Neural response to threat and reward among young adults at risk for alcohol use disorder

Abstract

Alcohol use disorder (AUD) is heritable. Thus, young adults with positive family histories represent an at‐risk group relative to those without a family history, and if studied at a time when both groups have similar levels of alcohol use, it provides an opportunity to identify neural processing patterns associated with risk for AUD. Previous studies have shown that diminished response to potential reward is associated with genetic risk for AUD, but it is unclear how threat may modulate this response. We used a modified Monetary Incentive Delay task during fMRI to examine neural correlates of the interaction between threat and reward anticipation in a sample of young adults with (n = 31) and without (n = 44) family histories of harmful alcohol use. We found an interaction (p = 0.048) between cue and group in the right nucleus accumbens where the family history positive group showed less differentiation to the anticipation of gaining $5 and losing $5 relative to gaining $0. The family history‐positive group also reported less excitement for trials to gain $5 relative to gaining $0 (p < 0.001). Family history‐positive individuals showed less activation in the left insula during both safe and threat blocks compared to family history‐negative individuals (p = 0.005), but the groups did not differ as a function of threat (p > 0.70). Young adults with, relative to without, enriched risk for AUD may have diminished reward processing via both neural and behavioural markers to potential rewarding and negative consequences. Neural response to threat may not be a contributing factor to risk at this stage.

We used a modified monetary incentive delay task to examine neural correlates of the interaction between threat and reward anticipation in a sample of young adults with (n = 31) and without (n = 44) family histories of harmful alcohol use. We found an interaction (p = 0.048) between cue and group in the right nucleus accumbens where the family history positive group showed less differentiation to the anticipation of gaining $5 and losing $5 relative to gaining $0.

1. INTRODUCTION

Alcohol use contributes to 3.8% of global deaths ¹ and therefore represents a major public health challenge. Identifying at‐risk individuals prior to the development of alcohol use disorder (AUD) could allow for early intervention. Since AUD is approximately 50% heritable, ² , ³ enriched risk via genetic predisposition presents an opportunity to study biological markers that indicate the likelihood of developing AUD. Common motives for alcohol use include enhancement of positive mood and coping with negative emotions, ⁴ and individual differences in sensitivity to these motives may indicate risk for harmful alcohol use. ⁵ Thus, one possibility is that young adults with enriched risk for AUD due to family history may display altered neurobiological response to potential positive or negative outcomes, reflecting sensitivity to enhancement or coping motives, and these altered positive and negative responses may interact.

One method to assess such neural differences is to examine brain activation during MRI while individuals complete a task involving potential positive (e.g., winning money) or negative outcomes (e.g., losing money or being startled). Neural response to reward‐based tasks occurs in brain regions that tend to be overlapping, including the ventral striatum, cingulate cortex and insula. ⁶ One of the brain regions most commonly associated with the anticipation of reward is the nucleus accumbens, ⁷ whereas the medial prefrontal cortex (mPFC) responds to the actual receipt of monetary reward. ⁸ , ⁹ , ¹⁰ , ¹¹ A region often associated with anticipation of negative outcomes is the anterior insula. ¹² , ¹³ Numerous studies have examined the neural response of adults with AUD during anticipation of winning money, ¹⁴ , ¹⁵ and a meta‐analysis of these studies indicated consistent reductions in activation of the nucleus accumbens among adults with AUD relative to healthy adults. ¹¹ Adults with greater levels of harmful alcohol use showed enhanced startle response during a task with a potentially negative outcome (i.e., a shock) ¹⁶ and adults with, relative to without, AUD showed increased activation in the anterior insula during the same task. ¹⁷ As these studies show, adults with AUD demonstrate altered neural responses to potential positive and negative outcomes, raising the possibility that these neural alterations may be present prior to the onset of AUD.

Several studies have investigated whether these altered neural responses may precede heavy drinking by studying young individuals with enriched risk via family history. For example, a study of children and adolescents showed that those with, relative to without, a family history of harmful alcohol use showed reduced activation in the nucleus accumbens when anticipating potential monetary rewards. ¹⁸ Similarly, greater synchrony of reward circuitry (including the accumbens) during early adolescence was a predictor of future risky drinking for females in a large longitudinal study. ¹⁹ Biological response to a threat of shock (i.e., strength of eye blink reflex) was stronger in young adults with, relative to without, a family history of AUD. ²⁰ There is therefore evidence that neural response to potential positive and negative outcomes may be associated with a family history of AUD.

Several aspects remain unexplored in relation to risk for AUD, given that evidence from healthy young adults shows that potential negative outcomes can modulate neural response to reward during modified monetary incentive delay tasks. For example, in a study examining testing whether threat and reward responses compete or add together, healthy young adults showed greater activation of the nucleus accumbens and insula during anticipation of a no‐reward ($0) condition in the presence of cues indicating a threat of a potential shock, ²¹ suggesting these processes may compete. In a similar study, nucleus accumbens activation was greater during the anticipation of winning money in the presence, relative to the absence of the threat of electric shock. ²² Effects of threat were also evident in the outcome phase of the task, where participants randomized to view aversive movie clips (e.g., violence) prior to completing the monetary incentive delay task, showed decreased mPFC activation when receiving money relative to participants randomized to view neutral movie clips (e.g. peaceful social interactions). ²³ Given that risk for AUD is associated with altered response to potential positive and negative outcomes, and that both reward‐seeking and stress are motives for alcohol consumption in young adults, it is possible that the interaction of these anticipatory states is altered in individuals with family history of AUD, but this hypothesis remains untested.

