Relationships between Generalized Impulsivity and Subjective Stimulant and Sedative Responses Following Alcohol Administration

Abstract

Impulsivity and subjective response to alcohol (SR; i.e., individual differences in sensitivity to pharmacologic alcohol effects) are both empirically supported risk factors for alcohol use disorder (AUD); however these constructs have been infrequently studied as related risk factors. The present investigation examined a self-report measure of impulsivity (i.e., the Barratt Impulsiveness Scale, version 11) in relation to acute alcohol effects (i.e., stimulant and sedative SR). Participants came from two cohorts of Chicago Social Drinking Project. Heavy and light drinkers from Cohort 1 (n=156) and heavy social drinkers from Cohort 2 (n=104) were examined using identical laboratory protocols following oral alcohol administration using a within-subject, double-blind, placebo-controlled laboratory study design. Self-reported impulsivity and, for comparison purposes, sensation seeking, were measured at baseline and SR was measured once prior to and four times following alcohol administration. More impulsive light, but not heavy drinkers, reported heightened stimulant SR following alcohol administration High impulsive, light drinkers reported stimulant SR at a magnitude similar to heavy drinkers, whereas low impulsive, light drinkers reported limited stimulant SR. The interaction between impulsivity and sensation seeking did not statistically predict stimulant SR and overall, impulsivity was a stronger predictor than sensation seeking. However, impulsivity was not statistically predictive of dampened sedative SR among light or heavy drinkers. These findings partially replicate and extend the recent literature linking self-reported impulsivity to heightened stimulant SR from alcohol. Future directions include longitudinal studies and research relating multiple facets of impulsivity to SR.

Excessive alcohol use is common and accounts for considerable economic burden to society (Sacks, Gonzales, Bouchery, Tomedi, & Brewer, 2015). Despite clinical and public health efforts, past-year alcohol use and frequent heavy drinking have increased by 11.2% and 29.9% respectively from 2001–2002 to 2012–2013. Further, alcohol use disorder (AUD) prevalence has increased from 8.5% to 12.7% in the same period (Grant et al., 2017). Thus, identifying and learning more about constructs related to heavy drinking and AUD are vital for ascertaining which drinkers are at particular risk and to inform intervention efforts.

Impulsivity is a heterogeneous construct that refers to a tendency toward immediate action with diminished regard for future consequences (Brewer & Potenza, 2008; Moeller, Barratt, Dougherty, Schmitz, & Swann, 2001; Caswell, Bond, Duka, & Morgan, 2015; Hamilton, Ansell, Reynolds, Potenza, & Sinha, 2013; Lane, Cherek, Rhoades, Pietras, & Tcheremissine, 2003; MacKillop et al., 2016). Impulsivity is associated with multiple alcohol-related outcomes including negative consequences of drinking and AUD symptoms (Coskunipar & Cyders, 2013; Stanford et al., 2009). Measurement of impulsivity can be complex but most often includes either self-report measures of impulsivity that assess aggregated perceptions of one’s behavior (e.g., rating the degree to which one typically acts quickly without thinking) or behavioral task measures, such as delay discounting (i.e., an inordinate tendency to favor immediate rewards) and response inhibition (i.e., inordinate difficulty withholding response to prepotent stimuli) (Jentsch et al., 2014; K. King, Patock-Peckham, Dager, Thimm, & Gates, 2014). Impulsivity is also a key factor for several clinical conditions, including AUD, attention deficit hyperactivity disorder (ADHD) (Hendershot et al., 2015), and other impulsive and compulsive disorders (Fineberg et al., 2014).

Like impulsivity, individual differences in sensitivity to alcohol’s pharmacological effects, as measured by self-reported subjective responses to alcohol (SR), relates to risk for AUD (for reviews see Morean & Corbin, 2010 and Quinn & Fromme, 2011). The Low Level of Response (LLR) model (Schuckit, 1994) posits that globally-reduced alcohol sensitivity confers AUD risk, perhaps due to weaker signals to slow down or stop drinking. Thus, by this theory, individuals who experience low response to alcohol’s effects may eventually drink heavily and incur heightened AUD risk (Morean & Corbin, 2010). In contrast, the Differentiator Model (DM; King, de Wit, McNamara, & Cao, 2011; Newlin & Thomson, 1990) posits that high-risk SR patterns include both heightened hedonic, stimulant and dampened aversive, sedative SR. Recent research has shown that greater sensitivity to hedonic, stimulant SR had stronger associations to AUD symptoms longitudinally than lower sensitivity to aversive, sedative SR (King, Hasin, O’Connor, McNamara, & Cao, 2016; King, McNamara, Hasin, & Cao, 2014). In light of these findings, researchers have sought to identify factors related to high-risk SR patterns. To date, a family history of alcohol problems and heavy drinking have been two commonly addressed factors (for reviews, see Quinn & Fromme, 2011; Morean & Corbin, 2010).

