The Moderating Role of Sex in the Relation between Cue-Induced Craving and Resting-State Functional Connectivity in the Salience Network of Non-Clinically Diagnosed Drinkers

Introduction: Previous research indicates a relation between craving and increased connectivity in the resting-state salience network. However, the link between cue-induced craving and connectivity in the salience network remains unclear. Further investigation is needed to understand the effect of sex on the relationship between cue-induced craving and the salience network. We investigated the role of sex in the association between the resting-state functional connectivity (RSFC) salience network and subjective cue-induced craving. Methods: Twenty-six males (mean age = 25.3) and 23 females (mean age = 26.0), with a score of 12 or higher on the alcohol use disorder identification test, were included in the current study. No significant difference in age was observed between males and females. Participants underwent a resting-state MRI scan for 6 min. Following the MRI scan, participants completed an alcohol cue-exposure task for 5.5 min to assess cue-induced craving using the desire to drink alcohol questionnaire. We applied independent component analysis methods to determine functional connectivity within the salience network. Subsequently, we investigated how cue-induced craving is related to the salience network’s RSFC and if this relationship is moderated by sex. Results: The association between the salience network and cue-induced craving was not statistically significant nor did we find a moderating effect for sex. Conclusion: The null findings in the study may be explained by a lack of power. Alternatively, alcohol use sex disparities may be more prevalent in the recreational/impulsive stage, whereas participants in our study were in the later stage of addiction.


Introduction
Over the last decade, sex differences in alcohol use and its harmful consequences have decreased [1,2]. Males continue to consume more alcohol, suffer more alcoholrelated injuries, and have more alcohol-related premature deaths than females [3]. However, disparities are closing as female alcohol consumption is increasing, whereas male alcohol use remains relatively stable. Nonetheless, females are underrepresented in most studies [4].
karger@karger.com www.karger.com/ear Furthermore, female drinkers seem to have a higher risk than male drinkers of developing alcohol use disorder (AUD) [5][6][7]. However, the mechanisms underlying these outcomes in females remain unclear. This highlights the importance of research on the mechanisms that explain potential sex differences in male and female alcohol drinkers to guide effective prevention and intervention strategies.
A key characteristic of AUD is craving: an intense, urgent, and abnormal desire characterized by longing, yearning, and a physiological need to take alcohol [8]. The presence of alcohol-related cues (cue exposure/cueinduced craving) may strengthen of the craving intensity [9]. Importantly, cue-induced craving has shown to be related to the severity of use [10] and relapse behavior [11,12] in subjects with substance use disorder (SUD). Sex differences in cue-induced craving in AUD populations or heavy drinkers have been identified. Females were found to have a stronger cue-induced craving for alcohol than males in several studies [13][14][15][16]. However, one study found no differences [17]. In addition, a previous study found no differences in neural cue-induced craving between male and female patients with AUD [18].
Previous studies suggested that males tend to have reward craving (positive reinforcement), which underlies the development of AUD and is related to the brain's resting-state functional connectivity (RSFC) salience network, as the salience network is explicitly involved in reward. Females are more inclined toward relief craving (negative reinforcement) [19][20][21][22] which could be related to the observation that females initially exhibit smaller responses in the nucleus accumbens (reward system) to drugs than males [23]. As such, sex differences in cueinduced craving may be related to sex differences in the RSFC salience network [24,25].
A potential mechanism that may explain strong cueinduced craving in individuals with an AUD is decreased connectivity within the salience network [26]. These suggested RSFC alterations in the salience network may contribute to the development and continuation of AUD [27][28][29] and include the insula, amygdala, dorsal anterior cingulate cortex (ACC), striatum, and ventral tegmental area in AUD patients [30]. Previous studies have shown that craving is associated with increased connectivity within the salience network in AUD participants [31]. Moreover, a stronger association was found between craving and RSFC of the salience network in individuals that did not complete treatment [32]. As such, these findings suggest that the association between RSFC and craving may provide a relevant biomarker for degeneration [32].
Decreased connectivity within the RSFC's salience network is not only found in people with AUD but also in nonclinical groups of problem drinkers, i.e., those that consume more than 4 (male) or 3 (female) glasses per day or more than 14 (male) or 7 (female) glasses per week, with 10 g of alcohol per glass in line with Dutch guidelines [33]. Both increased and decreased connectivity within the salience network has been reported in problem drinking populations, which may underlie differences in the risks for cue-induced craving [34][35][36]. It has furthermore been suggested that impairments in the salience network occur throughout the addiction cycle, both within the impulsive as well as compulsive stages of the disorder [37,38]. Unfortunately, evidence for the association between RSFC of the salience network and (cue-induced) craving in non-clinically diagnosed problem drinkers is currently lacking.
Therefore, the first aim of this study was to investigate the association between the RSFC of the salience network and cue-induced craving in problem drinkers. Second, this study aimed to identify sex differences in the relationship between RSFC of the salience network and craving in problem drinkers. It was hypothesized that enhanced connectivity of the salience networks would be associated with higher alcohol cue-induced craving and that this association would be stronger in males than females.

