Dopamine and Alcohol: A Review of in vivo PET and SPECT Studies

Alcohol consumption may lead to severe physical and mental health impairments. Approximately 3.3 million persons die annually worldwide due to alcohol-related health problems [1]. Prolonged alcohol use may result in the development and persistence of alcohol use disorder (AUD), which is associated with high cost burn on the society and high relapse probabilities for the individual [1]. The pathogenesis of AUD is considered to include a complex interplay of social, genetic, neurobiological, and familiar factors [2‒7]. Positron emission tomography (PET) studies allow a functional approach on a molecular level and have been used in the past decades to gain insights on the underlying neurochemical processes of AUD [8]. Several neurotransmitter systems are affected by alcohol consumption, and especially alterations of the dopaminergic (DA) system have been found to play a crucial role in the pathogenesis of AUD [8‒12].

To investigate different levels of alcohol consumption and AUD, we include in this review all PET studies investigating the DA system in acute alcohol consuming healthy controls (HCs), social drinkers, individuals at high risk, and patients with severe AUD [13]. As most of the included studies were conducted before the implementation of the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5), the categorical term “alcohol dependence (AD)” as defined by the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) is used when reporting the results of respective studies [13, 14]. AD corresponds to a moderate-to-severe AUD as defined by the DSM.

Individuals at high risk are defined as individuals determined by the Alcohol Use Disorder Test (AUDIT) to have a cutoff score of 8 points [15]. Participants with a lower AUDIT score are described as social drinkers with a low-to-moderate consumption, depending on the amount and regularity of alcohol intake often assessed with the timeline follow back test [16]. Additionally, the term “binge drinkers” describes individuals with at least four binge drinking episodes in the past month, and heavy drinkers are defined as persons who consume five or more drinks on any day in men (or at least 15 drinks per week) according to the “National Institute on Alcohol use and Alcoholism” (NIAAA) [17].

Specifically, we included all available studies investigating acute alcohol administration, as well as psychostimulant administration in AUD and its impact on dopamine release. Furthermore, PET studies investigating pre- and postsynaptic alterations of the DA transmitter system, such as impaired dopamine receptor (DR) availability and dopamine transporter (DAT) and DA synthesis capacity were reviewed. Due to the abundance of several confounders of the results of molecular imaging methods of the DA system, we evaluated several potentially confounding factors, such as nicotine consumption, age, and abstinence duration. To be able to track continuous changes in DA neurotransmission potentially associated with different levels of alcohol-associated behaviors and different brain regions involved in the development and persistence of AUD, we report PET results on striatal and extrastriatal regions of interest (ROI). Additionally, associations with clinical scales were reviewed to identify specific neurochemical correlates of the development of AUD. Furthermore, our understanding of whether DA changes are the result of adaptation processes or predisposing (e.g., epi/genetic) factors is limited. Therefore, we address the associations with genetic factors and evaluate the potential influence of sex/gender differences on the DA system in PET studies of persons with AUD.

To the best of our knowledge, this review is the first to explore molecular imaging results on the development and persistence of AUD, regarding clinical associations and confounding factors. This review seeks to give an overview of PET studies and their pitfalls in investigating AUD to provide implications and directions for future studies as well as their clinical consequences.

Acute alcohol consumption leads to a DA release in the striatum [18]. This DA release in the ventral striatum/nucleus accumbens contributes to the reinforcing effect of alcohol consumption and thus to repeated drug consumption [19, 20]. Striatal dopamine release as an acute effect of alcohol has been investigated in several in vivo PET studies (see Table 1) either after oral alcohol administration or intravenous alcohol infusion, in comparison with a placebo. Two PET scans, one before and one after the alcohol challenge, were used. We included 13 PET studies with a total of 271 study participants (see Table 1). Boileau et al. [19] (2003) conducted the first human study revealing a DA release in the nucleus accumbens of social drinkers after an oral alcohol challenge. Most of the following studies were able to replicate this striatal DA release after oral or intravenous alcohol administration – measured indirectly via pre-to-post reduction of radiotracer binding to D2 receptors, which is assumed to be caused by increasing competition between the radiotracer with endogenous DA for binding to the DA receptor during the second PET scan [21‒26]. Most of these studies included HCs with a low level of alcohol consumption or moderate social drinkers; one study investigated heavy drinkers [23], whereas one other study included participants with AUD [22].

Regardless of drinking behavior, most studies (9 out of 13) were able to confirm a DA release after alcohol intake (in comparison with placebo). Whereas most studies reported a significant DA release in striatal regions [19, 21‒26, 30], only one study reported a significant DA release in both the prefrontal and striatal regions [27]. However, in four studies, no DA releasing effect of alcohol administration was reported [28, 29, 31, 32]. Two of these studies focused their investigation on the extrastriatal (mostly prefrontal) brain regions [28, 29, 33‒35] instead of the striatal regions, which may have contributed to these results. In any case, the strongest effects were observed in the ventral striatum/nucleus accumbens [21‒23, 25, 30], which is consistent with findings from animal studies [18].

Taken together, robust evidence supports the striatal DA release after acute alcohol administration among social drinkers, particularly in the ventral striatum. Contrasting findings on the effects of acute alcohol intake on extrastriatal regions have been reported; therefore, further investigations are needed for clarity. As should be noted, most studies included HCs with low-to-moderate alcohol consumption. A comparison of social drinkers with teetotalers would help better understand differences in the acute effects of alcohol on the DA release and its differential effects on the alcohol consumption patterns. Future studies should include this approach in their study concept and during the recruitment process.

