Adamma Aghaizu*, Sonali Wayal*, Anthony Nardone, Victoria Parsons, Andrew Copas, Danielle Mercey, Graham Hart, Richard Gilson, Anne M Johnson

Transcription

1 Sexual behaviours, HIV testing, and the proportion of men at risk of transmitting and acquiring HIV in London, UK, 3: a serial cross-sectional study Adamma Aghaizu*, Sonali Wayal*, Anthony Nardone, Victoria Parsons, Andrew Copas, Danielle Mercey, Graham Hart, Richard Gilson, Anne M Johnson Summary Background HIV incidence in men who have sex with men (MSM) in the UK has remained unchanged over the past decade despite increases in HIV testing and antiretroviral therapy (ART) coverage. In this study, we examine trends in sexual behaviours and HIV testing in MSM and explore the risk of transmitting and acquiring HIV. Methods In this serial cross-sectional study, we obtained data from ten cross-sectional surveys done between and 3, consisting of anonymous self-administered questionnaires and oral HIV antibody testing in MSM recruited in gay social venues in London, UK. Data were collected between October and January for all survey years up to 8 and between February and August thereafter. All men older than 6 years were eligible to take part and fieldworkers attempted to approach all MSM in each venue and recorded refusal rates. Data were collected on demographic and sexual behavioural characteristics. We analysed trends over time using linear, logistic, and quantile regression. Findings Of 3 86 questionnaires collected between and 3, we excluded 985 (4 had completed the survey previously or were heterosexual reporting no anal intercourse in the past year, and 86 did not provide samples for antibody testing). Of the 876 eligible MSM recruited, 5 () were HIV positive, with no significant trend in HIV positivity over time. 3 (53 of 55) of HIV-positive MSM had undiagnosed infection, which decreased non-linearly over time from 3 (45 of 3) to (5 of 6; p= ), while recent HIV testing (ie, in the past year) increased from 6% (63 of 997) to 6% (467 of 777; ). The increase in recent testing in undiagnosed men (from 9% to 67%, ) and HIV-negative men (from 6% to 6%, ) suggests that undiagnosed infection might increasingly be recently acquired infection. The proportion of MSM reporting unprotected anal intercourse (UAI) in the past year increased from 4 (53 of 87) to 5 (394 of 749; ) and serosorting (exclusively) increased from (7 of 3) to 8% (77 of 6369; ). 68 (%) of 57 participants had undiagnosed HIV and reported UAI in the past year were at risk of transmitting HIV. Additionally 59 (%) had diagnosed infection and reported UAI and non-exclusive serosorting in the past year. Although we did not collect data on antiretroviral therapy or viral load, surveillance data suggests that a small proportion of men with diagnosed infection will have detectable viral load and hence might also be at risk of transmitting HIV. 633 () of 364 participants were at high risk of acquiring HIV (defined as HIV-negative MSM either reporting one or more casual UAI partners in the past year or not exclusively serosorting). The proportions of MSM at risk of transmission or acquisition changed little over time (p= 96 for MSM potentially at risk of transmission and p= 75 for MSM at high risk of acquiring HIV). Undiagnosed men reporting UAI and diagnosed men not exclusively serosorting had consistently higher partner numbers than did other MSM over the period (median ranged from one to three across surveys in undiagnosed men reporting UAI, two to ten in diagnosed men not exclusively serosorting, and none to two in other men). Lancet HIV 6; 3: e43 4 Published Online July 4, 6 S35-38(6)337-6 See Comment page e4 *Contributed equally HIV and STI Department, National Infection Service, Public Health England, London, UK (A Aghaizu MSc, A Nardone PhD); and Centre for Sexual Health and HIV Research, University College London, London, UK (S Wayal PhD, V Parsons MSc, A Copas PhD, D Mercey FRCP, Prof G Hart PhD, R Gilson FRCP, Prof A M Johnson MD) Correspondence to: Prof Anne M Johnson, Centre for Sexual Health and HIV Research, University College London, London WCE 6JB, UK Anne.Johnson@ucl.ac.uk Interpretation An increasing proportion of undiagnosed HIV infections in MSM in London might have been recently acquired, which is when people are likely to be most infectious. High UAI partner numbers of MSM at risk of transmitting HIV and the absence of a significant decrease in the proportion of men at high risk of acquiring the infection might explain the sustained HIV incidence. Implementation of combination prevention interventions comprising both behavioural and biological interventions to reduce community-wide risk is crucial to move towards eradication of HIV. Funding Public Health England. Introduction In the UK, men who have sex with men (MSM) are at the highest risk of acquiring HIV. The annual number of new HIV diagnoses in MSM has increased from 83 in to 37 in 3., Two studies have shown that estimated HIV incidence over this period remained stable (increasing slightly), and is now similar to the annual number of new HIV diagnoses (8 new infections estimated in 3). 3,4 One study 3 used a backcalculation approach based on CD4 cell count at diagnosis; the other study is a dynamic model of sexual behaviours. 4 Given the greatly increased uptake of HIV testing and antiretroviral therapy (ART) in the past decade, which should reduce transmission by reducing Vol 3 September 6 e43