The goal of this study is to determine whether young adults with enriched risk for AUD show altered neural responses to the interactive effects of potential negative and positive outcomes (e.g., being scared while winning money). Since heavy drinking could confound the response, our goal was to study an enriched risk group and a control group who currently had similar, low levels of alcohol use. However, we also wanted the risk of increasing alcohol use to be proximal, so that some of the individuals may increase levels of alcohol use within a year, as the current study was part of a larger parent study that was longitudinal in nature to identify neural predictors of future use. This would be unlikely if the sample was too young, so we aimed to study young adults aged 18–22 in the parent study. Neural response to reward anticipation has been well‐characterized using the Monetary Incentive Delay (MID) task (see ¹¹ , ²⁴ , ²⁵ , ²⁶ , ²⁷ ), and our study used the addition of sustained threat of a scream (MID‐Scream; see ²⁸ ) to examine the interaction of anticipation of threat and anticipation of reward. We hypothesized that the family history positive group versus the family history negative group would have less activation in the nucleus accumbens during the anticipation of monetary reward, and less activation in the mPFC during receipt of money; threat would diminish both responses. We also anticipated that the family history positive, relative to the family history negative, group would sustain higher activation levels in the insula during the anticipation of threat.

2. MATERIALS AND METHODS

2.1. Procedures

Participants were recruited through advertisements for a study (approved by the Colorado Multiple Institutional Review Board) examining neural correlates of parental and sibling harmful alcohol use. Participants were screened over the phone to determine eligibility prior to enrollment. Eligibility criteria were ¹ being able to undergo MRI, ² aged 18–22 (one individual turned 23 prior to the MRI scan), ³ using alcohol ⁴ scoring less than seven on the Alcohol Use Disorder Identification Test (indicating non‐hazardous drinking ²⁹ ), ⁵ not using medications affecting the hemodynamic response, ⁶ not seeking treatment for an alcohol use disorder, ⁷ not being treated for a psychiatric disorder, ⁸ scored less than 11 on the Cannabis Use Disorder Identification Test‐Revised ³⁰ or less than eight uses of cannabis per month, ⁹ not using tobacco regularly (< 20 cigarettes per week), ¹⁰ less than 10 lifetime uses of illicit drugs, ¹¹ no prescription medication misuse in the past 12 months, ¹² no past head trauma with unconsciousness lasting more than 10 minutes and ¹³ not pregnant. The objective of the substance use criteria (e.g., <8 uses of cannabis per month) was to reduce the likelihood that participants had heavy use patterns (e.g., more than 2 uses/week) prior to enrollment. However, some participants who reported low levels of use over the phone were found to have a current or past alcohol or cannabis use disorder when they completed a diagnostic interview (see Table 1).

TABLE 1.

Participant characteristics and substance use by family history group.

	FH‐ (n = 44)	FH+ (n = 31)
Variable	n (%)		Test statistic
Sex
Male	26 (59.09)	7 (22.58)	χ 2 (1) = 8.41
Female	18 (40.91)	24 (77.42)
Race
American Indian/Alaska Native	0 (0.00)	1 (3.23)	Fisher's p = .12
Asian	4 (9.09)	1 (3.23)
Black/African American	2 (4.55)	1 (3.23)
White	30 (68.18)	27 (87.10)
Biracial	6 (13.64)	0 (0.00)
No Response/“Unknown”	2 (4.54)	1 (3.23)
Ethnicity
Hispanic	11 (25.00)	7 (22.58)	Fisher's p = 1.00
Non‐Hispanic	32 (72.73)	24 (77.42)
No Response/“Unknown”	1 (2.27)	0 (0.00)
Alcohol Use Disorder
Past	1 (2.27)	3 (9.68)	Fisher's p = .300
Current	0 (0.00)	1 (3.23)	Fisher's p = .413
Cannabis Use Disorder
Past	1 (2.27)	1 (3.23)	Fisher's p = 1.00
Current	1 (2.27)	1 (3.23)	Fisher's p = 1.00
Attention Deficit Hyperactivity Disorder	0 (0.00)	1 (3.23)	Fisher's p = .413
High Motion in Scanner	3 (4.00)	2 (2.67)	Fisher's p = 1.00

	M (SD)		Test Statistic
Depression	4.09 (4.63)	6.45 (7.55)	t(46) = 1.55
Anxiety
State Anxiety	30.64 (7.94)	37.35 (8.77)	t(73) = 3.46
Trait Anxiety	33.34 (8.79)	41.10 (9.69)	t(73) = 3.61
Current Age	20.96 (1.37)	21.02 (1.33)	t(73) = 0.20
Alcohol Use Disorder Identification Test	4.61 (2.28)	4.55 (2.59)	t(73) = 0.12

	Median (IQR)		Test Statistic
Percent Days Drank Alcohol in Past Year a	10 (13.75)	7 (15.00)	W = 610.50
Number of Binge Episodes in Past Year b	2.50 (9.00)	2.00 (4.50)	W = 708.00
Total Drinks on Average Drinking Occasion c	3 (2.00)	3 (1.00)	W = 657.00
Lifetime cigarettes	0 (9.00)	0 (9.00)	W = 582.00
Percent of Scan Data Censored (% TRs) d	3.82 (6.91)	4.12 (6.99)	W = 675.00

Note: Bold font indicates significance at p < .05.

