If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
National Centre for Scientific Research UMR 5287 – Institut de Neurosciences Cognitives et Intégratives d'Aquitaine, University of Bordeaux, Bordeaux, FranceÉcole pratique des hautes études, Paris Sciences et Lettres Research University, Paris, France
National Centre for Scientific Research UMR 5287 – Institut de Neurosciences Cognitives et Intégratives d'Aquitaine, University of Bordeaux, Bordeaux, FranceÉcole pratique des hautes études, Paris Sciences et Lettres Research University, Paris, France
National Centre for Scientific Research UMR 5287 – Institut de Neurosciences Cognitives et Intégratives d'Aquitaine, University of Bordeaux, Bordeaux, FranceCentre Hospitalier Charles Perrens, Bordeaux, France
Centre National de la Recherche Scientifique UMR 6033 – Sleep, Addiction and Neuropsychiatry, University of Bordeaux, Bordeaux, FranceCentre Hospitalier Charles Perrens, Bordeaux, FranceCentre Hospitalier Universitaire Bordeaux, Bordeaux, France
National Centre for Scientific Research UMR 5287 – Institut de Neurosciences Cognitives et Intégratives d'Aquitaine, University of Bordeaux, Bordeaux, FranceCentre Hospitalier Universitaire Bordeaux, Bordeaux, France
National Centre for Scientific Research UMR 5287 – Institut de Neurosciences Cognitives et Intégratives d'Aquitaine, University of Bordeaux, Bordeaux, FranceÉcole pratique des hautes études, Paris Sciences et Lettres Research University, Paris, France
Substance use disorders (SUDs) are major contributors to morbidity and mortality rates worldwide, and this global burden is attributable in large part to the chronic nature of these conditions. Increased mood variability might represent a form of emotional dysregulation that may have particular significance for the risk of relapse in SUD, independent of mood severity or diagnostic status. However, the neural biomarkers that underlie mood variability remain poorly understood.
Methods
Ecological momentary assessment was used to assess mood variability, craving, and substance use in real time in 54 patients treated for addiction to alcohol, cannabis, or nicotine and 30 healthy control subjects. Such data were jointly examined relative to spectral dynamic causal modeling of effective brain connectivity within 4 networks involved in emotion generation and regulation.
Results
Differences in effective connectivity were related to daily life variability of emotional states experienced by persons with SUD, and mood variability was associated with craving intensity. Relative to the control participants, effective connectivity was decreased for patients in the prefrontal control networks and increased in the emotion generation networks. Findings revealed that effective connectivity within the patient group was modulated by mood variability.
Conclusions
The intrinsic causal dynamics in large-scale neural networks underlying emotion regulation play a predictive role in a patient’s susceptibility to experiencing mood variability (and, subsequently, craving) in daily life. The findings represent an important step toward informing interventional research through biomarkers of factors that increase the risk of relapse in persons with SUD.
Global burden of disease estimates have consistently ranked substance use disorders (SUDs) as among the greatest contributors to morbidity and mortality rates worldwide (
Alcohol consumption’s attributable disease burden and cost-effectiveness of targeted public health interventions: A systematic review of mathematical models.
GBD 2016 Alcohol and Drug Use Collaborators The global burden of disease attributable to alcohol and drug use in 195 countries and territories, 1990–2016: A systematic analysis for the Global Burden of Disease Study 2016.
). Next to comorbidity, immediate determinants (e.g., intrapersonal high-risk situations including negative emotional states, coping skills) and covert antecedents (e.g., craving, lifestyle) can contribute to relapse (
). Although these factors may each contribute independently to the chronic nature of SUD, investigations of their potential associations or interactions are lacking. In particular, a robust link has been established between the development, persistence, and severity of SUD and difficulties in emotion regulation (
Emotion regulation as a transdiagnostic treatment construct across anxiety, depression, substance, eating and borderline personality disorders: A systematic review.
The role of anxiety sensitivity and difficulties in emotion regulation in posttraumatic stress disorder among crack/cocaine dependent patients in residential substance abuse treatment.
Comparative efficacy and acceptability of 21 antidepressant drugs for the acute treatment of adults with major depressive disorder: A systematic review and network meta-analysis.
Efficacy and tolerability of antidepressants in the treatment of adolescents and young adults with depression and substance use disorders: A systematic review and meta-analysis.
) and were not developed to target pathophysiological mechanisms associated with addiction. The lack of knowledge of the neural biomarkers underlying emotion dysregulation in persons with SUD, therefore, represents an important missing link to identifying individual vulnerabilities that may reliably predict treatment response and relapse risk (
) investigating the intrinsic functional architecture in SUD have reported decreased resting-state functional and structural connectivity between prefrontal and subcortical regions, which might result in impaired top-down control of emotion-generating regions. In addition, decreased resting-state functional connectivity between and within prefrontal regulatory networks and disrupted default mode network connectivity to other large-scale networks, such as the executive control network (ECN) and salience network, have been observed in SUD. Studies examining large-scale networks in emotion dysregulation found decreased intrinsic connectivity with the default mode network and ECN in anxiety disorders (
). Taken together, these findings suggest that decreased functional connectivity within regulatory networks, on the one hand, and emotion-generating/processing networks, on the other, might explain poor emotion regulation in SUD. However, previous studies have essentially been limited to examining functional connectivity between different networks and its correlation with diagnostic status (
). Very recent investigations have begun to move beyond this traditional approach to develop more robust models to predict emotion regulation, treatment response, and substance use through advanced analytic approaches such as machine learning and cross-validation (
Rostral anterior cingulate network effective connectivity in depressed adolescents and associations with treatment response in a randomized controlled trial.
Rostral anterior cingulate network effective connectivity in depressed adolescents and associations with treatment response in a randomized controlled trial.
