Trends in condom use among MSM in the United States: the role of antiretroviral therapy and seroadaptive strategies

Abstract

Objective

Evaluate changes in condomless anal sex at last sex among men who have sex with men (MSM) and assess if these changes are associated with the adoption of serosorting and biomedical prevention.

Design

The National HIV Behavioral Surveillance is a crosssectional survey done in up to 21 cities in 2005, 2008, 2011 and 2014.

Methods

MSM were recruited through venue-based sampling. Among men reporting at least one male partner, we evaluated changes in condomless anal sex at last sex with a partner with (1) HIV-concordant (proxy for serosorting) or (2) HIV-discordant (discordant/unknown) status. We hypothesized that if concordant condomless sex was increasing while discordant was stable/declining, the increases could be driven by more men attempting to serosort. We used generalized estimating equations assuming a Poisson distribution and robust variance estimator to explore whether temporal changes in the outcomes varied by selected characteristics. We also assessed changes in condomless anal sex by antiretroviral therapy (ART) use among HIV-positive MSM.

Results

Among 5371 HIV-positive MSM, there were increases in concordant (19% in 2005 to 25% in 2014, P < 0.001) and discordant condomless sex (15 to 19%, P < 0.001). The increases were not different by ART use. Among 30 547 HIV-negative MSM, concordant (21 to 27%, P < 0.001) and discordant condomless sex (8 to 13%, P < 0.001) increased.

Conclusion

Our data suggest that condom use decreased among MSM and that the trends are not explained by serosorting or ART. Promotion of condoms and increased access to preexposure prophylaxis are vital to ensure that the benefits of ART in reducing transmission of HIV are not undermined.

Introduction

Condom use affects the likelihood of HIV sexual transmission when a contact is made between an infected and susceptible individual [1] and is one of the key indicators measured in behavioral surveillance and prevention research. Although condoms can reduce the risk of HIV transmission, they do not eliminate risk and are often not used consistently [2]. Some MSM attempt to decrease their HIV risk by engaging in seroadaptive practices such as serosorting and seropositioning [3]. Seroadaptive practices have been shown to reduce the risk of HIVacquisition compared with having no strategy but increase the risk of infection compared with consistent condom use [3]. Other biomedical prevention strategies are now available, such as treatment as prevention and preexposure prophylaxis (PrEP) [4]. Some men may also choose these strategies instead of using condoms.

Previous analyses from CDC’s National HIV Behavioral Surveillance (NHBS) showed that condom use among MSM declined 20% from 2005 to 2011 [5]. This follow-up analysis includes new data (2014) and investigates if the increases in condomless sex were associated with use of other perceived prevention strategies. We investigated trends in concordant condomless sex at last sex as a proxy for serosorting and trends in the adoption of the insertive or receptive role when having condomless sex as an indication of seropositioning. We compared the trends in condomless sex among HIV-positive men on antiretroviral treatment (ART) versus not on ART to investigate if the overall increases were mainly due to reliance among HIV-positive MSM on the protective role of ART. Finally, we investigated the association of PrEP use and condomless sex.

Methods

Serial crosssectional data for NHBS among MSM are collected every 3 years. Data are from MSM recruited for interviews and HIV testing through venue-based, time–space sampling in 2003–2005 (referred to as 2005), 2008, 2011 and 2014. The first MSM cycle of NHBS in 2005 included the following cities: Atlanta, Georgia; Baltimore, Maryland; Boston, Massachusetts; Chicago, Illinois; Denver, Colorado; Fort Lauderdale, Florida; Houston, Texas; Los Angeles, California; Miami, Florida; Newark, New Jersey; New York City, New York; Philadelphia, Pennsylvania; San Diego, California; San Francisco, California and San Juan, Puerto Rico. The second MSM cycle of NHBS in 2008 included all the cities in the first cycle except Fort Lauderdale, Florida; and the following cities: Washington District of Columbia, Dallas, Texas; Detroit, Michigan; New Orleans, Louisiana; Nassau-Suffolk, New York; St. Louis, Missouri; and Seattle, Washington. The third and fourth MSM cycles of NHBS included all the cities in the second cycle except for St. Louis, Missouri. NHBS eligibility criteria and procedures have been previously published [6,7]. Activities were approved by local institutional review boards in each participating city and by the CDC.