Although impulsivity and SR are both independent risk factors for AUD, they have rarely been studied in relation to each other. Across multiple domains, more impulsive individuals tend to neglect or discount indicators of punishment and focus inordinately on reward (Patterson & Newman, 1993; Potts, George, Martin, & Barratt, 2006). If this general, impulsive tendency toward elevated reward and diminished punishment also applies to alcohol use specifically, impulsive individuals with high-risk SR patterns may be particularly prone to negative outcomes. Moreover, SR patterns may help to explain why impulsivity confers future risk for negative alcohol-related outcomes.

Recent studies in heavy drinkers have attempted to relate facets of impulsivity to high-risk SR patterns. In a sample of young adults who tended to engage in heavy drinking, greater response inhibition difficulties on a Cued Go/No-Go Task (Marczinski & Fillmore, 2003) following oral alcohol were significantly related to elevated stimulant but not dampened sedative SR while blood alcohol concentrations (BACs) were at peak (target BAC=0.08%) (Quinn & Fromme, 2016). Among drinkers with current alcohol dependence given intravenous (IV) alcohol administration, greater delay discounting was associated with heightened stimulant SR while BACs were rising (Westman, Bujarski, & Ray, 2016). Contrary to initial hypotheses, Westman and colleagues (2016) also found a positive relation between delay discounting and heightened sedative SR in both the alcohol and placebo conditions, indicating that more impulsive individuals’ perception of sedation was due to factors beyond pharmacologic alcohol effects. In another IV alcohol study, young adult heavy drinkers who endorsed more attention deficit hyperactivity disorder (ADHD) symptoms (of which impulsivity is a hallmark), reported greater stimulant but not dampened sedative SR while BACs rose (Hendershot et al., 2015). In contrast, retrospective, cross-sectional data from young adult heavy drinkers showed that positive urgency (i.e., self-reported impulsivity following strong positive affect) measured using the UPPS Impulsive Behavior Scale (Whiteside & Lynam, 2001; Cyders et al., 2007) correlated significantly with retrospective, self-reports of dampened sedative SR (Wardell, Quilty, & Hendershot, 2015).

In addition to studies of impulsivity and SR in heavy drinkers, examining the relationships between impulsivity and SR in lower risk, light drinkers is important as it has the advantage of removing the confound of heavy alcohol exposure, which affects both SR (Morean & Corbin, 2010) and impulsivity (see Leeman, Grant, & Potenza 2009). Indeed, one recent study showed that among light social drinkers, self-reported impulsivity on the Barratt Impulsiveness Scale (BIS-11; Patton, Stanford, & Barratt, 1995) related to heightened stimulant and dampened sedative SR after IV alcohol administration as BACs were rising and then maintained at peak using a clamping approach to maintain consistent BAC (Leeman et al., 2014). A similar relationship was shown between self-reported impulsivity on the BIS-11 and dampened sedative SR among light social drinkers using retrospective self-report data (Berey, Leeman, Pittman, & O’Malley, 2017). However, few studies have examined relations between impulsivity and SR among both light and heavy drinkers in the same paradigm.

While relationships between impulsivity and negative alcohol-related outcomes have been demonstrated, the combination of impulsivity and sensation seeking may carry the greatest degree of risk (Ersche et al., 2013; Ersche, Turton, Pradhan, Bullmore, & Robbins 2010). Sensation seeking is defined as the pursuit of “varied, novel, complex, and intense sensations and experiences, and the willingness to take physical, social, legal, and financial risks for the sake of such experiences” (Zuckerman 1994, p. 27, emphasis in original). Like impulsivity, heightened sensation seeking on the Zuckerman SSS-V (Zuckerman, Eysenck, & Eysenck, 1978) has been associated with difficulties resisting prepotent responses and has been related to elevated reward from alcohol (Fillmore, Ostling, Martin, & Kelly, 2009). However, sensation seeking does not necessarily entail lack of regard for consequences as impulsivity does. Further, impulsivity tends to have stronger relationships to alcohol-related problems whereas sensation seeking relates more closely to frequency and quantity of consumption, thus the constructs are related, yet distinct (Dick et al., 2010; MacKillop et al., 2016). Accordingly, it is important to determine whether high-risk SR patterns relate to impulsivity, to sensation seeking, or both.

The primary goal of this study was to examine relationships between self-reported impulsivity and SR in heavy and light drinkers utilizing laboratory-based oral alcohol administration data from the Chicago Social Drinking Project (CSDP) (King et al., 2011; 2016; Roche, Palmeri, & King, 2014). Based on prior research (Berey et al., 2017; Leeman et al., 2014), we hypothesized that more impulsive, light drinkers would self-report heightened stimulant and dampened sedative SR in the present study. Based on prior human laboratory findings involving other facets of impulsivity (Hendershot et al., 2015; Westman et al., 2016) and retrospective, self-report results (Wardell et al., 2015), we predicted parallel relationships between self-reported impulsivity and heightened stimulant and dampened sedative SR among heavy drinkers. Due to greater risk of negative alcohol-related outcomes among heavy drinkers, we further predicted stronger relationships between self-reported impulsivity and SR among heavy compared to light drinkers.

