Growing acceptability of Undetectable = Untransmittable but widespread misunderstanding of transmission risk: Findings from a very large sample of sexual minority men in the United States

Abstract

Background :

Several studies have demonstrated no linked HIV transmissions in serodifferent sexual encounters where the partner with HIV has an undetectable viral load (VL). As a result, awareness and dissemination of treatment as prevention (TasP), and movements like “Undetectable = Untransmittable” (U=U), have grown.

Setting:

We conducted an online cross-sectional survey from November, 2017 through September, 2018 to gather data from a total of 111,747 sexual minority men (SMM) in the U.S.

Methods:

Participants provided sociodemographic data, and answered questions regarding biomedical status, HIV and STI prevention behaviors, drug use, condomless anal sex, and perceived accuracy of the U=U message. We conducted analyses to understand factors associated with perceived accuracy of U=U stratified by HIV status.

Results:

Overall, 53.2%% of the sample perceived U=U as accurate, with the highest rates among HIV-positive SMM (83.9%), followed by HIV-negative (53.8%) and status-unknown (39.0%) SMM. Multivariable models showed a 2-3% increase per month in accuracy beliefs among SMM. Consistent with prior work, there was greater heterogeneity among HIV-negative and unknown men, with several factors differentiating perceived accuracy, compared to SMM with HIV. Perceived transmission risk levels with undetectable partners were skewed well above accurate levels, and perceptions of transmission risk was associated with lower perceived accuracy of U=U.

Conclusions:

Public confidence in TasP and U=U are growing, but clear, unequivocal messaging about the effectiveness of U=U is critical. Due to misunderstandings of risk, language that focuses on protective benefits rather than transmission risks may reach more people and allow better comparisons with PrEP and condoms.

Introduction

Scientific consensus has been building for many years around the notion that successful antiretroviral treatment (ART) of HIV could suppress viral load (VL) to levels that would limit if not wholly avert HIV transmission, a phenomenon now known as treatment as prevention (TasP)¹. In 2008, scientists at the Swiss National AIDS Commission were among the first to publicly state that individuals living with HIV who were consistently using ART and had an undetectable VL presented with a significant reduction in transmission risk,² sparking a debate and a call to action to provide more empirical data³. Initially, an analysis of annual surveillance data from San Francisco demonstrated that as VL levels within a defined community (i.e., community VL) decreased, HIV incidence within that community also decreased⁴. Three years later, the HIV Prevention Trials Network (HPTN) presented the results of HPTN-052, a Phase-III randomized controlled study that followed 1,763 serodiscordant couples for over a decade. The study initially demonstrated a reduction of 96% in the likelihood of transmission occurring when ART was initiated early⁵ and later analyses demonstrated no linked HIV transmissions within couples where the HIV-positive partner had durably suppressed VL^6,7.

While these landmark studies provided evidence for Treatment as Prevention (TasP), less was known in terms of serodifferent sexual minority men (SMM) couples and condomless anal sex (CAS). The PARTNER1 trial included 340 serodifferent male couples and demonstrated no linked transmissions when the HIV-positive partner’s VL was suppressed⁸. Two subsequent trials of serodifferent male couples—Opposites Attract with 234 couples⁹ and PARTNER2 with 783 couples—have confirmed no linked transmissions when the partner with HIV was undetectable¹⁰. The PARTNER2 findings were in the context of 76,088 recorded CAS acts, leading the authors to conclude: “the risk of HIV transmission in gay couples through condomless sex when HIV viral load is suppressed is effectively zero.”¹⁰

Although scientific consensus regarding TasP has increased rapidly with these findings, the science of VL suppression and transmission risk have not been well understood or accepted among the general public and key groups living with HIV. In 2016, the Prevention Access Campaign launched the slogan “Undetectable = Untransmittable” or “U=U” to promote advocacy for and education about these scientific breakthroughs. However, even with a recent commentary published by officials at the National Institute of Allergy and Infectious Diseases supporting TasP as scientifically sound,¹¹ an official endorsement from the Centers for Disease Control that there is “effectively no risk,”¹² and advocacy from more than 926 organizations across 99 countries,¹³ acceptance of the science surrounding VL transmission has been mixed. Dissemination and acceptance of this science through U=U and other messaging is critical for HIV prevention and a central component of the U.S. initiative toward “Ending the HIV Epidemic” by 2030¹⁴. Thus, a greater understanding of barriers to uptake of this information among the general public is vital.