Measures

Information on the HIV status of partners was only available for the last sexual partner. Therefore, we used information on the last anal sex act (condom use, insertive or receptive position) and self-reported HIV status of the participant and his last sex partner to construct the analyses outcomes. HIV status was assessed with the following questions: “What was the result of your most recent test”? And “Before your test on [DATE] did you ever test positive for HIV”? Self-reported HIV-positive were those who answered positive to the first question or “yes” to the second question. Self-reported HIV-negative were those who answered negative to the first question. And self-reported unknown status were those participants who answered “don’t know” or never obtained results to the first question and “no” to the second question. Participants were asked if they knew the HIV-status of their last sex partner at the time of last sex and, if so, what was their partner’s HIV-status. The following condomless anal sex outcomes were analyzed [8]: (1) any; (2) concordant (partner of the same HIV status, proxy for serosorting), (3) discordant (partner of discordant/ unknown status), (4) receptive and (5) insertive. In 2005, 12% had missing information on the most recent partner, and it was assigned using imputation. In 2015, date of last sex for last main and casual partners was collected. Based on the most recent date, we selected the last sexual partner. When it was unclear between the last type because of similar dates or because it was missing date were imputed. The proportion of casual and main partners for imputation determined from entire sample combining all cycles (2005–2014). For each person for which we imputed, we simulated a uniform (0,1) random variable. If that random number was less than or equal to the computed proportion of main partners in the pooled sample the person unknown partner taken to be a main partner otherwise it was a casual partner.) Data imputation was not necessary for other years of data. We hypothesized that if the proportion of men reporting concordant condomless sex increased over time while discordant sex remained stable or declined and that the overall increases in condomless sex could be driven by increased reliance on serosorting strategies. We evaluated the position at last sex as a proxy for seropositioning. Receptive condomless sex with a discordant partner at last sex among HIV-positive MSM and insertive condomless sex at last sex among HIV-negative MSM was considered lower risk than the opposite role or practicing both positions at last sex. If the increases in condomless sex were because of an increased reliance on seropositioning, we expected the lower risk practice to increase while the other behaviors would remain stable or decrease. Outcomes are presented separately by the self-reported HIV status of the participants.

Analysis

The analysis included all MSM who had a sex partner in the past 12 months, irrespective of whether they had an HIV test through NHBS. All cities that participated in any of the four cycles of NHBS were included. Separate Poisson models using Generalized Estimating Equations (GEE) were used to test for a linear trend between 2005, 2008, 2011 and 2014 for each of the outcomes. All models include year, age, race and city. Models for concordant and discordant condomless sex included interactions terms for year by age and race if significant (defined as P < 0.05). Year was treated as a continuous variable. Additional models were run among HIV-positive individuals for concordant and discordant condomless sex including an interaction term for year by ART. Among MSM of unknown status, we did not evaluate seroadaptive practices. To account for PrEP, we conducted a sensitivity analysis excluding men who reported being on PrEP in 2011 and 2014 for any, discordant and concordant condomless sex.

Results

The percentage of black participants and those recruited at bars and clubs increased from 2005 to 2014. Other characteristics of the sample remained unchanged (Supplemental Table, http://links.lww.com/QAD/A926).

Concordant and discordant sex

Among HIV-positive MSM, condomless sex at last sex increased from 34% in 2005 to 45% in 2014 (P < 0.001). There were increases in both concordant (P < 0.001) and discordant condomless sex (P < 0.001), although the former was more common (Table 1). These increases did not differ significantly by age, race or ART use (Supplemental Fig., http://links.lww.com/QAD/A926). However, in 2014 most HIV-positive MSM were on ART (90%) and of those reporting discordant condomless sex, 88% were on ART (258/293).

Table 1.

Sexual behaviors at last sex among sexually active MSM by self-reported HIV status, National HIV Behavioral Surveillance, United States, 2005–2014.