Regarding relationships to time and limb effects, it was essential to include a time variable in our analyses as stimulant effects predominate on the ascending limb and sedative effects predominate on the descending limb of the blood alcohol curve. We predicted significant three-way interactions of self-reported impulsivity by heavy/light drinker group by time, at minimum because subjective effects are at lowest magnitude at the earliest and latest timepoints, making observation of relationships to impulsivity less likely at those times. However, the few prior results (Leeman et al., 2014; Hendershot et al., 2015; Westman et al., 2016) are equivocal regarding the extent that relationships between self-reported impulsivity and subjective effects are limited to the rising or declining limbs of the BAC curve or observable at both. Thus, we made no specific predictions regarding localization of relationships between impulsivity and subjective response to a particular limb of the BAC curve.

A secondary goal was to address whether sensation seeking had unique relationships to SR and if those endorsing both self-reported impulsivity and sensation seeking were particularly likely to show high-risk SR patterns. Considering prior findings that those endorsing both sensation seeking and impulsivity tend to report the most severe substance-related outcomes (Ersche et al., 2010; Ersche et al., 2013), it was important to determine whether a similar pattern applied to high-risk SR patterns..

The first CSDP cohort included light and heavy drinkers, enabling determination of the extent to which hypothesized relationships pertained to both groups. The second CSDP cohort enrolled heavy drinkers only, enabling replication of results involving heavy drinkers from the first cohort.

Method

Participants

Data for this secondary analysis were culled from the two cohorts of the Chicago Social Drinking Project (CSDP; King et al., 2011; 2016; Roche et al., 2014). Among the heavy drinkers (HD) and light drinkers (LD) who participated in Cohort 1, 156 completed a five-year re-examination between 2009 and 2011. During the same time interval, a second cohort of 104 HD was engaged in an identical laboratory protocol (see Roche et al., 2014, for details). For both cohorts, the impulsivity and sensation seeking questionnaires were given as part of a take-home packet completed outside the laboratory sessions. Participants in Cohort 1 completed the impulsivity questionnaire at the 5-year re-examination as the measure was not administered at the initial testing session. Participants in Cohort 2 completed the impulsivity questionnaire at the initial testing session.

Across both cohorts, heavy drinkers were non-alcohol dependent (DSM-IV) individuals who reported consuming between 10 and 40 standard alcoholic drinks per week and engaged in regular heavy episodic drinking (i.e., heavy drinking on 1–5 occasions per week, on average, as their predominant adult pattern). Cohort 1 also included light drinkers who were also non-alcohol dependent drinkers but self-reported consuming between 1–5 standard drinks per week with infrequent heavy drinking (i.e., 5 or fewer times in the past year, but allowing for one past six- month interval of up to twice weekly heavy drinking). Eligibility was determined using the alcohol and other substance dependence modules of the Structured Clinical Interview for DSM-IV – non-patient version (SCID-IV; First, Spitzer, Gibbon, & Williams, 2002) and the Timeline Follow Back (TLFB; Sobell & Sobell, 1992). Individuals with a past or current major medical or psychiatric condition including alcohol or other substance dependence, with the exception of nicotine dependence, were excluded.

Procedure

The study used a within-subjects, double-blind, placebo-controlled laboratory design. Participants received an ethanol dose (0.8 grams per kilogram [g/kg] of body weight) or placebo beverage in counterbalanced order. The alcohol dose consisted of 190-proof ethanol and the placebo beverage contained 1% ethanol as a taste mask. Each beverage also contained a flavored drink mix and sucralose-based sugar substitute to add flavor and mask the taste of alcohol. Adjustments were made for women to receive an approximate 85% dose of ethanol compared to men to account for differences in total body water affecting blood alcohol concentrations (Frezza et al., 1990; Sutker, Tabakoff, Goist, & Randall, 1983). Participants were excluded at baseline if they weighed more than 210 pounds or less than 110 pounds. Because some participants may have gained weight during the interval between initial testing and the 5-year re-examination, participants who weighed up to 250 pounds at the 5-year re-examination were retained.

To reduce expectancies, the Alternative Substance Paradigm was employed (Conrad, McNamara, & King, 2012). As such, the participant was informed that the beverage may contain a stimulant, sedative, alcohol, or placebo, at varying doses, and/or two substances in combination. Manipulation checks of the Alternative Substance Paradigm support that the deception was effective (Brumback, McNamara, Cao, & King, 2017; Conrad et al., 2012). Specifically, in a secondary analysis of data from the CSDP, Brumback and colleagues (2017) found that over one-third of heavy and light drinkers were not able to correctly identify alcohol as the only active ingredient in the alcohol session or likewise identify placebo. Further, King and colleagues (2011; 2016) found no significant effects of dose order on subjective responses.

