Conventional serosurveys are unable to determine the duration of HIV infection and therefore our knowledge of trends in HIV incidence over the last 10–15 years has been based largely on successive point prevalence measurements. Monitoring HIV incidence and identifying recently acquired infections are nevertheless of key importance both on a population and an individual basis: first, it enables recent developments in the epidemic to be tracked and timely interventions to be made to reduce transmission; second, it may allow recognition of local epidemics; third, it means individual treatment strategies can be adjusted according to whether infection is recent or longstanding; fourth, it simplifies tracing the sexual partners of newly diagnosed cases if the number to be traced can be reduced through knowing the interval since HIV was contracted.
STARHS is a technique that identifies specimens from individuals recently infected with HIV by exploiting the difference in sensitivity between a highly sensitive anti-HIV enzyme immunoassay and one that has been modified to diminish its sensitivity in a defined manner (detuned). Typically, the ‘detuned’ assay becomes reactive 4–6 months later than the sensitive screening assay, although this interval can be adjusted by using different assay thresholds. Only specimens previously confirmed to be anti-HIV-1 positive by a stringent anti-HIV confirmatory algorithm are tested in the ‘detuned’ assay. When coupled with appropriate information on the population that tested negative for HIV infection the STARHS assay can be used to measure HIV incidence in defined populations [1–3].
It is thought that HIV transmission in England and Wales due to sex between men was undiminished throughout the first half of the 1990s, and that the rate may have increased between 1995 and 1997 when there was an upsurge in homosexually acquired gonorrhoea [4,5]. It has been postulated that the introduction of highly active combination antiretroviral therapy (ART) in 1996/1997 should have led to a decline in HIV incidence because the decrease in the viral load of patients on combination ART would be expected to make them less infectious . However, modelling suggests that while usage of ART may decrease incidence, even modest increases in risk behaviour could overcome any beneficial effect on HIV transmission , as has been observed in San Francisco . In the light of this uncertainty, it is important to monitor HIV transmission in a substantial sample that is minimally biased.
In England and Wales there is a highly developed service of walk-in clinics where most cases of sexually transmitted infections (STI) are diagnosed and treated . This is particularly so for STI in subgroups of the population at greatest risk of HIV, such as men who have sex with men (MSM). An estimated 95% of HIV infections amongst MSM are diagnosed through STI clinic attendance (personal communication, J. Mortimer, Communicable Disease Surveillance Centre, UK). Using discarded routine syphilis serology specimens from MSM attending STI clinics as part of an ongoing large-scale unlinked anonymous survey of HIV infections  we sought to estimate HIV incidence in at risk MSM between 1995 and 2001. Had the widespread introduction of effective ART significantly affected the incidence of HIV in the population of MSM, we would expect to observe this effect, especially in those MSM attending STI clinics and included in our national sample.
Abbott 3A11 assay
All specimens were tested as previously described . Briefly, a 1 : 20 000 dilution of each anti-HIV-1 positive specimen was made in assay diluent, of which 200 μl was incubated with an antigen-coated bead at 40°C (± 1°C) for 30 min (± 1 min). A peroxidase anti-human IgG conjugate with an OPD substrate was used to detect bound antibody. The optical density (OD) given by each specimen was determined at 492 nm. A Parallel Processing Centre and 3A11 assay kits (Abbott Laboratories, Maidenhead, UK) were used. Each run incorporated triplicate tests on high and low positive control specimens and a calibrator specimen supplied by the Centers for Disease Control and Prevention (CDC), Atlanta, Georgia, USA. Standardized optical densities (SOD) were determined for each of the specimens by dividing the specimen OD by the median OD of the calibrator specimens after the subtraction of the median Abbott negative control OD from both. If the SOD of a test specimen was < 1.5, it was re-tested in triplicate, each replicate from individually prepared dilutions. A median SOD of < 0.75 from the re-tests signifies seroconversion within the previous 129 days (95% confidence interval, 109–149 days).
