Daily relations among affect, urge, targeted naltrexone, and alcohol use in young adults

Abstract

Heavy drinking among young adults is a serious public health problem. Naltrexone, an opioid antagonist, has been shown to reduce drinking in young adults compared to placebo and can be taken on a targeted (i.e., as needed) basis. Understanding risk factors for drinking and naltrexone effects within-person in young adults may help to optimize the use of targeted naltrexone. The current study was a secondary analysis of daily diary data from 127 (n=40 female) young adults (age 18-25) enrolled in a double-blind clinical trial of daily (25 mg) plus targeted (25 mg) naltrexone versus placebo. Hierarchical linear models were used to examine the effects of daily affect, urge, and taking targeted medication on same-day risk of drinking to intoxication (defined as estimated blood-alcohol-concentration, BAC≥.08g%). Results indicated urge significantly mediated within-person positive affect–drinking relations on a daily level. Specifically, positive affect was associated with greater urge to drink, which in turn was associated with greater odds of BAC≥.08g%. Furthermore, days of greater positive affect and urge were associated with taking a targeted dose of medication, which reduced the likelihood of intoxication by nearly 23% in the naltrexone group compared to placebo. Gender and family history of alcohol dependence were examined as moderators of these daily level effects. These results provide further evidence of naltrexone’s ability to reduce alcohol consumption in young adults and identify potential within-person risk processes related to heavy drinking that could inform alcohol-related interventions for this population.

Introduction

Heavy drinking among young adults is a serious public health problem. Across national surveys of young adults, an estimated 33-44% endorse past 30-day heavy drinking (i.e., ≥4 drinks for women or ≥5 for men) (SAMHSA, 2014; White & Hingson, 2014). Alcohol use in this population has been associated with serious negative consequences including health risk behaviors (e.g., unplanned/unprotected sex), lower academic performance, injury and death (Hingson, Zha, Weitzman, 2009; SAMHSA, 2011; White & Hingson, 2014). Furthermore, prospective research suggests that young adults who are heavy drinkers are more likely to develop long-term alcohol-related problems (Jennison, 2004). These findings underscore the importance of identifying effective interventions to reduce heavy drinking in young adults.

Recent evidence suggests that naltrexone, an opioid antagonist medication approved by the U.S. Food and Drug Administration (FDA) for the treatment of alcohol dependence, is effective for reducing drinks per drinking day (Leeman et al., 2008; Miranda et al., 2014) and the percentage of drinking days with an estimated blood alcohol concentration (eBAC) meeting the legal limit of intoxication (≥.08g%) (O’Malley et al., 2015) in young adult samples. In particular, reducing BAC to a “lower risk” level (i.e., below the legal limit) on drinking days may be a clinically important outcome for heavy-drinking young adults, given the risk of negative consequences and impairment from intoxication (Curtin & Fairchild, 2003; Jewett, Shults, Banerjee, Bergen, 2015; Stubig et al., 2012; WHO, 2007). Additionally, estimated BAC is likely to be a more sensitive measure of heavy drinking than the standard numerical limit often used in research (≥5 standard drinks for men, ≥4 for women), which does not consider body weight or duration of consumption.

While naltrexone has shown promise for reducing heavy drinking in young adults, additional research is needed to understand the mechanisms through which daily and targeted naltrexone exert their effects. Naltrexone may be an especially attractive intervention for young adults because it can be taken on a “targeted” basis prior to drinking. This dosing strategy may be more suitable or preferable to young adults who prefer to take medication “as needed” and may be more willing to reduce their drinking than to completely abstain (Dimeff, 1999; Epler, Sher, Loomis, & O’Malley, 2009).

Although not specific to young adults, secondary analysis of a randomized clinical trial of targeted versus daily naltrexone or placebo in heavy-drinking adults (ages 18-60) indicated that naltrexone may work, at least in part, by reducing the odds of heavy drinking on days characterized by high positive and high negative affect (Kranzler et al., 2004). In addition, higher levels of desire to drink were reported on days with higher positive or negative mood, and heavy drinking was more likely on days with higher desire, suggesting a mediating role of urge in mood–drinking relations. While these earlier findings provide important information about risk processes related to heavy drinking, the temporal sequencing of mood, urge, and drinking could not be determined because they were assessed concurrently. Furthermore, the extent to which these same relationships hold in heavy-drinking young adults is unknown.