Since neuroadaptive changes in DA neurotransmission following drug-related DA release in the ventral striatum/nucleus accumbens and other brain areas have been reported [20, 36], it is of interest to investigate DA release after substance intake in persons with AUD and HCs. Due to a drug-induced excess DA release in persons who consume alcohol chronically, counter-regulatory neuroadaptive changes have been hypothesized to limit or even blunt DA release elicited by alcohol. To be able to investigate DA release in detoxified patients with AUD without exposing them to alcohol, psychostimulants can be used as they are known to cause drug-induced striatal DA release. Psychostimulants used for this purpose are amphetamine, which has been associated with ventrostriatal DA release in HCs [37, 38], and methylphenidate (MP), which increases striatal DA levels by blocking DAT and is therefore associated with subjective rewarding effects [39, 40]. Several studies have investigated the effect of psychostimulants on DA release in patients with AD and HCs. A baseline PET scan and a PET scan after the administration of the psychostimulant provide indirect information on DA release due to the difference in the radiotracer binding potential (BP). Most of the psychostimulant studies (seven PET studies, n = 205) used MP, whereas few studies used amphetamine to investigate their effects on the DA system in AD (see Table 2). Only one study used remifentanil, which is an opioid, as a psychostimulant. All studies suggested a DA release after the administration of the psychostimulant [40, 41, 42], [43‒46]. Most studies (6 out of 7) measuring in the striatum reported a DA release in this region, whereas only two studies observed an extrastriatal DA release after the psychostimulant challenge [42, 45]. When comparing persons with AD and HCs, some studies (4 out of 7) observed a lower DA release in participants with AD than in HCs [41, 42, 44, 45]. This lowered DA release in patients with AD was particularly observed in the striatum [44, 45] and in one study specifically in the sensorimotor striatum [41]. Furthermore, one of the studies found a lowered DA release in the extrastriatal (cortical, prefrontal) regions of persons with AD [42]. The finding of lowered striatal DA release in AD may be a neurobiological correlate of the blunted reward system in AD associated with a lower reactivity of DA neurons and hence a reduced DA release elicited by stimuli associated with natural and substance-related rewards [36, 47]. In contrast, other studies were unable to find differences in striatal DA release after psychostimulant administration between patients with AD or binge drinkers and HCs [43, 46]. With regard to the lack of differences in DA release in binge drinkers, the significant differences may be apparent in participants with manifest diagnosis of AD but not in binge drinkers as the clinical differentiation of binge drinker and HC with low consumption might not be that clear [46]. The other study that was unable to observe significant differences between patients with AD and HCs applied was the study in which remifentanil, an opioid, was used. Therefore, the use of an opioid could have contributed to the varying results [43].

Taken together, evidence suggests that striatal DA release could be due to the administration of psychostimulants, such as MP and other amphetamines, in persons with AD and HCs. Further mixed evidence suggests a lowered DA release in persons with AD compared to HCs (3 out of 7 studies) [42, 44, 45]. Altogether, findings are not clear yet as to whether the psychostimulant-induced DA release is lower in patients with AD than in HCs and whether the potential differences reliably correlate with clinical alcohol-related measures.

The availability of striatal DA D2 and D3 receptors has been investigated mostly among persons with AD and HCs in several studies conducted in the past decades. Three single-photon emission tomography (SPECT) and 11 PET studies were found to have investigated DA D2 and/or D3 receptors in AD. These studies included 379 participants (191 abstinent alcohol-dependent participants, 169 HCs, and 19 individuals at high risk; duplicates using the same sample were excluded).

Volkow et al. [55] (1996) were the first to conduct an in vivo PET study on AD, observing a reduced D2 receptor availability among participants with AD, compared to HCs. Furthermore, Repo et al. [54] (1999) investigated DA D2 receptor availability in striatal and extrastriatal ROIs but did not find significant differences between patients with AD and HCs.

This finding is consistent with the findings of Guardia et al. [53] (2000) who did not find differences between D2 receptor availabilities in the basal ganglia of persons with AD and HCs. Interestingly, the authors found an association between higher levels of DA D2 receptor availabilities and early relapse in patients with AD, compared to those who abstained from alcohol. Additionally, patients with AD presenting with physical withdrawal symptoms had higher levels of striatal DA D2 receptors than patients with AD without these symptoms [52]. This study did not include HCs. Volkow et al. [51] (2002a) conducted two studies using 11C-raclopride and observed reduced DA D2 receptor availabilities in patients with AD, compared to HCs. As the same group of patients with AD was evaluated in two different stages of detoxification, these studies reported lowered D2 receptor availabilities in the caudate and putamen in patients within 1 month of abstinence and persisting reductions in D2 receptor availabilities in the caudate but not the putamen after 2–4 months of abstinence. In directly comparing patients with AD at the stages of early and late detoxification, no significant changes were observed [51]. In 2004, Heinz et al. [50] showed reduced DA D2 and 3 receptor availabilities in the ventral striatum in patients with AD, compared to HCs. In accordance with these findings, Martinez et al. [41] (2005) reported significant reduction in densities of DA D2 receptors measured using 11C-raclopride in the striatum of patients with AD and HCs. Methodologically, unlike the aforementioned studies, Martinez et al. [41] (2005) used the sensorimotor, associative, and limbic striata as functional subdivisions in their study and were able to observe reductions in all three subregions. The limbic striatum corresponds with the anatomic ventral striatum, the associative striatum corresponds with the precommissural dorsal putamen, and the caudate and sensorimotor striatum corresponds with the postcommissural caudate and putamen, as described primarily by Mawlawi et al. [68, 69] (2001). These functional divisions of the striatum were defined according to the review on preclinical studies by Joel and Weiner (2000) [70].

Rominger et al. [49] (2012) were the first to use 18Ffallypride, but did not find significant differences in striatal DA D2/3 receptor binding between persons with AD and HCs. Interestingly, they found significant differences among several extrastriatal brain regions. Moreover, they found an age-related loss of DA D2/3 receptors, which was more pronounced in AD, than in the control group. In comparison, Erritzoe et al. [48] (2014) specifically investigated dopamine D3 receptor availabilities via 11C-PHNO in striatal and extrastriatal ROIs and reported higher dopamine D3 receptor availabilities in the hypothalamus of patients with AD than in that of HCs. Additionally, they did not find striatal alterations in the group comparison. Spitta et al. [33] (2021) found significantly lower dopamine D2/3 receptor availability in the rostral anterior cingulate cortex and left ventrolateral prefrontal cortex of persons with AD than in those of individuals at high risk. This study and that of Gleich et al. [71] (2020), using the same sample, included a more dimensional approach by investigating three levels of alcohol consumption in their 18F-fallypride study. Compared to HCs and individuals at high risk, patients with AD had reduced dopamine D2/3 receptor levels in the caudate. The most recent 11C-PHNO PET study by Chukwueke showed generally reduced dopamine D3 receptors in patients with AUD, compared to HCs, in all the examined striatal and extrastriatal regions (pallidum, substantia nigra). Albeit, only the reductions in the sensorimotor striatum were significant, although the effect did not remain significant after Bonferroni corrections [34].