Organon Teknika Vironostika Assay
As for the Abbott assay, the OTV assay (Organon Teknika Vironostika anti-HIV 1 assay, bioMérieux, Cambridge, UK; product code 259605) required a 1 : 20 000 dilution of each specimen to be prepared. A three-step dilution process using Dilsum II diluent was used, and 200 μl of diluted specimen was added to a designated well of the kit's HIV antigen coated microtitre plate and incubated at 37°C (± 1°C) for 30 min (± 1 min). Following four washes, 150 μl of peroxidase labelled antibody was added to each well and incubated at 37°C (± 1°C) for 30 min (± 1 min). The plate was washed as previously described and 150 μl substrate added. The plate was incubated at 25°C (± 1°C) for 11–13 min and 150 μl of stop solution was added. Absorbances were read at 405 nm. Similarly to the Abbott 3A11 assay, a calibrator specimen, a low positive control and a high positive control (all supplied by CDC) were run in triplicate in each assay, along with the internal kit controls. An SOD was calculated for each specimen tested, as described for the Abbott 3A11 assay. Any specimen that gave an SOD < 2.00 was retested in triplicate, the dilutions for each replicate being prepared independently. A threshold of a median retest SOD of 0.75 was used for the OTV assay because the associated window period of 133 days (95% confidence interval, 127–140 days) during which recent infection is inferred has been shown to be close to that of the Abbott assay (129 days). This permitted annual incidences to be calculated even though not all specimens from any particular year were tested by the same assay.
An Unlinked Anonymous HIV Prevalence Serosurvey (UAPS) of the residues of specimens collected for syphilis serology from attendees at 15 representative STI clinics of the approximately 200 clinics in England, Wales and Northern Ireland has been in place since 1990 . A syphilis test is recommended for all first time attendees at STI clinics who are sexually active and for all new episodes of suspected STI . However, syphilis testing practice may vary between clinics and a voluntary confidential HIV test is not necessarily offered at the same time. Information accompanying each survey specimen included exposure category, age group, sex, whether or not HIV infection had been diagnosed, and whether the patient had an acute STI, or was known to have AIDS at the time the specimen was collected.
Between 1995 and 2001 a total of 43 100 leftover serum specimens from MSM who attended participating STI clinics were tested for anti-HIV, and 3565 were confirmed to be anti-HIV-1 positive.
All available anti-HIV-1 positive specimens were tested using the STARHS assay. However, specimens from individuals whose HIV infection had previously been diagnosed (n = 1740) were excluded from the incidence calculation as they could have been receiving ART, and this may affect the specificity of the assay. Specimens from 21 individuals with an AIDS defining illness were also excluded from the incidence calculation as they, too, may be falsely classified as recent by the STARHS assay . Specimens collected during the period 1995–1999 were tested predominantly using the Abbott 3A11 assay. However, the withdrawal by Abbott of the 3A11 assay meant that a small number of specimens from each of these years, and all specimens from years 2000 and 2001, could be tested only by the OTV assay. For ‘means of incidence’ calculations the STARHS seroconversion window period used was that associated with the most used assay for a particular year: 129 days for 1995–1999; and 133 days for 2000 and 2001.
During each year some anti-HIV-1 positive specimens were unavailable or insufficient for further testing. In total, 159 specimens eligible for STARHS studies were unavailable for testing, almost all from the first 3 years of the study. The missing specimens were allocated as either recent or long-standing infections in the same proportion as those specimens available for STARHS testing. The assignment of specimens unavailable for testing was performed within the age group and testing centre of the specimen. The following proportions of specimens were unavailable: 1995, 92/325 (28.3%); 1996, 42/279 (15.1%); 1997, 16/268 (5.9%); 1998, 2/267 (1.5%); 1999, 1/196 (0.5%); 2000, 2/195 (1.03%); 2001, 4/274 (1.5%).
Annual incidence is given by the formula: I = (n/N) (365/T) (100), where I is the annual incidence, n is the number of recent infections, N is the susceptible population (the recent infections plus those testing negative for anti-HIV) and T is the mean number of days between seroconversion in the sensitive and less sensitive assays . The results were also adjusted to allow for a 10% quarterly repeat attendance rate among those testing negative for HIV .
