The number of persons aged 13 years and older newly infected with HIV in 2010 in the United States was estimated to be 47 500, and 88% of all new HIV infections were attributed to sexual transmission [1]. Recent estimates suggest that approximately four transmissions occur per 100 HIV-infected persons annually [2]; however, nationally representative data on sexual risk behaviours have not been available since the late 1990s [3]. Thus, national data on the sexual risk behaviour of HIV-infected persons are needed to inform prevention efforts to reduce sexual transmission of HIV and the acquisition of other sexually transmitted infections (STIs).

Providing antiretroviral therapy (ART) improves individual health outcomes and observational evidence and a randomized clinical trial have demonstrated that taking ART substantially reduces the risk of heterosexual HIV transmission [4,5]. Although this finding has profound implications for HIV prevention, access to ART does not eliminate all the barriers that may interfere with the elimination of HIV transmission, including delayed diagnosis, lack of continuous care, suboptimal adherence, drug resistance and subsequent increases in risky sexual behaviour (i.e. risk compensation) [6–10]. Findings from prior investigations of the association between ART provision and increased sexual risk behaviour are mixed [11–14]. Soon after effective therapy became widely available in developed countries, evidence suggested a resurgence of HIV and other STIs, especially syphilis, among MSM [13,15]. Some data, including meta-analyses, have indicated no increases in sexual risk behaviour among persons who received ART or were virally suppressed, but the prevalence of unprotected sex was higher among HIV-infected adults who believed that ART or viral suppression protects against transmission [12,14].

In this report, we present data (overall and stratified by viral suppression) to describe sexual risk behaviours in a national sample of HIV-infected adults receiving medical care in the United States. Given the increased probability of HIV transmission when individuals are not virally suppressed, we also assess the association between sexual risk behaviour and viral suppression.

The Medical Monitoring Project (MMP) uses a three-stage sampling design to obtain nationally representative, annual cross-sectional samples of HIV-infected adults receiving outpatient medical care for HIV [16]. For the 2009 data collection cycle, U.S. states and territories were sampled, followed by outpatient facilities providing HIV care, and finally by HIV-infected adults aged 18 years and older who reported at least one medical care visit in a participating facility during January–April 2009. For state and territory samples, probability of selection was proportionate to AIDS prevalence; for provider samples, probability of selection was proportionate to HIV-infected patient census. Data were collected via face-to-face interviews and medical record abstractions during June 2009–May 2010. All sampled states and territories participated in MMP [California (including Los Angeles County and San Francisco), Delaware, Florida, Georgia, Illinois (including Chicago), Indiana, Michigan, Mississippi, New Jersey, New York (including New York City), North Carolina, Oregon, Pennsylvania (including Philadelphia), Puerto Rico, Texas (including Houston), Virginia, and Washington]. Of 603 sampled facilities, 461 participated in MMP (facility response rate, 76%); of 9338 sampled persons, 51% (n = 4127) were interviewed and had their medical records abstracted, resulting in a combined facility-patient response rate of 39% [17]. Data were weighted on the basis of known probabilities of selection at state or territory, facility and patient levels [18]. In addition, using information collected on all sampled facilities and 88% of sampled patients, we conducted an analysis to compare respondents and nonrespondents. Data were then weighted to adjust for nonresponse by using predictors of response, including facility size, facility type (public or private), patient race and ethnicity, time since HIV diagnosis and age group [18,19]. We analysed data on 4094 participants whose diagnosis was made at least 12 months earlier and weighted the data to represent 408 092 HIV-infected adults in care in the United States during January–April 2009. The Centers for Disease Control and Prevention (CDC) has determined that MMP is a public health surveillance activity [20]. Because MMP is not considered research, it is not subject to human subjects regulations including federal institutional review board (IRB) [21]. Despite this designation, voluntary informed consent was obtained from all participants who completed an interview. Several participating states, territories and facilities obtained local IRB approval to conduct MMP and also obtained informed consent to abstract information from patients’ medical records.

Among HIV-infected persons in care whose diagnosis occurred at least 12 months earlier, 46% (95% CI, 42–51) were MSM, 24% (95% CI, 21–26) were MSW, 27% (95% CI, 23–30) were WSM, 2% (95% CI, 1–2) were transgender and less than 1% (95% CI, 0.4–1.1) were WSW. Approximately 40% of persons were aged 40–49 years, 41% were non-Hispanic blacks and 50% had more than a high school education (Table 1). In the past 12 months, 89% had been prescribed ART, and 59% had been virally suppressed at all measurements; 68% had ever received an AIDS diagnosis. We found statistically significant differences in the sociodemographic and clinical characteristics of MSM, MSW and WSM except for continuous healthcare coverage in the past 12 months, length of time since HIV diagnosis and number of viral load measurements in the past 12 months.