The notion that affect and urge motivate drinking is central to models of alcohol use and substance use in general (Baker et al., 2004; Conger 1956; Cooper, Frone, Russell, Mudar, 1995), and investigations of within-person associations in young adult samples have shown that increases in negative (e.g., Hussong, 2007; O’Hara et al., 2014; Park, Armeli, & Tennen, 2004; Simons et al., 2005; Simons, Dvorak, Batien, & Wray, 2010; Simons, Wills, Neal, 2014) and positive affect are related to alcohol use (Collins et al., 1998; Simons et al., 2010; Simons, Wills, Neal, 2014). However, very few of these investigations have studied clinical populations of heavy-drinking young adults (e.g., O’Hara et al., 2014; Park et al., 2004), and the extent to which pharmacotherapy influences associations among mood, urge, and alcohol use in this population is unknown.

The current study builds on previous work by using hierarchical linear models to examine whether targeted naltrexone influences within-person relations among mood, urge, and alcohol use in a sample of young adults enrolled in an 8-week clinical trial of counseling and pharmacotherapy (see O’Malley et al., 2015). Participants were randomized to receive double-blind naltrexone or placebo on a daily (25 mg) and “as needed” targeted basis (25 mg), totaling up to 50 mg daily. In the primary analyses, naltrexone did not reduce drinking frequency (percent days abstinent) or percent heavy drinking days (based on ≥4 drinks for women or ≥5 for men) but was shown to significantly reduce the percentage of drinking days with an eBAC≥.08g% (O’Malley et al., 2015). Our current investigation builds on this finding by examining within-person how affect and urge influence risk for intoxication (eBAC≥.08g%) during treatment and explores possible within-person mechanisms that may contribute to the effects of targeted naltrexone.

Based on prior findings (Kranzler et al., 2004), we hypothesized that higher negative and positive affect would independently predict greater odds of intoxication, measured by eBAC≥.08g%, and that heightened positive or negative affect would indirectly influence intoxication risk by increasing the urge to drink. Additionally, we expected heightened affect or urge would predict when participants took a targeted dose of medication as a potential coping strategy in anticipation of a drinking event, and we expected that naltrexone would attenuate the risk of eBAC≥.08g% in response to these states more than placebo. Given evidence that alcohol use among young adults is influenced by gender and family history of alcohol dependence (Courtney & Polich, 2009; SAMHSA, 2014; Warner, White, Johnson, 2007), we also explored whether the within-person effects were influenced by these person-level factors. Identifying cues of risk and within-person effects of targeted naltrexone may help to inform the optimal use of targeted dosing.

Method

Participants

This study is a secondary analysis of data from 140 young adults enrolled in a randomized, double-blind placebo-controlled clinical trial of naltrexone in combination with brief motivational counseling for reducing drinking (see O’Malley et al., 2015). Eligible participants were 18-25 years old, reported at least 4 heavy drinking days (i.e., ≥4 drinks/women or ≥5 drinks/men) in the prior 4 weeks, were able to read English, had no significant cognitive impairment, and were not currently pregnant or breastfeeding. Participants were excluded if they had a current serious psychiatric illness or a past 12-month DSM-IV diagnosis of drug dependence other than nicotine. Of the 140 randomized participants, 127 provided daily diary data and were included in the current analyses. Baseline characteristics are summarized by medication condition in Table 1. Participants were 21.4 years old on average (SD=2.2); 68.5% were male; 37.8% had a family history of alcohol dependence; and the majority self-identified as white (78.0%) and single (97.6%).

Table 1.