Each participant attended two 4–5 hour laboratory sessions separated by at least 48 hours. Prior to each session, the participant was instructed to abstain from alcohol and medications for at least 48 hours, and caffeine, cigarettes, and food for 3 hours. Upon arrival, s/he consumed a low-fat snack (20% of daily calories) to avoid hunger effects on mood state and to reduce the possibility of alcohol-induced nausea. Thirty minutes later, each participant completed baseline subjective and objective measures as part of the larger CSDP. Next, s/he consumed the assigned beverage out of a clear cup and through a straw over a 15-minute period in the presence of a research assistant. The beverage was divided into two equal portions that were consumed over a 5-minute interval with a 5-minute rest period between portions. After beverage consumption, measures were administered at repeated time points over the remaining three hours. Between time points, the participant was allowed to relax and watch television or movies and/or read magazines provided by the study. After the session was completed, the study provided transportation to the participant’s home address. This study was approved by the University of Chicago Internal Review Board.

Measures

Barratt Impulsivity Scale (BIS-11; Patton et al.,1995). The BIS-11 is 30-item Likert-type scale (1 = Rarely/Never to 4 = Almost Always/Always) that assesses self-reported impulsivity with scores ranging from 30 (lowest impulsivity) to 120 (highest impulsivity). The BIS-11 has been found to be a reliable and valid measure (Patton et al., 1995). Similar to prior studies investigating relationships between impulsivity and SR to alcohol (Berey et al., 2017; Leeman et al., 2014), the main dependent variable was the BIS-11 total score.

Sensation Seeking Scale (SSS Form V; Zuckerman et al., 1978; Zuckerman, 1996). The SSS is a reliable and valid dichotomous 40-item scale used to assess multiple dimensions of sensation seeking including aspects of thrill and adventure seeking, experience seeking, disinhibition, and boredom susceptibility. As with the BIS-11, SSS total score was used in these analyses.

Biphasic Alcohol Effects Questionnaire (BAES; Martin, Earleywine, Musty, Perrine, & Swift, 1993). The BAES is a reliable and valid 14-item unipolar rating scale (1–10) used to assess the extent each participant feels the stimulating (e.g., stimulated, talkative, up, etc.) and sedating (e.g., sedated, sluggish, inactive, etc.) effects of alcohol (Martin et al., 1993). The current study modified the typical BAES instructions by not divulging beverage contents and including baseline assessments (i.e., pre-beverage administration) so that changes in stimulant and sedative SR could be assessed (Rueger, McNamara, & King, 2009). This method of BAES administration has been found to be reliable and valid (Rueger et al., 2009).

Drug Effects Questionnaire (DEQ; Johanson & Uhlenhuth, 1980). The DEQ is a postbeverage visual analog rating scale and valid measure of the effects of alcohol (Morean et al., 2013). The main dependent variables derived from this measure were like (e.g., do you like the effects you are feeling now?), and want more (e.g., would you like more of what you just consumed?).

Family History-Research Diagnostic Criteria (FH-RDC; Andreasen, Endicott, Spitzer, & Winokur, 1977). The FH-RDC is a valid, reliable two-generational biological family tree measure that assesses family history of AUD (Andreasen et al., 1977). Participants indicated which family members had alcohol use problems, including the participant’s paternal and maternal grandparents, aunts, and uncles and the participant’s parents and siblings. Follow-up questions concerning the specific type of problems related to alcohol use (e.g., legal, health, marital/family, occupational, other [e.g., fighting, losing friends], and receiving treatment) were asked about all family members who were identified as having alcohol problems. Family history negative (35%) individuals were defined as those with no history of AUD within the past two generations. Family history positive (45%) were those with at least one primary biological relative with an AUD within the past two generations. Participants who were adopted or unable to provide sufficient information were identified as family history unknown (20%).

Data Analytic Plan

Normality was assessed for continuous variables. Because self-reported impulsivity and sensation seeking were treated as continuous, these variables were mean-centered by subtracting the grand mean from each individual’s total score to reduce concerns of multicollinearity resulting from creating interaction terms (Aiken, West, & Reno, 1991).The main dependent variables were difference scores for BAES stimulant and sedative SR and DEQ like and want, calculated by subtracting the ratings at each time point for the placebo session from the alcohol session (i.e., at 0, +30, +60, +120, +180 mins.). Because the DEQ was administered only after alcohol administration, difference scores for the DEQ like and want variables did not include a zero time point. This subtraction procedure eliminated the need for a variable capturing beverage condition, as a main effect and component of interaction terms, thus enhancing statistical power.

Linear mixed effects (LME) models were used to examine relationships between self-reported impulsivity and SR and between the interactive effects of self-reported impulsivity and sensation seeking on SR. Random intercept and slope models and multiple covariance structures were considered for each of the four dependent variables (i.e., subjective stimulation, sedation, liking and wanting more alcohol). Maximum likelihood estimation was used for all models, permitting the fit of nested models to be compared using the Akaike Information Criterion (AIC). Preliminary analyses were conducted using SPSS, Version 25. LME models were tested using SAS, Version 9.4.

