HIV treatment as prevention among men who have sex with men in the UK: is transmission controlled by universal access to HIV treatment and care?
HIV and STI Department, Public Health England, Centre for Infectious Disease Surveillance and Control, London, UK
Correspondence: Dr Alison E. Brown, HIV and STI Department, Public Health England, Centre for Infectious Disease Surveillance and Control, 61 Colindale Avenue, London NW9 5EQ, UK. Tel: 00 44 208 3277559; fax: 020 82007868; e-mail: email@example.com
In the UK, free HIV care is provided through dedicated HIV clinics. Using the national cohort of men who have sex with men (MSM) with diagnosed HIV infection and estimates of the number of undiagnosed men, we assessed whether high retention in HIV care and treatment coverage is sufficient to reduce HIV transmission.
Antiretroviral therapy (ART) uptake and viral load distribution among diagnosed men were analysed by treatment status and CD4 count for the period between 2006 and 2010. A multi-parameter evidence synthesis (MPES) method was used to estimate the size of the undiagnosed population. The viral load distribution among newly diagnosed untreated men was applied to the undiagnosed population. Infectivity was defined as a viral load > 1500 HIV-1 RNA copies/mL.
Between 2006 and 2010, ART coverage among all HIV-infected MSM (diagnosed and undiagnosed) increased from 49 to 60%, while the proportion of infectious men fell from 47 to 35%. Over the same period, the number of all HIV-infected MSM increased from 30 000 to 40 100 and the number of infectious MSM remained stable at 14 000. Of the 14 000 infectious MSM in 2010, 62% were undiagnosed, 33% were diagnosed but untreated, and 5% received ART. Extending ART to all diagnosed HIV-infected MSM with CD4 counts < 500 cells/μL in 2010 would have reduced the overall proportion of infectious men from 35 to 29% and halving the proportion who were undiagnosed would further have reduced this to 21%.
High ART coverage in the UK has reduced the infectivity of the HIV-diagnosed population. However, the effectiveness of treatment as prevention will be limited unless the undiagnosed population is reduced through frequent HIV testing and consistent condom use.
The demonstration that prevention of HIV transmission was possible through antiretroviral therapy (ART) was described as ‘a defining moment for HIV control’ in mid-2011  and the ‘breakthrough of the year’ for scientific research in December 2011 . ART has been shown to be highly efficacious at preventing transmission from HIV-infected to HIV-negative heterosexual partners, providing that the HIV-positive partner adheres to treatment so that their viral load becomes undetectable .
While the benefit of HIV treatment for HIV-negative heterosexuals in HIV-serodiscordant relationships has been demonstrated, there is uncertainty surrounding the clinical benefits of ‘early’ ART (starting ART before the CD4 count reaches 500 cells/μL) for HIV-positive individuals. The START trial  is designed to identify the impact of early ART initiation among the HIV-diagnosed population on the development of AIDS-related and non-AIDS-related illnesses. Additional measurable outcomes include adherence, the development of antiretroviral resistance and patient satisfaction. This work will directly inform future national and international treatment policies.
However, even if early treatment for the diagnosed HIV-positive population is found to be clinically beneficial, there is continued uncertainty about a population-level prevention effect  that could be achieved if ART coverage were high enough to reduce ‘community viral load’ (CVL) . This challenge appears to be greatest in the USA where, in 2008, only 63% of HIV-infected men who have sex with men (MSM) were established in care following HIV diagnosis and 48% retained in care year on year . Consequently, it was estimated that only 19% of the 1.1 million adults living with HIV in the USA had an undetectable viral load, reducing their risk of transmission. It is suggested that if engagement and retention in HIV care following diagnosis were improved so that the 76% of those requiring ART in the USA received it , then the CVL might be reduced below the level needed to sustain transmission [7-9].
The UK is well placed to explore the relationship between ART coverage and transmission. HIV is endemic in MSM, with an estimated prevalence of 9% in London and 3% elsewhere . Provision through the National Health Service (NHS) has enabled free, universal access to HIV testing, treatment and specialized care services. Consequently, in 2010, 98% of all HIV-positive patients were established in care 3 months after diagnosis  and at least 95% of all patients were retained in care year on year . Treatment coverage was high, with over 90% of MSM diagnosed and meeting treatment criteria receiving ART in 2010 .