Taken together, most PET studies reveal significant reductions in striatal dopamine D2/3 receptor availabilities among recently detoxified persons with AD, compared to HCs, which is consistent with findings of reviews and meta-analyses on in vivo PET studies [8, 9, 11, 28, 72]. This finding is consistent with findings from a series of studies showing a blunted dopamine D2 receptor availability in active alcohol users [33, 73] but not in persons with a longer duration of abstinence [49, 51]. Some animal models even suggest that recovery of DA neurotransmission during prolonged abstinence leads to a hyperdopaminergic state [74]. Since alterations in DA receptor sensitivity and density might recover with abstinence [74, 75], duration of abstinence seems to substantially influence the measurement of DR availabilities in PET and SPECT studies as it varies greatly among studies reported in this review. Most in vivo studies among persons with AD were conducted during early abstinence, but reduction in dopamine D2 and D3 receptor densities have been reported in studies among persons with different durations of abstinence (mean duration: 57 [2–79] days) since their last alcohol consumption. One study reported elevated densities of dopamine D3 receptors and the longest duration of abstinence (>400 days) among patients with AD [48]. In three of the studies, a second PET scan was performed for patients with AD and prolonged abstinence (after 7–14 days, after 28 days, after 68 days and in one of the studies a third scan after 12 months was performed) [49, 51, 52]. As should be noted, sample sizes with follow-up PET scans were small as only a portion of the patients with AD abstained from alcohol. Ebert et al. [52] (2002) did not find any differences in striatal BP between the first and second PET scans (2 and 28 days) of patients with AD. Another study shows evidence of a potential recovery of DA D2/3 receptors during prolonged abstinence in patients with AD (one scan within 24 h, a second scan after 7–14 days, and a third scan 12 months later in abstinent patients with AD) [49]. Nonetheless, Volkow et al. [51] (2002a) reported a persistent striatal D2/3 receptor reduction in patients with AD after 4 months of abstinence. Whether dopamine D2/3 receptors recover significantly during prolonged abstinence is currently not clarified; therefore, more long-term PET studies are needed to resolve this challenge.

In observing the subregions of the striatum, several studies found a reduction of dorsostriatal BP_ND (putamen and nucleus caudate) [40, 51, 71]. Other studies were able to show ventrostriatal impairments in patients with AD, compared to HCs [41, 50]. In a review, Koob and Volkow [76] (2010) argue that along with the development of a more habitual, automated drinking pattern, neuroadaptive changes may shift from ventrostriatal to dorsostriatal brain regions. Therefore, impairments of dorsal or ventral dopamine D2/3 receptors in the striatum may reflect different stages of AUD. This hypothesis is supported by several animal experimental studies investigating subregions of the striatum in addiction [77‒79]. In contrast, the evidence of extrastriatal dopamine D2/3 receptor impairments is mixed as only few studies (4 out of 14 PET studies) have investigated extrastriatal ROIs [33, 43, 48, 49]. Clearly, more studies are needed to improve our understanding of the potential cortical DA alterations and their involvement in the pathogenesis of AUD. For an overview of SPECT and PET studies investigating DA receptor availabilities, see Table 2.

Changes in the availability of presynaptic DATs in patients with AD have been investigated in several studies, mostly using SPECT but not PET. These studies have led to the provision of mixed evidence.

We found eight SPECT studies and one PET study investigating DAT among 286 participants (132 persons with AD, 14 heavy drinkers, and 140 HCs; duplicates removed). As a transmembrane protein, DAT is responsible for the reuptake of dopamine from the synaptic cleft [80, 81].

Tiihonen et al. [62] (1995) were the first to investigate DAT in AD. They found reduced DAT density in patients with AD, compared to HCs. Other studies reported similar findings of reduced DAT density in AD [54, 61]. In contrast, Volkow et al. [55] (1996) did not find significant alterations of DAT in patients with AD, compared to HCs. This discrepancy may be due to differences in the duration of abstinence. Laine et al. [61] (1999) found a reduction in DAT density during alcohol withdrawal, and this DAT density increased to a normal level after 4 weeks of abstinence. Furthermore, Heinz et al. [60] did not find significant differences in DAT densities in the caudate and putamen between persons with AD and HCs. Again, this finding may be explained by the duration of abstinence in this study cohort. Data were obtained after 3–5 weeks of abstinence in patients with AD, and two long-term abstinent patients had been included [60]. This finding might imply that DAT density increases with prolonged duration of abstinence. In contrast, Cosgrove et al. [59] reported an elevated DAT density in the striatum of heavy drinkers, compared to HCs.

More recent studies by Yen et al. [57, 58] show reductions in DAT density in the striatum, putamen, and caudate of early abstinent patients with AD, compared to HCs. The findings were more pronounced in patients with AD without comorbid depression [58]. This finding is consistent with the findings of Grover et al. [56] who found reduced DAT density in the putamen and striatum of patients with AD. In this study, patients with AD had been abstinent for a maximum of 30 days. See Table 2 for an overview of DAT studies among patients with AD.

Taken together, 6 out of 9 studies reported reduced DAT density in persons with AD [54, 56‒58, 61, 62]. Furthermore, 2 studies found no difference in DAT density between patients with AD and HCs [55, 60], and one study reported elevated DAT density in heavy drinkers, compared with HCs [59]. Currently, whether low DAT density implies a hypodopaminergic state remains unclear. Additionally, the duration of abstinence does not completely explain the differences among these studies. Reductions in DAT densities have been observed in early abstinence (up to 30 days) [56‒58, 61] and in long-term abstinence (≥40 days) [54, 62]. Clearly, further studies are needed to investigate whether DAT densities are reduced in early abstinence and if they increase during prolonged abstinence in AUD.

Dopamine synthesis capacity is a marker of presynaptic DA neurotransmission and allows for the quantification of DA turnover with 6-[18F]-fluorodopa (FDOPA) PET. Findings from five PET studies on dopamine synthesis capacity had been reported in the literature; these studies included 121 participants (58 persons with AD and 63 HCs). The first in vivo FDOPA PET study, conducted by Tiihonen et al. [67] (1998), reported higher DA synthesis capacity in the right caudate and bilateral putamen of persons with AD considered to be “type 1” (i.e., a usually higher age of onset and frequent depression), compared to HCs, indicating an increased dopamine turnover in AD. In contrast, Heinz et al. [66] (2005) found no differences in DA synthesis capacity of patients with AD, compared to HCs. Findings from this study are consistent with those from Kienast et al. [65] (2013) who found no significant differences in DA synthesis capacity in the ventral striatum and amygdala of patients with AD. Moreover, Deserno et al. [63] (2015) observed no impairments in the striatal DA synthesis capacity of patients with AD, compared to HCs. Yet, Kumakura et al. [64] (2013) reported a reduced DA synthesis capacity in the left caudate in recently detoxified patients with AD, compared to HCs.

Taken together, most studies (3/5) found no differences in DA synthesis capacity via FDOPA between persons with AD and HCs. The question as to whether DA synthesis is altered in prolonged abstinence cannot be answered from currently available studies. However, future studies may have to elucidate potentially hypo- or hyperdopaminergic states in prolonged abstinence. An overview of PET studies investigating DA synthesis capacity in persons with AD and HCs is provided in Table 2.

Clinical correlations are essential for evaluating the functional relevance of impairments of the DA neurotransmitter system. Regarding the acute effect of alcohol on the DA system, two of the abovementioned studies reported a positive correlation between DA release and subjective measures of “drug-liking” [27, 29] or self-reported intoxication [23]. No associations were observed between the extent of DA release and drinking patterns [24, 31] or the AUDIT score [26]. Interestingly, Urban et al. [25] (2010) found positive clinical correlations between ventrostriatal DA release and subjective activation and drinking patterns in men but not in women, which will be further discussed in the section sex/gender differences. Yoder et al. [30] (2016) observed a significant DA release following alcohol administration in non-treatment seeking patients with AD but not in HCs, whereas Kegeles et al. [22] (2018) found no differences between patients with AUD and HCs with or without family history of AUD.

