Negative Cognitive Style Interacts with Negative Life Events to Predict First Onset of a Major Depressive Episode in Adolescence via Hopelessness

Abstract

The hopelessness theory of depression is a prominent account of depression that posits that individuals with a negative inferential style are more likely to become hopeless when they experience negative life events (NLEs), and that hopelessness is a proximal cause of depression. There is strong evidence supporting the role of a negative inferential style in the pathogenesis of major depression; however, substantially less is known about the proposed role played by hopelessness. The cornerstone hypothesis of hopelessness theory, that hopelessness is a proximal cause of major depression, is untested. Similarly, a small number of studies have generated inconclusive evidence that hopelessness mediates the relationship between a negative inferential style, NLEs, and depressive symptoms. The current study tested whether hopelessness mediates the relationship between a negative inferential style-NLEs interaction and (i) first onset of a major depressive episode (MDE) and (ii) depressive symptoms in a fully prospective design. A diverse sample of 249 adolescents, aged 12–13 years, were assessed at baseline and at two or more follow-ups over approximately 2.5 years. Self-report as well as life event and diagnostic interviews assessed inferential style, NLEs, hopelessness, depressive symptoms and diagnosis. Moderated mediation analyses indicated that hopelessness mediated the relationship between a negative inferential style-NLEs interaction and (a) first onset of a MDE as well as (b) depressive symptoms at higher levels of multiple types of NLEs. The current study demonstrates the validity of the hopelessness theory of depression and its continued clinical relevance in predicting depression in adolescence.

Introduction

Depression is one of the most prevalent psychological disorders, with a 12-month prevalence rate of major depressive disorder (MDD) in the United States estimated at 8.2% in adolescents and 6.6% in adults (Kessler et al., 2012; Kessler et al., 2003). Typically, depression follows a remitting, relapsing course; 50% of those who experience a first depressive episode and 80% of those who experience a second depressive episode will experience subsequent episodes (Burcusa & Iacono, 2007). The high prevalence, chronicity and early onset contribute to the significant burden associated with MDD. Among all psychological disorders, 40.5% of years lost through disability are attributable to depressive disorders (Whiteford et al., 2013). Furthermore, MDD directly increases the likelihood of serious adverse outcomes, such as suicidal behavior and cardiovascular disease (Oquendo et al., 2003; Van der Kooy et al., 2007).

Rates of depression rise dramatically in adolescence and multiple risk factors have been identified, such as: being female, reporting low socio-economic status, and experiencing stressful life events (Hankin et al., 1998; Kendler, Karkowski, & Prescott, 1999; Nolen-Hoeksema, 2001). Not only do rates of depression increase in adolescence, but adolescent-onset of MDD is associated with increased recurrence and more severe depressive symptoms in both adolescence and adulthood (Harrington, Fudge, Rutter, Pickles, & Hill, 1990; Kim-Cohen et al., 2003). Consequently, understanding the factors that shape the pathogenesis of MDD in adolescence is critical for early intervention, treatment and prevention strategies that can improve long-term physical and mental health outcomes (Patel, Flisher, Hetrick, & McGorry, 2007).

The hopelessness theory of depression (Abramson, Metalsky, & Alloy, 1989) is a prominent cognitive model of depression that focuses on dysfunctional cognitive processes that contribute to and maintain core features of MDD. The central tenet of hopelessness theory is that a negative cognitive style increases the risk of experiencing hopelessness when individuals are exposed to negative life events (NLEs). Hopelessness, in turn, is considered to be a proximal cause of MDD and, in particular, a specific subtype of depression: hopelessness depression (Abramson et al., 1989). A negative cognitive style is conceptualized as attributing negative life events (e.g., loss of a friendship) to causes that are internal (e.g. “I’m a difficult person to like”), stable (e.g., “I will continue to be difficult to like”) and global (e.g., “All my social interactions are affected by the fact that I am difficult to like”), and inferring negative consequences (e.g., “I’ll never get married and have children because no one will like me”) and negative personal characteristics (e.g., “I’m worthless”) from the occurrence of the negative life events. This negative cognitive interpretative framework is defined as a negative inferential style (Abramson et al., 1989). There is substantial evidence that negative inferential styles prospectively predict both depressive symptoms and depressive episodes in both adult and adolescent samples (Abela, 2001; Alloy et al., 2006; Gibb et al., 2001; Metalsky & Joiner, 1992); for a review, see Alloy, Salk, Stange, and Abramson (2017).

Rose and Abramson (1992) extended the hopelessness theory of depression to account for developmental antecedents of a negative inferential style. Negative childhood experiences were hypothesized to play a key role in the development of a negative inferential style (Rose, Abramson, Hodulik, Halberstadt, & Leff, 1994). Emotional maltreatment was held to be particularly salient, compared with physical or sexual abuse, because it occurs chronically and is frequently characterized by attacks on the personal characteristics of the victim (Rose & Abramson, 1992). There is significant evidence supporting the role of NLEs, and emotional maltreatment in particular, in the development of a negative inferential style, in addition to hopelessness and depression (Gibb, Alloy, Abramson, & Marx, 2003; Gibb & Abela, 2008; Gibb et al., 2001; Hankin, 2005; Paredes & Calvete, 2014).

