Reward Processing in Alcohol-Dependent Patients and First-Degree Relatives: Functional Brain Activity During Anticipation of Monetary Gains and Losses

BACKGROUND: According to the reward de ﬁ ciency syndrome and allostatic hypotheses, hyposensitivity of mesocorticolimbic regions to non – alcohol-related stimuli predisposes to dependence or is long-lastingly enhanced by chronic substance use. To date, no study has directly compared mesocorticolimbic brain activity during non – drug reward anticipation between alcohol-dependent, at risk, and healthy subjects. METHODS: Seventy- ﬁ ve abstinent alcohol-dependent human subjects (mean abstinence duration 957.66 days), 62 healthy ﬁ rst-degree relatives of alcohol-dependent individuals, and 76 healthy control subjects without family history of alcohol dependence performed a monetary incentive delay task. Functional magnetic resonance imaging data of the anticipation phase were analyzed, during which visual cues predicted that fast response to a target would result in monetary gain, avoidance of monetary loss, or a neutral outcome. RESULTS: During gain anticipation, there were no signi ﬁ cant group differences. During loss anticipation, abstinent alcohol-dependent subjects showed lower activity in the left anterior insula compared with healthy control subjects without family history of alcohol dependence only (Montreal Neurological Institute [MNI] 2 25 19 – 5; t 206 = 4.17, familywise error corrected p = .009). However, this effect was no longer signi ﬁ cant when age was included as a covariate. There were no group differences between abstinent alcohol-dependent subjects and healthy ﬁ rst-degree relatives or between healthy ﬁ rst-degree relatives and healthy control subjects during loss anticipation, respectively. CONCLUSIONS: Neither the neural reward de ﬁ ciency syndrome nor the allostatic hypotheses are supported by the results. Future studies should investigate whether the incentive salience hypothesis allows for more accurate predictions regarding mesocorticolimbic brain activity of subjects with alcohol dependence and healthy individuals during reward and loss anticipation and further examine the neural substrates

mesocorticolimbic system persisting throughout abstinence, as repeated drug use sensitizes dopamine release in the VS and thus reinforces drug intake and salience attribution to drug-associated cues (20). While phasic dopamine release as elicited by drug cues cannot be measured in vivo in humans, some empirical support for this hypothesis is provided by studies in [abstinent (20)(21)(22)] individuals with alcohol dependence (23,24), reporting increased functional activity in the VS, related mesocorticolimbic regions, and the insula in response to alcohol-associated stimuli compared with healthy control subjects.
Alternatively, the reward deficiency hypothesis suggests that an at least partly heritable deficit in the brain reward system including its dopaminergic input predisposes subjects to consume drugs of abuse or excessively engage in other activities such as food consumption, gambling, or high-risk sports because they strongly stimulate otherwise impaired dopaminergic neurotransmission and thus temporally compensate for the deficit (25,26).
Conversely, the allostatic hypothesis (27)(28)(29)(30) postulates that chronic substance use both decreases the sensitivity of the reward system to drug-related and natural rewards and recruits anti-reward networks in the sense of an opponent process (31,32). These neuroplastic changes are hypothesized to persist even during protracted abstinence (27)(28)(29)(30), though some studies described rather fast recovery of dopaminergic neurotransmission within the first days of alcohol abstinence (11,33) or even a hyperdopaminergic state after protracted abstinence (34).
A paradigm widely used to study brain activity during nonsubstance reward processing is the monetary incentive delay task (MIDT) (35), allowing dissociation between reward anticipation and receipt (36). During gain and loss anticipation, healthy subjects show robust activations in the VS, an important receptor region of dopaminergic afferents from the midbrain, as well as in the ACC and anterior insula (AI) (36,37). MIDT studies comparing abstinent alcohol-dependent (AD) subjects to healthy control subjects during functional magnetic resonance imaging (fMRI) have provided contradictory findings (36). While some found no differences during the anticipation phase (38)(39)(40), most studies, including a meta-analysis (41), reported comparatively decreased activity in AD subjects in the VS, insula, and frontal brain regions, among others (5,20,(42)(43)(44).
Whether neuronal differences are a consequence of chronic alcohol consumption or due to predisposition to alcohol dependence has been debated. One strategy to disentangle the contribution of underlying factors has been the examination of high-risk subjects, namely healthy first-degree relatives of alcohol-dependent individuals [family history-positive (FH 1 ) subjects (45)]. For instance, according to the reward deficiency syndrome hypothesis (26), FH 1 subjects, and (abstinent) individuals with alcohol dependence should exhibit similarly decreased neural activation to non-substance-related rewards, resulting in a chronic dysphoric state that can solely be reversed by substance use or other highly rewarding activities (26). While most MIDT studies in FH 1 subjects (46)(47)(48)(49) found no functional differences to healthy subjects during the anticipation phase, other studies showed comparatively increased (45,50,51) or decreased (52) activity in relatives. Differences existed in the VS, ACC, and insula, among others (45,51,52).
Findings from previous studies only allow for limited conclusions regarding the role of neural responses to non-drug reward cues as a risk factor for alcohol dependence as they 1) mainly had small sample sizes (5,20,39,(42)(43)(44)(45)(46)49,(51)(52)(53) and 2) did not directly compare neural responses between FH 1 and AD subjects. It thus remains to be elucidated whether similar functional abnormalities occur in both groups.
To this end, AD subjects with alcohol dependence according to the DSM-IV (54), FH 1 subjects, and healthy control subjects without a family history of alcohol dependence (family history-negative [FH 2 ] subjects) were examined using fMRI during an MIDT. To our knowledge, this is the first direct comparison of AD, FH 1 , and FH 2 subjects during the anticipation of monetary gains and losses. In addition to whole-brain analyses, region of interest (ROI) analyses were also performed. Based on the theoretical importance of the mesocorticolimbic dopamine system and the insula in reward processing and on a recent meta-analysis on anticipatory functional activity in healthy subjects (37), the VTA, VS, ACC, and AI were defined as ROIs.
Departing from the reward deficiency syndrome (26) and the allostatic hypotheses (27)(28)(29)(30), we expected that 1) AD subjects would show lower neuronal activity than FH 2 subjects, 2) FH 1 subjects would show lower activity than FH 2 subjects, and 3) AD subjects would show lower activity than FH 1 subjects in the ROIs during gain and loss anticipation, compared with anticipation of neutral outcomes, respectively.

