Tackling heterogeneity: Individual variability of emotion decoding deficits in severe alcohol use disorder.

Abstract

Background:

Severe alcohol use disorder (SAUD) is associated with social cognition deficits. Patients with SAUD are impaired for the recognition of emotional facial expressions, particularly at early stages of abstinence. These deficits damage interpersonal relations and increase relapse risk. However, uncertainties still abound on their variation across emotions and on the heterogeneity of emotional impairments across patients. We addressed these questions by exploring how the deficit varies according to emotions’ type/intensity and patients’ heterogeneity.

Methods:

Sixty-five recently detoxified patients with SAUD and 65 matched healthy controls performed the Facial Emotion Recognition Test, assessing the ability to identify six emotions (anger, contempt, disgust, fear, happiness, sadness) displayed by morphed faces with various intensities. Accuracy scores and detection thresholds were collected for each emotion. Beyond group comparisons, multiple single-case analyses determined the percentage of patients presenting decoding deficits for each emotion.

Results:

When current depression and anxiety symptoms were controlled for, patients did not present a general emotion decoding deficit, but were rather characterized by specific deficits for disgust/contempt in accuracy, and for disgust in detection threshold scores. Single-case analyses showed that only a third of patients presented a clinically significant emotional deficit.

Conclusions:

Patients with SAUD only present emotional decoding deficits for specific interpersonal emotions (disgust/contempt) when subclinical psychopathological states are controlled for, and show no general emotional impairment. This goes against the proposal of a generalized social cognition deficit in this population. This group effect moreover masks a massive heterogeneity across patients, which has implications at experimental and clinical levels.

1. Introduction

The currently prevailing theoretical frameworks of severe alcohol use disorder (SAUD), namely the dual-process models (McClure & Bickel, 2014; Stacy & Wiers, 2010), have conceptualized this addictive state as a progressive loss of inhibitory control over consumption (related to an under-activated prefrontal / reflective system) combined with an increased attraction towards alcohol-related stimuli (related to an over-activated striato-limbic / impulsive system). Empirical data have largely supported the existence of this imbalance between impaired cognitive abilities (Brion et al., 2017a; Stavro et al., 2013) and frontal lobe functioning (Arcurio et al., 2015; Bühler & Mann, 2011) on the one hand, and increased activation of the impulsive system (Field & Cox, 2008; Maurage et al., 2020; Wrase et al., 2007) on the other hand. Following this view, most current therapeutic programs aim at reestablishing equilibrium between the two systems, by increasing cognitive control and / or reducing automatic attraction (Keshavan et al., 2014; Rolland et al., 2019).

However, dual-process models have neglected other processes involved in SAUD, among which emotional and interpersonal disorders occupy a key position. Indeed, recently detoxified patients with SAUD are not exclusively characterized by an imbalance between controlled and automatic processes. SAUD is also an affective and relational disease: strong comorbidities exist between SAUD and mood disorders, most patients with SAUD presenting negative emotionality, as well as depressive and anxious symptomatology (Schuckit, 2006). SAUD is also associated with familial, professional, and relational problems (Hunter-Reel et al., 2009). Moreover, it has been suggested that patients with SAUD present reduced social network (Chou et al., 2011), limited social support (Ammon et al., 2008), and persistent social difficulties (Levola et al., 2014). A more inclusive conceptualization of SAUD should, therefore, encompass these socio-emotional impairments. A recent renewal of the dual-process approach, the triadic perspective (Flaudias et al., 2019; Noël et al., 2013), has initiated this inclusion by identifying the insular system (notably involved in interoceptive and affective abilities) as a third important component. However, these conceptual proposals still fail to fully integrate socio-emotional processes in their framework of addictive disorders. This might be related to the fact that evidence-based studies have long neglected these processes, as the experimental exploration of social cognition in SAUD has only emerged recently.