Only a handful of published studies to date have documented levels of and barriers to TasP and U=U acceptability among the general public. A Canadian study conducted between 2012 and 2015 assessed belief in TasP among 774 SMM found that support was greater among HIV-positive men (47.7%), with nearly all (93.9%) of HIV-negative men reporting to be skeptical and unaware of TasP¹⁵. A study of 732 SMM in New York City conducted in 2016-17 also found HIV-positive men to perceive TasP as more effective (58.3%) in comparison to HIV-negative men (34.4%)¹⁶. In the largest study to date with over 12,200 SMM conducted in 2016-17,¹⁷ researchers found that far fewer (30.0%) HIV-negative and unknown-status SMM perceived the message as accurate compared to HIV-positive SMM (64.0%). These findings highlight that the potential HIV-preventive benefit of U=U is hindered due to gaps in acceptance by HIV status. In addition to noting disparities by HIV status, research has also identified changes over time, with one Australian study of 1316 SMM noting an increasing belief in TasP from 2.6% in 2013 to 13.1% in 2015¹⁸. These and other studies have noted other factors being associated with greater beliefs in TasP and U=U, including higher engagement in CAS, sex with serodifferent partners, more interaction with HIV prevention or treatment services (e.g., more frequent HIV testing, use of PrEP, being prescribed and adherent to ART), and drug use, as well as higher rates of belief among Black and Latino SMM^16,19,20.

Since its launch in 2016, U=U’s message has been gaining momentum. Data suggest both the awareness and acceptance of the science of viral suppression among SMM has been increasing over time. Yet, a broader understanding of viral suppression and belief in the U=U message is needed for the greatest benefit ^15,18. Therefore, the present study aimed to expand on prior work to provide ongoing surveillance of accuracy beliefs in the U=U message among SMM, compare changes over time, describe factors associated with these beliefs, and examine how beliefs in the U=U message are associated with perceived risk of HIV transmission during CAS with undetectable partners.

Methods

Participants and Procedures

We gathered data during a 10-month span from November, 2017 through September, 2018 as part of ongoing efforts to screen participants into several research studies for SMM. We used online venues to advertise for these studies across the U.S., including targeted banner advertisements on social media sites, pop-up and inbox advertisements on popular geotargeted sexual/dating networking apps, and targeted banner advertisements on web-based sexual/dating networking websites. Those who clicked any of the ads containing images and text (e.g., “get a free at-home HIV test mailed to you”; “receive up to $275 for joining”) were directed to the secure survey. Those aged 18 or older were directed to a page that contained informed consent while those who were aged 13 to 17 were directed to an assent page, and the study received a waiver of parental consent. The informed consent/assent indicated the survey had no incentive, but they would be screened for multiple studies at once for which they could be compensated if they were eligible and enrolled. All procedures were approved by the Institutional Review Board of the City University of New York.

Measures

Biomedical status.

Participants reported their HIV status as positive, negative, or unknown (“I don’t know”), after which HIV-negative and status-unknown individuals were asked if they were currently prescribed PrEP and HIV-positive individuals were asked if their most recent VL test was undetectable, detectable, or unknown (“Not sure/don’t remember”). We combined the answers to form five biomedical status groups: (1) HIV-negative/unknown, on PrEP; (2) HIV-negative, not on PrEP; (3) HIV status unknown, not on PrEP; (4) HIV-positive, undetectable; and (5) HIV-positive, detectable or unsure.