Study Design
The present study is a secondary analysis of baseline data collected as part of a randomized controlled study, of which the behavioral results and other MRI data have been published elsewhere [39,40] (see Fig. 1 for details).

Participants
Fifty-seven heavy drinkers were recruited from the local Amsterdam community and the Psychology Faculty of the University of Amsterdam. However, resting-state data were present for forty-nine participants. Thirty-six participants were smokers, and 13 were nonsmokers (see Table 1). After providing informed consent, participants received an online screening questionnaire to assess age, the alcoholic beverage of preference, drug use frequency in the last 12 months, drug use severity using the drug use disorder identification test (DUDIT) [41], and alcohol use severity using the alcohol use disorder identification test (AUDIT) [42]. We did not screen participants for mental disorders and current drug use. All these instruments are wellvalidated self-report questionnaires [41][42][43][44]. Inclusion criteria were participants aged between 18 and 40, a sufficient understanding of the Dutch language, an AUDIT score of 12 or higher, and a preference for drinking beer or wine. Participants were excluded when their DUDIT score was 12 or higher, indicating a The Role of Sex in the Relation between Salience Network and Cue-Induced Craving SUD other than AUD. All subjects received EUR 100 monetary compensation or research participation credits upon study completion. This study was approved by the Ethics Review Board of the Faculty of Social and Behavioral Sciences, University of Amsterdam (ERB number: 2016-DP-7014).

Procedures
Participants filled in an online questionnaire on the test day to assess alcohol intake in the preceding 14 days using the timeline follow-back assessment (TLFB) [44]. We asked participants not to drink 12 h before the experiment, which we tested using a breathalyzer. None of the participants had a positive result from the test. After completing the questionnaires, the MRI scan was run. The resting state in the present study was a fixation cross. The instruction was to let the participant's mind wander. However, the mind wander was not assessed afterward with a questionnaire.

Magnetic Resonance Imaging Data Acquisition
Brain imaging data were scanned using a 3.0 Tesla Achieva fullbody scanner (Philips Medical System, Best, The Netherlands) with a 32-channel SENSE head coil. Echo planar images of the entire brain were taken, with 37 ascending axial slices in total (3 mm × 3 mm × 3 mm voxel size; slice gap 3 mm; TR/TE 2,000 ms/28 ms; matrix 80 × 80). Additionally, a T1-3D high-resolution anatomical scan (TR/TE 8.2/3.7; matrix 240 × 187; 1 mm × 1 mm × 1 mm voxel) with transverse slices was taken. The flip angle of both MRI sequences was 76°.

