Introduction
Endophenotypes are trait-related illness biomarkers observable in individuals at familial risk that have attracted great research attention as they may aid prevention, early detection and development of targeted treatments (Gottesman and Gould, Reference Gottesman and Gould2003). Aberrant affective cognition is a putative endophenotype for affective disorders involving attention to, encoding and memory of stimuli with affective salience, as well as the regulation of the emotional response to such stimuli (Elliott et al., Reference Elliott, Zahn, Deakin and Anderson2011). In particular, current neurocognitive models of affective disorders put negative bias in affective cognition at the core of illness development (Phillips et al., Reference Phillips, Ladouceur and Drevets2008; Roiser, Reference Roiser2013; Malhi et al., Reference Malhi, Byrow, Fritz, Das, Baune, Porter and Outhred2015).
Considerable research on euthymic patients with unipolar disorder (UD) and bipolar disorder (BD) point to aberrant affective cognition as a trait marker (Leppänen, Reference Leppänen2006; de Almeida Rocca et al., Reference de Almeida Rocca, van den Heuvel, Caetano and Lafer2009; Mercer and Becerra, Reference Mercer and Becerra2013; Miskowiak and Carvalho, Reference Miskowiak and Carvalho2014). Replicated findings are negative emotional bias and general deficits in the recognition of emotional faces (Leppänen, Reference Leppänen2006; de Almeida Rocca et al., Reference de Almeida Rocca, van den Heuvel, Caetano and Lafer2009; Mercer and Becerra, Reference Mercer and Becerra2013; Miskowiak and Carvalho, Reference Miskowiak and Carvalho2014) although bias in attention and memory is also observed (Leppänen, Reference Leppänen2006; de Almeida Rocca et al., Reference de Almeida Rocca, van den Heuvel, Caetano and Lafer2009; Mercer and Becerra, Reference Mercer and Becerra2013; Miskowiak and Carvalho, Reference Miskowiak and Carvalho2014). Two recent studies have demonstrated reduced emotion regulation ability in remitted BD patients (Rive, Reference Rive2015; Kjærstad et al., Reference Kjærstad, Vinberg, Goldin, Køster, Støttrup, Knorr, Kessing and Miskowiak2016).
Studies of adult individuals at familial risk provide emerging evidence of aberrant affective cognition as risk markers of affective disorder (Miskowiak and Carvalho, Reference Miskowiak and Carvalho2014; Miskowiak et al., Reference Miskowiak, Kjærstad, Meluken, Petersen, Maciel, Köhler, Vinberg, Kessing and Carvalho2017). However, abnormalities seem to especially characterize individuals at risk of UD (Le Masurier et al., Reference Le Masurier, Cowen and Harmer2007; Miskowiak et al., Reference Miskowiak, Glerup, Vestbo, Harmer, Reinecke, Macoveanu, Siebner, Kessing and Vinberg2015) rather than individuals at risk of BD (de Brito Ferreira Fernandes et al., Reference de Brito Ferreira Fernandes, Gigante, Berutti, Amaral, de Almeida, de Almeida Rocca, Lafer and Nery2016; McCormack et al., Reference McCormack, Green, Rowland, Roberts, Frankland, Hadzi-Pavlovic, Joslyn, Lau, Wright, Levy, Lenroot and Mitchell2016). Maladaptive coping strategies (e.g. emotion-oriented) have been observed in high-risk individuals (Vinberg et al., Reference Vinberg, Froekjaer and Kessing2010, Green et al., Reference Green, Lino, Hwang, Sparks, James and Mitchell2011). Nevertheless, findings are inconsistent and there is a paucity of studies of genetically well-defined adult risk samples. Additional methodological issues include differences in paradigms and analyses, a shortage of studies investigating performance across domains and – especially for the UD literature – studies where high-risk individuals are compared directly with affected probands.
Assessment of monozygotic (MZ) twins represents a strong methodology for endophenotype research because MZ twins share 100% of their segregating genes and early environment (Boomsma et al., Reference Boomsma, Busjahn and Peltonen2002). In discordant twin pairs, the unaffected twin is at ultra-high risk because of being genetically identical to the affected twin. This is reflected by the high concordance rates of 23–67% in UD and 44–62% in BD MZ twin pairs (McGuffin et al., Reference McGuffin, Katz, Watkins and Rutherford1996; Sullivan et al., Reference Sullivan, Neale and Kendler2000). Only one previous MZ twin study has investigated the association of aberrant affective cognition with familial risk defined as having a co-twin with a history of depression. This study revealed negative bias in attention to and deficits in the recognition of facial expressions in MZ high-risk twins (Miskowiak et al., Reference Miskowiak, Glerup, Vestbo, Harmer, Reinecke, Macoveanu, Siebner, Kessing and Vinberg2015).
