Men who have sex with men (MSM) have been the predominantly affected risk group since the earliest days of the US HIV epidemic and continue to be the risk group most commonly diagnosed with HIV in the national HIV Surveillance System . Recent data indicate that rates of reported HIV diagnoses have been increasing by over 8% annually since 2001, with the sharpest increases occurring among the youngest (aged 13–24 years) and oldest (aged ≥45 years) MSM and black MSM . MSM are the only risk group with increasing HIV incidence since 2000 . Recently, calls have emerged for a renewed focus on developing novel HIV prevention approaches for US MSM .
An earlier report that most HIV infections among young MSM in the Netherlands are attributable to main sex partners , and a recent study  from Africa documenting that most HIV transmissions among male–female couples in Zambia and Rwanda are from cohabiting sex partners, led us to question the relative role of transmissions from main sex partners in the United States. We developed a model of HIV transmission based on survey data and HIV test results from respondents to Center for Disease Control and Prevention (CDC)'s National HIV Behavioral Surveillance System (NHBS) among MSM and other published and publicly available data [7,8].
The primary data sources were a five-city subset of the NHBS data from the 2003 to 2005 cycle, which interviewed and conducted HIV testing among MSM recruited using venue–time–space sampling  in five US cities (Baltimore, Los Angeles County, Miami, New York, and San Francisco and the Greater Bay Area) [10,11], and the HIVNET Vaccine Preparedness Study (VPS), which collected data from MSM in six US metropolitan areas (Boston, Chicago, Denver, New York, San Francisco, and Seattle) from 1995 to 1997 (National Institute of Allergy and Infectious Diseases, National Institutes of Health, Department of Health and Human Services. Vaccine Preparedness Study, Uninfected Participant Cohort, FHCRC-HIVNET Statistical Center Version, March 2000). Because HIV testing was offered only during limited time periods in the five NHBS cities, not all men who completed the interview were offered HIV testing. During the period when HIV testing was offered, 2267 men completed the survey and 1777 (79% of these participants) agreed to be HIV tested. A total of 3617 MSM participated in the VPS study.
NHBS data were used to estimate numbers of MSM with main and casual partners in a year, average numbers of casual partners for men who had them, proportional distribution of partner serostatus, proportional distribution of sex episodes that involved receptive anal intercourse (RAI), insertive anal intercourse (IAI), or oral sex, proportion of anal intercourse that was unprotected by condoms, the prevalence of HIV by partner type (main versus casual), and the proportion of unrecognized HIV infection. Data on sex behaviors were from last sex; respondents were asked separately about last sex with both main and casual partners, when applicable. VPS data were used to approximate the number of sexual episodes per year for men with main and casual sex partners; because of the specific information available from VPS, we estimated sexual episodes per year in the following way. To estimate sexual episodes per year with main partners, we used data about number of oral or anal sex acts in the prior 6 months from the baseline interview and from 6, 12, and 18-month follow-up interviews for men who reported a single main partner. To estimate sexual episodes with causal partners, we used data from the same interview periods from men who had one or more casual partners, but not a main partner during the prior 6 months. Six-month estimates for episodes were annualized by multiplying by two.
We tested for differences in the distribution of demographic factors between NHBS respondents who were and were not tested for HIV and between NHBS respondents and VPS participants using χ 2 test. Published data were used to estimate per anal sex episode risk of HIV transmission [7,8]. Estimates of HIV transmission from the VPS data were updated, using the final study data from the VPS study; earlier estimates were published before the VPS data were finalized.