)]. Despite their novelty from the point of view of neuroimaging, these studies remain limited to assessments of symptom averages or total severity scores. The association of brain biomarkers with the core phenomena of emotion dysregulation, notably increased emotional variability (
) and examines their direct link to changes in the daily-life variability of emotional states as they are experienced by persons with SUD in real time. Specifically, resting-state functional magnetic resonance imaging (MRI) was used to test effective (causal) connectivity within 4 predefined large-scale networks underlying emotion generation and regulation (
) and included 2 lateral prefrontal networks (N1 and N2) implicated in working memory, language, and attention, and 2 subcortical networks (N3 and N4), including the amygdala and insula, involved in emotion generation/processing and interoception. By assessing effective connectivity in these networks in patients with SUD and healthy control subjects, we examine the link between effective connectivity and mood variability assessed in the natural contexts of daily life by ecological momentary assessment (EMA). We specifically hypothesized that 1) persons with SUD would experience greater mood variability than healthy control participants (
); 2) persons with SUD would demonstrate decreased connectivity in the prefrontal control (N1 and N2) and emotion generation (N3 and N4) networks compared with healthy control subjects (
); and 3) mood variability in SUD would be negatively associated with connectivity within the regulatory networks (N1 and N2) and positively associated with connectivity in the emotion generation/processing networks (N3 and N4).
Methods and Materials
Participants
A total of 126 individuals (86 patients with SUD and 40 healthy control subjects) participated in the EMA study. Eighty-four individuals (54 patients and 30 healthy control subjects) of the full sample also completed an MRI examination. See the Supplement for further description of participants and inclusion criteria.
Procedure
DSM-IV-TR diagnoses of SUDs were established based on criteria assessed by the Mini International Neuropsychiatric Interview French Version 5.0.0 (
The Mini-International Neuropsychiatric Interview (M.I.N.I.): The development and validation of a structured diagnostic psychiatric interview for DSM-IV and ICD-10.
). After verifying eligibility, all participants completed a clinical assessment battery and were trained to operate a study-dedicated smartphone (Samsung Galaxy S [Samsung Electronics] with a 10.6-cm screen, 12-point font size). The surveys occurred over a 7-day period, 5 times per day at random intervals within 5 equal time epochs from morning to evening (approximately every 3 hours). Participants recruited for the primary investigation involving EMA were also allowed to participate in the ancillary MRI study. The subsample of individuals who received an MRI examination did so within 48 hours before completing clinical testing and EMA. Financial compensation was provided with a maximum of €100 in purchase vouchers to complete both the EMA and MRI phases of the study. The feasibility and validity of the EMA protocol used in this investigation have previously been demonstrated in patients with SUD (
Effects of anxiety and mood disorders on craving and substance use among patients with substance use disorder: An ecological momentary assessment study.
Participants were prompted to respond to diverse questions concerning their emotions, behaviors, experiences, and environmental contexts at each EMA survey. In particular, they were asked to describe the maximum level of craving since the previous assessment on a 7-point scale ranging from 1 (no desire to use substances) to 7 (extreme desire to use substances). They were also asked if they had used the substance that initiated their treatment since the previous EMA assessment, followed by questions about the quantity (if any) of the consumed substance and other descriptive information. Participants were asked about any other form of a psychoactive substance used during that time period that was not the target of treatment (alcohol, tobacco, cannabis, opiates, cocaine, amphetamine, and other substances). Based on the mood circumplex model (
), a 7-point scale was used at each EMA assessment to assess the continuum of mood experienced by the participant at that moment, ranging from 1 (very sad) to 7 (very happy).
)] was used to account for the multilevel structure of the EMA data that involved within-person variance in mood, craving, and substance use, as well as between-person variance in clinical and sociodemographic characteristics. Mood variability was based on calculating the average standard deviation of mood scores acquired for each day of the 7-day EMA period. These within-day standard deviation coefficients were then examined for their association with the averages of craving for that same day and with within-day sums of substance use occasions (for both the treated substance and any substance). The within-person coefficients calculated for each day of the EMA sampling period were also extracted for use in neuroimaging analyses.
The anatomical scans were preprocessed with the Computational Anatomy Toolbox implemented in SPM12 (http://www.neuro.uni-jena.de/cat/index.html#VBM). See the Supplement for further description of preprocessing. We tested for group differences in gray matter volume of the network-specific regions by applying two-sample t tests with group as a factor and masked by each network. No differences in gray matter volume were determined between patients and healthy control subjects in any of the regions of interest (ROIs) of the 4 networks.
Step 2: Image Preprocessing
Resting-state functional MRI and T1-weighted MRI images were preprocessed using fMRIPrep 20.2.1 (
). This meta-analysis identified 4 large-scale neural networks using a meta-analytic hierarchical clustering approach (N1: 10 nodes; N2: 8 nodes; N3: 8 nodes; N4: 9 nodes) (illustrated in Figure S1). The Montreal Neurological Institute coordinates of all ROIs can be found in Table 1 and are illustrated in Figure 2. N1 and N2 are mainly related to emotion regulatory processes, while N3 and N4 are involved in emotion generative and physiological processes. We extracted blood oxygen level–dependent time series from each network’s predefined functional brain regions for each subject. See the Supplement for further description.