Central to the hopelessness theory is a diathesis-stress component, whereby those with a negative inferential style become hopeless when exposed to NLEs (Abramson et al., 1989). Studies of both adolescents and adults have demonstrated support for the diathesis-stress component of the hopelessness theory of depression (Liu, Kleiman, Nestor, & Cheek, 2015). Beyond the hopelessness theory of depression literature, there is a convincing body of evidence indicating that NLEs frequently precipitate a depressive episode (Kessler, 1997; Tennant, 2002). Data supporting the relationship between NLEs and depression are predominantly based on studies examining the occurrence of recent, discrete, episodic stressors (Hammen, 2005). However, there is also strong evidence that both chronic and early life stress are associated with depression (Vrshek-Schallhorn et al., 2014; 2015). There is also considerable evidence that NLEs that are severe, interpersonal (i.e. victimization, loss or separation) or dependent (i.e., NLEs to which an individual actively contributes, such as conflict with partner due to irritability) are particularly depressogenic (Hamilton et al., 2016; Hammen, 2005).

Several central components of the hopelessness theory of depression are empirically supported, particularly the role played by a negative inferential style (both interacting with NLEs and independently) in predicting both depressive symptoms and episodes (Liu et al., 2015). These studies have, in effect, been partial tests of the hopelessness theory, given that they omit the core construct in hopelessness theory: hopelessness. Despite the prominence of the hopelessness theory of depression over the last 25 years, the cornerstone hypothesis remains untested: namely, that individuals with a negative inferential style experiencing NLEs are more likely to develop hopelessness and that hopelessness leads to clinically significant depression.

There are no data evaluating the complete hopelessness theory in predicting first onset of a major depressive episode, although several studies have tested whether hopelessness mediates the relationship between negative inferential styles, NLEs, and depressive symptoms. Overall, these studies report inconsistent results, with some studies supporting the mediating role of hopelessness in predicting depressive symptoms (Feng & Yi, 2012; Hong, Gwee, & Karia, 2006; Metalsky & Joiner, 1992; Russell, Haeffel, Hankin, Maxwell, & Perera, 2014), whereas others do not (Abela, 2001) or report inconclusive results (Hankin, Abramson, & Siler, 2001). Inconsistent results may, in part, be caused by methodological limitations and differences, such as the failure to test temporally distinct mediation models in which the mediator (hopelessness) is assessed at a time point that falls between the times at which inferential styles/NLEs and depression are assessed (Hankin et al., 2001; Metalsky & Joiner, 1992). Other methodological differences include age of participants, since young children may not yet have the cognitive resources necessary to display hopelessness (Hankin et al., 2001), and reliance on undergraduate samples where results may be attributable to sample specific characteristics (Feng & Yi, 2012; Hong et al., 2006; Metalsky & Joiner, 1992). However, discrepant results also may be due to the failure to model NLEs as a moderator of the relationship between negative inferential style and hopelessness, and via hopelessness, depression. Given that hopelessness is the central construct in the hopelessness theory of depression, further research is required to test whether hopelessness mediates the relationship between negative inferential style and depression for individuals experiencing higher levels of NLEs, particularly in more generalizable community samples of adolescents that assess first onset of MDD.

The Present Study

To our knowledge, the current study is the first complete and prospective test of whether individuals with a negative inferential style (Time 1) encountering NLEs (Time 2) develop hopelessness (Time 3) and subsequent depression (Time 5+) in a diverse sample of adolescents. Based on the hopelessness theory of depression, we hypothesized that those with a more negative inferential style exposed to different types of NLEs would become more hopeless and that increased hopelessness would, in turn, lead to the first onset of a major depressive episode as well as increased depressive symptoms. Analyses accounted for potentially confounding factors, such as NLEs in early childhood, gender, race, socio-economic status and baseline hopelessness and depression.

Consequently, two hypotheses were tested. Time 3 Hopelessness will mediate the relationship between Time 1 negative inferential style and first onset of a major depressive episode (Time 5+; Hypothesis 1) and subsequent depressive symptoms (Time 5+; Hypothesis 2) at higher levels of Time 2 NLEs (the moderator), controlling for baseline hopelessness and depressive symptoms, gender, race, socio-economic status and stressful childhood life events that occurred prior to Time 1.

Methods

Participants

Participants were drawn from the Adolescent Cognition and Emotion (ACE) project, a prospective, longitudinal study of adolescent-onset depression. Youth aged 12–13 and their mothers or primary female caregivers (hereafter referred to as mothers because 93% were the adolescent’s biological mother) were recruited from the Philadelphia-area. Youth aged 12–13 were recruited because they were entering an age range where they were at heightened risk for developing depression. Recruitment involved mailings and follow up calls to families with children attending Philadelphia-area public and private middle schools (68% of the sample) as well as advertisement in Philadelphia-area newspapers (32% of the sample) inviting youth and mothers to participate. Adolescents also needed to self-identify as Caucasian, African-American, or biracial, because examining racial differences in onset of depression also was a goal of the ACE project. Participants were excluded if either the adolescent or mother (a) had insufficient English reading/speaking skills to complete the assessments, (b) had a severe psychiatric, developmental, medical, or learning disorder or (c) if both parties did not agree to participate. Complete details on recruitment and sample characteristics have been published elsewhere (Alloy et al., 2012).

The sample for the present study included adolescents who had completed at least the baseline study session (Time 1) as well as two or more consecutive follow-up sessions. Diagnostic data used in the current study included diagnostic assessments at any time point across the study (Time 1 – Time 15). Given that the aims of this study were to predict prospective onset of depression, participants who had a past or current diagnosis of MDD at either baseline (Time 1), first, or second follow-up (Times 2 or 3) were excluded from the present analyses. Following the exclusion of participants with a diagnosis, complete data were available for 434 adolescents at baseline, 280 at first follow-up, 204 at second follow-up and 173 at all time points. Missing data were handled statistically and models predicting to MDE and depressive symptoms were based on 249 and 244 participants, respectively. Summary statistics are reported for the 173 adolescents with complete data; 56% were female, 44% received free or reduced price lunch, and 53% self-identified as African-American. Independent sample t-tests indicated that the subsample retained in the study did not differ at baseline from the remainder on number of stressful childhood life events occurring prior to Time 1 (p =.70), hopelessness (p =.22), negative inferential style (p =.87) or depression (p =.25). Chi-squared analyses indicated no difference between the subsample used for analyses and the excluded sample were observed on sex (p =.41) or race (p =.51); however, a trend towards significance was observed for socio-economic status (SES), with fewer individuals with lower SES present at all data points in the subsample than expected (p = .06).