To characterize new HIV infections, which would be expected to occur among a hypothetical cohort of 3652 HIV-negative MSM with the behavioral characteristics of the NHBS MSM respondents, we summed the estimated number of transmissions for each combination of partner type (main or casual), sex act (RAI, IAI, and oral sex), and reported partner serostatus (negative, positive, and unknown):
where N is the number of men reporting having main or casual partners in the last year, S is the number of annual sex acts in the last year, U is the proportion of sex acts unprotected by condoms, H is the estimated true prevalence of HIV infection among partners, R is the per act risk of HIV transmission, i is the type of sex partner (main or casual), j is the reported partner HIV status (negative, positive, and unknown), and k is the type of sex act (receptive anal, insertive anal, and receptive oral)
The data sources and detailed information about model inputs are summarized in Table 1 ([7,11–13]). Total annual sex acts with main and casual partners were apportioned to one of nine categories [three levels of sex type (k) by three levels of partner's reported HIV status (j)]. Total numbers of sex acts with main and casual partners were multiplied, separately, by the proportion of sexual episodes that occurred with each stratum of sex type* reported partner HIV status. The result was an estimated annual number of RAI, IAI, and oral sex acts, by partner type and serostatus.
The total numbers of unprotected anal sex episodes were estimated by multiplying partner type–sex type–seroconcordance-specific condom usage rates (U) by the total number of sex acts (S) in the stratum. For oral sex, we assumed, on the basis of published literature, that the rates of condom use during oral sex were low enough to be negligible .
The number of unprotected sex acts with HIV-infected partners was estimated by multiplying the number of unprotected sex acts by the partner type–sex act-specific rates of HIV prevalence (H, as determined by HIV testing of respondents). The estimated new HIV infection was calculated by multiplying the total number of unprotected sex acts with infected partners by the per-act risk of HIV transmission for that act (R).
Thus, estimates of annual HIV infections were calculated and attributed to main or casual partners, RAI, IAI, or oral sex and perceived partner serostatus. The 95% confidence intervals (CIs) around the proportions were calculated using standard formulae, assuming a normal distribution. The annual HIV incidence rate was calculated by dividing the total number of predicted HIV transmissions by the total number of men, with 95% CIs.
We repeated the modeling process stratified by demographic variables when there were significant differences between the NHBS and VPS respondents (age, race, and education). The stratified models used stratum-specific estimates of partner numbers, distribution of sex acts, condom use rates, HIV prevalence, and undiagnosed HIV infection from NHBS data. We summed across strata to generate overall estimates from the stratified analysis. We tested for a trend in prevalence of transmission from main partners, from RAI, and from partners of unknown serostatus across age groups using the Cochran–Armitage test (SAS version 9.2; SAS Institute Inc., Cary, North Carolina, USA).
To determine how sensitive our model was to plausible changes in input parameters, we conducted sensitivity analyses (Table 2). We varied key input parameters to upper and lower 95% CIs and considered alternative data sources for estimates of annual sex episodes with main and casual partners. We constructed an overall scenario that favored transmission from casual partners, conducted an analysis restricted to the two cities (New York and San Francisco) where both NHBS and VPS data were collected, and constructed an analysis using only data from participants who were also HIV tested as part of the study. We also tested an alternate measure of seroprevalence by using direct estimates of seroprevalence in four groups (main and casual partners who reported being HIV-negative or HIV-unknown). Finally, we varied each individual variable according to an appropriate probability distribution and used Monte Carlo simulation (@RISK, Palisade Corporation, Ithaca New York, USA) to generate 95% CIs around the estimate of transmission from main partners and conducted a Monte Carlo simulation allowing all variables to vary according to their respective probability distributions.
To address the concern that the distribution of type of sex act might be correlated with frequency of sex (e.g., that men who reported higher numbers of sex acts might have a higher proportion of RAI), we used VPS data to model the relationship between number of annual sex acts and the proportion of sex which was RAI.
A total of 3652 MSM respondents to the NHBS survey provided complete information on recent sex partners and were included in the analysis group. Men who provided complete information differed statistically from the 1777 NHBS respondents tested for HIV in the five cities by age group, educational attainment, and city of recruitment (Table 3), but the magnitude of the differences was small. Included NHBS respondents differed from MSM in the VPS study by age group, race, and educational attainment.
