Emotional Face Identification in Youths With Primary Bipolar Disorder or Primary Attention-Deficit/Hyperactivity Disorder

Abstract

Objective

Bipolar disorder (BD) and attention-deficit/hyperactivity disorder (ADHD) are often comorbid or confounded; therefore, we evaluated emotional face identification to better understand brain/behavior interactions in children and adolescents with either primary BD, primary ADHD, or typically developing controls (TDC).

Method

Participants included individuals 7 to 17 years of age (overall sample mean age 12.40 ± 3.01 years), with “narrow-phenotype” pediatric BD (n = 30) or ADHD (n = 38), or typically developing controls (TDC) with no psychiatric disorders themselves or in their first-degree relatives (n = 41). In the BD group, comorbid diagnoses were allowed; however, youth in the ADHD group were excluded for comorbid mood or anxiety disorders. Patient groups were not excluded for psychotropic medication use. Emotional face identification was assessed using the computerized Diagnostic Analysis of Non-Verbal Accuracy (DANVA).

Results

Participants with BD made significantly more identification errors on child happy faces than either TDCs (p = .03) or participants with ADHD (p = .01). Furthermore, youth with BD (0.33 ± 0.55) were more likely than youth with ADHD (0.11 ± 0.31) to make errors on low-intensity child happy faces (p = .05) but not high-intensity happy faces (p = NS). Participants with BD and ADHD made significantly more total errors in child face labeling than did TDCs, although participants with BD and ADHD did not differ from one another.

Conclusion

Our data suggest that youths with BD have specific alterations in emotional face identification of happy faces, an important finding that supports theories that response to positively valenced emotional stimuli may be especially salient in BD. Clinical trial registration information—Brain Imaging and Computer Games in Children With Either Bipolar Disorder, ADHD, Anxiety or Healthy Controls (BBPP); http://clinicaltrials.gov/; NCT01570426.

Disentangling attention-deficit/hyperactivity disorder (ADHD) from bipolar disorder (BD) in children is difficult for clinicians and researchers alike, because of overlapping clinical presentations and diagnostic criteria. For example, a child who presents with hyper-activity, impulsivity, low frustration tolerance, distractibility, and who is talkative may be diagnosed with ADHD by one clinician, with BD by another, and with ADHD and BD by a third. Overlapping symptoms have been implicated in the rising diagnostic rates of both disorders.^1,2 Specifically, the percentage of children and adolescents discharged from U.S. psychiatric hospitals with a BD diagnosis surged from less than 10% in the mid-1990s to more than 20% in the mid-2000s.³ Similarly, epidemiological data indicate that the prevalence of ADHD among children has increased from 4% to 6% in 2000, to 8% to 10% currently.^4-6 An alternative approach is needed whereby greater understanding of the brain– behavior interactions underlying youths with either BD or ADHD leads to bio-behavioral markers for more precise diagnosis and treatment.

In this vein, emotional face identification provides a window into one of the most basic aspects of emotion regulation—i.e., categorizing another's emotional state. The human face is an important emotional stimulus, as people are hard-wired from birth to attend to faces.⁷ From a developmental perspective, emotional face identification involves the bottom-up evaluation of emotionally evocative stimuli, which typically develops before top-down regulation of attention and processing of visual stimuli.^8-10 Emotional face identification holds the potential to advance our understanding of how these fundamental emotional processes are altered in children with primary BD versus primary ADHD.^11-13

Prior studies have shown behavioral alterations in emotional face identification in youths with BD versus typically developing controls (TDC). Youths with BD make significantly more errors than TDCs in identifying child and adult facial emotions on the computerized Diagnostic Analysis of Nonverbal Accuracy (DANVA), and require greater intensity of emotion to correctly identify facial emotions than TDCs, as assessed using multi-morph tasks whereby the intensity of emotional face expression is varied from neutral to 100%.^14-16 Emotional face identification may represent a trait marker for pediatric BD, as children at elevated risk for BD, by having a first-degree relative with BD, have emotional face identification deficits similar to those of youths with BD themselves.¹⁷

Functional magnetic resonance imaging (fMRI) studies have demonstrated aberrant frontoamygdala-striatal neural activity among children and adults with BD when processing happy faces.^18-25 For example, adults with BD have significantly increased left amygdala and striatal activity compared to adult TDCs when processing happy faces.^19,23,24 Because happy faces are rewarding in adult TDCs,^26,27 and because individuals with BD show aberrant striatal activity when anticipating and processing rewards^28,29 and processing happy faces,^19,23,24 some speculate that individuals with BD may find happy faces less socially rewarding or even potentially aversive. However, this hypothesis remains untested.