	2005		2008		2011		2014		Adjusted P valuea
	n	(%)	n	(%)	n	(%)	n	(%)
HIV-positive MSM
Any condomless anal sex (CAS)	489	34.2	410	37.3	495	39.8	703	44.5	<0.001
Concordantb CAS	279	19.0	231	21.1	318	25.6	401	25.4	<0.001
Receptive onlyc	101	7.3	71	6.5	109	8.8	151	9.6	<0.001
Any insertived	178	12.3	160	14.6	209	16.9	250	15.8	<0.001
Discordante CAS	210	15.0	177	16.1	174	14.0	301	19.0	<0.001
Receptive onlyc	102	6.8	87	7.9	90	7.3	167	10.6	<0.001
Any insertived	108	7.7	90	8.2	84	6.8	134	8.5	0.3426
Total	1440	100.0	1101	100.0	1244	100.0	1586	100.0
HIV-negative MSM
Any CAS	2522	28.7	2304	32.8	2472	34.7	2998	40.5	<0.001
Concordantb CAS	1828	21.2	1576	22.4	1700	23.9	2023	27.4	<0.001
Insertive onlyf	854	9.9	699	9.9	740	10.4	847	11.5	<0.001
Any receptiveg	972	11.3	877	12.5	960	13.5	1174	15.9	<0.001
Discordante CAS	692	7.6	724	10.3	768	10.8	971	13.1	<0.001
Insertive onlyf	387	4.2	369	5.3	404	5.7	521	7.0	<0.001
Any receptiveg	304	3.4	354	5.0	363	5.1	448	6.1	<0.001
Total	8943	100.0	7049	100.0	7131	100.0	7424	100.0
Unknown HIV-status MSM
Any CAS	259	27.2	362	32.8	324	37.0	240	38.5	<0.001
Total	1068		1103		878		626

Note: Denominator may vary for certain outcomes because of missing data. Numbers and percentages for 2005 include imputed values.

^aSeven models included for HIV-positive and HIV-negative MSM on the following outcomes: any condomless sex at last anal sex; concordant condomless anal sex, concordant insertive condomless anal sex, concordant receptive condomless anal sex, discordant condomless anal sex, discordant receptive condomless anal sex and discordant insertive condomless anal sex. For MSM of unknown HIV-status, only one model is presented for any condomless anal sex. We used generalized estimating equations models using a robust variance estimate and assuming a Poisson distribution to test for a linear trend between 2005, 2008, 2011 and 2014 for each of the outcomes. All models include year, age, race and city.

^bCondomless anal sex at last sex with a partner of the same HIV status.

^cRespondent assumed receptive role during last sexual episode.

^dRespondent assumed insertive role during last sexual episode and includes those reporting both insertive and receptive sex.

^eCondomless anal sex at last sex with a partner of discordant or unknown HIV status.

^fRespondent assumed insertive role during last sexual episode.

^gRespondent assumed receptive role during last sexual episode and includes those reporting both insertive and receptive sex.

Overall (all years), 41% of HIV-negative MSM reported a discordant partner (37% with a partner of unknown status and 4% with an HIV-positive partner). Among HIV-negative MSM condomless sex, at last anal sex increased from 29% in 2005 to 41% in 2014 (P < 0.001) (Table 1). Both concordant and discordant condomless sex increased (both P < 0.001). The increases did not vary by race. The increase did vary by age (P = 0.003) and was greatest among MSM ages 18–24 years. (Fig. 1).

Fig. 1. Concordant condomless sex among HIV-negative MSM by age, National HIV Behavioral Surveillance, United States, 2005–2014.

Adjusted increase in concordant condomless sex per every 3 years was for 18–24 years 16% (CI: 12%, 20%), 25–29 years 9% (CI: 5%, 12%), 30–39 years 9% (CI: 5%, 12%) and 40 years and older 7% (3%, 11%).

Insertive and receptive sex

Among HIV-positive MSM, an increase was noted for discordant receptive condomless sex (P < 0.001) while the percentage engaging in discordant insertive sex (P = 0.34) remained unchanged. Among HIV-negative MSM, both insertive and receptive anal sex with a discordant partner increased from 2005 to 2014 (both P < 0.001).