Subjects and Procedures
A total of 235 subjects (87 AD, 66 FH 1 , 82 FH 2 ) participated in the study. Twenty-two subjects (12 AD, 4 FH 1 , 6 FH 2 ) were excluded from data analyses due to missing or poor-quality MRI data. In total, we thus included data of 213 subjects (75 AD, 62 FH 1 , 76 FH 2 ). Subjects gave written informed consent prior to participation, and the study was conducted in accordance with the Declaration of Helsinki and approved by all local ethics committees.
AD subjects reported a diagnosis of alcohol dependence according to DSM-IV (54) and at least 7 days of abstinence at the time of screening. With an average abstinence duration of 957.66 days (n = 64, SD = 1436.87), the majority of the AD sample was in protracted abstinence. FH 1 subjects reported not being affected by DSM-IV (54) alcohol dependence themselves but having at least one affected first-degree relative. In case of uncertainty about the relative's diagnosis, we administered the Family History Assessment Module (55,56). FH 2 subjects reported the absence of alcohol dependence according to DSM-IV (54) in themselves and any first-degree relative.
Participants were excluded if they had any DSM-IV (54) Axis I psychiatric disorder other than alcohol dependence (AD group), nicotine dependence or abuse, single past depressive episodes, or anxiety and adjustment disorders, as assessed via the German screening version of the Structured Clinical Interview for DSM-IV Axis I (57). As verified by drug urine tests, subjects had not consumed psychotropic substances (cannabinoids, amphetamines, cocaine, opioids) within 1 month Reward Processing in Alcohol Dependence prior to participation. Further exclusion criteria were current antidepressant or neuroleptic medication, claustrophobia, physical diseases potentially interfering with the examinations or having an influence on the parameters of interest, MRI contraindications, insufficient knowledge of German language, and current breastfeeding or pregnancy. Significant age differences were present between AD and FH 1 subjects and between AD and FH 2 subjects, respectively (Table 1).

