Daily Associations among Alcohol Intoxication, Partner Familiarity, Participant Effortful Control and Urgency and PrEP Uptake on Sexual Behavior in Men Who Have Sex with Men

Abstract

The purpose of this study was to examine the effects of alcohol intoxication and its interaction with contextual or situation (partner familiarity) and individual differences variables (effortful control, urgency, and whether taking pre-exposure prophylaxis (PrEP) medication) on sexual behaviors in men who have sex with men (MSM), a subgroup for whom HIV continues to be a major public health problem in the U.S. The participants were 236 men recruited from two northeastern U.S. cities and aged 21-50, M=27.8). These men participated in a 6-week (2, 3-week sampling bursts) experience sampling method (ESM) study. The ESM data were collected via use of software installed on the participant’s own or study-provided mobile phone. Individual differences variables were measured by participants’ completing questionnaires measuring effortful control and urgency, and the participant’s self-report of whether he was currently taking PrEP. The ESM data pertained to sexual behavior as well as situation variables of familiarity of relevant sexual partners and number of standard alcohol drinks consumed. The results generally were consistent with hypotheses, as alcohol intoxication showed a curvilinear relation to the occurrence of condomless anal intercourse. Furthermore, the likelihood of occurrence of condomless anal sex increased with increased familiarity of the sexual partner. Similarly, taking PrEP increased the likelihood of occurrence of condomless anal sex. At the same time, alcohol’s effects were moderated by all three individual differences variables as expected, but the prediction that partner familiarity would moderate alcohol’s effects on the occurrence of condomless sex was not supported. Clinical implications of the findings center on the application of the data to HIV prevention programs toward inclusion of more empirically-supported, nuanced information on the relation between acute alcohol intoxication and sexual behavior, Directions for further research address the need for additional testing and refinement of a person x situation approach to alcohol and sexual behavior. Furthermore, it is argued that it is important to refine further the concept of sexual risk in the context of taking PrEP. and to conduct more detailed, multivariate studies of the relation between taking PrEP and patterns of sexual behavior.

Despite considerable advances in HIV prevention interventions, HIV and other sexually transmitted infections (STIs) remain significant public health problems (Jiang et al., 2020). Data suggest that, in the United States, the extent of the public health problem is most acute in specific subgroups of the population, one of which is men who have sex with men (MSM; Centers for Disease Control and Prevention, 2020). Accordingly, the National Institutes of Health (U.S.) Office of AIDS Research 2021-2025 Strategic Plan emphasizes primary and secondary prevention efforts that target subpopulations that remain heavily affected by HIV/AIDS (National Institutes of Health, 2019). In this regard, increasing the effectiveness of behavioral interventions, or combined behavioral and biomedical interventions, requires a person-centered approach to their development and delivery (Brewer et al., 2019; Pantelic et al., 2018). In this case, a person-centered approach means that the design and implementation of interventions consider individual difference variables and the contexts in which these variables are manifested over time as determinants of health-related behaviors such as HIV/STI-related sexual risk behaviors.

Alcohol intoxication is one contextual or situation factor whose association with sexual behaviors has been studied intensively in multiple populations (e.g., MSM, college students, young adults) since HIV was identified as a major public health problem in the early 1980s. Alcohol consumption is a particularly important situation factor to consider in understanding the sexual behavior of MSM because that group tends to have a higher prevalence of alcohol use than the general population (Maisto & Simons, 2016; Medley et al., 2016). Overall, experimental studies have provided consistent evidence that alcohol intoxication, at least at moderate to high BACs, has a causal, facilitative relationship to risky decisions about sexual behavior (i.e., to incur HIV or another STI) in analogue contexts, primarily condomless sexual intercourse (Scott-Sheldon et al., 2016). However, evidence from longitudinal studies of event-level sexual risk behaviors in the natural environment in multiple participant populations has been far less clear and consistent. Weinhardt and Carey (2000) were among the first to identify such inconsistency in their systematic literature review, a finding which has been further supported in more recent published data (Freeman, 2016; Wray et al., 2019). These inconsistencies in the data suggest that the alcohol intoxication-condomless sex relation may be altered by other moderating variables. Another possibility that has been addressed considerably less frequently is that the relation between degree of alcohol intoxication and condomless sex at the event level over time may not be linear. In this regard, Simons, Simons, Maisto, Hahn, and Walters (2018) found in an experience sampling method (ESM) study that the relation between intoxication and condomless sex in young adults was characterized as an accelerating curve, with a steep increase in the likelihood of condomless sex as an individual’s BAC surpasses his or her average level of intoxication on a given day. Simons et al.’s (2018) data are consistent with the results of experimental studies, which have shown alcohol effects on sexual risk outcomes to be absent or weak until moderate to high BACs are reached. Given the accumulated evidence, efforts to improve primary and secondary HIV/STI prevention interventions may benefit from identifying contextual and person variables that exert the most powerful and consistent moderating effects on the alcohol intoxication-sexual behavior relation.

A situation factor that has long been identified as an important moderator of alcohol’s relation to condomless sex is sexual partner relationship (Cooper, 2010; Vanable et al., 2004). “Partner relationship” in the alcohol-sexual behaviors literature typically has been operationalized by the participant rating how well he/she knows a sexual partner. Data from within-persons studies of young adult heterosexual participants present a complex picture. Some studies have shown that alcohol is associated with a higher likelihood of condomless sex with less familiar but not with more familiar partners (Brown & Vanable, 2007; Kiene et al., 2009; LaBrie et al., 2005). Other studies have reported the opposite (Scott-Sheldon et al., 2010) or no relation between alcohol and condomless sex (Walsh et al., 2014). Finally, Simons et al. (2018) found that alcohol’s effects on sexual behaviors decreased with increasingly familiar sexual partners. The within-person relation among alcohol intoxication, partner relationship, and sexual behaviors has not been investigated in MSM.

In contrast, several cross-sectional, mixed methods retrospective studies of diverse samples of generally younger MSM have been published (Boyer et al., 2019; Martinez & Jonas, 2019; Storholm et al., 2017; Vasilenko et al., 2019). Overall, these studies are consistent in showing that condomless sex occurs considerably more frequently in self-reported “primary relationships” or with familiar partners than in casual or less familiar partners, due primarily to perceived less risk for contracting HIV or other STIs in the former partner types (Storholm et al., 2017). Unfortunately, these studies provide little data on whether partner relationship moderates the association between alcohol intoxication and frequency of sexual risk behaviors. However, it would seem that if partner relationship does have a moderation effect, it would be centered on increasing the likelihood of unprotected sex with less familiar partners, because of its causing less perceived risk in casual partners (cf. alcohol myopia theory, Steele & Josephs, 1990) or lowering the sexual inhibitions that may accompany sex with an unfamiliar partner (Parsons et al., 2004). Alcohol intoxication may also provide a context in which it is more likely that communication about sexual practices with an unfamiliar partner is less effective or does not happen at all (Cooper, 2010). In fact, in one of the few studies of alcohol’s relation to condomless sex in MSM and partner type, Vanable et al.’s (2004) cross-sectional survey showed that alcohol was not related to use of a condom with a primary partner, but that at a moderate dose or higher (4 drinks or more) alcohol was associated with condomless sex with a casual partner. The current study is the first to provide longitudinal, event-level data on alcohol intoxication, partner familiarity, and sexual behaviors in MSM, which allows the examination of partner relationship and alcohol intoxication in combination as within-person contextual factors associated with sexual behaviors.