Sociodemographic characteristics.

We recorded whether participants were recruited from a social media website, a sexual/dating/social networking website (herein referred to as a “networking site”), or a sexual/dating/social networking app (herein referred to as a “networking app”). Participants self-reported their age, Hispanic/Latino ethnicity, racial identity based on U.S. Census categories, gender identity, sexual orientation, zip code (which we recoded into the four primary regions of the U.S.), and their relationship status—for men in relationships, we asked the HIV status of their main partner, which we subsequently recoded into a seroconcordant or serodifferent relationship status. Participants who reported an unknown HIV status were recoded as being in a serodifferent relationship.

HIV and STI prevention and treatment.

HIV-negative and unknown-status participants reported the frequency with which they received HIV testing, which we recoded into a trichotomous variable of testing in the past six months, more than six months, or never tested.

ART adherence.

HIV-positive participants respond to a single validated item for assessing antiretroviral adherence²¹ ranging from 1 (very poor) to 6 (excellent), which was trichotomized into a variable indicating excellent adherence, less than excellent adherence, and currently not being on ART.

Club drug use.

Participants were asked whether they had used cocaine, crack, crystal meth, ecstasy, GHB, and ketamine in the past 6 months, and we recoded responses into a dichotomous indicator of any recent club drug use.

Recent CAS.

Participants were asked the number of casual male sexual partners they had in the prior 6 months and the number of times they engaged in insertive and receptive anal sex with and without a condom with these partners. We created a dichotomous indicator of any CAS with a casual male partner in the prior 6 months.

Month of completion.

To examine change over time, we created a variable indicating the month the survey was completed (range: 1-10).

Perceived accuracy of the Undetectable = Untransmittable message.

We relied on measures used in prior research,¹⁷ in which participants were asked, “With regard to HIV-positive individuals transmitting HIV through sexual contact, how accurate do you believe the slogan Undetectable = Untransmittable is?” Responses were on a Likert-type scale from 1 (Completely inaccurate) to 4 (Completely accurate) as well as a fifth option (I don’t know what “undetectable” means). Within multivariable models, the fifth category was excluded and higher scores indicate greater perceived accuracy of U=U.

Perceived risk of transmission with an undetectable partner.

Participants were asked, “What is the risk that an HIV+ man who is currently undetectable could transmit HIV sexually to his partner through topping?” The question was repeated for receptive sex, replacing the final word with “bottoming.” Participants responded on a sliding scale from 0% to 100% in intervals of 1%, with anchors only at the extremes of “no risk” and “complete risk.”

Statistical Analyses

We used SPSS 24 to examine sociodemographic characteristics of the sample and compared those characteristics between the HIV-negative, HIV-positive, and unknown status participants using chi-square tests of independence. Next, multivariable regression analyses were run with cross-sectional responses to perceived accuracy of the message as the outcome; one model was run for SMM who identified as HIV-positive and another for those who identified as HIV-negative/unknown-status. Because of the 4-point nature of the scale response, we ran an ordinal (i.e., proportional odds) logistic regression of the accuracy ratings for each, excluding those who responded that they were unsure what undetectable meant. Finally, we examined perceived accuracy of U=U in relation to perceived risk of transmission during condomless sex with undetectable male partners using chi-square tests of independence.

Results

Overall, 204,816 individuals submitted an eligible age and reached the survey consent/assent, of whom 157,587 provided consent/assent and 123,373 completed the survey in its entirety. We removed 9,427 responses that were duplicate responses of previous surveys. An additional 2,199 individuals did not provide valid regional data, identify as male, report male-identified casual or main partners, or identify as a sexual minority. A total of 111,747 SMM provided full data on the measures focused on in this manuscript and thus constituted the analytic sample.