Far fewer studies have evaluated emotional face processing in children with primary ADHD.³⁰ Youths with ADHD are less accurate in identifying facial emotional expression than TDCs even when presented with situational or contextual clues.^31-34 Participants who are ADHD hyperactive/impulsive also make more errors than TDCs in identifying emotional faces, particularly angry and sad faces.³² Compared to TDCs, children at elevated risk for ADHD by having elevated teacher ratings of ADHD also make significantly more errors on emotional face identification.³⁵ Neuroimaging studies show that youths with ADHD have amygdala hyper-activation when processing fearful faces versus TDCs or completing subjective fear ratings of neutral faces versus youths with BD, severe mood dysregulation (SMD; chronic nonepisodic irritability and ADHD-like symptoms of hyper-activity), or TDCs.^12,36 These results support pathophysiological models implicating altered noradrenergic pathways in ADHD.³⁷

Unfortunately, virtually no studies have directly compared emotional face identification among youths with primary BD versus those with primary ADHD. To the best of our knowledge, only Guyer et al. have examined emotional face identification among youths with BD versus a combined ADHD/conduct disorder (CD) sample (n = 35), showing that youths with BD made significantly more face identification errors than youths with ADHD/CD.³⁸ Conclusions about those with primary ADHD were limited in this study because the combined ADHD/CD group included youths with either ADHD (n = 18), CD (n = 7), or both ADHD and CD (n = 10).³⁸

To address this knowledge gap, we evaluated child and adult emotional face identification ability among youths with either primary BD, primary ADHD, or TDC. We hypothesized that youths with BD would make more errors on happy face identification than TDC or ADHD groups, given prior work suggesting BD involves neural alterations to happy faces and to positively valenced emotional stimuli.^13,18-25 We hypothesized that youths with ADHD would make more errors on fearful faces versus BD and TDC groups, based on research showing that ADHD involves brain/behavior alterations in fear processing.^17,39

METHOD

Participants

Children and adolescents 7 through 17 years of age were enrolled in a study approved by the Institutional Review Boards of Bradley Hospital and Brown University, after informed parental consent and child assent.

All participants were evaluated for psychopathology using the Child Schedule for Affective Disorders and Schizophrenia, Present and Lifetime version (K-SADS-PL) administered by either a board-certified child/adolescent psychiatrist (D.P.D.) or licensed clinical psychologist (K.L.K.; k > 0.85) to parents and youths separately.⁴⁰ During the K-SADS, comorbid diagnoses for the BD group were assessed by asking about symptoms and associated impairment during periods of generally euthymic mood to avoid counting manic or depressive symptoms twice.

For all groups, inclusion criteria were as follows: age between 7 and 17 years; English fluency; and a consenting parent/guardian. Exclusion criteria were as follows: Wechsler Abbreviated Scale of Intelligence Full-Scale IQ (WASI FSIQ) ≤70; substance/alcohol abuse or dependence within the last 2 months; autism spectrum disorders or primary psychosis; and medical/neurological conditions that mimic either BD or ADHD.⁴¹

BD (n = 30) inclusion criteria consisted of meeting DSM-IV-TR criteria for BD, including history of at least one episode of hypomania (≧4 days) or mania (≧7 days) wherein the child exhibited abnormally elevated or expansive mood and three or more DSM-IV criterion “B” mania symptoms. All participants with BD had type I BD, although type II was not exclusionary. Children presenting with irritable mood only (i.e., without elated or expansive mood) were not included. Thus, the BD group met Leibenluft's “narrow phenotype” pediatric BD criteria.⁴² Youths with primary BD were not excluded for comorbid conditions (e.g., ODD, ADHD).