Abbreviations: FH+ = family history positive, FH‐ = family history negative, IQR = interquartile range.

^a FH + ranged from 1–75%, and FH‐ ranged from 0–60%.

^b FH + ranged from 0–50, FH‐ ranged from 0–50.

^c FH + ranged from 0–24, FH‐ ranged from 0–14.

^d FH + ranged from 0–21.76, and FH‐ ranged from 0–36.06%.

Eligible participants were invited to participate, and they provided informed consent. They reported their current age, sex, race, ethnicity and alcohol consumption behaviours. Participants were administered a semi‐structured interview to assess psychiatric diagnoses by trained clinicians (Structured Clinical Interview for DSM‐5 Disorders Reserach Version ³¹ ). We created dichotomous variables for both past and current alcohol use disorder and cannabis use disorder.

At the beginning of the study visit, participants completed a second MRI safety screener to ensure MRI eligibility, an alcohol breathalyser and a saliva drug screener, and females took a urine pregnancy test. Study data were collected from March 2019 through June 2021 and managed using REDCap electronic data capture tools ³² hosted at the University of Colorado. REDCap is a secure, web‐based application designed to support data capture for research studies, providing: 1) an intuitive interface for validated data entry; 2) audit trails for tracking data manipulation and export procedures; 3) automated export procedures for seamless data downloads to common statistical packages; and 4) procedures for importing data from external sources.

2.2. Measures

2.2.1. MID‐scream task

During the MRI scan, participants completed the MID‐Scream task that was designed to examine anticipation of threat (e.g., sustained anticipation of aversive stimuli) modulated neural response to potential monetary rewards; full details of this task are described elsewhere. ²⁸ Briefly, during the anticipation phase of the task, participants were shown one of three cues indicating the opportunity to gain $5, avoid losing $5, or gain $0 (no change in earnings). The screen had a coloured border indicating whether the trial was in the safe condition (yellow) or threat condition (blue), where the threat condition indicates the possibility of hearing a scream and seeing a scary face. After being presented with the cue and condition combination, participants were briefly shown a target and were instructed to press a button as quickly as possible. During the outcome phase of the task, participants were provided feedback on their responses to the target during each trial. Successful responding during the circle cue results in gaining $5, while unsuccessful responding during the triangle cue results in losing $5. Success or failure during the square cue results in no change in earnings (i.e., gain $0); see Figure 1. A trial includes the anticipation plus outcome phase. Safe and threat conditions occurred in blocks of nine consecutive trials. There were six blocks per run. There were two runs of the task. The blocks alternated between safe and threat. Jittering was used between cue and target, and target and feedback (for each delay, the mean was 4 seconds with a range of 2–6 seconds, and this was balanced across conditions). The block with the scream had an extra trial that was not included in the analysis since we were interested in the threat of the scream rather than the scream itself. The scream was presented once per run to maintain the threat. Thus, there were 18 trials total of each of the six combinations (gain $5‐threat, gain $5‐safe, lose $5‐threat, lose $5‐safe, gain $0‐threat, gain $0‐safe). Each run lasted 13 minutes. Participants were informed that they would get to keep their earnings from the task.

FIGURE 1.

This figure depicts a schematic of the MID‐scream task. The top portion indicates how the shapes correspond to potential monetary earnings and how the coloured border corresponds to the threat. The bottom portion depicts an example of a trial during a threat block. During this trial, the participant was presented with the possibility to gain $5 (circle cue) during sustained threat of hearing a scream and seeing a scary face (blue condition). In this trial, the participant heard the scream and saw the scary face. Upon successfully pressing the button while the target was shown, the participant was provided feedback that they were successful, had gained $5, and shown their cumulative earnings, during the outcome phase of the task. The time depicted is an example of the sequence, but the time between cue and target was jittered from 2–6 seconds, and the time from the target to feedback was jittered 2–6 seconds

2.2.2. Self‐report ratings

Following the MRI scan, participants were asked to report how they felt in response to each cue and condition combination and to hear the scream during the MID‐Scream task. Using a scale of 0–10, they reported how much they liked each combination, and how excited and nervous each combination made them feel.

2.2.3. Family history of harmful alcohol use

Using the Family History Assessment Module, ³³ participants reported whether first‐degree relatives (i.e., parents and siblings) had experienced legal, occupational or social problems due to their alcohol use, given that parental and sibling alcohol use is associated with alcohol use risk (e.g., ³⁴ , ³⁵ , ³⁶ ). Those with first‐degree relatives with harmful alcohol use only (n = 11) or both harmful alcohol and other drug use (n = 20) were coded as family history positive. All others were coded as family history negative (n = 44).

2.2.4. Additional measures

Depression was assessed using the Beck Depression Inventory, ³⁷ a 21‐item measure assessing the intensity of depression. Anxiety was assessed with the Spielberger State–Trait Anxiety Inventory, ³⁸ a 40‐item self‐report measure assessing state anxiety (anxiety taking place at a given time) on 20 items and trait anxiety (the relatively stable disposition to anxiety) on 20 items.