Sociodemographic characteristics of the sample are presented in Table 1. Looking by HIV status (without consideration of biomedical status), we saw significant sociodemographic differences between the three groups on all variables examined. Perceived accuracy of U=U is fully broken down in Table 1—collapsing the categories, we saw that 53.8% of HIV-negative men and 39.0% of HIV status-unknown men perceived U=U to be somewhat or completely accurate compared with the vast majority (83.9%) of HIV-positive men.

Table 1.

Demographic characteristics and comparisons by HIV status.

	Full Sample (N = 111,747)		HIV-Negative (N = 72,088)		HIV-Positive (n = 16,392)		Unknown Status (n = 23,267)
	n	%	n	%	n	%	n	%
Biomedical status					Not testeda
HIV-negative/unknown on PrEP	14250	12.8	14250	19.8	0	0.0	0	0.0
HIV-negative not on PrEP	57838	51.8	57838	80.2	0	0.0	0	0.0
HIV status unknown not on PrEP	23267	20.8	0	0·0	0	0.0	23267	100.0
HIV-positive undetectable	13915	12.5	0	0·0	13915	82.2	0	0.0
HIV-positive detectable/unsure	2477	2.2	0	0·0	2477	14.6	0	0.0
Recruitment source					χ2(4) = 255.53***
Social media/other	5683	5.1	3780	5.2	536	3.3	1367	5.1
Networking app	102718	91.9	66430	92.2	15199	92.7	21089	91.9
Networking website	3346	3.0	1878	2.6	657	4.0	811	3.0
Ethnicity					χ2(2) = 453.28***
Hispanic/Latino	27105	24.3	16495	22.9	3728	22.7	6882	29.6
Not Hispanic/Latino	84642	75.7	55593	77.1	12664	77.3	16385	70.4
Race					χ2(10) = 1964.65***
Black	15607	14.0	8850	12.3	3743	22.8	3014	13.0
White	69618	62.3	46577	64.6	9463	57.7	13578	58.4
Asian/Pacific Islander/Native Hawaiian	5407	4.8	3877	5.4	343	2.1	1187	5.1
Native American/Alaskan Native	1875	1.7	1153	1.6	231	1.4	491	2.1
Multiracial	11784	10.5	7279	10.1	1647	10.0	2858	12.3
Other	7456	6.6	4352	6.0	965	5.9	2139	9.2
Gender					χ2(2) = 151.09***
Cisgender Male	110691	99.1	71277	98.9	16376	99.9	23038	99.1
Transgender Male	1056	0.9	811	1.1	16	0.1	229	0.9
Sexual orientation identity					χ2(6) = 1688.65***
Gay	88497	79.2	56849	78.9	14624	89.2	17024	73.2
Queer	2846	2.6	2072	2.9	264	1.6	510	2.2
Bisexual	19562	17.5	12666	17.6	1444	8.8	5452	23.4
Straight	842	0.8	501	0.7	60	0.4	281	1.2
Region					χ2(10) = 679.34***
Northeast	21318	19.1	14843	20.6	2762	16.8	3713	16.0
Midwest	19775	17.7	12884	17.9	2452	15.0	439	19.1
South	39629	35.5	24199	33.6	6575	40.1	8855	38.1
West	29829	36.7	19441	27.0	4501	27.5	5887	25.3
U.S. Possession	1153	1.0	693	1.0	99	0.6	361	1.6
Military Overseas	43	0.03	28	0.03	3	0.01	12	0.04
Relationship status					χ2(4) = 10879.61***
Single	77774	69.6	49135	68.2	11006	67.1	17634	75.8
Partnered, seroconcordant	19826	17.7	17613	24.4	2213	13.5	0.0	0.0
Partnered, serodifferent	14147	12.7	5341	7.4	3173	19.4	5633	24.2
U=U Message Accuracy					χ2(8) = 14628.39***
Completely accurate	24654	22.1	13693	19.0	8403	51.3	2558	11.0
Somewhat accurate	36942	33.1	25083	34.8	5337	32.6	6522	28.0
Somewhat inaccurate	18843	16.9	12862	17.8	1339	8.2	4642	20.0
Completely inaccurate	22073	19.8	15527	21.5	1203	7.3	5343	23.0
Unsure what undetectable means	9235	8.3	4923	6.8	110	0.7	4202	18.1

^***p < 0.001.