Measures

Diagnostic Assessment

The Schedule for Affective Disorders and Schizophrenia for School-Age Children-Epidemiologic Version (K-SADS-E), Fifth Edition (Orvaschel, 1995a) is a semi-structured diagnostic interview that assesses DSM-IV-TR Axis I (DSM) and Research Diagnostic Criteria (RDC) psychopathology (American Psychiatric Association, 2013; Spitzer, Endicott, & Robins, 1978). The K-SADS-E has demonstrated good test-retest and interrater reliability (Orvaschel, 1995b). The K-SADS-E was administered by the same interviewer first to the mother and then to the adolescent at Time 1 to assess current and lifetime Axis-I psychopathology, and was re-administered at follow-ups to assess subsequent diagnoses. Summary ratings for the K-SADS-E were based on the interviewer’s “best estimate” clinical judgment from the combined information from the mother and youth report. Interviewers were clinical psychology postdoctoral fellows and PhD students, individuals with master’s degrees in counseling or clinical psychology, and post-BA research staff who were blind to other project data. Training consisted of a three-day instructional session followed by ~200 hours of didactic instruction, role-play, observation and practice to ensure standardization (for further details, see Alloy et al., 2012). Inter-rater reliability for Project ACE was κ = .85 for 120 pairs of ratings (five raters for each of 24 diagnoses from 10 K-SADS-E interviews).

Depressive Symptoms

The Children’s Depression Inventory (CDI) is a valid, reliable self-report measure of current depressive symptoms in youth (Kovacs, 1992). It consists of 27 items, scored on a three-point scale ranging from zero to two. Items were summed, with higher scores indicative of higher levels of depressive symptoms. The CDI was administered at Time 1 and Time 5. For those who were not present at Time 5, the next available CDI after Time 5 was used. Cronbach’s α for Time 1 and Time 5 was .85 and .86, respectively.

Hopelessness

The Hopelessness Scale for Children (HSC) (Kazdin, French, Unis, Esveldt-Dawson, & Sherick, 1983) is a self-report measure of current beliefs of hopelessness. It consists of 17 true or false questions (“When things are going badly, I know that they won’t be bad all of the time”; “I never get what I want, so it’s dumb to want anything.”). Higher summary scores indicate increased hopeless thinking. The HSC has demonstrated good internal consistency (Spirito, Williams, Stark, & Hart, 1988) and validity (Kazdin, Rodgers, & Colbus, 1986). It was administered at Time 1 and Time 3. In the current sample, the internal consistency of the HSC was α = .59 at baseline assessment (Time 1) and .72 at second follow-up, the timepoint at which it was used as a mediator (Time 3).

Emotional Maltreatment

The Childhood Trauma Questionnaire Short Form (CTQ) (Bernstein et al., 2003) is a 28 item self-report questionnaire shown to be a valid retrospective assessment of abuse and neglect over the previous six months. Adolescents responded to items describing maltreatment with responses ranging from 1 (Never True) to 5 (Very Often True). The CTQ yields five subscales-Emotional Abuse, Physical Abuse, Sexual Abuse, Emotional Neglect, and Physical Neglect. Project ACE only utilized the Emotional Abuse and Emotional Neglect subscales to specifically measure emotional maltreatment. These subscales have demonstrated good internal consistency (α’s =.89) in adolescent samples (Bernstein et al., 2003). The CTQ also demonstrates good concurrent validity with severity of psychopathology (Spinhoven et al., 2014). The CTQ was administered at Time 2 and Time 3. For participants who were not present at Time 2, data from Time 3 were used as an estimate of emotional maltreatment in between Time 1 and Time 3 (n = 99). In the current sample, the internal consistency of the emotional abuse and emotional neglect subscales of the CTQ were α = .80 and .82.

Adolescent Cognitive Style

The Adolescent Cognitive Style Questionnaire-Modified (ACSQ-M; (Alloy et al., 2012) evaluates the style of inferences made by adolescents about the causes, consequences, and self-worth implications of NLEs. Three domains of NLEs (achievement, interpersonal and appearance) are assessed. When presented with each of the 12 hypothetical negative events (4 events for each domain), adolescents are invited to make inferences about the causes (internal/external, stable/unstable, and global/specific), consequences, and self-worth implications of each event. Each item is scored on a scale ranging from one to seven, with higher scores indicative of a more negative inferential style. A total negative composite score was calculated as the mean score across dimensions of stability, globality, consequences, and self-worth implications and across the achievement, interpersonal, and appearance domains with scores ranging from 1 to 7. The ACSQ-M is internally consistent with good test-retest reliability (Alloy et al., 2012). The internal consistency in this sample for total negative composite score at Time 1 was excellent (α = .94).

Childhood Life Events

The Children’s Life Events Scale – parent report (CLES-PR) measures the occurrence of stressful life events during childhood, as reported by a parent (Crossfield, Alloy, Gibb, & Abramson, 2002). Mothers responded “yes” or “no” for 50 childhood events that occurred in their adolescent child’s life prior to Time 1 deemed to be moderately-to-majorly stressful, including physical and sexual abuse, bereavement, poor school performance, achievement failures and negative emotional feedback. A total score was calculated by summing all affirmative answers, with higher scores indicating greater experiences of stress. The CLES-PR has demonstrable predictive validity and has been associated with the later development of negative inferential styles (Crossfield et al., 2002).