All statistical analysis was performed in STATA 7 (StataCorp. 2001; Stata statistical software release 7.0, College Station, Texas, USA). National data on uptake of ART , AIDS mortality  and diagnoses of gonorrhoea in MSM  were used to aid interpretation of the HIV incidence findings. Gonorrhoea diagnoses were selected as a marker of risky sexual behaviour because of its much higher incidence than syphilis in the UK throughout the study period, and thus its relatively higher sensitivity for this purpose.
The overall annual HIV incidence in MSM attending STI clinics ranged from 1.5% in 1999 to 3.3% in 1996, with a median of 2.4% (Table 1). There was no significant upward or downward trend. For London clinic attendees HIV incidence varied from 1.7% in 1999 to 3.9% in 1996 (Fig. 1), with a median of 3.0%. Apart from 1995 and 1997, when rates were similar, the London incidence was usually two or more times greater than for clinic attendees outside London, where the median was 0.97% (Fig. 1).
Recent infections were identified in all age groups (Fig. 2) although, due to small numbers, data are not shown for those aged less than 20 years, and age data was not available for 50 men (2.7%). The highest incidence (4.2%) was identified in those aged 35–44 years in 1998. In 2001, the highest incidence was again found in those aged 35–44 (3.2%) and the lowest in those aged less than 20 years (1.4%). In most years those aged between 25 and 44 years had the highest incidence.
When specimens from the 21 patients with an AIDS defining condition at the time of their HIV diagnosis were tested one (4.8%) gave a result indicative of recent infection.
Of the 1804 HIV infected individuals whose specimens were eligible for this unlinked anonymous HIV incidence study 1154 (64%) of them remained undiagnosed after their clinic visit and, of these 462 (40%) had an associated acute STI.
Other surveillance findings relevant to the interpretation of the HIV incidence data are summarized in Fig. 3 [10,14,15]. The number of cases of gonorrhoea diagnosed in MSM increased from 1836 in 1999 to 3509 in 2001 . In addition, diagnoses of infectious syphilis have increased in this group from under 30 cases each year between 1995 and 1997, to over 350 cases in 2001. AIDS mortality in MSM dropped significantly from 1995 to 1998 and has remained relatively stable since then. From 1997 to 2001 it is estimated that less than 5% of those ever diagnosed as HIV infected and living with HIV in England and Wales failed to attend for care in a given year  and over 60% of HIV-infected MSM who had had their HIV infection diagnosed were receiving ART during this period (Fig. 3) .
Our findings provide several key insights into the recent evolution of the HIV epidemic in MSM in England, Wales and Northern Ireland. The most important finding is that the hope that the increasing use of effective combination ART since 1996 would lead to a significant reduction in the transmission of HIV has proved unfounded. Our study shows that during a period of increasingly widespread use of effective ART the HIV incidence in MSM attending STI clinics has persisted at between two and three per hundred per year.
A second finding is that, although HIV incidence has been greatest within London, incident infections have been found elsewhere throughout the study period. In the most recent year (2001) the respective HIV incidences inside and outside London were 3.1% and 1.0%.
Third, contrary to general expectations, recent HIV infections have been occurring at similar levels in MSM of all age groups between 20 and 44 years. The general opinion has been that continuing HIV transmission is principally due to younger MSM, less aware of the safe sex messages of the late 1980s and early 1990s, adopting high-risk behaviours and practices . In most years, however, incidence has been highest either in those aged 35–44 years or those aged 25–34 years. Our data indicate that interventions to reduce risky behaviour and practices need to be intensified nationwide and be targeted at older, as well as younger, MSM.
Last, the increasingly effective use of ART has coincided with a very substantial increase in the number of cases of homosexually acquired gonorrhoea and syphilis, particularly since 1999 [15,17]. This suggests that among high risk MSM any fall in HIV transmission as a result of reduced infectivity due to ART has been offset by an increase in high-risk behaviour. If, as has been proposed , only a small increase in risk behaviour is needed to negate the benefits derived from ART then our data suggest that a further increase in the level of high-risk behaviour may lead to future increases in HIV incidence.