The aim of the present study was to investigate if aberrant affective cognition constitutes an endophenotype for affective disorder by comparing several key domains of affective cognition between (1) affected MZ twins in partial remission; (2) unaffected high-risk MZ twins with a co-twin history of affective disorder; (3) and low-risk MZ twins with no personal or first-degree history of affective disorder. We hypothesized that both affected twins and – to a lesser degree – high-risk twins would have general deficits in the recognition of emotional faces, negative bias in face processing (i.e. reduced attention to and/or recognition of happy v. fearful faces) and/or in self-referential memory, reduced ability to down-regulate emotions in social scenarios and maladaptive habitual coping compared with low-risk twins.
Methods
Participants and recruitment
A nationwide record linkage of the Danish Twin Registry (Skytthe et al., Reference Skytthe, Christiansen, Kyvik, Bødker, Hvidberg, Petersen, Nielsen, Bingley, Hjelmborg, Tan, Holm, Vaupel, McGue and Christensen2013), the Danish Psychiatric Central Research Register (Mors et al., Reference Mors, Perto and Mortensen2011) and the Danish Civil Registration System (Pedersen, Reference Pedersen2011) identified eligible twins by the following criteria: (1) monozygozity; (2) concordance or discordance of UD or BD from January 1995 to June 2014; (3) age 18–50 years (at the register linkage date 1 June 2014). Exclusion criteria were birth weight <1.3 kg, current severe somatic illness, history of brain injury, current substance abuse, current mood episode (i.e. Hamilton Depression Rating Scale (HDRS-17) or Young Mania Rating Scale (YMRS) > 14), pregnancy and dizygosity. Unaffected twin pairs were excluded if having first-degree relatives with organic mental disorder, schizophrenia spectrum disorders or affective disorders. Zygosity was estimated with the twin likeness questionnaire from the Danish Twin Registry. However, pair-wise DNA tests were conducted if zygosity was considered uncertain.
Procedure
This study is part of a nationwide cross-sectional assessor-blind study of putative epigenetic, cellular, neurocognitive and imaging endophenotypes for affective disorders in MZ twins. Participants were invited to a single day assessment from 8.30 a.m. to 3–7 p.m. After informed consent the following assessments were conducted chronologically: mood ratings using the HDRS-17 (Hamilton, Reference Hamilton1967) and YMRS (Young et al., Reference Young, Biggs, Ziegler and Meyer1978), biological material sampling, a semi-structured diagnostic interview using the Schedules for Clinical Assessment in Neuropsychiatry (SCAN) (Wing et al., Reference Wing, Babor, Brugha, Burke, Cooper, Giel, Jablenski, Regier and Sartorius1990), two hours of neurocognitive testing and – for a subgroup – functional magnetic resonance imaging. All assessors were blinded for the diagnostic status of participants during data collection.
Questionnaires
Participants completed the following questionnaires: The Major Depression Inventory (MDI) (Bech et al., Reference Bech, Rasmussen, Olsen, Noerholm and Abildgaard2001), The State-Trait Anxiety Inventory form Y (STAI-Y) (Spielberger, Reference Spielberger1989), the Eysenck Personality Questionnaire (EPQ) (Eysenck, Reference Eysenck1975) and The Coping Inventory of Stressful Situations (CISS) (Endler and Parker, Reference Endler and Parker1990).
Assessment of non-affective cognition
Premorbid verbal intelligence was estimated with the Danish Adult Reading Task (DART) (Nelson and O'Connell, Reference Nelson and O'Connell1978). A brief assessment of non-affective cognitive performance was conducted using the Screen of Cognitive Impairment in Psychiatry (SCIP-D) (Purdon, Reference Purdon2005) and The Trail Making Test parts A and B (TMT A/B) (Army Individual Test Battery, 1944).
Assessment of affective cognition
Emotional processing was investigated with the following computerized tests from the Emotional Test Battery [P1Vital® Oxford Emotional Test Battery (ETB), 2017]: The Faces Dot-Probe Task, the Facial Expression Recognition Task and the Emotional Categorization and Recall Tasks. Emotion reactivity and regulation were assessed with the Social Scenarios Task (Goldin et al., Reference Goldin, Manber-Ball, Werner, Heimberg and Gross2009; Kjærstad et al., Reference Kjærstad, Vinberg, Goldin, Køster, Støttrup, Knorr, Kessing and Miskowiak2016).