2 Research in context Evidence before this study We searched PubMed for articles published in English up to May 3, 6, with the terms sexual behaviour, MSM, homosexuality or male, trends, HIV, HIV infections or HIV antibodies or HIV seropositivity or saliva or incidence or prevalence, UK, and Great Britain. Only one other study has examined trends in sexual behaviours in UK men who have sex with men (MSM; recruited from gyms across London) between 998 and 8, which found lower rates of MSM reporting unprotected anal intercourse (UAI) overall than in this study, but also an increase in the proportion of men reporting UAI and to serosort. No studies reported trends in sexual behaviours in the UK in more recent years and none showed trends in numbers of sexual partners in this population. One crosssectional study in HIV-positive MSM recruited from HIV clinics between and showed a lower prevalence of UAI (38%) and serosorting (8%). Another cross-sectional study in 8 also found a lower prevalence of serosorting and strategic positioning during UAI in MSM in Scotland with % in HIV-positive and in HIV-negative MSM. Three modelling studies that used multiple national surveillance databases show no reduction in undiagnosed HIV infection and an increase in HIV testing. Additionally, similar patterns were found in community-based surveys done in England and Scotland. Added value of this study Our data indicate changes in sexual risk behaviours with increasing rates of UAI and serosorting, with serosorting considered to be a risk reduction strategy. Our findings emphasise the importance of core groups in the epidemiology and control of HIV infection in MSM in the UK, with one in MSM identified as being potentially at risk of transmitting HIV and one in four at risk of acquiring HIV. An increasing proportion of MSM with undiagnosed HIV infection might have a recently acquired infection (ie, acquired within the past months) during which the individual is most infectious. This finding, coupled with the high partner numbers of a core group of MSM potentially at risk of transmitting HIV, and the sustained proportion at men at risk of acquiring the infection, could explain the persistent HIV incidence in the UK, despite increases in HIV testing and antiretroviral therapy coverage. Implications of all the available evidence Growing evidence shows that test-and-treat interventions alone are not sufficient to reduce HIV incidence at the population level. Combination prevention interventions will be crucial for countries with similar epidemics in MSM. viral load, sustained HIV incidence supports the notion that risk behaviours have increased over this period. 5 Since the introduction of ART in the mid-99s, studies suggest the prevalence of high-risk sexual behaviours in MSM is increasing (at least partly) because of treatment optimism, relating both to the greatly reduced morbidity and mortality associated with the infection, and the reduced risk of transmission from a positive partner (which was discovered after ). 6,7 Few behavioural studies are able to explore trends in sexual behaviours in detail, in particular examining seroadaptive behaviours relating to confirmed versus perceived HIV status. Guidelines suggest that MSM should test for HIV once a year and at least every 3 months if having unprotected anal intercourse (UAI) with new or casual partners.,8 In this Article, we examine trends in both sexual risk and HIV testing behaviours against a background of targeted prevention and testing initiatives in MSM recruited from community venues in London over the past 4 years. With half of all new HIV diagnoses in the UK occurring in London, these trends can be used to understand the role of behavioural change and testing in driving continued HIV transmission. Methods Study population and data collection In this serial cross-sectional study, we obtained data from the Gay Men s Sexual Health Survey a regular community-based survey done since 996. The survey methods have been described in detail elsewhere. 9, Briefly, from to 3, ten surveys were done. For each survey, fieldworkers visited bars, clubs, and saunas across London, UK, over 3 months inviting MSM to self-complete short, anonymous questionnaires on demographic and sexual behaviour characteristics and to provide oral fluid specimens for HIV antibody testing (OraSure Technologies, Bethlehem, PA, USA). Participants were recruited between October and January for all survey years up to 8 and between February and August in and 3. A barcode linked specimens to the corresponding questionnaire. Participants were informed that the specimens would be tested for research purposes only and results would not be returned to them. All participants were advised to attend a health-care setting for a named HIV test if they wanted to know their status. All men aged 6 years and older in the study venues were eligible to take part and fieldworkers attempted to approach as many people as possible and recorded refusal rates. Ethical approval was granted each year by the UCL research ethics committee (/58). Verbal consent for anonymous saliva samples and selfcompletion of questionnaires was obtained to ensure anonymity of all participants. Procedures Oral fluid samples, collected with the OraSure kit, were tested for HIV- antibody at Public Health England (London) with GACELISA (Abbott Laboratories, Maidenhead, UK). All samples were tested for total e43 Vol 3 September 6

