ICGP AWARD WINNING PAPER
Emotion recognition processing as early predictor of response to 8-week citalopram treatment in late-life depression Paulo R. Shiroma1,3, Paul Thuras1,3, Brian Johns2,3 and Kelvin O. Lim1,3 1
Mental Health Service Line, Minneapolis VA Medical Center, Minneapolis, MN, USA Department of Psychiatry, North Memorial Medical Center, Minneapolis, MN, USA 3 Department of Psychiatry, University of Minnesota Medical School, Minneapolis, MN, USA Correspondence to: P. R. Shiroma, E-mail: [email protected] 2
Background: In subjects with depression, exposure to antidepressants improves recognition of positive emotions. This phenomenon, which occurs early in the course of treatment, has been proposed as the initial step in the mechanism of action to subsequent therapeutic effects of antidepressants. To this date, it has not been well examined among older depressed patients. Method: Older subjects with non-psychotic major depressive disorder were treated with citalopram in an 8-week open-label study. The main predictor of response and remission was the change in emotion recognition between baseline and day 7. Covariates included executive functions, baseline anxiety level, medical comorbidity, level of subjective stress, serum citalopram level, and level of social support. Results: Twenty-seven patients were considered for final analysis. Overall, accuracy of emotion recognition significantly improved between baseline (75%) and day 7 (83%) (X2 = 34.50, df = 1, p < 0.001). Improvement to identify happy expressions occurred at 25% and 50% intensity with ceiling effect at 0%, 75%, and 100%. Change in emotion processing was marginally significant in predicting antidepressant response at day 56. Multivariate analysis showed that emotion processing is a significant predictor of response and remission when considered along with perceived level of social support. Conclusions: Recognition of mildly intense happy expression, which improved early in the course of citalopram treatment, predicts subsequent antidepressant response and remission when considered along with perception of social support. Further studies would be necessary to examine specific neural substrates in the affective network involved in the acute therapeutic action of antidepressant in late-life depression. Copyright # 2014 John Wiley & Sons, Ltd. Key words: depression; old age; older adults; citalopram; emotions History: Received 16 December 2013; Accepted 25 February 2014; Published online 4 April 2014 in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/gps.4104
Introduction Antidepressants are one of the most commonly prescribed medications in the USA (National Center for Health Statistics (US) 2011). One of the largest and longest studies ever carried out to evaluate depression treatment, the Sequenced Treatment Alternatives to Relieve Depression (STAR*D) study (Trivedi et al., 2006), showed that about 30% of patients achieved remission after first-line antidepressant citalopram. The STAR*D Copyright # 2014 John Wiley & Sons, Ltd.
study also suggested that any antidepressant or combination given late in treatment appears to work less effectively for attaining and sustaining remission than when given earlier in a sequence of antidepressant treatments (Rush et al., 2006; Warden et al., 2007). If treating clinicians knew the probability of response earlier in the course of treatment, they would be in a better position to advise patients about various treatment options, therefore limiting harmful neurobiological effects and poor outcome secondary to Int J Geriatr Psychiatry 2014; 29: 1132–1139
Emotion recognition in late-life depression
1133
enduring depressive symptoms. This situation could be particularly advantageous for the treatment of late-life depression (LLD) where treatment is often challenged by age-related sensitivity to medications and medical comorbidity (Lotrich and Pollock, 2005). Predictors of antidepressant drug response in LLD have rendered mixed results with research mainly concentrated in the role of executive dysfunction and the presence of deep white matter hyperintensities in T2-weighted brain magnetic resonance imaging. The study of emotion processing in the cause, maintenance, and relapse of affective disorders is largely limited to mid-life adult population (Rubinow and Post, 1992; Mikhailova et al., 1996; Bouhuys et al., 1999; Lawrence et al., 2004; Fu et al., 2007). Experiments in middle-aged healthy (Harmer et al., 2003; Harmer et al., 2004) and depressed subjects (Gorlyn et al., 2008; Harmer et al., 2009b) have shown improvement of emotion recognition biases after acute administration of antidepressant drugs (i.e., citalopram and reboxetine). Interestingly, increased accurate recognition of happy faces at 2 weeks of treatment predicts clinical outcome at 6 weeks of continued antidepressants (Tranter et al., 2009). Older adults with psychiatric conditions have reported less negative affect intensity than their younger counterparts, suggesting an interaction between aging and psychopathology on emotion processing (Cheavens et al., 2008). In this study, we investigated early changes in facial emotion recognition among older depressed subjects treated with citalopram as a predictor of subsequent response after 8 weeks of treatment.
Subjects with 17-item Hamilton Depression Rating Scale (HDRS17) score ≥18 were accepted in the study. Diagnoses of major depressive disorder, single or recurrent, were established by the Operational Criteria (OPCRIT) Diagnostic System (Azevedo et al., 1999). OPCRIT is a system of computer programs to be completed by trained clinicians from diagnostic interview and case records. OPCRIT generates diagnoses according to 12 operational diagnostic systems including the Diagnostic and Statistical Manual of Mental Disorders and the International Classification of Diseases. OPCRIT was performed on all participants by the principal investigator (P. R. S.). Exclusion criteria were as follows: Folstein mini mental status examination score ≤ 24; previous failure within the last year to at least two pharmacologically different antidepressants given for at least 6 weeks at the maximum tolerated dose; antidepressant exposure within 2 weeks (6 weeks for fluoxetine) prior to baseline assessment; and concurrent use of antipsychotics, mood stabilizers, or anticonvulsants. Patients were also excluded if they had substance use disorder within 6 months of assessment; positive urine toxicology screen test; history of psychosis, bipolar disorder, post-traumatic stress disorder, or acute stress disorder; electroconvulsive therapy within previous 6 months; Parkinson’s disease, dementia, seizures, or other central nervous system-related disorders; history of traumatic brain injury; and any clinically unstable medical illness or its treatment (e.g., interferon) that could compromise the patient’s ability to tolerate the study or likely interfere with the procedures or results.
Methods and materials
Rating scales and procedures
Participants
Baseline assessments included clinical and demographic information, 15-item Geriatric Depression Scale (GDS; Yesavage et al., n. d), Wechsler Adult Intelligence Scale—digit symbol substitution test (Wechler, 2008), and Stroop test (Treisman and Fearnley, 1969). Other covariates of antidepressant response included baseline anxiety (Hamilton Anxiety Scale; Hamilton, 1959), cardiovascular risk factors (cerebrovascular risk factor assessment [CRFA]; Wolf et al., 1991), level of social support (Interpersonal Support Evaluation List [ISEL]; Cohen and Hoberman, 1983), and severity of recent stress (4-item Perceived Stress Scale; Cohen et al., 1983). Subjects then received citalopram 10 mg/day for the first 7 days followed by 20 mg/day for the next 7 days. Any further dose adjustment was based on tolerability and clinical response by the treating clinician. Citalopram is recommended as the first-line treatment
Veterans aged 55 years and older were recruited from primary care and psychiatric outpatient clinics at the Minneapolis VA Medical Center over a 2-year period. Patients who received a prescription of citalopram for depression were identified through electronic notifications to the research team by the clinical pharmacist. Before initiating citalopram treatment, patients had a brief phone interview explaining the purpose of the study. Those patients interested and eligible to participate were invited for a face-to-face screen assessment. The Minneapolis VA Medical Center Institutional Review Board approved the study, and written informed consent was obtained from all subjects before participation. Participants received monetary compensation for their time committed to the study. Copyright # 2014 John Wiley & Sons, Ltd.