Moreover, some studies observed the clinical correlations of psychostimulant-induced DA release, which was operationalized as the reduction in the radiotracer BP following drug administration, including subjective effects of the substance (feeling “high, happy, restlessness, drug good”) [44]. Spreckelmeyer et al. [43] (2011) observed a significant correlation of the extent of reduced receptor binding due to opioid administration with higher AUDIT scores in the ventral striatum of patients with AD. Furthermore, Volkow et al. [45] (2013) found a significant negative correlation between striatal DA release and relative metabolism (glucose utilization) in different brain regions, including the ACC. This correlation was found to be stronger in persons with AD, and thus the authors argue that the modulation of the DA signals may be impaired in AD [45]. In contrast, other studies found no clinical correlations between psychostimulant-induced DA release and alcohol-related behaviors [41, 42, 46].

Regarding chronic alcohol consumption, DA alterations, and related clinical symptoms, reduced dopamine D2/3 receptor availability has been associated with craving symptoms [50]. In contrast, Martinez et al. [41] found no correlation between craving symptoms and D2/3 receptors. Craving symptoms are of special clinical interest as they have been associated with relapse prediction in AUD and other substance use disorders [82‒84]. Furthermore, two studies have reported an association between alcohol craving and striatal DA synthesis capacity in AD [66, 64].

Another frequently reported clinical item is addiction-related “symptom severity” measured on the “Alcohol dependence scale” (ADS) [85, 86]. Gleich et al. [71] (2020) found the ADS score to be associated with reduced dopamine D2/3 receptor availability in the nucleus caudate, but many other studies found no correlation between an impaired striatal DA system and ADS [34, 41, 42]. Erritzoe et al. [48] (2014) observed a positive correlation between lifetime alcohol intake and dopamine D3 receptor availability in AD. Moreover, one study reported an association between DAT density and years of alcohol abuse in patients with AD, but this effect was no more significant when age was included as a nuisance regressor [55]. Yen et al. [57] (2015) found a significant positive correlation between harm avoidance and DAT availability in patients with AD.

The correlation between striatal DA turnover and craving symptoms in AD has been reported by two studies currently, suggesting a potential clinical relevance [64, 66]. More precisely, DA synthesis capacity in the bilateral putamen was inversely correlated with craving symptoms in recently abstinent patients with AD [66]. Further, Kumakura et al. [64] found a reduced DA synthesis capacity in the ventral striatum to be correlated with higher craving symptoms in recently detoxified patients with DA, compared to HCs.

Taken together, the evidence for substance-specific symptoms and DA neurotransmission is mixed and needs further evaluation. Acute DA release seems to be associated with intoxication-related subjective symptoms, such as “drug-liking,” and some studies also reported similar effects in psychostimulant studies. Symptom severity and craving symptoms have been associated with dopamine D2/3 receptor availability, and craving symptoms have been associated with DA turnover in AD. Therefore, craving symptoms specifically and their impact on relapse probabilities seem to be of importance and should be a target variable for future studies. Finding relevant clinical associations of alterations of the DA system in alcohol-consuming individuals is complicated by the limited sample sizes of PET and SPECT studies and has to be considered when interpreting study findings.

In vivo imaging of neurochemical pathways/functions and molecular targets in the brain with PET and SPECT is based on the administration of “tracers” consisting of two components, the carrier molecule and the radioactive label “built-in” or “attached to” the carrier molecule [87, 88]. The carrier molecule determines the kinetics of the tracer in the body after intravenous injection. It is selected such that the tracer molecules traverse the blood-brain barrier and then enter the neurochemical pathway/function of interest or bind to the molecular target to be investigated with high affinity and selectivity. The decay of the radioactive label aids the detection and localization of the tracer molecules in the brain from outside the body with a PET or SPECT camera [89‒91]. For PET, the radioactive labels are positron emitters, such as carbon-11 (C-11) and fluorine-18 (F-18), with physical half-life time of 20 min and 110 min, respectively. For SPECT, the radioactive labels most often are gamma emitters, such as iodine-123 (I-123), with physical half-life time of about 13 h. Both PET and SPECT provide three-dimensional images of the regional tracer concentration throughout the whole brain.

For many neurochemical pathways/functions and molecular targets, several tracers have been developed [92, 93]. For example, tracers for D2 receptor imaging include the PET tracers, C-11-raclopride and F-18-fallypride, and the SPECT tracer, I-123-IBZM [94‒96]. Tracers for the same target can differ strongly, with respect to affinity and selectivity for the target, and with respect to the permeability of the blood-brain barrier for exchange of the tracer between blood and tissue [97, 98]. Tracers with high affinity for the target (e.g., F-18-fallypride for the D2 receptor) might allow more reliable characterization of the target in brain regions with low target density (e.g., cortical brain regions) but might require long acquisition times in brain regions with high target density (e.g., the striatum) [99]. Therefore, the choice of the most appropriate tracer for the research question to be investigated is of great relevance. The use of different tracers for the same target adds variability that usually is not of primary interest but complicates the comparison and summary assessment of different studies.

A major strength of PET and SPECT is their excellent sensitivity for detection of the tracer molecules such that nM to pM concentrations of the tracer in the brain are sufficient for imaging [88]. This provides considerable flexibility in the choice of carrier molecules for specific neurochemical pathways/functions and targets because pharmacological or toxicological issues that may prevent the use of a molecule are rare. For example, I-123-labeled ioflupane, widely used as a tracer for SPECT imaging of the DAT, is a chemical analog of cocaine [100]. However, the dose of I-123-ioflupane typically used for DAT-SPECT in humans is 0.3 µg, which is several orders of magnitude below cocaine doses for recreational drug use [101].

For applications in humans, PET provides better spatial resolution and better statistical image quality (less noise) than SPECT [89]. Limited spatial resolution can hamper the demarcation of ROIs in the PET or SPECT images. Related to this, limited spatial resolution causes underestimation of the tracer concentration in small brain structures [102, 103]. The underestimation is more pronounced when the brain structure is smaller. Consequently, normal between-subjects variability in size and structure of the head can cause additional between-subjects variability of no interest in PET and SPECT of the brain that increases the risk of both false-negative and false-positive findings [104]. These limitations are usually much more pronounced in SPECT than in PET, given that spatial resolution is usually much lower in SPECT than in PET of the brain (≥8 mm in SPECT vs. ≥4 mm in PET) [89, 105, 106]. Therefore, small brain structures, such as the nucleus accumbens and small striatal subregions (e.g., the sensorimotor part), are far more difficult to assess with SPECT than with PET.