3 immunoglobulin (IgG) to assess the specimen quality, apart from samples collected in, when a two-stage approach was used: first screening with a modified enzyme immunoassay then by rescreening positive specimens with an enzyme immunoassay and a western blot (HIV blot.; Genelabs, Redwood City, CA, USA). We defined undiagnosed infection as a participant who had a positive OraSure specimen and reported that they had never had an HIV test, or they perceived themselves to be negative or didn t know, or the result of their last test was negative. We defined a casual partner as a partner with whom UAI (condomless) was reported once only and defined a regular partner as a partner with whom the participant had had UAI more than once in the past year. Exclusively serosorting was defined as having UAI only with partners of presumed same HIV status in the past year. Status is referred to as presumed because the HIV status of partners was self-reported by respondents, as established with the question, In the past year, how many men that you had active/passive anal intercourse without a condom did you know had the same HIV status as you? MSM potentially at risk of transmitting HIV were defined as men with undiagnosed HIV reporting UAI in the past year or with diagnosed HIV reporting UAI and not exclusively serosorting in the past year. In the latter group, most men might have had undetectable viral load as a result of ART and might therefore not have been at risk of transmitting, but information on ART and viral load were not collected in this study. We discuss the implications of the absence of these data for the interpretation of our findings later. We defined MSM at high risk of acquiring HIV as men who were HIV negative and reported one or more casual UAI partners or were not exclusively serosorting in the past year. All other HIV-negative men were regarded as being at lower risk of acquiring HIV than men at high risk. Statistical analysis We analysed data using Stata (version 3.). Analyses were stratified by HIV status. We examined the significance of trends over time using linear, logistic, and quantile regression, adjusted for age, with survey year modelled as a linear term. For trends in HIV testing, overall HIV positivity, and undiagnosed HIV, we also adjusted for education, employment, and ethnicity, and assessed linearity with a likelihood ratio test relative to a model with survey year as a categorical variable. Characteristics of MSM at potential risk of transmitting and at high risk of acquiring HIV were explored with a multivariable model controlling for the year of survey as a linear term (odds ratios [OR] for year not shown). Factors significant to p< 5 in univariable analyses were included in the multivariable model. Role of the funding source The funder contributed to the study design, data collection, data analysis, data interpretation, and writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication. Total Year p value* HIV positive All MSM (5/ 876) HIV positive, undiagnosed Tested for HIV (ever) 3 (53/55) All MSM 79% (984/ 568) HIV negative 78% (7886/ 6) HIV positive, undiagnosed 79% (48/57) % tested for HIV (in the past year) All MSM (489/ 568) HIV negative (43/ 6) HIV positive, undiagnosed 4 (4/57) % (3/6) 3 (45/3) 6 (69/997) 6% (567/9) 57% (/35) 6% (63/997) 6% (4/9) 9% (/35) % (5/39) 49% (74/5) 69% (9/97) 67% (77/5) 77% (57/74) (4/97) (37/5) (/74) % (6/43) 3 (4/6) 76% (788/35) 7 (683/9) 57% (4/4) 36% (375/35) 37% (334/9) % (9/4) % (77/43) 4 (78/77) 78% (/4) 77% (955/48) 8% (6/76) 39% (55/4) 39% (48/48) 46% (35/76) (77/377) 4% (73/77) 78% (65/363) 76% (94/89) 8% (6/73) (576/363) (498/98) 49% (36/73) (95/53) 9% (56/93) 8% (95/487) 79% (4/33) 77% (43/56) 4 (634/487) 4 (563/33) 39% (/56) (/53) 3 (68/97) 8 (95/4) 8% (77/95) 8 (58/68) 4 (54/4) 4 (4/95) 56% (38/68) (67/6) 8% (46/66) 89% (97/9) 87% (88/935) 9 (4/45) 5% (56/9) (485/935) 5 (4/45) 8% (8/965) 9% (4/8) 9% (86/958) 89% (787/88) 88% (/4) 56% (53/958) 57% (5/88) 5% (/4) (6/78) (5/6) 9% (79/777) 9% (68/674) 9% (/4) 6% (467/777) 6% (46/674) 67% (6/4) p< p= p< 4 Data are % (n/n), unless otherwise specified. Denominators vary because of incomplete data on all variables. Percentages might not add up to totals because of rounding. MSM=men who have sex with men. *Adjusted for age, education, ethnicity, and employment. p value for association between outcome and survey year (categorical) because data showed evidence of departure from linearity. Table : HIV status and HIV testing in MSM in London, 3 Vol 3 September 6 e433

4 Total Year p value* UAI in the past year HIV negative 46% (465/ 39) HIV positive, diagnosed 6% (556/934) HIV positive, undiagnosed 5 (68/497) All 47% (5475/ 57) Exclusively serosorted in the past year HIV negative % (94/966) HIV positive, diagnosed 6% (5/86) HIV positive, undiagnosed (64/438) All % (3/ 466) (448/58) 49% (4/84) 5 (4/45) 4 (53/87) (8/7) % (8/83) (8/4) (7/3) 4 (5/38) 47% (35/74) 59% (4/7) 46% (587/83) (78/5) (3/68) 9% (6/64) (97/47) Reported UAI with partners of unknown or discordant HIV status in the past year HIV negative (748/966) HIV positive, diagnosed (59/86) HIV positive, undiagnosed 3 (46/438) All % (53/ 466) Number of UAI partners in the past year HIV negative 4 (6 8) HIV positive, diagnosed 9 (3 7) ( 5) HIV positive, undiagnosed 4 (7 5) ( ) All ( 6) % (8/7) 7% (/83) 3% (3/4) % (53/3) 9 (3 ) 4 7 ( 8) ( ) 6 (3 5) (4 7) % (/5) (6/68) 4 (9/64) % (56/47) 6 (5 9) 4 6 (4 7) ( ) 3 8 (6 5) ( 3) 9 (6 9) (378/97) 58% (46/79) 5 (/39) 4 (445/5) (43/844) % (5/7) 9% (3/34) (6/95) % (73/844) 3 (4/7) 38% (3/34) % (/95) (5 ) 8 3 (4 9) ( 7) 9 (8 4) 8 (8 7) 46% (567/35) 6% (58/94) 4 (33/75) 47% (658/44) % (4/44) (6/88) % (8/67) % (67/99) % (35/4) (6/88) (7/67) % (78/99) 4 (4 9) 7 9 ( 5) ( 4) 5 3 (34 6) ( ) ( ) 4 (5/73) 66% (65/98) 57% (38/67) 4 (65/338) (98/67) (8/9) % (6/6) (3/9) (/67) 3 (3/9) 4 (6/6) % (57/9) 3 (7 ) 7 (5 6) ( 5) 6 5 ( 4) ( ) (9 5) 49% (633/9) 5 (7/9) 6% (3/5) 5% (735/47) (7/55) (3/4) % (9/46) (3/35) % (9/55) 8% (35/4) 37% (7/46) % (8/35) 7 () 9 3 (45 ) ( 5) 4 5 (7 8) ( 4 5) 5 (7 ) 46% (43/97) 66% (8/5) 56% (35/63) 49% (548/5) (86/87) 7% (3/4) (7/53) (4/984) (36/87) 3 (4/4) 3 (8/53) % (94/984) 3 (4 3) 3 5 (36 8) ( 5) 4 (3 9) ( ) 8 (3 8) 47% (43/9) 6 (74/6) 4 (8/4) 48% (5/78) (/838) (3/5) % (8/39) (4/98) (48/838) (3/5) (7/39) (86/98) 3 (4 6) 8 (4 ) ( 5) 5 (3 3) ( ) 4 (4 8) 5% (43/843) 6% (34/55) (/) 5% (468/99) 7% (93/75) (/47) (3/7) 6% (7/789) 6% (/75) (5/47) (4/7) (3/789) 4 (3 6) 6 (3 8) ( ) ( ) (8 9) 5% (39/647) 6 (5/8) 6 (4/) 5 (394/749) 7% (5/554) (/69) 38% (6/6) 8% (77/639) 6% (86/554) 8% (9/69) (/6) (7/639) 9 ( 5) 9 7 ( 5) ( ) 7 4 (3 9) ( 5) 9 (4 ) p= p= 93 p= 6 p= 33 p= 433 p= p= 73 p= 8 p= 77 p= Data are % (n/n), mean (SD), or median (IQR), unless otherwise reported. Denominators vary because of incomplete data on all variables. Percentages might not add up to totals because of rounding. MSM=men who have sex with men. UAI=unprotected anal intercourse. *Adjusted for age. Table : Sexual behaviour in MSM by HIV status, 3 e434 Vol 3 September 6