A number of factors suggest that our study sample was sufficiently representative to allow cautious extrapolation of our findings to the population of all MSM at risk of HIV infection. High-risk behaviour in MSM has been associated consistently with STI clinic attendance [18,19] and one-third of MSM surveyed in the 2000 National Survey of Sexual Attitudes and Lifestyles had attended an STI clinic in the 5 years prior to interview (personal communication, C.H. Mercer, A.M. Johnson, K.A. Fenton, B. Erens, K. Wellings, Royal Free and University College Medical School, UK 2003). These findings are consistent with observations that over 95% of clinical diagnoses of HIV in MSM are reported from the network of STI clinics in the country and, where this information is available, a very high proportion of reported HIV-infected MSM attended an STI clinic prior to the visit at which their HIV infection was diagnosed (personal communication, J.Y. Mortimer, Communicable Disease Surveillance Centre, 2003).
The 15 STI clinics that have been included in the UAPS are broadly representative of clinics that are attended by MSM both inside and outside London. The general increase in gonorrhoea diagnoses in MSM during the period 1998–2001 was seen also in the 15 survey clinics. The increase in syphilis testing in 2001 in response to outbreaks could have changed the sampling frame so that men at lower risk of HIV infection became more likely to be included in the survey, leading to a fall in the observed HIV prevalence. However, the prevalence of previously undiagnosed HIV infection did not fall despite a 20% rise in the number of men included in the sample . This steady prevalence in 2001, despite a larger sample, suggests that the population surveyed remains representative of those MSM who are behaviourally vulnerable to HIV infection. Although our prevalence monitoring survey was not designed to monitor HIV incidence directly, it is difficult to imagine a better design with which to apply the STARHS technique to measure changes in HIV incidence in a large group of MSM that were broadly representative of those at greatest risk of HIV infection.
The STARHS technique is well suited to application to specimens from UK MSM attending STI clinics as part of the UAPS. Firstly, the design of the UAPS minimizes participation bias associated with voluntary confidential testing: barring the very few objections to allow their specimen to be entered into the UAPS, all serum specimens collected for syphilis screening are included. Consequently, the population is not skewed by willingness to have a voluntary confidential HIV test, includes both clinically diagnosed and undiagnosed infections, and provides a more accurate picture of HIV prevalence and incidence than studies based on voluntary confidential testing. Furthermore, the attached demographic information allowed us to exclude specimens which may otherwise have led to an overestimate of incidence, such as those from patients receiving ART and those with an AIDS defining condition, whose specimens have been associated with false-positive HIV incidence results . Because the actual numbers of MSM attending each participating STI clinic and receiving a negative HIV test are known, we have an accurate denominator for the calculation of HIV incidence. Finally, STARHS is fully validated for testing specimens from individuals infected with HIV-1 subtype B, and previous studies have shown that the HIV infections in MSM in England and Wales are almost exclusively of that subtype .
The use of two different assays within the STARHS algorithm was regrettable but unavoidable because of the lack of availability and subsequent withdrawal of the Abbott 3A11 assay. Unpublished studies by the CDC have shown each assay to have similar performance characteristics. Unfortunately, a small proportion of specimens were unavailable for incidence testing and this could have led to a bias in our results, especially if a number of recent infections occurred in a particular subset of the missing specimens. We have assumed that there was no difference between the number of recent infections in those specimens available for testing and those unavailable, and have reassigned missing or insufficient specimens as recent or long-standing infections in the same proportions as those observed for specimens available for testing from the same age group and centre, rather than using an overall recent infection rate. However, had all the missing specimens been recent and come from the same subset of the population, this would have altered our incidence estimates. By reassigning specimens in the proportion seen in a matched peer group allocated by centre and age group, rather than according to the overall incidence rate we believe we have minimized the risk of bias in our estimates. In any case, when we performed the same analysis without reassigning the specimens there was no significant difference in our results (data not shown).
The factors that most drive the continuing epidemic of HIV infection among MSM in the UK have yet to be elucidated. Factors that are likely to counter the potential benefits of ART at a population level and contribute to the persisting high HIV incidence include large increases each in the prevalence of HIV infection, high-risk sexual behaviour, and incidence of other STI that facilitate HIV transmission. However, the single most important factor may be the continuing existence of substantial numbers of undiagnosed HIV infections. Direct estimates suggest that in 2001 approximately 4200 (22%) HIV-infected MSM in the UK were unaware of their infection .
