Neural Response to Pleasant Pictures Moderates Prospective Relationship Between Stress and Depressive Symptoms in Adolescent Girls

Abstract

Adolescent girls are at increased risk for depression, which is thought to result from the interaction of biological vulnerabilities and life stressors common to adolescent girls. A blunted late positive potential (LPP) to emotional stimuli (i.e., pleasant and unpleasant) has been associated with depressive symptoms and risk. The current study of adolescent girls examines the moderating effects of the LPP, a candidate biomarker of depression, of the link between life stress and increases in depressive symptoms over 1 year. We measured LPP to pleasant and unpleasant pictures from the International Affective Picture Set among 143 predominantly Caucasian adolescent girls ages 8 to 14, who also reported on the frequency of common life stressors. Self-reported depressive symptoms were assessed both at baseline and 1 year after the initial lab visit. The LPP to pleasant pictures moderated the relationship between baseline life stressors and the change in depressive symptoms. Specifically, life stress was associated with increases in depressive symptoms when the LPP to pleasant pictures was blunted, whereas life stress was associated with decreases in depressive symptoms when the LPP to pleasant pictures was potentiated. These effects showed some specificity to family and school-related stressors and to anhedonic and efficacy-related depressive symptoms. A similar pattern, though not statistically significant, was found for the LPP to unpleasant pictures. Together, these findings suggest that the LPP to pleasant pictures may represent a useful biomarker in identifying individuals at greatest risk of experiencing depressive symptoms following stress.

Notes

¹ Studies that examine the LPP to semantic self-relevant information using the Self-Referential Encoding Task find that depression and depression risk factors are associated with a blunted LPP to positive words and a potentiated LPP to negative words (Auerbach et al., 2015; Speed et al., 2016; Webb et al., 2017). This suggests that the link between depression and the LPP to self-referential information would be better understood using a positivity attenuation or negative potentiation model. However, the current study focuses on the LPP elicited by nomothetic emotional scenes.

² In a review of depression measures for children and adolescents, Klein and colleagues reported that, across multiple studies, the CDI has good reported internal consistency, generally good reported test–retest reliability, strong concurrent validity when compared to other common measures (e.g., Children’s Depression Rating Scale) and is sensitive to change in symptoms as in a treatment study (Klein, Dougherty, & Olino, 2005). However, this review also notes that the CDI factor loadings on the CDI subscales vary as a function of age, indicating that the CDI subscales may not be consistent across all samples. Similarly, the factor structure of the CDI has also been found to vary between child and parent report (Cole, Hoffman, Tram, & Maxwell, 2000).

³ The following IAPS pictures were used in this task. Unpleasant: 1050, 1052, 6571, 1205, 1200, 1300, 1304, 1930, 2458, 9600, 2691, 2703, 2800, 2811, 2900, 3022, 6190, 6213, 6231, 6510. Pleasant: 1462, 1710, 1750, 1811, 2070, 2091, 2092, 2224, 2340, 2345, 2347, 7325, 8031, 8200, 8461, 8496, 8497, 8370, 7400, 7330. Neutral: 5395, 7026, 7130, 7190, 7175, 2514, 7038, 2580, 5390, 7090, 5500, 5731, 5740, 7100, 5900, 7000, 7002, 7009, 7010, 7039. The means and standard deviations of IAPS normative arousal scores for the selected pictures are as follows: unpleasant = 6.36 ± 0.55, pleasant = 5.22 ± 0.82, and neutral = 3.03 ± 0.63 (Lang et al., 2008).

⁴ Because the LPP to pleasant minus neutral appeared to be of greatest magnitude on either side of the midline, we repeated moderation analyses using the residual LPP to pleasant pictures calculated as the averaged amplitude at left and right lateral sites P3, PO3, P4, and PO4. These analyses replicated the patterns we found when calculating the LPP to pleasant at midline site Pz.