5 Results 3 86 questionnaires were collected between and 3. The proportion of MSM who responded ranged between 5% and 7% each year. Venue data were missing for 93 participants (two participants in, three participants in 8, and all 95 participants in 3). In the 93 participants for whom data were available, 578 (8%) were recruited from bars, 636 () from clubs, and 77 (6%) from saunas. We excluded 4 questionnaires from men who completed the survey previously or were heterosexual reporting no anal intercourse in the past year, leaving questionnaires. Additionally, questionnaires from 86 men () were excluded because they did not provide samples for antibody testing, resulting in 876 questionnaires. Men who did and did not give samples were similar in age, education, and employment status (data not shown), but differed slightly by ethnic origin: 5 (6%) of 856 men who did not provide a sample were black compared with 374 () of 84 men who did. The demographic characteristics of 876 men included were similar in each of the ten surveys (data not shown). Overall, the median age was 33 years (range 6 8 years) and (86%) of 84 men were white. Across all study years, 5 () of 876 of participants were HIV positive, the proportion ranged from 8% in to in 6 (table ). 3 of HIV-positive MSM ( of the entire sample) were undiagnosed, which decreased (non-linearly) over the period from 3 in to in 3. During this period, recent HIV testing (in the past year) in all men increased from 6% to 6%. Recent testing increased in undiagnosed MSM at a similar rate, from 9% to 67%. The proportion of MSM ever having had an HIV test increased from 6 in to 9% in 3. HIV positivity varied by recruitment venue type with a similar prevalence in MSM in bars (8% [74 of 9] diagnosed, [387 of 9] undiagnosed) and clubs (6% [89 of 375] diagnosed, [6 of 375] undiagnosed), and the highest prevalence in saunas (% [7 of 67] diagnosed, 9% [57 of 67] undiagnosed). The proportion of MSM reporting UAI in the past year increased from 4 in to 5 in 3 (table, figure). This increase was significant in HIV-negative MSM, increasing from to 5% and in HIV-positive, diagnosed MSM, increasing from 49% to 6. Numbers in undiagnosed MSM were small and no clear change was recorded (table ); prevalence of UAI in undiagnosed MSM varied between 4 and 6. The proportion of MSM who exclusively serosorted increased overall from in to 8% in 3; in HIV-negative men (with other presumed HIV-negative men) this increased from to 7%. In HIV-positive, diagnosed men and in HIV-positive undiagnosed men (who reported UAI with presumed HIV-negative men) the numbers were small and changes in the proportion serosorting were not significant (table ). Notably, (8 of 653) of men who perceived themselves to be HIV negative had undiagnosed HIV in 3, with no significant trend over time (data not shown). The proportion of men reporting UAI with partners of unknown or discordant status decreased from % in to in 3 overall, from % to 6% in HIVnegative men, and from 3% to in HIV-positive, undiagnosed men. In HIV-positive, diagnosed men the changes were not significant. Over the 4 years of study, the mean number of sexual partners in the past year was consistently highest in diagnosed HIV-positive MSM and increased significantly from 4 7 (SD 8) partners in to 9 7 ( 5) in 3, having peaked at 3 5 (36 8) in 6. Between and 3, 59 diagnosed MSM reported UAI (table 3) and were not exclusively serosorting, some of whom might have been at risk of transmitting HIV. A further 68 undiagnosed HIV-positive MSM reported UAI (table 3). Together, they represented of MSM overall (table 3). The overall proportion of MSM potentially at risk of transmitting HIV remained stable over the 4 year period, as did the proportion of men in this group who were diagnosed and undiagnosed. Both diagnosed and undiagnosed MSM potentially at risk of transmitting HIV had consistently higher numbers of UAI partners than did all other MSM, but the mean number of partners increased over time for all groups. In 3, undiagnosed men at risk of transmitting reported a mean of 6 UAI partners (SD 6 ) and median of 5 (IQR ) UAI partners in the past year; diagnosed MSM, of whom some might be at risk of transmitting, reported a mean of 4 UAI partners (SD 3 ) and median of ten UAI partners (IQR 8) compared with other MSM with UAI partners (SD 3 ) and one UAI partner (IQR ). This shows 3 7% 9% 8% 4 % 8% % 5 % 56 8% 3 9% % 47 No unprotected anal intercourse Diagnosed (exclusively serosorting) Undiagnosed (exclusively serosorting with presumed negative partners) Diagnosed (not exclusively serosorting) Undiagnosed (not exclusively serosorting with presumed negative partners) HIV negative MSM (exclusively serosorting) HIV negative MSM (not exclusively serosorting) Figure: Population prevalence of unprotected anal intercourse and serosorting by HIV status in MSM in London, and 3 MSM=men who have sex with men. Vol 3 September 6 e435