Preliminary models were tested including participant sex, family history, and age, however these variables were dropped for parsimony as they did not affect the statistical significance of the key variables of interest (i.e., heavy or light drinker group, self-reported impulsivity and time). The same approach was used for the Cohort 2 data with the exception that there was no drinker group distinction.

Following these preliminary models, Cohort 1 models included three predictor variables: (1) light versus heavy drinker group; (2) self-reported impulsivity and (3) time (five time points for BAES subscales: 0, +30, +60, +120, +180 min. and four time points for DEQ items as there is no baseline measure: +30, +60, +120, +180 min). Based on expected heterogeneity in stimulant and sedative SR across time points, the possibility of non-linear change in stimulant and sedative SR over time was addressed by testing quadratic and linear effects. All two-way interactions were tested along with three-way interactions of drinker group by impulsivity by time (linear or quadratic). Non-significant interactions were eliminated, leaving only main effects and significant interactions in final models reported here.

Following the main analyses, additional models were evaluated to test impulsivity by sensation seeking interactions as predictors of SR variables. Models to test both impulsivity and sensation seeking in relation to SR were conducted somewhat differently. In order to avoid testing complicated four-way interactions, for which statistical power would have been limited, we first verified whether any effects of impulsivity on SR applied across drinker group. Provided effects applied across drinker group, we planned to drop drinker group from the impulsivity and sensation seeking models. With this approach, a three-way interaction (time x impulsivity x sensation seeking) remained the most complicated term in our models.

Results

Participants

Overall, participants in Cohorts 1 and 2 were primarily male (53% and 63%, respectively) and Caucasian (76% and 77%, respectively). However, participants in Cohorts 1 and 2 differed on several sociodemographic, drinking, and impulsivity characteristics (see Table 1). A higher percentage of heavy drinkers in Cohorts 1 and 2 reported a family history of alcohol problems (44% and 45%, respectively) relative to light drinkers (34%). Further, Cohort 2 participants were significantly younger due to our use of only 5-year re-evaluation data for Cohort 1. The light drinkers had more education than the heavy drinkers from both cohorts. Drinking levels for all variables were significantly higher among the heavy drinkers in Cohort 2 versus Cohort 1, and heavy drinkers in both Cohorts had significantly higher drinking levels than the light drinkers in Cohort 1. Cohort 1 heavy drinkers had significantly higher mean BIS-11 scores than Cohort 1 light drinkers. Cohort 2 heavy drinkers had the highest mean BIS-11 scores across all groups, which approximated the normative adult score of 62 reported by Stanford and colleagues (2009). Heavy drinkers in both Cohorts had significantly higher sensation seeking mean scores than light drinkers, but heavy drinkers in Cohorts 1 and 2 did not significantly differ in their sensation seeking scores.

Table 1.

Sample Descriptives

	Cohort 1		Cohort 2
General Characteristics	Light Drinkers (n = 70)	Heavy Drinkers (n = 86)	Heavy Drinkers (n = 104)
Age, in years,	31.1 (3.5), 26 – 40a	30.2 (2.9), 26 – 39a	24.9 (2.3), 21 – 29b
Education level, in years,	16.5 (2.1), 12 – 22a	15.7 (1.5), 12 – 20b	15.6 (1.5), 12 – 20b
Family History of Alcohol Use Disorder	24 (34%)	38 (44%)	47 (45%)
Caucasian	47 (67%)	71 (83%)	80 (77%)
Male sex	34 (49%)	49 (57%)	65 (63%)
BIS-11 (self-reported impulsivity) total score	53.1 (10.4), 35 – 74a	58.6 (9), 41 –83b	63.3 (8.8), 43 – 82c
SSS-V total score	19.1 (6), 4–33a	24.7 (5.6) 10–35b	25.6 (5) 12–38b
Alcohol-related variables: past month drinking
Drinking days per month	7.9 (5.5), 0 – 28a	12.4 (6.7), 1–28b	14.8 (4.7), 6 – 28c
Standard drinks per drinking day	1.9 (0.7), 0 – 4.33a	4.8 (2.3), 1 – 12.5b	6.1 (2.8), 2.2 – 16.9c
Heavy drinking days	0.54 (1.1), 0 – 7a	5.6 (4.1), 0 – 23b	8.9 (2.9), 4 – 17c

Note: Data are M (SD), range or No. (%), as appropriate. Drinker groups denoted with different superscript letters differed significantly at p < . 05. Self-reported impulsivity was measured using the Barratt Impulsiveness Scale, version 11 (Patton et al., 1995). The possible range of scores on the BIS-11 is 30–120. Sensation Seeking was measured using the Sensation Seeking Scale – Form V (Zuckerman et al., 1978; Zuckerman, 1996). The possible range of scores on the SSS-V is 0–40.