British HIV Association (BHIVA) treatment guidelines have broadly mirrored US guidelines; in 2008 [14, 15] both recommended ART be initiated before CD4 counts fell below 350 cells/μL. In 2009, US guidelines recommended that treatment began at < 500 cells/μL  and, in 2012, these were further updated to consider ART for all HIV-infected individuals, partially in response to the demonstration of the prevention effect of treatment. A similar change is now indicated in the UK for people who wish to protect their negative partners from infection  and the World Health Organisation has recently recommended treatment to start at < 500 cells/mm3 to prevent transmission .
We used comprehensive HIV surveillance data and estimates of undiagnosed infection to describe the viral load distribution among MSM living with HIV in the UK and to estimate the proportion of men who are at risk of passing on their HIV infection by treatment and CD4 status. We define ‘infectiousness’ as a viral load > 1500 HIV-1 RNA copies/mL . The potential for a population-level reduction in HIV transmission through treatment as prevention among MSM is discussed.
Information relating to the HIV-diagnosed MSM population was obtained from three sources for the years 2006−2010. The first source was new HIV diagnoses  reported by clinicians and laboratory staff. The second was a survey of people accessing care at all HIV out-patient sites ; demographic, exposure and clinical information (including ART uptake, CD4 counts and viral loads) relating to each patient's most recent attendance within a calendar year was also collected. The third source was the CD4 surveillance scheme which monitors levels of immunosuppression among diagnosed HIV-infected patients for over 90% of haematology laboratories . Limited patient identifiers (surname soundex, sex and date of birth) were used to link individual records across data sets across years, and for the sensitivity analysis, to estimate establishment and retention in care.
All records had information about ART, 94% had CD4 information, and 93% had viral load information. The distribution of CD4 and viral load values for the treated and untreated populations were applied to the treated and untreated populations who had these values missing. Establishment and retention in care were assumed to be 100%, with the exception of the sensitivity analysis described below. HIV-positive MSM were defined as men who probably acquired their infection through sexual contact with other men.
Estimating the size of the undiagnosed population
The number of MSM living with an undiagnosed HIV infection has previously been estimated annually using a Bayesian multi-parameter evidence synthesis (MPES) statistical model [22-24]. This technique combines data from an unlinked anonymous HIV seroprevalence survey of MSM attending sentinel sexually transmitted infection (STI) clinics [25, 26] with estimates of the UK MSM population (3.4%) , estimates of the number of diagnosed HIV-infected MSM and data from behavioural surveys. Credible intervals (CIs; 95%) demonstrate the level of uncertainty around these estimates.
Estimating infectivity and community viral load
There are currently no data on the effect of ART on HIV transmission through anal sex. However, it is biologically plausible that treatment will also reduce the risk of HIV transmission between serodiscordant MSM. We take a threshold of viral load > 1500 copies/mL as the definition for infectivity, in line with a meta analysis of transmission risk through sexual exposure .
The number and proportion of infectious men were calculated for those diagnosed, stratified by treatment status and CD4 count (< 350, 350–500 and > 500 cells/μL). The viral load and CD4 distributions among newly diagnosed men who did not receive treatment in a given year were applied to the estimated undiagnosed population. The community viral load was calculated using the median viral load and the interquartile range (IQR).
Three separate sensitivity analyses were undertaken. The first assessed the impact upon community viral load should the ‘upper’ or ‘lower’ estimate of undiagnosed prevalence be taken. The second assessed the impact on community viral load should ’infectivity’ be redefined as having a detectable viral load (> 50 copies/mL). In the third analysis, the results were adjusted to account for the proportion not established in care (2% ) and not retained in care year on year (5% ).
Alternative treatment and testing scenarios
Four scenario analyses were undertaken to explore the impact of changing treatment and testing thresholds on the proportion of infectious HIV-infected MSM. In turn, these assessed the impact on the proportion of infectious MSM if (a) all MSM living with HIV infection received ART according to current national guidelines (i.e. CD4 < 350 cells/μL); (b) all MSM received ART; (c) the undiagnosed HIV-infected population was halved through increased HIV testing (without changing the CD4 threshold for ART); and (d) the undiagnosed population was halved, combined with treating the population with CD4 counts < 500 cells/μL.