Several individual differences variables also warrant consideration in longitudinal event-level analyses of sexual risk behaviors. Two such variables, effortful control and urgency, stand out as particularly important according to dual process models of self-regulatory behavior (Wiers et al., 2007), which have been influential in recent years in guiding research on sexual behavior (Ames et al., 2013; Grenard et al., 2013). The first of the dual processes is a deliberate, slower, more effortful subsystem, and the other is a quicker, more reactive and impulsive subsystem (Lieberman, 2007; Wiers et al., 2007; Wills et al., 2007). Research has shown that constructs loading on the reactive subsystem, such as urgency, are positively associated with both alcohol use and condomless sex (Hahn et al., 2016; Simons et al., 2009; Zapolski et al., 2009). Urgency is the tendency to act impulsively under extreme emotions (Anderson, Palfai, Maisto, & Simons, 2020 (Cyders & Smith, 2008). In contrast, the deliberative subsystem tends to be negatively related to substance use and condomless sex in adolescents and college students (Simons et al., 2010; Wills et al., 2013). Effortful control, which refers to the regulation of the self toward the achievement of future goals rather than immediate outcomes (Evans & Rothbart, 2007; Walters & Simons, 2020), is a construct that reflects the deliberative subsystem. If the reactive and deliberative self-regulatory subsystems act in this way, then it would be expected that both urgency and effortful control moderate the relation between alcohol intoxication and the occurrence of condomless sex. This prediction may be considered further in the context of partner familiarity. Effortful control and urgency would be expected to exert their respective attenuating and enhancing effects on the occurrence of condomless sex with less familiar partners. Studies of heterosexual college students have shown that deliberate control constructs such as executive function and working memory attenuate the positive association between alcohol intoxication and condomless sex (Abbey et al., 2006; Simons et al., 2018). On the other hand, there is little evidence that reactive system constructs like impulsivity or urgency moderate the alcohol intoxication-condomless sex association. At the between-person level, Semple and colleagues (2006) reported that urgency potentiated the association between methamphetamine use and condom use among HIV-positive MSM. In contrast, Simons, Maisto, and Wray (Simons et al., 2010) found significant main effects of urgency on failure to protect from STDs or pregnancy but did not find evidence of urgency moderating the effect of either alcohol or marijuana use on failure to protect from STDs or pregnancy in a college student sample. Simons et al. (2018) reported one of the few within-person analyses of experience sampling method (ESM) data on intoxication and condomless sex that examined the moderating effects of effortful control and reactivity (operationalized by combining items from scales measuring impulsivity and distractibility) in college students. Consistent with previous research, they found that effortful control did attenuate the alcohol intoxication and condomless sex relation but did not find moderating effects of reactivity. Simons et al. did not include partner familiarity in their study. Both effortful control and urgency were included in the current study in order to test if the findings of previous studies showing main effects of each variable on sexual behavior replicate in MSM and to provide initial data on whether the two variables moderate alcohol intoxication or partner familiarity effects on the occurrence of condomless sex in MSM.

An individual differences factor that may be of particular importance but that has not been studied intensively within persons in the sexual behavior of MSM is whether they are taking pre-exposure prophylaxis (PrEP) medication for HIV. PrEP consists of subsets of two or three antiretroviral drugs that have been developed to manage HIV that may be taken daily or, in one formulation, “on demand” before and after a given sexual event to prevent HIV seroconversion. Two such drug combinations have been approved by the Federal Drug Administration (FDA) for prescription in the United States. The first is a combination of the medications emtricitabine and tenofovir disoproxil fumarate, whose trade name is Truvada^™. The second is emtricitabine and tenovir alafenamide, with trade name Descovey^™. When taken as prescribed, PrEP is a highly effective non-condom method of protection against contracting HIV (Liu et al., 2016).

Although early PrEP trials did not show an increase in condomless anal sex in individuals on PrEP compared to those not on the medication, emerging evidence from participants not enrolled in clinical trials suggests that MSM may become less risk averse in their sexual behaviors because of the presumptive protection that PrEP offers (Hojilla et al., 2018; Luehring-Jones et al., 2019; Shuper et al., 2017). Such data suggest that taking PrEP may add to alcohol’s facilitative effects on engaging in sexual behaviors that typically have been identified as “risky” in the HIV literature. Although there are little data on alcohol’s moderation of the relation between taking PrEP and condom use (Freeborn et al., 2017), it also is possible that, among individuals on PrEP who do still use condoms at least some of the time, the likelihood of condom use may decrease when drinking (Fontenot et al., 2020), implying an interaction effect between alcohol intoxication and PrEP use on condom use. Further, it is important to understand how the relation among dual systems constructs such as urgency/impulsivity interact with alcohol intoxication in the context of being on PrEP.

Use of Experience Sampling Methodology

Taking a Person X Situation approach to study sexual behavior necessitates use of a research method that can capture both rapidly changing and more stable characteristics and their interaction over time. Ideal for this purpose among currently available methodologies is the use of ESM in a longitudinal design. Although MSM have been identified as a group that is at relatively high risk to contract HIV, there have been only several longitudinal studies of sexual behaviors that have used ESM with MSM. These studies have shown that use of ESM is feasible and acceptable to different subgroups of MSM (Duncan et al., 2017; Miner et al., 2019; Wray et al., 2019) and thus was used in the current study’s design.

Purpose of the Current Study

The purpose of this research was to conduct an ESM study to increase understanding of the connection between several individual difference and situation variables and the likelihood of occurrence of different sexual behaviors in adult MSM. Specifically, degree of alcohol intoxication and partner familiarity were examined as situation factors. It was hypothesized that both alcohol intoxication and partner familiarity would show main effects on the likelihood of occurrence of condomless anal intercourse (CAI). Based on the Simons et al. (2018) data, alcohol’s effect of increasing the likelihood of CAI was expected to be most pronounced at moderately high to high levels of intoxication, in the shape of an accelerating quadratic function. There is not consistent evidence to support a prediction of an interaction between intoxication and relationship familiarity, although studies of MSM cited earlier imply that alcohol is more likely to increase the likelihood of CAI with less familiar partners but have little effect with more familiar partners. Three between-persons or individual differences factors also were studied. Based on past research in heterosexual or sexuality not specified samples, it was hypothesized that effortful control would be inversely associated with CAI, whereas urgency was expected to be associated with higher rates of CAI. Furthermore, effortful control was expected to attenuate alcohol’s positive relation with the likelihood of occurrence of CAI. Research on moderating effects of urgency on relationships between alcohol and condom use has been limited. Given mixed findings, urgency was not expected to alter alcohol’s within-person association with CAI. The third between-persons factor studied, whether the individual reports that he is taking PrEP, was hypothesized to have a main, enhancing effect on likelihood of occurrence of CAI. It also was hypothesized that taking PrEP has synergistic effects with alcohol such that it enhances alcohol intoxication’s overall positive, curvilinear relation with likelihood of occurrence of CAI.