NHBS respondents reported information from most recent sex with a total of 2395 main sex partners and 2850 casual sex partners (Table 4). Forty-four percent of NHBS respondents reported both main and casual partners in the 12 months before interview. Main partners were significantly less likely to be of unknown HIV serostatus (P < 0.0001). Men were significantly less likely to use condoms with main partners (versus casual partners) for RAI and IAI with partners who were reported to be HIV negative (43 versus 73% for RAI, 43 versus 71% for IAI; P < 0.0001 for both comparisons) or with unknown serostatus (57 versus 76% for RAI, 59 versus 75% for IAI; P < 0.0001 for both comparisons); condom use was similar with main and casual HIV-positive partners (P = 0.76 and P = 0.87 for RAI and IAI, respectively). Men were more likely to have had RAI with their main partner (40%) than with their casual partner (26%) at last sex (P < 0.0001 by χ 2). Mean annual number of main partners was 1.5 for men with only main partners and 1.3 for men with both main and casual partners; 25% of men with only main partners and 10% of men with both main and casual partners reported more than one main partner. Among NHBS respondents, the proportion of men who had only main partners who were truly HIV-infected was 24% and of men who had any casual partner it was 26%; in the same groups, proportions of HIV-infected men who were unaware of their HIV infection were 43 and 49%, respectively. We estimated that NHBS respondents had 194 875 annual sex acts with main partners and 173 450 annual sex acts with casual partners.
Most transmissions (68%) were estimated to be from main partners in the primary analysis and in models stratified by age, race, and educational attainment (Fig. 1a). There was a significant trend in proportion of transmissions from main partners across age groups by Cochran–Armitage test (two-tailed P = 0.002, Fig. 1a). There was no trend in proportion of transmissions from RAI or from partners of unknown serostatus across age groups. By sex type, most transmissions were estimated to arise from RAI; among all men, most transmissions were estimated to arise from partners believed to be HIV-negative (Fig. 1b). Among transmissions from main partners, most were from partners reported to be HIV negative; among transmissions from casual partners, most were from partners of unknown HIV status (Fig. 1c).
The predicted annual HIV incidence for the overall analysis was 2.2% (CI 1.7–2.7%). In sensitivity analyses (Table 2), the highest and lowest estimates of the proportion of HIV transmissions from main partners were 74 and 52%, respectively.
In addressing our concern that the distribution of type of sex act might be correlated with frequency of sex, we found that among VPS respondents, frequency of sex was significantly (P < 0.0001) but not meaningfully associated with proportion of sex acts, which were RAI. Only 1% in the variability of proportion of sex acts, which were RAI was explained by frequency of sex (r 2 = 0.01).
Our finding that most HIV transmissions among this group of urban MSM arose from sex with main partners was driven by several factors. First, men have more sex with their main partners than with their casual partners each year, over 10% more by our estimates. Second, men were 14% more likely to have RAI, which carries a higher per-act risk of HIV transmission, with their main partners than with their casual partners. Third, men in our study, and others [15–19], were less likely to use condoms for anal sex with their main partners; partner serostatus–sex type-specific rates of sex not protected by condoms were 16–31% higher with main partners than casual partners in our data. Taken together, these factors result in more episodes of unprotected anal sex, especially RAI, with main partners than with casual partners. Although prevalence of HIV infection and of undiagnosed HIV infection were marginally lower in main than casual partners in our model (Table 2), these differences were too small to overcome the larger differences favoring more unprotected RAI acts with main partners.
Our finding of a decreasing trend by age in transmission from main partners is driven by higher proportions of undiagnosed HIV infection in younger men; a previous analysis  of NHBS data showed that the prevalence of undiagnosed HIV was as high as 79% among younger MSM. There may be differences in how younger men define main partners or differences in the length of relationships with main partners. It is likely that, for many older men, main partners may have been established for many years. Given the frequency of HIV testing among MSM , a main partner of long duration is less likely to have unrecognized HIV infection, and the risk of acquiring HIV from that partner would be lower.