Int J Geriatr Psychiatry 2014; 29: 1132–1139
P. R. Shiroma et al.
1134
for LLD given its efficacy and tolerability compared with other antidepressants with less potential for drug–drug interactions (Alexopoulos et al., 2001). The effect of citalopram on the severity of depressive symptoms was measured by the HDRS17 (Hamilton, 1960) score at week 8, with response defined as a ≥50% decrease from the baseline score and remission defined as a final score ≤7. The percent decrease in HDRS17 and 15-item GDS rating scales was also examined. Prospective follow-up on antidepressant response was ascertained together with medication-related side effects (Frequency, Intensity, and Burden of Side Effects Rating Scale; Wisniewski et al., 2006) at days 3, 10, 14, 21, 28, 35, 42, and 56. At day 7, trough citalopram serum concentration (MedTox Scientific, Inc., St. Paul, MN) was obtained as a covariate of change in emotion recognition. DNA extraction for genetic analysis was also obtained (not reported here). Facial emotion recognition task
Facial emotion recognition task (FERT) was administered prior to the onset of citalopram treatment and at day 7. Facial images were taken from the Facial Expressions of Emotion: Stimuli and Tests (Young et al., 2002) and presented by Superlab Pro on a laptop computer in a quiet room to minimize distracting noise. Written instructions about the purpose and procedures of the test were presented on the laptop screen followed by a brief practice session. A research staff was present during sessions to answer questions. FERT was delivered in two blocks. Block #1 was composed of facial images (three men and three women) depicting 12 neutral expressions and 48 happy expressions. Faces were randomly presented at 25%, 50%, 75%, and 100% of emotional intensity and at 500- or 5000-ms duration of exposure. The intensity of emotions and duration of exposures were counterbalanced in each session. Each facial image was followed by a blank screen with a white crosshair in the center to prompt response. Although no time limit to response was imposed, subjects were asked to identify as quickly and as accurately as possible whether the facial expression was happy or not by pressing a labeled key on a response box (DirectIN High Speed), which was automatically coded as accurate or inaccurate identification. A new fresh image was shown 2 s after the participant’s response. Block #2 followed the same randomized sequence of images as the first block to control for practice effect. Following completion of FERT, participants had a gender recognition task (male or female). Thirty male Copyright # 2014 John Wiley & Sons, Ltd.
and 30 female faces depicting 48 emotional expressions and 12 neutral expressions were randomly presented following the same procedures as FERT. This task served as an active experimental control for aspects of facial perception unrelated to emotion recognition such as visual and attentional processing. Statistical analysis Statistical analyses were performed using SPSS v. 19. Chi-square statistics were used to examine differences between differences in accuracy on the recognition task (e.g., duration, block order, and intensity). Change in accuracy by intensity was examined using analysis of variance. A mixed-effects regression model was used to examine change in symptoms over time. Multivariate logistic regression analysis was used to examine predictors of remission and response for the indicators of depressive symptoms (HDRS17 and GDS). Multiple regression analyses were used to test for interaction effects in predictors of symptom change. The level of significance (alpha) was set at 0.05. Results Sixty-two subjects were screened; 32 were excluded (n = 22) or refused to participate (n = 10). Demographic and clinical characteristics of the study sample (n = 30) are presented in Table 1. The sample was composed only of men in late adulthood, mostly Caucasians, unemployed, and with high school or higher education. Four subjects were hospitalized for depression, and one subject had attempted suicide. The average onset of the first major depressive episode occurred at middle age. Major depressive disorder was mostly recurrent, with current episode prolonged by several weeks, moderate to severe in intensity, and comorbid with moderate level of anxiety. The sample had a probability of stroke in 10 years of 11.2% based on the CRFA (range 2.6–87.9%), similar to a previous study of LLD (Alexopoulos et al., 2009). Twenty-seven subjects considered for analyses completed FERT at baseline and at day 7, had detectable serum level of citalopram, and had at least one HDRS17 measure at or after day 10. The overall response and remission rates at day 56 for these 27 subjects were 46.6% and 40.0%, respectively. There was a significant reduction in HDRS17 and GDS scores during treatment (F = 34.07, df = 9, p < 0.0001 and F = 23.803, df = 9, p < 0.0001, respectively). Int J Geriatr Psychiatry 2014; 29: 1132–1139
Emotion recognition in late-life depression
1135
Table 1 Demographic and clinical characteristics of study sample Age (years), mean (SD) Age (years) at first MDE, mean (SD) Family history of mental illness, no. Mood disorder Other psychiatric disorder Past history of substance use, no. Alcohol Illicit drugs Duration of current episode (months), mean (SD) At least one previous antidepressant trial during current episode, no. Yes No Recurrent MDD, no. Yes No HAM-A at baseline, mean (SD) HDRS17, mean (SD) GDS15, mean (SD) MMSE, mean (SD) Color-Word Stroop, mean (SD) Serum citalopram level at day 7 (ng/mL), mean (SD) ISEL score at baseline (range = 0–36), mean (SD) PSS-4 score at baseline (range = 0–12), mean (SD) CRFA score at baseline, mean (SD)
65.3 (7.0) 49.4 (19.8) 9 15 11 6 26.2 (26.3) 18 12 19 11 19.0 (5.4) 23.4 (3.3) 11.4 (1.9) 28.0 (1.28) 28.9 (5.9) 28.1 (16.9) 19.8 (7.3) 9.1 (2.0) 11.4 (4.7)
MDE, major depressive episode; MDD, major depressive disorder; HAM-A, Hamilton Anxiety Scale; HDRS17, 17-item Hamilton Depression Rating Scale; GDS15, 15-item Geriatric Depression Scale; MMSE, mini mental state examination; ISEL, Interpersonal Support Evaluation List; PSS-4, 4-item version of Perceived Stress Scale; CRFA, cerebrovascular risk factor assessment.
Overall, accuracy of emotion recognition significantly improved between baseline (75%) and day 7 (83%) (X2 = 34.50, df = 1, p < 0.001). We found nonsignificant changes in accuracy between blocks #1 (75%) and #2 (76%) (X2 = 0.24, df = 1, p = 0.62) at baseline or at day 7 (block #1 = 83.1% and block #2 = 84.2%, X2 = 0.36, df = 1, p = 0.54). As we found no practice effect over identification of facial expressions, all subsequent analyses combined the results from blocks #1 and #2 as a single trial.
We analyzed the effect of intensity of stimuli and length of exposure over the accuracy to identify emotions. We found non-significant association between accuracy of responses and the length of exposure of each facial stimuli (X2 = 0.37, df = 1, p = 0.54). Improvement in FEST performance varied by the intensity of facial expressions (Figure 1). At 25% intensity, accuracy changed from 22% to 40% (t = 3.53, df = 24, p = 0.002); at 50% intensity, accuracy improved from 69% to 89% (t = 5.00, df = 24, p < 0.001). Minimum improvement in accuracy occurred at 0%, 75%, and 100% (F(1, 74) = 1.14, p = 0.29). Thus, all subsequent analyses considered the change of emotion recognition as the combined accuracy responses from 25% and 50% stimuli intensity. There was a significant difference in reaction time from baseline (mean = 606.63 ms, SD = 33.68) to day 7 (mean = 581.23 ms, SD = 34.15; t = 2.7512, df = 52, p = 0.008). As subjects were not limited on the time to response, we also analyzed whether there was any speed–accuracy trade-offs during FERT, meaning higher accuracy achieved with longer time to respond. We found that reaction time had a significant negative correlation with emotion recognition (r = 0.46, p = 0.01) where greater accuracy correlated with faster reaction time. Thus, no speed–accuracy trade-off was found. Serum level of citalopram measured at day 7 shows a non-significant correlation with changes in both emotion recognition (r = 0.21, p = 0.31) or HDRS17 score (r = 0.18, p = 0.45) from baseline to day 7. We analyzed changes in accuracy of gender recognition to control for visual and attentional processing. We found a significant but small increase in gender recognition from baseline (94%) to day 7 (97%) (X2 = 12.00, df = 1, p < 0.001, φ = 0.084). A non-significant correlation was found between gender recognition task and FERT (r = 0.24, p = 0.35).
Figure 1 Change in the accuracy of emotion recognition from baseline to day 7 by the intensity of facial stimuli.
Copyright # 2014 John Wiley & Sons, Ltd.
Int J Geriatr Psychiatry 2014; 29: 1132–1139
P. R. Shiroma et al.
1136
We conducted univariate and multivariate analyses with change in FEST treated both as a categorical and as a continuous variable. The categorization of accuracy change produces easily understandable results and facilitates the study of interaction effects. Analyses of accuracy change as a continuous variable decreases the likelihood to lose information between variables because of categorization. Univariate analysis showed improved accuracy as a marginally significant predictor of antidepressant response at day 56 (weight coefficient = 0.043; SE = 0.024; Wald X 2df ¼1 ¼ 3:04, p = 0.08; odds ratio = 1.044; 95% CI = 0.995–1.095) where a 10% improvement in FEST by day 7 increases the likelihood of response by 44%. Similar results were found for predicting remission at day 56 (weight coefficient = 0.046; SE = 0.025; Wald X 2df ¼1 = 3.41, p = 0.065; odds ratio = 1.047; 95% CI = 0.997–1.100) where a 10% improvement in FEST by day 7 increases the likelihood of response by 47%. Covariates including HDRS17 baseline score, accuracy in gender recognition, age, Hamilton Anxiety Scale score, Stroop test, digit symbol substitution test, citalopram serum level, ISEL, Perceived Stress Scale, CRFA, and mini mental status examination were non-significant predictors of response or remission (data not shown). Figure 2 shows the study sample grouped in tertiles on the basis of the percentage of improvement recognition from baseline to day 7. We found a significant association with change in HDRS17 scores over time (F = 3.58, df = 2, p = 0.045). Subjects with no improvement had a mean HDRS17 score decreased by 5 points during the course of treatment. Subjects with improved accuracy from baseline up to 25% and those with >25% had a decrease in HDRS17 score of 15 and 14 points at the study end, respectively.