2.3. Imaging acquisition

Magnetic resonance imaging scans were conducted using a Siemens 3.0 Tesla Skyra magnet with a 20‐channel head coil. Functional images were acquired using BOLD signal across 40 axial slices with TR = 2000 ms, TE = 30 ms, flip angle = 77°, acquisition matrix = 74 mm x 74 mm voxels, 40 axial 3 mm thick/0 mm gap slices and multiband factor of 2.

2.4. fMRI data preprocessing

Imaging data were preprocessed using Analysis of Functional NeuroImages version 22.2.01 (AFNI ³⁹ ). We first converted raw scanner data into AFNI‐compatible formats. Using @SSwarper, anatomical data were nonlinearly warped to the Montreal Neurological Institute (MNI) standard space and were skull‐stripped and deobliqued. Task data were deobliqued to match anatomical data. Time points greater than 0.3 mm Euclidean distance of framewise displacement were censored from analyses. We also censored time points with greater than 10% of voxels displaying outliers in activation levels. We used an 8 mm kernel for blur. Each run was scaled to produce a mean voxel intensity of 100. For regression, we examined six cue conditions (gain $5‐threat, gain $5‐safe, lose‐$5‐threat, lose‐$5‐safe, gain$0‐threat, gain$0‐safe) and 12 outcome conditions (trial success and failure for each of the six cue conditions). Separate regressions were conducted for the cue and outcome phases because of the high correlation between conditions (e.g., “gain $5 cue” is strongly correlated with “win $5 outcome”). We used a hemodynamic response model with 1‐second block and an amplitude of 1 for each of these 18 conditions. For the main effect of the threat block, we applied a 96‐second block model with an amplitude of one, starting when the screen indicated if the participant was about to begin a safe or threat block, and terminating when the block was over. For each of those regression analyses, we included six regressors of no interest to control for motion of translation and rotation in the x, y and z dimensions. The final voxel resolution was 3 mm isotropic.

2.5. Data analytic plan

Data analyses were conducted in R version 4.1.2. We assessed significant differences between family history positive and family history negative groups on demographic, substance use, depression, anxiety variables and behavioural task data. We used Student and Welch two‐sample t‐tests and Wilcoxon rank sum tests for continuous variables and Fisher's Exact Test and Pearson Chi‐Square tests for categorical variables. Linear mixed‐effects models and post hoc comparisons using factorial analysis of variance (ANOVA) with Kenward–Roger approximation assessed the main and interaction effects of family history group, cue and condition on participant MID‐Scream reaction times and success rates, and self‐report ratings following the task.

2.5.1. Regions of interest (ROI) analyses

Left and right nucleus accumbens masks were defined using the Harvard‐Oxford Subcortical Structural Atlas (https://identifiers.org/neurovault.image:1700), and the left and right medial prefrontal cortex masks were defined by creating 8 mm diameter spheres centred on the MNI coordinates of x = ± 6, y = 49, z = −8 in AFNI. ⁴⁰ The nucleus accumbens and mPFC is the regions most strongly associated with the cue and outcome phases of the task, respectively. ⁹ , ²⁶ , ⁴¹ The insula masks were defined using the Brainnetome Atlas BN_Atlas_246_3mm.nii ⁴² using parcels 167 and 168 for the left and right, respectively. The anterior insula is a region commonly associated with substance use and threat processing. ¹² , ¹⁷ , ⁴³ Values were extracted from each mask using AFNI's 3dROIstats command. For the bilateral nucleus accumbens, we assessed the main effects of group, cue and condition during the anticipation phase, as well as the interaction effects of group‐by‐cue, group‐by‐condition and group‐by‐cue‐by‐condition. In the bilateral medial prefrontal cortex, we assessed the same effects during the outcome phase of the task. In the bilateral insula, we assessed the main effects of group and block (safe and threatening blocks of the task) and we assessed the interaction effects of group‐by‐block.

Analyses were conducted using the nlme package in R software. The dependent variable in analyses was the beta weights for each task effect. Group, cue and condition were entered as fixed‐effects in restricted maximum likelihood linear mixed‐effects models, with participants entered as the random effect variable. The models were assessed in R followed by factorial ANOVA with Kenward‐Roger approximation. Post hoc comparisons of least‐square mean with Tukey‐adjusted p‐values were conducted using the lsmeans package in R to assess the significant main and interaction effects. To account for the power to detect interactions being approximately half of the main effects, ⁴⁴ an alpha level of 0.10 was applied to interactions, an approach similarly used in Kirk‐Provencher et al. ⁴⁵ For the ROIs, sensitivity analyses were conducted to examine potential confounding factors of lifetime alcohol or cannabis use disorders, sex, depression and trait anxiety by entering the covariates into adjusted linear mixed effects models.