Method

Participants

Participants were 236 men aged 21 to 50 (M =27.84, SD = 6.89) who were recruited from the Syracuse, NY and Boston, MA areas using flyers, advertisements, and social networking sites (e.g. Facebook, Grindr, Tinder). Approximately 64.96% were White, 13.68% were Black, 7.26% were Asian, 3.85% were mixed race, and 9.40% were designated other. Approximately 18.80% were Hispanic/Latinx. Average yearly income was $32068.57 (SD = $25902.58; median = $28000.00). Of these, 199 were included in the analytic model. Twenty-two were excluded because they reported zero instances of oral or anal intercourse for the analyses. In addition, three individuals were excluded due to missing level 2 (L2) data. Twelve were excluded because they failed to respond to at least 33% of the random prompts during an analysis burst. These 15 individuals did not differ from the remaining 199 participants in age, urgency, effortful control, average partner familiarity, average level of intoxication, PrEP use, site, or average frequency of having any sex.

To be eligible for participation all individuals were required to meet the following inclusion criteria: 21-50 years old and a moderate or heavy drinker (according to the Quantity Frequency Variability; Cahalan et al., 1969) based on pattern of drinking in the last three months. The Cahalan et al. measure classifies individuals as heavy drinkers if they report a range of patterns of use, e.g., frequency of drinking typically 1-2 drinks of any alcohol beverage 3 or more times a day to having typically 5-6 drinks 2-3 times a month. Similarly, “moderate drinkers” could be individuals who report typically having 1-2 drinks twice a day to typically having 5-6 drinks once a month. Participants also had to be sexually active with men (at least once a month in the 3 months prior to study enrollment), self-reported score of at least a 3 (“equally heterosexual or homosexual experience(s)” on the Kinsey Scale ((three or higher on the Kinsey Scale; Kinsey et al., 1948), and not be in a committed monogamous relationship. To ensure participants’ suitability for the primary study that included an alcohol administration component, exclusion criteria included: self-reported current medical or psychiatric problems or their reports on the Brief Symptom Inventory (Derogatis & Melisaratos, 1983), use of a vitamin/herb or medication for which the use of alcohol is contraindicated, current alcohol or other substance use disorder, alcohol treatment within the past 3 years, substance use disorder or mental health treatment in the past 3 months, or any lifetime history of treatment for bipolar disorder or schizophrenia. Three previous papers have been reported from this ongoing study (Luehring-Jones et al., 2019; Simons, Maisto, & Palfai, 2019; Tahaney, Palfai, Luehring-Jones, Maisto, & Simons, 2019), one of which cited analyses of ESM data. The latter paper concerned the reliability and validity of ESM data on sexual behavior and related variables in this sample and has no conceptual overlap with the current paper.

Procedure

Individuals interested in the study called the laboratory and completed initial screening by telephone with a trained research assistant. If eligible, participants were invited to attend two laboratory sessions followed by the ESM study. Participants who completed both laboratory sessions and indicated that they would like to participate in the ESM study were trained in the use of Metricwire Inc. software (https://metricwire.com), the momentary assessment software that was installed on either their cellular phone or one provided by the study. There were two types of surveys: a morning survey (~ 5 minutes) and a random survey (~ 2 minutes). A practice morning and random survey were completed by participants in the laboratory. The surveys were programmed to begin immediately after enrollment in the ESM study.

Participants were asked to complete a morning assessment shortly after waking (morning survey). The morning survey was prompted at 8 a.m. and a reminder was sent at 10 a.m. The morning survey did not expire on a given day. In this study, the median time to complete the survey was approximately 9:02 a.m., the 75^th percentile was 10:18 a.m., and the 95^th percentile was 2:35 pm. About two-thirds of the morning surveys were completed at 10 a.m. or earlier. Random assessments were prompted at random times within 2-hour blocks from 10 a.m. and 2 a.m. If surveys were not completed within 15 minutes of receiving the initial notice, a reminder notification was initiated. After 30 minutes the random surveys expired. The validity of the sampling design used is supported by previous research (Armeli et al., 2003; Simons et al., 2010; Swendsen et al., 2000). Nine random surveys were prompted each day.

Participants completed questionnaires for a total of 42 days that were divided into 2 bursts of 3 weeks with a 3 week break in between to minimize fatigue. Research assistants provided updates to participants regarding their progress (e.g., percentage of surveys complete, amount of money earned) periodically throughout each burst. Participants were compensated $1.00 per completed random assessment, $3.00 per completed morning assessment, plus a $23.00 bonus for completing a week of morning assessments (i.e., up to $100 week or ~$14.29 per day). Compensation was provided in cash at the end of each burst.

Measures

Baseline

Effortful Control.

Effortful control was assessed by three scales: (1) a 7-item measure of planfulness (Kendall & Williams, 1982), (2) a 6-item measure of problem solving (Wills et al., 2001), and (3) a 10-item measure of goal setting (Neal & Carey, 2005). Higher scores indicate greater effortful control (α = 0.79). Confirmatory factor analysis indicates that these scales load onto a reliable effortful control factor that is moderately inversely correlated with indices of reactivity (Simons et al., 2016; Walters et al., 2018).

Urgency.

Urgency was assessed by the short version of the Urgency, Premeditation, Perseverance, Sensation seeking, and Positive urgency scale (SUPPS-P; Cyders et al., 2014). The scale includes four items assessing tendency to act rashly when positively aroused (positive urgency) and four items assessing tendency to act rashly when negatively aroused. For this study, a unidimensional urgency construct was used (Cyders & Smith, 2007; Peterson et al., 2018), defined by the mean of four positive and three negative urgency items (α = 0.83).¹

PrEP.

Participants indicated whether they were currently taking PrEP (0 = No, 1 = Yes).

Experience Sampling

Sexual Behavior.

Sexual behavior for each night was assessed during the self-initiated morning assessments. An initial item assessed any sexual activity (How many male partners did you engage in sexual activity with last night [e.g., kissing, fondling, intercourse]?). Subsequently, for each partner, follow-up questions assessed instances of (1) oral sex, (2) insertive anal intercourse, and (3) receptive anal intercourse. The oral sex item was dichotomous (Yes, No). The anal intercourse items had three response options (No; Yes, NOT with a condom the whole time; Yes, with a condom the whole time). A nominal sexual behavior variable was created for each night with five mutually exclusive categories (No sex, Minimal Sex Contact, Oral Sex, Anal Intercourse with a Condom, and Condomless Anal Intercourse [CAI]). The categories were formed to reflect the highest level of HIV risk engaged in with any partner each night. Hence, the Minimal Sex Contact category indicates the individual reported some form of sexual activity but not oral sex or anal intercourse. The Oral Sex category signifies oral sex but not anal intercourse occurred. The Anal Intercourse with a Condom category signifies that anal intercourse occurred and there were not any instances of CAI.