Characteristics of the 127 participants with daily diary data

	Naltrexone (N=61)	Placebo (N=66)
Demographics
Age, mean (SD)	21.6 (2.2)	21.2 (2.0)
Gender (male), N (%)	43 (70.5%)	44 (66.7%)
Single marital status, N (%)	59 (96.7%)	65 (98.5%)
Race
White, N (%)	49 (80.3%)	50 (75.8%)
African American, N (%)	5 (8.2%)	3 (4.5%)
Multiracial, N (%)	2 (3.3%)	4 (6.1%)
Asian, N (%)	2 (3.3%)	2 (3.0%)
American Indian, N (%)	1 (1.6%)	0 (0.0%)
Other, N (%)	2 (3.3%)	7 (10.6%)
Highest Education
High school or less, N (%)	7 (11.5%)	11 (16.7%)
Some college or associates degree, N (%)	33 (54.1%)	38 (57.6%)
College or advanced degree, N (%)	21 (34.4%)	17 (25.7%)
Family History of Alcohol Dependence
Positive, N (%)	24 (39.3%)	24 (36.4%)
Negative/Unknown, N (%)	37 (60.7%)	42 (63.6%)
Baseline Alcohol Use
Percent drinking days, mean (SD)	56.7 (21.8)	49.8 (13.8)
Drinks per drinking day, mean (SD)	6.7 (2.9)	6.8 (2.5)

Note: Baseline characteristics did not differ by medication condition, with the exception of a trend level difference in baseline drinking frequency (percent drinking days), p=.07. The original report (O’Malley et al., 2015) presents demographic data on the 128 participants who started treatment.

Procedures

The Yale University Institutional Review Board approved all study procedures. Participants were recruited primarily through Facebook and fliers seeking research volunteers who wanted to cut down on their alcohol use. Participants were screened by phone or online surveys, and eligible individuals were invited for an interview, where study procedures were explained and written informed consent was obtained. Participants received up to $415 in total for completing appointments and assessments.

Medication

This was a double-blind placebo-controlled study in which young adults were randomly assigned to receive either naltrexone (25 mg daily + 25 mg targeted) or placebo (placebo daily + placebo targeted) for 8 weeks. To maximize tolerability, participants were instructed to take a single targeted dose of medication 2 hours prior to drinking for the first week and to add the daily dose beginning in the second week. From the second week through the end of treatment, participants were instructed to take targeted medication as needed prior to drinking in addition to a daily dose.

Baseline Measures

Participants provided baseline demographic information, including age, gender, family history of alcohol dependence, and measures of alcohol consumption (quantity and frequency) during the preceding three months at baseline.

Daily Diaries

Participants completed once daily web-based assessments (DatStat^TM) for 8 weeks. Surveys were brief reports (<5 minutes) assessing mood, desire to drink, and number of drinks consumed, among other measures. Participants were instructed to complete the daily diary assessment as early in the day as possible after waking. Daily automated email prompts were delivered from the DatStat system to remind participants to complete the daily diary. Participants were compensated for completing the daily reports ($2 for each day up to $120, and a $25 bonus for completing all assessments for two consecutive weeks during treatment, up to $100). Participants were asked to provide mood ratings at the time of the assessment and respond about urge intensity, alcohol use, and medication retrospectively during the previous day. Thus, consecutive reports could be selected to model time-lagged effects of mood earlier in the day predicting urge and alcohol use later in the day. In total, 80.9% of reports were completed, although completion rates varied by subject (M=75.8%, SD=28.6%). If participants missed an assessment, they could select an option to complete a “make-up survey” (26.6% of reports) that did not include a mood assessment but provided an opportunity to respond about their urge to drink, alcohol use, and medication use up to two days prior, to limit retrospective recall bias. If participants were unable to access a computer, they were given paper assessments to complete each day (5.8% of reports).

Affect

Participants reported their current mood at the time of the survey with 5 items rated from 1 (not at all) to 7 (extremely). Positive affect items included: “enthusiastic,” “excited,” and “happy.” Negative affect items included “distressed” and “sad.” Scores on items were averaged to create a composite daily rating of positive and negative affect (α=.92 and .73 for positive affect and negative affect, respectively).

Urge

Participants reported on their urge to drink during the previous day. Urge to drink was measured with three items: “I felt like I could have really used a drink,” “The idea of drinking was appealing,” and “I really didn’t feel like drinking” (reverse-scored). Participants rated agreement on these items from 1 (strongly disagree) to 9 (strongly agree). Scores were averaged to create a composite rating of urge to drink (α=.86).

Alcohol use

Participants reported the number of standard drinks of alcohol they consumed in total during the previous day and the times that they started and finished drinking. The primary outcome of interest was a daily measure of drinking to intoxication, defined as estimated BAC≥.08g% on drinking days. Estimated BAC (eBAC) was based on the number of standard drinks reported, duration of drinking, and total body water (calculated from gender, age, height and weight) (Curtin & Fairchild, 2003).