Two other findings bear special attention. First, transmissions from IAI represented 28% of estimated infections; the ratio of RAI to IAI transmissions in our model (about 2.5) is lower than would be predicted only on the basis of the relative transmission efficiencies of RAI versus IAI (about 4.0). The difference is explained in the model by lower condom use for IAI than RAI, especially with HIV-positive partners (Table 3). Strategic positioning may play an important role in transmission dynamics, and prevention counselors should continue to emphasize that the risk of IAI, although lower than the risk of acquisition from RAI, is still an important concern for MSM. Second, among main partners, transmissions mostly arise from partners believed to be negative (partners with unrecognized infection), whereas among casual partners, most transmissions arise from partners of unknown serostatus. Assumptions about serostatus in main partnerships may be especially problematic with respect to decisions about condom use within main partnerships.
Our data are subject to several limitations. First, our model synthesizes data collected in different cities during different time periods. However, in analyses stratified by race and limited to the two cities where both studies collected data, our main conclusion was unchanged. Because data were collected differently in NHBS and VPS, there was some misalignment of our definitions of main and casual partners: VPS did not allow estimation of annual sexual episodes separately for sex with main and casual partners for men who had both types of partners.
Second, data on HIV prevalence and unrecognized HIV infection are an important component of the model. It is possible that proportion of men with unrecognized HIV may be erroneously high because men who know they are HIV-infected may have reported in the interview that they were HIV negative . We estimated the true prevalence of HIV infection and the prevalence of unrecognized HIV infection among partners of our modeled cohort by assuming that the NHBS respondents were themselves reflective of the true HIV prevalence and prevalence of unrecognized HIV infection in their sex partners. This assumption could be problematic if venue-attending men often had male sex partners who did not attend venues and would not have been eligible for recruitment. The use of age-specific and race-specific estimates of HIV prevalence and undiagnosed prevalence require the assumption that most partnerships are similar with respect to age and race. If this assumption were violated, it would be more likely to affect the race- and age-specific estimates; the overall estimates would still be valid as long as the types of sex partners our respondents sought attended venues and were eligible for sampling in the NHBS study.
Third, our model does not account for some other, important aspects that impact HIV transmission among MSM. For example, infection with sexually transmitted diseases (STDs) increases the per-act risk of HIV transmission and acquisition ; STD infections have increased among MSM in the United States in recent years [21,22]. Our model did not account for the impact of STDs on HIV transmission risk. The VPS data used to estimate HIV transmission risks were collected before the broad availability of HAART; HAART prescription, when associated with suppression of viral load, is expected to substantially reduce the per-episode risk of HIV transmission . Also, we did not account for varying infectivity of HIV-infected persons in different stages of HIV disease (acute infection versus latent period), for certain harm-reduction practices such as premature withdrawal, or for the potential impact of circumcision  on HIV transmission rates.
Our modeling approach did not allow us to censor men at the time of infection; therefore, our model may have overestimated transmissions. Because the likelihood and magnitude of this issue would be related to number of sex acts, it would tend to differentially overestimate transmissions from main partners. Our assumption that men reporting more than one male sex partner per year contributed the same number of transmissions as men reporting one male partner may also be problematic. Men who have serial main partners may actually be at greater risk of HIV infection, so our assumption may lead to an underestimate of transmissions in main partners.
Neither NHBS nor VPS collected data on duration of partnerships. If most HIV transmissions occur from main sex partners, then substantial rates of main partner change would be prerequisite to sustaining onward HIV transmission among MSM. In our study, a substantial proportion of men with main partners reported more than one main partner in the previous 12 months. Previous studies [25,26] in cross-sectional groups of MSM have documented mean duration of current primary relationships of 17 and 40 months. For men who have main partnerships of longer duration, the risk of HIV transmission from main partners would likely decrease over time. The frequency of sex within main partnerships declines over time in relationship . Also, it is likely that the longer relationships (especially monogamous relationships) last, the more likely it is that both partners would have accurate knowledge of HIV serostatus. Of note, structural interventions to promote stability of relationships between men have been suggested as an approach to addressing the HIV epidemic in MSM ; if such approaches are implemented and are successful in promoting stability, longer duration of primary relationships may be an important mediator in reducing HIV incidence and would be expected to decrease the proportion of infections attributed to main partners in our model.