^a= Data are descriptive of sample.

Table 2 reports the stratified regression results for the ordinal accuracy belief outcome, as well as the unadjusted percentage who endorsed belief in U=U (accurate or completely accurate) for each subgroup comparison to translate these findings into more meaningful indicators of practical significance. Among HIV-negative/unknown SMM, those on PrEP had substantially higher beliefs in the accuracy of U=U than those not on PrEP. Similarly, in the model for HIV-positive SMM, those who reported their most recent VL result was undetectable had substantially higher perceptions of the accuracy of U=U. Across both models, there were some notably consistent findings. The odds of moving up 1 point on the accuracy scale (which ranged from 1-4) increased by 2.0% per month for HIV-negative/unknown men and 3.0% by month for HIV-positive men. Men recruited from networking apps had lower perceived accuracy than those from social media, queer-identified men had higher and bisexually-identified men had lower perceived accuracy than gay-identified men, and men partnered in a serodifferent relationship had higher perceived accuracy than single men.

Table 2.

Multivariable analyses examining perceived accuracy of U=U message.

	HIV-Negative or Unknown (n = 86,230)				HIV-Positive (n = 16,282)
	Rated accurate	B	AOR	AOR 95% CI	Rated accurate	B	AOR	AOR 95% CI
Recruitment month (1 through 10)	-	0.01	1.01**	[1.0, 1.01]	-	0.02	1.02*	[1.00, 1.04]
Age (per 10 years)	-	0.01	1.01	[1.0, 1.02]	-	−0.12	0.89***	[0.87, 0.92]
Negative or Unknown Status (ref. = currently on PrEP)	79.1%				-
Negative, not on PrEP	52.1%	−1.02	0.36***	[0.35, 0.37]	-	-	-	-
HIV status unknown	47.4%	−0.92	0.40***	[0.38, 0.42]	-	-	-	-
Positive Status (ref. = Undetectable)	-				86.7%
Positive, detectable	-	-	-	-	71.1%	−0.76	0.47***	[0.43, 0.51]
Recruitment source (ref. = social media)	55.8%				86.1%
Networking app	55.6%	−0.08	0.92**	[0.87, 0.97]	84.5%	−0.22	0.81*	[0.68, 0.96]
Networking website	51.1%	−0.09	0.91*	[0.82, 1.0]	80.2%	−0.26	0.77**	[0.62, 0.97]
Ethnicity (ref. = Not Hispanic/Latino)	56.0%				84.4%