Predicting stimulant SR

In Cohort 1, there was a statistically significant three-way interaction of drinker group by impulsivity by quadratic time (Table 2). To illustrate the significant three-way interaction, BIS-11 scores were divided by quartile and stimulant SR was plotted across time points among high and low impulsive (i.e., the highest and lowest quartiles), heavy and light drinkers (Figure 1). Based on this plot, low impulsive, light drinkers reported less stimulant SR than high impulsive, light drinkers and heavy drinkers regardless of impulsivity level. Post-hoc LMEs were tested to confirm this observation and determine whether this interaction applied on the ascending limb (+30 time point), on the descending limb (+120 time point), or both. A median split was used to divide the sample by impulsivity level for these post-hoc analyses. Low impulsive, light drinkers were compared with all other participants and interactions were tested with time. The quadratic time by low impulsive, light drinker versus others interaction was statistically significant (β = 1.96, SE = .46, p < .001). Two further LMEs were tested to compare stimulant SR at the +30 time point and the +120 time point with baseline levels. The time by low impulsive, light drinker versus others interaction was statistically significant for both comparisons (+30 time point: β = − 4.39, SE = 1.38, p = .002; +120 time point: β = 3.57, SE = 1.04, p = .001).

Table 2.

Final linear mixed effects models of stimulant subjective response (alcohol minus placebo) from laboratory alcohol administration sessions during five-year reexamination among participants in Cohort 1 of the Chicago Social Drinking Project

Effect	β	SE	p value
Stimulant response final model
Linear Time (five time points: +0 through +180)	5.32	1.30	<.0001
Quadratic Time	−1.42	0.28	<.0001
Group (heavy versus light drinker)	−1.25	2.09	0.548
BIS-11 (self-reported impulsivity) score at year 5	0.17	0.15	0.246
Quadratic time x impulsivity x group	−0.12	0.04	0.003

Notes: BIS-11 scores were grand mean-centered. All lower-order interactions were included in the model, however were omitted from the table for parsimony. The final model included random intercept and slope effects and utilized a heterogeneous autoregressive covariance structure. Self-reported impulsivity was measured using the Barratt Impulsiveness Scale, version 11 (Patton et al., 1995).

Figure 1.

Stimulant subjective response (alcohol minus placebo) over five time points in laboratory alcohol administration sessions at five-year re-examination phase among participants whose self-reported impulsivity (BIS-11) score is in the highest or lowest quartile in the sample. Data are shown for light (LD; n = 70) and heavy drinker (HD; n = 86) groups.

We then tested models to address whether participants reporting both high impulsivity and sensation seeking were particularly likely to report elevated stimulant SR and whether relationships between self-reported impulsivity and stimulant SR were unique to impulsivity or also applicable to sensation seeking. Because the disinhibition subscale of the SSS is related to impulsivity, we examined bivariate correlations among the BIS-11 and the SSS with and without the disinhibition subscale among participants in both cohorts. There was a significant positive correlation between the BIS-11 and SSS among participants in both cohorts with (Cohort 1: r=.38, p<.01; Cohort 2: r=.25, p<.01) and without the disinhibition subscale (Cohort 1: r=.33, p<.01; Cohort 2: r=.23, p<.05), confirming that removing the disinhibition subscale of the SSS did not affect the relationship between impulsivity and sensation seeking substantially. Further, when both impulsivity and sensation seeking main effects were included in a model to predict stimulant effects without any interaction terms, impulsivity (β = 0.29, SE = .08, p = .001) was a significant predictor of stimulant effects but sensation seeking was not (β = 0.22, SE = .14, p = .122).

Subsequently, we confirmed that a significant interaction of impulsivity by quadratic time for stimulant SR was still observable when drinker groups were combined (β = −0.09, SE = .02, p < .001). Having made this confirmation, we combined drinker groups for the purposes of these models including both impulsivity and sensation seeking. The three-way interaction of impulsivity by sensation seeking by time was not statistically significant (β = 0.002, SE = .003, p = .556). A subsequent model with the three-way interaction removed showed that an impulsivity by quadratic time interaction predicted stimulant SR (β = −0.07, SE = .02, p = .002) but sensation seeking by quadratic time did not (β = −0.06, SE = .04, p = .095).

Similar to findings among heavy drinkers in Cohort 1, more impulsive heavy drinkers in Cohort 2 did not show heightened stimulant SR across time (β = −0.02, SE = .08, p = .826).

Predicting sedative SR

In contrast with stimulant SR, for sedative SR, the three-way interaction of drinker group by impulsivity by time (neither linear nor quadratic) was not significant (β = 0.02, SE = .04, p = .613). Further, there were no significant effects of any kind involving impulsivity (main effect or two-way interactions). The best fit model for sedative SR included a significant drinker group by quadratic time effect (β = −0.97, SE = .43, p = .027), indicating dampened sedative SR among heavy drinkers across time, as previously reported by King et al. (2016). An illustrative plot of sedative SR among participants with the highest and lowest BIS-11 scores can be found in Figure 2.