Life Event Interview

NLEs were measured using a semi-structured interview, the Life Events Interview (LEI; Safford et al., 2007). Both adolescents and their mothers completed separate versions of a self-report questionnaire, the Adolescent Life Events Questionnaire (ALEQ; Hankin & Abramson, 2002), that ascertain the occurrence of 63 NLEs common in adolescence. Subsequently, the LEI was conducted to determine whether events endorsed on the ALEQ by adolescents and/or their mothers met event definition criteria for inclusion and occurred during the outlined time period. Objective criteria are designed to reduce the effect of a negative reporter bias. Life events are rated by severity and categorized as interpersonal (e.g., romantic break-up), achievement (e.g., failed a test), independent (e.g., family member died) or dependent (e.g., fought with friend). Independent and dependent life events distinguish events that occur independently of the participant’s actions and events that individuals can contribute towards. The current study examined total number of NLEs, as well as the total number of major, independent and interpersonal NLEs at Time 2. Both the ALEQ and LEI are valid and reliable instruments (Hankin, 2008; Safford et al., 2007).

Procedure

At baseline (Time 1), adolescents completed a diagnostic interview (K-SADS-E), and questionnaires assessing depressive symptoms (CDI), cognitive style (ACSQ), hopelessness (HSC), and their mothers completed a report of stressful childhood life events (CLES-PR) and demographic information. Participants were scheduled to complete questionnaire measures of hopelessness (HSC) and depressive symptoms (CDI) every 6 months (Time 1, Time 2, Time 3, Time 4…) and measures of negative inferential style and diagnostic interviews at every 12 month assessment (Time 1, Time 3, Time 5…). All types of NLEs were assessed every 6 months (Time 2, Time 3, Time 4, Time 5…) after baseline. NLEs (Total number, major, interpersonal and independent) were assessed using the semi-structured interview (LEI) while emotional maltreatment was assessed using the CTQ. Participants returned to the lab for at least two follow-up visits scheduled at either 6- or 12-month intervals, with time to first follow-up (Time 2) occurring approximately 7 months after the Time 1 baseline (M=7.09; SD=2.02), time to second follow-up (Time 3) approximately 11 months after Time 2 (M=11.56; SD=4.49), and time to third follow-up (Time 5) approximately 12 months after Time 3 (M=12.02; SD=4.17). To ensure a fully prospective design, participants who experienced a MDE prior to Time 5 were excluded from these analyses. When predicting to first onset, all available diagnostic assessments that occurred at Time 5 or later were used; on average, diagnostic data covered a time period of 3.95 years (SD = 619 days) from Time 1. Likewise, follow-up depressive symptoms were assessed at Time 5 or at the next available follow-up.

Data Analysis Plan

Descriptive statistics and correlations for complete data were conducted in SPSS (v24) (IBM Corp, 2016). Primary analyses were conducted using Mplus Version 7.4 (Muthén & Muthén, 1998–2015). Full Information Maximum Likelihood (FMIL) was used to handle missing data and reduce the likelihood of biased parameter estimates (Graham, 2009). When predicting to a binary outcome, robust weighted least squares (WLSMV) estimates were used and when predicting to a continuous outcome, robust maximum likelihood (MLR) estimates were used. Indirect effects were estimated based on 5000 bootstrapped samples. Moderated mediation analyses examined whether hopelessness (Time 3) mediated the relationship between the inferential style-NLEs interaction and depressive symptoms (Time 5+) as well as a major depression diagnosis (Time 5+). Moderated mediation analyses controlled for baseline depressive symptoms (Time 1), gender, socio-economic status, race, and stressful childhood events (Time 1). Further, the direct effect of NLEs on both outcome variables (Time 5+ Depressive Symptoms and Time 5+ MDE) were estimated and included within moderated mediation models. Within the moderated mediation models, the moderation component differed from the moderated mediation component because it controlled for baseline hopelessness, instead of baseline depressive symptoms. Stressful childhood life events, gender, socio-economic status and race were included as covariates because they have been associated with hopelessness and/or depression. Three estimates of model fit were used to assess whether the moderated mediation models fit the data well: chi-squared estimate of goodness of fit, Comparative Fit Index (CFI) and Root Mean Square Error of Approximation (RMSEA). Chi-squared estimate of goodness of fit statistics are not appropriate indices of model fit when using WLSMV and, consequently, were not reported.

Results

Descriptive Statistics

Descriptive statistics and bivariate correlations for the main study variables are presented in Table 1. Descriptive statistics reported are based on the sample of 173 adolescents with complete data on demographic (sex, race and socio-economic status), diagnostic and life event interview (K-SADS-E, LEI) as well as self-report measures (ACSQ, CTQ, HSC, CDI). Twenty-six participants were diagnosed with first onset of a MDE at their third or subsequent follow-up assessments (Time 5+), based on DSM-IV-TR (major depressive episode) or RDC (Major Definite or Major Probable) criteria.

Table 1.