Partner Relationship.

Relationship with each sexual partner was assessed during the morning assessment. A single item asked participants to report how well they knew each partner. The item was rated on a 7-point scale (1 = not at all, 7 = very well). The partner relationship corresponding to the target sexual behavior was used in the analysis. Partner relationship is not observed (i.e., missing) on days when no sexual behavior occurred. Hence, to allow estimating the model with all categories, we replaced missing values with random variates from a distribution defined by each person’s relationship mean and SD.² Conceptually, this distribution reflects each participant’s potential sexual partners, which have been assessed over the course of the 6-week monitoring period. On any given day, as a function of desire, deliberation, and opportunity, a participant is choosing to engage in sexual activity or not with a partner from this observed distribution. On days when the participant does NOT have sex, the population of potential partners exists, and is defined by the observed population of sexual partners over time. On days when the participant DOES have sex, the partner is actualized from the population of potential partners. The act of engaging in sex shifts the probability distribution of potential partners to a single deterministic outcome. That is, once the behavior with a partner occurs its “probability” becomes 1.0, and this is the observed datapoint on a sexually active day. This approach allows the inclusion of the no sex category, which otherwise would be dropped by listwise deletion due to the missing partner relationship data. Partner relationship is person-mean centered in the analysis. Person-mean centering and the replacement of missing values with random variates from the person’s observed distribution of partners allows for a consistent interpretation of the effects as a function of within-person deviations from their own mean.

Intoxication.

Alcohol intoxication was assessed by random in situ assessments of number of standard drinks consumed during the past 30 minutes (0–7 or more), an estimate of BAC (Matthews & Miller, 1979) derived from next morning assessments of total number of drinks (0 –24) and hours spent drinking (0 –12) the previous night, and perceived intoxication the previous night assessed the next morning (1–7 scale: 1 = not at all; 7 = extremely). The mean of the three standardized variables (i.e., nightly drinks from random prompts, BAC estimate from morning assessments, and perceived intoxication from morning assessments) provided the estimate of and is henceforth in this paper called “intoxication.” The composite exhibited excellent reliability at both the within- (McDonald’s Omega = .85) and between- person levels (McDonald’s Omega = .89; Geldhof et al., 2014). There were two substantial outliers on the subject mean (i.e., more than 3.29 standard deviations from sample mean) and these were recoded to one unit greater than the nearest outlying value (Tabachnik & Fidell, 2001).

Analysis Plan

We estimated a mixed multinomial model with days (level 1 [L1]) nested within persons (L2) in Mplus 8.4 (Muthén & Muthén, 2019). The model included 5 nominal outcomes: No Sex, Minimal Sex Contact, Oral Sex, Anal Intercourse with a Condom, and Condomless Anal Intercourse (CAI). The outcomes are mutually exclusive and reflect the highest degree of sexual risk behavior for the night. Hence, Minimal Sex Contact indicates no anal or oral sex occurred, Oral Sex implies no anal intercourse occurred, and Anal Intercourse with a Condom implies no CAI occurred. At L1, intoxication, intoxication², partner relationship, intoxication X partner relationship, elapsed days in the study (a time² term was also tested but was not significant and hence dropped), and six orthogonal day of the week indicators were predictors. At L2, age, site, PrEP, urgency, and effortful control were predictors. Finally, the model tested cross-level interactions between the L2 predictors PrEP, Urgency, and effortful control on the L1 intoxication, intoxication^2, and relationship effects. We trimmed cross-level effects on the intoxication² (i.e., 3-way interactions) that had p-values > .10 to create a more parsimonious model. Interactions in non-linear models are difficult to interpret because the significance and form of the interaction also depends on levels of covariates (Buis, 2010; Norton et al., 2004). Hence, we graph all interactions and report relative risk ratios, which have a consistent multiplicative interpretation irrespective of the conditional mean (Buis, 2010). The continuous L1 variables were centered at the person-mean and the L2 variables were centered at the grand-mean to differentiate within- and between- person processes. The categorical intercepts were random effects and allowed to freely covary. We tested for random slopes as well. However, initial estimation of the model showed that variances in the L1 slopes across persons were 0 or close to 0. As a result, not all could be estimated. Due to this evidence of a lack of substantial variation, all L1 slopes were fixed effects.

Results

Protocol Compliance

Participants in the analysis sample completed 92% (SD = 0.08, median = .96) of the self-initiated morning assessments. The response rate for the random assessments was 67% (SD = 0.14, median = .70; response rate before midnight = 71% (SD = .15, median = .73). Participants provided an average of 41.3 analysis days (range 8 – 61).³

Descriptive Statistics

Table 1 includes descriptive statistics. At the between-persons level, effortful control and urgency were moderately inversely correlated. Effortful control was modestly positively associated with tending to know sexual partners better and being more likely to use PrEP. Conversely, urgency was not significantly associated with partner relationship and exhibited a modest inverse association with PrEP use (see Table 2). Participants in the analysis sample had some form of sexual activity on 19.73% of days. Anal sex occurred on 10.46% of days (n = 861). Of these, condoms were used 33.57% of the time (n = 289). CAI occurred on 6.95% of days (n = 572) and oral sex (no anal sex) occurred on 4.74% of days (n = 390). As expected, men taking PrEP were less likely to use condoms during anal intercourse (88% of intercourse occasions were condomless vs 59% [CAI for men not taking prep], χ2(1) =62.31, p < .001). Participants drank on 55.87% of days and consumed an average of 4.99 drinks (SD = 4.38, range = 1.00-24.00) per drinking day.

Table 1.

Summary Statistics

Variable	N	M/%	SD	Min	Max	Skew	Kurtosis
Site ([location omitted for blind review])	236	23%
Age (years)	236	27.84	6.89	21.00	50.00	1.26	4.12
Effortful control	233	3.96	0.49	2.14	4.93	−0.61	3.25
Urgency	233	1.84	0.59	1.00	3.57	0.44	2.50
PrEP	236	22%
SM Partner rel.	222	3.88	1.76	1.00	7.00	0.25	2.00
SM Intoxication	236	−0.01	0.41	−0.57	1.88	1.74	7.16
SP Any sex	236	0.18	0.15	0.00	0.93	1.44	5.84
SP Minimal sex	236	0.04	0.07	0.00	0.59	3.94	23.19
SP Oral sex	236	0.04	0.06	0.00	0.34	2.02	8.24
SP Anal with condom	236	0.04	0.06	0.00	0.51	4.01	24.98
SP CAI	236	0.06	0.09	0.00	0.70	2.63	13.54
Intoxication	10271	0.00	0.89	−0.63	5.93	2.19	8.92
Partner rel.	1764	4.43	2.30	1.00	7.00	−0.22	1.53

Note. SM = subject mean across days from the ESM data, SP = subject percent of days from the ESM data, rel. = relationship, CAI = condomless anal intercourse

Table 2.