We constructed a multivariate logistic regression model to predict response at day 56. Independent effect of change in emotion recognition (FEST), perceived social support (ISEL) at baseline, and an interaction term of FEST by ISEL were included. Analysis showed that change in emotion recognition by day 7 and social support became significant predictors of response (weight coefficient = 0.072; SE = 0.033; Wald X 2df ¼1 = 4.68, p = 0.03; odds ratio = 1.075; 95% CI = 1.007–1.148 for emotion recognition change; weight coefficient = 0.165; SE = 0.083; Wald X 2df ¼1 = 3.92, p = 0.048; odds ratio = 1.179; 95% CI = 1.002–1.388 for social support). This was also observed in the model predicting depression remission (weight coefficient = 0.082; SE = 0.036; Wald X 2df ¼1 = 5.17, p = 0.023; odds ratio = 1.085; 95% CI = 1.011–1.164 for accuracy change; weight coefficient = 0.181; SE = 0.088; Wald X 2df ¼1 = 4.25, p = 0.039; odds ratio = 1.199; 95% CI = 1.009–1.424 for social support). The interaction of emotion recognition and social support did not meet significance for response (p = 0.088) or remission (p = 0.071). When categorization of emotion recognition was entered, subjects with improved accuracy >25% from baseline had the highest likelihood to reduce HDRS17 scores over time compared with those with no improvement adjusted for level of social support. ISEL scores also became a significant predictor of change in depressive symptoms adjusted for emotion recognition. A multiple regression analysis found that the interaction between change in FEST and ISEL was significant (β = 0.507, t = 3.38, p = 0.003; Figure 3). Exploratory analysis revealed a significant change in HDRS17 score by day 7 (F(2, 25) = 7.04, p = 0.004). However, decreased scores in HDRS17 did not significantly correlate with FEST improvement (p = 0.14) or
HDRS17
Days of Treatment with Citalopram Figure 2 Repeated Hamilton Depression Rating Scale (HDRS17) score over time by groups based on the percentage of improvement recognition of happy facial expressions from baseline to day 7.
Copyright # 2014 John Wiley & Sons, Ltd.
Int J Geriatr Psychiatry 2014; 29: 1132–1139
Emotion recognition in late-life depression
1137
Negative or no improvement
25% improvement
Figure 3 Effect of categorization of emotion recognition task by perception of social support (Interpersonal Support Evaluation List [ISEL]) at baseline over likelihood of change in depression severity measured by Hamilton Depression Rating Scale (HDRS).
became a significant predictor of response (p = 0.59) or remission (p = 0.31) by day 56. Using mixed regression models, we also examined the impact of change in emotion recognition on reduction of depressive symptoms during different segments from 7 days post-initiation of citalopram. We found that the effect was maximized from 7 to 14 days (F(1, 52.3) = 9.47, p = 0.003) but was no longer significant for the segment from 14 to 21 days (F(1, 25.6) = 1.21, p = 0.281). Discussion Our results extend to older adults the growing evidence that antidepressants exert a change in emotion processing prior to the improvement of mood symptoms as we found an increased recognition of facial expression of happiness as early as 1 week of continued treatment with citalopram. Moreover, emotion recognition along with perceived social support became significant predictors of subsequent antidepressant response and remission. Tranter and colleagues (Tranter et al., 2009) found a significant correlation between accurate recognition of happy faces at 2 weeks of treatment and clinical outcome at 6 weeks of treatment. The authors suggested that early changes in emotional processing may underlie clinical response to antidepressants; however, the emotion task performed at 2 weeks opened the possibility that changes in emotion recognition might be secondary to improvements in mood rather than direct antidepressant effect. Our study addressed this Copyright # 2014 John Wiley & Sons, Ltd.
issue by conducting an emotional task at baseline and then at 1 week of treatment. Although both depression scores and gender recognition improved at 1 week of treatment, they did not correlate with changes in emotion recognition nor were predictors of response or remission at the study end. These findings further support the hypothesis that early change in emotion recognition due to antidepressants is an unrelated phenomenon to symptomatic or visual/ attention processing. According to this view, Harmer and colleagues (Harmer et al., 2009a) have proposed that the effects of antidepressants on the processing of facial expressions are seen rapidly, but the translation into improved subjective mood takes time as the patient learns to interpret social signals as positive, leading to social participation, which eventually could improve mood symptoms of depression. In contrast, inaccurate recognition of facial emotions would lead to appraise social situations more negatively, minimize social interaction, and even elicit rejection by others perpetuating a depressive state. A “cognitive neuropsychological” model of depression (Roiser et al., 2012) posits that negative emotion processing biases play a central causal role in the development of depression and may even be possible to predict which patients will benefit most from which treatments on the basis of neural responses to negative stimuli. In our study, improved recognition of happy expressions and higher social support at baseline increased the likelihood of better antidepressant outcomes. It is conceivable that corrected emotion processing mechanisms Int J Geriatr Psychiatry 2014; 29: 1132–1139
P. R. Shiroma et al.
1138
may exert therapeutic action by acting upon a supportive social network. In fact, age-related changes in the ability to interpret the feelings of others influenced the quality of social relationships in later life (Malatesta et al., 1987), and the subjective social support was reported as an important clinical factor in predicting shorter time to remission among geriatric depressed patients (Bosworth et al., 2002). Depression may be characterized primarily by difficulties in the processing of positive affect (Surguladze et al., 2004). Neuroimaging studies identified a low activation in subcortical and limbic regions in response to happy faces (Lawrence et al., 2004; Fu et al., 2007), and depression has been associated with decreased memory for happy faces (Gilboa-Schechtman et al., 2002). However, other studies found no deficits in depressed patients (Kan et al., 2004; Leppanen et al., 2004; Gollan et al., 2008). These discrepant results could be in part due to methodological inconsistencies that include the use of different facial stimulus sets and varied emotion processing tasks. The numbers of choices available in several cognitive tasks have demonstrated to influence older adult’s performance (Fozard et al., 1994). For instance, choosing one label from six overlapping constructs may tap cognitive rather than emotion processing (Wieser et al., 2006). In this study, the task to identify facial expressions required a simple YES/NO labeling of target emotion (e.g., happy face), which avoids high demands on cognitive effort but remained sensitive to detect changes in emotion recognition. The emotional intensity of the facial expression is also a factor to consider when measuring emotion recognition deficits (Kohler et al., 2003). Although depressed subjects may require significantly greater intensity to correctly identify happy expression compared with controls (Yoon et al., 2009), happiness is an easier emotion to detect than other emotions because of the presence of a smile (Dailey et al., 2002). Consistent with previous studies (Surguladze et al., 2004; Csukly et al., 2009), we found that moderately intense but still recognizable emotional stimuli captured the changes in emotion recognition, whereas neutral or high-intensity stimuli may result in a ceiling effect. Our study has several limitations. The open-label design, small sample size, and lack of placebo group limit stringent analysis of clinical factors previously reported to affect antidepressant response in LLD (e.g., executive dysfunction and severity of anxiety). However, the current study did not intend to test the efficacy of citalopram per se but rather to investigate the early changes in emotion recognition as predictor of subsequent antidepressant outcomes. Concurrent Copyright # 2014 John Wiley & Sons, Ltd.
interventions such as psychological treatment were not controlled and could have influenced response, remission, and even improvement in emotion recognition. Despite the small sample size, the current report extends previous finding in younger population to older adults on how antidepressants exert a therapeutic action through the modification of affective processing. Finally, it is unknown whether an antidepressant with a different pharmacological profile than citalopram would have modified facial recognition in a similar way. Tranter and colleagues (Tranter et al., 2009) did not find a significant difference in facial emotion recognition between citalopram (serotonergic) and reboxetine (noradrenergic), suggesting that in younger populations, antidepressants do not exert differential effects on emotion processing. In conclusion, accuracy recognition of moderately intense happy expression improved after 1 week of treatment with citalopram and was a significant predictor of later response and remission in LLD. Whether emotion processing in depression is a necessary initial step of improvement in mood symptom or an epiphenomenon during a depressive episode is unclear. Future studies should include development of interventions to target and improve emotion processing, which ultimately may enhance depression treatment. Emotion activation tasks in neuroimaging studies could also identify specific circuits in the affective network that relate to the early changes in emotion recognition. Conflict of interest None declared. Key points
• •
Recognition of mildly intense happy expression improved early in the course of citalopram treatment among older adults with depression. Improvement in emotion recognition along with perception of social support predicts subsequent antidepressant response and remission.