Our data are not generalizable to all MSM in the United States. Both source studies recruited sexually active MSM in urban areas; our findings do not apply to MSM in more rural areas. If MSM in our hypothetical cohort migrated to areas of varying HIV prevalence or undiagnosed HIV infection, our model would likely misrepresent incidence. Although diverse venues such as retail establishments and social organizations were included in the NHBS sampling frame , most NHBS participants were recruited from bars or dance clubs . It is likely that MSM who do not attend NHBS venues, or who attend such venues infrequently, have patterns of behavioral risk that are different from those recruited to NHBS.
Despite these limitations, we believe that our model is a reasonable one. Our model estimated an annual HIV incidence of 2.2% (CI 1.7–2.7%). A meta-analysis  of published estimates of HIV incidence from 1995 to 2005 reported a weighted mean incidence of 2.3% (CI 2.0–2.65%) for MSM recruited in community settings. The median transmission rate derived from the serologic algorithm for HIV seroconversion (STARHS) from the same respondents in the five NHBS cities who were included in our analysis was 2.3% (range by city, 1.2–8.0%) . The comparability of our estimates of HIV incidence derived by modeling and other estimates of HIV incidence derived using biological markers of recent infection suggests that our model makes reasonable assumptions.
Our data suggest that existing HIV prevention interventions should address issues related to risk within male partnerships and issues of testing and disclosure, and interventions that focus on MSM couples should also be considered. To date, most HIV prevention interventions tested in MSM (80%) have been group level interventions . Many of our estimated transmissions arose in partnerships, which were believed by the participants to be seronegative concordant relationships. It is clear that men, whether they believe themselves to be HIV negative or know that they are HIV positive, do communicate a negative serostatus to sex partners [30,31]. Couples who believe themselves to be concordant seronegative may also benefit from improved knowledge of serostatus and risk reduction strategies. Also, the current MSM cycle of the NHBS system is interviewing MSM in 21 US cities during 2008; in the 2008 cycle, all men were offered HIV testing, and data on concurrency and duration of sexual partnerships were collected. This analysis should be repeated with data from this broader group of cities, to determine whether our finding is robust over time and in a more heterogeneous group of US cities.
A recent commentary  on HIV prevention noted that, ‘for too long, AIDS activists, academics, and national and international institutions have given insufficient emphasis to aligning prevention priorities with epidemic transmission dynamics’. Such is the case with available HIV prevention interventions for MSM in the United States. Although over 50% of new US HIV infections are among MSM, only eight of 36 (22%) HIV-prevention interventions designated as ‘best-evidence’ prevention interventions by the CDC have been tested with MSM . New HIV prevention interventions for US MSM are needed and should be responsive to the current behavioral risk patterns and transmission characteristics of the US MSM epidemic. Our data suggest that the development and testing of HIV prevention interventions for MSM couples, a historically understudied area of HIV prevention, may be important as we approach the fourth decade of the US HIV epidemic.
The authors express thanks to the Principal Investigators of the NHBS surveillance sites for the MSM data collection, who participated in the protocol development, oversaw the collection and local management of behavioral and HIV testing data, and provided critical review of the manuscript: Baltimore, Colin Flynn; Los Angeles County, Trista Bingham; Miami, Marlene LaLota; New York, Alan Neaigus; San Francisco and the Greater Bay Area, H. Fisher Raymond.
The authors express thanks to Dr Eric Vittinghoff, who provided updated estimates of HIV transmission risk for use in the model and thoughtful critique of the manuscript.