Correlation matrix

Variable	1.	2.	3.	4.	5.	6.	7.	8.	9.	10.	11.
1. Site	1
2. Age	0.17*	1
3. Effortful control	−0.07	0.08	1
4. Urgency	0.03	−0.06	−0.43***	1
5. PrEP	−0.12	0.03	0.15*	−0.14*	1
6. SM Partner rel.	0.01	0.08	0.14*	0.06	−0.05	1
7. SM Intoxication	0.14*	0.06	−0.10	0.08	−0.10	0.12	1
8. SP Any sex	−0.09	0.03	0.08	0.04	0.12	0.38***	0.25***	1
9. SP Minimal sex	−0.15*	0.02	−0.00	−0.02	0.03	0.28***	0.08	0.57***	1
10. SP Oral sex	−0.21**	0.02	0.03	0.03	0.07	0.07	−0.03	0.42***	0.18**	1
11. SP Anal with condom	0.14*	0.02	0.10	−0.01	−0.13*	0.09	0.20**	0.38***	−0.09	−0.11	1
12. SP CAI	−0.01	0.01	0.04	0.08	0.23***	0.28***	0.23***	0.69***	0.09	0.02	0.06

Note. N = 219 to 236. Site is coded 0 = [location omitted for blind review], 1 = [location omitted for blind review]. PrEP is coded 0 = No, 1 = Yes. SM = subject mean across days from the ESM data, SP = subject percent of days from the ESM data, rel. = relationship, CAI = Condomless anal intercourse

^*p<0.05

^**p<0.01

^***p<0.001

Mixed Effect Multinomial Regression Model

A mixed multinomial regression analysis of sexual behavior (N = 199, days = 8227) tested within-person, between-person, and cross-level associations (see planned analysis section for additional detail). The urgency X intoxication² probability value was > .10 across equations and hence was trimmed from the model for parsimony. The results with No Sex as the baseline category and CAI as the baseline category are presented in Tables 3 and 4, respectively. In Table 3, the effects indicate how each outcome changes relative to the No Sex outcome. For example, the relative risk ratio (RRR) of 2.04 for the effect of intoxication on CAI indicates that the relative risk of CAI versus No Sex increases by a factor of 2.04 for a 1-unit increase in intoxication (given the other variables are held constant at a mean of zero). In Table 4, the effects indicate how each outcome changes relative to the CAI outcome. For example, the RRR of 0.72 for the effect of partner relationship on Anal Intercourse with a Condom indicates that the relative risk of Anal Intercourse with a Condom versus CAI changes by a factor of .72 (i.e., becomes smaller) for every unit increase in partner relationship (given the other variables are held constant at a mean of zero). Figures 1 to 6 display the marginal effects. The RRR just described is reflected in the ratios of the probabilities. In Figure 2, the probability of Anal Intercourse with a Condom at the relationship mean is ~.04 and the probability of CAI is ~.08 (relative risk = .5). As partner relationship increases by 1 unit the probability of Anal Intercourse with a Condom decreases to about .03 and the probability of CAI increases to ~.085 (relative risk = ~.35). Hence the relative risk changed by a factor of .70 (.35 / .50), which corresponds to the RRR of .72 (difference due to rounding error and random effects).

Table 3.

Multilevel Multinomial Regression – No Sex Reference Category

	Outcome
	Minimal Sex Contact (n = 371)				Oral Sex (n = 390)				Anal Intercourse with Condom (n = 289)				Condomless Anal Intercourse (n = 572)
Variable	B	SE	p	RRR	B	SE	p	RRR	B	SE	p	RRR	B	SE	p	RRR
Within-person
Intoxication	0.71	0.15	<.001	2.03	0.58	0.12	<.001	1.79	0.94	0.16	<.001	2.55	0.71	0.13	<.001	2.04
Intoxication 2	−0.13	0.07	.044	0.88	−0.32	0.12	.007	0.73	−0.21	0.06	.001	0.81	−0.22	0.06	<.001	0.80
Partner rel	0.14	0.07	.034	1.15	−0.06	0.04	.143	0.94	−0.20	0.07	.004	0.82	0.14	0.05	.008	1.15
Partner rel x Intox	−0.07	0.05	.204	0.94	0.16	0.06	.009	1.17	−0.07	0.05	.124	0.93	−0.05	0.05	.363	0.96
Cross-level (Person x Situation)
EC x Intox	−0.05	0.05	.301	0.95	0.04	0.05	.432	1.04	0.09	0.07	.183	1.09	0.04	0.04	.361	1.04
Urgency x Intox	0.07	0.03	.018	1.07	−0.02	0.03	.481	0.98	−0.10	0.04	.021	0.91	0.01	0.03	.700	1.01
PrEP x Intox	0.71	0.34	.038	2.04	−0.09	0.31	.769	0.91	−0.17	0.41	.689	0.85	0.37	0.24	.128	1.44
EC X Part rel	−0.06	0.03	.073	0.94	0.03	0.02	.133	1.03	0.02	0.03	.443	1.02	0.03	0.02	.304	1.03
Urgency x Part rel	−0.05	0.02	.018	0.96	0.02	0.02	.176	1.02	−0.02	0.02	.247	0.98	0.03	0.01	.034	1.03
PrEP x Part rel	−0.29	0.14	.035	0.75	0.22	0.08	.008	1.25	−0.15	0.18	.425	0.86	−0.14	0.08	.096	0.87
EC X Intox2	0.07	0.02	<.001	1.07	0.00	0.03	.909	1.00	−0.05	0.03	.082	0.95	−0.02	0.02	.394	0.98
PrEP x Intox2	−0.44	0.15	.004	0.64	−0.66	0.43	.126	0.52	0.14	0.17	.418	1.15	−0.31	0.14	.022	0.73
Between-person
Site	−0.67	0.34	.048		−0.71	0.25	.005		0.77	0.30	.011		0.30	0.32	.344
Age	−0.01	0.02	.697		0.01	0.01	.400		−0.01	0.02	.466		0.00	0.02	.984
EC	−0.01	0.06	.866		0.04	0.04	.369		0.13	0.05	.010		0.08	0.06	.191
Urgency	0.02	0.05	.607		0.04	0.03	.305		0.09	0.04	.042		0.11	0.05	.030
PrEP	0.28	0.31	.360		0.30	0.26	.241		−0.97	0.30	.001		1.21	0.27	<.001

Note. N (days) = 8227, N (persons) = 199. Akaike Information Criteria (AIC) = 10987.05, Bayesian Information Criteria (BIC) = 11758.72. Site is coded 0 = [location omitted for blind review], 1 = [location omitted for blind review]. Day of the week and elapsed day are included as covariates but not depicted due to space limitations. Intox = Intoxication, EC = effortful control, Part rel = Partner relationship, RRR= relative risk ratio. Urgency and EC are scaled by a factor of 5. Intoxication and relationship are centered at the person mean. All between-person variables are centered at grand mean. Categories are mutually exclusive. Minimal sex contact category refers to nights that some sexual behavior occurred but no oral sex or intercourse occurred. The oral sex category refers to nights that oral sex but no anal intercourse occurred.