Acknowledgements Dr Shiroma was supported by the Minneapolis VA Center for Epidemiological and Clinical Research (CECR), a VA Clinical Research Center of Excellence, and the Mental Health Service Line, Minneapolis VA Medical Center. Int J Geriatr Psychiatry 2014; 29: 1132–1139
Emotion recognition in late-life depression
References Alexopoulos GS, Katz IR, Reynolds CF, 3rd, et al. 2001. Pharmacotherapy of depression in older patients: a summary of the expert consensus guidelines. J Psychiatr Pract 7(6): 361–376. Alexopoulos GS, Murphy CF, Gunning-Dixon FM, et al. 2009. Serotonin transporter polymorphisms, microstructural white matter abnormalities and remission of geriatric depression. J Affect Disord 119(1-3): 132–141. Azevedo MH, Soares MJ, Coelho I, et al. 1999. Using consensus OPCRIT diagnoses. An efficient procedure for best-estimate lifetime diagnoses. Br J Psychiatry J Ment Sci 175: 154–157. Bosworth HB, McQuoid DR, George LK, Steffens DC. 2002. Time-to-remission from geriatric depression: psychosocial and clinical factors. Am J Geriatr Psychiatry Off J Am Assoc Geriatr Psychiatry 10(5): 551–559. Bouhuys AL, Geerts E, Gordijn MC. 1999. Depressed patients’ perceptions of facial emotions in depressed and remitted states are associated with relapse: a longitudinal study. J Nerv Ment Dis 187(10): 595–602. Cheavens JS, Zachary Rosenthal M, Banawan SF, Lynch TR. 2008. Differences in emotional experience and emotion regulation as a function of age and psychiatric condition. Aging Ment Health 12(4): 478–487. Cohen S, Kamarck T, Mermelstein R. 1983. A global measure of perceived stress. J Health Soc Behav 24(4): 385–396. Cohen S, Hoberman HM. 1983. Positive events and social supports as buffers of life change stress1. J Appl Soc Psychol 13(2): 99–125. Csukly G, Czobor P, Szily E., Takacs B, Simon L. 2009. Facial expression recognition in depressed subjects: the impact of intensity level and arousal dimension. J Nerv Ment Dis 197(2): 98–103. Dailey MN, Cottrell GW, Padgett C, Adolphs R. 2002. EMPATH: a neural network that categorizes facial expressions. J Cogn Neurosci 14(8): 1158–1173. Fozard JL, Vercryssen M, Reynolds SL, Hancock PA, Quilter RE. 1994. Age differences and changes in reaction time: the Baltimore Longitudinal Study of Aging. J Gerontol 49(4): P179–89. Fu CH, Williams SC, Brammer MJ, et al. 2007. Neural responses to happy facial expressions in major depression following antidepressant treatment. Am J Psychiatry 164(4): 599–607. Gilboa-Schechtman E, Erhard-Weiss D, Jeczemien P. 2002. Interpersonal deficits meet cognitive biases: memory for facial expressions in depressed and anxious men and women. Psychiatry Res 113(3): 279–293. Gollan JK, Pane HT, McCloskey MS, Coccaro EF. 2008. Identifying differences in biased affective information processing in major depression. Psychiatry Res 159(1-2): 18–24. Gorlyn, M., Keilp, J.G., Grunebaum, M.F., et al. 2008. Neuropsychological characteristics as predictors of SSRI treatment response in depressed subjects. J Neural Transm (Vienna, Austria: 1996) 115(8): 1213–1219. Hamilton M. 1960. A rating scale for depression. J Neurol Neurosurg Psychiatry 23: 56–62. Hamilton M. 1959. The assessment of anxiety states by rating. Br J Med Psychol 32(1): 50–55. Harmer CJ, Bhagwagar Z, Perrett DI, et al. 2003. Acute SSRI administration affects the processing of social cues in healthy volunteers. Neuropsychopharmacol Off Publ Am Coll Neuropsychopharmacol 28(1): 148–152. Harmer CJ, Goodwin GM, Cowen PJ. 2009a. Why do antidepressants take so long to work? A cognitive neuropsychological model of antidepressant drug action. Br J Psychiatry J Ment Sci 195(2): 102–108. Harmer CJ, O’Sullivan U, Favaron E, et al. 2009b. Effect of acute antidepressant administration on negative affective bias in depressed patients. Am J Psychiatry 166(10): 1178–1184. Harmer CJ, Shelley NC, Cowen PJ, Goodwin GM. 2004. Increased positive versus negative affective perception and memory in healthy volunteers following
Copyright # 2014 John Wiley & Sons, Ltd.
1139 selective serotonin and norepinephrine reuptake inhibition. Am J Psychiatry 161(7): 1256–1263. Kan Y, Mimura M, Kamijima K, Kawamura M. 2004. Recognition of emotion from moving facial and prosodic stimuli in depressed patients. J Neurol Neurosurg Psychiatry 75(12): 1667–1671. Kohler CG, Turner TH, Bilker WB, et al. 2003. Facial emotion recognition in schizophrenia: intensity effects and error pattern. Am J Psychiatry 160(10): 1768–1774. Lawrence NS, Williams AM, Surguladze S, et al. 2004. Subcortical and ventral prefrontal cortical neural responses to facial expressions distinguish patients with bipolar disorder and major depression. Biol Psychiatry 55(6): 578–587. Leppanen JM, Milders M, Bell JS, Terriere E, Hietanen JK. 2004. Depression biases the recognition of emotionally neutral faces. Psychiatry Res 128(2): 123–133. Lotrich FE, Pollock BG. 2005. Aging and clinical pharmacology: implications for antidepressants. (THERAPEUTIC REVIEW). Malatesta CZ, Izard CE, Culver C, Nicolich M. 1987. Emotion communication skills in young, middle-aged, and older women. Psychol Aging 2(2): 193–203. Mikhailova ES, Vladimirova TV, Iznak AF, Tsusulkovskaya EJ, Sushko NV. 1996. Abnormal recognition of facial expression of emotions in depressed patients with major depression disorder and schizotypal personality disorder. Biol Psychiatry 40(8): 697–705. National Center for Health Statistics (US). 2011. Roiser JP, Elliott R, Sahakian BJ. 2012. Cognitive mechanisms of treatment in depression. Neuropsychopharmacol Off Publ Am Coll Neuropsychopharmacol 37(1): 117–136. Rubinow DR, Post RM. 1992. Impaired recognition of affect in facial expression in depressed patients. Biol Psychiatry 31: 9: 947–953. Rush AJ, Trivedi MH, Wisniewski SR, et al. 2006. Acute and longer-term outcomes in depressed outpatients requiring one or several treatment steps: a STAR*D report. Am J Psychiatry 163(11): 1905–1917. Surguladze SA, Young AW, Senior C, et al. 2004. Recognition accuracy and response bias to happy and sad facial expressions in patients with major depression. Neuropsychology 18(2): 212–218. Tranter R, Bell D, Gutting P, et al. 2009. The effect of serotonergic and noradrenergic antidepressants on face emotion processing in depressed patients. J Affect Disord 118(1-3): 87–93. Treisman A, Fearnley S. 1969. The Stroop test: selective attention to colours and words. Nature 222(5192): 437–439. Trivedi M, Rush AJ, Wisniewski S, et al. 2006. Evaluation of outcomes with citalopram for depression using measurement-based care in STAR*D: implications for clinical practice. Am J Psychiatry 163(1): 28–40. Warden D, Rush AJ, Trivedi MH, Fava M, Wisniewski SR. 2007. The STAR*D Project results: a comprehensive review of findings. Curr Psychiatry Rep 9(6): 449–459. Wechler D. 2008. Wechsler Adult Intelligence Scale (4th edn). Technical and interpretive manual. Pearson Assessment: San Antonio,TX. Wieser MJ, Muhlberger A, Kenntner-Mabiala R, Pauli P. 2006. Is emotion processing affected by advancing age? An event-related brain potential study. Brain Res 1096(1): 138–147. Wisniewski SR, Rush AJ, Balasubramani GK, et al. 2006. Self-rated global measure of the frequency, intensity, and burden of side effects. J Psychiatr Pract 12(2): 71–79. Wolf PA, D’Agostino RB, Belanger AJ, Kannel WB. 1991. Probability of stroke: a risk profile from the Framingham Study. Stroke 22(3): 312–318. Yesavage JA, Brink TL, Rose TL, et al. 1983. Development and validation of a geriatric depression screening scale: a preliminary report. J Psychiatr Res 17(1): 37–49. Yoon KL, Joormann J, Gotlib IH. 2009. Judging the intensity of facial expressions of emotion: depression-related biases in the processing of positive affect. J Abnorm Psychol 118(1): 223–228. Young AW, Perrett D, Calder AJ, Sprengelmeyer R, Ekman P (eds) 2002. Facial Expressions of Emotion Stimuli and Tests (FEEST). Thames Valley Test Company, Bury St. Edmunds: Suffolk, England.