6 Total Year p value* Potentially at risk of transmitting HIV Undiagnosed HIV-positive MSM reporting UAI in the past year All MSM % (68/ 57) MSM reporting UAI (68/5475) Mean (SD) number of UAI partners Median (IQR) of UAI partners 7 6 (3 ) ( 5) % (4/87) (4/53) 3 (4 4) ( ) (4/83) 7% (4/587) 6 5 ( ) 3 ( ) % (/5) (/445) 5 5 ( ) ( 5) Diagnosed MSM reporting UAI and not exclusively serosorting in the past year All MSM % (59/ 57) MSM reporting UAI (59/5475) Mean (SD) number of UAI partners Median (IQR) of UAI partners Total 7 8 (39 5) 5 ( 5) All MSM (57/ 57) No reported risk of transmitting HIV All MSM 9 ( 43/ 57) MSM reporting UAI 9% (4948/5475) Mean (SD) number of UAI partners Median (IQR) of UAI partners 6 (9 ) At high risk of acquiring HIV All MSM (633/ 57) All HIV-negative MSM (633/364) Mean (SD) of UAI partners Median (IQR) of UAI partners Tested for HIV in the past year 4 ( 8) 5% (88/573) % (/87) (/53) 3 4 ( 3) 7 ( 5) (46/87) 96% (4/87) 9% (467/53) (3 3) (8/87) 6% (8/74) 8 (5 7) ( ) 3 (79/34) % (6/83) (6/587) 6 (7 6) 4 ( 4) (58/83) 9 (5/83) 9% (59/587) 5 (5 8) (99/83) 6% (99/59) 5 ( 8) 37% (8/97) % (4/5) (4/445) 3 8 (7 4) 7 (3 6) (45/5) 96% (98/5) 9% (4/445) 4 (8 ) (36/5) 6% (36/97) 3 4 (9 ) 4 (/34) % (33/44) (33/658) (5 8) % (6/44) (6/658) 3 9 (3 4) 4 ( 8) (59/44) 96% (345/44) 9% (599/658) 5 (6 9) (34/44) 7% (34/55) 3 8 (8 ) 47% (57/337) (38/338) 6% (38/65) 4 (8 9) % (3/338) (3/65) (3 ) 4 ( 5) (69/338) 9 (69/338) 89% (536/65) 5 (7 3) % (88/338) (88/) 4 5 (3 9) 5% (4/84) % (3/47) (3/735) 7 4 (8 9) 3 ( ) % (35/47) (35/735) 9 5 ( 5) 5 ( ) (67/47) 9 (45/47) 9% (668/735) (7 ) 6% (377/47) 9% (377/38) 4 8 ( ) 5 (99/376) (35/5) 6% (35/548) 7 (8 ) (4/5) 7% (4/548) 8 8 (56 ) 5 ( 6) 7% (75/5) 9 (4/5) 86% (473/548) 7 (7 ) (7/5) (7/953) 4 (8 ) 5% (/7) % (8/78) (8/5) 3 5 (4 4) (3/78) 6% (3/5) 4 (7 5) 5 ( 5) (49/78) 9 (9/78) 9% (473/5) 7 (8 ) % (38/78) (38/939) 3 7 (8 4) 58% (38/38) % (/99) % (/468) 4 ( 7) % (5/99) (5/468) 7 (5 6) ( ) (6/99) 97% (893/99) 9 (44/468) 7 (5 6) % (3/99) (3/883) 3 8 (6 ) 7% (43/3) % (4/749) (4/394) 6 (6 ) 3 ( ) (9/749) (9/394) 4 (3 ) ( 8) (33/749) 96% (76/749) 9% (36/394) (3 ) % (54/749) (54/676) 6 (5 ) 7 (/53) < Data are % (n/n), mean (SD), or median (IQR), unless otherwise specified. Denominators vary because of incomplete data on all variables. Percentages might not add up to totals because of rounding. MSM=men who have sex with men. UAI=unprotected anal intercourse. *Adjusted for age. MSM with undiagnosed HIV who reported UAI in the past year or MSM with diagnosed HIV who reported UAI and not exclusively serosorting. In the past year. Data on antiretroviral therapy or viral load not available, so we were unable to ascertain if men in this group were on treatment and had undetectable viral loads and therefore not at risk of transmitting HIV. All MSM not including those who were potentially at risk of transmitting HIV who provided information on number of UAI partners. HIV-negative MSM reporting one or more casual UAI partner or not exclusively serosorting in the past year. Table 3: Proportion of MSM potentially at risk of transmitting and acquiring HIV, the number of UAI partners in the past year, and recent testing (in the past year) in men at high risk of acquiring HIV, 3 the skewed distribution of partner numbers, with no increase over time for most men (medians largely unchanged), but an increase in partners for the proportion at the upper end of the distribution, which increased the mean. Multivariable analyses revealed that MSM were more likely to be at risk of transmitting HIV (than were all other MSM) if they were older, of black ethnic origin, had a higher number of casual UAI partners in the past year, or had attended a genitourinary medicine clinic during the past year (table 4). e436 Vol 3 September 6

7 HIV-negative MSM were considered at high risk of HIV if they reported UAI with one or more casual partners or reported not exclusively serosorting in the past year. Overall, of HIV-negative men were at high risk (table 3). HIV testing in the past year increased in MSM in this group (table 3). Men at higher risk of acquiring HIV were more likely to have had a higher number of casual partners or have been diagnosed with a sexually transmitted infection in the past year than other HIV-negative MSM (table 4). Discussion We report that the prevalence of HIV remains high at in MSM in our London surveys. The uptake of HIV testing increased substantially in the period 3, and the proportion of HIV that is undiagnosed decreased concomitantly. Despite these changes, which might have been expected, combined with improved uptake of treatment to reduce trans mission, HIV incidence remains high and unchanged. 3,4 Our data show that irrespective of the positive changes in testing uptake, risk behaviour has increased over this period characterised by increased UAI and increasing numbers of sexual partners, particularly in HIV-positive men and men who are at risk of transmission. Serosorting, which has increased substantially over the past 4 years, is a risky practice, particularly in HIV-negative men, because (in 3) of men who perceive themselves to be HIV negative are HIV positive, inadvertently putting others at risk. We have identified and characterised a subgroup at risk of transmitting infection, particularly undiagnosed men reporting UAI (one in 43 MSM) and a larger group at risk of acquiring infection (one in five HIV-negative MSM) in whom maintenance of the epidemic could be occurring. Because not all diagnosed MSM are on treatment (69% in [Yin Z, Public Health England, personal communication], 9% in 3 ) and not all men on treatment have undetectable viral loads (9 in 3), a proportion of individuals diagnosed reporting UAI and not exclusively serosorting are also likely to be at risk of transmission. Furthermore, increased uptake of recent testing (in the past year), combined with evidence of undiagnosed HIV-positive men who have recently tested negative, suggest that an increasing proportion of the undiagnosed men might have been recently infected and could pose a high risk of transmission. This study examined long-term trends in undiagnosed HIV, testing, UAI, serosorting, and partner numbers by HIV status in MSM in London. We report on the changes in behaviours and testing alongside other available information on testing and ART uptake., Some HIV-positive individuals have been reported to change their behaviour shortly after diagnosis and we have presented differences in behaviour by HIV infection status and further explore a large