Table 1Montreal Neurological Institute Coordinates of Regions of Interest Identified by Meta-analysis
Network
Region
BA
Coordinates
x
y
z
1
LSFG
8
0
24
50
RMFG
8
40
24
42
RIPL
40
58
−52
38
LIPL
40
−58
−50
44
LMFG
10
−36
52
−2
LMFG
6
−42
14
48
RMFG
11
42
46
−8
RINS
13
36
16
6
RACC
23
2
−22
30
RPreCun
7
10
−64
36
2
LIFG
47
−46
24
−8
LSFG
6
−4
10
62
RIFG
47
50
28
−8
LSTG
39
−46
−52
28
LMTG
–
−54
−34
−2
LMFG
6
−44
6
50
LSFG
9
−30
48
26
LCaudate
–
−16
10
12
3
LAMY
–
−22
−4
−16
RAMY
–
24
−4
−18
RFFA
37
40
−46
−18
RThal
–
6
−26
0
LFFA
37
−38
−54
−14
LPHG
27
−22
−28
−4
BMedFG
10
0
54
−10
LIOG
19
−42
−76
−6
4
LPostCG
2
−58
−22
32
LINS
13
−44
−4
10
LSPL
7
−28
−52
56
RPostCG
2
62
−22
30
LCun
18
−10
−76
22
LMOG
19
−48
−74
2
RThal
–
10
−26
−4
RPreCun
19
28
−60
38
RPCC
30
16
−56
16
B, bilateral; BA, Brodmann area; BMedFG, bilateral medial frontal gyrus; L, left; LAMY, left amygdala; LCaudate, left caudate; LCun, left cuneus; LFFA, left fusiform face area; LIFG, left inferior frontal gyrus; LIOG, left inferior occipital gyrus; LINS, left insula; LIPL, left inferior parietal lobule; LMFG, left middle frontal gyrus; LMOG, left middle occipital gyrus; LMTG, left middle temporal gyrus; LPHG, left parahippocampal gyrus; LPostCG, left postcentral gyrus; LSFG, left superior frontal gyrus; LSPL, left superior parietal lobule; LSTG, left superior temporal gyrus; R, right; RACC, right anterior cingulate cortex; RAMY, right amygdala; RFFA, right fusiform face area; RIFG, right inferior frontal gyrus; RINS, right insula; RIPL, right inferior parietal lobule; RMFG, right middle frontal gyrus; RPCC, right posterior cingulate; RPostCG, right postcentral gyrus; RPreCun, right precuneus; RThal, right thalamus.
Figure 2Regions of interest definition and fully connected model. The initial model for each network assumed a fully connected intrinsic architecture within (A) network 1 (N1, 10 nodes, 100 connections), (B) network 2 (N2, 8 nodes, 64 connections), (C) network 3 (N3, 8 nodes, 64 connections), and (D) network 4 (N4, 9 nodes, 81 connections). Regions of Interest, represented in circles, are color-coded for various brain regions (e.g., frontal, temporal, parietal, etc.). BA, Brodmann area; LAMY, left amygdala; LCaudate, left caudate; LCun, left cuneus; LFFA, left fusiform face area; LIFG, left inferior frontal gyrus; LIOG, left inferior occipital gyrus; LINS, left insula; LIPL, left inferior parietal lobule; LMFG, left middle frontal gyrus; LMTG, left middle temporal gyrus; LPHG, left parahippocampal gyrus; LPostCG, left postcentral gyrus; LPreCun, left precuneus; LSFG, left superior frontal gyrus; LSPL, left superior parietal lobule; LSTG, left superior temporal gyrus; MedFG, medial frontal gyrus; RACC, right anterior cingulate cortex; RAMY, right amygdala; RFFA, right fusiform face area; RIFG, right inferior frontal gyrus; RINS, right insula; RIPL, right inferior parietal lobule; RMFG, right middle frontal gyrus; RPCC, right posterior cingulate; RPostCG, right postcentral gyrus; RPreCun, right precuneus; RThal, right thalamus.
spDCM was performed using DCM12 as implemented in SPM12. DCM tests the directed functional interactions between brain regions. Connectivity between ROIs is measured in hertz. Positive values indicate an excitatory connection, while negative values represent an inhibitory connection. See the Supplement for further description of spDCM.
Step 5: Parametric Empirical Bayes Model
On the second level, we were interested in modulating the connectivity between neural networks by mood variability. For this, we used the hierarchical parametric empirical Bayes framework for DCM (
Four parametric empirical Bayes analyses were conducted: 1) testing the group difference in each effective connectivity between the SUD and control groups; 2) testing the linear relationship between each effective connectivity and the mood variability score (main regressor of interest) within the SUD group; 3) testing the linear relationship between each effective connectivity and the mood variability score (main regressor of interest) within the SUD group corrected for lifetime depression (reported in the Supplement); and 4) testing the linear relationship between each effective connectivity and the mood variability score (main regressor of interest) within the SUD group corrected for lifetime depression and substance type (reported in the Supplement). In all parametric empirical Bayes models, age, sex, and mean frame-to-frame displacement were modeled as regressors of no interest. Only effects (i.e., changes in effective connectivity) that show a posterior probability > .95 are reported.
Step 6: Prediction—Cross-validation
In a final step, leave-one-out cross-validation (LOOCV) was performed to determine the robustness of observed effect sizes. This procedure assessed the predictive validity of mood variability, i.e., whether the effect size was large enough to predict a left-out patient’s mood variability score from the connectivity within the 4 networks within the SUD group. LOOCV was performed for each significant connection separately.
Results
Demographic and Clinical Results
Table 2 presents the sociodemographic and clinical characteristics of the sample. Craving intensity was greater in patients than in control subjects and for the cannabis relative to the alcohol group, and use of the substance at the origin of treatment was more frequent for the nicotine group than the other substance groups. Mood negativity and mood variability were significantly greater in patients than in the control group. Those in the alcohol group had greater negative mood than those in the nicotine group. No significant sex differences in mood variability were found between men and women in the patient (t = −0.459, p = .647) or healthy control (t = 1.074, p = .290) groups. The subsample who completed the MRI examination did not differ from those who participated only in EMA for variables presented in Table 2 except that healthy control subjects who received an MRI were more compliant than those who did not (96% vs. 89%). This MRI group also included proportionately more men than the EMA-only group.