Medication

Participants reported whether they took a daily dose (yes/no) and a targeted dose (yes/no) of medication during the previous day, as well as the time that they took the dose. For analyses examining the effect of taking a targeted dose of medication above and beyond the daily dose, dose was coded as 1=targeted plus daily dose vs. 0=daily dose only. Participants also reported their reasons for taking a targeted dose of medication, selecting all that applied from the following: “in anticipation of high-risk drinking,” “in a high-risk situation,” “as a preventative measure, just in case,” or “to comply with the study.”

Data Analysis

A series of multilevel models were tested using Hierarchical Linear Modeling (HLM) version 6.02 software. Daily reports during the 8-week treatment period comprised level-1 variables nested within person at level-2. Descriptive statistics were used to characterize predictors of interest by medication condition, gender, and family history of alcohol dependence (Table 2). Analyses were limited to consecutive reports within the treatment period to model lagged effects isolating earlier mood as a predictor of later drinking outcome. Reports were excluded from analyses where the lagged mood rating occurred after drinking began (n=191), leaving n=4863 reports for the current analyses. Descriptive analyses were used to verify the temporal sequencing of affect, medication, and alcohol use in the reports retained for analysis. The median times were as follows, mood assessment: 11:00 a.m.; time of first drink: 8:00 p.m.; time of daily dose: 8:00 a.m.; time of targeted dose: 7:00 p.m. It was not possible to identify the time of day for urge since it was reported for the previous day (no time specified).

Table 2.

Descriptive statistics for predictors of interest across the treatment period

	Variable, mean (SD)
	Negative Affect	Positive Affect	Urge
Medication Condition
Naltrexone (N=61)	2.13 (1.01)	4.12 (1.20)	4.34 (1.35)
Placebo (N=66)	2.14 (0.90)	4.00 (1.31)	4.20 (1.12)
Gender
Men (N=87)	2.16 (0.92)	4.10 (1.18)	4.40 (1.26)
Women (N=40)	2.08 (1.04)	4.00 (1.42)	3.98 (1.14)
Family History of Alcohol Dependence
Positive (N=48)	2.16 (1.02)	3.83 (1.42)	4.36 (1.37)
Negative (N=79)	2.12 (0.91)	4.21 (1.12)	4.21 (1.15)

Note: Values reflect average person-level means across the 8 week treatment period. Values were not statistically different by medication condition, gender, or family history of alcohol dependence (ps>.05). Ranges: negative and positive affect 1-7, urge 1-9.

All models were set with a random intercept and random effects for predictors of interest (affect, urge, medication) to allow regression coefficients to vary across individuals. Including random effects for each predictor of interest (affect, urge, medication) resulted in significant model improvements compared to models with these effects fixed (Hox, 2010; Raudenbush & Bryk, 2002), χ2=7.18-15.93, df=2, p<.05. Models were run using full information maximum likelihood estimation. All models controlled for time (measured as treatment day), gender, and family history of alcohol dependence, based on research indicating differential benefit from naltrexone based on family history (Krishnan-Sarin, Krystal, Shi, Pittman, O’Malley, 2007; Monterosso et al., 2001). All models included six contrast variables for days of the week to control for weekly drinking trends. Results are presented from population-average models with robust standard errors.

In the first set of analyses, we tested the mediated effect of affect on drinking outcome through urge. Analyses specified a Bernoulli distribution with a logit link function to model the binary drinking outcome (eBAC≥.08g% on drinking day, yes/no). Of the n=1833 days where any drinking was reported, 36.6% (n=671) were days on which eBAC≥.08g%. Next, we examined whether affect and urge predicted taking a targeted dose of medication and the effect of this dose on drinking outcome. The indirect effects of positive affect and urge on drinking outcome through targeted dosing were evaluated separately for naltrexone and placebo to test the moderated mediation effect.