group of undiagnosed MSM. Additionally, we were able to identify HIV-positive (particularly undiagnosed) MSM reporting behaviours conducive to transmission. These data will be useful in modelling studies, because we are able to provide key variables such as rate of partner change and the proportion of the population at risk. One limitation of the study is that we had no information on the timing of infection in relation to contact with sexual partners, or how many diagnosed individuals were on treatment and had undetectable viral load for MSM at risk of transmitting HIV. In HIV diagnosed men potentially at risk of transmitting, the proportion diagnosed with a sexually transmitted infection in the past year was double that of other MSM (39% vs ), which might have increased their risk of transmission. Second, the increase in recent testing in undiagnosed MSM is only suggestive of undiagnosed infection being increasingly recently acquired, because we do not know the length of infection in men that were not tested. Third, the surveys were convenience samples and might not be generalisable, comparable or both, over time. This might also have restricted the ability to detect trends over time. Response rates varied between 5% and 7% and we are unable to say how non-responders differed in risk. The demographic characteristics of the of men who refused a test were broadly the same as men who did take a test, although we are unable to infer differences in HIV status. Also, the self-reported behaviour and testing data could be subject to recall bias. Furthermore, new webbased or app-based methods to meet partners have become increasingly popular, and MSM who use these services could differ from individuals visiting bars, clubs, and saunas. A study comparing MSM recruited to online and offline behavioural surveillance studies showed that individuals using web-based methods were younger, less likely to identify as gay, less likely to use condoms with casual partners, and less likely to be tested for HIV. 3 Additionally, London-based MSM might not be representative of MSM in the UK; the estimated HIV prevalence in London is one in MSM compared with one in 8 in England and Wales outside London. 4 However, because of the low MSM population prevalence, it is not feasible to obtain a true probability sample. Unlike many convenience samples or internet samples, we did have a clear sampling frame and calculated a response rate. Data from the National Survey of Sexual Attitudes and Lifestyles (NATSAL) show that in, 6 6% (9 CI ) of MSM had attended a gay club or bar in the past year, and in, the proportion was 5 ( ; Mercer C, UCL, personal communication). 77 ( ) of gay MSM, had attended such venues in the past year. Comparison of data from MSM in convenience sample surveys and the NATSAL show that convenience sample surveys are likely to overestimate rates of sexually transmitted infection diagnoses and HIV testing, but that these differences are smaller in MSM who identify themselves as gay, 5 which suggests that our findings Vol 3 September 6 e437

8 MSM potentially at risk of transmitting HIV* MSM at higher risk of acquiring HIV % (n/n) OR (9 CI) AOR (9 CI) % (n/n) OR (9 CI) AOR (9 CI) Age, years 6 4 (4/583) (469/554) 5 34 (7/4735) 85 ( 3 6) ( ) 7% (46/437) 84 ( 74 95) 94 ( 79 ) (95/3635) 9 ( ) 67 ( 8 3 9) (745/39) 74 ( 64 85) 83 ( 69 ) (7/455) 94 ( 3 89) 55 ( ) % (5/45) 59 ( 49 7) 6 ( 49 8) 65 % (/8) 48 ( 6 3 5) 89 ( 7 ) % (9/8) 3 ( ) 39 ( 5 4) p value < < 3 Ethnic origin White (434/9973) (69/89) Black % (4/356) 86 ( 4 4 ) 6 ( ) 6% (76/87) 6 ( 8 38) Asian (/34) 7 ( 38 37) 83 ( 43 64) (73/3) 9 ( 69 8) Southeast Asian % (5/4) 55 ( 3 35) 7 ( 8 8) (36/97) 65 ( 45 94) Mixed/other (36/689) ( 85 7) ( 8 79) 7% (68/68) 9 ( 9 3) p value < 4 Years education after age 6 years None (69/35) 8% (34/6) Up to years 6% (9/888) ( 9 66) 3 ( 94 85) 8% (459/648) 99 ( 84 7) 9 ( 7 3) 3 years or more (36/7459) 77 ( 59 ) 9 ( 68 3) (637/674) 83 ( 7 95) 8 ( 68 99) Still in full-time education (3/8) 69 ( 44 6) 99 ( 6 63) 7% (3/779) 97 ( 79 9) 85 ( 64 4) p value Employed No 6% (87/58) 7% (34/47) Yes (437/ ) 75 ( 59 95) 97 ( 74 7) (89/99) 89 ( 78 ) ( 84 4) p value Age at first anal intercourse <6 years No (4/9393) 6% (76/849) Yes 8% (/463) 78 ( 43 ) 7 ( 63) 3 (46/8) 44 ( 6 64) ( 94 33) p value < 54 < 4 Casual UAI partners in the past year < % (68/964) (7/838) 5 % (8/633) 6 73 ( ) 5 5 ( ) 77% (46/36) 9 7 (7 7) 7 9 (5 4 9) 6 (64/343) 49 ( ) 9 83 ( ) 89% (8/45) 48 (3 7 9) 54 4 ( ) > 3 (5/33) 9 9 ( ) 77 ( ) 9% (4/54) 7 ( ) 69 8 ( ) p value < < < < Sexually transmitted infection in the past year No (3/939) (938/86) Yes % (/83) 3 5 ( ) 4 ( 99 55) (67/66) 3 ( 8 59) 43 ( 68) p value < 58 < < Attended a genitourinary medicine clinic in the past year No % (54/699) % (75/5937) Yes 7% (367/556) 97 ( ) 8 ( 45 6) 3% (33/437) 65 ( 5 8) ( 97 5) p value < < < Percentages might not add up to totals because of rounding. MSM=men who have sex with men. UAI=unprotected anal intercourse. OR=odds ratio. AOR=adjusted odds ratio. NA=not applicable. *Includes MSM with undiagnosed HIV who reported UAI in the past year and MSM with diagnosed HIV who reported UAI and not exclusively serosorting in the past year, compared with all other MSM. Includes HIV-negative MSM who either report one or more UAI casual partners or not exclusively serosorting in the past year, compared with all other HIV-negative MSM. Denominators vary because of incomplete data on all variables. Adjusted for age and year of survey. Multivariable model includes variables that were significant (ie, p< 5) in the univariable analysis. Table 4: Factors associated with potential risk of transmitting and high risk of acquiring HIV in MSM, 3 data combined might be generalisable to MSM who identify as gay. By obtaining trends from similar venues over an extended timeframe, we were able to make comparisons over time. Lastly, some participants might not have accurately disclosed their status, which could potentially have inflated our estimate of undiagnosed participants. However, we believe non-disclosure was kept to a minimum because the self-completed survey was entirely anonymous. e438 Vol 3 September 6