Descriptive Statistics and Bivariate Correlations of Study Variables

Measure	2	3	4	5	6	7	8	9	10	11	12	13	14	15
1: MDD Dx	.40***	.24**	.16*	.12	−.02	−.13‡	.11	.19*	.09	.06	.20**	.13	.19*	.05
2: CDI T5+	-	.38***	.37***	.15*	.14‡	.06	.15*	.39***	.20**	.15‡	.42***	.23**	.40***	.23*
3: HSC T3		-	.37***	.15‡	.12	.08	.32***	.27***	.10	.21**	.20**	.11	.20**	.05
4: CTQ T2			-	.11	−.02	−.05	.26**	.41***	.16*	.32***	.37***	.06	.39***	.01
5: Sex				-	−.02	.09	−.08	.12	−.13‡	.07	.25**	.24**	.29***	.15*
6: SES					-	.43***	−.05	.02	−.03	.07	.04	.06	−.01	.08
7: Race						-	−.03	−.01	−.11	.03	.05	.19*	.04	.23**
8: HSC T1							-	.28***	.12‡	.04	.05	−.03	.06	−.04
9: CDI T1								-	.15*	.13‡	.26***	.08	.24**	.06
10: ACSQ-M T1									-	−.07	.16*	.00	.16*	−.01
11: CLES-PR T1										-	.07	.05	.07	.07
12: Total											-	.66***	.96***	.55***
13: Independent												-	.65***	.86***
14: Interpersonal													-	.54***
15: Major														-
Mean	6.52	2.86	15.97	.56	.44	.53	2.94	6.54	2.43	9.92	11.01	2.32	7.37	1.42
(SD)	6.04	2.51	6.01	.50	.50	.50	1.96	5.59	.78	4.39	6.75	2.13	5.15	1.65

Note: N = 173; MDD Dx: Major Depression DSM or RDC Diagnosis; CDI: Children’s Depression Inventory; HSC: Hopelessness Scale for Children; CTQ: Childhood Trauma Questionnaire Emotional Abuse and Emotional Neglect subscales; SES: Socio-economic status; ACSQ-M: Adolescent Cognitive Style Questionnaire – Modified; CLES-PR: Child Life Events Scale – Parental Report; Total: Total Number of Negative Life Events; Independent: Total Number of Independent Negative Life Events; Interpersonal: Total Number of Interpersonal Negative Life Events; Major: Total Number of Major Negative Life Events; T1: Time 1; T2: Time 2; T3: Time 3; T5+: Time 5 or first subsequent time point.

Probability: ‡ p<.10

^*p<.05;

^**p<.01;

^***p<.001

Hypothesis 1: Moderated Mediation Analyses predicting to First Major Depressive Episode

Moderated mediation analysis tested whether adolescent hopelessness (T3) mediated the relationship between negative inferential style (T1) and first onset of a MDE (T5+) at levels of five NLEs (T2): emotional maltreatment, total number of NLEs, total number of major NLEs, total number of interpersonal NLEs, and total number of independent NLEs, see Table 2. Overall, estimates of model fit indicated that the moderated mediation models fit the data well: emotional maltreatment (CFI = .988; RMSEA = .032 (90% CI: .001, .12)), total number of NLEs (CFI = .989; RMSEA = .029 (90% CI: .001, .116)); major NLEs (CFI = .945; RMSEA = .049 (90% CI: .001, .128)); interpersonal NLEs (CFI = .994; RMSEA = .029 (90% CI: .001, .112)) and independent NLEs (CFI = .985; RMSEA = .032 (90% CI: .001, .117)).

Table 2.

Hopelessness Mediates the Relationship Between T1 Negative Inferential Styleand First-onset of Major Depressive Episode for Different Values of the Moderators, Negative Life Events

Outcome: T5+ First Major Depressive Episode (N = 249)
a’ path: predicting to T3 Hopelessness
	T2 Negative Life Events
	CTQ	Total	Major	Interpers	Indep
Variable	b(se) p	b(se) p	b(se) p	b(se) p	b(se) p
T1 NIS	.06(.06)	.06(.06)	.11(.07)	.06(.06)	.11(.07)
T1 Hopelessness	.25(.09)**	.28(.09)**	.29(.10)**	.29(.09)**	.29(.09)**
T1 CLES-PR	.10(.08)	.16(.08)	.17(.08)*	.16(.08)*	.17(.08)*
T2 NLE	.20(.10)*	.09(.07)	−.01(.06)	.09(.07)	.04(.06)
Sex (Female)	.17(.13)	.15(.15)	.24(.14)	.15(.15)	.20(.14)
Race (Black)	.06(.14)	.05(.13)	.07(.14)	.06(.14)	.05(.14)
SES (Low SES)	.22(.13)	.22(.14)	.22(.13)	.23(.14)	.22(.13)
Interaction	.16(.07)*	.21(.06)**	.15(.06)*	.19(.06)**	.15(.06)*
R2 accounted for	21.5%	19.8%	17.0%	19.1%	17.1%
b’ path: predicting to T5+ First Major Depressive Episode
	CTQ	Total	Major	Interpers	Indep
Variable	b(se) p	b(se) p	b(se) p	b(se) p	b(se) p
T1 NIS	.04(.12)	.08(.12)	.10(.11)	.08(.23)	.09(.11)
T1Depressive Sxs	.09(.14)	.06(.13)	.11(.13)	.07(.13)	.09(.13)
T1 CLES-PR	−.04(.10)	−.02(.10)	−.02(.10)	−.01(.10)	−.02(.10)
T2 NLE	.03(.14)	.12(.11)	.01(.12)	.10(.11)	.07(.13)
Sex (Female)	.51(.22)*	.46(.23)*	.52(.22)*	.46(.23)*	.50(.23)*
Race (Black)	−.45(.24)	−.41(.23)	−.40(.23)	−.41(.23)	−.42(.23)
SES (Low SES)	.17(.26)	.12(.25)	.12(.25)	.13(.25)	.13(.25)
T3 Hopelessness	.20(.11)	.21(.10)*	.22(.10)*	.21(.10)*	.21(.10)*

^*p <.05;

^**p < .01;

^***<.001;

NIS = Negative Inferential Style; CLES-PR = Parent Report of Stressful Childhood Life Events; NLE = Negative Life Events; Interaction = T1 NIS *T2 NLE; Total = Total Number of NLE; Major = Total Number of Major NLE; Interpers = Total Number of Interpersonal NLE; Indep = Total Number of Independent NLE.