2.5.2. Exploratory analysis

To assess whether group differences in neural activation were evident in other brain regions, we conducted exploratory whole‐brain analyses using a linear mixed effects model in AFNI (3dLME program) to examine the main effects of group, cue and condition, and interaction effects of group‐by‐cue, group‐by‐condition and group‐by‐cue‐by‐condition. We did this for the cue phase only, and not for the outcome phase. We used a minimum voxel‐wise threshold of p < 0.001 and a family‐wise corrected error of α = .05 as assessed by a Monte Carlo simulation using AFNI's 3dClustSim command with spatial autocorrelation function correction parameters of 0.57, 7.04 and 14.59. This indicated a minimum cluster size of k ≥ 49 voxels. For identified clusters, we examined effects in R. As neural activation during anticipation of reward and during sustained threat were our primary interests, we assessed the magnitude of the effect of the group during the gain $5 cue (reward anticipation) and during the threatening block. We generated separate effect size maps for reward and threat as a function of the family history group using Cohen's d, as recommended ⁴⁶ for transparency in neuroimaging research.

3. RESULTS

3.1. Participant characteristics

Participants were aged 18–23, mostly female (56%) and White (76%). The family history positive group (n = 31) had a greater proportion of females (χ²(1) = 8.41, p < 0.05) and reported greater state and trait anxiety (both t ₇₃ > 3.0, p < 0.05) than the family history negative group (n = 44). The groups did not significantly differ in remaining characteristics or alcohol consumption behaviours (Table 1). While one family history‐positive individual met criteria for AUD, the diagnosis was mild and in early remission, indicating no current heavy use.

3.2. Behavioural outcomes of the MID‐scream task

3.2.1. Reaction time and success rate

Reaction time and success rate for trials indicated that participants were engaged and motivated by the task cues (see Supporting Information).

3.2.2. Self‐report ratings

There were no significant group‐by‐cue‐by‐condition interactions for liking, excitement, or nervousness (all F _2,365 < 0.20, p > 0.80; Table S1). There were significant main effects of the condition on liking and nervousness (both F _2,365 > 26.0, p < 0.001; see Table S1), such that participants reported liking the safe condition more than the threat condition and were more nervous during the threat condition than the safe condition.

There were significant group‐by‐cue interactions on liking, excitement and nervousness (all F _2,365 > 2.0, p < 0.070; Table S1); see Figure 2. Compared to the family history negative group, the family history positive group reported a smaller relative difference in liking cues to gain $5 versus gain $0. The family history positive group was significantly less excited during cues to gain $5 than the family history negative group (p = 0.032) and also showed a greater relative change for excitement to gain $5 versus gain $0. Compared to the family history negative group, the family history positive group reported a smaller relative difference in nervousness for cues to gain $5 and avoid losing $5 relative to cues to gain $0. See Supporting Information for full results of self‐report ratings following the MID‐Scream task.

FIGURE 2.

This figure depicts the results of self‐report ratings linear mixed effects modelling. Values indicate the average ratings, ** indicates significance of p < 0.001, “ns” indicates non‐significance, and all error bars represent the standard error of the least‐squares means. Panel A: the graph depicts the significant group‐by‐cue interaction for liking (F _2,365 = 6.86, p = 0.001) with significant contrasts. Panel B: the graph depicts the significant group‐by‐cue interaction for excitement (F _2,365 = 8.93, p < 0.001) with significant contrasts. Panel C: the graph depicts the group‐by‐cue interaction for nervousness (F _2,365 = 2.83, p = 0.060) with significant contrasts

3.3. Region of interest analyses

3.3.1. Insula

There were no significant group‐by‐block interactions in the left or right insula (see Table 2). There was a significant main effect of the block on neural activation bilaterally in the insula (both F _1,73 > 11.0, p = 0.001) where participants demonstrated greater activation during the sustained threat blocks relative to the safe blocks. In the left insula, there was a significant main effect of group (F _1,73 = 8.51, p = 0.005) where the family history positive group showed lower neural activation during the task (Figure 3).

TABLE 2.

Factorial ANOVA results for the left and right regions of interest.

Region of interest	Left				Right
	Effect	F	df	p	Effect	F	df	p
Anticipation phase
Nucleus Accumbens	Intercept	173.52	1,73	<.001	Intercept	131.11	1,73	<.001
	Group	0.87	1,73	.355	Group	0.28	1,73	.598
	Cue	139.60	2,365	<.001	Cue	137.33	2,365	<.001
	Condition	4.09	1,365	.044	Condition	1.23	1,365	.268
	Group x Cue	2.47	2,365	.086	Group x Cue	3.05	2,365	.048
	Group x Condition	0.06	1,365	.809	Group x Condition	0.12	1,365	.728
	Group x Cue x Condition	0.41	2,365	.667	Group x Cue x Condition	0.15	2,365	.859
Outcome phase
Medial prefrontal cortex	Intercept	2.43	1,73	.123	Intercept	2.47	1,73	.120
	Group	1.85	1,73	.178	Group	0.22	1,73	.639
	Outcome	44.80	2,365	<.001	Outcome	37.85	2,365	<.001
	Condition	0.20	1,365	.656	Condition	0.80	1,365	.373
	Group x Outcome	2.53	2,365	.081	Group x Outcome	2.62	2,365	.074
	Group x Condition	1.80	1,365	.180	Group x Condition	0.003	1,365	.958
	Group x Outcome x Condition	0.40	2,365	.671	Group x Outcome x Condition	0.25	2,365	.782
Blocks
Insula	Intercept	57.94	1,73	<.001	Intercept	78.38	1,73	<.001
	Group	8.51	1,73	.005	Group	1.67	1,73	.201
	Block	11.27	1,73	.001	Block	11.95	1,73	.001
	Group x Block	0.09	1,73	.759	Group x Block	0.03	1,73	.861

Note: Bold font indicates significance at p < .05, and italic font indicates interaction significance at p < .10.