Social cognition refers to the processes required to reliably perceive, understand, regulate, and react to social stimulations emerging from other individuals and the environment (Green et al., 2015). An influential classification (Green et al., 2008) has identified five subcategories of social cognition abilities: Theory of Mind, social perception, social knowledge, attributional bias and emotional processing / decoding. Beyond memory and executive functions, the exploration of social cognition has improved the understanding of the associated deficits and relapse factors in SAUD (Bora & Zorlu, 2017; Le Berre, 2019; Onuoha et al., 2016). Impairments have been documented for Theory of Mind (e.g., Bosco et al., 2014; Maurage et al., 2015; Thoma et al., 2013), social perception (e.g., Maurage et al., 2012; 2016), social knowledge (e.g., Amenta et al., 2013; Maurage et al., 2013; Uekermann et al., 2007) and attribution bias (Pabst et al., 2020a). However, the social cognition subcategory that has been by far the most investigated in SAUD is emotional decoding (i.e., the ability to identify and label the emotions presented by other individuals). This focus is justified by the crucial role played by emotional decoding abilities in interpersonal functioning (Feldman et al., 1991), making it relevant in a large range of psychopathological states (Power & Dalgleish, 1997). Since the publication of a seminal study describing a reduced ability to identify the emotional states expressed by human faces in SAUD (Philippot et al., 1999), more than fifty papers have explored emotional decoding through a large variety of experimental designs (Donadon & Osório, 2014). Most results focused on facial expressions decoding and confirmed that patients with SAUD need more emotional intensity to detect others’ affective states, particularly for negative emotions (see Bora & Zorlu, 2017; Castellano et al., 2015 for meta-analyses). This impairment has been generalized to all types of emotional stimuli, including voice prosody and body posture (Maurage et al., 2009). Emotional decoding is thus now considered a key deficit in alcohol-related disorders, which is further reinforced by the fact that such impairments have been documented on the whole spectrum of alcohol consumption, including acute alcohol intoxication (Kamboj et al., 2013; Khouja et al., 2019), sub-clinical consumption patterns (Lannoy et al., 2019; Leganes-Fonteneau, 2020), and Korsakoff syndrome (Brion et al., 2017b; 2018).

Since this decoding ability is crucial to understand other individuals’ state of mind and feelings, social cognition impairments in SAUD play a key role in the reduction of interpersonal bonds and in the interpersonal problems frequently observed in this population (Freeman et al., 2018; Kornreich et al., 2002). Socio-emotional disturbances are thus not secondary variables in SAUD treatment: social support constitutes a major abstinence factor (Gordon & Zrull, 1991), and more than one-third of relapses are directly attributed to persistent negative affects or altered social network (Zywiak et al., 2003). Negative emotionality and disrupted interpersonal functioning are thus at the heart of this addictive disorder. They are strongly involved in its persistence (Sliedrecht et al., 2019) and their rehabilitation constitutes a promising therapeutic avenue to improve clinical outcomes (Rupp et al., 2017).

However, despite the consensus that emotion decoding is impaired in SAUD, several debates persist regarding the characteristics and extent of this deficit. Recent review papers (Bora & Zorlu, 2017; Le Berre, 2019) have notably raised two questions still to be addressed, related to the variation of this deficit according to:

(1) the type and intensity of the emotions to be decoded. First, the strong deficit observed for negative emotions decoding (e.g., Kornreich et al., 2013) might be generalized to positive emotions (Maurage et al., 2011) or might be identical across all emotions (D’Hondt et al., 2015; Frigerio et al., 2002). As most previous studies did not explore a wide range of emotions, this differential deficit across emotions remains to be clarified. Second, most previous works have used full-blown emotions, which present a low ecological value (emotions in everyday life being most often subtle or mixed, e.g., D’Hondt et al., 2015; Townshend & Duka, 2003). Only a few earlier works have used stimuli with lower intensity or morphed faces in SAUD, and they have led to mixed results: some reported a global deficit (e.g., Kornreich et al., 2001) while others identified a specific deficit for some emotions (e.g., Frigerio et al., 2002; Kumar et al., 2011; O’Daly et al., 2012) or even an absence of deficit for morphed faces (e.g., Trick et al., 2014). The impact of emotional type / intensity on emotional deficit decoding in SAUD thus remains to be established;

(2) the patients’ individual differences, as previous studies have focused on a mere group comparison between patients and healthy controls, considering patients as a homogeneous population. However, data on cognitive functioning indicate that patients with SAUD do not constitute a unitary group, as cognitive deficits concern around 50% of them (Alarcon et al., 2015) and are strongly modulated by coexisting psychopathological symptoms (D’Hondt et al., 2018). The same heterogeneity might be at stake regarding interpersonal difficulties, as an analysis of individual performances during a Theory of Mind task showed that only half of patients presented a genuine deficit (Maurage et al., 2015). Another study explored this heterogeneity in socio-emotional disorders using a cluster analytic approach (Maurage et al., 2017) and showed that patients with SAUD present a wide variety of socio-emotional profiles, around 25% of them reporting no emotional or interpersonal problems. There is thus an urgent need to go beyond the classical group approach through individual analyses specifying the inter-individual variation of emotional deficits. Indeed, the group differences reported in previous studies might mask a massive heterogeneity explaining the discrepancies observed in earlier studies, but also hampering the application of emotion decoding rehabilitation programs in clinical settings.