Table 4.

Multilevel Multinomial Regression – Condomless Anal Intercourse Reference Category

	Outcome
	No Sex (n = 6605)				Minimal Sex Contact (n = 371)				Oral Sex (n = 390)				Anal Intercourse with Condom (n = 289)
Variable	B	SE	p	RRR	B	SE	p	RRR	B	SE	p	RRR	B	SE	p	RRR
Within-person
Intoxication	−0.71	0.13	<.001	0.49	0.00	0.17	.992	1.00	−0.13	0.15	.401	0.88	0.23	0.17	.193	1.25
Intoxication 2	0.22	0.06	<.001	1.25	0.09	0.09	.290	1.10	−0.10	0.13	.463	0.91	0.01	0.08	.886	1.01
Partner rel	−0.14	0.05	.008	0.87	0.00	0.10	.964	1.00	−0.20	0.07	.006	0.82	−0.33	0.09	<.001	0.72
Partner rel x Intox	0.05	0.05	.363	1.05	−0.02	0.07	.773	0.98	0.21	0.08	.009	1.23	−0.03	0.06	.701	0.98
Cross-level (Person x Situation)
EC x Intox	−0.04	0.04	.362	0.97	−0.08	0.05	.131	0.92	0.00	0.06	.956	1.00	0.06	0.07	.436	1.06
Urgency x Intox	−0.01	0.03	.701	0.99	0.06	0.04	.096	1.06	−0.03	0.03	.318	0.97	−0.11	0.05	.022	0.90
PrEP x Intox	−0.37	0.24	.127	0.69	0.35	0.42	.408	1.41	−0.46	0.35	.189	0.63	−0.53	0.43	.216	0.59
EC X Part rel	−0.03	0.02	.305	0.98	−0.08	0.05	.110	0.92	0.00	0.04	.952	1.00	−0.01	0.04	.903	1.00
Urgency x Part rel	−0.03	0.01	.034	0.97	−0.07	0.03	.008	0.93	−0.01	0.02	.731	0.99	−0.05	0.03	.051	0.95
PrEP x Part rel	0.14	0.08	.096	1.14	−0.16	0.19	.397	0.85	0.36	0.12	.004	1.43	−0.01	0.21	.961	0.99
EC X Intox2	0.02	0.02	.395	1.02	0.09	0.03	.001	1.09	0.01	0.03	.676	1.01	−0.03	0.03	.307	0.97
PrEP x Intox2	0.31	0.14	.022	1.36	−0.13	0.22	.550	0.88	−0.35	0.44	.428	0.71	0.45	0.20	.023	1.56
Between-person
Site	−0.31	0.32	.343		−0.98	0.47	.037		−1.02	0.40	.011		0.46	0.39	.242
Age	0.00	0.02	.984		−0.01	0.02	.755		0.01	0.02	.602		−0.01	0.02	.570
EC	−0.08	0.06	.191		−0.09	0.08	.240		−0.04	0.07	.579		0.05	0.07	.477
Urgency	−0.11	0.05	.030		−0.08	0.06	.166		−0.07	0.05	.185		−0.02	0.06	.782
PrEP	−1.22	0.27	<.001		−0.93	0.36	.009		−0.92	0.35	.008		−2.18	0.42	<.001

Note. N (days) = 8227, N (persons) = 199. Akaike Information Criteria (AIC) = 10987.04, Bayesian Information Criteria (BIC) = 11758.71. Site is coded 0 = [location omitted for blind review], 1 = [location omitted for blind review]. Day of the week and elapsed day are included as covariates but not depicted due to space limitations. Intox = Intoxication, EC = effortful control, Part rel = Partner relationship, RRR= relative risk ratio. Urgency and EC are scaled by a factor of 5. Intoxication and relationship are centered at the person mean. All between-person variables are centered at grand mean. Categories are mutually exclusive. Minimal sex contact category refers to nights that some sexual behavior occurred but no oral sex or intercourse occurred. The oral sex category refers to nights that oral sex but no anal intercourse occurred.

Figure 1.

Marginal within-person effects of intoxication on sexual behavior

Figure 6.

Marginal effects of the cross-level interactions between PrEP and intoxication (upper panel) and partner relationship (lower panel) on sexual behavior

Figure 2.

Marginal within-person effects of partner relationship on sexual behavior

Within-person Associations

Overall, intoxication exhibited a curvilinear inverted-U association with Oral Sex, Anal Intercourse with a Condom, and CAI. In this regard, alcohol increased the likelihood of each respective sexual behavior as level of intoxication increased to about 1.7 standard deviations above the average level of intoxication for an individual. Beyond that point, however, the likelihood of the indicated sexual behaviors decreased. The association with the Minimal Sex category differed and was consistent with the Simons et al. (2018) data on condomless sex and with our hypothesis, with the probability of minimal sex contact increasing with intoxication level (see Figure 1). Further, consistent with our hypothesis, increased partner familiarity was associated with increased probability of CAI relative to all outcomes aside from Minimal Sex Contact (see Table 4 and Figure 2). However, there was little evidence of interactions between partner relationship and intoxication (see Table 3 and Figure 3).

Figure 3.

Marginal effects of the within-person interaction between partner relationship and intoxication on sexual behavior

Between-persons Associations

Effortful control was associated with an increased proportion of Anal Intercourse with a Condom relative to No Sex. In contrast, urgency was associated with an increased proportion of CAI as well as Anal Intercourse with a Condom relative to No Sex (see Table 3). PrEP use was associated with an increased proportion of CAI relative to all other categories (see Table 4). Sexual behavior did not vary as a function of age.

Cross-level Interactions

The within-person intoxication associations were qualified by several cross-level interactions (see Figures 4-6). For example, the positive association between intoxication and the probability of engaging in Minimal Sex Contact relative to No Sex increased as a function of effortful control (see Table 3 and Figure 4 – upper panel). Effortful control was associated with an increased likelihood of engaging in Minimal Sex Contact relative to CAI at the highest levels of intoxication (see Table 4 and Figure 4 – upper panel). In contrast, for men lower in urgency intoxication was associated with an increased likelihood of using a condom when having anal intercourse (see Table 4 and Figure 5 – upper panel). Men who took PrEP exhibited a declining probability of Minimal Sex Contact at higher levels of intoxication (see Figure 6 - upper panel). In addition, the association between intoxication and the likelihood of using a condom during anal intercourse varied as a function of PrEP use (see Table 4 and Figure 6 – upper panel). Men on PrEP exhibited a more pronounced curvilinear (inverted-U) association between intoxication and CAI.