ADHD (n = 38) inclusion criteria were: meeting DSM-IV-TR criteria for any subtype of ADHD. ADHD exclusion criteria were the presence of any current or lifetime mood (i.e., major depression or BD) or anxiety disorders. Participants with ADHD were not excluded for the presence of BD in their first-degree relatives (n = 5/38 [13%]) or the presence of comorbid behavioral disorders (e.g., ODD, CD, etc.); however, ADHD was their primary diagnosis.

Because this was not a treatment study, participants with BD and ADHD remained on their outpatient medication regimens; however, those taking ADHD stimulant medications were asked, but not required, to hold those medications for 4 drug half-lives before behavioral testing, as such medication holidays are common in standard clinical care on weekends or holidays.

TDC participant (n = 41) inclusion criteria were absence of current or lifetime psychiatric illness or substance abuse/dependence, and absence of first-degree relatives with history of psychiatric illness.

Measures

Overall functional impairment in the BD and ADHD groups was assessed using the Children's Global Assessment Scale (CGAS).⁴³ Participants with BD mania and depressive symptoms during the week of DANVA testing were assessed using the Young Mania Rating Scale (YMRS) and Children's Depression Rating Scale (CDRS).^44,45 ADHD symptoms for both patient groups were assessed using the Conners ADHD Rating Scale–Parent Report (ADHD index, t-scores).⁴⁶ Socioeconomic status (SES) was assessed via parent report on the four-factor Hollingshead scale.⁴⁷

DANVA Emotional Face Identification Task

Emotional face identification ability was assessed using the DANVA.⁴⁸ The DANVA comprises two subtests: adult and child emotional faces. Each subtest included 24 standardized photographs of models (12 male, 12 female) displaying one of four facial emotions (happy, sad, angry, fearful) at two levels of intensity (high and low). Faces were presented for 2 seconds, and participants had to choose which of the four emotions listed was expressed in the photograph. Both DANVA subtests have been standardized and demonstrate adequate construct validity, internal reliability (Cronbach α = 0.64–0.81), and test–retest reliability for children and adolescents 4 to 16 years old.⁴⁸ Outcome variables for each subtest include total errors per emotion, total errors at each intensity level, and total errors across all emotion conditions. Specific types of errors can also be examined (e.g., mislabeling a face as sad rather than angry).

Data Analysis

The Statistical Package for Social Sciences (version 19.0; SPSS Inc., Chicago, IL) was used for all analyses. We evaluated between-group differences in demographic variables (age, sex, FSIQ, and SES) using analysis of variance (ANOVA) or χ² tests as appropriate.⁴⁹ We examined potential between-group differences in ADHD symptoms and functional impairment using separate analysis of covariance (ANCOVA), covarying for significant between-group differences in FSIQ and SES.

We conducted four primary analyses of DANVA data: total child face errors, child face errors by emotion type (four emotions), total adult face errors, and adult face errors by emotion type (four emotions). Examination of child and adult total errors used general linear models (GLM) univariate analyses (GLM analysis of covariance [ANCOVA]), whereas examination of child and adult face errors by emotion type (four emotions each) used GLM multivariate analyses (GLM multivariate analysis of covariance [MANCOVA]). All models included FSIQ and SES as covariates. Main effects and subsequent pairwise comparisons were examined. Significance was set at p < .05 with Bonferonni correction for multiple comparisons. Effect size was evaluated using Cohen's d.⁵⁰ Post-hoc analyses were conducted to explore the effects of comorbid ADHD and medication.

RESULTS

Participants

There were significant between-group differences for both SES (F_2,106 = 7.57, p = 0.001) and FSIQ (F_2,106 = 9.41, p < .001), but not age, gender, or ethnicity. Specifically, participants with TDC had significantly higher average incomes and higher FSIQs than participants with BD and ADHD, who did not significantly differ from one another (Table 1).

TABLE 1.