Monetary Incentive Delay Task
We administered a modified version of the MIDT (35) using Presentation (Neurobehavioral Systems) (Supplemental Methods and Figure S1). In each trial, subjects saw a cue indicating a potential gain, potential loss, or neutral trial, respectively (anticipation phase). After a variable delay, they had to respond to a target as quickly as possible. Subsequent feedback (outcome phase) indicated whether a trial was successful (i.e., 11V in gain trials, 0V in loss trials) or unsuccessful (i.e., 0V in gain trials, 21V in loss trials). The outcome for neutral trials was always 0V. The MIDT had a total duration of approximately 12 minutes and consisted of 25 gain, 25 loss, and 25 neutral trials presented in a pseudo-randomized order, with the same condition occurring maximally twice in succession.
To avoid decreased anticipatory activity at the beginning of the MIDT (36) and to ensure learning of the association between the cues and their respective outcomes, participants performed one practice session each outside and inside the scanner. An adaptive algorithm was applied in the in-scanner practice session and during the main task, leading to each subject being successful in approximately 70% of trials (Supplemental Methods).

Data Analysis
Demographic and behavioral MIDT data were analyzed using SPSS Statistics 23 and 28 (IBM Corp.), while fMRI analyses were performed with SPM12 (The Wellcome Centre for Human Neuroimaging) in MATLAB (versions 2013b and 2021a; The MathWorks, Inc.). Data analyses were preregistered in the Open Science Framework (https://osf.io/3gnk9) (58).
Behavioral Data. Group differences regarding sociodemographic and psychometric data were examined using c 2 tests or analyses of variance, respectively. For MIDT reaction time (RT), success rate, and outcome analyses, we used univariate analyses of variance. Significant main effects of group were followed by post hoc independent-samples t tests or, in case of variance inequality, Welch tests. In case of significant main effects of anticipation condition, we performed post hoc t tests for dependent samples. For exploratory analyses including age as a covariate, see the Supplement.
fMRI Data. After preprocessing (Supplemental Methods), we performed firstand second-level analyses in the context of the general linear model. At the first level of the event-related design, 3 anticipation conditions (gain, loss, neutral), target presentation, and 5 feedback conditions (gain, unsuccessful gain attempt, neutral feedback, loss avoidance, unsuccessful loss avoidance attempt) were modeled using stick functions convolved with the hemodynamic response function implemented in SPM. In addition, we included 3 regressors each for translational and rotational head motion.
For each participant, the linear contrast images gain . neutral anticipation and loss . neutral anticipation were computed and subsequently used in the random-effects second-level analysis. Here, the 3 groups (AD, FH 1 , FH 2 ) were included as regressors. MRI scanner, handedness, and smoking status were included as covariates of no interest (Supplemental Methods). To rule out the possibility that neural group differences were caused or masked by systematically different success rates (36) and thus by systematically different motivation to actively achieve a positive outcome, we included success rate in MIDT gain or loss trials as a covariate (Supplemental Discussion).
Besides whole-brain analyses, we conducted ROI analyses using small-volume correction for each group comparison (AD vs. FH 2 , FH 1 vs. FH 2 , AD vs. FH 1 ) for the contrasts gain . neutral anticipation and loss . neutral anticipation. Based on previous MIDT studies (37) and using an in-house tool (59), the VS, AI, ACC, and VTA were defined as ROIs (Supplemental Methods, Figure S2, and Tables S1 and S2).
A familywise error (FWE)-corrected significance level of p = .05 was used in all whole-brain and ROI analyses. Activations within AD, FH 1 , and FH 2 groups during gain and loss anticipation were assessed using one-sample t tests, group differences (AD vs. FH 2 , FH 1 vs. FH 2 , AD vs. FH 1 ) for both contrasts (gain . neutral anticipation, loss . neutral anticipation) using independent-samples t tests. For exploratory analyses related to success rate, age, craving measures, abstinence duration, and the MIDT outcome phase, see the Supplement.