The basic multilevel model is shown below.

$$\begin{array}{cc}\hfill & \mathrm{Log}[{\mathrm{P}}_{\text{it}}∕{\text{1-P}}_{\text{it}}]={\pi }_{0\mathrm{i}}+{\pi }_{1\mathrm{i}}\left({\text{PositiveAffect}}_{\text{it}}\right)+{\pi }_{2\mathrm{i}}\left({\text{NegativeAffect}}_{\text{it}}\right)+{\pi }_{3\mathrm{i}}\left({\text{Urge}}_{\text{it}}\right)+\hfill \\ \hfill & {\pi }_{4\mathrm{i}}\left({\text{Dose}}_{\text{it}}\right)+{\pi }_{5\mathrm{i}}\left({\text{Time}}_{\text{it}}\right)+{\mathrm{e}}_{\text{it}}\hfill \\ \hfill & {\pi }_{0\mathrm{i}}={\beta }_{00}+{\beta }_{01}\left({\text{MeanPositiveAffect}}_{\mathrm{i}}\right)+{\beta }_{02}\left({\text{MeanNegativeAffect}}_{\mathrm{i}}\right)+{\beta }_{03}\left({\text{MeanUrge}}_{\mathrm{i}}\right)+\hfill \\ \hfill & {\beta }_{04}\left({\text{Gender}}_{\mathrm{i}}\right)+{\beta }_{05}\left({\text{FamilyHistory}}_{\mathrm{i}}\right)+{\mathrm{u}}_{0\mathrm{i}}\hfill \\ \hfill & {\pi }_{1\mathrm{i}}={\beta }_{10}+{\mathrm{u}}_{1\mathrm{i}}\hfill \\ \hfill & {\pi }_{2\mathrm{i}}={\beta }_{20}+{\mathrm{u}}_{2\mathrm{i}}\hfill \\ \hfill & {\pi }_{3\mathrm{i}}={\beta }_{30}+{\mathrm{u}}_{3\mathrm{i}}\hfill \\ \hfill & {\pi }_{4\mathrm{i}}={\beta }_{40}+{\mathrm{u}}_{4\mathrm{i}}\hfill \\ \hfill & {\pi }_{5\mathrm{i}}={\beta }_{50}\hfill \end{array}$$

Where P_it is the probability of person i having an eBAC≥.08g% on day t; π_0i is the intercept, which estimates person i’s log odds of eBAC≥.08g% on day t when all other predictors are 0; π_1i – π_5i are the partial within-person regression coefficients for person i; and e_it is a random residual component, which is the error term. Level-1 continuous variables were centered on person-specific means and the person means were entered at level-2. This allows for disaggregation of the within-person and between-person effects of the predictors on drinking outcome (Raudenbush & Bryk, 2002). For example, PositiveAffect_it represents how much the positive affect rating on day t for person i differed from their usual rating averaged across all reports (MeanPositiveAffect_i).

To examine the effect of targeted naltrexone versus targeted placebo on drinking outcome, cross-level interactions were tested where medication (naltrexone vs. placebo) was added as a level-two predictor of the level-one coefficient for dose (targeted + daily vs. daily):

$${\pi }_{4\mathrm{i}}={\beta }_{40}+{\beta }_{41}\left({\text{Medication}}_{\mathrm{i}}\right)+{\mathrm{u}}_{4\mathrm{i}}$$

where β₄₁ is the coefficient for the effect of naltrexone on the dose–drinking relationship. Reports from the first week of medication titration when participants were instructed to take only the targeted dose (n=126) were omitted from this primary analysis, but were the focus of separate exploratory analyses about the effect of targeted naltrexone alone without a daily dose (see results below).

Lastly, cross-level interactions were tested where gender (male=1, female=0) and family history of alcohol dependence (positive=1, negative=0) were added as moderators of these level-1 effects to see if the strength of the relations differed by gender or family history status.

Results

Over ninety percent of randomized participants provided daily diary data within treatment. Demographic characteristics did not differ significantly between the analyzed sample (n=127) and the 13 participants who were enrolled but did not provide daily diary data. However, participants who did not provide data consumed significantly more drinks per drinking day in the 30 days prior to baseline (M=8.6, SD=3.6) than those who were retained in treatment (M=6.8, SD=2.7), t(138)=−2.34, p=.02.