In the Faces Dot-Probe Task, pairs of happy-neutral, fearful-neutral or neutral-neutral faces were presented horizontally unmasked (100 ms) or masked (17 ms). One of the faces was immediately replaced by two dots presented either vertically (:) or horizontally (· ·). Participants were instructed to indicate the orientation of the dots as quickly and accurately as possible by pressing labelled keys on the keyboard. The task consisted of 32 trials of six conditions: masked or unmasked happy-neutral pairs, fearful-neutral pairs and neutral-neutral pairs. Eight blocks of unmasked and eight blocks of masked trials were presented in an alternating order, where each block consisted of 12 trials including all three types of face pairs (Murphy et al., Reference Murphy, Downham, Cowen and Harmer2008).
For the Facial Expression Recognition Task, pictures of faces taken from Ekman and Friesen (Reference Ekman and Friesen1976) expressing one of six basic emotions: anger, disgust, fear, happy, sad and surprised were displayed morphed at 10% intensity levels between a neutral face (0%) and a full emotional face (100%). Participants were instructed to determine the emotional expression as quickly and accurately as possible by pressing the corresponding key on the keyboard. Participants viewed 250 faces presented in randomized order for 500 ms, immediately replaced by a black screen. The faces consisted of four examples of every emotion at each intensity level plus a neutral face for every emotion (Harmer et al., Reference Harmer, Shelley, Cowen and Goodwin2004).
The Emotional Categorization and Recall Tasks involved the presentation of 45 positive and 45 negative personality characteristics taken from Anderson's list of personality-trait words (Anderson, Reference Anderson1968). Words were matched on length and ratings of frequency and meaningfulness, translated to Danish and displayed for 500 ms in a randomized order, separated by a fixation cross displayed for 500 ms. Participants were instructed to categorize personality trait words as referring to likeable or dislikeable attributes as quickly and accurately as possible. This categorization was performed in a self-referential manner where participants imagined overhearing someone talking about them, using these words about them. A recall task was given 15 min after completion, where participants were asked to state words remembered from the categorization task within 5 min.
Finally, the Social Scenarios Task involved the presentation of highly positive and negative social scenarios by short written paragraphs followed by associated self-beliefs statements (e.g. you are outstanding, you don't fit in). Each scenario consisted of 11 sentences describing the situation (3s each), 10 self-beliefs (3s each) and 10 corresponding emotion ratings. Participants were instructed to either react naturally or dampen their emotional responses by cognitive reappraisal and to judge their emotional state on a scale from 1 to 100 representing the degree of discomfort/sadness or pleasure/happiness. The first scenario was neutral followed by two scenarios of the same valence with alternate react/dampen conditions. After assessing sexual orientation, participants were given one of two versions involving attraction to/rejection by a man or a woman.
Equipment
The Facial Expression Recognition and Emotional Categorization Tasks were administered on a Dell PP18l laptop computer using Superlab Pro version 1.05. The Faces Dot-Probe and Social Scenarios Tasks were administered on a Lenovo T450s laptop computer using E-Prime Version 2.0.
Data analysis
In the primary analysis, the three groups were compared on the following variables of interest: (1) attentional vigilance to fear and happiness, (2) discrimination of facial expressions in general and of happiness and fear specifically, (3) emotional reactivity and down-regulation to positive and negative scenarios; (4) categorization and recall of positive and negative self-referential words, and (5) self-reported habitual coping strategies. In the secondary explorative analyses, we compared (1) complete twin pairs grouped according to the pairwise history of affective disorders as concordant, discordant and unaffected twin pairs and (2) affected and unaffected twins from discordant twin pairs. The two-high-threshold model was applied to obtain a measure of discrimination accuracy of facial expressions (d′) by the formula: [(number of hits + 0.5)/(number of targets + 1)] − [(number of false alarms + 0.5)/(number of distractors + 1)] (Corwin, Reference Corwin1994) and response bias by the formula: [(number of false alarms + 0.5)/(number of distractors + 1)]/(1 – discrimination accuracy)] (Chronaki et al., Reference Chronaki, Garner, Hadwin, Thompson, Chin and Sonuga-Barke2015). Vigilance scores were calculated as the response time latency identifying probes after neutral faces v. emotional faces.