Hispanic/Latino	54.0%	−0.02	0.98	[0.95, 1.02]	84.5%	−0.01	0.99	[0.91, 1.08]
Race/ethnicity (ref. = Black)	54.7%				81.7%
White	56.3%	−0.04	0.97	[0.92, 1.0]	85.6%	0.18	1.20***	[1.11,1.30]
Asian or Hawaiian Native or Pacific Islander	55.4%	−0.09	0.92**	[0.86, 0.97]	83.3%	−0.09	0.92	[0.74, 1.14]
Native American or Alaskan Native	50.6%	−0.17	0.84**	0.76, 0.94]	82.4%	0.00	1.00	[0.77, 1.29]
Multiracial	53.9%	−0.08	0.92**	[0.87, 0.97]	84.2%	0.13	1.14*	[1.01, 1.29]
Other	53.2%	−0.03	0.97	[0.90, 1.04]	83.7%	0.07	1.07	[0.91, 1.26]
Gender (ref. = cisgender male)	55.5%				84.4%
Transgender male	57.9%	−0.01	0.99	[0.87, 1.12]	87.5%	0.14	1.15	[0.45, 3.19]
Sexual Orientation Identity (ref. = Gay)	58.2%				84.8%
Queer	68.1%	0.49	1.63***	[1.51, 1.77]	90.5%	0.52	1.69***	[1.32, 2.18]
Bisexual	42.6%	−0.46	0.63***	[0.61, 0.65]	79.4%	−0.22	0.80***	[0.73, 0.89]
Straight	34.5%	−0.71	0.49***	[0.42, 0.57]	80.7%	−0.11	0.89	[0.55, 1.49]
Region by Zip Code (ref. = Northeast)	57.2%				84.9%
Midwest	55.5%	0.04	1.04*	[1.00, 1.09]	84.2%	0.04	1.05	[0.94, 1.16]
South	53.8%	−0.03	0.97	[0.94, 1.01]	83.4%	−0.06	0.94	[0.86, 1.03]
West	56.8%	0.02	1.01	[0.98, 1.05]	85.7%	0.05	1.05	[0.96, 1.16]
U.S. Possession	46.2%	−0.08	0.93	[0.81, 1.05]	80.6%	0.03	1.03	[0.79, 1.56]
Military Overseas	47.2%	−0.3	0.74	[0.41, 1.35]	100.0%	−0.41	0.66	[0.10, 4.27]
Relationship status (ref. = single)	54.2%				83.6%
Partnered, seroconcordant	58.5%	0.06	1.07***	[1.03, 1.10]	84.2%	−0.04	0.96	[0.90, 1.05]
Partnered, serodifferent	61.0%	0.26	1.30***	[1.24, 1.35]	87.3%	0.17	1.19***	[1.10, 1.29]
Recent club drug use (ref. = no)	54.4%				84.9%
Yes	61.0%	0.16	1.17***	[1.14, 1.21]	83.4%	0.03	1.03	[0.97, 1.10]
Recent CAS (ref. = no)	51.8%				84.8%
Yes	55.8%	0.11	1.12***	[1.08, 1.17]	84.4%	0.02	1.02	[0.90, 1.15]
Frequency of HIV testing (ref. = past 6 months)	62.7%				-
More than 6 months ago	49.3%	−0.32	0.72***	[0.70, 0.75]	-	-	-	-
Never	38.6%	−0.66	0.52***	[0.49, 0.54]	-	-	-	-
ART Medication Adherence (ref. = excellent)	-				87.9%
Less than excellent	-	-	-	-	82.1%	−0.47	0.63***	[0.59, 0.67]
Not prescribed ART	-	-	-	-	67.5%	−0.94	0.39***	[0.34, 0.45]