Sample Characteristics

Variable	BD (n = 30)	ADHD (n = 38)	TDC (n = 41)	Significance
Age (years), mean ± SD	13.03 ± 2.99	12.08 ± 2.78	12.24 ± 3.23	F2,106 = 0.93
Sex: male, n (%)	20 (67)	22 (58)	16 (39)	χ2(2,N = 109) = 5.83
Full scale IQa	105.4 (11.0)	110.0 (13.9)	117.0 (8.8)	F2,106 = 9.41***
Ethnicity: white, n (%)	24 (80)	27 (71)	33 (81)	χ2(2,N = 109) = 1.20
Socioeconomic status (Hollingstead)b	41.1 (13.8)	38.6 (16.4)	50.6 (12.7)	F2,106 = 7.57**
K-SADS Diagnosis (past 6 mo), n (%)
Manic episode	27 (90)	0	–
Depressive episode	15 (50)	0	–
ADHD	18 (60)	38 (100)	–
Oppositional defiant disorder	21 (70)	8 (21)	–
Conduct disorder	2 (7)	0	–
Generalized anxiety disorder	4 (13)	0	–
Social phobia	0	0	–
Simple phobia	6 (20)	0	–
Separation anxiety disorder	2 (7)	0	–
Posttraumatic stress disorder	1 (3)	0	–
Obsessive-compulsive disorder	3 (10)	0	–
Young Mania Rating Scale, mean ± SD	8.5 ± 7.3	–	–
Children's Depression Rating Scale, mean ± SD	29.4 ± 12.0	–	–
Children's Global Assessment Scale, mean ± SD	63.4 ± 15.3	72.9 ± 13.8	–
Conners ADHD index t score,c mean ± SD	72.0 ± 12.5	68.8 ± 10.2	–	F3,64 = 1.81
Medication, n (%)
Lithium	11 (37)	0	–
Atypical neuroleptic	18 (60)	0	–
Antiepileptic	5 (17)	0	–
Antidepressant	5 (17)	0	–
Stimulantd	11 (37)	33 (87)	–
α-Agonist ADHD medication	5 (17)	4(11)	–

Note: K-SADS = Child Schedule for Affective Disorders and Schizophrenia.

^aFull Scale IQ is reported as standard scores on the Wechsler Abbreviated Scale of Intelligence (WASI).

^bSocioeconomic status is reported as total score on the Hollingstead.

^cData were not available for all participants for these variables: participants with bipolar disorder (BD) (n=29); participants with attention-deficit/hyperactivity disorder (ADHD) (n=37).

^dTwo participants with BD and 4 participants with ADHD were taking stimulant medication when tested.

^**p ≤ .01

^***p ≤ .001.

During the week of DANVA testing, the BD sample was overall euthymic by clinician-administered mood ratings (YMRS: 8.5 ± 7.3, CDRS: 29.4 ± 12.0). Among participants with BD, 70% (21/30) were euthymic (YMRS≤12 CDRS<40), 10% (3/30) depressed (YMRS≤12 CDRS≥40), 13% (4/30) hypomanic (YMRS=13–24 CDRS<40), 7% (2/30) mixed (YMRS>12 CDRS≥40), and none were manic (YMRS>25, CDRS<40). Participants with BD were significantly more impaired with lower CGAS scores than participants with ADHD (F_3,64 = 6.49, p = .01). BD and ADHD groups did not differ on ADHD symptom scores (F_3,64 = 1.81, p = .18), (ADHD Index t-scores: BD: 72.0 ± 12.5; ADHD: 68.8 ± 10.2). At testing, 18% (2/11) of BD and 12% (4/33) of participants with ADHD were taking their stimulant medications (χ² = 0.31, p = .58).

DANVA Child Face Labeling Errors

Total Errors

On total child face identification errors, we found a significant main effect of group (F_4,104 = 6.20, p = .003). Post-hoc pairwise com parisons showed that participants with BD (p = .03, Cohen's d = 0.69) and ADHD (p = .003, Cohen's d = 0.65) made signicantly more total errors on child faces than TDCs, but patient groups did not differ significantly from one another.