Table 2Description of Sociodemographic, Clinical, and EMA Variables for the Sample
Association of Mood Variability With Craving and Substance Use
Although within-day averages of mood and craving were not associated in the overall sample, γ = 0.272, SE = 0.180, p > .05, mood variability was significantly linked to increased craving intensity, γ = 0.517, SE = 0.256, p < .05. Moreover, when these analyses used the average daily severity of mood scores as covariates, the results confirmed the independent association between mood variability and craving intensity, γ = 0.559, SE = 0.255, p < .05. Subsequent analyses demonstrated that the association of mood variability and craving intensity was significant among patients with an SUD, γ = 0.763, SE = 0.329, p < .05, but not among healthy control subjects. For this reason, the link between mood variability and effective connectivity was investigated only in patients with SUD. The association of mood variability and craving intensity did not vary in magnitude across the alcohol, nicotine, or cannabis groups. Finally, the magnitude of the within-person association between mood variability and craving intensity in patients was examined as a function of comorbid diagnoses. While no effect was observed for lifetime psychotic or anxiety disorders, those with lifetime depression had a significantly stronger association between mood variability and craving intensity (Table 3). The within-person association of mood variability and craving was not associated with current comorbid diagnoses, and mood variability was not directly associated with the use of the treated substance or the use of other substances. Of note, sex had an impact on the within-person association between mood variability and craving, with men showing a more pronounced association than women (coefficient = −0.593, p < .001).
Table 3Within-Day Mood Variability and Craving Intensity as a Function of Lifetime Depression
The group differences (SUD minus control) in effective connectivity within each network are shown in Figure 3. Positive values (green) indicate increased connectivity for the patients compared with the control group. In contrast, negative values (red) indicate reduced connectivity for the patients compared with the control group. The direction of effective connectivity (excitatory/inhibitory) between 2 regions is reported in Table 4. See the Supplement for further description on the group comparisons.
Figure 3Spectral dynamic causal modeling results of group differences (patients > control subjects). Effective connectivity of group differences within each network (N1–N4). Green/red colors indicate positive/negative connectivity for patients compared with control subjects. Effect sizes that survived a ≥95% posterior confidence criterion are shown in color. The matrices can be interpreted as effective connectivity from column (source) to row (target). BA, Brodmann area; LAMY, left amygdala; LCaudate, left caudate; LCun, left cuneus; LFFA, left fusiform face area; LIFG, left inferior frontal gyrus; LINS, left insula; LIOG, left inferior occipital gyrus; LIPL, left inferior parietal lobule; LMFG, left middle frontal gyrus; LMOG, left middle occipital gyrus; LMTG, left middle temporal gyrus; LPHG, left parahippocampal gyrus; LPostCG, left postcentral gyrus; LSFG, left superior frontal gyrus; LSPL, left superior parietal lobule; LSTG, left superior temporal gyrus; MedFG, medial frontal gyrus; RACC, right anterior cingulate cortex; RAMY, right amygdala; RFFA, right fusiform face area; RIFG, right inferior frontal gyrus; RINS, right insula; RIPL, right inferior parietal lobule; RMFG, right middle frontal gyrus; RPCC, right posterior cingulate; RPostCG, right postcentral gyrus; RPreCun, right precuneus; RThal, right thalamus.
Table 4Direction of Effective Connectivity Between 2 Regions (Group Comparisons)
Network and Effect of Connectivity
Direction of Connectivity
Source
Target
Interaction With Group
Effect Size in Hz
Network 1
Inhibition
LMFG_BA6
→
LIPL_BA40
−
−0.08