FIGURE 3.

This figure depicts the results of the insula linear mixed effects modelling for the safe and threatening blocks during the MID‐scream task. * indicates significance at p < .05, “ns” indicates non‐significance, and all error bars represent the standard error of the least‐squares means. FH‐ = family history negative and FH + = family history positive. Panel A: the coronal view of the brain depicts the left insula highlighted in red. The graph depicts the main effects of group (F _1,73 = 8.51, p = 0.005; indicated by the * on the graph) and block (F _1,73 = 11.27, p = 0.001); the group‐by‐block interaction was not significant. Panel B: the coronal view of the brain depicts the right insula highlighted in red. The graph depicts the effects of group and block on neural activation; only block showed a significant main effect (F _2,365 = 11.95, p = 0.001), while the main effect of group (indicated by the “ns” on the graph) and the group‐by‐block interaction were not significant

3.3.2. Nucleus Accumbens

During the anticipation phase of the task in the left and right nucleus accumbens, there were no significant group‐by‐condition or group‐by‐cue‐by‐condition interactions (see Table 2). In the left accumbens only, there was a significant main effect of condition (F _1,365 = 4.09, p = 0.044), where participants demonstrated greater activation during the safe relative to threat condition (Figure 4). Additionally, there was a significant main effect of cue (both F _2,365 > 137.0, p < 0.001) and a group‐by‐cue interaction (both F _2,365 > 2.0, p < 0.09) on neural activation bilaterally in the accumbens (Figure 4). Participants demonstrated greater activation during cues to gain $5 relative to gain $0 and avoid losing $5 and during cues to avoid losing $5 relative to gain $0. The relative differences in activation between cues to gain $5 and gain $0 and between cues to avoid losing $5 and gain $0 were smaller for the family history positive group compared to the family history negative group. The groups did not significantly differ on activation during cues to gain $5, avoid losing $5, or gain $0 (all p > 0.50).

FIGURE 4.

This figure depicts the results of the nucleus accumbens linear mixed effects modelling for the anticipation phase of the MID‐scream task. Values on the graphs indicate the average relative difference in activation between cues, ** indicates significance at p < .001, and all error bars represent the standard error of the least‐squares means. Panel A: the coronal view of the brain depicts the left nucleus accumbens highlighted in red. The graph depicts the significant group‐by‐cue interaction (F _2,365 = 2.47, p = 0.086) with significant contrasts. Panel B: the coronal view of the brain depicts the right nucleus accumbens highlighted in red. The graph depicts the significant group‐by‐cue interaction (F _2,365 = 3.05, p = 0.048) with significant contrasts

3.3.3. Medial prefrontal cortex

During the outcome phase of the task, in the bilateral medial prefrontal cortex, there were no significant main effects of condition, nor significant group‐by‐condition or group‐by‐outcome‐by‐condition interactions (see Table 2). There was a significant main effect of outcome (both F _2,365 > 37.0, p < 0.001) and a group‐by‐outcome interaction (both F _2,365 > 2.0, p < 0.09) on activation bilaterally (Figure 5). Participants demonstrated greater activation during outcomes of winning $5 relative to gaining $0 and avoiding the loss of $5. The relative difference in activation between outcomes of winning $5 and gaining $0 was smaller for the family history positive group bilaterally compared to the family history negative group. The groups did not significantly differ on activation during the winning $5, gaining $0 or the loss of $5 outcomes (all p > 0.10).

FIGURE 5.

This figure depicts the results of the medial prefrontal cortex linear mixed effects modelling for the outcome phase of the MID‐scream task. Lose $5 = avoiding losing $5 outcome, values on the graphs indicate the average relative difference in activation between outcomes, * indicates significance at p < .05, ** indicates significance at p < .001, “ns” indicates non‐significance, and all error bars represent the standard error of the least‐squares means. Panel A: the axial view of the brain depicts the left medial prefrontal cortex highlighted in red. The graph depicts the significant group‐by‐outcome interaction (F _2,365 = 2.53, p = 0.081) with significant contrasts. Panel B: the axial view of the brain depicts the right medial prefrontal cortex highlighted in red. The graph depicts the significant group‐by‐outcome interaction (F _2,365 = 2.62, p = 0.074) with significant contrasts

3.3.4. Sensitivity analyses

Results showed that the significant main and interaction effects in the unadjusted models were not altered after including lifetime alcohol use disorder, cannabis use disorder, sex, trait anxiety, or depression in the adjusted models. None of these variables had significant associations with neural activation in the bilateral insula, left nucleus accumbens and bilateral medial prefrontal cortex (all F _1,69 < 4.0, p ≥ .05), while trait anxiety was significantly associated with neural activation in the right nucleus accumbens (F _1,69 = 4.86, p = .031); see Supporting Information.

3.4. Exploratory analyses

Whole‐brain analyses did not identify any significant group‐by‐cue‐by‐condition effects. One cluster was identified with a significant effect of group and condition during the anticipation phase of the task. Regions in the cluster include the left operculum parietal 1, retroinsular cortex and area PFcm of the inferior parietal lobule. In this cluster, there was a significant main effect of cue (F _2,365 = 23.31, p < 0.001) and a significant group‐by‐condition interaction (F _1,365 = 19.96, p < 0.001; Table S6 and Figure S2); see Supporting Information.

