Facial emotion recognition performance influences executive control impairment in Anorexia Nervosa: an exploratory study

Executive control encompasses a set of mechanisms that allow individuals to plan complex goal-directed behaviours, flexibly adjusting their performance and adapting to changing environmental and internal states (Alvarez & Emory, 2006).

Various deficits in executive control have been identified in the Eating Disorders (EDs) literature, in particular, as regards difficulties in set-shifting (e.g., Kittel, Brauhardt, & Hilbert, 2015). Deficits in executive control – including inhibitory control, cognitive flexibility, decision-making, and working memory – may be risk or maintenance factors for EDs (Manasse et al., 2015). Nonetheless, evidence is somehow mixed in several aspects (Wu et al., 2014), considering various socio-demographic and clinical confounders. Significant methodological weaknesses (e.g., use of a single executive control measure, omission of relevant covariates) in the current literature represent one reason for lack of consensus. For example, considering Anorexia Nervosa (AN), children and adolescents populations show little, if any, impairment in set-shifting ability (Lang et al., 2014). This might be due to the cognitive disturbances derived from alterations in the brain structure detectable in the adulthood among people with AN, and possibly exacerbated by starvation, which contribute to the maintenance of AN after recovery (Fonville et al., 2014). Similarly, evidence on impairments underlying poor executive control in Bulimia Nervosa (BN) is not conclusive (Van den Eynde et al., 2011). In addition, individual differences in clinical characteristics in ED samples provide further interpretative challenges, as these could distinctively affect each individual’s cognitive performance. For example, age of onset might influence inefficiencies in set-shifting that are apparent in the adult, since they do not appear to be as pronounced in children. This may suggest that set shifting difficulties in adult AN are the result of starvation or indicative of longer duration of illness (Lang et al., 2014).

In particular, Facial Emotion Recognition (FER), that is, the ability to accurately identify and name emotions of other people, has been shown compromised in women with AN (Caglar-Nazali et al., 2014). The facets of social-cognition that appear to be compromised in this patient group include difficulties identifying basic and complex emotions, with large to moderate effects found for poor FER (Cardi et al., 2015), particularly in women with AN. Although evidence remains not entirely convincing and definitive, difficulties in interpreting emotions may contribute to, and exacerbate, AN clinical symptoms (Treasure, Corfield, & Cardi, 2012). Also, for BN, the limited literature has found some difficulties in recognizing specific emotions (e.g., Kuhnpast, Gramann, & Pollatos, 2012; Aloi et al., 2017). In particular, authors using pooled subjects with BN and AN have reported contradictory findings, showing global difficulties recognising emotions in adolescents (Zonnevijlle-Bender, van Goozen, Cohen-Kettenis, van Elburg, & van Engeland, 2002), but not in adults (Cardi et al., 2015). However, considering BN only, some investigations have found slight difficulties recognising surprise (Kessler, Schwarze, Filipic, Traue, & von Wietersheim, 2006; Legenbauer, Vocks, & Ruddel, 2008), and anger, often misinterpreted as fear (Kuhnpast, Gramann, & Pollatos, 2012), but no problems recognising other emotions.

However, an association between impaired FER and executive control has been found in several other serious mental disorders, including alcohol dependence (Trick et al., 2014), schizophrenia (Bortolon, Capdevielle, & Raffard, 2015), and bipolar disorders (David et al., 2014). Prefrontal dysfunction may lead to impairments in executive control and FER (Adolphs, 2002). Emotion face perception is a complex process that cannot be related to a single neural event taking place in a single brain region, but rather implicates an interactive network with distributed activity in time and space. Traditional models have often assumed that facial expression and executive control are processed along separate pathways. However, evidence from fMRI suggests that emotional processing can strongly affect brain systems responsible for both face recognition and memory, with fear-related modulations of face processing driven by amygdala signals implicating not only fusiform cortex, but also earlier visual areas in occipital cortex involved in cognitive responses (Vuilleumier & Pourtois, 2007). Deficits in FER and executive control could thus be associated also in EDs. Consistently, determining whether the impairment in EDs is cognitive-affective may have key therapeutic implications, as these require different intervention strategies (Legenbauer, Vocks, & Rüddel, 2008). It might well be that, also in people with EDs, set-shifting and FER are jointly compromised, at least as regards the impairment of certain emotions. Nonetheless, to our knowledge, no empirical research exists, explicitly addressing this issue. We thus tested the hypothesis that difficulties in set-shifting experienced by people with AN or BN may be influenced by impaired FER. We used a real world sample of women with EDs, rigorously assessing diagnostic and clinical features, but also executive control and emotional recognition abilities. It was hypothesized that, after adjusting for important sociodemographic and clinical characteristics, impaired FER would influence executive control performance in people with AN or BN.