Errors by Emotion Type

Examining child face errors by emotion type, we found a significant effect of group (Wilks’ λ = 0.76, p = .001), with a univariate main effect of group on errors on happy (F_2,104 = 5.00, p = .008), angry (F_2,104 = 3.60, p = .03), sad (F_2,104 = 3.01, p = .05), and fearful faces (F_2,104 = 4.96, p = .009) (Table 2 and Figure 1).

TABLE 2.

Between-Group Differences in Emotional Face Identification Errors

	Group, mean ± SD			Analysis
	BD (n = 30)	ADHD (n = 38)	TDC (n = 41)	F	p	Post-hoc	Cohen's d
Child Face Errors
Total	4.47 ± 2.86	4.74 ± 3.79	2.85 ± 1.67	6.20	.003	ADHD>TDC, p = .003	0.65
						BD>TDC, p = .03	0.69
Happy	0.45 ± 0.68	0.13 ± 0.34	0.15 ± 0.36	5.00	.008	BD>TDC, p = .03	0.55
						BD>ADHD, p = .01	0.60
Angry	2.07 ± 1.31	1.84 ± 1.20	1.42 ± 1.05	3.60	.03	BD>TDC, p = .04	0.55
Sad	0.83 ± 0.95	0.61 ± 0.72	0.41 ± 0.84	3.01	.05	BD>TDC, p = .05	0.47
Fearful	1.10 ± 1.03	1.53 ± 0.80	0.88 ± 0.84	4.96	.009	ADHD>TDC, p = .009	0.79
Adult Face Errors
Total	6.37 ± 2.65	6.39 ± 4.43	4.80 ± 1.76	2.25	.11
Happy	0.80 ± 0.85	0.50 ± 1.06	0.27 ± 0.50	1.39	.25
Angry	1.90 ± 1.32	1.92 ± 1.24	1.85 ± 1.15	0.06	.94
Sad	1.90 ± 1.35	1.21 ± 1.38	1.12 ± 1.05	2.51	.09
Fearful	1.77 ± 1.38	2.13 ± 1.21	1.56 ± 1.14	2.08	.13

Note: ADHD = attention-deficit/hyperactivity disorder; BD = bipolar disorder; TDC = typically developing control.

FIGURE 1.

Errors on child facial emotions by group. Note: Means adjusted for Full Scale IQ and socio-economic status. ^aPediatric bipolar disorder (BD) significantly different from controls. ^bPediatric BD significantly different from attention-deficit/hyperactivity disorder (ADHD). ^cADHD significantly different from controls.

Post-hoc pairwise comparisons of happy face errors showed that participants with BD made significantly more errors on happy faces than participants with either ADHD (p = .01, Cohen's d = 0.60) or TDC (p = .03, Cohen's d = 0.55), but participants with ADHD and TDC did not differ.

To further probe this BD specific deficit, we examined the effect of emotional face intensity on child happy faces errors. We found that youths with BD (0.33 ± 0.55) were more likely than those with ADHD (0.11 ± 0.31) to make errors on low-intensity child happy faces (t_43.50 = 2.04, p = .05, Cohen's d = 0.49), but not high-intensity child happy face errors (BD 0.13 ± 0.43; ADHD 0.03 ± 0.16; t_35.41 = 1.28, p = NS).

Post-hoc pairwise comparisons of angry face errors revealed that youths with BD made significantly more errors than TDCs (p = .04, Cohen's d = 0.55), but not versus youths with ADHD. Youths with ADHD did not differ from TDCs.

Post-hoc pairwise comparisons of sad face errors showed that participants with BD made significantly more errors than TDCs (p = .05, Cohen's d = 0.47), but not significantly versus participants with ADHD, nor did participants with ADHD differ from TDCs.

Post-hoc pairwise comparisons of fearful face errors demonstrated that participants with ADHD made significantly more errors than TDC (p = .009, Cohen's d = 0.79), but not than participants with BD, nor did participants with BD differ from TDCs.

DANVA Adult Face Labeling Errors

Total Errors

On total adult face identification errors, there was no main effect of group (F_4,104 = 2.25, p = NS); therefore, we did not conduct post-hoc pairwise analyses.