Figure 2.

Sedative subjective response (alcohol minus placebo) over five time points in laboratory alcohol administration sessions at five-year re-examination phase among participants whose self-reported impulsivity (BIS-11) score is in the highest or lowest quartile in the sample. Data are shown for light (LD; n = 70) and heavy drinker (HD; n = 86) groups.

Similarly, among heavy drinkers in Cohort 2, there were no significant effects involving impulsivity (main effect or two-way interactions).

Predicting Alcohol Reward

In Cohort 1, there was a statistically significant main effect whereby greater impulsivity predicted DEQ like ratings (β = 0.41, SE = .17, p = .015), such that more impulsive participants reported greater alcohol liking regardless of group and time. However, no interactions involving impulsivity, time and group significantly predicted DEQ Like ratings. As previously reported by King et al. (2011), heavy drinkers endorsed liking the effects of the alcohol more than light drinkers, β = −10.61, SE = 3.35, p = .002. Among participants in Cohort 2, no effects involving impulsivity (main effect or any interactions) significantly predicted DEQ Like ratings.

In Cohort 1, impulsivity was not related significantly to DEQ want more ratings, β = 0.49, SE = .26, p = .062. As previously reported by King et al. (2011), heavy drinkers in cohort one reported significantly higher DEQ want more ratings relative to light drinkers, β = −20.04, SE = 5.18, p = <.001. Diverging from Cohort 1 results, there was a significant main effect whereby more impulsive heavy drinkers in Cohort 2 reported significantly higher DEQ want more ratings (β = .79, SE = .31, p =.01). There were no significant interactions involving impulsivity in these models though.

Discussion

Results from the current study extend prior research relating impulsivity to SR. Similar to Leeman and colleagues’ (2014) findings, self-reported impulsivity on the BIS-11 (Patton et al., 1995) was related to elevated stimulant SR in light drinkers. Subsequent analyses showed that this relationship was driven by high impulsive, light drinkers reporting stimulant SR of a magnitude similar to heavy drinkers.

Unlike prior results (Berey et al., 2017; Leeman et al., 2014), self-reported impulsivity on the BIS-11 (Patton et al., 1995) was not related to dampened sedative SR among light drinkers. Further, and contrary to study hypotheses, more impulsive heavy drinkers did not experience heightened stimulant or dampened sedative SR compared to less impulsive heavy drinkers following high-dose alcohol in either cohort. Null results regarding stimulant and sedative SR among Cohort 2 heavy drinkers occurred despite significantly higher levels of impulsivity than Cohort 1 light and heavy drinkers. It is possible that relationships between impulsivity and SR among heavy drinkers are not captured well by self-reported impulsivity and are instead characterized best by behavioral facets of impulsivity. Supporting this possibility, prior results suggest a link between greater delay discounting and elevated stimulant SR among alcohol dependent individuals (Westman et al., 2016) and between response inhibition difficulties and stimulant SR in a sample including heavy drinking young adults (Quinn & Fromme, 2016). It is important to note that overall, self-reported impulsivity levels were not inordinately high, even among the younger, heavy drinkers in Cohort 2. Perhaps among heavy drinkers, impulsive tendencies must reach clinically notable levels in order to be linked to elevated stimulant SR, as suggested by findings from Hendershot et al. (2015). Alternatively, affect-laden impulsivity (i.e., urgency), which is captured by the UPPS Impulsive Behavior Scale (Whiteside & Lynam, 2001; Cyders et al., 2007) to a greater extent than the BIS-11 (Patton et al., 1995), may be most relevant to high-risk SR patterns (Wardell et al., 2015). Notably, Wardell et al. (2015) found that cross-sectionally, retrospective reports of dampened sedative SR correlated significantly with positive urgency on the UPPS (Whiteside & Lynam, 2001; Cyders et al., 2007). Relationships between urgency and high-risk SR patterns should be addressed in future laboratory alcohol administration studies.

Based on prior findings that co-occurring impulsivity and sensation seeking is particularly problematic (Ersche et al., 2010; 2013), we addressed the possibility that high levels of both impulsivity and sensation seeking would show the strongest relationships to high-risk SR patterns. This was not confirmed as we found no significant interaction between impulsivity and sensation seeking in predicting stimulant SR. Further, we demonstrated that elevated stimulant SR had a stronger relationship to self-reported impulsivity than sensation seeking in the Cohort 1 sample.