The main aim of the present study is, therefore, to address these two questions through the exploration of emotional decoding abilities in a large sample of recently detoxified patients with SAUD, with a design allowing the exploration of the deficit’s variation according to emotion type / intensity and patients’ heterogeneity. We hypothesized that patients with SAUD will be particularly impaired for the processing of subtle / low intensity emotional signals, and negative emotions signaling interpersonal threat (e.g., anger, disgust; Bora & Zorlu, 2017; Pabst et al., 2020b). We also expected that a large heterogeneity would exist across patients, some presenting generalized emotion decoding deficits while others would present totally preserved abilities (Maurage et al., 2017).

2. Material and Methods

2.1. Participants

Sixty-five recently detoxified patients, diagnosed with SAUD according to DSM-5 criteria (American Psychiatric Association, 2013), and 65 matched healthy controls (HC) took part in the study (see Table 1 for demographic data). Considering the effect sizes (0.65-0.67) reported in recent meta-analyses (Bora & Zorlu, 2017; Castellano et al., 2015), our study had a power of more than 0.97 to detect between-group differences in global emotion recognition abilities. SAUD diagnosis was established during a comprehensive interview performed by a trained psychiatrist. This semi-structured interview was also used to precisely determine medical history, alcohol consumption characteristics, and psychopathological comorbidities. Participants with SAUD were recruited during their third week of detoxification (Saint-Luc Hospital, Brussels, Belgium) and had all abstained from alcohol for at least 14 days. They were free of any other psychiatric diagnosis, including past / present other substance use, addictive disorder (except nicotine dependence) and depression / anxiety disorders. Their mean alcohol consumption before detoxification was 209.4 grams of pure ethanol per day (SD=114.8). The mean number of previous detoxification treatments was 1.11 (SD=1.58) and the mean duration of SAUD was 13.15 years (SD=10.99). HC were free of any history of psychiatric disorder or drug / substance abuse (excluding nicotine dependence). The mean alcohol consumption in the control group was 17.8 grams of pure ethanol per day (SD=19.2), and their mean AUDIT score [classically used as a global screening of alcohol consumption in the general population (Saunders et al., 1993)] was 3.52 (SD=1.83), all HC presenting a low-risk alcohol consumption (i.e., AUDIT score lower than 8). HC were free of any medication but 42 participants with SAUD still received low doses of benzodiazepines (M=12.38 milligrams of diazepam per day, SD=21.31). Education level was assessed according to the number of years of education completed since starting primary school. Exclusion criteria for both groups included major medical problems, neurological disease, and polysubstance use. All participants were right-handed, as evaluated by the Oldfield test (Oldfield, 1971). Participants were informed with full details regarding the aims of the study, provided written informed consent to take part in the study, and were tested individually in a quiet room. The study was approved by the Ethical Committee of the Medical School (UCLouvain) and carried out according to the Declaration of Helsinki, as revised in 2008. Participants were not paid for their participation. This experiment was part of a larger project, in collaboration with the GDR 3557, investigating social cognition impairments in SAUD (e.g., Maurage et al., 2016).

Table 1.

Demographic and psychopathological characteristics of participants with severe alcohol use disorder (SAUD) and healthy control participants (HC): mean (SD).

	SAUD (N=65)	HC (N=65)
Gender ratio (M/F) NS	32/33	32/33
Age (in years) NS	48.89 (9.04)	48.60 (9.41)
Education level (in years) NS	13.31 (2.39)	13.98 (2.98)
BDI1, *	9.08 (6.95)	2.06 (2.81)
STAI-A2, *	39.49 (13.42)	29.86 (7.69)
STAI-B2, *	46.41 (12.11)	35.92 (8.26)

¹BDI = Beck Depression Inventory (Beck and Steer, 1987).

²STAI = State (A) and Trait (B) Anxiety Inventory (Spielberger et al., 1983).

NS= No significant group difference

^*p<0.001 for group comparison

2.2. Measures and procedure

2.2.1. Control measures

Validated self-completion questionnaires were used to assess the presence of depressive (Beck Depression Inventory; Beck and Steer, 1987) and anxious (State and Trait Anxiety Inventory; Spielberger et al., 1983) symptoms or traits.