Int J Geriatr Psychiatry 2014; 29: 1132–1139
Emotion recognition processing as early predictor of response to 8-week citalopram treatment in late-life depression Paulo R. Shiroma1,3, Paul Thuras1,3, Brian Johns2,3 and Kelvin O. Lim1,3 1
Mental Health Service Line, Minneapolis VA Medical Center, Minneapolis, MN, USA Department of Psychiatry, North Memorial Medical Center, Minneapolis, MN, USA 3 Department of Psychiatry, University of Minnesota Medical School, Minneapolis, MN, USA Correspondence to: P. R. Shiroma, E-mail: [email protected] 2
Background: In subjects with depression, exposure to antidepressants improves recognition of positive emotions. This phenomenon, which occurs early in the course of treatment, has been proposed as the initial step in the mechanism of action to subsequent therapeutic effects of antidepressants. To this date, it has not been well examined among older depressed patients. Method: Older subjects with non-psychotic major depressive disorder were treated with citalopram in an 8-week open-label study. The main predictor of response and remission was the change in emotion recognition between baseline and day 7. Covariates included executive functions, baseline anxiety level, medical comorbidity, level of subjective stress, serum citalopram level, and level of social support. Results: Twenty-seven patients were considered for final analysis. Overall, accuracy of emotion recognition significantly improved between baseline (75%) and day 7 (83%) (X2 = 34.50, df = 1, p < 0.001). Improvement to identify happy expressions occurred at 25% and 50% intensity with ceiling effect at 0%, 75%, and 100%. Change in emotion processing was marginally significant in predicting antidepressant response at day 56. Multivariate analysis showed that emotion processing is a significant predictor of response and remission when considered along with perceived level of social support. Conclusions: Recognition of mildly intense happy expression, which improved early in the course of citalopram treatment, predicts subsequent antidepressant response and remission when considered along with perception of social support. Further studies would be necessary to examine specific neural substrates in the affective network involved in the acute therapeutic action of antidepressant in late-life depression. Copyright # 2014 John Wiley & Sons, Ltd. Key words: depression; old age; older adults; citalopram; emotions History: Received 16 December 2013; Accepted 25 February 2014; Published online 4 April 2014 in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/gps.4104
Introduction Antidepressants are one of the most commonly prescribed medications in the USA (National Center for Health Statistics (US) 2011). One of the largest and longest studies ever carried out to evaluate depression treatment, the Sequenced Treatment Alternatives to Relieve Depression (STAR*D) study (Trivedi et al., 2006), showed that about 30% of patients achieved remission after first-line antidepressant citalopram. The STAR*D Copyright # 2014 John Wiley & Sons, Ltd.
study also suggested that any antidepressant or combination given late in treatment appears to work less effectively for attaining and sustaining remission than when given earlier in a sequence of antidepressant treatments (Rush et al., 2006; Warden et al., 2007). If treating clinicians knew the probability of response earlier in the course of treatment, they would be in a better position to advise patients about various treatment options, therefore limiting harmful neurobiological effects and poor outcome secondary to Int J Geriatr Psychiatry 2014; 29: 1132–1139
Emotion recognition in late-life depression
1133
enduring depressive symptoms. This situation could be particularly advantageous for the treatment of late-life depression (LLD) where treatment is often challenged by age-related sensitivity to medications and medical comorbidity (Lotrich and Pollock, 2005). Predictors of antidepressant drug response in LLD have rendered mixed results with research mainly concentrated in the role of executive dysfunction and the presence of deep white matter hyperintensities in T2-weighted brain magnetic resonance imaging. The study of emotion processing in the cause, maintenance, and relapse of affective disorders is largely limited to mid-life adult population (Rubinow and Post, 1992; Mikhailova et al., 1996; Bouhuys et al., 1999; Lawrence et al., 2004; Fu et al., 2007). Experiments in middle-aged healthy (Harmer et al., 2003; Harmer et al., 2004) and depressed subjects (Gorlyn et al., 2008; Harmer et al., 2009b) have shown improvement of emotion recognition biases after acute administration of antidepressant drugs (i.e., citalopram and reboxetine). Interestingly, increased accurate recognition of happy faces at 2 weeks of treatment predicts clinical outcome at 6 weeks of continued antidepressants (Tranter et al., 2009). Older adults with psychiatric conditions have reported less negative affect intensity than their younger counterparts, suggesting an interaction between aging and psychopathology on emotion processing (Cheavens et al., 2008). In this study, we investigated early changes in facial emotion recognition among older depressed subjects treated with citalopram as a predictor of subsequent response after 8 weeks of treatment.
Subjects with 17-item Hamilton Depression Rating Scale (HDRS17) score ≥18 were accepted in the study. Diagnoses of major depressive disorder, single or recurrent, were established by the Operational Criteria (OPCRIT) Diagnostic System (Azevedo et al., 1999). OPCRIT is a system of computer programs to be completed by trained clinicians from diagnostic interview and case records. OPCRIT generates diagnoses according to 12 operational diagnostic systems including the Diagnostic and Statistical Manual of Mental Disorders and the International Classification of Diseases. OPCRIT was performed on all participants by the principal investigator (P. R. S.). Exclusion criteria were as follows: Folstein mini mental status examination score ≤ 24; previous failure within the last year to at least two pharmacologically different antidepressants given for at least 6 weeks at the maximum tolerated dose; antidepressant exposure within 2 weeks (6 weeks for fluoxetine) prior to baseline assessment; and concurrent use of antipsychotics, mood stabilizers, or anticonvulsants. Patients were also excluded if they had substance use disorder within 6 months of assessment; positive urine toxicology screen test; history of psychosis, bipolar disorder, post-traumatic stress disorder, or acute stress disorder; electroconvulsive therapy within previous 6 months; Parkinson’s disease, dementia, seizures, or other central nervous system-related disorders; history of traumatic brain injury; and any clinically unstable medical illness or its treatment (e.g., interferon) that could compromise the patient’s ability to tolerate the study or likely interfere with the procedures or results.
Methods and materials
Rating scales and procedures
Participants
Baseline assessments included clinical and demographic information, 15-item Geriatric Depression Scale (GDS; Yesavage et al., n. d), Wechsler Adult Intelligence Scale—digit symbol substitution test (Wechler, 2008), and Stroop test (Treisman and Fearnley, 1969). Other covariates of antidepressant response included baseline anxiety (Hamilton Anxiety Scale; Hamilton, 1959), cardiovascular risk factors (cerebrovascular risk factor assessment [CRFA]; Wolf et al., 1991), level of social support (Interpersonal Support Evaluation List [ISEL]; Cohen and Hoberman, 1983), and severity of recent stress (4-item Perceived Stress Scale; Cohen et al., 1983). Subjects then received citalopram 10 mg/day for the first 7 days followed by 20 mg/day for the next 7 days. Any further dose adjustment was based on tolerability and clinical response by the treating clinician. Citalopram is recommended as the first-line treatment
Veterans aged 55 years and older were recruited from primary care and psychiatric outpatient clinics at the Minneapolis VA Medical Center over a 2-year period. Patients who received a prescription of citalopram for depression were identified through electronic notifications to the research team by the clinical pharmacist. Before initiating citalopram treatment, patients had a brief phone interview explaining the purpose of the study. Those patients interested and eligible to participate were invited for a face-to-face screen assessment. The Minneapolis VA Medical Center Institutional Review Board approved the study, and written informed consent was obtained from all subjects before participation. Participants received monetary compensation for their time committed to the study. Copyright # 2014 John Wiley & Sons, Ltd.