⁵ To better examine what aspects of our model contribute to the current results, we conducted a series of supplemental analyses, as follows: To address concerns about the confounding effects of age, we conducted exploratory analyses, replicating our core moderation model with age as a covariate. With the inclusion of this age covariate, the LPP to pleasant pictures continues to significantly interact with baseline ALEQ to predict residualized CDI at follow-up, F(1, 138) = 6.86, p = .01, and the same interaction using the LPP to unpleasant pictures now reaches statistical significance, F(1, 138) = 4.19, p = .04. Furthermore, we analyzed the bivariate associations between age and our model variables. We find that age is significantly positively correlated with CDI depression scores and ALEQ life stress scores. We also find that age is significantly negatively correlated with the LPP to negative pictures but is not significantly related to the LPP to positive pictures. Finally, we conducted the recommended three-way interaction models, adding age as a moderator of the original core moderation models. For the LPP to positive pictures, we find that age does not significantly moderate the effects of either core moderation model (i.e., age does not significantly interact with the LPP, ALEQ stressors, or their interaction term). For the LPP to negative pictures, we find that there is a significant three-way interaction between the LPP, ALEQ stressors, and age such that the interactive effects of the LPP and life stressors in predicting depression were significant only for participants closest the mean age of the sample. This analysis is exploratory; however, one possible interpretation is there may be a sensitive developmental period around the mean age of our sample (12.90 years old) during which a blunted LPP to negative pictures conveys more risk for stress-induced depression. Anxiety has been associated with LPP magnitude (MacNamara, Ferri, & Hajcak, 2011), and thus anxiety symptoms may present a confound to the current analyses. To examine this potential confounding effect, we repeated our core moderation models with baseline anxiety symptoms, as measured by the total child self-reported anxiety at baseline using the Screen for Child Anxiety Related Disorders (Birmaher et al., 1997). With the inclusion of this covariate, the LPP to pleasant pictures continues to significantly interact with baseline ALEQ to predict residualized CDI at follow-up, F(1, 138) = 7.14, p = .01, and the same interaction using the LPP to unpleasant pictures is trending toward, but does not achieve, statistical significance, F(1, 138) = 3.53, p = .06. It is common to divide the LPP into early (~300–600 ms) and late (~600–1,000ms) components, as these have been suggested to represent separable components (Foti, Hajcak, & Dien, 2009). To address the question of whether there was a significant effect of early (300–600 ms) versus late (600–1,000 ms) LPP, we conducted a supplemental replication our core moderation analyses separately for each of these time windows. The early LPP to pleasant pictures continues to significantly interact with baseline ALEQ to predict residualized CDI at follow-up, F(1, 139) = 5.88, p = .02, and the same interaction using the early LPP to unpleasant pictures approaches but does not reach statistical significance, F(1, 139) = 3.27, p = .07. The late LPP to pleasant pictures also continues to significantly interact with baseline ALEQ to predict residualized CDI at follow-up, F(1, 139) = 6.66, p = .01, and the same interaction using the late LPP to unpleasant pictures now reaches statistical significance, F(1, 139) = 4.01, p = .05. To understand whether the interaction effect with life stress was driven by the response to emotional or neutral pictures, we repeated out analyses, including each interaction term separately. These analyses indicate that the LPP to the emotional picture and not the neutral picture drives the interaction effect with ALEQ to predict depressive symptoms. Using pleasant pictures, ALEQ × LPP to pleasant pictures significantly predicts CDI (β = −0.39, p = .02), whereas ALEQ × LPP to neutral pictures does not (β = 0.17, p = .28). Similarly, using unpleasant pictures, ALEQ × LPP to unpleasant pictures is trending toward significance (β = −0.30, p = .10), whereas ALEQ × LPP to neutral pictures does not approach significance (β = 0.10, p = .57). To assess whether behavioral measures of interference predict similar outcomes as the LPP, we replicated our core moderation models, substituting in reaction time (RT) measures of interference (i.e., residualized RT on pleasant trials adjusting for RT on neutral trials, residualized RT on unpleasant trials adjusting for RT on neutral trials) as the moderator in place of the LPP. These models found that moderating effects of interference on pleasant picture trials approach but do not reach statistical significance, F(1, 110) = −2.72, p = .10. Moderating effects of Interference × Unpleasant Pictures does not significantly moderate this relationship, F(1, 110) = 0.11, p = .74. These analyses suggest that the moderating effects of the LPP cannot be comparably indexed using behavioral measures of emotional interference, though interference on pleasant picture trials trends in the same direction.

Additional information

Funding

This work was supported by the National Institute of Mental Health [5R01MH097767-04].