In accordance with our first hypothesis, T3 adolescent hopelessness mediated the relationship between the interaction of negative inferential style at T1 and multiple NLEs at Time 2 and first onset of a MDE. An indirect conditional effect was only observed at high levels of: total number of NLEs, (b = .02; 95%CI: .003, .05), major NLEs (b = .08; 95%CI: .01, .20), interpersonal NLEs (b = .08; 95%CI: .001, .19), and independent NLEs (b = .08; 95%CI: .001, .21). Importantly, indices of moderated mediation suggest that hopelessness mediated this effect for total NLEs (b = .01, 95%CI: .001, .02), major NLEs (b = .03, 95%CI: .01, .08), and interpersonal NLEs (b = .01, 95%CI: .001, .03), but not for emotional maltreatment (b = .01, 95%CI: −.002, .02) or independent NLEs (b = .02, 95%CI: −.001, .07). Therefore, for total, major and interpersonal NLEs, hopelessness mediated the relationship between a negative inferential style and subsequent MDE onset at higher levels of NLEs only.

Time 3 hopelessness predicted to first onset of a MDE across all models, except emotional maltreatment where a trend level association was observed (p = .06). Being female was associated with a greater likelihood of a MDE across all models. As expected, the interaction of negative inferential style (T1) and all types of NLEs predicted to later adolescent hopelessness (T3), controlling for baseline hopelessness and other confounds. Likewise, baseline hopelessness consistently predicted hopelessness. A main effect was observed for emotional maltreatment predicting to hopelessness; however, no main effect was observed for other NLEs. Significant main effects of negative childhood life events (T1 CLES-PR) predicting to hopelessness were reported for major, interpersonal and independent NLEs models, with a trend level association observed in the model utilizing total number of NLEs.

Hypothesis 2: Moderated Mediation Analyses predicting to Depressive Symptoms

The second hypothesis tested whether a similar relationship could be observed when predicting to depressive symptoms at third, or subsequent, follow-up. Results can be seen in Table 3. Moderated mediation analyses tested whether T3 adolescent hopelessness mediated the relationship between negative inferential style (T1) and depressive symptoms (T5+) at different levels of T2 NLEs. Estimates of model fit indicated that the moderated mediation models fit the data well for: emotional maltreatment (χ²(3) = 3.35, p > .05; CFI = .997 and RMSEA = .02 (90% CI: .00, .11), total number of NLEs (χ²(3) = 5.05, p > .05; CFI = .982 and RMSEA = .05 (90% CI: .00, .13), major NLEs (χ²(3) = 2.54, p > .05; CFI = .994 and RMSEA = .03 (90% CI: .00, .12), interpersonal NLEs (χ²(3) = 4.03, p > .05; CFI = .990 and RMSEA = .04 (90% CI: .00, .12), and independent NLEs (χ²(3) = 3.68, p > .05; CFI = .993 and RMSEA = .03 (90% CI: .00, .12).

Table 3.

Hopelessness Mediates the Relationship Between T1 Negative Inferential Styleand Time 5+ Depressive Symptoms for Different Values of the Moderator, Negative Life Events

Outcome: T5+ Depressive Symptoms (N = 244)
a’ path: predicting to T3 Hopelessness
	T2 Negative Life Events
	CTQ	Total	Major	Interpers	Indep
Variable	b(se) p	b(se) p	b(se) p	b(se) p	b(se) p
T1 NIS	.07(.06)	.07(.06)	.12(.06)	.07(.06)	.12(.06)
T1 Hopelessness	.25(.08)**	.30(.08)***	.31(.09)***	.31(.07)***	.31(.09)***
T1 CLES-PR	.09(.08)	.16(.09)	.18(.08)*	.17(.08)*	.18(.08)*
T2 NLE	.23(.09)**	.11(.07)	−.01(.06)	.12(.07)	.05(.06)
Sex	.18(.13)	.16(.14)	.13(.12)	.15(.14)	.10(.12)
Race	.05(.13)	.03(.14)	−.10(.14)	.04(.14)	−.08(.13)
SES	.20(.13)	.19(.13)	.12(.13)	.20(.14)	.12(.13)
Interaction	.14(.06)*	.22(.06)***	.17(.06)**	.20(.06)**	.17(.06)**
R2accounted for	23.2%	22.2%	19.3%	21.2%	19.3%
b’ path: predicting to T5+ Depressive Symptoms
	CTQ	Total	Major	Interpers	Indep
Variable	b(se) p	b(se) p	b(se) p	b(se) p	b(se) p
T1 NIS	.11(.06)	.10(.05)	.13(.06)*	.10(.05)	.13(.06)*
T1Depressive Sxs	.19(.07)**	.18(.06)***	.24(.06)***	.20(.06)**	.24(.06)***
T1 CLES-PR	−.01(.06)	.03(.06)	.03(.06)	.04(.06)	.04(.06)
T2 NLE	.20(.08)*	.28(.07)***	.15(.08)	.25(.06)***	.15(.08)*
Sex	.12(.12)	.03(.12)	.13(.12)	.02(.12)	.09(.12)
Race	−.01(.13)	−.05(.13)	−.10(.14)	−.05(.13)	−.08(.13)
SES	.10(.13)	.11(.12)	.12(.13)	.12(.12)	.12(.12)
T3 Hopelessness	.20(.08)*	.24(.08)**	.25(.08)**	.24(.08)**	.25(.08)**
R2accounted for	24.4%	28.6%	23.8%	27.2%	24.2%

^*p <.05;

^**p < .01;

^***<.001;

NIS = Negative Inferential Style;CLES-PR = Parent Report of Stressful Childhood Life Events; NLE = Negative Life Events; Sxs = Symptoms; Interaction = T1 NIS *T2 NLE; Total = Total Number of NLE; Major = Total Number of Major NLE; Interpers = Total Number of Interpersonal NLE; Indep = Total Number of Independent NLE.