9 To our knowledge, few UK studies exist that examine trends in sexual behaviours in MSM in the community by HIV status, and none that have reported trends in MSM partner numbers in detail by HIV transmission risk. Most are cross-sectional data from earlier rounds of surveys included in this study.,6,7 In a study of sexual behaviour of gay men who used gyms in London between 998 and 8, Lattimore and colleagues 8 found a lower proportion of MSM reporting UAI (37%) than we did (5% in 8), but also an increase in UAI with partners of the same status, particularly in HIV-negative MSM, from % in 998 to % in 8. A study by McDaid and colleagues 9 on serosorting and strategic positioning during UAI in MSM in Scotland found that, although these behaviours were occurring (in % of HIV-positive and of negative MSM in 8), they were inconsistently performed. Both of these studies found increased HIV testing (ever and recent) in MSM. Continuing high levels of undiagnosed infection in MSM in the community have also been reported in Scotland of HIV-positive MSM in Scotland were undiagnosed in. Notably, the characteristics of HIV epidemics in MSM in several other countries are similar to that in the UK. For example, reports show that despite increases in ART coverage and testing, transmission in France and the USA is sustained at a high level,, probably because of increased risk behaviours similar to those shown in this study. This study emphasises the importance of core groups in the epidemiology and control of HIV infection in the MSM community in the UK. The data show changes in sexual risk behaviours of MSM in London over the past 4 years with more reporting UAI and an increase in serosorting as a risk reduction strategy. As expected, distinct differences in risk behaviours of MSM by HIV status exist, with HIV-positive men describing the highest risk. A subgroup of these men are infectious, particularly those who are undiagnosed. Coupled with high partner numbers and the one in five HIV-negative men at high risk of acquisition, this subgroup of men are likely to disproportionately be the drivers of the sustained incidence over the past decade. The benefits of serosorting might be outweighed by increased partner numbers, inconsistent practice, and incorrect perceived serostatus, as shown by the high proportion of undiagnosed men who incorrectly perceive their HIV status as negative. Additionally, the rise in testing rates in undiagnosed men suggests these infections are increasingly recently acquired, when people might be most infectious. Modelling studies have shown that reducing the number of undiagnosed infections and subsequently treating these individuals will have the greatest effect on HIV incidence. 3,4 Undiagnosed HIV infection is prevalent in the MSM community, particularly in saunas where nearly one in ten men were undiagnosed. Community-level interventions in settings such as bars, clubs, and saunas have been shown to be successful. 5 Furthermore, self-sampling and self-testing are acceptable to MSM 6 and now available in the UK, which could promote testing at more regular intervals, and would be important in earlier detection of infection to reduce transmission, particularly in men who do not attend sexual health clinics often. Anecdotal evidence suggests that an increase in new diagnoses and infections in London could also partly be attributable to other behaviours not studied here, such as an increase in recreational drug use. 7 MSM attending the central London CODE clinic (a clinic that specialises in sexual health for men who use drugs for sex chemsex) prefer to use internet sites and apps that specialise in barebacking (the common term for UAI) to find partners, with an average of five partners per encounter. 7 Further work is needed to design interventions that reach the users of these sites. Although HIV testing is increasing and the coverage of ART is high in people diagnosed, the prevalence of high-risk behaviours in MSM visiting gay social venues remains high. Treatment as prevention strategies alone are unlikely to have a substantial effect on HIV incidence in the UK because of transmission from men with primary infection and undiagnosed cases. 8 We have shown here that a large proportion of undiagnosed infections are now recently acquired infections and some of these infections are probably primary infections. Additionally, modelling studies have found the epidemiological effect of earlier diagnosis and treatment to be offset by increases in risk behaviours. 9,3 Preexposure prophylaxis (PrEP) might help prevent outbreaks. However, PrEP relies on MSM perceiving themselves at risk and choosing to test. In our study, a third of undiagnosed MSM had not been tested in the past year. Finally, serosorting, in which the status of the partner is presumed, is unsafe because of incorrect perception of serostatus. Thus, public health authorities urgently need to focus on implementing and evaluating combined behavioural and biomedical interventions (such as test-and-treat and PrEP programmes) that focus on the group of men potentially at risk of HIV transmission and HIV-negative men with behavioural patterns that put them at high risk of acquisition. The social and cultural mixing of these groups will need to be considered as part of the design of risk reduction strategies (eg, targeting of younger MSM who might be less aware of the risks and less able to protect themselves). The findings presented in this Article are an important contribution to the growing evidence that testing and treatment strategies alone are not sufficient to reduce HIV incidence at the population level. Combination prevention working closely with affected communities to reduce community-wide risk by both behavioural and biological interventions is crucial if we are to move towards eradication of HIV. Vol 3 September 6 e439