Overall, 62% reported oral, anal or vaginal sex with one or more partners in the past 12 months; 13% reported an STI; and 56% reported not having been tested for STIs. In addition, 24% drank alcohol and 11% used drugs before or during sex (Table 2). MSM were significantly more likely than MSW or WSM to have had two or more partners and to have drunk alcohol or used drugs before or during sex; MSM, compared with MSW, were significantly more likely to report an STI. MSW, compared with MSM or WSM, were significantly more likely to report that they had not been tested for an STI; compared with WSM, MSW were more likely to have used noninjection drugs and report exchanging sex for money or goods. WSM, compared with MSM, were significantly more likely to report no sex partners, and they were more likely than MSW to report an STI.

We found that the majority of HIV-infected adults receiving care in the U.S. did not engage in sexual risk behaviours that have the potential to transmit HIV. However, of the 12% of persons who engaged in unprotected vaginal or anal sex with at least one partner of negative or unknown HIV status, approximately half did so while not virally suppressed. This analysis provides the first nationally representative estimates of sexual risk behaviour since the HIV Cost and Utilization Study, which was conducted in 1996, in the pre-ART era [3]. Our estimates – that 30% of HIV-infected MSM engaged in unprotected vaginal and anal sex and that 13% did so with a partner of negative or unknown HIV status – are lower than estimates from a recent meta-analysis indicating that 43% of HIV-infected MSM engaged in unprotected anal sex and that 26% did so with a partner of unknown or negative status [14]. However, a study that sampled HIV-infected persons from clinics and community organizations in four large U.S. cities found that 16% of MSM, 13% of MSW and 19% of WSM had unprotected sex with partners of negative or unknown HIV status, findings that are more consistent with our estimates of 13, 9 and 15%, respectively [27].

Since ART became widely available, concerns have been raised that the perceived reduction in sexual transmission risk secondary to viral suppression might lead to increased sexual risk behaviour. The 2012 recommendation by the U.S. Department of Health and Human Services [28] to provide ART regardless of CD4⁺ T-lymphocyte count further emphasized the importance of understanding the relationship between viral suppression and sexual risk behaviour. In the HIV Prevention Trials Network (HPTN) 052 study [5], which demonstrated that early initiation of ART reduces the likelihood of heterosexual HIV-1 transmission, condom use was extremely high among participants. However, condoms and behavioural-risk reduction interventions were provided to all participants; therefore, this level of behavioural risk reduction may not be achieved in real-world settings [10]. Most previous studies evaluating the association between ART use and sexual risk behaviour in the United States, among both MSM and heterosexual men and women, have not found ART use to be associated with increased sexual risk behaviour [12,14]. As our data are cross-sectional, we were unable to evaluate changes in risk behaviour before and after ART initiation. However, we found no evidence that people receiving ART were more likely to engage sexual risk behaviour than those who were not. In fact, we found that persons who were virally suppressed were less likely to engage in vaginal or anal sex, unprotected vaginal or anal sex, and unprotected vaginal or anal sex with a partner of negative or unknown HIV status. Furthermore, those persons who were prescribed ART were less likely to engage in risk behaviours, regardless of viral suppression, which suggests that ART use mediated the relationship between viral suppression and sexual behaviours. This finding may be a result of prevention interventions delivered in the care setting, possible ART side effects that could decrease libido (e.g. depression or anxiety) or an indication that persons who are prescribed ART differ from those who are not. Although our analysis was not designed to address the causal relationship between ART use, viral suppression and sexual risk behaviour, it indicates that virally suppressed persons were less likely to engage in risky sexual behaviours. In other words, those individuals who were not virally suppressed, and therefore at a greatest risk of transmitting HIV [5], were also more likely to engage in sexual behaviours that have the potential to transmit HIV. This suggests that persons not on ART and not virally suppressed may be in need of targeted prevention interventions.