The statistical quality of PET and SPECT images depends on the total number of radioactive decays detected for the image generation. Given that the sensitivity to detect the radioactive decays in the participant’s head is about 100 times higher for PET than for SPECT (typically 1 of 100 decays is detected in conventional PET vs. 1 of 10,000 decays in conventional SPECT) [89], the acquisition duration of the scan required to achieve acceptable statistical image quality is considerably shorter in PET than in SPECT. This results in reduced time resolution of SPECT to image the time course of the regional tracer concentration in the brain after intravenous injection, which limits the possibilities of (semi-)quantitative characterization of the status of the neurochemical pathway/function or molecular target of interest by tracer kinetic modeling [107].

Tracer kinetic modeling is often based on a (simplified) mathematical model of how the status of the neurochemical pathway/function or molecular target of interest affects the time course of the tracer in the brain [108‒117]. Within this mathematical model, the status of the neurochemical pathway/function or molecular target is characterized quantitatively by a set of parameters, including rate constants and BP [114]. The aim of tracer kinetic modeling was to determine these parameters, separately for each ROI. The time course of the tracer in the brain after intravenous injection is measured by “dynamic” PET or SPECT imaging, that is, a sequence of PET or SPECT images (“frames”). The acquisition of the first image usually starts with the tracer injection. The duration of the acquisition of the first and of the following images is rather short (e.g., 10 s) to reliably track the rather rapid change in the tracer concentration in the brain during the first minutes. The acquisition duration of the subsequent image frames, when tracer concentration in the brain changes progressively slower, typically increases stepwise (up to a maximum of 5 or 10 min) [118, 119]. The total duration of the dynamic imaging strongly varies between tracers. With tracers that show reversible binding to the D2 receptor, the dynamic image sequence should capture the time point at which binding of the tracer to the receptor in the ROI is in (transient) equilibrium with release of tracer from the receptor [115]. The time point of (transient) equilibrium depends on the affinity of the tracer for the target and on the density of the target. To reliably capture the (transient) equilibrium for the high-affinity D2 receptor tracer, F-18-fallypride, in the striatum (high receptor density), dynamic PET acquisition typically covers 3 h after tracer injection (often with breaks) [118, 120]. The time course (“time activity curve”) of the tracer in each ROI is obtained by computing the mean tracer concentration in the ROI separately in each image frame. The time activity curve of the ROI depends on the status of the neurochemical pathway/function or molecular target of interest. However, the time activity curve of the ROI strongly depends on the supply of tracer to the brain via the arterial blood, the “input function.” Additionally, the latter relationship is considered by the mathematical model for tracer kinetic modeling. Therefore, to estimate the model parameters for characterizing the status of the neurochemical pathway/function or molecular target, the time course of the tracer in arterial blood during the entire dynamic image sequence needs to be known. This requires sampling of arterial blood during the entire dynamic image sequence. Usually, a combination of automatic blood sampling (with a dedicated pump) during the first minutes after tracer injection followed by manually drawing arterial blood samples until the end of the dynamic imaging sequence is used [121]. If the tracer undergoes peripheral metabolism, resulting in radioactive metabolites in the blood [122], the amount of unmetabolized (parent) tracer must be determined for each blood sample (e.g., using high-performance liquid chromatography) since the input function is given by the time course of the unmetabolized (parent) tracer in arterial blood [123]. In fact, the input function is given by the free fraction (not bound to plasma proteins) of the unmetabolized tracer in plasma since only free tracer molecules in arterial plasma are available for transport into the brain tissue during a capillary passage. Therefore, the free fraction of unmetabolized tracer must be determined as well [124].

In PET or SPECT of molecular targets, such as the D2 receptor and DAT, using a tracer that shows reversible binding to the target, tracer kinetic modeling usually aims at estimating the BP. The (in vivo) BP was first defined by Mintun et al. [125] as BP = B_avail/K_D, where B_avail is the density of specific binding sites (e.g., D2 receptors) available for binding of the tracer, and 1/K_D is the affinity of the tracer (e.g., F-18-fallypride) for the binding site [114]. Between-subjects differences or within-subjects changes in BP are usually assumed to indicate between-subjects differences or within-subjects changes in availability of binding sites, if the affinity of the tracer for the binding site does not differ between subjects/conditions and is not affected by interventions (e.g., administration of psychostimulants). Estimation of the BP requires dynamic imaging, starting from the time of tracer injection and long enough to reliably capture the (transient) equilibrium in tissue, sampling of arterial blood during the whole dynamic imaging session, and correcting each blood sample for radioactive metabolites and plasma protein binding of unmetabolized tracer molecules. This is not only “expensive” (in terms of scan time and staff requirements) but also presents considerable discomfort to patients or study participants. Furthermore, this approach is prone to accumulate (statistical and systematically) errors from various measurements [126].

Therefore, simplified methods have been developed to estimate surrogates of the BP. So-called reference tissue methods 119, 127-131 do not require blood sampling but instead “compare” the time activity curve in the ROI to the time activity curve in a “reference” region in the brain [9, 10]. The reference region is assumed (i) to be void of the imaging target (e.g., the D2 receptor) and (ii) to show the same non-displaceable (by complete blocking of the target) tracer kinetics as the ROI [128, 132]. Reference region methods usually aim at estimating the “non-displaceable” BP (BP_ND) that differs from the actual BP by a factor, that is, given by the (unknown) free fraction of non-displaceable tracer in tissue (that is, tracer in tissue not bound to the target) [114]. Therefore, for between-subjects and between-conditions differences of the “non-displaceable” BP (BP_ND) to indicate between-subjects differences or (within-subjects) between-conditions change of the target density, further assumptions ought to be made (in addition to the assumptions made when interpreting the actual BP). Additional assumptions include neglect of between-subjects differences and (within-subjects) between-conditions’ changes of the tracer kinetics in the reference region. The same assumption is made for the concentration of non-displaceable tracer in the ROI.

Conventional reference tissue methods [130, 133] require dynamic imaging to measure time activity curves in the ROI and in the reference region and, therefore, are still “expensive” in terms of total scan time. Ratio methods can be used to estimate the non-displaceable BP from a single static image acquisition (of rather short duration, typically 5–20 min) at a predefined late time point after tracer injection, typically by computing the ratio of the tracer concentration in the ROI to the tracer concentration in the reference region minus 1 in the late image [133, 134]. This eliminates the need for dynamic imaging to fully monitor time activity curves from tracer injection until after (transient) equilibrium [133]. Ratio methods are often used with SPECT since the limited time resolution of SPECT does not allow reliable tracking of the fast kinetics of the tracer in the brain early after intravenous injection [135]. The fixed time point for the single static image acquisition is usually chosen with the intention to image at the (transient) equilibrium. When using ratio methods, further assumptions are made (in addition to the assumptions of conventional reference tissue methods) to attribute differences and changes in the corresponding BP_ND estimate to differences and changes in the target density. By assumption, the impact of between-subjects variability of the time point at which the (transient) equilibrium is reached can be neglected, and interventions do not shift the (transient) equilibrium on the time axis. Furthermore, BP_ND estimates from ratio methods can be sensitive to changes in regional cerebral blood flow and to changes (of the shape) in the input function (e.g., due to differences in peripheral clearance of the tracer from blood to urine).