Int J Geriatr Psychiatry 2014; 29: 1132–1139
P. R. Shiroma et al.
1134
for LLD given its efficacy and tolerability compared with other antidepressants with less potential for drug–drug interactions (Alexopoulos et al., 2001). The effect of citalopram on the severity of depressive symptoms was measured by the HDRS17 (Hamilton, 1960) score at week 8, with response defined as a ≥50% decrease from the baseline score and remission defined as a final score ≤7. The percent decrease in HDRS17 and 15-item GDS rating scales was also examined. Prospective follow-up on antidepressant response was ascertained together with medication-related side effects (Frequency, Intensity, and Burden of Side Effects Rating Scale; Wisniewski et al., 2006) at days 3, 10, 14, 21, 28, 35, 42, and 56. At day 7, trough citalopram serum concentration (MedTox Scientific, Inc., St. Paul, MN) was obtained as a covariate of change in emotion recognition. DNA extraction for genetic analysis was also obtained (not reported here). Facial emotion recognition task
Facial emotion recognition task (FERT) was administered prior to the onset of citalopram treatment and at day 7. Facial images were taken from the Facial Expressions of Emotion: Stimuli and Tests (Young et al., 2002) and presented by Superlab Pro on a laptop computer in a quiet room to minimize distracting noise. Written instructions about the purpose and procedures of the test were presented on the laptop screen followed by a brief practice session. A research staff was present during sessions to answer questions. FERT was delivered in two blocks. Block #1 was composed of facial images (three men and three women) depicting 12 neutral expressions and 48 happy expressions. Faces were randomly presented at 25%, 50%, 75%, and 100% of emotional intensity and at 500- or 5000-ms duration of exposure. The intensity of emotions and duration of exposures were counterbalanced in each session. Each facial image was followed by a blank screen with a white crosshair in the center to prompt response. Although no time limit to response was imposed, subjects were asked to identify as quickly and as accurately as possible whether the facial expression was happy or not by pressing a labeled key on a response box (DirectIN High Speed), which was automatically coded as accurate or inaccurate identification. A new fresh image was shown 2 s after the participant’s response. Block #2 followed the same randomized sequence of images as the first block to control for practice effect. Following completion of FERT, participants had a gender recognition task (male or female). Thirty male Copyright # 2014 John Wiley & Sons, Ltd.
and 30 female faces depicting 48 emotional expressions and 12 neutral expressions were randomly presented following the same procedures as FERT. This task served as an active experimental control for aspects of facial perception unrelated to emotion recognition such as visual and attentional processing. Statistical analysis Statistical analyses were performed using SPSS v. 19. Chi-square statistics were used to examine differences between differences in accuracy on the recognition task (e.g., duration, block order, and intensity). Change in accuracy by intensity was examined using analysis of variance. A mixed-effects regression model was used to examine change in symptoms over time. Multivariate logistic regression analysis was used to examine predictors of remission and response for the indicators of depressive symptoms (HDRS17 and GDS). Multiple regression analyses were used to test for interaction effects in predictors of symptom change. The level of significance (alpha) was set at 0.05. Results Sixty-two subjects were screened; 32 were excluded (n = 22) or refused to participate (n = 10). Demographic and clinical characteristics of the study sample (n = 30) are presented in Table 1. The sample was composed only of men in late adulthood, mostly Caucasians, unemployed, and with high school or higher education. Four subjects were hospitalized for depression, and one subject had attempted suicide. The average onset of the first major depressive episode occurred at middle age. Major depressive disorder was mostly recurrent, with current episode prolonged by several weeks, moderate to severe in intensity, and comorbid with moderate level of anxiety. The sample had a probability of stroke in 10 years of 11.2% based on the CRFA (range 2.6–87.9%), similar to a previous study of LLD (Alexopoulos et al., 2009). Twenty-seven subjects considered for analyses completed FERT at baseline and at day 7, had detectable serum level of citalopram, and had at least one HDRS17 measure at or after day 10. The overall response and remission rates at day 56 for these 27 subjects were 46.6% and 40.0%, respectively. There was a significant reduction in HDRS17 and GDS scores during treatment (F = 34.07, df = 9, p < 0.0001 and F = 23.803, df = 9, p < 0.0001, respectively). Int J Geriatr Psychiatry 2014; 29: 1132–1139
Emotion recognition in late-life depression
1135
Table 1 Demographic and clinical characteristics of study sample Age (years), mean (SD) Age (years) at first MDE, mean (SD) Family history of mental illness, no. Mood disorder Other psychiatric disorder Past history of substance use, no. Alcohol Illicit drugs Duration of current episode (months), mean (SD) At least one previous antidepressant trial during current episode, no. Yes No Recurrent MDD, no. Yes No HAM-A at baseline, mean (SD) HDRS17, mean (SD) GDS15, mean (SD) MMSE, mean (SD) Color-Word Stroop, mean (SD) Serum citalopram level at day 7 (ng/mL), mean (SD) ISEL score at baseline (range = 0–36), mean (SD) PSS-4 score at baseline (range = 0–12), mean (SD) CRFA score at baseline, mean (SD)
65.3 (7.0) 49.4 (19.8) 9 15 11 6 26.2 (26.3) 18 12 19 11 19.0 (5.4) 23.4 (3.3) 11.4 (1.9) 28.0 (1.28) 28.9 (5.9) 28.1 (16.9) 19.8 (7.3) 9.1 (2.0) 11.4 (4.7)
MDE, major depressive episode; MDD, major depressive disorder; HAM-A, Hamilton Anxiety Scale; HDRS17, 17-item Hamilton Depression Rating Scale; GDS15, 15-item Geriatric Depression Scale; MMSE, mini mental state examination; ISEL, Interpersonal Support Evaluation List; PSS-4, 4-item version of Perceived Stress Scale; CRFA, cerebrovascular risk factor assessment.
Overall, accuracy of emotion recognition significantly improved between baseline (75%) and day 7 (83%) (X2 = 34.50, df = 1, p < 0.001). We found nonsignificant changes in accuracy between blocks #1 (75%) and #2 (76%) (X2 = 0.24, df = 1, p = 0.62) at baseline or at day 7 (block #1 = 83.1% and block #2 = 84.2%, X2 = 0.36, df = 1, p = 0.54). As we found no practice effect over identification of facial expressions, all subsequent analyses combined the results from blocks #1 and #2 as a single trial.
We analyzed the effect of intensity of stimuli and length of exposure over the accuracy to identify emotions. We found non-significant association between accuracy of responses and the length of exposure of each facial stimuli (X2 = 0.37, df = 1, p = 0.54). Improvement in FEST performance varied by the intensity of facial expressions (Figure 1). At 25% intensity, accuracy changed from 22% to 40% (t = 3.53, df = 24, p = 0.002); at 50% intensity, accuracy improved from 69% to 89% (t = 5.00, df = 24, p < 0.001). Minimum improvement in accuracy occurred at 0%, 75%, and 100% (F(1, 74) = 1.14, p = 0.29). Thus, all subsequent analyses considered the change of emotion recognition as the combined accuracy responses from 25% and 50% stimuli intensity. There was a significant difference in reaction time from baseline (mean = 606.63 ms, SD = 33.68) to day 7 (mean = 581.23 ms, SD = 34.15; t = 2.7512, df = 52, p = 0.008). As subjects were not limited on the time to response, we also analyzed whether there was any speed–accuracy trade-offs during FERT, meaning higher accuracy achieved with longer time to respond. We found that reaction time had a significant negative correlation with emotion recognition (r = 0.46, p = 0.01) where greater accuracy correlated with faster reaction time. Thus, no speed–accuracy trade-off was found. Serum level of citalopram measured at day 7 shows a non-significant correlation with changes in both emotion recognition (r = 0.21, p = 0.31) or HDRS17 score (r = 0.18, p = 0.45) from baseline to day 7. We analyzed changes in accuracy of gender recognition to control for visual and attentional processing. We found a significant but small increase in gender recognition from baseline (94%) to day 7 (97%) (X2 = 12.00, df = 1, p < 0.001, φ = 0.084). A non-significant correlation was found between gender recognition task and FERT (r = 0.24, p = 0.35).
Figure 1 Change in the accuracy of emotion recognition from baseline to day 7 by the intensity of facial stimuli.
Copyright # 2014 John Wiley & Sons, Ltd.
Int J Geriatr Psychiatry 2014; 29: 1132–1139
P. R. Shiroma et al.
1136
We conducted univariate and multivariate analyses with change in FEST treated both as a categorical and as a continuous variable. The categorization of accuracy change produces easily understandable results and facilitates the study of interaction effects. Analyses of accuracy change as a continuous variable decreases the likelihood to lose information between variables because of categorization. Univariate analysis showed improved accuracy as a marginally significant predictor of antidepressant response at day 56 (weight coefficient = 0.043; SE = 0.024; Wald X 2df ¼1 ¼ 3:04, p = 0.08; odds ratio = 1.044; 95% CI = 0.995–1.095) where a 10% improvement in FEST by day 7 increases the likelihood of response by 44%. Similar results were found for predicting remission at day 56 (weight coefficient = 0.046; SE = 0.025; Wald X 2df ¼1 = 3.41, p = 0.065; odds ratio = 1.047; 95% CI = 0.997–1.100) where a 10% improvement in FEST by day 7 increases the likelihood of response by 47%. Covariates including HDRS17 baseline score, accuracy in gender recognition, age, Hamilton Anxiety Scale score, Stroop test, digit symbol substitution test, citalopram serum level, ISEL, Perceived Stress Scale, CRFA, and mini mental status examination were non-significant predictors of response or remission (data not shown). Figure 2 shows the study sample grouped in tertiles on the basis of the percentage of improvement recognition from baseline to day 7. We found a significant association with change in HDRS17 scores over time (F = 3.58, df = 2, p = 0.045). Subjects with no improvement had a mean HDRS17 score decreased by 5 points during the course of treatment. Subjects with improved accuracy from baseline up to 25% and those with >25% had a decrease in HDRS17 score of 15 and 14 points at the study end, respectively.