The association between viral suppression and sexual behaviours was not statistically significant for all groups. MSM who were virally suppressed were significantly less likely to engage in all three measures of sexual risk behaviour than HIV-infected MSM who were not virally suppressed. For MSW and WSM, the trend was the same in regard to vaginal or anal sex, but none of the differences were statistically significant for MSW. The only statistically significant association for WSM was engagement in vaginal or anal sex. Although this null result could, in part, be explained by the smaller sample size of these groups, our findings are consistent with those of other studies that found no association between viral suppression and risk behaviour [29,30].

Compared with MSW, MSM in our analysis were more likely to report high-risk sexual behaviour, which is consistent with previous studies [3,29,31]. This increased prevalence of risk behaviours, combined with the increased per-act transmissibility of HIV through anal intercourse (both insertive and receptive) [31], highlights the pressing need for effective prevention efforts for MSM. Implementation of prevention efforts will require careful consideration of biological, network, social and structural factors (e.g. stigma, discrimination and homophobia) among MSM [32].

Combination prevention approaches, including behavioural risk reduction counselling, and access to ART and medical care should be available to all HIV-infected adults to maximize the individual and public health benefit. However, recent data on the HIV care continuum indicate that additional work is needed at every step in the continuum: expanded HIV testing, increased numbers of HIV-infected adults linked to and retained in care, and increased numbers of persons who are prescribed ART and ultimately achieve viral suppression [33]. The fact that only 25% of HIV-infected persons in the United States are virally suppressed highlights the need for action [34]. In addition, the delivery of behavioural risk reduction interventions in the clinical setting is essential. Medical visits provide an opportunity to discuss risk behaviours, reinforce prevention messages, diagnose and treat STIs, and emphasize medication adherence. Effective evidence-based interventions that can be delivered by providers as brief messages during clinic visits are available [35], but too few providers discuss HIV prevention with their patients [33]. STI screening and treatment, when indicated, is recommended by guidelines and is likely to reduce HIV transmission [36], but data indicate that too few patients are being screened by their providers [37]. Combination prevention approaches will require additional efforts to engage and retain HIV-infected persons in care and improve the delivery of risk reduction interventions in clinical settings.

Our analysis has limitations. First, MMP data reflect the experiences of HIV-infected adults receiving medical care in the United States; thus, our results cannot be generalized to persons unaware of their HIV infection or to those not receiving medical care. Sexual behaviours and viral suppression status are likely to differ among persons in care and persons not in care [38]. Also, we did not collect information on whether persons were aware of their current viral load status, which could influence subsequent risk behaviour. In addition, given less than optimal response rates, nonresponse bias is possible. Furthermore, because compliant patients may have been more likely to participate in MMP, it is possible that results may differ for less compliant patients. However, we collected information on sampled patients (sex, age, race, length of time since diagnosis) and facilities (HIV patient case load) and were able to compare the characteristics of respondents and nonrespondents. Our estimates were then weighted to minimize nonresponse bias. In addition, our probabilistic sampling frame was rigorously constructed, geographically diverse, included both urban and rural clinics, public and private facilities, providers who see many and few HIV-infected patients and jurisdictions that have high, medium and low HIV prevalence. Furthermore, empirical research indicates that low response rates are not necessarily indicative of nonresponse bias, particularly when probabilistic samples are drawn from rigorously constructed frames [39]. Next, because this is a cross-sectional analysis and we did not collect dates for the sexual behaviour data (beyond knowing they occurred in the 12 months before the interview), we cannot determine the exact timing of sexual risk behaviour with respect to viral suppression. Finally, sexual risk behaviour data were collected during face-to-face interviews, so social desirability bias may have led to underreporting of some behaviours [40].

The following are the contributions of the authors to this analysis: C.L.M, M.F. and J.S.: analysis conception, data analysis, wrote the article; E.F., P.H., C.J. provided statistical and data analysis support; J.F., L.B., E.E.V, C.S., A.D.M, P.S., A.L., J.M. and J.H. were involved in conception, design and/or implementation of the MMP. All authors contributed to data interpretation, article writing and/or review and editing. The authors thank the MMP patients, facilities and Community and Provider Advisory Board members. They also thank the 2009 MMP staff: http://www.cdc.gov/hiv/pdf/research_mmp_studygroupmembers_2009.pdf. The authors are grateful to Ms. Marie Morgan for her editorial assistance and the reviewers at AIDS for their valuable input on the manuscript.

The findings and conclusions in this report are those of the authors and do not necessarily represent the views of the CDC.

Funding for the MMP is provided by the CDC.