Taken together, PET and SPECT methodologies can have a profound impact on the utility of the outcome measure (e.g., the BP) to characterize between-subjects differences in the molecular target of interest. As a rule of thumb, more “expensive” methods are more reliable for this task since less assumptions are required. Tracer kinetic modeling of tissue time activity curves acquired by dynamic PET or SPECT together with arterial blood sampling and correction of radioactive metabolites and binding to plasma proteins is the gold standard. However, this method is quite “expensive.” Ratio methods are particularly “cheap” but depend on many assumptions. A high risk of both false-positive and false-negative findings exists, if one or more of these assumptions are violated. Conventional reference tissue approaches using dynamic imaging data without blood sampling are in between [128, 132]. Additionally, this applies to F-18-FDOPA PET, which is widely used to investigate DOPA metabolism in the brain [136, 137]. Simplified methods, including reference tissue methods, assume irreversible trapping of decarboxylated FDOPA metabolites in the brain, neglecting possible (e.g., age-related) loss of vesicular capacity to store decarboxylated FDOPA metabolites [136, 138‒140]. To account for the latter, “expensive” dynamic F-18-FDOPA PET with arterial blood sampling and separation of metabolites in the blood samples is required [138].

Computing the difference in a given outcome measure (e.g., the BP) between two short-term repeat imaging sessions (e.g., after minus before administration of a psychostimulant) does not only eliminate between-subjects variability of the target of interest at baseline (before stimulation) but also reduces the impact of systematic errors caused by violation of the assumptions associated with the chosen modeling approach. Together, this is expected to result in increased statistical power, which might explain the fact that the reported findings on DA release in AUD appear more consistent than the reported findings on D2 receptor availability in AUD.

Tracer kinetic modeling most often is performed on the level of ROIs, that is, on time activity curves of the ROIs obtained by averaging the tracer concentration over all image voxels in each ROI, separately in each image frame. There is considerable variability with respect to the ROIs selected for the analysis among studies. When investigating the status of the DA system in the striatum, most studies used “anatomical” ROIs, such as the putamen, caudate nucleus, and nucleus accumbens, as defined by e.g., the WFU Pick Atlas (http://fmri.wfubmc.edu/-software/PickAtlas) [71, 140] or the Talairach and Tournoux Atlas [50, 141]. Other studies implemented “functional” subdivisions of the striatum, such as the sensorimotor, associative, and limbic parts [34, 41, 43, 48, 68, 69, 142‒144]. This variability complicates the comparison of the findings among studies. Additional variability results from using unilateral ROIs, separately in each brain hemisphere versus using bilateral ROIs, averaging the imaging signal in the considered ROI over both hemispheres. Compared to bilateral analyses, unilateral ROIs double the number of analyses and, therefore, require that statistical issues associated with multiple testing are addressed particularly carefully [33]. Strict application of Bonferroni correction might not be most appropriate for this purpose, particularly when the considered ROIs are not independent [33, 34, 44]. In the latter case, the risk of rejecting a true effect because it does not remain significant after Bonferroni correction might increase.

When reference region methods are used for tracer kinetic modeling, additional variability results from using different reference regions. Most studies of the DA system used the cerebellum as reference region [48]. However, the cerebellum may undergo volume loss (atrophy) in chronic AUD [145, 146] and, therefore, might not be the most suitable reference region in PET and SPECT studies in AUD. Some recent studies used the bilateral superior longitudinal fasciculus, as defined by the white-matter tractography atlas of the Laboratory of Brain Anatomical MRI of Johns Hopkins University [147, p. 200] as reference region in PET with F-18-fallypride [33, 71, 142, 143]. The rationale for this choice was that the bilateral superior longitudinal fasciculus might improve the statistical power of F-18-fallypride PET to detect group difference of D2 receptor availability [148], presumably due to lower between-subjects variability of no interest of the F-18-fallypride kinetics in the bilateral superior longitudinal fasciculus compared to the cerebellum.

The smallest meaningful size of the ROI depends on the sensitivity of the selected tracer kinetic modeling method, with respect to statistical noise. (Non-linear) tracer kinetic modeling methods often converge and provide reliable results with rather large ROIs only (e.g., whole putamen). (Linear) reference tissue methods can be quite stable, with respect to statistical noise in the PET or SPECT images and, therefore, can be used with small ROIs down to the level of single image voxels [149]. In the latter case, tracer kinetic modeling results in three-dimensional parametric maps. In these parametric maps, the intensity of a given voxel represents a specific outcome parameter of tracer kinetic modeling (e.g., BP_ND) for the time activity curve of this voxel. Parametric maps can be compared between subjects and/or conditions using voxel-based statistical testing to identify effects without a prior hypothesis about their localization in the brain [150]. ROI-based analyses based on a priori hypotheses and exploratory voxel-based testing are rather complementary approaches than being mutually exclusive alternatives.

Age-related changes in the DA system have been reported in several studies [151], and a recent review revealed significant effects of age on DRs and transporters but not on DA synthesis capacity [152]. Although, DA storage capacity seems to be affected by an age-related decline in some studies [138].

Several studies seem to report an age-related loss of DAT [153, 154]. Furthermore, age-related reduction in D2/3 availability in the population has been a robust finding in several studies over the past decades. Besides age-related reductions in DA receptor densities in the striatum, extrastriatal D2-like receptors seem to be affected as well [155, 156]. As shown in a recent meta-analysis by Karrer et al. [152] (2017), age seems to have an enormous effect on D1- and D2-like receptors. Karrer et al. [152] (2017) have calculated the extent of the age-related loss of dopamine D2 receptor in the striatum per decade, age-related loss of DAT, and DA synthesis capacity to be 8%, 8.9%, and 1.4%, respectively, based on 89 PET studies. In comparison, alcohol-related effects on D2 receptor availability vary greatly among the studies and have, for example, been calculated to be a 20% reduction [41] or 12.3% reduction [34] in the sensorimotor striatum of persons with AD. According to Gleich et al. [57] (2020), differences may be calculated to be about 7.5% reduction in the putamen and 6.9% in the caudate of patients with AD. The DAT reduction in persons with AD may be calculated to be approximately 19.6%.