We explored effect size differences ⁴⁷ to depict how the groups compared during the anticipation of threat the threatening blocks of the task, and also during the anticipation of winning money when presented with the gain $5 cue. During the threatening blocks, the family history positive relative to the family history negative group showed less activation across several regions of the brain, with one significant cluster with a medium‐large effect size located bilaterally in the dorsal anterior cingulate cortex, extending bilaterally to the caudal dorsomedial prefrontal cortex (p = 0.001, d = 0.75; Figure S3). The groups showed non‐significant differences in activation in numerous areas during the anticipation of gaining $5, with some areas showing medium effect sizes (i.e., d > 0.5) for group differences (Figure S4); see Supporting Information.

4. DISCUSSION

The present study tested the hypothesis that young adults with a family history of harmful alcohol use differ in reward processing while under sustained threat compared to those without such a family history. Using the MID‐Scream task, ²⁸ we examined the interactions between family history group and neural activation in response to anticipation of threat and monetary reward in threat‐ (i.e., insula) and reward‐related (i.e., accumbens and mPFC) brain regions. Our family history‐positive population did demonstrate evidence for diminished reward signalling, but we did not find support for the hypothesized interaction between family history, threat and reward. Our results indicate that, in the absence of current heavy alcohol use, individuals with enriched risk may not differ from those without such risk on sensitivity to the threat but demonstrate diminished salience to potential rewards.

Consistent with previous research (e.g., ¹⁸ , ⁴⁸ , ⁴⁹ ), our results indicate that those with positive family histories have less neural differentiation between rewarding and neutral cues and outcomes, than those without such family histories. These results suggest that the family history positive group has dampened monetary reward salience irrespective of unpredictable threat. We found the family history positive group, versus the family history negative group, demonstrated smaller relative differences in neural activation in the nucleus accumbens and the medial prefrontal cortex during the anticipation and notification of winning money relative to earning no money, respectively. We further found that the family history positive group reported liking and being excited during the anticipation of winning money relative to winning no money, less so than the family history negative group. The family history positive group reported feeling less nervous during the anticipation of winning money relative to earning no money, compared to the family history negative group. These results are consistent with the findings of a recent meta‐analysis ¹¹ that found that individuals with alcohol use disorder demonstrated less neural activation in several reward and incentive brain regions, including the ventral striatum, during anticipation of reward and receipt of reward during the MID task. Zeng and colleagues ¹¹ posit that decreased activation in these regions among individuals with alcohol use disorder indicates dysfunction in incentive salience to conventional rewards (e.g., money). Taken together, our results indicate that family history positive young adults may not be as sensitive to the rewarding salience of monetary reward tasks as family history negative individuals.

Contrary to our hypothesis, the groups did not differ on neural activation in the insula as a function of induced anticipatory anxiety, despite the family history positive group reporting higher levels of state‐ and trait‐anxiety. While previous research has indicated that individuals with higher levels of anxiety, ⁵⁰ and individuals with alcohol use disorder ¹⁷ show increased insular activation while under unpredictable threat, the study populations of those studies differed from ours. Specifically, our sample consisted of young adults who had low levels of psychopathology overall, and low drinking levels overall. Those studies also used shock, versus a scream, as the aversive event. This may indicate that young adults with familial risk of developing an alcohol use disorder do not differ from those without such family histories in their sensitivity to threat. It may be that anticipation of the threat becomes altered following chronic patterns of heavy drinking (e.g., the substance‐induced model ⁵¹ ). For example, we previously showed no difference in negative emotion reactivity among young adults with a family history of harmful alcohol use. ⁴⁵ Continued hazardous or harmful alcohol use is associated with the development of depression ⁵² and anxiety, ⁵¹ and results in negative affectivity ⁵³ and individuals may then continue to engage in harmful alcohol use due to the reduction of anxious states following the consumption of alcohol. ⁵⁴ It is also possible that in our sample, those with enriched risk for developing an alcohol use disorder may not experience altered reactivity to prolonged threat due to resilience. ⁴⁹ It may be that by this age if individuals with enriched risk have not gone on to develop harmful patterns of use, they may be more resilient to environmental risk factors including threatening events. However, should family history‐positive individuals go on to develop hazardous alcohol use behaviours in the future, we may expect to see altered insular functioning, given the association with alcohol craving, ⁵⁵ alcohol use disorder, ¹⁷ and risky decision making. ⁵⁶ , ⁵⁷ , ⁵⁸

At the same time, negative outcomes, such as the actual loss of money do not appear to differentially alter reward responses in family history positive individuals. This may indicate that at‐risk young adults may be appropriately sensitive to the experience of negative outcomes even though they do not currently have patterns of chronic heavy alcohol use. There were also no three‐way group‐by‐cue‐by‐condition interactions in the accumbens during the cue phase or the mPFC during the outcome phase. Meanwhile, the dampened response to the possibility of losing money relative to neutral outcomes and less self‐reported nervousness during anticipation of winning or losing money among the family history positive group suggests dysfunction in the salience of potential negative consequences. Thus, it appears that a family history of positive individuals demonstrates decreased reward response to both potential rewarding and negative outcomes. The results of our study suggest that family history‐positive individuals have a predisposition to dysfunctional reward processing of potential conventional rewards and losses, even in the absence of hazardous alcohol use patterns. The lack of natural reward response has garnered speculation that such individuals may engage in excessive sensation seeking to achieve these rewarding effects, including using substances at hazardous levels. ⁵⁹