Data analyses were conducted in SAS 9.4 (SAS Institute Inc.). Continuous variables were examined with mixed model analysis of variance with random effects for twin pairs to account for dependence within these. In an analysis of the ten intensity levels of happy and fearful faces, we used logistic regression techniques with nested random effects for twin and subject to jointly consider the different levels. Categorical variables were also compared using logistic regression, where within twin-pair dependence was adjusted for using GEE estimates of the standard errors. Group comparisons were considered as fixed factors. Affected and unaffected co-twins of discordant twin pairs were compared using parametric paired samples t tests, with the exception of non-parametric tests for intensity levels of facial expressions. Experimental paradigms involving repeated measurements (i.e. emotional expression of faces, positive and negative words and react and dampen conditions) were modelled using nested random effects for twin pairs and subjects. All primary analyses were performed unadjusted and adjusted for HDRS-17 scores (i.e. depressive symptoms). Two sensitivity analyses were conducted. Firstly, we tested the influence of BD by excluding participants with or at risk of BD from analyses. Secondly, we tested the influence of antidepressants by excluding participants with current antidepressant usage from analyses. The significance level was set to ɑ = 0.05 and hypotheses tests were two-sided. Analyses were not corrected for multiple comparisons given the explorative nature of the study.
Results
Participants
Of 476 eligible MZ twins identified by the register linkage, we invited 408 twins to take part in the study between December 2014 and January 2017. Of these, 44 were excluded: (1) two were dead; (2) four were dizygotic; (3) ten had a birth weight <1300 g; (4) three had a history of severe head trauma; (5) three had severe somatic illness (chronic myelogenous leukaemia, severe anorexia nervosa, multiple sclerosis); (6) three had a current alcohol abuse; (7) five had a current mood episode; (8) three were pregnant; (9) two had Asperger syndrome. Additionally, nine low-risk twins were excluded due to having a first-degree relative with a psychiatric disorder. Of the remaining 364 eligible participants, 209 (affected: n = 119, high-risk: n = 50, low-risk: n = 40) agreed to participate (inclusion rate 57%). We excluded one high-risk twin and four affected twins from the analysis after interviews because of being diagnosed with, or being at risk of, schizophrenia spectrum disorders. Due to a technical issue with the test computer, affective cognition data were lost for 21 participants (affected: n = 12, high-risk: n = 5, low-risk: n = 4). The final sample of 183 MZ twins (affected: n = 103, high-risk: n = 44, low-risk: n = 36) consisted of 79 complete twin pairs and 25 twin individuals whose co-twin was not included in the analyses. Of the 79 complete twin pairs, 22 were concordant, 40 were discordant and 17 were unaffected twin pairs. Among the 44 high-risk twins, 34 (77%) had a co-twin diagnosed with UD and 10 (23%) with BD.
Affected, high-risk and low-risk groups were well-balanced with respect to age, sex, years of education and premorbid IQ (ps ⩾ 0.16, Table 1). However, affected twins exhibited more depression and anxiety symptoms and scored higher on neuroticism than high- and low-risk twins (Ps < 0.001, Table 1). Importantly, there were no group differences in non-affective cognitive performance (ps ⩾ 0.15, Table 2).
Demographic and clinical characteristics of affected, high-risk and low-risk MZ twins are presented as estimated group means with confidence intervals by the proc mixed procedure in SAS accounting for within twin-pair dependence. Significant post-hoc pairwise comparisons are indicated by directions of effects
MZ, Monozygotic; CI, Confidence intervals; NA, Not Applicable; SSRI, Selective Serotonin Reuptake Inhibitor; SNRI, Seretonin-Norephinephrine Reuptake Inhibitor; TCA, Tricyclics Antidepressant; HDRS-17, Hamilton Depression Rating Scale; AF, Affected Twins; HR, High-risk twins; LR, Low-risk twins; YMRS, Young Mania Rating Scale; MDI, Major Depression Inventory; STAI, State and Trait Anxiety Inventory; EPQ, Eysenck Personality Questionnaire.
All three-way group comparisons of affected (n = 103), high-risk (n = 44) and low-risk twins (n = 36) are reported with p values whereas significant post-hoc pairwise comparisons are indicated by directions of effects.
a Eleven participants with dyslexia were excluded.
b Attentional deficit and hyperactivity disorder, eating disorder and adjustment disorder.
c Lamotrigine, Valproate, Lithium.
d Data were missing for 12 participants from EPQ.
Non-affective cognitive performance in affected, high-risk and low-risk MZ twins are presented as estimated group means with confidence intervals by the proc mixed procedure in SAS accounting for within twin-pair dependence
MZ, Monozygotic; CI, Confidence intervals; TMT-A, Trail Making Task-A; TMT-B, Trail Making Task-B; SCIP-D, Screen for Cognitive Impairment in Psychiatry, Danish; VLT-I, Verbal Learning Test Immediate; WMT, Working Memory Test; VFT, Verbal Fluency Test; VLT-D, Verbal Learning Test Delayed; PST, Processing Speed Test.