LMFG_BA6
→
RIPL_BA40
−
−0.09
RPreCun_BA7
→
RMFG_BA8
−
−0.10
RPreCun_BA7
→
RIPL_BA40
−
−0.10
RMFG_BA8
→
RPreCun_BA7
−
−0.07
RMFG_BA8
→
LMFG_BA10
−
−0.08
RMFG_BA11
→
LIPL_BA40
−
−0.15
RMFG_BA11
→
RIPL_BA40
−
−0.15
RIPL_BA40
→
RIPL_BA40
−
−0.23
LMFG_BA6
→
LMFG_BA6
+
0.12
RPreCun_BA7
→
LSFG_BA8
+
0.11
RINS_BA13
→
RMFG_BA11
+
0.09
RINS_BA13
→
RINS_BA13
+
0.16
RACC_BA23
→
RACC_BA23
+
0.10
LIPL_BA40
→
RPreCun_BA7
+
0.10
LIPL_BA40
→
RMFG_BA11
+
0.07
LIPL_BA40
→
RINS_BA13
+
0.10
Excitation
LMFG_BA6
→
RPreCun_BA7
−
−0.08
LMFG_BA6
→
LSFG_BA8
−
−0.11
LMFG_BA6
→
RMFG_BA8
−
−0.11
LMFG_BA6
→
LMFG_BA10
−
−0.14
LMFG_BA6
→
RACC_BA23
−
−0.09
LSFG_BA8
→
RPreCun_BA7
−
−0.09
LSFG_BA8
→
RACC_BA23
−
−0.10
RMFG_BA8
→
RINS_BA13
−
−0.11
RACC_BA23
→
LMFG_BA10
−
−0.08
RIPL_BA40
→
LSFG_BA8
−
−0.10
LMFG_BA10
→
LMFG_BA6
+
0.12
LMFG_BA10
→
LIPL_BA40
+
0.12
LIPL_BA40
→
RIPL_BA40
+
0.09
RIPL_BA40
→
RPreCun_BA7
+
0.13
RIPL_BA40
→
RMFG_BA8
+
0.17
Network 2
Inhibition
LSFG_BA9
→
LMFG_BA6
−
−0.09
LSFG_BA9
→
LSFG
−
−0.15
LSFG_BA9
→
LIFG_BA47
−
−0.10
LSFG_BA9
→
LMTG
−
−0.09
LSTG_BA39
→
LSFG
−
−0.09
LSTG_BA39
→
LMTG
−
−0.11
LMTG
→
LMTG
−
−0.24
LCaudate
→
LSFG_BA9
−
−0.15
LCaudate
→
LIFG_BA47
−
−0.12
LSFG
→
LSFG
+
0.14
RIFG_BA47
→
RIFG_BA47
+
0.11
LMTG
→
LMFG_BA6
+
0.10
LMTG
→
LIFG_BA47
+
0.15
LMTG
→
RIFG_BA47
+
0.20
LCaudate
→
LSTG_BA39
+
0.11
Excitation
LIFG_BA47
→
LSFG_BA9
−
−0.11
RIFG_BA47
→
LIFG_BA47
−
−0.09
LMFG_BA6
→
LSTG_BA39
+
0.11
LSFG
→
LSTG_BA39
+
0.08
Network 3
Inhibition
MedFG_BA10
→
MedFG_BA10
−
−0.11
LIOG_BA19
→
RAMY
−
−0.15
RFFA_BA37
→
LAMY
+
0.15
LPHG_BA27
→
LIOG_BA19
+
0.30
LAMY
→
LPHG_BA27
+
0.12
RAMY
→
RFFA_BA37
+
0.11
LIOG_BA19
→
RFFA_BA37
+
0.18
RThal
→
MedFG_BA10
+
0.10
RThal
→
RFFA_BA37
+
0.08
Excitation
LFFA_BA37
→
LFFA_BA37
−
−0.16
LFFA_BA37
→
RFFA_BA37
−
−0.10
RFFA_BA37
→
LIOG_BA19
−
−0.16
RFFA_BA37
→
RThal
−
−0.12
LPHG_BA27
→
RThal
−
−0.19
LAMY
→
LFFA_BA37
−
−0.08
LAMY
→
LAMY
−
−0.15
LAMY
→
RAMY
−
−0.13
RAMY
→
RAMY
−
−0.12
LIOG_BA19
→
LIOG_BA19
−
−0.17
MedFG_BA10
→
LIOG_BA19
+
0.08
LFFA_BA37
→
LPHG_BA27
+
0.19
LFFA_BA37
→
RThal
+
0.18
LPHG_BA27
→
LPHG_BA27
+
0.12
Network 4
Inhibition
LSPL_BA7
→
LMOG_BA19
−
−0.13
LINS_BA13
→
LMOG_BA19
−
−0.13
LINS_BA13
→
RPreCun_BA19
−
−0.11
LCun_BA18
→
LPostCG_BA2
−
−0.17
LCun_BA18
→
LSPL_BA7
−
−0.14
LMOG_BA19
→
RPostCG_BA2
−
−0.10
RPreCun_BA19
→
RPreCun_BA19
−
−0.23
RPCC_BA30
→
RPostCG_BA2
−
−0.09
RThal
→
LINS_BA13
−
−0.12
LPostCG_BA2
→
LINS_BA13
+
0.08
LPostCG_BA2
→
LCun_BA18
+
0.10
LPostCG_BA2
→
LMOG_BA19
+
0.18
LPostCG_BA2
→
RPCC_BA30
+
0.10
LPostCG_BA2
→
RThal
+
0.08
RPostCG_BA2
→
LINS_BA13
+
0.10
LSPL_BA7
→
LINS_BA13
+
0.13
LINS_BA13
→
LINS_BA13
+
0.10
RPreCun_BA19
→
RPostCG_BA2
+
0.11
RThal
→
LPostCG_BA2
+
0.12
RThal
→
RPostCG_BA2
+
0.14
RThal
→
LSPL_BA7
+
0.09
RThal
→
LCun_BA18
+
0.24
RThal
→
LMOG_BA19
+
0.28
Excitation
RPostCG_BA2
→
RPostCG_BA2
−
−0.14
RPostCG_BA2
→
LMOG_BA19
−
−0.09
LINS_BA13
→
LPostCG_BA2
−
−0.11
LINS_BA13
→
RPostCG_BA2
−
−0.23
LINS_BA13
→
LCun_BA18
−
−0.16
LINS_BA13
→
RPCC_BA30
−
−0.12
LCun_BA18
→
RPreCun_BA19
−
−0.08
LMOG_BA19
→
LPostCG_BA2
−
−0.09
LMOG_BA19
→
LINS_BA13
−
−0.26
LMOG_BA19
→
RThal
−
−0.08
LPostCG_BA2
→
RPreCun_BA19
+
0.12
RThal
→
RPCC_BA30
+
0.08
BA, Brodmann area; LAMY, left amygdala; LCaudate, left caudate; LCun, left cuneus; LFFA, left fusiform face area; LIFG, left inferior frontal gyrus; LIOG, left inferior occipital gyrus; LINS, left insula; LIPL, left inferior parietal lobule; LMFG, left middle frontal gyrus; LMOG, left middle occipital gyrus; LMTG, left middle temporal gyrus; LPHG, left parahippocampal gyrus; LPostCG, left postcentral gyrus; LSFG, left superior frontal gyrus; LSPL, left superior parietal lobule; LSTG, left superior temporal gyrus; MedFG, medial frontal gyrus; RACC, right anterior cingulate cortex; RAMY, right amygdala; RFFA, right fusiform face area; RIFG, right inferior frontal gyrus; RINS, right insula; RIPL, right inferior parietal lobule; RMFG, right middle frontal gyrus; RPCC, right posterior cingulate; RPostCG, right postcentral gyrus; RPreCun, right precuneus; RThal, right thalamus.