4.1. Limitations

Our study had some limitations. Most individuals in our study were white, so the homogeneity of our sample may decrease generalizability. Similarly, most family‐history‐positive individuals were female, so the sex distribution of our sample may be a limitation that reduces the generalizability of our findings. Future studies should explore group‐by‐sex interactions, but we were underpowered in this study. It is possible that our sample was resilient or comprised of individuals motivated to refrain from harmful drinking patterns. For example, our sample may demonstrate resilience as this sample represents a group of at‐risk individuals who, despite having begun to drink alcohol, have not largely developed any hazardous drinking patterns, thus may have surpassed the window during which we would expect such young adults to develop harmful patterns of use. We used monetary rewards, and thus using alcohol or other appetitive rewards may have led to different results. We recruited individuals who had either a first‐degree relative with AUD or another SUD, but the results may differ if we had only recruited individuals with a family history of AUD. While all participants with a positive family history reported harmful alcohol use in the affected family member, some also reported harmful use of other drugs, and it remains unclear how this heterogeneity impacts our results. Finally, our study design did not counterbalance the order of presenting the threat and safe blocks; all participants experienced a threat block first. While we do not have reason to suspect this to have compromised the signal of the threat condition, we cannot rule out the possibility that it altered the response in some way.

4.2. Conclusions

Through employing a model of both unpredictable, sustained threat and reward processing in a modified MID task, we studied an at‐risk group for alcohol use disorder to address questions about neural processing that may affect the likelihood of developing harmful alcohol use patterns. The results did demonstrate important group differences based on family history. Most notably, the results of this study show that family history positive individuals did not differ from family history negative individuals in neural response in the insula during unpredictable threat, but did differ as a group by demonstrating diminished neural response to monetary incentives during the MID task in the nucleus accumbens and medial prefrontal cortex. These results may provide an important clue to how alcohol use disorder develops with regard to emotional processing. Our results suggest that at‐risk individuals, who have not developed harmful alcohol use patterns, do demonstrate a diminished response to monetary reward, while notably not differing during a simulated anxious state. While this would require further study, our results may provide evidence that even in the absence of current harmful alcohol use patterns, there is a predisposition in enriched‐risk individuals through an altered reward pathway. With further study in more diverse samples, we may find that altered response to unpredictable threats may be a neural adaptation that follows from chronic, heavy alcohol use rather than a preexisting condition.

AUTHOR CONTRIBUTIONS

Conceptualization: Katelyn Kirk‐Provencher, Joshua Gowin. Data curation: Katelyn Kirk‐Provencher, Keinada Andereas, Rosa Hakimi, and Joshua Gowin. Formal analysis: Katelyn Kirk‐Provencher and Joshua Gowin. Funding acquisition: Joshua Gowin. Investigation: Anne Penner and Joshua Gowin. Methodology: Anne Penner and Joshua Gowin. Project administration: Joshua Gowin. Resources: Joshua Gowin. Supervision: Joshua Gowin. Visualization: Katelyn Kirk‐Provencher. Roles/writing—original draft: Katelyn Kirk‐Provencher, Keinada Andereas, and Rosa Hakimi. Writing—review and editing: Katelyn Kirk‐Provencher, Keinada Andereas, Rosa Hakimi, Anne Penner, and Joshua Gowin.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

Supporting information

Table S1. Factorial ANOVA Results for Reaction Time, Success Rate, and Self‐Report Ratings on MID‐Scream fMRI Task.

Figure S1. Reaction Time and Success Rate.

Table S2. Mean Reaction Times and Self‐Report Ratings by Group for Cue and Condition Trials.

Table S3. Sensitivity Analyses: Adjusted Factorial ANOVA Results for the Left and Right Insula for the Safe and Threatening Blocks of the MID‐Scream Task.

Table S4. Sensitivity Analyses: Adjusted Factorial ANOVA Results for the Left and Right Nucleus Accumbens for the Anticipation Phase of the MID‐Scream Task.

Table S5. Sensitivity Analyses: Adjusted Factorial ANOVA Results for the Left and Right Medial Prefrontal Cortex for the Outcome Phase of the MID‐Scream Task.

Table S6. Exploratory Analysis: Factorial ANOVA Results for Significant Effect of Group and Condition on Activation During Whole‐Brain Analysis at k ≥ 49 (p ≤ .001).

Figure S2. Whole‐Brain Analysis: Effect of Group and Condition on Neural Activation.

Figure S3. Effect Size Map: Group Comparison During Anticipation of Winning (Gain $5).

Figure S4. Effect Size Map: Group Comparison During Threat Blocks.

Kirk‐Provencher KT, Hakimi RH, Andereas K, Penner AE, Gowin JL. Neural response to threat and reward among young adults at risk for alcohol use disorder. Addiction Biology. 2024;29(2):1‐13. doi: 10.1111/adb.13378

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.