With respect to AD, Guardia et al. [53] (2000) showed an age-dependent loss of dopamine D2 receptors in persons with AD and HCs. Additionally, Rominger et al. [49] (2012) showed that the age-related loss of DR2/3 availability was more pronounced in patients with AD than in HCs. This finding warrants further investigation and emphasizes the importance of matching patients with AUD and HCs by age and including age as a nuisance regressor in the analyses, as recommended in recent studies [33, 43, 53, 71, 157].

Tobacco use disorder is a common comorbidity of AUD [158]. In most of the studies included in this review, most patients with AD were smokers. Dopamine D2/3 receptor availability seems to be affected by smoking status. Fehr et al. [159] (2008) showed that nicotine-dependent smokers had significantly lower dopamine D2/3 receptor availabilities than non-smokers. Albrecht et al. [160] (2013) investigated the effects of smoking status on D2/3 receptors in social drinkers and patients with AD and found a reduction in D2/3 receptor availabilities in smokers, regardless of their alcohol consumption habits. This finding was replicated by Wiers et al. [161] (2017), showing significantly lower dopamine D2/3 receptor availabilities in the caudate and putamen of smokers than in those of non-smokers.

DAT availability seems to be affected by smoking status as well, as confirmed by a recent meta-analysis study [162]. Cosgrove et al. [59] reported a lower DAT density in heavy drinking smokers than in heavy drinking non-smokers, indicating the possible effect of smoking status on individuals at high risk of developing AD.

Other comorbidities, such as substance abuse, may influence the DA system [11]. In nearly all cited SPECT and PET studies, participants with other substance use disorders – despite presence of tobacco use disorder – were excluded. Therefore, smoking status is relevant and should be reported in the descriptive statistics, and reference groups should be matched by duration of abstinence or at least last tobacco use before scanning. Preferably, smoking status should be included as a covariate in PET data analyses targeting the DA system, as recommended by some studies [33, 71].

There seem to be sex/gender differences in alcohol-related behaviors and their neurobiological correlates as indicated by a recent review of preclinical and clinical studies [163]. However, only a few PET studies indicate the possibility of differences in the DA neurotransmitter system between men and women [25, 164]. Other studies report no sex differences [22]. More precisely, Pohjalainen et al. [164] (1998) observed lower striatal D2 receptor availabilities and greater age-related D2 receptor decline in women than in men. Urban et al. [25] found higher DA release in young men than in women after exposure to similar amounts of (body weight-adapted) oral alcoholic beverage. In contrast, Kegeles et al. [22] (2018) found no effect of sex/gender on striatal DA release after acute alcohol intake in patients with AUD and HCs.

Interestingly, a recent study on chronic alcohol consuming rhesus macaques found reduced striatal DA release in male monkeys but not in female monkeys [165]. Additionally, the dorsostriatal dopamine D2/3 receptor function was lower in male alcohol-drinking monkeys than in the control group but not in females. This finding raises the question of potentially similar sex/gender differences in the DA neurotransmitter system of humans. This question is of particular interest as the number of women with AUD has rapidly increased over the past years. Additionally, this effect seems to strongly contribute to the rising prevalence of AUD [166, 167]. Sex/gender differences in PET neuroimaging studies have been recently reviewed, and this field is considered to be severely understudied [168]. In the reviewed PET studies on acute alcohol consumption and DA release, 25.8% of the study participants (70 females, N = 271) were females. Furthermore, 8.8% of the study participants (106 females, N = 930) in studies investigating AD (DR, DAT, DA synthesis, and psychostimulant studies) were female. This finding emphasizes that further studies with higher numbers of female participants are needed to better understand the neurobiological mechanisms in women and clarify potential sex/gender differences in the DA system and their contribution to the development of AUD.

Currently, our understanding is limited, regarding whether DA system impairments are solely due to adaptive processes, resulting from excess DA in prolonged alcohol consumption or reflect predisposing factors that are, for example, genetically determined. Several DA polymorphisms have been reported and critically reviewed, with respect to their functional relevance [169]. Regarding dopamine D2 receptor function, a possible effect of the ANKK1 TaqIA allele status (rs1800497) has been discussed (for a review, see Gluskin and Mickey [170], 2016). More precisely, the minor allele (A1) seems to be associated with lowered dopamine D2 receptor availability. This potential genetic effect was first reported by Blum et al. [171] (1990); however, results vary, and several studies did not replicate this finding. Regarding AUD, postmortem studies have reported a link between ANKK1 TaqIA A1 allele status and dopamine D2/3 availability in patients with AD [172]. However, the finding regarding the blunted DR sensitivity in persons with AD was not replicated [75]. Additionally, no significant effect, regarding dopamine D2 receptor availability, was observed in persons with AD [75].

Moreover, reduced DAT availability has been linked to alterations in VNTR polymorphism in some studies. Heinz et al. [60] found a reduced DAT density in the putamen to be associated with the heterozygous VNTR polymorphism of the DAT gene. Furthermore, Grover et al. [56] found reduced DAT availabilities to be associated with the VNTR methylation status but not with the VNTR genotype.

Only few PET/SPECT studies have investigated genetic factors, which may be due to the resulting power issues that stem from small sample sizes, a high number of small subgroups due to different genotypes, and different phenotypes that need to be assessed [75]. The mean sample size of all PET and SPECT studies included in this review was 27 participants (N = 1,201, 45 PET/SPECT studies, see Tables 1 and 2). This sample size is small, when considering smaller subgroups (AUD/HC). One option to acquire larger datasets for combined genetic and molecular imaging studies is collaboration among research groups and pooling of different datasets.

Currently, only few studies have investigated the DA system using structural or functional MRI or MRS in patients with AUD or acute alcohol consumption. In fMRI studies, alcohol-related cues seem to evoke greater reactivity than neutral cues in the mesocorticolimbic circuit of patients with AUD and hyperactivations in cortical regions, such as the medial PFC and ACC of AD, compared to HCs (for meta-analysis, see Zeng et al., 2022 [173]). However, multimodal studies combining findings of an altered reward system via structural or functional MRI with DA correlates via PET were found to be scarce.

Kienast et al. [65] (2013) observed a significant correlation between DA synthesis capacity in the left amygdala and activation in the left ACC in HCs but not in persons with AD. Moreover, in persons with AD, functional connectivity between amygdala and ACC during processing of aversive cues was reduced, which may limit emotion control [65]. Deserno et al. [63] (2015) found no differences between patients with AD and HCs, regarding activation in the ventral striatum due to reward prediction errors or striatal DA synthesis capacity. However, DA synthesis capacity was only significantly correlated with functional activation in the same brain area in HC but not in persons with AD.