Group comparisons of affected (n = 103), high-risk (n = 44) and low-risk twins (n = 36) are reported with p values.
a Data were missing for one participant from the TMT-B, PST (SCIP) and SCIP-D total.
Attention to and recognition of emotional faces
Comparing vigilance scores with zero across all participants revealed subliminal attentional avoidance of happy faces (t = −4.16, df = 180, P < 0.001) and a tendency of supraliminal attentional avoidance of happy faces (t = −1.92, df = 180, p = 0.056), but no attentional bias or avoidance of fearful faces (ps ⩾ 0.70). We found group differences in subliminal vigilance to happy faces (F = 4.7, df = 2, 177, p = 0.01) and supraliminal vigilance to fearful faces (F = 3.6, df = 2, 177, p = 0.03), in the absence of differences in subliminal vigilance to fearful faces or supraliminal vigilance to happy faces (ps ⩾ 0.60, Table 3 and Fig. 1). These group differences were driven by subliminal attentional avoidance of happy faces (p = 0.003) and supraliminal attentional avoidance of fearful faces (p = 0.009) in high-risk v. affected twins. Differences prevailed after adjusting for depressive symptoms (ps ⩽ 0.02).
High-risk twins showed attentional avoidance of emotional faces. Vigilance scores are calculated as response time difference in ms. identifying probes shown after emotional faces over neutral faces (i.e. positive values represent attention bias towards emotional faces, negative values represent attentional avoidance). Unmasked condition: Face stimuli shown for 100 ms, masked condition: Face stimuli shown for 17 ms. Data are presented as estimated means of vigilance scores from the mixed procedure in SAS. Error bars represent standard error of the mean. *p < 0.05.
Affective cognitive performance in affected, high-risk and low-risk MZ twins are presented as estimated group means with confidence intervals by the proc mixed procedure in SAS accounting for within twin-pair dependence. Group comparisons are reported with p-values unadjusted for depressive symptoms. Subheadings in the test result column indicate within subject factors in repeated measurements analysis
MZ, Monozygotic; CI, Confidence intervals; FDOT, Faces Dot-Probe Task; FERT, Facial Expression Recognition Task; SST, Social Scenarios Task; CISS, Coping Strategies in Stressful Situations; ECAT, Emotional Categorization Task; EREC, Emotional Recall task.
Group comparisons of affected (n = 103), high-risk (n = 44) and low-risk twins (n = 36) are reported with unadjusted p values and degrees of freedom. Subheadings in the test result column indicate within-subjects factors in repeated measurements analysis.
a Data were missing for two affected twins from FDOT, one affected twin from FERT, four affected twins, one high-risk and three low-risk twins from ECAT, three affected, one high-risk and three low-risk twins from EREC, one affected and one low-risk twin from SST, 13 affected and seven high-risk twins from CISS.
b Scores of discrimination accuracy tend to 1, 0, and −1 reflecting better than chance, close-to-chance and worse-than-chance.
c Eleven participants with dyslexia were excluded.
Participants generally displayed a positive bias in recognition of facial expressions as reflected by increased discrimination accuracy (F = 205.7, df = 1, 181, p ⩽ 0.001) and shorter response times (F = 339.4, df = 1, 180, p ⩽ 0.001) to happy than fearful faces, although participants were more likely to categorize faces as fearful than happy (F = 67.0, df = 181, p ⩽ 0.001). There were no group differences in discrimination accuracy, speed during recognition or response bias of fear and happy or other emotional expressions (ps ⩾ 0.19, Table 3). Analysis of the ten intensity levels of fearful and happy faces revealed no group by task interactions (ps ⩾ 0.99).
Self-referential categorization and memory
Participants generally displayed a positive bias in categorization of self-referential personality trait words as reflected by shorter response times to positive than negative words (F = 53.8, df = 1, 163, p < 0.001), but there were no differences in categorization accuracy or recall of positive v. negative words (ps ⩾ 0.17). We found no group differences in speed or accuracy during categorization or recall of positive and negative self-referential personality trait words, with or without adjustment for depressive symptoms (ps ⩾ 0.08, Table 3).