Preexposure prophylaxis use

PrEP use among HIV-negative MSM was 0.5% in 2011 and 3.5% in 2014. When excluding men on PrEP from the analyses, the increasing trend in any concordant and discordant condomless anal sex remained unchanged.

Discussion

The analysis explored whether documented increases in condomless sex among MSM were associated with adoption of other prevention strategies. We found that among HIV-negative MSM condomless sex increased with both concordant and discordant partners, and there was no indication that seroadaptive behaviors were associated with the increases. In the years studied, PrEP use was too low to contribute to the increases in condomless sex and excluding MSM on PrEP from the analyses did not change the results. Although discordant condomless receptive anal sex was rare among HIV-negative MSM, our data suggest that this practice has increased. While we found increases in concordant and discordant condomless sex among HIV-positive MSM, we also found evidence of seroadaptive behaviors among HIV-positive MSM; the increase in discordant condomless anal sex was only statistically significant for receptive sex but not insertive sex, which carries higher risk of HIV transmission. The data also suggest that HIV treatment does not explain the increase in condomless sex among HIV-positive MSM; however, most HIV-positive MSM were on ART.

These findings corroborate previous reports of increases in condomless sex among MSM in the United States [9]. We cannot establish if these trends are contributing to the documented increase in HIV incidence among MSM. However, the increase in concordant condomless sex among HIV-negative MSM was highest among the youngest age group, among whom increases in the number of new HIV diagnoses have occurred [10]. Men may perceive themselves and their partners to be HIV-negative; however, many men who are HIV-positive are not aware of their infection, and awareness is lowest among the youngest age groups [11]. Although men could be choosing other prevention strategies such as PrEP, this strategy was uncommon and did not explain the increases in condomless sex. Men with discordant partners could be choosing not to use condoms if their HIV-positive partner is on ART. However, most discordant partnerships among HIV-negative MSM in this analysis were with a partner of unknown status as opposed to an HIV-positive partner. There may be other reasons for the increases in condomless sex that were not explored such as changing social norms around condom use [12].

Similar increases in condomless sex have been reported from other developed countries [13–17]. Mathematical modeling from the United Kingdom [18] and the Netherlands [19] suggests that reductions in HIV incidence due to ART and earlier HIV diagnosis have been offset by increases in condomless sex among MSM.

The findings in this report are subject to limitations. NHBS data are from MSM who were recruited at venues in cities with high AIDS burden. Thus, results may not be generalizable to all MSM. Further, analyses were based on self-reported data and may be subject to social-desirability bias. Several studies have previously documented misreporting of HIV status [20]. However, trends by HIV-status were similar, and we do not believe the potential biased introduced by this would affect the conclusion of our analyses. It is not possible to fully exclude a methodological change in NHBS contributing to our findings. One major change in NHBS methods was the inclusion of sexual behavior questions during eligibility screening starting in 2011. Requiring disclosure of sexual behaviors at the time of screening could have differentially selected participants who were more comfortable with disclosing their sexuality. Another possible explanation for our findings is that MSM may be more willing to disclose their risk behaviors in later years if stigma associated with HIV infection or homosexuality is decreasing, for which some evidence exists [21,22]. Finally, data are not weighted to account for the complex sampling methodology used to recruit MSM. Point estimates may therefore be biased by over-represented or under-represented subgroups of the population. We did not present behaviors by partner type, because condomless sex with male partners is also risky [23].

Our data suggest that condom use has decreased among MSM and that the trends are not explained by serosorting, seropositioning, PrEP use or HIV treatment and should continue to be monitored. The promotion of condom use among HIV-negative as well as HIV-positive MSM remains vital to ensure that the benefits of ART in reducing transmission of HIV are not undermined. However, MSM comprise diverse populations that vary in sociodemographic and behavioral characteristics and have different prevention needs and preferences. In this new era of HIV prevention, MSM have more tools available to them than ever before. There are new strategies, such as PrEP for those who are negative and at high risk and treatment as prevention for HIV-positive men, and previously recognized methods that can substantially reduce risk such as using condoms consistently and correctly. As no single strategy provides complete protection in real-world use, multiple approaches are needed to reduce new HIV infections.