Urge as a Mediator of Affect–Drinking Relations

Figure 1 displays the relations among positive affect, urge, and drinking outcome, eBAC≥.08g%. Negative affect was not significantly related to urge (B=.03, SE=.06, p=.54) or drinking outcome (B=−.05, SE=.07, p=.50) and was retained in the model but not shown in the figure. Coefficients depict the pooled within-person slopes adjusted for other covariates. Positive affect was significantly related to later drinking (c path), such that individuals were more likely to drink to intoxication on days when their positive affect was higher than usual. Specifically, a one-unit increase from within-person average positive affect predicted a 14% increase in the odds of drinking to intoxication (eBAC≥.08g%). Urge was also positively associated with drinking (b path), such that a one-unit increase in urge was associated with a 57% increase in the odds of drinking to intoxication (eBAC≥.08g%).

Figure 1.

Urge as a mediator of daily affect–drinking relations.

This figure illustrates daily within-person relations among affect, urge, and drinking outcome, eBAC≥.08g%. Coefficient estimates can be exponentiated to derive odds ratios for affect-drinking and urge-drinking relations.

The indirect effect of affect on drinking through urge was estimated as the product of the path between affect and urge (a path) and the path between urge and drinking outcome (b path). Standard errors and confidence intervals around the indirect effect (ab) were computed using the RMediation program (Tofighi & MacKinnon, 2011). The indirect effect of positive affect on drinking through urge was statistically significant, ab=.11, SE=.02, 95%CI=.07-.16 and the relation between positive affect and drinking was fully mediated by urge; positive affect was no longer a significant predictor of eBAC≥.08g% (c’path) once accounting for the indirect effect through urge.

Examination of cross-level interactions with gender indicated that the relation between positive affect and eBAC≥.08g% (c path) was weaker for men than women, B=−.27, SE=.11, p=.02, while the association between urge and eBAC≥.08g% (b path) was stronger for men than women, B=.24, SE=.09, p=.01. Gender did not significantly moderate the associations between positive or negative affect and urge (a path, ps>.62). Examination of cross-level interactions with family history of alcohol dependence indicated that none of these relations were significantly influenced by family history status (positive versus negative) (ps>.13).

Effect of Targeted Naltrexone on Affect–Drinking and Urge–Drinking Relations

Next, we examined the influence of a targeted dose of medication on affect–drinking and urge–drinking relations (Figure 2). Negative affect was not significantly related to taking a targeted dose of medication, B=−.004, SE=.04, p=.91, (not shown in Figure 2). Of the n=4737 days included in the primary analyses, daily dose alone was taken 60.0% of the time (n=2840), a targeted dose was taken in addition to a daily dose 24.2% of the time (n=1145), no medication was taken 13.2% of the time (n=629), and medication information was not reported 2.6% of the time (n=123). On days of eBAC≥.08g%, a targeted dose was taken 60.4% of the time. The most commonly cited reasons for taking a targeted dose were: “to comply with the study” (n=815), “in anticipation of high risk drinking” (n=565), “as a preventative measure” (n=281), and “in a high-risk situation” (n=118).

Figure 2.

Targeted medication as a mediator of daily urge–drinking and affect–drinking relations. Urge and positive affect predict taking a targeted dose of medication, and medication condition (naltrexone versus placebo) moderates the effect of taking a targeted dose on drinking outcome, eBAC≥.08g%.

We examined taking a targeted dose of medication as a mediator of urge–drinking and positive affect–drinking relations. The effect of taking a targeted dose of medication was significantly moderated by medication condition (v), such that taking a targeted dose was associated with greater odds of eBAC≥.08g% in the placebo B=.48, SE=.23, p=.04 but not naltrexone condition B=.28, SE=.24, p=.25 (Figure 3). The indirect effects of urge and positive affect on drinking through targeted dose differed by medication condition (moderated mediation). Specifically, the indirect effects of urge and positive affect on drinking outcome were significant in the placebo (a₁b=.24, SE=.12, 95%CI=.01-.48; a₂b=.07, SE=.04, 95%CI=.01-.16) but not naltrexone condition (a₁b=.16, SE=.14, 95%CI=−.11-.44; a₂b=.02, SE=.03, 95%CI=−.02-.09). In other words, naltrexone buffered against the indirect effects of urge and positive affect on drinking to intoxication (eBAC≥.08g%) through targeted dosing.

Figure 3.