^*p < 0·05;

^**p < 0·01;

^***p < 0·001.

B = unstandardized beta; AOR = Adjusted Odds Ratio. Results are from ordinal regression of the 4-point accuracy rating, whereas percentages rated accurate is reported as the unadjusted within-group percentage who rated the message as “accurate” or “completely accurate” for practical interpretation.

Within the model for HIV-negative and status-unknown men, we found that Black men had higher perceived accuracy than men who identified as Asian, Hawaiian Native, or Pacific Islander, those who identified as Native American or Alaskan Native, and those who identified as Multiracial, but did not differ from White men. Straight-identified men had lower perceived accuracy than gay-identified men, men from the Midwest had higher perceived accuracy than those from the Northeast, and men partnered in seroconcordant relationships had higher perceived accuracy than single men. We also found that recent club drug use and recent CAS were both significantly associated with higher perceived accuracy; HIV testing less often than every 6 months or never testing was associated with significantly lower perceived accuracy compared to men who tested every 6 months or more often.

Within the model for HIV-positive men, fewer significant differences emerged. Men from a networking site had significantly lower perceived accuracy than those from social media. Also, self-reported adherence less than excellent or not being prescribed ART were associated with significantly lower perceived accuracy compared with men who reported excellent adherence.

Finally, we found that perceived accuracy of U=U was strongly linked to perceived risk of HIV transmission during CAS with a known undetectable partner. Figures 1 and 2 show the distribution of perceived risk within a histogram for the overall sample and within box-and-whisker plots stratified by perceived accuracy of U=U, with standard notation for median and quartiles and single points for means within each group. As shown in the histogram on the left of Figure 1, only 10% of the sample overall perceived transmission risk to be zero when the undetectable partner was insertive during CAS and, in Figure 2, this was slightly higher at 14% when the undetectable partner was receptive. However, these proportions were 31.0% and 39.0%, respectively, among the group who perceived U=U to be completely accurate. Perceived accuracy of U=U was associated with significantly lower perceived risk of HIV transmission during CAS with an undetectable partner whether he is in the insertive (χ² = 31223.38, p < 0.001) or receptive (χ² = 25704.95, p < 0.001) position.

Figure 1.

Distribution of reported risk of HIV transmission during condomless anal sex when an undetectable HIV-positive individual is the insertive partner for the sample overall (histogram) and stratified by perceived accuracy of U=U (box-and-whisker plots).

Figure 2.

Distribution of reported risk of HIV transmission during condomless anal sex when an undetectable HIV-positive individual is the receptive partner for the sample overall (histogram) and stratified by perceived accuracy of U=U (box-and-whisker plots).

Discussion

In the present study, we examined perceived accuracy of U=U among SMM in the U.S. to inform ongoing TasP implementation efforts to reduce HIV transmission. Similar to prior studies, the current findings demonstrated that HIV-positive men continue to be more likely to consider TasP messaging as accurate compared to HIV-negative and unknown-status men. Among HIV-positive men, four in five considered the message to be somewhat or completely accurate.

Compared to previous work^15–17 we found that across all biomedical statuses, SSM were more likely to perceive U=U as somewhat or completely accurate, suggesting a positive shift in acceptance. In the present study, 53.8% of HIV-negative men and 39% of HIV status-unknown men perceived U=U to be somewhat or completely accurate. These rates of perceived accuracy show quite a shift for HIV-negative/unknown SMM since a study in 2016-17 with over 12,200 SMM,¹⁷ where fewer than one-third (30%) of HIV-negative and unknown-status SMM reported the message as accurate. Additionally, we observed the odds of a 1-point increase on the accuracy scale (ranging from 1-4) being 2% higher per month for HIV-negative/unknown men and 3% by month for HIV-positive men.

Even with these advancements, there are SMM—particularly HIV-negative—who continue to rate U=U messaging as inaccurate. The present results echo prior studies showing that men who report more risk behavior and more engagement with HIV treatment and prevention were more likely to believe in the accuracy of U=U¹⁷. Specifically, recent CAS was a strong predictor of higher perceived accuracy of U=U among the HIV-negative SMM. Because reporting higher accuracy of U=U may indicate better understanding of the nuanced per-act risk associated with different types of sexual behavior, aggregate measures of CAS that do not delineate subtypes with higher and lower risk may be higher among this group despite the potential for reduced risk of transmission overall. Relatedly, more frequent HIV testing was associated with significantly higher perceived accuracy among the HIV-negative SMM. These finding suggest that HIV prevention services continue to have the intended effect of increasing HIV prevention knowledge.

More striking, the growing acceptance of TasP messaging has led to an almost ubiquitous belief that TasP is accurate among SMM living with HIV, contributing to the growing gap between U=U perceptions among HIV-positive and HIV-negative or HIV-status-unknown men. In the present study, 83.9% of HIV-positive men perceived U=U to be somewhat or completely accurate, compared to other research that found lower rates of acceptance among this group (47.7%¹⁵, 58.3%¹⁶, and 36% ¹⁷). Among the SMM living with HIV, the two most salient factors associated with accuracy beliefs were about treatment engagement—those who reported their most recent VL test was detectable or were unsure what their most recent result was and those who either reported suboptimal adherence or not being prescribed ART had significantly lower accuracy beliefs than those who reported an undetectable result and excellent adherence. One potential reason for this is simply lack of knowledge or understanding around the underlying risk levels of different behaviors, which may be driven by lower engagement in and contact with HIV care. Another possibility is that this group reacts differently to the potential social and personal consequences of U=U, which has yet to be empirically explored in published work.