2.2.2. Experimental measure

Participants performed the Facial Emotion Recognition Test (TREF; Gaudelus et al., 2015), assessing the ability to identify six emotions (anger, contempt, disgust, fear, happiness, sadness) displayed by facial expressions. Emotions were expressed by six adult actors (three females) and emotional stimuli were morphed (from neutral face to 100% emotional expression) to evaluate nine emotional intensity levels (i.e., from 20% to 100%), with a total of 54 stimuli (6 emotions X 9 intensity levels). Stimuli (Figure 1) were randomly and individually presented on a Dell Latitude E5530 laptop located at 60cm from the participant’s eyes, sustaining a visual angle of 11.4 X 13.3° (stimuli dimensions: 12 X 14cm). Each stimulus was accompanied by a six-label list on the right side of the screen (reminding the six emotional labels). Each emotional label was associated with a response key corresponding to the emotion’s first letter (stickers with these letters were placed on the same line of the keyboard to facilitate the answer). Participants had to indicate which emotion was presented by pressing the corresponding button with their dominant hand. Each stimulus was presented for 10s, but participants had no time limit to answer. Two scores were computed for each emotion: accuracy (i.e., the percentage of correct responses) and detection threshold (i.e., the minimal intensity at which the emotion is constantly detected without error). This threshold was obtained by averaging the intensity of the first correct recognition (i.e., first morphing level leading to a correct answer for an emotion), and the intensity from which the emotion is perfectly recognized (namely the intensity from which the emotion is always detected correctly), with the following formula: (first threshold + perfect threshold)/2. Global scores (averaging the scores for each emotion) were also computed for accuracy and detection threshold. For accuracy, higher scores represent better emotion decoding (i.e., a higher rate of correct detection) while for detection threshold, higher scores represent reduced performance (i.e., higher emotional intensity needed to correctly detect the emotion).

Figure 1.

Stimuli used in the experimental task, each of the six emotions (anger, contempt, disgust, fear, happiness, sadness) being represented in nine stimuli of increasing intensity (from 20% to 100%).

2.3. Data analytic plan

Statistical analyses were performed using IBM SPSS Statistics (Version 22.0. Armonk, NY: IBM Corp.) for group analyses and Singlims (Crawford and Howell, 1998) for multiple single-case analyses. The following strategy was used. First, between group comparisons were performed on demographic (age, gender, and educational level) and psychopathological (depression, and state / trait anxiety) characteristics. The variables for which significant group differences were observed (i.e., depression and anxiety scores, see below) were included as covariates in the following analyses. Note that, whereas the assumption for performing these covariate analyses were generally met¹, the independence between the independent variable (i.e., groups) and the covariates (i.e., psychopathological states) was obviously not respected due to the significant group differences for depression / anxiety symptoms. Second, two repeated measures analysis of variance (ANOVA) were performed without and with covariates, respectively on TREF accuracy and detection thresholds, with groups (SAUD, HC) as between-subjects factor and emotion (Anger, Contempt, Disgust, Fear, Happiness, Sadness) as within-subjects factor. The Greenhouse-Geisser correction was applied when appropriate. For each ANOVA, significant main effects and interactions were followed by univariate contrasts (post-hoc independent samples t-tests). Pearson’s correlational analyses were also performed to test the links between emotional decoding performance, psychopathological measures and alcohol-related variables. Multiple single-case analyses were then conducted on TREF accuracy and detection threshold to estimate the percentage of patients with SAUD presenting a significant deficit, beyond group comparisons. To do so, each participant with SAUD was compared with a group of five HC (individually matched with the patient for age, gender and education level) using Crawford’s adapted t-test (Crawford and Garthwaite, 2005; Crawford et al., 2003), namely t=[(mean patient - mean group)/(SD group √((N group + 1)/N group))] with one-tailed p-value. This allowed determining the percentage of participants with SAUD presenting significantly reduced performance (reduced accuracy and / or increased detection threshold) for each emotion separately and for mean scores. Finally, we also determined the percentage of participants with SAUD presenting significant deficits for each emotion and total score, according to the guidelines proposed in the original TREF study (Gaudelus et al., 2015). This validation study indeed determined that scores below 61.58% indicated clinical impairments.

3. Results

3.1. Demographic and psychopathological measures

As illustrated in Table 1, there were no significant group differences for age [t(128)=0.18, p=0.857], gender, or education level [t(128)=1.43, p=0.156], but the two groups significantly differed for depression [t(128)=7.54, p<0.001], state anxiety [t(128)=5.02, p<0.001], and trait anxiety [t(128)=5.77, p<0.001].

3.2. Experimental measures

These results are reported in Table 2.

Table 2.

Accuracy score (percentage of correct answers) and detection threshold for participants with severe alcohol use disorder (SAUD) and healthy control participants (HC) at the Facial Emotion Recognition Test (for each emotion and mean score): mean (SD).