Errors by Emotion Type

Examination of errors by facial emotion type on adult faces did not show a main effect of group (Wilks’ λ 0.90, F_{8, 202} = 1.39, p = NS); therefore, we did not examine univariate analyses.

Post-Hoc Exploration of Potential Medication Effects

To assess potential medication effects, GLM MANCOVAs were conducted on child subtest total errors, given prior results suggesting differences on child, but not adult faces. There was no effect of medication status (i.e., medicated versus not medicated) on child face total errors for participants with BD or ADHD (BD: F_3,26= 0.02, p = .88; ADHD: F_3,34 = 0.24, p = .6). In youths with BD, we found no differences on child face total errors among those taking atypical neuroleptics (n = 18) versus not (n = 12) (F_3,26 = 0.02, p = .9) or child face errors by emotion type (Wilks’ λ = 0.99, F_4,23 = 0.03, p = 1.0). Similarly, no effect was found when comparing youths with BD taking lithium (n = 11) versus not (n = 19) on child face total errors (F_3,26 = 0.15, p = .7) or child errors by emotion type face (Wilks’ λ = 0.76, F_4,23 = 1.81, p = .2).

Post-Hoc Examination of ADHD Diagnostic Comorbidity

To explore potential effects of ADHD, we conducted a post-hoc comparison of participants with BD with comorbid ADHD (BD + ADHD, n = 18) and participants with BD without ADHD (BD-ADHD, n = 12) versus the ADHD-only and TDC samples. We found significant main effects of group on child angry (F_3,103 = 2.73, p = .05), sad (F_3,103 = 3.51, p = .02), fearful (F_3,103 = 3.89, p = .01), and happy faces (F_3,103 = 3.35, p = .02) (Table 3).

TABLE 3.

Effect of Attention-Deficit/Hyperactivity Disorder (ADHD) Comorbidity on Child Face Identification Errors

	Group, mean ± SD				Analysis
Child Face Errors	BD+ADHD (n=18)	BD-ADHD (n = 12)	ADHD only (n = 38)	TDC (n = 41)	F	p	Post-hoc	Cohen's d
Total	5.11 ± 3.12	3.50 ± 2.20	4.74 ± 3.79	2.85 ± 1.67	4.67	.004	BD + ADHD > TDC, p = .03	0.90
							ADHD > TDC, p = .007	0.65
Happy	0.44 ± 0.70	0.50 ± 0.67	0.13 ± 0.34	0.15 ± 0.36	3.35	.02	BD + ADHD = BD–ADHD = ADHD = TDC	N/A
Angry	2.28 ± 1.60	1.75 ± 0.62	1.84 ± 1.20	1.41 ± 1.05	2.73	.05	BD + ADHD > TDC, p = .05	0.64
Sad	1.11 ± 0.90	0.42 ± 0.90	0.61 ± 0.72	0.41 ± 0.84	3.51	.02	BD + ADHD > TDC, p = .01	0.80
Fearful	1.27 ± 0.21	0.84 ± 0.27	1.53 ± 0.80	0.87 ± 0.15	3.89	.01	ADHD > TDC, p = .021	1.15

Note: ADHD = attention-deficit/hyperactivity disorder; BD = bipolar disorder; IDC = typically developing control.

Post-hoc pairwise comparisons on child angry faces showed that participants with BD+ADHD differed significantly from TDCs (p = .05, Cohen's d = 0.64), but not from participants with BDADHD or ADHD-only. Similarly, on child sad faces, participants with BD+ADHD differed significantly from TDCs (p = .01, Cohen's d = 0.80), but not from participants with BD-ADHD or ADHD-only. For child fearful faces, participants with ADHD-only made significantly more errors than TDCs (p = .02, Cohen's d = 1.15), but not than participants with BD+ADHD or BDADHD. There were no significant pairwise differences on child happy faces errors.

On child face total identification errors, we found amain effect of group (F_3,103 = 4.67, p = .004). Post-hoc pairwise comparisons showed that participants with BD+ADHD (p = .03, Cohen's d = 0.90) and participants with ADHD-only (p = .007, Cohen's d = 0.65) made significantly more total child face errors than TDCs, but not versus participants with BD-ADHD. There were no significant differences between participants with BD+ADHD and ADHD-only.