From a public health perspective diagnosed infections should represent a reduced risk overall of transmission to uninfected sexual contacts. Diagnosis provides access to advice on safer sexual behaviour and to ART capable of reducing subjects’ plasma HIV RNA concentration to undetectable levels that, in turn, reduces infectivity. The clinical effectiveness of ART is evident in the considerable reductions in AIDS mortality (Fig. 3), especially since the introduction of protease inhibitors from 1997 onwards, and provides indirect support for the notion that diagnosed and treated cases are likely to be less infectious. Furthermore, those diagnosed cases not on therapy are likely to be those with low or undetectable plasma HIV RNA as they are offered regular plasma HIV RNA monitoring with prompt ART should the need arise. Consequently, with few exceptions, diagnosed cases should present substantially reduced infectious hazard. The exceptions include individuals who elect not to receive ART when indicated, those who develop elevated HIV RNA levels between HIV RNA measurements, and those whose treatment fails due to resistance, poor pharmacokinetics, or low drug adherence. Rarely, there may be cases with discordantly high levels of HIV RNA in semen  and infection with an acute STI may lead to a temporary increase in semen viral load, increasing infectiousness, although data exist to indicate that this may not occur in patients receiving effective combination ART .
On the other hand, even if ART and behavioural change among those whose HIV infection has been diagnosed were highly effective at reducing transmission, there exists a significant pool of MSM whose HIV infections remain undiagnosed. They cannot benefit from measures that stem from early diagnosis and the optimal clinical care that would also reduce the infection risk they present to their HIV negative partners. Over 60% of MSM with prevalent diagnosed HIV infection, that otherwise might present an increased infectious hazard, are receiving some form of effective ART. We therefore speculate that a similar proportion of undiagnosed infections may be at increased risk of passing on their infection and would be in need of ART and behavioural risk reduction to ameliorate this risk. As they will mostly be unaware of their HIV infection they may be less likely to adopt less risky behaviour and cannot ensure that their partners are of concordant HIV status. Indeed, among those in our survey that remained undiagnosed after their clinic visit 40% attended with a concomitant acute STI, indicating recent high-risk behaviour, and these cases represent crucial lost opportunities to diagnose HIV infection. Included among the undiagnosed are those with acute primary HIV infections whose typically high viraemia prior to anti-HIV seroconversion represents both potentially high infectivity  and a hard challenge for diagnosis and prevention.
Our data highlight the continuing need for enhanced surveillance of the epidemic of HIV in MSM in the UK. The use of the STARHS technique, coupled with comprehensive surveillance, allows monitoring of the epidemic in real time and enables failures of prevention programmes to be identified and remedied. Although the use of ART has been associated with a considerable decline in AIDS mortality and new AIDS diagnoses it has not yet been associated with a reduction in the incidence of HIV infection in MSM in the UK. Renewed efforts to promote safer sex and reduce high-risk sexual behaviour in all MSM and to decrease the proportion of HIV infections that are not clinically diagnosed, including those with an acute primary HIV infection, would not only provide the benefits of clinical care and timely therapy to those individuals, but should also have a vital public health impact.
The authors thank A. Johnson and colleagues (UK National Survey of Sexual Attitudes and Lifestyles, NATSAL) and J. Mortimer (Communicable Disease Surveillance Centre) for providing unpublished data that supports interpretation of our findings. We also thank A. Brown, C. McGarrigle, I. Simms and K. Fenton for helpful information, and P. Mortimer, J. Mortimer, B. Evans and N. Macdonald for helpful comments on earlier drafts.
Sponsorship: Supported by the Medical Research Council and the UK Department of Health. N.O. is supported by the Medical Research Council (Strategic grant No G0000775).