Between 2006 and 2010, the number of MSM living with a diagnosed HIV infection increased from 20 800 to 29 900. The estimated number of undiagnosed MSM rose from 9300 (95% CI 6800–12 800) in 2006 to 10 300 (95% CI 5500–16 800) in 2010. Adding the observed number of diagnosed HIV-infected MSM to the estimated number of undiagnosed HIV-infected MSM, the overall number of MSM living with HIV rose from 30 000 to 40 100 over the same period (Fig. 1).
There were no significant differences in demographic and treatment profiles between HIV-diagnosed individuals with complete and missing CD4 and viral load values (P < 0.4). Among MSM with a diagnosed HIV infection, ART coverage was 71% (14 800) in 2006 and rose to 80% (24 000) by 2010. Overall, 60% of all MSM living with HIV (including the undiagnosed) received treatment in 2010 (Fig. 1).
The number and proportion of infectious MSM and community viral load
The estimated number of infectious MSM remained unchanged at around 14 100 between 2006 and 2010 (Fig. 1). In 2010, 35% (14 000/40 100) of all men living with HIV (both undiagnosed and diagnosed) were estimated to have a viral load > 1500 copies/mL and were considered ‘infectious’ (Table 1). This proportion decreased from 47% in 2006 (14 100/30 100).
Table 1. Community viral load and estimated number and proportion with viral load > 1500 and > 50 copies/mL among HIV-infected men who have sex with men (MSM) in the UK in 2010
Diagnosed and untreated
> 500 cells/μL
< 350 cells/μL
*Estimated using multi-parameter evidence synthesis (MPES) methodology .
In 2010, 3% (700/24 000) of treated MSM had a viral load > 1500 copies/mL compared with 79% (4600/5800) among diagnosed MSM not receiving treatment (Table 1). Among the undiagnosed population, the equivalent figure was estimated to be 85% (8700/10 300).
Among the treated population, the median viral load was 40 copies/mL (IQR 39–49 copies/mL), with the same value observed among the overall diagnosed population (IQR 40–52 copies/mL). A median of 12 900 copies/mL (IQR 2300–50 800 copies/mL) was found among the diagnosed, untreated population (Table 1).
Distribution of the population with viral loads > 1500 copies/mL
In 2010 among the 14 000 MSM with viral loads > 1500 copies/mL of whom 62% (8700) were estimated to be undiagnosed. A further 5% (700) received treatment, 16% (2300) were untreated with a CD4 count > 500 cells/μL, 12% (1600) were untreated with a CD4 count between 350 and 500 cells/μL and 5% (700) were untreated with a CD4 count < 350 cells/μL (Fig. 2).
Sensitivity analyses were conducted using the lower and upper CIs of the undiagnosed infection estimate. When the lower 95% CI of 5500 undiagnosed HIV-infected MSM was taken, the overall proportion of infectious MSM was calculated to be 24%, with individuals with undiagnosed infection constituting 48% of the infectious population. When the upper 95% CI of 16 800 undiagnosed HIV-infected MSM was taken, the overall proportion of infectious MSM was estimated to be 48%, with individuals with undiagnosed infection making up 74% of the infectious population.
When adjustments were made to account for the small proportions not established (2%) and not retained in care (5%), 38% (15 400) of all HIV-infected MSM were estimated to have a viral load > 1500 copies/mL in 2010. Following these adjustments, among the 15 400 men with a viral load > 1500 copies, 57% (8700) were estimated to be undiagnosed, 5% (700) to be treated and the remaining 38% (6000) to be diagnosed but untreated.
If a detectable viral load > 50 copies/mL was taken to be ‘infectious’, the proportion of all infectious MSM living with HIV was 46% (18 300) in 2010 (Table 1). Among MSM with a detectable viral load, 54% (10 000) were estimated to be undiagnosed, with 16% (3000) receiving treatment and the remaining 29% (5300) diagnosed but untreated.
Alternative treatment and testing scenarios
In 2010, if all the 40 100 MSM living with HIV infection had received ART according to current national guidelines (i.e. at a CD4 count < 350 cells/μL), the overall proportion (and number) of MSM with viral loads > 1500 copies could have been reduced slightly from 35% (14 000) to 34% (13 600). Extending ART to all with CD4 counts < 500 cells/μL could further have reduced the proportion with viral loads > 1500 copies to 29% (11 600). Halving the undiagnosed HIV-infected population through increased HIV testing (without changing the CD4 threshold for ART) could have reduced the proportion of MSM with viral loads > 1500 copies to 27% (11 000). Halving the undiagnosed population, combined with treating the population with CD4 counts < 500 cells/μL, could have reduced infectivity to 21% (8400).