	SAUD (N=65)	HC (N=65)
Accuracy score
Anger NS	53.85 (18.03)	56.69 (23.15)
Contempt *	40.68 (19.96)	50.43 (20.23)
Disgust **	53.85 (14.06)	64.24 (10.11)
Fear NS	77.61 (17.84)	80.88 (13.95)
Happiness NS	85.98 (15.99)	86.81 (11.26)
Sadness NS	71.80 (16.44)	74.89 (17.93)
Mean score NS	63.96 (9.84)	68.99 (8.19)
Detection threshold score
Anger NS	62.46 (14.74)	62.08 (20.29)
Contempt NS	69.62 (13.47)	66.62 (12.22)
Disgust **	65.69 (15.51)	53.54 (9.99)
Fear NS	43.38 (17.12)	39.85 (14.87)
Happiness NS	32.62 (15.55)	32.46 (10.65)
Sadness NS	50.92 (15.93)	46.31 (16.54)
Mean score NS	54.15 (8.08)	50.15 (7.10)

NS= No significant group difference

^*p<0.01 for group comparison

^**p<0.001 for group comparison

3.2.1. Accuracy scores

In the statistical model without covariates, main effects were found for Group [F(1,128)=10.03, p=0.002, η²_p=0.073], patients with SAUD presenting lower scores than HC, and emotion [F(5,640)=128.88, p<0.001, η²_p=0.502], without significant Group X Emotion interaction [F(5,640)=2.11, p=0.063, η²_p=0.016].

However, when the covariates were included in the analysis², the main Group effect disappeared [F(1,125)=1.27, p=0.261, η²_p=0.010], but a significant main effect of emotion [F(5,625)=8.57, p<0.001, η²_p=0.064] and a Group X Emotion interaction [F(5,625)=2.51, p=0.029, η²_p=0.020] were found.

Regarding the emotion effect, Happiness was related to higher scores than Anger [t(129)=15.63, p<0.001], Contempt [t(129)=19.32, p<0.001], Disgust [t(129)=16.94, p<0.001], Fear [t(129)=4.51, p<0.001] and Sadness [t(129)=6.73, p<0.001]; Fear was related to higher scores than Anger [t(129)=11.94, p<0.001], Contempt [t(129)=15.40, p<0.001], Disgust [t(129)=12.49, p<0.001] and Sadness [t(129)=3.14, p=0.002]; Sadness was related to higher scores than Anger [t(129)=8.49, p<0.001], Contempt [t(129)=12.84, p<0.001] and Disgust [t(129)=7.56, p<0.001]; Contempt was related to lower scores than Anger [t(129)=4.88, p<0.001] and Disgust [t(129)=6.60, p<0.001], which did not differ [t(129)=1.81, p=0.073].

Importantly, the Group X Emotion interaction was explained by the fact that groups only differed for Contempt [t(128)=2.76, p=0.007] and Disgust [t(128)=4.84, p<0.001], patients with SAUD presenting lower performance than HC, while groups did not significantly differ for other emotions: Anger [t(128)=0.78, p=0.436], Fear [t(128)=1.16, p=0.246], Happiness [t(128)=0.34, p=0.733] and Sadness [t(128)=1.02, p=0.307].

3.2.2. Detection threshold scores

In the statistical model without covariates, main effects were found for Group [F(1,128)=8.96, p=0.003, η²_p=0.065], patients with SAUD presenting higher detection thresholds than HC, and emotion [F(5,640)=121.01, p<0.001, η²_p=0.486], with a Group X Emotion interaction [F(5,640)=3.15, p=0.008, η²_p=0.024].

However, when the covariates were included in the analysis, the main Group effect disappeared [F(1,125)=0.93, p=0.336, η²_p=0.007], but a significant main effect of emotion [F(5,625)=10.66, p<0.001, η²_p=0.079] and a Group X Emotion interaction [F(5,625)=3.12, p=0.009, η²_p=0.024] were found.

Regarding the emotion effect, Happiness was related to lower detection threshold than Anger [t(129)=17.44, p<0.001], Contempt [t(129)=24.46, p<0.001], Disgust [t(129)=16.03, p<0.001], Fear [t(129)=5.42, p<0.001] and Sadness [t(129)=8.78, p<0.001]; Fear was related to lower detection threshold than Anger [t(129)=11.09, p<0.001], Contempt [t(129)=15.23, p<0.001], Disgust [t(129)=10.68, p<0.001] and Sadness [t(129)=4.09, p=0.002]; Sadness was related to lower detection threshold than Anger [t(129)=7.00, p<0.001], Contempt [t(129)=10.88, p<0.001] and Disgust [t(129)=6.13, p<0.001]; Contempt was related to higher detection threshold than Anger [t(129)=3.31, p=0.001] and Disgust [t(129)=4.97, p<0.001], which did not differ [t(129)=1.35, p=0.180].