The present study was the first to assess both U=U beliefs and perceived transmission risk during sex with undetectable partners, and links lower belief in the accuracy of U=U to a fundamental misunderstanding of the underlying rates of transmission with undetectable partners. It is critical to note that reaching high rates of undetectable status has been shown to be the most effective method for curbing the HIV epidemic, in addition to having health and social benefits for people living with HIV themselves^8,15,22,23. Moreover, uptake of TasP and U=U for HIV-negative individuals is highly cost-effective, requiring no more than an accurate understanding of risk and the ongoing successful treatment of people living with HIV. Despite this, prior work shows much higher public awareness of and confidence in PrEP and condoms than VL suppression for risk reduction^16,19,24,25.

Communication about the science of VL suppression must be unequivocal to ensure the risk-reducing benefits of VL suppression reach their full potential for effectiveness and to reduce HIV stigma. Psychological literature suggests that framing is important, and a switch from change in risk to the degree of protection may also prove critical to increasing success of U=U messaging. Clear, unequivocal messaging about the effectiveness of U=U is critical—based on the fundamental misunderstanding of base rates of risk and psychological literature on framing, we recommend language that uses relative rather than overall risk (i.e., no risk to complete reduction in risk) and uses protection-enhancing rather than risk-reducing framing. In other words, describing U=U as being “100% effective” in protecting against HIV transmission may enhance acceptability among the groups who have been slower to accept the message to date.

Study Strengths and Limitations

We were able to recruit a large and diverse sample of SMM from across the U.S., though data collected was cross-sectional and self-reported, and thus we were unable to confirm factors like reported HIV status or levels of ART adherence for SMM who reported were HIV-positive. Although we showed perceptions being somewhat higher for each subsequent month of data collection, our cross-sectional data cannot longitudinally elucidate the within-person changes in perceived accuracy of the U=U message over time. While our sample was large, we had a relatively small proportion who identified as transgender men and no female-identified participants, preventing us from assessing beliefs among these groups or the general population. Finally, while several precautions were taken to safeguard the online recollection and validity of our data and we have a large sample, the online survey was anonymous and drew heavily from app-based recruitment and thus may not generalize to the broader population of SMM in the U.S.

Conclusions

Among SMM in the U.S., we found a greater proportion believe the message compared to similar prior research, with the overall proportion who believed the U=U message exceeding half for the first time in the published literature. In addition to showing a monthly trend for increasing accuracy beliefs, we demonstrated growing consensus around the accuracy among SMM living with HIV. At the same time, men who showed less engagement with HIV care, as evidenced by a recent detectable VL, ART non-adherence, or not being prescribed ART, had significantly lower beliefs. Several factors were associated with accuracy beliefs among HIV-negative SMM, with those more engaged in HIV prevention (i.e., on PrEP, more frequent HIV testing) and those engaging in potentially higher risk behaviors (i.e., CAS, drug use) rating the message as more accurate. In this study, perceived risk of HIV transmission with an undetectable partner was skewed well above zero, even among those who rated U=U as completely accurate. Individuals often misperceive and exaggerate base rates of risk,²⁶ and a focus on the effectiveness in averting risk (i.e., a relative risk reduction or degree of effectiveness) would mirror language used for PrEP and condoms and allow comparability among them. Taken together, these findings suggest people believe antivirals are more effective as PrEP than as TasP and highlight the need for clear and compelling messaging about there being zero risk of transmission with a partner who is durably suppressed and adherent to treatment. Finally, providers are a necessary partner in disseminating TasP and U=U science, and data suggest that even physicians may be more strongly influenced by relative risk information than absolute risk²⁷.