Behavioral Data
We found a main effect of anticipation condition on RT (Tables 2 and 3): consistent with a basic assumption underlying MIDT (36), participants responded faster in gain and loss compared with neutral trials, respectively. In addition, RT across groups was lower in gain than in loss trials. Moreover, the main effect of groups was significant ( Figure 1A), with AD and FH 1 subjects responding more slowly than FH 2 subjects across anticipation conditions, respectively. AD and FH 1 subjects did not differ regarding RT. The interaction between group and condition was not significant.
A similar pattern emerged regarding success rate (Tables 2  and 3): across groups, the success rate was higher in gain than

Reward Processing in Alcohol Dependence
Biological Psychiatry --, 2022; -:---www.sobp.org/journal Reward Processing in Alcohol Dependence in loss trials. In addition, we found a main effect of group ( Figure 1B). Post hoc t tests revealed that AD subjects were less successful than FH 2 subjects, while AD and FH 1 subjects as well as FH 1 and FH 2 subjects did not differ, respectively. The interaction between group and condition was not significant.

fMRI Data
As expected (37,60), confirming MIDT effectiveness, FH 2 subjects showed significant activation in the VTA, VS, ACC, and AI, among others, during both gain and loss anticipation contrasted against neutral trials, respectively (Figure 2; Tables S5 and S8).

Gain Anticipation
During gain anticipation (gain . neutral anticipation), we found no differences in activity between AD, FH 1 , and FH 2 subjects, neither within the ROIs (Figure 3) nor in whole-brain analyses ( Figure 2; Tables S3-S5). Exploratory ROI and whole-brain analyses 1) including age as covariate, with and without success rate as covariate, respectively, and 2) on subsamples of age-matched AD versus FH 1 subjects or AD versus FH 2 subjects, respectively, did not reveal any group differences. Instead, we found significant reductions in neural activity with increasing age in the VS ROI When exploring the role of the substantial variation in abstinence duration in AD subjects, we found no significant correlation (n = 64, r s = 20.19, p = .14) between mean activity in the bilateral VS ROI during gain anticipation and abstinence duration (Supplemental Results and Figure S4).

Loss Anticipation
During loss anticipation (loss . neutral anticipation), AD subjects exhibited significantly decreased neuronal activation in the left AI ROI compared with FH 2 subjects (MNI 225 19 25; t 206 = 4.17, p FWE = .009) (Figure 4). There were no significant differences in activity within the ROIs between AD and FH 1 subjects or between FH 1 and FH 2 subjects, respectively. Whole-brain analyses did not reveal any significant group differences, although, descriptively, neural activity spanned fewer brain regions in AD than in FH 1 and FH 2 subjects, respectively ( Figure 2; Tables S6-S8).
When exploratively 1) including age as covariate in wholebrain and ROI analyses, with and without success rate as covariate, respectively, and 2) repeating whole-brain and ROI analyses on subsamples of age-matched AD versus FH 1 subjects or AD versus FH 2 subjects, we neither found a difference between AD and FH 2 subjects in left AI nor any other group difference. Instead, with increasing age, BOLD signals decreased in left putamen (whole-brain,  ).

Reward Processing in Alcohol Dependence
Biological Psychiatry --, 2022; -:---www.sobp.org/journal In a post hoc analysis exploring the substantial variation in abstinence duration in AD subjects, we found no significant Spearman's rank correlation between mean activity in the bilateral AI ROI during loss anticipation and abstinence duration (n = 64, r s = 20.05, p = .68) (Supplement and Figure S7).

DISCUSSION
During the anticipation of monetary gains or losses, FH 1 subjects did not exhibit reduced or increased neural responses in mesocorticolimbic regions in comparison to FH 2 and AD subjects, respectively. In contrast, AD subjects showed decreased activity in the left AI during loss anticipation compared with FH 2 subjects. However, this finding was not significant when age was included as a covariate. Taken together, our findings are neither in accordance with the reward deficiency hypothesis (26) nor the allostatic hypothesis (27)(28)(29)(30), at least when the latter refers to persistent hyposensitivity of the reward system to non-drug-related stimuli during prolonged abstinence.