Checking the data distribution for IED adjusted total errors and EK-60F scores, we could detect two outliers based on interquartile range (iqr) rule (Tukey, 1977; Dawson, 2011), bringing values whose ratio exceeded 3 (‘severe outliers’). These participants were excluded from analyses, resulting in an overall sample of 58 women (AN = 19; BN = 20; HCs = 19). Consistent with our study hypothesis, we first explored the attributes of EDs, also in terms of performance in executive control and FER. Thus, Table 1 shows the univariate comparisons for sociodemographic and clinical characteristics across AN, BN and HCs subgroups. Mean age was comparable with participants usually in their late twenties, and not different in terms of living and family conditions, education and employment status. History of childhood abuse was rare, and BMI was obviously uneven between AN and BN. Women with AN performed significantly worse than HCs in terms of IED adjusted total errors. As regards EK-60F total and subscales scores, HCs scored significantly better than AN and BN only on fear subscale (subscale range 0–10). In addition, for measures describing current symptoms dimensions, that is, means of BDI-II, TAS-20, SCL-90 GSI, and EDI-3 subscales, women with AN and BN scored significantly worse than HCs.

Then, in order to explore the role of other factors, including FER performance and ED diagnostic subgroup, associated with IED adjusted total errors, we fitted several models that included significant attributes from the univariate analyses, and known to be relevant from the literature (Table 2). The first model considering the EDs diagnostic subgroups compared with HCs showed that IED set shifting performance was associated with AN (p = 0.008). However, in the second model, also taking into account age, EK-60F fear subscale, depression as measured by BDI-II, and alexithymia as measured by TAS-20, this association could not be confirmed. In addition, the performances of the two models in terms of AIC were comparable.

This study aimed to investigate the influence of FER abilities on impaired executive control in women with EDs. As a measure of executive control, we used the IED, a standardized and validated test that is not timed and so may be less affected by potential deficits in speed of processing, and that was previously successfully used among women with EDs (e.g., Galimberti et al., 2012). This is critical, since the effect sizes for different measures of set-shifting among women with EDs across the studies differ significantly from each other (Wu et al., 2014). We could initially uncover an association between IED set-shifting adjusted total errors and just AN, that was not found in the previous studies. We possibly sampled a slightly older clinical population that might have had sufficient time to develop cognitive disturbances. However, taking into account different correlates, including fear recognition ability, this association was no longer significant. Fear recognition ability is likely to have a role on impairment in executive control in women with AN as much as alexithymia (Brewer et al., 2015) and other clinical variables (Wu et al., 2014) may have. As a whole, our findings are consistent with the cognitive-interpersonal maintenance model of AN (Treasure & Schmidt, 2013), as far as executive control is shown to be compromised in women with AN. Nonetheless, they provide incremental knowledge jointly exploring, for the first time, the influence of impaired interpretation of facial emotions and weak set shifting, an important measure of executive control. In sum, further research should pay closer attention to the evaluation of several potential confounders on executive control (Abbate-Daga et al., 2015).

These limitations notwithstanding, our study provides preliminary insight about the influence of FER abilities on impaired executive control in women with AN. They may not be able to recognize complex emotional states, and this may provide clues on their executive control abilities, affecting the maintenance of severe EDs symptoms. This may have important implications that can be included in specialized treatment programmes and interventions aimed at improving socio-emotional processing in AN and updating the current programs dealing with emotion functioning (Tchanturia et al., 2015).