Importantly, the Group X Emotion interaction was explained by the fact that group only differed for Disgust [t(128)=5.31, p<0.001], patients with SAUD presenting lower performance than HC, while groups did not significantly differ for other emotions: Anger [t(128)=0.12, p=0.902], Contempt [t(128)=1.33, p=0.186], Fear [t(128)=1.26, p=0.211], Happiness [t(128)=0.07, p=0.945] and Sadness [t(128)=1.62, p=0.108].

3.2.3. Exploratory correlational analyses

- Test for medication effect: No significant correlation was observed between experimental results (global accuracy and decision threshold) and benzodiazepine consumption (all ps>0.05).

- Test for the link between emotion decoding performance and alcohol-related variables: No significant correlation was observed between experimental results and alcohol-related variables (i.e., number of grams of ethanol per day before detoxification, number of previous detoxifications, duration of SAUD; all ps>0.05).

- Test for the link between emotion decoding performance and psychopathological symptoms: Among HC, accuracy was negatively correlated with depressive symptoms (r=−0.32, p=0.009). Among patients with SAUD, accuracy and detection threshold were correlated with depressive (r=−0.24, p=0.049 and r=0.28, p=0.03, respectively) and state anxiety (r=−0.32, p=0.008 and r =0.29, p=0.02, respectively) symptoms. All other correlations were non-significant (all ps>0.05).

3.2.4. Multiple single-case analysis

- Percentage of impaired individuals with SAUD in multiple single-case analyses: When each individual with SAUD was compared with a matched group of five HC, 26.15% of patients with SAUD presented a significant global deficit for accuracy (comparison on the mean score across emotions), with 30.77%-18.46%-4.62%-4.62% of patients presenting a significant deficit for 1-2-3-4 emotions, respectively, and 41.53% of patients without deficit for any specific emotion. Regarding detection thresholds, 16.92% of patients with SAUD presented a significant global deficit, with 38.46%-12.31%-6.15% of patients presenting a significant deficit for 1-2-3 emotions, respectively, and 43.08% of patients without deficit for any specific emotion.

When independently exploring the results related to each emotion (Figure 2), 30.77% of patients with SAUD had significantly lower accuracy scores for disgust, 16.92% for fear and sadness, 13.85% for contempt and happiness, and 7.69% for anger. This heterogeneity across patients with SAUD is further illustrated by the accuracy scores related to each individual for each emotion, depicted in the Supplementary Figure. Regarding detection thresholds, 36.92% of patients with SAUD had a significantly higher threshold (i.e., needed a higher emotional intensity for correct identification) for disgust, 13.85% for fear, 12.31% for contempt, 9.23% for happiness, 7.69% for sadness and 1.53% for anger.

Figure 2.

Percentage of patients with SAUD presenting significant decoding impairment in accuracy (left part) and detection threshold (right part) for each emotion (anger, contempt, disgust, fear, happiness, sadness), as determined through multiple single-case analyses (Crawford’s adapted t-tests).

- Percentage of impaired individuals with SAUD in task norms comparisons: 41.54% of individuals with SAUD exhibited a global score lower than the cut-off score identified in the validation study as indicating a clinically significant emotion recognition deficit.

4. Discussion

We investigated how the previously documented emotional decoding deficit in SAUD varies according to the type / intensity of the emotions presented and to patients’ heterogeneity. We addressed these two questions in a large sample of recently detoxified patients with SAUD, through the validated Facial Emotion Recognition Test allowing the simultaneous exploration of six basic emotions and including nine morphing levels for each emotion. Moreover, a multiple single-case analysis determined the variability of this deficit across patients. This multi-level and multi-method approach led to two main results, offering key insights on emotional disturbances in SAUD.