Gain Anticipation
We found no significant functional differences between FH 1 and FH 2 subjects in VTA, VS, ACC, or AI during gain anticipation. While this is in line with a number of previous studies (46)(47)(48)(49), it stands in contrast to others reporting functional differences of inconsistent direction between FH 1 and FH 2 subjects during gain anticipation (45,(50)(51)(52).
Reinforced by a comparatively large sample size, our finding questions the validity of the reward deficiency syndrome hypothesis (26) and the concomitant assumption that Values are presented as mean 6 SD. AD, abstinent alcohol-dependent subjects; FH 1 , healthy first-degree relatives of alcohol-dependent individuals; FH 2 , healthy control subjects without family history of alcohol dependence; Gain, in gain trials; Loss, in loss trials; Neutral, in neutral trials; RT, reaction time.
a Values refer to a subsample of 208 subjects (72 AD, 61 FH 1 , 75 FH 2 ) (Supplemental Methods). Reward Processing in Alcohol Dependence hyposensitivity of the mesocorticolimbic reward system to non-substance-related rewards predisposes to alcohol dependence, at least when assuming a genetic contribution to reward deficiency shared between (subsequent) patients and relatives of alcohol-dependent patients. However, unaffected family history-positive individuals may differ from (subsequently) addicted individuals: Volkow et al. (61) observed increased dopamine D 2 receptors in the VS of unaffected relatives of alcohol-dependent patients, who, in contrast, tended to display reduced dopamine D 2 receptors at least during early abstinence (62)(63)(64). Whether, instead, a hypersensitivity of mesocorticolimbic regions to alcohol-related stimuli (9,18,19) plays a role in relatives remains to be elucidated. So far, consistent with the incentive salience hypothesis (9,13,18,19), increased alcohol cue-elicited activity in the insula, among other regions, has been found in heavy drinkers regardless of familial exposure (65). For AD and FH 2 subjects, we found comparable neuronal activation during the anticipation of monetary gains. While in line with 3 previous studies (38)(39)(40), this is inconsistent with the majority of MIDT research in AD subjects (5,20,(42)(43)(44) and a meta-analysis (41) reporting decreased activity in the VS and insula, among others. Importantly, most previous studies reporting group differences assessed patients with rather short abstinence durations [1-3 weeks; e.g., (5,20,42,44)], while in the current study, average abstinence duration was about 3 years (Supplemental Results). Neuroadaptations elicited by chronic alcohol intake have, at least in subgroups (66,67), been shown to reverse within the first weeks or days of abstinence (33,68,69). Conversely, delayed recovery of dopaminergic neurotransmission following detoxification is associated with subsequent relapse (11,69). In our long-term abstinent sample, dopamine neurotransmission and dopamine-dependent reward anticipation (70) might thus long have recovered from such neuroadaptations, and we cannot rule out reduced mesolimbic responses in acutely abstinent individuals. Our finding that abstinence duration was not correlated with mean reward anticipatory activity in the VS does not stand in contrast to this hypothesis, as only 2 subjects with a known abstinence duration of ,1 month were included (Supplement).

Loss Anticipation
During loss anticipation, FH 1 subjects did not show decreased activity in any ROI compared with FH 2 subjects. This is inconsistent with 2 previous studies (51,52), but in line with most studies reporting no group differences in insula (46,48,52) or VS activation (46,48) between relatives and individuals without family history of alcohol dependence. As loss anticipation in the MIDT may be best understood as anticipation of potential loss avoidance, this finding, again, argues against the reward deficiency syndrome hypothesis.
However, AD subjects exhibited decreased activity in the left AI compared with FH 2 subjects during loss anticipation. The AI has been linked to action initiation in response to subjectively salient stimuli (77) and impairments in insula

Reward Processing in Alcohol Dependence
Biological Psychiatry --, 2022; -:---www.sobp.org/journal function have repeatedly been associated with addictive behavior (13,14,17). However, the decreased AI activation was most likely driven by AD subjects' being significantly older than FH 2 subjects rather than by their diagnosis of alcohol dependence, as it was no longer significant when age was included as a covariate.
Apart from the AI, we found no significant differences in neural activation between AD and FH 2 subjects in any other ROI. This is inconsistent with 2 previous findings (5,20), but in line with several MIDT studies in AD subjects (38,39,42,43) and suggests that the most robust finding is that of no group differences during loss anticipation.