We constructed a multivariate logistic regression model to predict response at day 56. Independent effect of change in emotion recognition (FEST), perceived social support (ISEL) at baseline, and an interaction term of FEST by ISEL were included. Analysis showed that change in emotion recognition by day 7 and social support became significant predictors of response (weight coefficient = 0.072; SE = 0.033; Wald X 2df ¼1 = 4.68, p = 0.03; odds ratio = 1.075; 95% CI = 1.007–1.148 for emotion recognition change; weight coefficient = 0.165; SE = 0.083; Wald X 2df ¼1 = 3.92, p = 0.048; odds ratio = 1.179; 95% CI = 1.002–1.388 for social support). This was also observed in the model predicting depression remission (weight coefficient = 0.082; SE = 0.036; Wald X 2df ¼1 = 5.17, p = 0.023; odds ratio = 1.085; 95% CI = 1.011–1.164 for accuracy change; weight coefficient = 0.181; SE = 0.088; Wald X 2df ¼1 = 4.25, p = 0.039; odds ratio = 1.199; 95% CI = 1.009–1.424 for social support). The interaction of emotion recognition and social support did not meet significance for response (p = 0.088) or remission (p = 0.071). When categorization of emotion recognition was entered, subjects with improved accuracy >25% from baseline had the highest likelihood to reduce HDRS17 scores over time compared with those with no improvement adjusted for level of social support. ISEL scores also became a significant predictor of change in depressive symptoms adjusted for emotion recognition. A multiple regression analysis found that the interaction between change in FEST and ISEL was significant (β = 0.507, t = 3.38, p = 0.003; Figure 3). Exploratory analysis revealed a significant change in HDRS17 score by day 7 (F(2, 25) = 7.04, p = 0.004). However, decreased scores in HDRS17 did not significantly correlate with FEST improvement (p = 0.14) or
HDRS17
Days of Treatment with Citalopram Figure 2 Repeated Hamilton Depression Rating Scale (HDRS17) score over time by groups based on the percentage of improvement recognition of happy facial expressions from baseline to day 7.
Copyright # 2014 John Wiley & Sons, Ltd.
Int J Geriatr Psychiatry 2014; 29: 1132–1139
Emotion recognition in late-life depression
1137
Negative or no improvement
25% improvement
Figure 3 Effect of categorization of emotion recognition task by perception of social support (Interpersonal Support Evaluation List [ISEL]) at baseline over likelihood of change in depression severity measured by Hamilton Depression Rating Scale (HDRS).
became a significant predictor of response (p = 0.59) or remission (p = 0.31) by day 56. Using mixed regression models, we also examined the impact of change in emotion recognition on reduction of depressive symptoms during different segments from 7 days post-initiation of citalopram. We found that the effect was maximized from 7 to 14 days (F(1, 52.3) = 9.47, p = 0.003) but was no longer significant for the segment from 14 to 21 days (F(1, 25.6) = 1.21, p = 0.281). Discussion Our results extend to older adults the growing evidence that antidepressants exert a change in emotion processing prior to the improvement of mood symptoms as we found an increased recognition of facial expression of happiness as early as 1 week of continued treatment with citalopram. Moreover, emotion recognition along with perceived social support became significant predictors of subsequent antidepressant response and remission. Tranter and colleagues (Tranter et al., 2009) found a significant correlation between accurate recognition of happy faces at 2 weeks of treatment and clinical outcome at 6 weeks of treatment. The authors suggested that early changes in emotional processing may underlie clinical response to antidepressants; however, the emotion task performed at 2 weeks opened the possibility that changes in emotion recognition might be secondary to improvements in mood rather than direct antidepressant effect. Our study addressed this Copyright # 2014 John Wiley & Sons, Ltd.
issue by conducting an emotional task at baseline and then at 1 week of treatment. Although both depression scores and gender recognition improved at 1 week of treatment, they did not correlate with changes in emotion recognition nor were predictors of response or remission at the study end. These findings further support the hypothesis that early change in emotion recognition due to antidepressants is an unrelated phenomenon to symptomatic or visual/ attention processing. According to this view, Harmer and colleagues (Harmer et al., 2009a) have proposed that the effects of antidepressants on the processing of facial expressions are seen rapidly, but the translation into improved subjective mood takes time as the patient learns to interpret social signals as positive, leading to social participation, which eventually could improve mood symptoms of depression. In contrast, inaccurate recognition of facial emotions would lead to appraise social situations more negatively, minimize social interaction, and even elicit rejection by others perpetuating a depressive state. A “cognitive neuropsychological” model of depression (Roiser et al., 2012) posits that negative emotion processing biases play a central causal role in the development of depression and may even be possible to predict which patients will benefit most from which treatments on the basis of neural responses to negative stimuli. In our study, improved recognition of happy expressions and higher social support at baseline increased the likelihood of better antidepressant outcomes. It is conceivable that corrected emotion processing mechanisms Int J Geriatr Psychiatry 2014; 29: 1132–1139
P. R. Shiroma et al.
1138
may exert therapeutic action by acting upon a supportive social network. In fact, age-related changes in the ability to interpret the feelings of others influenced the quality of social relationships in later life (Malatesta et al., 1987), and the subjective social support was reported as an important clinical factor in predicting shorter time to remission among geriatric depressed patients (Bosworth et al., 2002). Depression may be characterized primarily by difficulties in the processing of positive affect (Surguladze et al., 2004). Neuroimaging studies identified a low activation in subcortical and limbic regions in response to happy faces (Lawrence et al., 2004; Fu et al., 2007), and depression has been associated with decreased memory for happy faces (Gilboa-Schechtman et al., 2002). However, other studies found no deficits in depressed patients (Kan et al., 2004; Leppanen et al., 2004; Gollan et al., 2008). These discrepant results could be in part due to methodological inconsistencies that include the use of different facial stimulus sets and varied emotion processing tasks. The numbers of choices available in several cognitive tasks have demonstrated to influence older adult’s performance (Fozard et al., 1994). For instance, choosing one label from six overlapping constructs may tap cognitive rather than emotion processing (Wieser et al., 2006). In this study, the task to identify facial expressions required a simple YES/NO labeling of target emotion (e.g., happy face), which avoids high demands on cognitive effort but remained sensitive to detect changes in emotion recognition. The emotional intensity of the facial expression is also a factor to consider when measuring emotion recognition deficits (Kohler et al., 2003). Although depressed subjects may require significantly greater intensity to correctly identify happy expression compared with controls (Yoon et al., 2009), happiness is an easier emotion to detect than other emotions because of the presence of a smile (Dailey et al., 2002). Consistent with previous studies (Surguladze et al., 2004; Csukly et al., 2009), we found that moderately intense but still recognizable emotional stimuli captured the changes in emotion recognition, whereas neutral or high-intensity stimuli may result in a ceiling effect. Our study has several limitations. The open-label design, small sample size, and lack of placebo group limit stringent analysis of clinical factors previously reported to affect antidepressant response in LLD (e.g., executive dysfunction and severity of anxiety). However, the current study did not intend to test the efficacy of citalopram per se but rather to investigate the early changes in emotion recognition as predictor of subsequent antidepressant outcomes. Concurrent Copyright # 2014 John Wiley & Sons, Ltd.
interventions such as psychological treatment were not controlled and could have influenced response, remission, and even improvement in emotion recognition. Despite the small sample size, the current report extends previous finding in younger population to older adults on how antidepressants exert a therapeutic action through the modification of affective processing. Finally, it is unknown whether an antidepressant with a different pharmacological profile than citalopram would have modified facial recognition in a similar way. Tranter and colleagues (Tranter et al., 2009) did not find a significant difference in facial emotion recognition between citalopram (serotonergic) and reboxetine (noradrenergic), suggesting that in younger populations, antidepressants do not exert differential effects on emotion processing. In conclusion, accuracy recognition of moderately intense happy expression improved after 1 week of treatment with citalopram and was a significant predictor of later response and remission in LLD. Whether emotion processing in depression is a necessary initial step of improvement in mood symptom or an epiphenomenon during a depressive episode is unclear. Future studies should include development of interventions to target and improve emotion processing, which ultimately may enhance depression treatment. Emotion activation tasks in neuroimaging studies could also identify specific circuits in the affective network that relate to the early changes in emotion recognition. Conflict of interest None declared. Key points
• •
Recognition of mildly intense happy expression improved early in the course of citalopram treatment among older adults with depression. Improvement in emotion recognition along with perception of social support predicts subsequent antidepressant response and remission.