Association of Effective Connectivity With Mood Variability
The results of the relationship between mood variability and effective connectivity within each network in the patient group are shown in Figure 4. Positive values (green) indicate a positive relationship between effective connectivity and mood variability, while negative values (red) represent a negative association. The direction of effective connectivity (excitatory/inhibitory) between 2 regions is reported in Table 5.
Figure 4Spectral dynamic causal modeling results of the patient group. Effective connectivity of mood variability within each network (N1–N4). Green/red colors indicate a positive/negative relationship with mood variability. Effect sizes that survived a ≥95% posterior confidence criterion are shown in color. The connections with a group difference (patients > control subjects) are highlighted in dashed boxes. Number of incoming (target) and outgoing (source) connections are shown in the gray bar graphs along the sides. The matrices can be interpreted as effective connectivity from column (source) to row (target). BA, Brodmann area; LAMY, left amygdala; LCaudate, left caudate; LCun, left cuneus; LFFA, left fusiform face area; LIFG, left inferior frontal gyrus; LINS, left insula; LIOG, left inferior occipital gyrus; LIPL, left inferior parietal lobule; LMFG, left middle frontal gyrus; LMOG, left middle occipital gyrus; LMTG, left middle temporal gyrus; LPHG, left parahippocampal gyrus; LPostCG, left postcentral gyrus; LSFG, left superior frontal gyrus; LSPL, left superior parietal lobule; LSTG, left superior temporal gyrus; MedFG, medial frontal gyrus; RACC, right anterior cingulate cortex; RAMY, right amygdala; RFFA, right fusiform face area; RIFG, right inferior frontal gyrus; RINS, right insula; RIPL, right inferior parietal lobule; RMFG, right middle frontal gyrus; RPCC, right posterior cingulate; RPostCG, right postcentral gyrus; RPreCun, right precuneus; RThal, right thalamus.
BA, Brodmann area; LAMY, left amygdala; LCaudate, left caudate; LCun, left cuneus; LFFA, left fusiform face area; LIFG, left inferior frontal gyrus; LIOG, left inferior occipital gyrus; LINS, left insula; LIPL, left inferior parietal lobule; LMFG, left middle frontal gyrus; LMOG, left middle occipital gyrus; LMTG, left middle temporal gyrus; LPostCG, left postcentral gyrus; LPHG, left parahippocampal gyrus; LSFG, left superior frontal gyrus; LSPL, left superior parietal lobule; LSTG, left superior temporal gyrus; MedFG, medial frontal gyrus; MV, mood variability; RACC, right anterior cingulate cortex; RAMY, right amygdala; RFFA, right fusiform face area; RIFG, right inferior frontal gyrus; RINS, right insula; RIPL, right inferior parietal lobule; RMFG, right middle frontal gyrus; RPCC, right posterior cingulate; RPostCG, right postcentral gyrus; RPreCun, right precuneus; RThal, right thalamus.
Within N1, 50% inhibitory and 50% excitatory connections were positively or negatively related to mood variability. In N2, a similar pattern of results was observed with 55% inhibitory and 45% excitatory connections that were positively or negatively associated with mood variability. N3 demonstrated more inhibitory (72%) than excitatory (28%) connections to be linked to mood variability. In N4, 52% of inhibitory and 48% of excitatory connections were modulated by mood variability. Regarding the direction of the association between mood variability and effective connectivity, positive and negative relationships were more or less balanced across all connections within each network.
In terms of interconnectivity, we found that in N1, 32% of all connections showing an association with mood variability were projecting from frontal regions to frontal regions. In N2, frontal regions were mainly targeting frontal (30%) and temporal (30%) regions. In N3 and N4, the pattern of interconnectivity was less pronounced, with 16% of the connections projecting from limbic to limbic regions in N3, 24% of the connections projecting from parietal to parietal regions in N4.
In all networks, connections were determined that showed a reliable group difference (between SUD and controls) and a reliable relationship with mood variability in the SUD group (indicated in rectangles in Figure 4). In N1 and N2, 32% and 35% of all connections, respectively, demonstrated a reliable group difference and association with mood variability. In N3 and N4, 28% and 32% of all connections, respectively, differed between groups and were related to mood variability.
Supplementary analyses investigating mood variability—while controlling for the only significant comorbid diagnosis (lifetime depression) that demonstrated a positive relationship between mood variability and craving—revealed near-identical results (Figure S4; Tables S1, S3). Controlling for substance type and lifetime depression did not alter the previously reported findings (Figure S5; Tables S2, S4).
Leave-One-Out Cross-validation
Results of the LOOCV are shown in Figure 5 and reported in Table 6. LOOCV revealed that effect sizes were large enough to predict mood variability with an out-of-sample estimate for 4 connections in N1, 7 connections in N2, 4 connections in N3, and 7 connections in N4. This analysis indicates which specific connections within each network could significantly predict mood variability across patients with SUD.