P.S. conceived the analysis, conducted the statistical analyses, and drafted the manuscript. L.S. participated in the development of the analysis plan, assisted in preparation of the manuscript, and provided perspective on how local data collection practices may have affected variable quality. S.B. provided critical input into the specifications of the model, coordinated updating of the HIV transmission risks, and provided critical input on the manuscript. T.S. conceived the analysis with P.S., participated in the development of the analysis plan, and provided critical input on the manuscript. All authors reviewed and provided substantial input to revisions of the manuscript.
The findings and conclusions in this report are those of the authors and do not necessarily represent the views of the Centers for Disease Control and Prevention.
1. Centers for Disease Control and Prevention. HIV/AIDS Surveillance Report, 2005 (vol. 17)
. Atlanta: US Department of Health and Human Services, Centers for Disease Control and Prevention; 2006. pp. 1–46. http://www.cdc.gov/hiv/stats/hasrlink.htm
. [Accessed 30 January 2007].
2. CDC. Trends in HIV/AIDS diagnoses among men who have sex with men: 33 States, 2001–2006
3. Hall HI, Song R, Rhodes P, Prejean J, An Q, Lee LM, et al
. Estimation of HIV
incidence in the United States. J Am Med Assoc 2008; 300:520–529.
4. Jaffe HW, Valdiserri RO, De Cock KM. The reemerging HIV
/AIDS epidemic in men who have sex with men
. JAMA 2007; 298:2412–2414.
5. Davidovich U, de Wit J, Albrecht N, Geskus R, Stroebe W, Coutinho R. Increase in the share of steady partners as a source of HIV
infection: a 17-year study of seroconversion among gay men. AIDS 2001; 15:1303–1308.
6. Dunkle KL, Stephenson R, Karita E, Chomba E, Kayitenkore K, Vwalika C, et al
. New heterosexually transmitted HIV
infections in married or cohabiting couples in urban Zambia and Rwanda: an analysis of survey and clinical data. Lancet 2008; 371:2183–2191.
7. Vittinghoff E, Douglas J, Judson F, McKirnan D, MacQueen K, Buchbinder SP. Per-contact risk of human immunodeficiency virus transmission between male sexual partners. Am J Epidemiol 1999; 150:306–311.
8. Varghese B, Maher JE, Peterman TA, Branson BM, Steketee RW. Reducing the risk of sexual HIV
transmission: quantifying the per-act risk for HIV
on the basis of choice of partner, sex act, and condom use. Sex Transm Dis 2002; 29:38–43.
9. MacKellar DA, Gallagher KM, Finlayson T, Sanchez T, Lansky A, Sullivan PS. Surveillance of HIV
risk and prevention behaviors of men who have sex with men
: a national application of venue-based, time-space sampling. Public Health Rep 2007; 122(Suppl 1):39–47.
10. Sanchez T, Finlayson T, Drake A, Behel S, Cribbin M, Dinenno E, et al
. Human immunodeficiency virus (HIV
) risk, prevention, and testing behaviors: United States, National HIV
Behavioral Surveillance System – men who have sex with men
, November 2003–April 2005. MMWR Surveill Summ 2006; 55:1–16.
11. Centers for Disease Control and Prevention. HIV prevalence, unrecognized infection, and HIV testing among men who have sex with men: five U.S. cities, June 2004–April 2005
12. Rietmeijer CA, Lansky A, Anderson JE, Fichtner RR. The current good should precede the future best: a response to a response. AIDS Educ Prev 2002; 14:348–350.
13. Stone E, Heagerty P, Vittinghoff E, Douglas JM Jr, Koblin BA, Mayer KH, et al
. Correlates of condom failure in a sexually active cohort of men who have sex with men
. J Acquir Immune Defic Syndr Hum Retrovirol 1999; 20:495–501.
14. Page-Shafer K, Shiboski CH, Osmond DH, Dilley J, McFarland W, Shiboski SC, et al
. Risk of HIV
infection attributable to oral sex among men who have sex with men
and in the population of men who have sex with men
. AIDS 2002; 16:2350–2352.