Supporting the second hypothesis, T3 adolescent hopelessness mediated the relationship between the interaction of negative inferential style at T1 and NLEs at Time 2 and subsequent depressive symptoms at high levels of: emotional maltreatment (b = .29; 95%CI: .04, .77), total number of NLEs (b = .53; 95%CI: .19, 1.07), major NLEs (b = .56; 95%CI: .19, 1.21), interpersonal NLEs (b = .49; 95%CI: .16, 1.04), and independent NLEs (b = .55; 95%CI: .18, 1.21). Additionally, T3 adolescent hopelessness mediated the relationship at mean levels of: major NLEs (b = .22; 95%CI: .02, .59) and independent NLEs (b = .22; 95%CI: .02, .60). Unexpectedly, an indirect conditional effect was observed for low levels of total number of NLEs (b = −.29; 95%CI: −.75, −.03). Index of moderated mediation estimates indicate that the conditional indirect effect was mediated by hopelessness for each model: emotional maltreatment (b = .03, 95%CI: .006, .09), total NLEs (b = .06, 95%CI: .02, .12), major NLEs (b = .22, 95%CI: .06, .46), interpersonal NLEs (b = .08, 95%CI: .02, .16), and independent NLEs (b = .17, 95%CI: .05, .37). Further, hopelessness at Time 3, baseline depressive symptoms, and all types of NLEs at Time 2 all directly predicted T5+ depressive symptoms. Once again, a main effect was only observed for emotional maltreatment predicting to hopelessness. Likewise, major, interpersonal and independent NLEs models reported significant main effects of negative childhood life events (T1 CLES-PR) predicting to hopelessness, with a trend level association observed for the model using the total number of NLEs.

As mentioned in the Measures section, a trend towards significance was observed such that participants of lower SES at baseline were less likely to complete at least two follow-up sessions after baseline. Consequently, both moderated mediation analyses conducted above were repeated with SES declared as an endogenous variable in FIML and no changes in the overall pattern of significant results were observed.

Discussion

The hopelessness theory of depression is a clearly defined theoretical model of depression with considerable empirical support. Few studies, however, have used moderated mediation analyses to test the complete model of depression specified by hopelessness theory: specifically, that individuals with a negative inferential style who encounter negative life events are more likely to develop hopelessness and, via hopelessness, depression. Furthermore, surprisingly, no studies have tested the complete model when predicting to first onset of a diagnosed depressive episode. Our results showed that hopelessness does, in fact, mediate the relationship between a negative inferential style and depression (both depressive symptoms and first onset of a MDE) at higher levels of multiple, but not all, NLEs in a diverse, community sample of adolescents.

Complete Hopelessness Model of Depression Predicting First Onset of Major Depression

This study is the first to demonstrate that the complete hopelessness model of depression predicts first onset of a MDE. It is surprising that, despite more than 1,400 citations over 25 years, the cornerstone hypothesis of the hopelessness theory of depression has not been adequately tested (Liu et al., 2015). It is unclear why it has taken so long to test this hypothesis; however, it may be that: the need to carry out diagnostic interviews in a sample of adolescents sufficiently large to reliably predict first onset of MDE, the need to assess participants at three or more temporally distinct points, and/or the relatively recent proliferation of more advanced statistical approaches/software (required to run moderated mediation models) proved to be prohibitively high barriers to theory-testing. These results support the complete hopelessness theory of depression and confidence in the generalizability of our results is bolstered by the use of a fully prospective design in a socio-economically and racially diverse community sample of adolescents.

These findings support and build upon previous studies that have partially tested the hopelessness theory of depression’s ability to predict clinical depression. Previous research has shown that a negative inferential style directly predicts a depression diagnosis and that negative inferential style interacts with NLEs to predict first onset and recurrence of a MDE (Alloy et al., 2006; Gibb et al., 2001; Hankin, Abramson, Miller, & Haeffel, 2004). Our findings are also in line with previous complete tests of the hopelessness theory when predicting to depressive symptoms (Russell et al., 2014). Results of the current study suggest that hopelessness mediates the relationship between a negative inferential style and depression at higher levels of many forms of NLEs. This is consistent with the hopelessness theory of depression’s prediction that NLEs are necessary to activate a negative inferential style, thereby leading to hopelessness. Our findings are also in line with a number of previous studies reporting that hopelessness mediates the relationship between the interaction of a negative inferential style and NLEs and later depressive symptoms (Feng & Yi, 2012; Hong et al., 2006; Metalsky & Joiner, 1992; Russell et al., 2014) and more elaborate models that included additional constructs, such as social support (Panzarella, Alloy, & Whitehouse, 2006) and self-esteem (Metalsky, Joiner, Hardin, & Abramson, 1993).