Figure 4.

Marginal effects of the cross-level interactions between effortful control and intoxication (upper panel) and partner relationship (lower panel) on sexual behavior

Figure 5.

Marginal effects of the cross-level interactions between urgency and intoxication (upper panel) and partner relationship (lower panel) on sexual behavior

The within-person associations between sexual behavior and partner relationship also differed as a function of person characteristics. For example, men characterized by higher urgency exhibited a stronger positive association between partner relationship and the probability of CAI (relative to all categories aside from Oral Sex Only; see Table 4 and Figure 5 – lower panel). In contrast, use of PrEP was generally associated with a diminished association between partner relationship and sexual behavior (see Table 3 and Figure 6 – lower panel).

Discussion

This study’s main hypotheses were that alcohol intoxication has a curvilinear relation to the probability of occurrence of condomless anal sex in MSM, such that alcohol’s effects are weak when relative level of intoxication for an individual is low but increase in strength in an accelerated fashion as intoxication surpasses an individual’s average level. At the same time, we argued that alcohol’s relation to CAI is moderated in predictable ways by both situation and person variables. The results of this study overall were consistent with these hypotheses.

Within-person effects

As described earlier, the degree of intoxication showed a curvilinear relation to the probability of occurrence of each of the four sexual contact categories, but the relation between degree of intoxication and probability of occurrence of a sexual behavior was similar in shape to our prediction and that of Simons et al. (2018) only for minimal sexual contact. For the other three categories of sexual contact that were measured, the probability of their occurrence exhibited an initial increase before declining at the highest levels of relative intoxication. For example, the probability of CAI increased as intoxication increased to about 1.7 standard deviations above average for the person, and then declined as intoxication continued to increase. This inverted-U function may have been due primarily to degree of physical impairment as intoxication levels become extremely high, which presumably would have been less of an impediment for “minimal sexual contact.” In contrast, the Simons et al. study showed a slowly increasing likelihood of condomless intercourse as degree of intoxication approached and passed average level for the individual and then a rapidly accelerating increasing likelihood when intoxication reached and passed about 1.4 standard deviations above average. There are several differences between the study conducted by Simons et al. and this study, the most apparent being that the target population in this study was MSM whereas participants in the Simons et al. (2018) study included both men and women and were not selected for self-identified sexual orientation/behavior. Another potentially important difference between the two studies’ samples was age. The Simons et al. (2018) sample selected for young adults, and its mean age was younger than 20 years, which is younger than the U.S. legal drinking age. In this study, participants were eligible if they were 21-50 years old, and the sample had a mean age of about 27 years. Both variables, sexual orientation and age, may have influenced the extent to which varying amounts of alcohol consumption contribute to the initiation of sexual activity and the contexts in which it occurs.

Another situation variable, relationship familiarity, showed significant main effects. As expected, relationship familiarity increased the likelihood of occurrence of CAI and minimal sexual contact relative to no sex and decreased the likelihood of anal intercourse with a condom relative to both CAI and no sex. In the current study, partner familiarity was largely conceptualized as a situation factor that should influence perceived risk for the participant. We expected intoxication and partner relationship would interact, such that intoxication would exert greater effects with unfamiliar than familiar partners due to differences between the two in perceived risk (Parsons et al., 2004). One possible explanation of our unexpected finding is that the less familiar partners in this study may not have been perceived as “high risk” partners and thus not creating significant conflict in the individual about having sex with them and that alcohol might then attenuate (MacDonald et al., 2003; Steele & Josephs, 1990). Future within-person research on alcohol and sexual behaviors would benefit by including measurement of perceived partner risk along with a measure of partner familiarity (Newcomb et al., 2016).

Between-person and cross-level effects

As hypothesized, all three individual differences variables effortful control, urgency, and taking PrEP, showed significant main effects. Also as expected, alcohol main effects were qualified by interactions with individual differences variables. Effortful control modified the relation between alcohol intoxication and sexual behavior such that it tended to increase the likelihood of minimal compared to no sex with increasing levels of intoxication. Therefore, as hypothesized, a greater degree of effortful control was associated with a more likely occurrence of sexual behavior that is less likely to result in contracting HIV or another STI with increasing intoxication. However, effortful control did not buffer the association between intoxication and the likelihood of using a condom during intercourse as was observed in previous research with heterosexual individuals (Simons et al., 2018). Rather, urgency was associated with a decreased likelihood of using a condom during anal intercourse as intoxication increased. As reviewed earlier, this finding is consistent with dual process theory of self-regulation and is the first evidence from a within-person research design of an interaction effect on the relation between alcohol intoxication and condomless sexual behavior. Similarly, in a two-year longitudinal survey study of young MSM (Puckett, Newcomb, Garofalo, & Mustanski, 2017), a positive relation between internalized homophobia and the occurrence of condomless sex was evident only in participants who were higher in negative urgency (the tendency to act impulsively when experiencing negative emotion). Finally, the peak likelihood of CAI reached as intoxication rose was higher in individuals who were taking PrEP than the likelihood reached in those who were not on PrEP. As suggested earlier, this result may have been due to decreased condom use when drinking among PrEP users who nevertheless still used condoms at least some of the time (Fontenot et al., 2020).

Although we made no hypotheses regarding moderation effects among the individual differences variables and relationship familiarity, several were found that warrant mention. For individuals lower in urgency, partner familiarity seemed to have little effect on any of the sexual behaviors. However, for individuals higher in urgency, greater partner familiarity was associated with a higher likelihood of CAI. From the perspective of dual process self-regulation theory, the common belief that known and familiar partners pose less risk than those who are less familiar may have provided the context in which it seemed that little deliberation about the risk of CAI was required. In such a context, individuals higher in urgency likely were less inclined to consider factors other than familiarity of their partners in making decisions to engage in CAI. In contrast, with the exception of oral sex, associations between partner familiarity and sexual behaviors tended to be attenuated among individuals who were taking PrEP. Such a result follows from PrEP’s lifting the burden of risk of contracting HIV that CAI poses in a way similar to what greater partner familiarity seems to provide for many MSM.

The results of this study confirm hypothesized relations between degree of alcohol intoxication and moderators of those effects on sexual risk in MSM that have been observed in other populations. It is important to note that, for important reasons, the literature on alcohol and sexual risk has tended to focus on sexual behaviors thought to put individuals at risk for contracting HIV or other STIs rather than the broad range of sexual activity. This study showed that alcohol intoxication and moderating factors have important effects on a range of sexual behaviors and thus provides a fuller context for understanding intoxication as a determinant of sexual behavior that may incur risk of viral infection. All forms of sexual activity, including CAI, increased as a function of intoxication. However, the average effects of intoxication on the likelihood of CAI relative to other, deemed to be less risky, sexual behaviors were not significant. This shows the importance of evaluating alcohol effects on a range of sexual behaviors in order to differentiate effects of alcohol on social-sexual behavior broadly, versus effects on decisions about behaviors that may be associated with risk of contracting HIV or other STIs more specifically. The results suggest that drinking is a correlate of being with a potential sexual partner. The subsequent determinants of the sexual behavior that may occur are complex, resulting from the interplay of intoxication level, familiarity with the partner, other situational variables, and individual differences variables. Additional research along those lines would lead to more precise theoretical models and hypotheses about alcohol’s relation to sexual behaviors, including those that may incur greater risk of viral infection.