DISCUSSION

The principal finding of our study, the first to compare emotional face identification in youths with primary BD versus primary ADHD, was that participants with BD made significantly more errors while identifying child happy faces than either participants with ADHD or TDC. In this BD-specific deficit, youths with BD were more likely to make errors on low- rather than high-intensity happy faces. Participants with BD also made significantly more errors on child angry and sad faces than participants with TDC. Participants with ADHD made more errors on child fearful faces than participants with TDC. Both patient groups made more total face identification errors than TDCs on the child faces subtest. No significant group differences were found using adult faces. Taken together, this suggests that child happy face identification may be a specific behavioral marker of children with primary BD versus those with primary ADHD or TDC.

Our results are consistent with other behavioral studies of emotional face processing in youths with BD. Our finding that youths with BD made more total child face identification errors than TDCs aligns with previous DANVA studies in youths with BD.^14,17 Similar to McClure et al. (2005), our participants with BD made more identification errors on child happy and angry faces than TDCs.¹⁵ Perhaps most intriguing and unique is our finding that participants with BD made significantly more errors on child happy faces than either participants with ADHD or TDC, suggesting a BD-specific behavioral deficit in emotional face processing. Our finding that youths with BD were more likely to make errors on low- versus high-intensity happy faces complements research showing youths with BD require greater emotional intensity to correctly identify facial emotions compared to TDCs.¹⁶

Our data also advance what is known about emotional face identification in youths with ADHD.³⁰ Youths with ADHD made more total errors when identifying child faces than did TDCs, which is consistent with prior literature.^31-34,36 Others have shown that children and adults with ADHD have altered neural and behavioral responses to fearful faces versus TDCs.^36,39,51 Given that noradrenergic pathways regulate fear responses and have been implicated in the pathophysiology of ADHD, further work is needed to examine the relationship between fear processing and noradrenergic circuitry in youths with ADHD.^37,52 Our findings contrast with work suggesting that children with ADHD have particular difficulty with the identification of angry and sad faces.³² However, this discrepancy may result from use of different tasks, as Pelc used a morphed faces task probing “happiness,” “sadness,” “anger,” and “disgust,” whereas we used the nonmorphed DANVA, tapping happiness, anger, sadness, and fear.³² Most importantly, we extend the ADHD emotional face processing literature by comparing youths with ADHD those with another form of psychopathology—i.e., BD.

Our results differ from those of Guyer et al., the only other study to compare youths with BD versus ADHD. We found that both participants with BD and ADHD made more total child, but not adult, emotional face identification errors, whereas Guyer et al. found that participants with BD and SMD made more total errors in both child and adult emotional face identification than youths with ADHD/CD, anxiety/depression (ANX/DEP) or TDCs. They also found that participants with ADHD/CD were no different than TDCs on overall face identification ability. This discrepancy may result from Guyer's use of a combined ADHD and/or CD group, which may represent an extreme sample of youths who are not representative of youths with ADHD more generally.³⁸ Moving forward, differences in emotional face identification in youths with primary BD versus primary ADHD require replication.

Our behavioral results should also be examined within the context of the neuroimaging literature examining emotional face processing in BD. Compared to TDCs and individuals with other forms of psychopathology, adults and children with BD demonstrate aberrant neural activation and connectivity in fronto-amygdalastriatal regions when processing happy faces.^18-25 For example, atypical striatal activation has been shown during happy face processing and encoding among adults and youths with BD.^19,20,23,24 Furthermore, adults with BD and their unaffected first-degree relatives have increased medial prefrontal cortex activation versus TDCs when processing happy faces; however, only those with BD have increased activation in the putamen.⁵³ Finally. Blumberg et al. found significant increases in amygdala activation for adult participants with BD versus TDCs when passively viewing happy faces.¹⁹ Although TDCs find happy faces rewarding,^26,27 neuroimaging and behavioral studies of happy face processing in individuals with BD, including our own, suggest the need to determine whether individuals with BD find happy faces less socially rewarding or even aversive. However, this hypothesis requires further empirical testing.