Between 2006 and 2010, the high and increasing ART coverage reduced the proportion of all infectious men (diagnosed and undiagnosed) from 47 to 35%. Only 3% of those receiving ART had a viral load > 1500 copies/mL compared with 83% of those not receiving treatment. However, over the same time period, the number of MSM living with HIV rose from 30 000 to 40 100 and the absolute number of infectious MSM was unchanged. In 2010, an estimated two-thirds of the infectious population were undiagnosed, suggesting that HIV transmission is largely driven by this population. Consequently, while treatment as prevention is effective at reducing the infectivity of the treated population, it may not be effective at reducing HIV transmission unless it is combined with primary prevention to reduce the size of the undiagnosed population.
Our study provides a ‘snap-shot’ of the distribution of viral load among diagnosed and undiagnosed MSM living with HIV. Its strength relates to our use of observational data which are comprehensive (all MSM accessing HIV care are included) and which have almost complete information relating to ART, CD4 and viral load. These data enable us to explore the relationship between ART coverage and infectiousness within a complete population of HIV-infected MSM.
Only a small reduction in infectivity could be achieved through ensuring all diagnosed MSM are treated according to national guidelines. In 2010, only 5% of the infectious population met treatment guidelines but were not receiving ART and most of these men went on to start treatment within the next 6 months. While more substantial gains could be achieved through expanding ART to MSM with CD4 counts < 500 cells/μL, an even greater impact could be achieved by halving the undiagnosed population and extending ART to all MSM with a CD4 count < 500 cells/μL.
In addition to being infectious, the undiagnosed population is also more likely to have unprotected sex compared with their diagnosed counterparts  and to contain those MSM who are experiencing primary infection (characterized by elevated viraemia ). In the UK, while the coverage of HIV testing is 82% among MSM attending sexual health clinics , only 55% of HIV-negative MSM reported an HIV test in the last year . Further efforts are needed to improve the overall uptake and frequency of HIV testing among MSM.
Our results differ from those reported in other studies [8, 9], which concluded that rising ART coverage was associated with declining HIV transmission by inference from decreasing numbers of new HIV diagnoses against a background of increased HIV testing. While estimates of undiagnosed HIV infection are lower, trends in new HIV diagnoses do not necessarily reflect contemporaneous trends in HIV incidence, and interpretation of these other studies would have been facilitated by data on median age and CD4 count at diagnosis.
In the UK, surveillance data have demonstrated that the annual number of newly diagnosed MSM (2650–2900), their median age (35 years), and their median CD4 count at diagnosis (around 400 cells/μL) have remained stable between 2006 and 2010. This stability suggests that the numbers of new HIV diagnoses and underlying new HIV infections were approximately equal. Furthermore, a back calculation model based on CD4 count at HIV diagnosis in MSM has demonstrated that the number of new HIV infections among MSM has remained consistently between 2000 and 3000 cases .
While the overall mean viral loads in the UK (between 28 300 and 15 900 copies/mL from 2006–2010) and San Francisco (23 300 copies/mL)  were broadly similar, UK ART coverage was probably much higher. Moreover, the extent and completeness of UK data suggest that the estimate may be more accurate compared with a setting where access to care is suboptimal, over-representing the treated population. If mean viral load is lower in the UK compared with San Francisco, then the lack of change in HIV transmission observed so far in the UK is of greater significance.
While we used robust, complete surveillance data, some limitations exist with respect to the population of interest; the certainty on the number of undiagnosed HIV infections; methods for estimating infectivity; the proportions established and retained in care; surveillance data; and the lack of behavioural data. These are discussed in turn.
Firstly, the wide CIs surrounding the estimates of the number of undiagnosed MSM (4000–6500) may mask a true rise or decline in the number of undiagnosed infections. The sensitivity analysis demonstrated that the uncertainty surrounding the estimates is critical to the measurement of the estimated proportion of infectious MSM, which ranges from 24 to 48%, using the lower and upper CIs of the estimates, respectively. However, extended work on the MPES model has indicated that the level of undiagnosed HIV infections was stable between 2001 and 2008 in England and Wales . Furthermore, the method has been found to be robust  and its results have been corroborated with other estimates of the size of the undiagnosed MSM population [22, 30, 31].