Acknowledgements Dr Shiroma was supported by the Minneapolis VA Center for Epidemiological and Clinical Research (CECR), a VA Clinical Research Center of Excellence, and the Mental Health Service Line, Minneapolis VA Medical Center. Int J Geriatr Psychiatry 2014; 29: 1132–1139
Emotion recognition in late-life depression
References Alexopoulos GS, Katz IR, Reynolds CF, 3rd, et al. 2001. Pharmacotherapy of depression in older patients: a summary of the expert consensus guidelines. J Psychiatr Pract 7(6): 361–376. Alexopoulos GS, Murphy CF, Gunning-Dixon FM, et al. 2009. Serotonin transporter polymorphisms, microstructural white matter abnormalities and remission of geriatric depression. J Affect Disord 119(1-3): 132–141. Azevedo MH, Soares MJ, Coelho I, et al. 1999. Using consensus OPCRIT diagnoses. An efficient procedure for best-estimate lifetime diagnoses. Br J Psychiatry J Ment Sci 175: 154–157. Bosworth HB, McQuoid DR, George LK, Steffens DC. 2002. Time-to-remission from geriatric depression: psychosocial and clinical factors. Am J Geriatr Psychiatry Off J Am Assoc Geriatr Psychiatry 10(5): 551–559. Bouhuys AL, Geerts E, Gordijn MC. 1999. Depressed patients’ perceptions of facial emotions in depressed and remitted states are associated with relapse: a longitudinal study. J Nerv Ment Dis 187(10): 595–602. Cheavens JS, Zachary Rosenthal M, Banawan SF, Lynch TR. 2008. Differences in emotional experience and emotion regulation as a function of age and psychiatric condition. Aging Ment Health 12(4): 478–487. Cohen S, Kamarck T, Mermelstein R. 1983. A global measure of perceived stress. J Health Soc Behav 24(4): 385–396. Cohen S, Hoberman HM. 1983. Positive events and social supports as buffers of life change stress1. J Appl Soc Psychol 13(2): 99–125. Csukly G, Czobor P, Szily E., Takacs B, Simon L. 2009. Facial expression recognition in depressed subjects: the impact of intensity level and arousal dimension. J Nerv Ment Dis 197(2): 98–103. Dailey MN, Cottrell GW, Padgett C, Adolphs R. 2002. EMPATH: a neural network that categorizes facial expressions. J Cogn Neurosci 14(8): 1158–1173. Fozard JL, Vercryssen M, Reynolds SL, Hancock PA, Quilter RE. 1994. Age differences and changes in reaction time: the Baltimore Longitudinal Study of Aging. J Gerontol 49(4): P179–89. Fu CH, Williams SC, Brammer MJ, et al. 2007. Neural responses to happy facial expressions in major depression following antidepressant treatment. Am J Psychiatry 164(4): 599–607. Gilboa-Schechtman E, Erhard-Weiss D, Jeczemien P. 2002. Interpersonal deficits meet cognitive biases: memory for facial expressions in depressed and anxious men and women. Psychiatry Res 113(3): 279–293. Gollan JK, Pane HT, McCloskey MS, Coccaro EF. 2008. Identifying differences in biased affective information processing in major depression. Psychiatry Res 159(1-2): 18–24. Gorlyn, M., Keilp, J.G., Grunebaum, M.F., et al. 2008. Neuropsychological characteristics as predictors of SSRI treatment response in depressed subjects. J Neural Transm (Vienna, Austria: 1996) 115(8): 1213–1219. Hamilton M. 1960. A rating scale for depression. J Neurol Neurosurg Psychiatry 23: 56–62. Hamilton M. 1959. The assessment of anxiety states by rating. Br J Med Psychol 32(1): 50–55. Harmer CJ, Bhagwagar Z, Perrett DI, et al. 2003. Acute SSRI administration affects the processing of social cues in healthy volunteers. Neuropsychopharmacol Off Publ Am Coll Neuropsychopharmacol 28(1): 148–152. Harmer CJ, Goodwin GM, Cowen PJ. 2009a. Why do antidepressants take so long to work? A cognitive neuropsychological model of antidepressant drug action. Br J Psychiatry J Ment Sci 195(2): 102–108. Harmer CJ, O’Sullivan U, Favaron E, et al. 2009b. Effect of acute antidepressant administration on negative affective bias in depressed patients. Am J Psychiatry 166(10): 1178–1184. Harmer CJ, Shelley NC, Cowen PJ, Goodwin GM. 2004. Increased positive versus negative affective perception and memory in healthy volunteers following
Copyright # 2014 John Wiley & Sons, Ltd.
1139 selective serotonin and norepinephrine reuptake inhibition. Am J Psychiatry 161(7): 1256–1263. Kan Y, Mimura M, Kamijima K, Kawamura M. 2004. Recognition of emotion from moving facial and prosodic stimuli in depressed patients. J Neurol Neurosurg Psychiatry 75(12): 1667–1671. Kohler CG, Turner TH, Bilker WB, et al. 2003. Facial emotion recognition in schizophrenia: intensity effects and error pattern. Am J Psychiatry 160(10): 1768–1774. Lawrence NS, Williams AM, Surguladze S, et al. 2004. Subcortical and ventral prefrontal cortical neural responses to facial expressions distinguish patients with bipolar disorder and major depression. Biol Psychiatry 55(6): 578–587. Leppanen JM, Milders M, Bell JS, Terriere E, Hietanen JK. 2004. Depression biases the recognition of emotionally neutral faces. Psychiatry Res 128(2): 123–133. Lotrich FE, Pollock BG. 2005. Aging and clinical pharmacology: implications for antidepressants. (THERAPEUTIC REVIEW). Malatesta CZ, Izard CE, Culver C, Nicolich M. 1987. Emotion communication skills in young, middle-aged, and older women. Psychol Aging 2(2): 193–203. Mikhailova ES, Vladimirova TV, Iznak AF, Tsusulkovskaya EJ, Sushko NV. 1996. Abnormal recognition of facial expression of emotions in depressed patients with major depression disorder and schizotypal personality disorder. Biol Psychiatry 40(8): 697–705. National Center for Health Statistics (US). 2011. Roiser JP, Elliott R, Sahakian BJ. 2012. Cognitive mechanisms of treatment in depression. Neuropsychopharmacol Off Publ Am Coll Neuropsychopharmacol 37(1): 117–136. Rubinow DR, Post RM. 1992. Impaired recognition of affect in facial expression in depressed patients. Biol Psychiatry 31: 9: 947–953. Rush AJ, Trivedi MH, Wisniewski SR, et al. 2006. Acute and longer-term outcomes in depressed outpatients requiring one or several treatment steps: a STAR*D report. Am J Psychiatry 163(11): 1905–1917. Surguladze SA, Young AW, Senior C, et al. 2004. Recognition accuracy and response bias to happy and sad facial expressions in patients with major depression. Neuropsychology 18(2): 212–218. Tranter R, Bell D, Gutting P, et al. 2009. The effect of serotonergic and noradrenergic antidepressants on face emotion processing in depressed patients. J Affect Disord 118(1-3): 87–93. Treisman A, Fearnley S. 1969. The Stroop test: selective attention to colours and words. Nature 222(5192): 437–439. Trivedi M, Rush AJ, Wisniewski S, et al. 2006. Evaluation of outcomes with citalopram for depression using measurement-based care in STAR*D: implications for clinical practice. Am J Psychiatry 163(1): 28–40. Warden D, Rush AJ, Trivedi MH, Fava M, Wisniewski SR. 2007. The STAR*D Project results: a comprehensive review of findings. Curr Psychiatry Rep 9(6): 449–459. Wechler D. 2008. Wechsler Adult Intelligence Scale (4th edn). Technical and interpretive manual. Pearson Assessment: San Antonio,TX. Wieser MJ, Muhlberger A, Kenntner-Mabiala R, Pauli P. 2006. Is emotion processing affected by advancing age? An event-related brain potential study. Brain Res 1096(1): 138–147. Wisniewski SR, Rush AJ, Balasubramani GK, et al. 2006. Self-rated global measure of the frequency, intensity, and burden of side effects. J Psychiatr Pract 12(2): 71–79. Wolf PA, D’Agostino RB, Belanger AJ, Kannel WB. 1991. Probability of stroke: a risk profile from the Framingham Study. Stroke 22(3): 312–318. Yesavage JA, Brink TL, Rose TL, et al. 1983. Development and validation of a geriatric depression screening scale: a preliminary report. J Psychiatr Res 17(1): 37–49. Yoon KL, Joormann J, Gotlib IH. 2009. Judging the intensity of facial expressions of emotion: depression-related biases in the processing of positive affect. J Abnorm Psychol 118(1): 223–228. Young AW, Perrett D, Calder AJ, Sprengelmeyer R, Ekman P (eds) 2002. Facial Expressions of Emotion Stimuli and Tests (FEEST). Thames Valley Test Company, Bury St. Edmunds: Suffolk, England.
Int J Geriatr Psychiatry 2014; 29: 1132–1139