Emotional reactivity and regulation to social scenarios
In the ‘react’ conditions, negative and positive social scenarios elicited more unpleasant and pleasant emotional responses, respectively, compared with the neutral scenario (negative: F = 1273.7, df = 1, 169, p < 0.001; positive: F = 1533.6, df = 1, 168, p < 0.001). In the ‘dampen’ conditions, participants were generally able to dampen their negative and positive emotional responses (negative: F = 281.4, df = 1, 169, p < 0.001; positive: F = 207.4, df = 1, 167, p < 0.001). We found no significant group differences in emotional reactivity to negative (p = 0.08) or positive (p = 0.33) social scenarios. However, explorative pairwise comparisons suggested that affected twins experienced more unpleasant emotions than low-risk twins when instructed to ‘react’ in negative scenarios (t = 1.99, df = 169, p = 0.049). We found no group differences in the ability to down-regulate emotional reactions to negative or positive scenarios, with or without adjustment for depressive symptoms (ps < 0.26, Table 3).
Habitual coping strategies
We found group differences in the use of emotion-oriented (F = 7.8, df = 2, 159, p < 0.001) and task-oriented (F = 3.2 df = 2, 159, p = 0.04) coping (Table 3). These differences were driven by increased use of task-oriented coping in high-risk twins relative to affected twins (p = 0.01) and by more use of emotion-oriented coping in affected twins relative to both high-risk (p = 0.004) and low-risk (p = 0.001) twins. The greater use of emotion-oriented coping strategies in affected twins prevailed after adjustment for depressive symptoms (p = 0.03).
Sensitivity analysis
Omitting the BD subgroup rendered non-significant group differences in supraliminal vigilance to fearful faces (p = 0.06) and task-oriented coping (p = 0.14). Omitting participants on antidepressants rendered non-significant group differences in task-oriented coping (p = 0.31). All other outcomes remained statistically significant.
Secondary analysis
Comparable to the primary analysis, unaffected co-twins in discordant twin pairs showed supraliminal attentional avoidance of fearful faces (p = 0.002), and more use of task-oriented coping (p = 0.02) compared with the affected twins. Concordant twin pairs and affected twins from discordant twin pairs reported more use of emotion-oriented coping than low-risk twin pairs (p = 0.006) and unaffected co-twins (p = 0.001), respectively. Specifically for complete twin pair analyses, we found a group difference in speed when categorizing self-referential personality trait words (F = 3.3, df = 2, 143, p = 0.04). This difference was driven by slower responses in concordant twin pairs than unaffected twin pairs (p = 0.01). No other secondary analyses yielded group or group by task interaction effects.
Discussion
We investigated whether aberrant affective cognition is an endophenotype for the affective disorder by comparing MZ affected (n = 103), high-risk (n = 44) and low-risk (n = 36) twins. In contrast with our hypotheses, there were no abnormalities in affective cognition across affected and high-risk relative to low-risk twins. High-risk twins displayed attentional avoidance of emotional faces and more habitual use of task-oriented coping relative to affected twins but did not differ on these measures from low-risk twins. Affected twins showed more use of emotion-oriented coping than both high-risk and low-risk twins.
The absence of negative bias and deficits in facial expression recognition contrasts with increased attentional vigilance to fearful faces and face recognition difficulties in MZ twins at risk of UD (Miskowiak et al., Reference Miskowiak, Glerup, Vestbo, Harmer, Reinecke, Macoveanu, Siebner, Kessing and Vinberg2015). In particular, discrepancies may be due to our high-risk sample being younger (mean age 37 v. 47). However, the risk of onset may not decline with age (Vinberg et al., Reference Vinberg, Miskowiak and Kessing2013), suggesting that these samples are at comparable risks. Another difference was the exclusion of twins with any axis-1 diagnosis in the previous study. However, allowing for minor psychiatric disorders would be expected to be associated with more (rather than less) emotion processing abnormalities. Moreover, the previous study included twins at risk of UD only, compared with twins at risk of UD constituting 77% of all at-risk twins in the present study. Since our sensitivity analyses of twins at risk of UD rendered attentional avoidance of subliminally processed happy faces as significant, the effect of diagnosis does not explain the differences. Finally, the two studies differ in the statistical procedure as the present study accounted for covariance within twin pairs. This was not necessary in the previous study because of including only one twin per twin pair. Other findings point to a specific change in fear recognition in individuals at risk for UD with observations of both faster (Le Masurier et al., Reference Le Masurier, Cowen and Harmer2007) and slower (Watters et al., Reference Watters, Gotlib, Harris, Boyce and Williams2013) reaction time to fearful faces. One could argue that both findings reflect a negative bias with faster reaction time reflecting increased recognition, whereas slower reaction time could reflect more engagement in fearful expressions. However, in line with the present study, the slower reaction time to fearful faces in the large study by Watters and colleagues (high-risk: n = 101) could also be interpreted as absence of bias. Moreover, our results corroborate with three large studies that detected no deficits in face recognition in adult samples at risk of UD (Mannie et al., Reference Mannie, Bristow, Harmer and Cowen2007) and BD (de Brito Ferreira Fernandes et al., Reference de Brito Ferreira Fernandes, Gigante, Berutti, Amaral, de Almeida, de Almeida Rocca, Lafer and Nery2016; McCormack et al., Reference McCormack, Green, Rowland, Roberts, Frankland, Hadzi-Pavlovic, Joslyn, Lau, Wright, Levy, Lenroot and Mitchell2016). Overall, the finding of a negative bias in emotional face processing is exclusively reported in small studies (high risk: n = 13–25), while the absence of deficits and of negative bias are reported in larger studies (high-risk: n = 26–101). This raises concerns about spurious findings in the small studies and a publication bias (Porter et al., Reference Porter, Boden, Miskowiak and Malhi2017). In light of the present negative finding and the paucity of studies of adult first-degree relatives due to a predominant use of child and adolescent samples (e.g. Hanford et al., Reference Hanford, Sassi and Hall2016; Sharma et al., Reference Sharma, Barron, Le Couteur, Close, Rushton, Grunze, Kelly, Nicol Ferrier and Le Couteur2016), there is no strong support for abnormalities in the processing of emotional faces as an endophenotype.
