidence already at least as high as that in PROUD to a near-unprecedented 24.7%. The only group in which STI infection did not increase risk were receptive men who usually used condoms. However the numbers diagnosed with STIs in individual risk combinations were small and confidence intervals wide, so the exact figures may not be reliable.
Condom use was the variable most strongly related to HIV incidence. And it had a particularly strong protective effect, interestingly, in men who only took the insertive role. ‘Top’ men who did not often use condoms were at the same risk as ‘versatile or bottom’ men who did use them. Having condomless sex more than doubled the risk of HIV in men who were receptive – but it raised the risk of HIV sixfold in ‘tops’.
What about the converse? What difference did sex role make in itself? In men who rarely protected themselves with condoms, being versatile or bottom, as opposed to exclusively top, raised the risk of HIV by 66%. But in men who usually used condoms, being versatile or bottom was more than four times (344%) riskier than being exclusively top.
There is a similar pattern when it comes to numbers of partners. In the subgroup of men who had more than 20 partners in the last six months, HIV risk was raised by 26%. But having a larger number of partners almost doubled the risk of HIV in men who were only tops who did not often use condoms – or in tops who usually used condoms but had an STI.
In other words, HIV risk factors are in general additive but each individual risk factor has a much bigger impact at the bottom end of the risk scale than the top – when it came to the group with more than ten partners in six months, only exclusive tops who always used condoms and had no STIs had an HIV incidence lower than the clinic average. Any deviation from this meant that one’s background incidence jumped over the WHO PrEP threshold and in anyone who only used condoms occasionally rather than usually, their incidence was above the cost-effectiveness threshold of even a 64% effective PrEP programme.
HIV prevalence in Spain is in general higher than in the UK but the EMIS survey and other studies have found that HIV prevalence in Barcelona in gay men, at about 12%, is about the same as it is in London, so there is no reason to suppose that a similar London clinic survey would produce substantially different results.
This survey only included men in whom every behavioural risk was documented so may include a large number of men at equivalent risk. But at the very least this survey does establish that there is a sizeable proportion – maybe two-thirds of those with more than ten partners in six months – for whom PrEP would be a cost-effective measure right now.
Meulbroek M et al. Cohort of MSM as a useful tool to assure cost effectiveness during the implementation of PrEP. 15th European AIDS Conference, Barcelona. Abstract PE18/14. 2015. See here for conference programme and abstracts.