First, the emotional decoding deficit strongly varies across emotions. We replicated previous results showing higher performance for happiness than for negative emotions (e.g., D’Hondt et al., 2015; Wegrzyn et al., 2017), as well as stronger difficulties for contempt compared to other emotions (e.g., Langner et al., 2010), independently of the group. More centrally, patients presented a globally lower accuracy than HC in the initial analysis but, when current depressive / anxious symptoms were controlled for, this main group effect disappeared. The only remaining emotions for which patients presented reduced accuracy were contempt and disgust. Several studies have observed specific disgust processing disorders (Dursun et al., 2007; Quaglino et al., 2015), and such a difficulty to correctly identify disgust in faces at the early stages of abstinence is related to higher relapse risk and treatment dropout rates (Rupp et al., 2017). Moreover, a recent meta-analysis (Bora & Zorlu, 2017) identified disgust as the most impaired emotion in SAUD, which is reinforced here by the fact that disgust is the only emotion related to a deficit for both accuracy and detection threshold in SAUD. Conversely, contempt has been far less explored since the identification of contempt processing impairments in the very first study exploring emotional decoding in SAUD (Philippot et al., 1999). We show here that the lower decoding accuracy observed for disgust can be generalized to contempt. These two emotions are distinct from other negative ones (e.g., sadness, fear) as they are by essence related to interpersonal relations and social judgment (Bilewicz et al., 2017; Chapman et al., 2009). An impaired identification of these emotions carrying crucial cues for efficient social adaptation might thus foster maladaptive behaviors during social interactions and more global social cognition deficits. Most previous studies also identified anger processing impairments in SAUD (Bora & Zorlu, 2017) which is not found here although anger is also considered a key social emotion. A closer inspection of anger processing impairments suggests that patients present a biased anger processing rather than a mere decoding impairment, as they over-estimate anger in faces presenting other negative emotions (e.g., Frigerio et al., 2002; Maurage et al., 2008). This anger over-attribution is further reinforced by neuroimaging results showing increased cingulate activations during anger processing in SAUD (Park et al., 2015). As anger is a strong indicator of social threat, this over-estimation might lead patients to over-react aggressively to others’ facial expressions or to avoid social interactions (Pabst et al., 2020b). Applied to our results, this anger over-estimation explains why patients do not present any deficit for anger. However, the decoding performance for anger among HC was lower than the one observed in most earlier studies using the same task (Gaudelus et al., 2015; Lannoy et al., 2019, but see also Kitoko et al., 2019). This could be related to the fact that these studies focused on a younger and more educated control group, but it might partly explain the absence of group differences for anger in the present study. As a whole, the emotion decoding deficit in SAUD appears specific to disgust and contempt (two emotions strongly related to social interactions) rather than generalized for all emotions. It should however be noted that this key result might be partly related to (1) the emotions included in the experiment, as the inclusion of other emotional labels might have modified the results (e.g., including surprise stimuli might have reduced fear decoding performance, as these two emotions are frequently mixed); (2) a global difficulty effect, as disgust and contempt are among the emotions leading to the lowest accuracy and detection threshold performances (while disgust was related to similar performances than anger, for which no deficit was found in severe alcohol use disorder). Moreover, this result might appear at odds with several previous ones reporting a general emotional decoding impairment in SAUD, particularly for all negative emotions (Acharya & Dolan, 2012; Frigerio et al., 2002; Kornreich et al., 2001; Maurage et al., 2007; 2008). This could be explained by the fact that (1) these earlier studies focused on a limited set of emotions. Some deficits might thus have been misestimated in studies only presenting a limited emotional range. Particularly, disgust and contempt which, as shown here, are centrally involved in the deficit, have most often not been evaluated; (2) most earlier studies did not control for coexistent psychopathological symptoms, thus mixing the core deficit related to socio-affective impairments in SAUD with those related to current subclinical depressive or anxious symptoms. As emotional decoding impairments have been repeatedly documented in major depression and anxiety disorders (e.g., Berg et al., 2016; Dalili et al., 2015; Plana et al., 2014), and as patients with SAUD are frequently facing transitory psychopathological symptoms during detoxification (e.g., Hasin & Grant, 2015; Ribadier & Varescon, 2019), our results suggest that the influence of such current states should be controlled for to specify the emotional impairments related to SAUD, beyond the influence of depression / anxiety. It should nevertheless be underlined that, as the presence of clinically significant (i.e., diagnosed through DSM 5 criteria) mood disorders constituted an exclusion criterion, and as we only measured short-term subclinical depressive / anxiety symptoms through self-reported questionnaires, our study did not allow to disentangle the influence of current subclinical depressive / anxious symptoms from the potential effect of diagnosed mood disorder.

Of note, the emotional decoding deficit also varies according to the morphing level. A globally increased detection threshold is found among patients in the initial analysis, indexing that they need higher emotional intensity to identify emotions. These globally increased thresholds are in line with a previous study also using morphed stimuli among non-depressed patients with SAUD (D’Hondt et al., 2015). Here again though, when psychopathological symptoms where controlled for, this main group effect disappeared, disgust remaining the only emotion leading to an increased detection threshold. This finding confirms that, when using a controlled paradigm with a wide range of emotions in a large sample of patients, the emotion decoding deficit is not global but specific to disgust (for accuracy and detection threshold) and contempt (for accuracy). The absence of correlation between this emotion decoding deficit and medication suggests that this impairment is quite independent from current benzodiazepines intakes. Moreover, as no significant link has been observed between emotional impairments and alcohol-related variables, it can be proposed that socio-affective difficulties might constitute a core and stable characteristic of SAUD (not strongly related to the current intensity / duration of SAUD) or might even be present before the emergence of SAUD and thus constitute a premorbid risk factor for addictive disorders.