Figure 5Leave-one-out cross-validation results for each network (N1 to N4) in patients. Only significant effect sizes that were large enough to predict mood variability with an out-of-sample estimate are illustrated (p < .05). Green/red arrows indicate a positive/negative relationship with mood variability in patients within each network. The thickness of arrows indicates positive/negative effective connectivity. The size of regions of interest indicates the number of inputs/outputs. Regions of interest are color-coded for various brain regions. BA, Brodmann area; LAMY, left amygdala; LCaudate, left caudate; LCun, left cuneus; LFFA, left fusiform face area; LIFG, left inferior frontal gyrus; LINS, left insula; LIOG, left inferior occipital gyrus; LIPL, left inferior parietal lobule; LMFG, left middle frontal gyrus; LMOG, left middle occipital gyrus; LMTG, left middle temporal gyrus; LPHG, left parahippocampal gyrus; LPostCG, left postcentral gyrus; LSFG, left superior frontal gyrus; LSPL, left superior parietal lobule; LSTG, left superior temporal gyrus; MedFG, medial frontal gyrus; RACC, right anterior cingulate cortex; RAMY, right amygdala; RFFA, right fusiform face area; RIFG, right inferior frontal gyrus; RINS, right insula; RIPL, right inferior parietal lobule; RMFG, right middle frontal gyrus; RPCC, right posterior cingulate; RPostCG, right postcentral gyrus; RPreCun, right precuneus; RThal, right thalamus.
Table 6Results of Leave-One-Out Cross-validation Analyses
Direction of Connectivity
Network and Effect of Connectivity
Source
Target
r Value
p Value
Network 1
LMFG_BA6
→
LIPL_BA40
0.22
.05
RINS_BA13
→
LIPL_BA40
0.38
<.001
RACC_BA23
→
RPreCun_BA7
0.28
.02
RACC_BA23
→
RIPL_BA40
0.29
.02
Network 2
LMFG_BA6
→
LSTG_BA39
0.27
.02
LSFG
→
LMTG
0.36
<.001
LSFG_BA9
→
LIFG_BA47
0.33
.01
LSFG_BA9
→
LSTG_BA39
0.31
.01
LIFG_BA47
→
LSFG
0.25
.04
LSTG_BA39
→
LIFG_BA47
0.27
.03
Network 3
MedFG_BA10
→
RAMY
0.33
.01
RFFA_BA37
→
RThal
0.31
.01
LPHG_BA27
→
RFFA_BA37
0.26
.03
RAMY
→
LPHG_BA27
0.23
.05
Nework 4
RPostCG_BA2
→
LCun_BA18
0.24
.04
RPostCG_BA2
→
LMOG_BA19
0.30
.01
LCun_BA18
→
LSPL_BA7
0.28
.02
RPCC_BA30
→
LPostCG_BA2
0.30
.01
RPCC_BA30
→
RPostCG_BA2
0.35
<.001
LCun_BA18
→
RPreCun_BA19
0.23
.05
BA, Brodmann area; LCun, left cuneus; LIFG, left inferior frontal gyrus; LIPL, left inferior parietal lobule; LMFG, left middle frontal gyrus; LMOG, left middle occipital gyrus; LMTG, left middle temporal gyrus; LPHG, left parahippocampal gyrus; LPostCG, left postcentral gyrus; LSPL, left superior parietal lobule; LSFG, left superior frontal gyrus; LSTG, left superior temporal gyrus; MedFG, medial frontal gyrus; RACC, right anterior cingulate cortex; RAMY, right amygdala; RFFA, right fusiform face area; RINS, right insula; RIPL, right inferior parietal lobule; RPCC, right posterior cingulate; RPostCG, right postcentral gyrus; RPreCun, right precuneus; RThal, right thalamus.
SUD is often experienced as a chronic, lifelong disorder due to lasting alterations in brain functioning induced by substance use as well as due to deleterious effects of comorbid mood disorders (
) and may have transdiagnostic significance. Using EMA, mood variability (but not mood tonality or severity) was found to be associated with increased craving in individuals with SUD. These findings were significant regardless of comorbid diagnosis or type of SUD, but they were of greater magnitude among persons with a history of depression. Taken together, these findings provide novel insights into the mechanisms that maintain craving and may therefore contribute to the global burden of SUD (
Alcohol consumption’s attributable disease burden and cost-effectiveness of targeted public health interventions: A systematic review of mathematical models.
GBD 2016 Alcohol and Drug Use Collaborators The global burden of disease attributable to alcohol and drug use in 195 countries and territories, 1990–2016: A systematic analysis for the Global Burden of Disease Study 2016.
To our knowledge, these findings are also the first to demonstrate the underlying causal network dynamics associated with mood variability in the daily lives of individuals with SUD. Overall, effective connectivity between groups was decreased for patients in the prefrontal control networks and increased in the emotion generation networks compared with healthy control subjects. The findings further revealed that effective connectivity within the patient group was modulated by mood variability and that a subset of these connectivity parameters were significant predictors of mood variability. However, contrary to our hypothesis, no specific direction of association was observed between effective connectivity and mood variability within the networks. Our findings, therefore, support the notion that mood variability and craving in persons with SUD are linked to the dynamic network configurations within the prefrontal control and subcortical emotion-generating networks at rest.
Patients with SUD compared with control subjects showed a pattern of more reduced connectivity in the prefrontal control networks (N1 and N2). This is generally consistent with previous work (
), which showed that reduced connectivity within the ECN (centered on nodes in the dorsolateral/ventrolateral prefrontal cortex and the lateral posterior parietal cortex) relates to impaired functioning in cognitive control. The cause of this decreased connectivity within regulatory networks in SUD has been proposed to be an impairment in white matter integrity (
). In N2, the reward-related regions, such as the striatum (i.e., caudate), also demonstrated more reduced connectivity with the prefrontal regions in line with previous studies (
). In the emotion generation and interoceptive networks (N3 and N4), no clear pattern of change emerged in connectivity strength. Interestingly, the effect of group differences was more pronounced for the inhibitory coupling in N2 and N4. This finding might indicate weak communication from interoceptive regions such as the insula involved in the monitoring of internal bodily states to achieve or maintain homeostasis (
The intrinsic causal dynamics in large-scale neural networks implicated in emotion generation and regulation play a predictive role in the patients’ susceptibility to experiencing mood variability (and subsequently, craving) in daily life. Patients with high mood variability showed a pattern of equally inhibitory and excitatory connectivity within cognitive control (N1 and N2) and interoceptive (N4) networks. The higher interconnectivity between frontal regions in the control networks linked to mood variability may represent an altered ability to downregulate emotions adaptively, including processes such as directing attention away from the emotional stimulus, maintaining the goal of the regulation strategy, selecting goal-appropriate strategies, and reinterpreting the meaning of the emotional stimulus/situation, thereby contributing to the manifestation of symptoms of mood variability in SUD.