Cue-Induced Craving Paradigm
Following the MRI scan, participants were taken into a soundattenuated room for the cue-induced craving paradigm. The sound-attenuated room is a soundproof room to prevent participants from being distracted during the exposure task.
The cue-induced craving procedures consisted of three phases, all for a total of 5.5 min. The sessions started with the participants' completion of 14 items of the desire for alcohol questionnaire (DAQ) [45,46] to measure baseline craving before the exposure, followed by a short relaxation exercise. The computer-assisted alcohol memory retrieval paradigm was modified from a protocol by Hammarberg et al. (2009) and Khemiri et al. (2015) [47,48]. For the first phase, the participants were exposed to four alcohol pictures lasting 30 s for each picture. Subsequently, they were instructed to read out loud two personalized scripts related to a pleasant alcohol memory and situations associated with strong craving. Participants provided the 30 s script before the experiment. The script example is as follows: "I just took an exam"; "I am excited but still a bit tense and stressed"; "I drink a cold Heineken beer from the bottle." Finally, in the third phase, the participants were gradually exposed to real alcoholic beverages. They were instructed to open a box in front of them, Firdaus/Kleiboer/Huizink/Kaag take out the alcohol, pour the alcohol into a glass, smell the alcohol three times for 30 s each, and to drink the alcohol. The alcohol that was provided to the participants during the alcohol-exposure paradigm was the participant's preferred brand of wine or beer. Immediately after the last phase, the DAQ was administered to reassess craving after alcohol exposure (see Fig. 2). Cue-induced craving was subsequently calculated by subtracting the post-and pre-cue-induced scores of subjective cravings (shown in Fig. 2).

Pre-processing and Resting-State Network Analysis
We analyzed the MRI data and ran the pre-processing analysis using the FSL program version 6.0.3. There was no reason to exclude participants' data after quality checking. We analyzed the data of 49 participants in total. The pre-processing began with reorientating the data and extracting cerebral structural scans using the BET tool [49]. The independent component analysis (ICA) was performed with the multi exploratory linear optimized decomposition into independent components (MELODIC) tool as part of the FSL program using FSL GUI. In the Miscvis, we turned on the balloon help, progress watcher, and 10% of the brain/background threshold. Subsequently, in the data, we uploaded 49 reoriented resting-state data for 240 total volumes, 2 TR, and a 100 Hz high pass filter. Next, we used mean displacement for relative motion exclusion in the pre-stats using MCFLIRT. We used <4 mm movement as a reference for motion exclusion correction. Using a 4 mm cut off or less is considered standard practice, particularly when dealing with sensitive or high-resolution data, as it helps minimize the effect of motion artifacts in the analysis and make sure the data are as clean as possible [50,51]. For the slice timing correction, we used the default from the FSL, "regular up." We turned on the brain extraction of functional scans using spatial smoothing with a 5 mm Gaussian Kernel and registration. Subsequently, we inspected all the filtered data for their motion and registration, and we did not find any movements exceeding 4 mm.

ICA and Salience Network Identification
We restricted the ICA to 30 components and performed a group ICA to generate the 30 components in all participants. The components consisting of the salience network regions include the ACC and insula [52], dorsal striatum and ventral striatum, VTA [53], and orbitofrontal cortex (OFC) from Harvard Center for Morphometric Analysis [54][55][56][57] were identified by overlaying standard atlases from the FSL program for higher level analysis.
We averaged values of component 8 (medial salience network [MSN]), which contains the ACC and bilateral insula, and 21 (frontal salience network [FSN]), which has the ACC, OFC, and bilateral ventral striatum to be included in the dual regression analysis as the salience network. Subsequently, we obtained the individual time courses using dual regression analysis [58]. The individual time series data were then used in the statistical analysis explained below.

Statistical Analysis
The normality of distribution of age, alcohol use severity (AUDIT), alcohol intake, DUDIT, pre-cue exposure, and postcue exposure DAQ were tested by the Shapiro-Wilk test. The report was differentiated based on whether the variable was normally distributed. Means and standard deviations were reported if the variable was normally distributed, but median and IQR were used if it was not. An independent sample Student's t test was performed to assess sex differences between males and females for the normality of variable distribution. Mann-Whitney U test was used if the variables were not normally distributed. In addition, the level of drinking severity was also classified into three categories according to World Health Organization (WHO) guidelines [42]. An AUDIT score of 1-7 represents low-risk consumption. Scores of 8-14 indicate Firdaus/Kleiboer/Huizink/Kaag hazardous or harmful alcohol consumption, and a score of 15 or more suggests alcohol dependence or moderate-severe alcohol consumption. We performed a χ 2 test to demonstrate whether the number of severe drinkers between males and females differed.
Furthermore, we used repeated measures analysis of variance (RM ANOVA) to answer the research hypothesis. In the first step, we tested the effect of the exposure phase (pre-exposure vs. postexposure) on craving. In the second step, we added RSFC and sex into the same model to test the main and interaction effects of RSFC of the salience network and sex on cue-induced craving. There was no assumption violated in the RM ANOVA model.