Limitations
An important limitation of this study is that we cannot rule out the confounding influence of age. Indeed, exploratory analyses showed that activation in the VS and AI decreased with increasing age, which is in line with findings in healthy subjects (78). However, considering that older FH 1 subjects who did not develop alcohol dependence are likely particularly resilient, we did not find any group differences in mesolimbic activation when comparing age-matched subsamples of FH 1 and AD or FH 2 and AD subjects, respectively. In addition, since ROI analyses including age as a covariate did not reveal any significant group differences during gain or loss anticipation, we deem it unlikely that age differences between groups could Figure 2. Whole-brain activity (p , .05, familywise error [FWE] corrected) within healthy control subjects without family history of alcohol dependence (FH 2 ), healthy first-degree relatives of alcohol-dependent individuals (FH 1 ), and abstinent alcohol-dependent subjects (AD), respectively, during gain (gain . neutral anticipation; warm colors) and loss (loss . neutral anticipation; cold colors) anticipation in the monetary incentive delay task. The respective z coordinate is indicated above each layer. Figure 3. Estimated contrast weights for mean activity in the bilateral ventral striatum (VS) region of interest during gain anticipation (gain . neutral anticipation) in the monetary incentive delay task for healthy control subjects without family history of alcohol dependence (FH 2 ), healthy firstdegree relatives of alcohol-dependent individuals (FH 1 ), and abstinent alcohol-dependent subjects (AD), respectively. Asterisks represent means, lines within violin plots represent interquartile range, and width of violin plots represents distribution density.
Reward Processing in Alcohol Dependence have obscured group differences in neural responses-they rather appear to have amplified them. In light of this, we consider a comparable neuronal activation in mesocorticolimbic regions and the AI during gain and loss anticipation in all groups the most likely scenario.
Moreover, it remains unclear whether functional differences exist between FH 1 and FH 2 subjects with high-and low-risk drinking behaviors or between those prospectively developing and not developing alcohol dependence: even though findings are ambiguous (79)(80)(81), one study (52) reported decreased anticipatory activity in gain and loss trials in the VS only among relatives with low-risk drinking behavior. Moreover, recent studies suggest that neural activation in the VS, among others, during reward anticipation predicts future alcohol use in association with certain personality traits and other risk factors (82)(83)(84). Differences between such subgroups may have masked overall group differences and should be considered in future longitudinal studies.
Finally, the MIDT version used included only comparatively low reward and loss magnitudes, which might have limited possible group differences inand reliability ofmesolimbic activation (85). However, we observed robust whole-brain FWEcorrected VS activity in all groups. Moreover, 1V likely elicited a stronger mesolimbic response than similar magnitudes in MIDT versions using additional larger reward or loss magnitudes within the same task (86). For a detailed discussion of this aspect and of further limitations, see Supplemental Discussion.

CONCLUSIONS
Conclusively, neither the reward deficiency syndrome (26) nor the allostatic hypothesis (27)(28)(29)(30) is supported by the present findings, at least not insofar as the latter posits persisting hyposensitivity of the reward system to non-drug rewardindicating cues. Concomitantly, our data do not support the hypothesis that functional abnormalities in mesocorticolimbic regions are a predisposing factor for the development of alcohol dependence. The reduced activation in the left AI in AD subjects during loss anticipation compared with FH 2 subjects is likely due to group age differences and was no longer significant after controlling for age. Future research should investigate drug and non-drug reward cue processing over the course of abstinence and control for individual differences including sign-tracking behavior as a potential confound (87,88). This research could assess 1) whether the incentive salience hypothesis-compared with the reward deficiency syndrome and allostatic hypotheses-allows for more accurate predictions regarding mesocorticolimbic brain activity of AD and FH 2 subjects during reward and loss anticipation, and 2) which neural substrates underlie a predisposition to dependence. . Group difference in brain activity in the left anterior insula between abstinent alcohol-dependent subjects (AD) and healthy control subjects without family history of alcohol dependence (FH 2 ) during loss anticipation (loss . neutral anticipation) in the monetary incentive delay task. (A) Three-dimensional plot of activity differences (AD , FH 2 ) within regions of interest at a significance level of p , .001, uncorrected. (B) Estimated contrast weights for activity in the peak voxel of the group difference, for FH 2 , healthy first-degree relatives of alcohol-dependent individuals (FH 1 ), and AD, respectively. Asterisks represent means, lines within violin plots represent interquartile range, and width of plots represents distribution density.