This pattern of findings follows a recent trend in the literature in which significant prediction of stimulant SR has been reported by a variety of factors more frequently than significant prediction of blunted sedative SR. Examples of this trend include the aforementioned studies concerning ADHD symptoms (Hendershot et al., 2015) and response inhibition (Quinn & Fromme, 2016); a study involving trait sensitivity to reward and punishment (Morris, Treloar, Tsai, McCarty, & McCarthy, 2016); an investigation concerning rate of alcohol metabolism (Boyd & Corbin, 2018), and genetic studies investigating single nucleotide polymorphisms in GABRA2 (Arias et al., 2013). Although King and colleagues (2002; 2011) demonstrated that heavy versus light drinkers exhibited significantly lower sedative SR, identifying key factors related to sedative SR among heavy drinkers has been challenging. It may be that examining other facets of impulsivity (e.g., delay discounting; urgency) or combinations of facets in relation to SR is the best approach to determining whether impulsivity relates to dampened sedative SR.

In addition to stimulant SR, we examined questions regarding liking and wanting more alcohol as additional measures of alcohol reward. The present findings suggested that more impulsive individuals may like and want more alcohol, but this was a general tendency, not tied to the timing of alcohol administration and seemingly not related to distinctions between light and heavy drinkers. Further, these findings must be interpreted with caution as they were not replicated in Cohort 2. Nonetheless, findings involving liking and wanting more alcohol did not closely parallel findings regarding stimulant SR in Cohort 1. This suggests that the more impulsive light drinkers in this study did not necessarily experience stimulant SR as subjectively rewarding. As such, these results may provide evidence as to why more impulsive light drinkers do not progress to frequent heavy drinking.

In addition to an apparent lack of connection between stimulant response and hedonic effects among impulsive light drinkers, sedative SR was particularly pronounced among light drinkers overall. The sedation subscale of the BAES (Martin et al., 1993) measures alcohol effects typically viewed as aversive, and King and colleagues (2011; 2016) provided evidence that heightened sedative SR is associated with less heavy drinking and AUD symptomology over time. Perhaps a combination of not experiencing stimulant SR as pleasurable and elevated sedative SR among light drinkers was protective for more impulsive light drinkers in this study.

Overall, examining relationships between impulsivity and SR has clinical and public health implications. Impulsive individuals are at greater risk for heavy drinking and AUD (Coskunipar & Cyders, 2013; Stanford et al., 2009) and it is critically important to learn more about why. The present findings suggest a relationship between self-reported impulsivity and elevated stimulant response to alcohol, however only among light drinkers. As such, higher levels of impulsivity and alcohol stimulation among young adult light drinkers may help to identify those who are more likely progress into heavier drinking levels during middle and older age. While a minority of adults initiate heavy drinking during their late 20s and early 30s (e.g., Dom, D’haene, Hulstijn, & Sabbe, 2006), factors influencing this specific trajectory are less understood.

Moreover, considering these results along with prior findings (e.g., Hendershot et al., 2015; Leeman et al., 2014; Westman et al., 2016) leads to several future research directions concerning heavy drinkers. For example, there is value in determining whether high-risk SR patterns relate to self-reported impulsivity with a stronger affective component (urgency), to response and/or choice impulsivity measurable via cognitive tasks, or to higher/clinical levels of impulsivity. Discovering any of these linkages in future studies could point to intervention targets (reducing rewarding, stimulant and/or increasing aversive, sedative effects of alcohol) among subsets of heavy drinkers who score highly on these measures of impulsivity. Such findings are crucial given that impulsive individuals are often resistant to currently available interventions (Helstrom, Hutchison, & Bryan, 2007; Stevens et al., 2014).

This study is not without limitations. Although this study utilized an experimental design, we could not determine temporal precedence due to the cross-sectional nature of the data. Moving forward, longitudinal research will be needed to evaluate relationships between impulsivity and SR among adolescent and young adult drinkers. Specifically, it will be important to determine whether young impulsive individuals who are initially light drinkers experience stronger stimulant but less rewarding effects from alcohol and if these patterns change over time. While impossible to ascertain from the current study, relationships between stimulant SR and the type of self-reported impulsivity captured by the BIS-11 may manifest early on, but attenuate following the onset of heavy drinking. Longitudinal research is necessary to help identify whether impulsive individuals who go on to heavy drinking initially experienced heightened stimulant and/or dampened sedative SR and how these patterns change over time. Second, the parent study only contained one measure of self-reported impulsivity and one measure of a related construct (i.e., sensation seeking). Because impulsivity is a complex, multifaceted construct, future research should investigate multiple facets of impulsivity in relation to SR. Doing so will be critical to understand fully relationships among impulsivity, SR, and negative outcomes including the development of AUD.

Taken together, the current study partially replicates and extends previous research linking self-reported impulsivity to SR among light and heavy young adult drinkers. Future research must incorporate multiple facets of impulsivity to determine which specific facets convey differential risk for negative alcohol-related outcomes. Moreover, longitudinal study designs are needed to investigate whether links between impulsivity and SR (particularly stimulant SR) are observable early on in adolescence and young adulthood, along with what outcomes this linkage predicts and whether relationships between impulsivity and SR continue or dissipate over time.