Other in vivo multimodal studies measured brain metabolism (glucose use) via 18FDG-PET in persons with AD [94‒96]. Some studies reported reduced global brain metabolism and especially a hypometabolism in the frontal cortical regions of persons with AD, compared to HCs [174, 175]. Interestingly, Maillard et al. [176] found a hypermetabolism in the cerebellum and hippocampus of persons with AD who relapsed within 6 months and additionally in the anterior cingulate cortex of persons with AD who relapsed after 12 months. Further (preclinical) multimodal studies combining PET and MRI in AD investigated glutamatergic but not DA neurotransmission in animal models [177, 178].

To our knowledge, no other in vivo studies have investigated the DA system in AD using PET and its associations with functional MRI or other neurotransmitter systems measured with MRS. Since several studies indicate the potential effects of glutamate on GABAergic interneurons on striatal DA in healthy participants [179] and in mental disorders, such as schizophrenia [180, p. 199] [181], further studies assessing glutamate concentrations and their associations with, for example, striatal dopamine D2/3 receptors in AUD. Clearly, more multimodal studies are needed to embed the DA findings into the complex functional and structural neurobiological changes associated with the development and persistence of AUD.

With the aim to evaluate the quality of the reviewed studies in Table 2, ratings were conducted. Due to the heterogeneity of studies summarized in Table 1 (randomized and nonrandomized/single-group and multiple groups), we decided to not rate the studies due to the lack of one proper rating tool to rate all these type of studies (including an alcohol challenge as in Table 1).

To evaluate bias risk of included nonrandomized cross-sectional case-control studies (Table 2), the authors GS and MG independently assessed each study using the Newcastle-Ottawa Quality Score Scale (NOS) [182]. The two authors finally met with the aim to find a consensus about their individual evaluations (see Table 3). NOS rates nonrandomized case-control studies using eight items categorized into the three following groups: (1) selection of study participants, (2) population comparability, and (3) verification as to whether exposure or outcome includes any risk of bias, selection bias, or bias of response rate between the groups. The NOS sum score ranges from 0 to 9, and studies with scores ≥7 are considered high quality [9]. The selection group comprises four items, with 0–1 point, respectively, leading to a maximum score of 4 points. Comparability has two items with a maximum value of 1 point, leading to a maximum score of 2 points. Exposure comprises three items with a maximum value of 1 point, leading to a maximum score of 3 points.

As a result of the evaluation, we found 14/30 studies to be of high quality (NOS score 7 or higher, see Table 3, [9]). Importantly, this does not necessarily mean that studies below NOS score 7 are biased or of bad quality. It might be that information was just not conducted by the respective authors, thus risk of biased studies might be lower [183].

In this review, we evaluated 35 in vivo PET and 10 SPECT studies (N = 1,201) on DA neurotransmission related to different levels of alcohol consumption and stages of development of AUD. In most studies, acute alcohol consumption increased DA release in cortical (anterior cingulate, orbitofrontal, and prefrontal cortex) and subcortical (striatum, nucleus caudate, putamen, nucleus accumbens/ventral striatum) brain regions (9/13 studies, see Table 1). The administration of a psychostimulant to HCs and persons with chronic alcohol consumption habits (manifest AD or binge drinker) leads to a DA release in cortical (anterior cingulate, orbitofrontal, and prefrontal cortex) and subcortical (amygdala, pallidum, putamen, and [ventral] striatum) brain regions (7/7 studies, see Table 2). Reduced DA D2 and/or D3 receptor availabilities in patients with AD compared with HCs (9 of the 14 studies) was reported in subcortical (associative, limbic, sensorimotor striatum, [ventral] striatum, nucleus accumbens, hippocampus, insular cortex, putamen, thalamus, and thalamic cortex) and cortical (anterior cingulate and prefrontal cortex) (only in one study) brain regions (see Table 2). Reduced DAT availability in persons with AD compared to HC was reported in six out of nine studies [54, 56‒58, 61, 62]. However, the ability of DAT recovery during early abstinence remains unclear (see Table 2). DA synthesis capacity did not differ significantly between patients with AD and HCs (3/5 studies, see Table 2), although withdrawal symptoms have been associated with reduced DS (2/5 studies) [66, 64].

Currently, our understanding of the changes in the DA system as the sole result of a compensatory downregulation or genetic predisposition remains limited. One challenge limiting our understanding is the difficulty associated with genetic association studies on PET samples; one of such difficulties is the reduced power, resulting from the use of small sample sizes. Additionally, the influence of ROI and radiotracer selection is crucial, and the great variability within the studies contributes to reduced comparability of results. Moreover, smoking status and age should be included as covariates in the analyses of DA D2/3 receptor availabilities. As only 8.8% of study participants with AUD in PET studies were female, studies with a higher number of females are required to improve our understanding of the neurobiological mechanisms in women to clarify potential gender or sex differences.

In total, challenge studies (acute alcohol or psychostimulant administration) seem to be more consistent in their findings and might be less prone to the effects of confounders. For future acute alcohol or psychostimulant challenge studies, evaluation of the dose-dependent effect of alcohol consumption and severity of AUD (as determined by the DSM-5) on the DA system may be potentially beneficial [13].

The DA system seems to be differently impaired during the development and persistence of AUD. Yet only few studies compared the effect of different levels of alcohol consumption within one study. The recruitment of persons with different drinking patterns is one option to investigate specific differences related to the development and persistence of AUD [33, 71]. A dimensional approach to AUD, as outlined in DSM-5, may help improve the mapping of trajectories of clinical and neurobiological changes associated with increasingly severe substance use disorder. For example, studies correlating biological changes with clinical severity scales, such as the ADS [85, 86], may promote a dimensional approach. Gleich et al. [184] (2020) found symptom severity to be associated with reduced striatal DA D2/3 receptor availability among patients with AD, individuals at high risk, and HCs. However, dimensional approaches suffer from their own diagnostic uncertainties. Independent of dimensional and categorical approaches, long-term studies with larger samples are required to better evaluate the alterations during chronic consumption and prolonged abstinence. Such studies can help in tracking biological differences and changes among mild, moderate, and severe AUD. They can further identify factors contributing to the risk of relapse, promoting individualized treatment options.

The authors have no conflicts of interest to declare.

This study was supported by Grants from the German Research Foundation (Deutsche Forschungsgemeinschaft, DFG, FOR 1617: Grant Nos. HE 2597/14-1, HE 2597/15-1, HE 2597/14-2, HE 2597/15-2, GA 707/6-1, RA 1047/2-1, RA 1047/2-2; as well as DFG Grant No. CRC TRR 265). The data of this study are available on request from the corresponding author.

G.S. and A.H. conceived the study; G.S., M.G., R.B., and A.H. drafted the manuscript and supervised the study; G.S. performed the literature research; G.S. and M.G. created the tables. All authors revised the manuscript for important intellectual content and contributed to editorial changes in the manuscript. All authors read and approved the final manuscript.