Of particular interest in N2 are the effective connections of the striatum, a region that plays a key role in reward processing and addictive behavior. Bidirectional intrinsic changes of connection strength related to mood variability between the striatum and frontal/temporal regions were found. High mood variability was related to a decrease in connectivity from the frontal cortex to the striatum and an increase in connectivity from the striatum to the frontal and temporal cortex. These results expand on previous findings that linked emotion regulation, abstinence, craving, and subjective withdrawal to elevated connectivity between the ECN and reward network (
Within the emotion-generating network (N3), mood variability was linked to greater inhibitory connectivity, mainly originating from subcortical and limbic regions. This indicates that mood variability seems to modulate bottom-up, context-dependent signals important for emotional reactivity, the generation of emotional responses, and the reactivation of autobiographic memory processes. Of note, patients with high mood variability also showed reduced excitatory connectivity from regions involved in valuation (i.e., medial frontal gyrus) (
). Among all 4 networks, N4 is the least clear in terms of characteristic network dynamics (i.e., patterns of directionality). This aligns well with the notion that interoception in addiction (
) influences moment-to-moment information processing by identifying the most subjectively relevant stimuli that could arise from internal or external sources, thereby implementing a network-switching function between the default mode network and ECN (
The LOOCV results highlight the possibility that resting-state connectivity within all 4 networks might be particularly useful in predicting mood variability in daily life. To further support the clinical utility and predictive value of these connectivity alterations, it was important to consider whether the effects observed here were truly predictive of mood variability (or due to comorbid diagnoses or substance type) and that they constitute generalizable markers. Indeed, our findings did not differ when including lifetime depression and SUD type as control variables in our analysis, thus suggesting that differences in connectivity associated with mood variability are independent of these factors.
Several limitations of this investigation should be considered in interpreting the observed findings. First, as this represents the first study testing causal network dynamics in SUD in relation to mood variability, we did not differentiate between the different types of substances (i.e., cannabis, tobacco, or alcohol) in the overall analyses, and power considerations would not permit substance-specific analyses. However, the lack of significant alterations in the findings when controlling for substance type indicates that the magnitude of such differences is unlikely to be moderate or large. Related to this issue, it has to be noted that the findings might be limited in generalizability as a specific set of drugs has been examined, disregarding other drugs such as opioids. Second, we did not control for withdrawal states or duration of abstinence, which might affect changes in resting-state connectivity (
Large-scale resting-state DCM revealed systematic causal network dynamics in relation to mood variability among persons with SUD. As such, our study provides novel insights into how mood variability—as it naturally occurs in daily life—is associated with (and potentially predicted by) directional connectivity of emotion generative and regulatory brain networks. Patients with SUD compared with control subjects demonstrated differences in within-network effective connectivity, and these alterations were related to mood variability. Within each network, we determined connections that appear to predict mood variability in patients. Determining the causal nature of network dynamics underlying emotion regulation is an important step forward in informing intervention research through biomarkers of SUD risk factors.
Acknowledgments and Disclosures
This work was supported by Agence Nationale de la Recherche (Grant No. ANR-SH2-0012 [to JS]) and the Fondation Pour la Recherche Médicale (Grant No. DPA20140629807 [to JS]).
Further funding support for office and staff was provided by Hospital Charles Perrens (M. Au-riacombe for SANPSY). PI for SUD patient participants was M. Auriacombe, PI for healthy control participants was M. Allard.
The publication of this paper has been overshadowed by J. Swendsen's death in July. In deep sorrow, we dedicate this work to his memory.
The data that support the findings of this study are available from the corresponding author, CM, upon reasonable request.
The authors report no biomedical financial interests or potential conflicts of interest.
Alcohol consumption’s attributable disease burden and cost-effectiveness of targeted public health interventions: A systematic review of mathematical models.
The global burden of disease attributable to alcohol and drug use in 195 countries and territories, 1990–2016: A systematic analysis for the Global Burden of Disease Study 2016.
Emotion regulation as a transdiagnostic treatment construct across anxiety, depression, substance, eating and borderline personality disorders: A systematic review.
The role of anxiety sensitivity and difficulties in emotion regulation in posttraumatic stress disorder among crack/cocaine dependent patients in residential substance abuse treatment.
Comparative efficacy and acceptability of 21 antidepressant drugs for the acute treatment of adults with major depressive disorder: A systematic review and network meta-analysis.
Efficacy and tolerability of antidepressants in the treatment of adolescents and young adults with depression and substance use disorders: A systematic review and meta-analysis.
Rostral anterior cingulate network effective connectivity in depressed adolescents and associations with treatment response in a randomized controlled trial.
The Mini-International Neuropsychiatric Interview (M.I.N.I.): The development and validation of a structured diagnostic psychiatric interview for DSM-IV and ICD-10.
Effects of anxiety and mood disorders on craving and substance use among patients with substance use disorder: An ecological momentary assessment study.
00275-0&id=gr1.jpg){kind=link}
00275-0&id=gr2.jpg){kind=link}
00275-0&id=gr3.jpg){kind=link}
00275-0&id=gr4.jpg){kind=link}
00275-0&id=gr5.jpg){kind=link}