Albeit highly speculative, the relationship between our data indicating impaired emotional face identification is interesting in light of others’ work on social impairment in youths with either BD or ADHD. Compared to youths with ADHD and TDCs, youths with BD demonstrate greater impairments in parental and peer social interactions.⁵⁴ Furthermore, even asymptomatic youths with BD demonstrate significant interpersonal deficits compared to TDCs.⁵⁵ Although few studies have examined the relationship between impaired social skills and emotional face processing in youths with either BD or ADHD, a preliminary study by Rich et al. found a negative relationship between delayed face identification and parent-reported social reciprocity (e.g., social awareness, social information processing, etc.) in youths with BD.⁵⁶ Being effective in social relationships requires the ability to interpret others’ emotions and to respond appropriately. Thus, impaired emotional identification in youths with either primary BD or primary ADHD might lead to inappropriate behavioral responses, such as reactive aggression or intrusions on others. Further work is needed to examine the impact of deficits in emotional face processing on actual social impairments with peers and adults in youths with either BD or ADHD.

Our study has several limitations. First, most participants with BD were taking their usual psychiatric medications when tested, as this was not a treatment study, and it would be unethical to withdraw them from anti-manic or anti-depressant medications for a phenomenology study. Our post-hoc analyses did not find differential DANVA performance based on participants with BD's current use of lithium or atypical neuroleptics. To minimize the effect of stimulants, most participants with BD and ADHD were tested having held their stimulants for at least 4 drug half-lives. However, further examination of the effect of psychotropic medications on emotional face identification is needed.

Second, the BD sample as a whole was euthymic, although there was some heterogeneity. Additional work is warranted to examine the role of mood state on emotional face identi-fication in a sample of youths with BD sufficiently powered to examine all potential mood states, given exploratory analyses suggesting a relationship between depression symptoms and child sad face errors (see Supplement 1, available online).

Third, exploratory analyses showed that ADHD symptoms, regardless of child diagnosis, predicted child emotional face total, sad, and angry face errors, suggesting the need to better understand the role of inattention and hyperactivity/impulsivity in emotional face identification (see Supplement 1, available online).

Fourth, 70% of our BD sample had comorbid ODD, which, albeit consistent with rates of comorbid ODD in other studies of pediatric BD (range, 47–88%), is higher than that of Guyer et al. (41%), thus potentially limiting comparisons of these studies.^36,38,57,58 Furthermore, our BD groups had a higher rate of comorbid ODD than the ADHD (21%) group, and neither BD nor ADHD groups had sufficient numbers of participants with and without comorbid ODD for adequately powered evaluations of its effect on emotional face identification. We note that this is an important area for future investigation.

Finally, as with all behavioral tasks, the DANVA has some limitations. For example, its lack of neutral trials used as a control condition versus specific emotional exemplars (i.e., angry) reduces our ability to determine whether that would help tease apart whether our groups had specific difficulties with emotional face processing (e.g., emotional aspect of the task) versus more global difficulties with face processing. Although the DANVA's limited number of trials per emotion could potentially reduce power to detect group-by-emotion interactions, this was neither true in our study nor in a wealth of other studies.^14,15,17,38 Also, it is beyond the DANVA's scope to disentangle potential contributions of other cognitive processes, including working memory or attention, to emotional face processing, which might be especially relevant to ADHD. However, others, have shown deficits in emotional face processing among youths with ADHD even when accounting for deficits in cognitive control.⁵⁹ Future work is needed with behavioral tasks designed to elucidate the relationship between cognition and emotional face processing.

In summary, comparing emotional face identification ability in children and adolescents with either primary BD, primary ADHD, or TDCs, we found a BD-specific deficit on child happy faces, with youths with BD making more errors than both youths with ADHD and TDCs. Such results suggest a potential disorder-specific behavioral marker of BD in youths. Further research is needed to elucidate the neural underpinnings of emotional face identification, especially of happy faces, as a potential bio-behavioral marker of pediatric BD.