Group Characteristics
There was no difference in age between males (mean = 25.3) and females (mean = 26.0). Among the 26 male participants, 19 (73.1%) were moderate-severe drinkers, while the remaining 26.9% were considered to engage in hazardous alcohol consumption. Similarly, among 23 female participants, 20 (87.0%) were moderate-severe drinkers, while the remaining 13.0% were considered as hazardous alcohol consumers. Based on the χ 2 test result, the proportion of severe drinkers in the current study did not differ between males and females (χ 2 [1,49] = 0.72, p = 0.396) (shown in Table 1).
In addition, there were no significant differences between males and females regarding their age, heavy drinking status, smoking status, or drug use disorder. Moreover, we found that there were no significant differences in the craving scores at baseline between males (mean = 2.97, SD = 1.02) and females (mean = 3.17, SD = 0.64). Likewise, there were no differences in craving scores after cue exposure between males (mean = 3.54, SD = 1.10) and females (mean = 3.84, SD = 0.79). However, males reported more alcohol intake (standard drinks) than females per week.

Resting-State Salience Network Identification
The ICA succeeded in identifying the salience network. Components 8 and 21 were identified, showing the salience networks. Moreover, these components are referred to as the MSN and the FSN. The MSN contains the ACC and bilateral insula, and the FSN includes the ACC, OFC, and bilateral ventral striatum. We averaged both values of MSN and FSN to include in our regression model as the salience network.
To test the presence of overlap between the two components, we generated a combined mask and extracted activities to evaluate the overlap, as illustrated in Figure 3. We observed a slight overlap between the two components. As a precautionary measure, we re-evaluated our analysis using both MSN and FSN separately (see online suppl. material; for all online suppl. material, see https://doi.org/10.1159/000531090 for further details). The result indicated that, despite the overlap, there was no significant difference from our original findings. Therefore, we proceeded with the analysis using the average values obtained from MSN and FSN.