10 Contributors All authors contributed to the design of the study. AA and SW analysed the data and drafted the manuscript. All authors commented on drafts of the manuscript and approved the final version. Declaration of interests We declare no competing interests. Acknowledgments This study was funded by Public Health England. We thank Gary Murphy and Bharati Patel for doing the laboratory testing, Catherine Mercer for providing the NATSAL statistics, Zheng Yin for providing data on proportion of diagnosed men on antiretroviral therapy with detectable viral load in, and all participants for their valuable contributions. References Aghaizu A, Brown AE, Nardone A, et al. HIV in the United Kingdom 3 report: data to end. London: Public Health England, 3. Public Health England. National HIV surveillance data tables. London: Public Health England, 3. 3 Birrell PJ, Gill ON, Delpech VC, et al. HIV incidence in men who have sex with men in England and Wales : a nationwide population study. Lancet Infect Dis 3; 3: Phillips AN, Cambiano V, Nakagawa F, et al. Increased HIV incidence in men who have sex with men despite high levels of ART-Induced viral suppression: analysis of an extensively documented epidemic. PLoS One 3; 8: e Elford J. Changing patterns of sexual behaviour in the era of highly active antiretroviral therapy. Curr Opin Infect Dis 6; 9: Crepaz N, Hart TA, Marks G. Highly active antiretroviral therapy and sexual risk behavior: a meta-analytic review. JAMA 4; 9: Elford J. HIV treatment optimism and high-risk sexual behaviour among gay men: the attributable population risk. AIDS 6; 8: British HIV Association, British Association of Sexual Health and HIV, British Infection Society. UK national guidelines for HIV testing 8. London: British HIV Association, 8. 9 Aghaizu A, Mercey D, Copas A, et al. Who would use PrEP? Factors associated with intention to use among MSM in London: a community survey. Sex Transm Infect 3; 89: 7. Dodds JP, Mercey DE, Parry JV, Johnson AM. Increasing risk behaviour and high levels of undiagnosed HIV infection in a community sample of homosexual men. Sex Transm Infect 4; 8: Fox J, White PJ, MacDonald N, et al. Reductions in HIV transmission risk behaviour following diagnosis of primary HIV infection: a cohort of high-risk men who have sex with men. HIV Med 9; 7: Bolding G, Davis M, Hart G, Sherr L, Elford J. Where young MSM meet their first sexual partner: the role of the Internet. AIDS Behav 7; : Saxton P, Dickson N, Hughes A. Who is omitted from repeated offline HIV behavioural surveillance among MSM? Implications for interpreting trends. AIDS Behav 3; 7: Skingsley A, Yin Z, Kirwan P, et al. HIV in the UK: situation report 5. Incidence, prevalence and prevention. London: Public Health England, 5. 5 Prah P, Hickson F, Bonell C, et al. Men who have sex with men in Great Britain: comparing methods and estimates from probability and convenience sample surveys. Sex Transm Infect 6; published online March. DOI:.36/sextrans Dodds JP, Johnson AM, Parry JV, Mercey DE. A tale of three cities: persisting high HIV prevalence, risk behaviour and undiagnosed infection in community samples of men who have sex with men. Sex Transm Infect 4; 83: Williamson LM, Dood JP, Mercey DE, Johnson AM, Hart GJ. Increases in HIV-related sexual risk behavior among community samples of gay men in London and Glasgow: how do they compare? J Acquir Immune Defic Syndr 6; 4: Lattimore S, Thornton A, Delpech V, Elford J. Changing patterns of sexual risk behavior among London gay men: Sex Transm Dis ; 38: 9. 9 McDaid LM, Hart GJ. Serosorting and strategic positioning during unprotected anal intercourse: are risk reduction strategies being employed by gay and bisexual men in Scotland? Sex Transm Dis 3; 39: Wallace LA, Li J, McDaid LM. HIV prevalence and undiagnosed infection among a community sample of gay and bisexual men in Scotland, 5 : implications for HIV testing policy and prevention. PLoS One 4; 9: e985. Prejean J, Song R, Hernandez A, et al. Estimated HIV incidence in the United States, 6 9. PLoS One ; 6: e75. Le Vu S, Le Strat Y, Barin F, et al. Population-based HIV- incidence in France, 3 8: a modelling analysis. Lancet Infect Dis ; : Marks G, Crepaz N, Janssen RS. Estimating sexual transmission of HIV from persons aware and unaware that they are infected with the virus in the USA. AIDS 6; 6: Granich RM, Gilks CF, Dye C, De Cock KM, Williams BG. Universal voluntary HIV testing with immediate antiretroviral therapy as a strategy for elimination of HIV transmission: a mathematical model. Lancet 9; 373: Godin G, Naccache H, Cote F, Leclerc R, Frechette M, Alary M. Promotion of safe sex: evaluation of a community-level intervention programme in gay bars, saunas and sex shops. Health Educ Res 8; 3: Wayal S, Llewellyn C, Smith H, et al. Self-sampling for oropharyngeal and rectal specimens to screen for sexually transmitted infections: acceptability among men who have sex with men. Sex Transm Infect 5; 85: Kirby T, Thornber-Dunwell M. High-risk drug practices tighten grip on London gay scene. Lancet 3; 38:. 8 Brown AE, Nardone A, Delpech VC. WHO Treatment as Prevention guidelines are unlikely to decrease HIV transmission in the UK unless undiagnosed HIV infections are reduced. AIDS 4; 8: van Sighem A, Vidondo B, Glass T, et al. Resurgence of HIV infection among men who have sex with men in switzerland: mathematical modelling study. PLoS One ; 7: e Bezemer D, de Wolf F, Boerlijst MC, et al. 7 years of the HIV epidemic amongst men having sex with men in the Netherlands: an in depth mathematical model-based analysis. Epidemics ; : e44 Vol 3 September 6