The effect of targeted naltrexone vs. placebo on odds of drinking outcome, eBAC≥.08g%. Subjects were randomized to naltrexone (dosing: 25mg daily, 25mg targeted) or placebo.

Examination of these effects by gender or family history indicated no significant differences, with the exception that the association between heightened urge and taking a targeted dose of medication (a₁) was weaker for individuals with a positive family history of alcohol dependence (compared to a negative family history), B=−.26, SE=.11, p=.02. This effect was seen in the naltrexone (B=−.36, SE=.08, p<.001) but not placebo condition (B=.17, SE=.11, p=.12); however, drinking outcomes did not differ by family history status (B=.34, SE=.21, p=.11).

Given that young adults may prefer to take medication “as needed” instead of on a daily basis, we conducted initial exploratory analyses comparing drinking outcomes on days where only a targeted dose was taken (n=126) with days where no medication was taken (n=629). A similar moderating effect was seen, with targeted placebo resulting in significantly greater odds of eBAC≥.08g% than targeted naltrexone (54.7% vs. 27.5%), B=−1.66, SE=.48, p=.001.

Discussion

We examined daily relations among mood, urge, and drinking to intoxication (eBAC≥.08g%) in a sample of young adults enrolled in a double-blind clinical trial studying the efficacy of daily (25 mg) plus targeted (25 mg) naltrexone versus placebo. Results supported a mediating role of urge in positive affect–drinking relations. Specifically, positive affect predicted a greater urge to drink, which in turn was associated with greater odds of eBAC≥.08g%. Our study is the first to examine within-person effects of targeted naltrexone on drinking in heavy-drinking young adults. Results indicated that taking targeted naltrexone buffers against the indirect effects of positive affect and urge on drinking to intoxication (eBAC≥.08g%). These findings provide further evidence of naltrexone’s utility in reducing heavy drinking in young adults and highlight within-person processes related to alcohol use that could potentially be used to optimize this intervention.

The finding that within-person positive, but not negative, affect was a unique risk factor related to higher urge and drinking outcome in this sample aligns with other experience sampling research showing consistent relations between within-person positive affect and drinking (Collins et al., 1998; Simons et al., 2010; Simons, Wills, Neal, 2014) and mixed or less robust associations between negative affect and drinking (e.g., Collins et al., 1998; Simons et al., 2010). Heightened positive affect may be an especially salient risk factor for alcohol use in young adults, as others have identified that young adults often endorse drinking for enhancement or social motives (i.e., “to celebrate,” “to have a good time”) (Kuntsche, Knibbe, Gmel, & Engels, 2005). Additionally, several studies have noted that young adults who drink heavily endorse these enhancement motives more than others who drink less (e.g., Kairouz et al., 2002; Carey et al., 1993). The current findings extend prior work by providing greater temporal specificity—positive affect earlier in the day predicted later-day drinking to intoxication (eBAC≥.08g%)—and showing a possible indirect mechanism linking positive affect and drinking via increased urge. Together, these findings suggest that within-person increases in positive affect, especially when accompanied by high urge, are important risk indicators that can signal the need for intervention. Additionally, our results suggest that the risk processes for drinking may differ by gender, such that greater positive affect is more strongly related to eBAC≥.08g% for women and urge is more strongly related to eBAC≥.08g% for men. Further research is needed to understand how gender influences these risk processes to inform potential gender-specific interventions to reduce alcohol use.

Based on the results of the current study, targeted naltrexone may be an effective intervention for attenuating urge–drinking and positive affect–drinking relations in young adults. Taking a targeted dose of medication predicted greater risk of intoxication (eBAC≥.08g%) in the placebo but not naltrexone condition, while the risk of intoxication on days with only a daily dose was similar for both naltrexone and placebo conditions. One of the most commonly cited reasons for taking a targeted dose of medication was “in anticipation of high risk drinking.” Our findings may indicate that young adults were aware of an upcoming drinking episode and were relying on the targeted dose of medication to help reduce risk, and this was only effective when they were taking active medication.