15. Koblin BA, Chesney MA, Husnik MJ, Bozeman S, Celum CL, Buchbinder S, et al
. High-risk behaviors among men who have sex with men
in 6 US cities: baseline data from the EXPLORE Study. Am J Public Health 2003; 93:926–932.
16. Stueve A, O'Donnell L, Duran R, San Doval A, Geier J. Being high and taking sexual risks: findings from a multisite survey of urban young men who have sex with men
. AIDS Educ Prev 2002; 14:482–495.
17. Poppen PJ, Reisen CA, Zea MC, Bianchi FT, Echeverry JJ. Serostatus disclosure, seroconcordance, partner relationship, and unprotected anal intercourse among HIV
-positive Latino men who have sex with men
. AIDS Educ Prev 2005; 17:227–237.
18. Hart T, Peterson JL. Predictors of risky sexual behavior among young African American men who have sex with men
. Am J Public Health 2004; 94:1122–1124.
19. Choi KH, Han CS, Hudes ES, Kegeles S. Unprotected sex and associated risk factors among young Asian and Pacific Islander men who have sex with men
. AIDS Educ Prev 2002; 14:472–481.
20. Cohen MS. HIV
and sexually transmitted diseases: lethal synergy. Top HIV
Med 2004; 12:104–107.
21. Division of STD Prevention. Sexually Transmitted Disease Surveillance, 2005. Department of Health and Human Services. Atlanta: Centers for Disease Control and Prevention (CDC); 2006.
22. Ciesielski CA. Sexually transmitted diseases in men who have sex with men
: an epidemiologic review. Curr Infect Dis Rep 2003; 5:145–152.
23. Gray RH, Wawer MJ, Brookmeyer R, Sewankambo NK, Serwadda D, Wabwire-Mangen F, et al
. Probability of HIV
-1 transmission per coital act in monogamous, heterosexual, HIV
-1-discordant couples in Rakai, Uganda. Lancet 2001; 357:1149–1153.
24. Millett GA, Flores SA, Marks G, Reed JB, Herbst JH. Circumcision status and risk of HIV
and sexually transmitted infections among men who have sex with men
: a meta-analysis. JAMA 2008; 300:1674–1684.
25. Halkitis PN, Wilton L, Parsons JT, Hoff CC. Correlates of sexual risk taking behaviour among HIV
seropositive gay men in concordant primary partner relationships. Psychol Health Med 2004; 9:99–113.
26. Davidovich U, de Wit JB, Stroebe W. Assessing sexual risk behaviour of young gay men in primary relationships: the incorporation of negotiated safety and negotiated safety compliance. AIDS 2000; 14:701–706.
27. Blumstein P, Schwartz PW. American couples: money, work, sex. New York: William Morrow and Company; 1983.
28. Stall R, Duran L, Wisniewski SR, Friedman MS, Marshal MP, McFarland W, et al
. Running in place: implications of HIV incidence estimates among urban men who have sex with men in the United States and other industrialized countries
. AIDS Behav
2009 [Epub ahead of print, Feb 10, 2009].
29. Semaan S, Kay L, Strouse D, Sogolow E, Mullen PD, Neumann MS, et al
. A profile of U.S.-based trials of behavioral and social interventions for HIV
risk reduction. J Acquir Immune Defic Syndr 2002; 30(Suppl 1):S30–S50.
30. Wolitski RJ, Rietmeijer CA, Goldbaum GM, Wilson RM. HIV
serostatus disclosure among gay and bisexual men in four American cities: general patterns and relation to sexual practices. AIDS Care 1998; 10:599–610.
31. MacKellar DA, Valleroy LA, Behel S, Secura GM, Bingham T, Celentano DD, et al
. Unintentional HIV
exposures from young men who have sex with men
who disclose being HIV
-negative. AIDS 2006; 20:1637–1644.
32. Wilson D, Halperin DT. ‘Know your epidemic, know your response’: a useful approach, if we get it right. Lancet 2008; 372:423–426.