This research had the feature of studying the relation between self-report of taking PrEP and sexual behaviors in individuals over time. Such systematic study is important as well as innovative, because PrEP is a main component of biomedical and combined behavioral and biomedical approaches to primary prevention of HIV. In this study, PrEP’s main, facilitative effects on sexual activity, in particular CAI, stood out as most important. One possible reason for this result is that individuals who take PrEP simply do not regard condomless sex as incurring risk, in fact because they take PrEP (Storholm et al., 2017). Conversely, it may be that individuals who already are engaging in relatively high rates of condomless sex begin a regimen of PrEP in order to continue such sexual activity but with reduced risk of contracting HIV (Gafos et al., 2019). Similarly, Kessler et al.’s (2016) data showed that the best predictor of MSM’s willingness to take PrEP was actual risk of incurring HIV given their frequency of engaging in CAI. Consistent with these results, Fontenot et al.’s (2020) qualitative data showed that attitudes toward condom use in younger (ages 17-24) MSM were more positive in individuals who were not taking PrEP compared to those who were. Although this study’s data cannot address the relations between beliefs about and motives for taking PrEP and probability of occurrence of condomless sex, it is a critical topic for future research. It also is important to build on this study by addressing how situational factors, such as partner characteristics besides familiarity, are associated with sexual behavior when taking PrEP. For example, Newcomb et al. (2016) showed that MSM’s decisions about whether to engage in condomless anal sex were influenced by their partner’s disclosure of whether he was taking PrEP and whether he had an undetectable viral load. Finally, Fontenot et al.’s data suggest that there are individual differences between PrEP users about whether they believe that condom use still is a useful prevention measure. Overall, future research should be designed with the consideration that multiple factors likely determine decisions to take PrEP and sexual behavior while on it (DiCiaccio et al., 2020).

Clinical Implications

The results of this study have implications for the development or modification of HIV and other STI primary prevention interventions. First, the data reaffirm that, perhaps contingent on the population and context, moderately higher levels of intoxication than are typical for an individual are associated with decisions to engage in condomless anal sex among MSM. This finding is consistent with Maisto, Palfai, Vanable, Heath and Woolf-King’s (2012) findings in an alcohol administration, laboratory analogue experimental study of making decisions about engaging in condomless sex in MSM aged 21-50. Most critically, alcohol’s effects on sex and decision-making about sex remain under-addressed in specific ways, especially its influences in the context of other individual differences (e.g., effortful control and urgency) and situation (sexual and emotional arousal) variables, in otherwise evidence-based HIV primary prevention programs that the U.S. Centers for Disease Control and Prevention supports (Maisto & Simons, 2016). Therefore, HIV and other STI prevention programs that target individuals who consume alcohol or other drugs could be improved by including more specific, empirically-based information about how alcohol use together with contextual and individual differences factors specifically may affect sexual decision-making and behavior. The findings of a direct association between self-report of taking PrEP and likelihood of CAI also is important for HIV/STI prevention programs. Although PrEP provides outstanding protection from contracting HIV, medically it offers no protection against contracting other STIs (e.g., Storholm et al., 2017), although to date there is not strong evidence that individuals who begin taking PrEP and increase their frequency of CAI also increase their rate of contracting STIs besides HIV (Freeborn et al., 2017). Nevertheless, it seems important that STI prevention programs emphasize this fact, given that individuals may believe that the protection that PrEP offers against HIV generalizes to other STIs.

The finding of a direct relation between partner familiarity and likelihood of engaging in CAI is similarly of importance for primary prevention programs. In this regard, although MSM (and others) generally believe that having a “committed” sexual partner is a guarantor that having condomless sex with that person is safe, over two-thirds of new cases of HIV in MSM in the U.S. result from having sex with a committed partner (Hahn et al., 2019). As Hahn et al. (2019) note, such findings may be the result of factors such as serial monogamy and inaccurate assumptions about the sexual activity of the committed partner. Overall, the data show that HIV and other STI prevention programs should fully discuss what knowledge about a partner’s sexual activity is essential to have, regardless of how “familiar” the partner is, in order to sustain a safe sexual relationship.

Strengths and Limitations

There are some strengths of this study that are important to highlight. It is one of the few that report the findings of an extended period of ESM data on alcohol intoxication and multiple sexual behaviors in MSM and the only study to our knowledge that allowed comparisons between participants who report that they are taking PrEP and those who do not. There are also several weaknesses of this study that should be considered in the interpretation of its results. A total of 15 participants were excluded from the analyses due to missing data. Concerns about possible limits to the generalizability of findings are alleviated somewhat by the finding of no differences between these individuals and those included in the analyses in relevant variables at baseline. In addition, the large majority of analyses were of morning assessment data, which had a good compliance rate of 92%. However, a component of the intoxication variable was based on random prompt data, which showed compliance of 67%. This rate was not as high as in our previous studies Luehring-Jones et al., 2019; Simons et al., 2019; Tahaney et al., 2019) or in past studies reported by other labs that involved the collection of ESM data on alcohol use and sexual behaviors in MSM (e.g., Wray et al., 2016). One notable difference between this study and Wray et al. (2016) or other ESM studies is that this study’s design included a relatively large number (9) of daily random prompts that covered a wide range of hours of the day, with the last occurring by 2 a.m. In addition, the interpretation of PrEP effects and their implications would have been strengthened by data on how adherent the participant was to his prescription regimen. Finally, although this study’s sample was racially and ethnically diverse, it still under-represented African American and Hispanic MSM, who had the highest rates of new HIV infections in the U.S. 2018, followed by White MSM (https://www.hiv.gov/hiv-basics/overview/data-and-trends/statistics).

In conclusion, the results of this study highlight the importance of both individual differences and situation variables as correlates of sexual behaviors in MSM, and include new findings on the importance of taking PrEP usage into consideration in understanding MSM sexual behaviors. Future research should build upon this study by including variables in the protocol that may be correlates of taking PrEP and that may help to explain its association with different sexual behaviors. In addition, it seems important to further build on this paper’s report of assessment of situation variables besides intoxication level by including affect or sexual arousal. These two variables have been shown to be determinants of sexual risk and sexual behaviors in general (George, 2019; Maisto et al., 2012; Maisto & Simons, 2016) and specifically in other ESM and daily diary studies of alcohol and sexual behaviors in MSM (Grov et al., 2010; Wray et al., 2019).