1. Murphy G, Parry JV, Gupta SB, Graham C, Jordan LF, Nicoll AN, et al
. Test of HIV incidence shows continuing HIV transmission in homosexual/bisexual men in England and Wales. Commun Dis Public Health
2. McFarland W, Busch MP, Kellogg TA,
Rawal BD, Satten GA, Katz MH, et al
. Detection of early HIV infection and estimation of incidence using a sensitive/less-sensitive enzyme immunoassay testing strategy at anonymous counselling and testing sites in San Francisco. J Acquir Immune Def Syndr
3. Weinstock H, Dale M, Gwinn M, Satten GA, Kothe D, Mei J, et al
. HIV seroincidence among patients at clinics for sexually transmitted diseases in nine cities in the United States. J Acquir Immune Def Syndr
4. De Angelis D, Gilks WR, Day NE. Bayesian projection of the acquired immune deficiency syndrome epidemic. JRSS Series C
5. Hughes G, Simm I, Rogers PA, Swan AV, Catchpole M. New cases seen at genitourinary medicine clinics: England 1997. Commun Dis Rep CDR
6. Vernazza PL, Eron JJ, Fiscus SA, Cohen MS. Sexual transmission of HIV: Infectiousness and prevention. AIDS
7. Blower SM, Gershengorn HB, Grant RM. A tale to two futures: HIV and antiretroviral therapy in San Francisco. Science
8. Katz MH, Schwarcz SK, Kellogg TA, Klausner JD, Dilley JW, Gibson S, McFarland W. Impact of highly active antiretroviral treatment on HIV seroincidence among men who have sex with men: San Francisco. Am J Public Health
9. Johnson AM, Mercer CH, Erens B,
Copas AJ, McManus S, Wellings K, et al. Sexual behaviour in Britain: partnerships, practices, and HIV risk behaviours. Lancet
10. Unlinked Anonymous Surveys Steering Group. Prevalence of HIV and Hepatitis Infections in the United Kingdom 2001.
London: Department of Health; 2002.
11. Janssen RS, Satten GA, Stramer SL, Rawal BD, O'Brien TR, Weiblen BJ, et al
. New testing strategy to detect early HIV-1 infection for use in incidence estimates and for clinical and prevention purposes. JAMA
12. Doherty L, Fenton KA, Jones J, Paine TC, Higgins SP, Williams D, Palfreeman A. Syphilis: old problem, new strategy. BMJ
13. Gupta SB, Gill ON, Graham C, Grant AD, Rogers PA, Murphy G. What a test for recent infection might reveal about HIV incidence in England and Wales. AIDS
14. McHenry A, MacDonald N, Sinka K, Mortimer J, Evans B. National assessment of prevalent diagnosed HIV Infections. Commun Dis Public Health
15. PHLS, DHSS & PS and the Scottish ISD(D)5 Collaborative Group. Sexually Transmitted Infections in the UK: New Episodes Seen at Genitourinary Medicine Clinics, 1991 to 2001.
London: Public Health Laboratory Service, 2002.
16. Hays RB, Paul J, Ekstrand M, Kegeles SM, Stall R, Coates TJ. Actual versus perceived HIV status, sexual behaviors and predictors of unprotected sex among gay and bisexual men who identify as HIV-negative, HIV-positive and untested. AIDS
17. Fenton KA, Rogers PA, Simms I, Maguire H, Catchpole M. Increasing gonorrhoea reports - not only in London. Lancet
18. Dodds JP, Nardone A, Mercey DE, Johnson AM. Increasing high risk sexual behaviour amongst gay men in London 1996-1998: a repeated cross sectional questionnaire study. BMJ
19. Nardone A, Dodds JP, Mercey DM, Johnson AM. Active surveillance of sexual behaviour among homosexual men in London. Commun Dis Public Health
20. Parry JV, Murphy G, Barlow KL, Lewis K., Rogers PA, Belda FJ, et al
. National surveillance of HIV-1 subtypes for England and Wales: design, methods and initial findings. J Acquir Immune Def Syndr
21. Coombs RW, Speck CE, Hughes JP, Lee W, Sampoleo R, Ross SO, et al
. Association between culturable Human Immunodeficiency Virus type 1 in semen and HIV-1 RNA levels in semen and blood: evidence for compartmentalization of HIV-1 between semen and blood. J Inf Dis
22. Sadiq S, Taylor S, Kaye S, Bennett J, Johnstone R, Byrne P, et al
. The effects of antiretroviral therapy on HIV-1 RNA loads in seminal plasma in HIV-positive patients with and without urethritis. AIDS
23. Apoola A, Ahmad S, Radcliffe K. Primary HIV infection. Int J STD AIDS