The findings of the current study are, overall, at odds with other studies that have failed to find a mediating role for hopelessness (Abela, 2001; Hankin et al., 2001). These studies explicitly tested whether hopelessness mediated the relationship between a cognitive vulnerability – stress interaction and depression in samples of children and adolescents experiencing NLEs and found no evidence of hopelessness as a mediator. This discrepancy may be attributable to the use of less robust statistical methods by both studies reporting null results, with both studies following the precedent of Metalsky and Joiner (1992) in estimating mediation by testing whether a statistically significant relationship between a cognitive vulnerability-stress interaction and depression held after controlling for hopelessness and whether hopelessness remained a significant predictor. This method fails to estimate the conditional indirect effect of a cognitive vulnerability-stress interaction on depressive symptoms that occurs via hopelessness, and consequently, may not be an effective test of the hopelessness theory. Null results also may be attributable to sample characteristics (child samples that haven’t fully developed the capacity for hopelessness) or methodological issues (temporally indistinct assessment points) that characterize these previous studies. Additionally, replication of the mediating role of hopelessness in a socio-economically and racially diverse community sample in this study suggests that the hopelessness theory of depression generalizes beyond the undergraduate samples that predominate in previous research (Feng & Yi, 2012; Hong et al., 2006; Metalsky & Joiner, 1992). The results of the current study contribute to a growing number of studies that have found a mediating role for hopelessness when modeling the relationship between negative inferential style, NLEs, and depression.

Negative Life Events

This study used a mixture of both self-report and interview-based assessments of NLEs. Overall, there was a consistent pattern of results observed for the five different types of stressors, suggesting that multiple, albeit not all, forms of life stress interact with a negative inferential style to predict hopelessness and, via hopelessness, a MDE. Mediation only occurred at high levels of NLEs, suggesting that higher numbers of NLEs are required for the hopelessness pathway to depression to be activated for those with a negative inferential style. This finding may reflect the higher levels of stress required to precipitate first onset of a depressive episode, compared to recurrent depression, observed in previous studies (Hammen, 2005; Stroud, Davila & Moyer, 2008). However, hopelessness did not mediate the relationship for emotional maltreatment or independent life events, suggesting that both types of stressors increased the likelihood of a MDE via a mechanism other than hopelessness. For example, it may be that, since independent NLEs are, by definition, beyond an individual’s control, they do not activate a negative inferential style. Given the ubiquity and inevitability of experiencing NLEs, treatment research may fruitfully continue to focus on the development and dissemination of cognitive reappraisal and cognitive emotional regulation techniques that may promote resilience to NLEs in adolescents.

Emotional maltreatment is a particularly malignant NLE that has been shown to significantly and differentially predict depression (Spinhoven et al., 2010). Although the current study assessed emotional maltreatment occurring in the six months prior to each follow-up assessment, it is likely that this questionnaire captures variance attributable to earlier and possibly a more chronic course of emotional maltreatment. This effect is observable in the statistical significance of stressful childhood life events in predicting to hopelessness in models that use interview-based assessments of NLEs. This likely highlights the difficulty in parsing apart episodic from chronic emotional maltreatment due to their frequent co-occurrence. These results caution about the use of emotional maltreatment as a measure of episodic life stress. In the current study, all forms of life stress measured were found to have an independent and direct effect on depressive symptoms, but not on the likelihood of experiencing a depressive episode. This may reflect the difficulties detecting the effect in a small group of individuals diagnosed with a first depressive episode; however, it may also support previous studies reporting that observed cognitive vulnerabilities completely or partially mediate the relationship between NLEs and depression, at least for specific NLEs (Gibb & Abela, 2008; Gibb et al., 2001).

Strengths and Limitations

The use of a fully prospective design and sophisticated data-analytic methods in a racially and socio-economically diverse community sample of adolescents and testable hypotheses based on a sound theoretical framework are considerable strengths of this study. Conclusions, however, should be considered alongside the study’s limitations. A significant limitation in the current study is that a relatively small number of participants developed MDEs at any point beyond the third follow-up, although it is worth noting that results predicting to both MDEs and depressive symptoms are highly congruent, supporting the validity of these results. A second limitation of the study is that the effect sizes were small when predicting to MDEs. This suggests that, although significant, the pathway identified by hopelessness theory leading to MDE onset is only a partial explanation of why depression emerges in adolescence. Consequently, future research should test the pathway identified by the hopelessness theory of depression alongside other known genetic (e.g. 5-HTTLPR gene), personality (e.g. neuroticism), neuroendocrine (e.g. hypothalamic pituitary adrenal axis dysregulation), neuroimmunological (e.g. immune system abnormalities) and psychological factors (e.g. information-processing biases) associated with depression. A third limitation was that assessments occurred at six month intervals. Researchers have shown that negative inferential style interacts with NLEs to predict hopelessness/depression within five to six weeks (Abela, 2001; Metalsky & Joiner, 1992); thus, the temporal precedence of hopelessness before depression could not be definitely established in this study. However, the use of temporally distinct points of measurement satisfied criteria for mediation analyses and the multi-year research design did enable the study to detect the emergence of clinically relevant depression assessed via diagnostic interview. A final limitation is that NLEs from the second follow up assessment were used for participants who missed the first follow up (n = 99) and consequently, raises the possibility that recall of NLEs was biased by levels of hopelessness at second follow up for a proportion of participants; these data were included, however, because the larger sample size would improve the precision of the estimates used and the reliability of the results. This limitation is somewhat offset by the use of multiple measures of NLEs, particularly ones using more valid interview-based assessment with event definition criteria that protect against subjective report bias (Monroe, 2008).

Conclusion

This study provides strong support for the previously untested cornerstone hypothesis of the hopelessness theory of depression. We found that, as predicted, a negative inferential style interacts with negative life events to predict first onset of a major depressive episode as well as depressive symptoms in adolescence via hopelessness. This demonstrates the continued relevance of hopelessness theory in contributing to an explanation of depression.

General Scientific Summary.

Adolescents who possessed a negative cognitive style developed hopelessness when they were exposed to many types of stressful life events, particularly major and interpersonal negative life events. In turn, hopelessness was shown to predict the first diagnosis of depression in adolescents, thus providing strong support for hopelessness as one pathway to depression.