The absence of negative bias in the recall of self-referential words in high-risk twins is in line with previous findings in individuals at risk of UD (Le Masurier et al., Reference Le Masurier, Cowen and Harmer2007; Mannie et al., Reference Mannie, Bristow, Harmer and Cowen2007) and only subtle negative memory bias in twins at high-risk of UD (Miskowiak et al., Reference Miskowiak, Larsen, Harmer, Siebner, Kessing, Macoveanu and Vinberg2018). However, while we found no deficits in the categorization of emotional words, previous UD high-risk studies reported slower response times in general (Mannie et al., Reference Mannie, Bristow, Harmer and Cowen2007), poorer accuracy (Miskowiak et al., Reference Miskowiak, Larsen, Harmer, Siebner, Kessing, Macoveanu and Vinberg2018) and faster categorization of negative words (Le Masurier et al., Reference Le Masurier, Cowen and Harmer2007). Discrepancies may be due to small sample size and investigation of individuals at-risk of UD only. From these few studies conducted, there is no evidence suggesting that negative bias in the emotional recall is an endophenotype. Rather, negative memory bias seems to represent a scarring effect that to some degree is present in remitted patients (Mercer and Becerra, Reference Mercer and Becerra2013; Miskowiak and Carvalho, Reference Miskowiak and Carvalho2014).
Our demonstration of no emotion regulation difficulties contrasts with one report of less success in down-regulating positive emotions through reappraisal in first-degree relatives of BD probands (Kanske et al., Reference Kanske, Schönfelder, Forneck and Wessa2015). Discrepancies may be due to diagnosis specific effects and a small sample (high-risk: n = 17). However, this aspect of cognition is critically understudied as no other study has been conducted of individuals at familial high-risk. There is a need for future research on emotion regulation to clarify whether this constitutes an endophenotype for affective disorders.
The more frequent use of task-oriented coping in high-risk twins compared with affected twins is in keeping with one previous study demonstrating greater use of adaptive coping strategies (i.e. putting in perspective) in first-degree relatives compared with their BD probands (Green et al., Reference Green, Lino, Hwang, Sparks, James and Mitchell2011). However, familial risk of affective disorder has also been associated with increased use of maladaptive coping strategies (e.g. self-blame) compared with affected BD probands (Green et al., Reference Green, Lino, Hwang, Sparks, James and Mitchell2011) and individuals at low risk for the affective disorder (Vinberg et al., Reference Vinberg, Froekjaer and Kessing2010; Green et al., Reference Green, Lino, Hwang, Sparks, James and Mitchell2011). Discrepancies may relate to use of different assessment tools: the Coping Inventory in Stressful Situations (Vinberg et al., Reference Vinberg, Froekjaer and Kessing2010) and the Cognitive Emotion Regulation Questionnaire (Green et al., Reference Green, Lino, Hwang, Sparks, James and Mitchell2011). Although this preliminary evidence point to both adaptive and maladaptive coping being present in high-risk samples, there is a paucity of studies of this clinically relevant topic.