The threshold for infectivity was defined as a viral load > 1500 copies/mL for two reasons. Firstly, no sexual transmissions were found to occur below this level in an observational study of serodiscordant heterosexuals . It is biologically plausible that the preventative effect of ART on transmission through vaginal sex is also applicable to anal sex . It is noted that no data exist on the effect of ART on sex between men, and that the risk of HIV transmission through anal sex is higher compared with vaginal sex . However, this does not mean that the transmission threshold of a viral load > 1500 copies/mL is not meaningful for sex between men.
Secondly, if a detectable viral load was defined as the threshold of infectiousness, this would have produced an overestimate of infectivity as it would include MSM currently experiencing transient, low-level blips. This is supported by the sensitivity analysis, which found that changing the threshold of infectivity to > 50 copies/mL increased the proportion of ‘infectious’ MSM who were treated to 16%, compared with 5% using > 1500 copies/mL. It is also noted that other factors, which are not considered here, are also likely to impact upon the risk of transmission, including STI coinfection and insertive and receptive anal sex.
The true infectivity of the undiagnosed population is unknown, and the viraemia among untreated MSM in their first year of diagnosis was used as a proxy. The percentage of infectious MSM among the undiagnosed population is likely to have been higher than our estimate of 85% for two reasons. Firstly, highly infectious primary infections will have been concentrated among the undiagnosed; our estimate does not take this into account. Secondly, newly diagnosed MSM with low CD4 counts will rapidly start ART; the viraemia of newly diagnosed, untreated men will have been lowered by the prompt removal of the most infectious. Consequently, we are likely to be underestimating the infectivity of the undiagnosed population.
The rates for establishment and retention in care are likely to be underestimates because data on out-migration from the UK are not available. Conversely, the data also overestimate retention in care as they do not take account of the 4.9% of MSM who attend intermittently . Regardless, adjustments made for the small numbers not established or retained in care had minimal effect on infectivity in the sensitivity analysis. While estimates were applied from a published study relating to earlier years , 2009–2010 data suggest that retention rates have remained stable at 95%. Furthermore, using exact matching criteria, 92% of all MSM receiving HIV care in 2006 and not known to have died attended for care in 2010; less strict matching would increase this to an estimated 95%. Recent results from the 2013 BHIVA audit suggest that true estimates of loss to follow up are likely to be a maximum of 2.4% .
The methods draw information from an annual survey that links patients over time to create a prospective cohort. As part of the survey methodology, information is collected for the last patient attendance each year. Consequently, information relating to ART reflects the outcome of the clinic attendance. For a minority of patients who stopped ART temporarily or permanently at the clinic attendance, viral loads will reflect the recent treatment received. This may account for the 9% of untreated diagnosed individuals with a CD4 count > 500 cells/mL and the 3% of the estimated undiagnosed population (Table 1).
Sexual behaviour, specifically the frequency of unprotected anal sex and partner change, will also affect rates of ongoing transmission. This observational study was unable to account for this. While there is still uncertainty about the impact of diagnosis on long-term risk behaviour, previous studies indicate that HIV-infected MSM who are unaware of their infection are more likely to have higher risk behaviours compared with men with diagnosed infection . Consequently, the high proportion of infectious MSM who are undiagnosed are of greater concern.
In conclusion, we have demonstrated that widespread ART coverage reduces HIV infectivity at a population level. It is likely that large-scale ART coverage has had an impact on transmission, and that HIV incidence would have been higher without it. However, despite universal access to ART and care in the UK, treatment as prevention is unlikely to decrease HIV transmission at the population level unless the undiagnosed population can be substantially reduced through increasing both the coverage and frequency of HIV testing.
We thank all the clinicians and data providers for their efforts in reporting high-quality and timely data, without which these analyses could not have been conducted.
We also thank the staff of the HIV and STI Department, Health Protection Agency.
Particular thanks to Anne Presanis for giving permission to use the results of the MPES analysis and, together with Gary Murphy, for commenting on earlier drafts of the paper.
Conflicts of interest
The authors have no conflicts of interest to report.
AB had full access to the data and together with VD is responsible for its integrity. AB developed the idea, designed and conducted the analysis and undertook the main writing of the manuscript. ONG helped to develop the idea scientifically and from a public health perspective, and substantially contributed to the writing. VD helped develop the idea scientifically and from a public health perspective, contributed towards the methodology design and substantially contributed to the writing.