Discussion
In the present study, we analyzed baseline data of a randomized controlled study to investigate the effect of a working memory intervention on alcohol cue reactivity [39]. The study concerns secondary analyses on the MRI data collected before the cue-exposure task. The measurement order (MRI scan first, followed by the cue-exposure task) was chosen to reduce the risk of carryover effects of the exposure task, which was expected to increase the craving. Therefore, to ensure that this exposure paradigm did not influence the MRI data, the exposure paradigm was done after the questionnaire and MRI scanning procedure.
It was hypothesized that cue-induced craving would be positively associated with RSFC of the salience network and that this correlation would be stronger in males than in females. However, no association was found between the RSFC of the salience network and cue-induced craving. Moreover, contrary to our hypothesis, there were no sex differences in cue-induced craving. Lastly, the relationship between the RSFC of the salience network and cue-induced craving was not moderated by sex. In addition to our main finding, we also found that there was no significant difference in the RSFC of the salience network between males and females.
The salience network has been linked to the early stages of addiction, specifically the binge/intoxication phase, associated with the initial and impulsive phase of SUDs [59]. Despite the fact that the current participants were not diagnosed with AUD, the AUDIT scores reported by the participants did indicate that the majority of the participants could be categorized as potentially dependent drinkers. As such, the participants in the study may not have been in the initial (impulsive) stage of AUD. According to Koob's model [59], the striatum and the ventral tegmental area are more involved in the impulsive, early stages of substance use (binge/intoxication phase), whereas the executive control network is implicated more in the compulsive, later stages of SUD (negative withdrawal and anticipation phase) [38]. Thus, cue-induced craving in compulsive alcohol users (as the ones potentially included in the current study) may be related more to the RSFC of other networks (e.g., executive control network).
Alternatively, despite the pragmatic advantage of investigating the resting brain, such as the absence of a cognitive demand for the subject [60], the RSFC may not be the best biomarker for cue-induced craving. Although the cognitive demand during the restingstate functional MRI (fMRI) scan was absent, the brain can be involved in a number of different processes, such as worrying about something or daydreaming [61]. As a result, interpreting RSFC would be challenging if we were unclear about what the participant did during the resting-state period [60].
Finally, the current study did not provide evidence regarding the difference between males and females in the association between cue-induced craving and the resting salience network. It is likely that the lack of evidence for our hypothesis is due to the fact that sex differences are most noticeable in the early stages of addiction [23]. The majority of studies have found that females progress more quickly than males from their first exposure to a substance to an addiction stage. As we mentioned above, 79.59% of female participants in our study were alcohol dependent, and 73.08% of male participants were moderate-severe drinkers. Hence, our study population may not be in the stage of occasional use of alcohol anymore but rather in a later stage of addiction.
There are several limitations in our study that must be taken into consideration when interpreting the results. The current study had a sample size of 49, which would have been sufficient to detect intermediate to large effect sizes, but the effect sizes found in our study were very small. Moreover, we did not consider the menstrual phase or hormonal contraceptive use in females, as it has been reported that the menstrual cycle [62][63][64] and contraceptive use [65] may affect sensitivity to reward and substance-related stimuli. The previous study hypothesized that a sudden drop in hormone levels during the late luteal phase, which begins when progesterone levels drop and end on the day before the next menses, causes a decrease in endogenous dopamine activity, which in turn causes increased dopamine release in response to reward cues [66]. This could explain why women have a higher liking for alcohol consumption in the late luteal phase than in the follicular phase, which begins on the first day of menstruation and ends with ovulation [67].
While the current study suggests that RSFC is not associated with cue-induced craving for either males or females, this conclusion warrants replication in a larger sample size that could potentially include more diverse groups, including at-risk drinkers, drinkers with AUD, and light drinking controls. This way, a researcher could have a better point of view to capture the whole addiction cycle and to see differences in the associative strength between RSFC and cue-induced craving in both males and females in the different stages of addiction. It should also be noted that we only considered sex assigned at birth and not participants' gender-identity. According to the sex and gender equity in research (SAGER) guidelines, gender and gender-identity are equally important determinants of health and well-being as sex and should therefore be implicated in all research [68]. Finally, by controlling the menstrual cycle and focusing on other RSFC than the salience network, researchers may gain further insight into the association between RSFC and cue-induced craving and potential sex differences.

Conclusion
In summary, our study could not demonstrate that the salience network is unrelated to cue-induced craving in a population of non-clinically diagnosed heavy drinkers, for either males or females. It is important to mention that our findings in the current investigation are exploratory and could be used to generate future hypotheses regarding the salience network and cueinduced craving changes and how this relationship could differ between males and females in non-clinically diagnosed drinkers.

Statement of Ethics
This study protocol was reviewed and approved by the Ethics Review Board of the Faculty of Social and Behavioral Sciences, University of Amsterdam (ERB number: 2016-DP-7014). Moreover, the authors stated that the subjects in the recent study have given their written informed consent.

Conflict of Interest Statement
The authors declare no personal or financial conflicts of interest related to the results of this research.

Funding Sources
This work was supported by a research grant from the Amsterdam Brain and Mind Project, a joint initiative of the University of Amsterdam and VU University Amsterdam. Moreover, the salary of the Ph.D. student and the publication fee was provided by a research grant from the Endowment Fund for Education (LPDP) number 201802221012420, Indonesia's Ministry of Finance. Both grants (Amsterdam Brain and Mind Project and LPDP) had no involvement in this study design, analysis, and interpretation of data and the decision to submit the article for publication.

Data Availability Statement
In accordance with ethical and legal considerations, the data generated and analyzed in this study is not openly available. Requests for access to the anonymized and coded data must be submitted and reviewed by Anne Marije Kaag prior to release. Further inquiries can be directed to the corresponding author.