Attentional avoidance of emotional faces and greater use of task-oriented coping in high-risk twins may reflect compensatory strategies. Specifically, the genetic and environmental vulnerability of having a MZ co-twin with the affective disorder may promote the development of cognitive strategies that protect against illness onset. Markers specific to high-risk individuals that are not observed in affected or low-risk groups may be interpreted as markers of resilience (Frangou, Reference Frangou2011). Notably, recent functional imaging studies showed that resilience to BD was associated with the increased integration of the default mode network (Doucet et al., Reference Doucet, Bassett, Yao, Glahn and Frangou2017) and hyperconnectivity of the ventral visual stream during face processing (Dima et al., Reference Dima, Roberts and Frangou2016). Moreover, direct comparison of affected and unaffected twins in discordant pairs revealed similar findings compared with the primary analyses, pointing to differences being due to environmental factors as MZ twin pairs have almost identical genes. Replication of the findings of the increased use of emotion-oriented coping (e.g. get angry, become tense) in affected individuals across analyses makes this a robust marker associated with the history of mood episodes, in line with previous findings (Christensen and Kessing, Reference Christensen and Kessing2005).
It is a limitation that we did not include dizygotic twins since this enables investigation of the interaction between environmental and genetic influence on variance, preferable when investigating endophenotypes. Our population-based sampling strategy using public registers to identify eligible participants reduces selection bias but does not account for cases in which treatment is not sought. Foremost, we could have increased power in statistical inference by the inclusion of more high- and low-risk twins. However, as especially BD has a high heritability, the number of available discordant MZ twin pairs was limited (n = 63). Our sample is relatively small compared with other recent studies of individuals at familial risk (high-risk: n ≈ 100) (Watters et al., Reference Watters, Gotlib, Harris, Boyce and Williams2013; McCormack et al., Reference McCormack, Green, Rowland, Roberts, Frankland, Hadzi-Pavlovic, Joslyn, Lau, Wright, Levy, Lenroot and Mitchell2016), but large compared with studies using MZ twins as high-risk participants (Gourovitch et al., Reference Gourovitch, Torrey, Gold, Randolph, Weinberger and Goldberg1999; Hsu et al., Reference Hsu, Young-Wolff, Kendler, Halberstadt and Prescott2014; Miskowiak et al., Reference Miskowiak, Glerup, Vestbo, Harmer, Reinecke, Macoveanu, Siebner, Kessing and Vinberg2015). Participants taking psychotropic medication were not excluded, although antidepressants have shown to normalize disruptions in affective cognition (Harmer et al., Reference Harmer, Cowen and Goodwin2011). To account for this, we conducted sensitivity analyses excluding participants taking antidepressants. Finally, collapsing the two discrete diagnoses BD and UD hinders classification of diagnosis-specific endophenotypes needed to increase diagnostic precision. Nonetheless, we examined the risk of affective disorders as deficits in emotion processing have been found across distinct categorical disorders (e.g. BD, UD and anxiety disorders) (Kret and Ploeger, Reference Kret and Ploeger2015) and because UD and BD share genetic underpinnings (Cross-Disorder Group of the Psychiatric Genomics Consortium, 2013).
In conclusion, our findings provide no support for aberrant affective cognition as an endophenotype for affective disorders. Rather, high-risk twins’ attentional avoidance of emotional faces and greater use of task-oriented coping might reflect compensatory strategies that help them withstand disease onset despite their biological vulnerability. Implications for future research are increased focus on longitudinal comparisons of genetically well-defined risk samples with both affected and controls, using tools probing cognitive aspects of both resilience and risk.
Acknowledgements
We thank Mette Marie Støttrup for her help with data collection and Hanne Lie Kjærstad for her help with data collection and proofreading of the manuscript.
Financial support
IM's Ph.D. salary was funded by The Lundbeck Foundation (grant number R108-A10015) and the Hørslev Foundation is acknowledged for their financial contribution to running costs. The Lundbeck Foundation and Weimann Foundation are acknowledged for their contributions to KWM's salary to increase her time for research between 2012 and 2021.
Conflict of interest
CJH has received consultancy fees from P1vital ltd, Lundbeck, Servier, and Eli-Lilly and is a company director of Oxford Psychologists ltd. CJH has also received grant income from GSK, UCB, Janssen Inc, Lundbeck, Servier and Astra Zeneca. LVK reports having been a consultant for Lundbeck, AstraZeneca and Sunovion within the last 3 years. MV has received consultant fee from Lundbeck and AstraZeneca within the last 3 years. KWM reports having received consultancy fees from Lundbeck and Allergan in the past 3 years.
Ethical standards
The authors assert that all procedures contributing to this work comply with the ethical standards of local ethics committee (H-3-2014-003) and data protection agency (2014-331-0751) and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008.