The current findings extend prior work which has been limited to examining only the between-subject effects of targeted versus daily naltrexone (Hernandez-Avila et al., 2006; Kranzler et al., 2004; Kranzler et al., 2009) by identifying the benefit of taking a targeted dose in addition to a daily dose on a given day. In particular, given that targeted medication adherence varies at the daily level, it is especially important to model these relations within-person to better understand the effectiveness of the medication. Furthermore, our results provide preliminary evidence that young adults with a positive family history of alcohol dependence are less likely to take a targeted dose of naltrexone (in addition to a daily dose) in response to a heightened urge to drink than those without a family history of alcohol dependence. More research is needed to understand this effect and potential reasons (e.g., intention to drink, lack of planning for targeted dose). However, these individuals appeared to be somewhat protected against drinking to intoxication by the daily dose, given that drinking outcomes did not differ by family history status. Daily dosing was intended to reduce alcohol consumption in the event that targeted (i.e., as needed) dosing was missed or forgotten, suggesting that this dosing strategy may be particularly helpful for individuals with a positive family history of alcohol dependence.

These results also have important clinical implications related to optimizing targeted medication adherence. Because drinking days on which only daily medication doses were taken represent missed opportunities to take a targeted dose, efforts to promote anticipation of high-risk drinking situations may be useful. Our findings suggest that it may be especially important to encourage medication adherence on days of heightened positive affect given evidence that positive affect earlier in the day was associated with heightened urge and later risky alcohol use. Finding ways to promote adherence to naltrexone is particularly critical since several studies document generally low rates of medication adherence among individuals with alcohol problems (e.g., Garbutt, 2009; Weiss, 2004).

Targeted dosing alone may be a useful strategy for promoting medication adherence in some young adults given that they often prefer to take medication as needed (Epler et al., 2009). Our study provides preliminary evidence that taking a targeted dose (without a daily dose) reduced the odds of eBAC≥.08g% compared to days when no medication was taken. However, the present results cannot adequately address the utility of targeted naltrexone alone given that dosing was prescribed as daily + targeted dosing. Taking a targeted dose of medication (versus no medication) may be related to other factors (e.g., motivation to reduce drinking) that also influenced drinking behavior on these days. Thus, additional experimental research is needed to evaluate the effectiveness of targeted naltrexone alone in this population.

The present results should be interpreted with the following limitations in mind. First, the sample comprised a relatively homogenous group of young adult heavy drinkers who were willing to participate in a randomized clinical trial and were compensated for their participation. Thus, the results may not generalize to samples with different demographic characteristics or individuals attempting to stop or reduce drinking on their own. In particular, our sample was restricted to young adults without current serious psychiatric illness, and different mood–drinking associations might be seen in those with current mood or anxiety disorders. Second, results are limited due to missing responses on daily diary assessments and the temporal ordering of constructs assessed. While a strength of the study is that lagged assessments allowed for an examination of mood prior to drinking, the temporal sequence of urge with affect, medication use, and drinking are unknown because urge was assessed globally for the previous day. Predictors of interest (e.g., affect and urge) likely vary throughout the day and a single assessment within a day does not capture this variability. Future work would benefit from using more frequent assessments to help specify these causal processes with greater temporal precision. Third, the brief, limited measures of affect that were used may have influenced the lack of associations between negative affect and outcomes. Future work would benefit from replicating these findings using additional measures of affect. Fourth, our outcomes relied on self-reported daily alcohol use and estimated BAC levels. However, we believe that modeling drinking according to eBAC≥.08g% is likely to be more sensitive and clinically meaningful than other heavy drinking definitions used in prior studies. Lastly, we also relied on self-reported medication use each day, although participants were not incentivized for medication adherence and were informed that they could continue to receive counseling support and participate in the research study if they chose to stop medication.

Despite its limitations, the current study provides new information on the daily relations among mood, urge, and alcohol use in heavy-drinking young adults that can help to inform risk reduction in this population. Importantly, these results suggest that heightened positive affect and urge are signals for alcohol use in young adults and taking targeted naltrexone in response to these states can reduce the odds of drinking to intoxication (eBAC≥.08g%). Identifying optimal interventions for heavy-drinking young adults is an important priority to reduce long-term consequences and alcohol use problems.

Public Health Significance.

Positive affect predicts heightened urge and increased risk of drinking to intoxication (BAC≥.08g%) on a daily level among young adults. Taking targeted naltrexone buffers against the indirect effects of positive affect and urge on drinking to intoxication. Targeted naltrexone may be a useful intervention to reduce alcohol consumption and related problems in young adults.