Second, we proposed the first exploration of the inter-individual variability of emotion decoding deficit in SAUD. Previous studies have suggested that patients with SAUD do not constitute a unitary group regarding cognitive (Alarcon et al., 2015; Ihara et al., 2000) or Theory of Mind (Maurage et al., 2015) abilities, only 50% of the patients presenting a significant impairment when individually compared to matched controls. By applying the same approach, we showed that emotion decoding performance strongly varies across patients: the group effect observed for disgust was driven by only a third of the patients, presenting a significant individual deficit according to multiple single-case analyses. Although these analyses did not control for psychopathological symptoms, the percentage of impaired patients was even lower for other emotions, none of them leading to impairment rates higher than 20%. Beyond showing that the emotion decoding deficit is not generalized across emotions, we further prove that this deficit is far from being homogeneous in SAUD.

A global emotional deficit is only found in a restricted proportion of patients (26.15% of patients for accuracy, 16.92% for detection thresholds in the multiple single-case analyses), with a massive heterogeneity (further illustrated in the Supplementary Figure): a third of the sample present a deficit limited to one specific emotion, only a small proportion (less than 20%) being impaired for more than one emotion and more than 40% not presenting any deficit for a specific emotion. This important result aligns well with those from a previous study (Maurage et al., 2017) showing, through self-reported questionnaires, that patients with SAUD are heterogeneous regarding emotional and interpersonal difficulties: whereas some (particularly those with high depression / anxiety symptoms) reported massive impairments in self-related (i.e., alexithymia) and / or other-related (i.e., interpersonal difficulties) socio-affective abilities, a significant part of the sample reported limited or even no social cognition impairments. In the same vein, a study (Lannoy et al., 2019) using a similar approach to the present one among subclinical populations with alcohol use disorder (i.e., binge drinkers) led to an identical conclusion: a significant deficit was reported in less than a quarter of the binge drinkers for each emotion.

This leads to the proposal that patients with SAUD should no longer be seen as a unitary group when considering emotion decoding abilities, but rather as a sum of individuals presenting a wide variety of emotional profiles. At the experimental level, this result implies that future studies should go beyond the classical group comparison approach still widely used in SAUD research, to systematically determine which part of this population is concerned by this deficit. More importantly, future studies should focus on understanding this heterogeneity and investigate the specificities of patients presenting emotion recognition deficits. This recommendation applies to studies exploring affective but also psychological or neuropsychological processes in SAUD. Moreover, it would also be worth reconsidering previous studies on emotion decoding abilities to determine, beyond the group comparisons classically reported, the heterogeneity of the SAUD sample and therefore the strength of the reported results. At the clinical level, our results related to group comparisons, together with the large previous literature on social cognition deficits in SAUD, claim for the urgent development of rehabilitation programs focusing on these abilities. Social cognition deficits are currently under-evaluated and under-treated in clinical settings. Cognitive remediation programs have recently bloomed in addictive disorders (Rolland et al., 2019; Verdejo-Garcia, 2016), but these programs have been focused on the rehabilitation of purely cognitive abilities. In line with what has been developed in schizophrenia (Grant et al., 2017), structured neuropsychological programs should incorporate the evaluation and training of social cognition, as it could positively impact clinical outcomes of SAUD by breaking the vicious cycle linking social cognition deficits, social isolation, and alcohol consumption. Our multiple single-case analyses also showed that the emotion decoding deficit is specific for interpersonal emotions and strongly varies across patients, which claims for the individual evaluation of social cognition in SAUD, as well as for tailored rehabilitation proposals based on the determination of the idiosyncratic impairments presented by each patient. Finally, in view of the role played by psychopathological symptoms in cognitive (e.g., D’Hondt et al., 2018) and emotional deficits among patients with SAUD, as further illustrated by the present results, a specific screening for such current depressive or anxious states should be systematically conducted during detoxification, to allow their early and efficient treatment in this population, and consequently to reduce the extent of social cognition impairments.

Highlights.

• Severe alcohol use disorder (SAUD) is related to emotion decoding deficits

• We explored how this deficit individually varies with emotion type/intensity

• Patients with SAUD present a specific deficit for disgust and contempt decoding

• Only a third of patients with SAUD present a significant emotion decoding deficit

• The emotional deficit is specific for social emotions and heterogeneous in SAUD