Developmental Trajectories of Anxiety and Depression
Developmental trajectories of anxiety and depression in adolescents have been widely investigated with large-scale longitudinal studies. Growth mixture modeling (GMM) is a useful data-driven approach for discerning distinct developmental classes within large heterogeneous data sets and identifying those most at risk (Grimm et al.
2017; Olino et al.
2010). This approach has mostly been applied to developmental trajectories of depression. A recent meta-analysis of twenty longitudinal studies published between 2002 and 2015, found substantial heterogeneity in the development of depression during adolescence (Shore et al.
2018). Within these studies, between three and eleven distinct developmental trajectories were found, although a random pooled effect estimate identified three distinct groups. The largest group identified (56% of the pooled sample), were characterised by low or no depression throughout adolescence. The second group (26%) were characterised by moderate and stable levels of depression. While the third group (12%) showed high fluctuating levels of depression throughout adolescence. Risk factors included being female, having low socio-economic status, as well as multiple peer or family problems, and poor adjustment outcomes.
Age at onset has also been shown to be an important factor contributing to depression heterogeneity in adolescence. A recent study using data from 7543 adolescents who took part in the Avon longitudinal study, a 1991 UK birth cohort, found three distinct depression trajectories (Rice et al.
2019). The largest group (74%) were characterised by persistently low symptoms. The second largest group (17%) were characterised by late-onset (around 16 years of age) and increasing depression. While the smallest group (9%) showed early-onset (around 12 years of age) and increasing depression. Genome-wide analysis was conducted and polygenic risk scores based on different psychiatric disorders were able to distinguish between the groups. Both the late-onset and early-onset groups, relative to the low symptoms group, were associated with higher polygenic risk for major depressive disorder. Furthermore, the early-onset group was associated with higher polygenic risk for attention deficit hyperactivity disorder and schizophrenia, suggesting that this group may have a wider breadth of genetic psychiatric vulnerability. While this study highlights important biological pathways, cognitive mechanisms that could provide useful intervention targets were not investigated.
The development of anxiety in adolescence is less clearly understood. This is partly due to the large number of subtypes of anxiety (e.g., generalised anxiety, separation anxiety, and social phobia), which tend to show different peak ages of onset (Cummings et al.
2014). Previous longitudinal studies have found an overall trend for decreasing levels of anxiety from childhood to adolescence (Allan et al.
2014; McLaughlin and King
2015; Van Oort et al.
2009). However, this is likely to be symptom-specific, as panic disorder and social anxiety have been shown to increase during adolescence (Hale III et al.
2009). One cohort study conducted in a community sample of 2220 adolescents, found that the developmental course of anxiety symptoms decreased from late childhood to early adolescence, however slightly increased from mid-adolescence (generalised anxiety, separation anxiety, and social phobia) or late adolescence (panic disorder and OCD) onwards depending on the anxiety subtype (Van Oort et al.
2009). There is some consensus that anxiety predominately manifests during childhood and early adolescence, while depression develops in later adolescence and young adulthood (Hankin et al.
1998; Merikangas et al.
2010; Roza et al.
2003). Yet, anxiety and depression are highly overlapping and comorbid across adolescence (Ferdinand et al.
2005). Due to this, some have argued that both anxiety and depression be considered under one general factor reflecting ‘internalising disorders’ (Hankin et al.
2016). However, a longitudinal study of 1313 adolescents found that while anxiety and depression are highly comorbid, they are best described by parallel growth processes (Hale III et al.
2009).
An in-depth review of the literature on anxiety and depression comorbidity in youth led to the ‘Multiple Pathways Model’ (Cummings et al.
2014). These authors concluded that anxiety and depression are separate but meaningfully related constructs, which emerge largely from three distinct pathways. Pathway one refers to youth with a diathesis for anxiety (often separation or social anxiety), which develops into depression comorbidity if anxiety is left untreated. In this pathway, anxiety is likely to be severe and depression mild. Pathway two refers to youth with a shared diathesis for anxiety and depression who experience the disorders simultaneously, often manifesting with symptoms of depression and generalised anxiety. In this pathway, anxiety is likely to be severe and depression moderate. Pathway three refers to youth with a diathesis for depression who develop anxiety comorbidity resulting from depression-related impairment, such as peer victimisation or social isolation. The third pathway is the least common and is likely to represent older adolescents and young adults. Thus, evidence suggests that anxiety and depression often co-occur during childhood and adolescence, which can have a detrimental impact on subsequent development (Kaufman et al.
2001). A better understanding of the mechanisms that contribute to the onset and maintenance of early psychopathology in youth can help identify risk factors to target in interventions.
Discussion
The present study investigated the development of anxiety and depressive symptom trajectories and cognitive biases during adolescence. As hypothesised, the results showed that overall levels of anxiety and depression were low, yet depressive symptoms increased slightly at each wave. Furthermore, in line with our hypotheses, we found multiple class trajectories of anxiety and depression. Although we did not hypothesise the number of class trajectories, our results identified four distinct developmental classes. Analysis of the development of cognitive biases with regard to class membership was in line with expectations. The majority of the sample (‘Low symptoms’ group) showed a healthy trajectory characterised by consistently low levels of anxiety, and low, but slightly increasing levels of depression. In terms of cognitive biases, the ‘Low symptoms’ group showed low interpretation and memory bias across waves. The ‘Decreasing anxiety symptoms’ group showed moderate and decreasing levels of anxiety and stable low depression. Interestingly, this group showed decreasing interpretation bias, but increasing memory bias. The ‘Comorbid increasing symptoms’ group displayed simultaneously increasing levels of anxiety and depression and similarly showed increasing interpretation and memory bias. While the ‘Comorbid decreasing symptoms’ group showed relatively high levels of anxiety and depression that simultaneously decreased over time, as well as decreasing interpretation and memory bias. These findings shed light on the different pathways of anxiety and depressive symptoms in adolescence and the co-occurring development of cognitive biases.
Consistent with previous longitudinal studies, we found substantial heterogeneity in the developmental trajectories of anxiety and depressive symptoms across adolescence (Allan et al.
2014; Cummings et al.
2014; McLaughlin and King
2015; Miers et al.
2013; Olino et al.
2010; Shore et al.
2018; Rice et al.
2019; Van Oort et al.
2009). Our findings indicate that anxiety and depressive symptoms are highly comorbid throughout this period (Hankin et al.
2016). However, there was evidence to suggest that anxiety and depression trajectories develop as distinct parallel growth processes, which supported previous research (Cummings et al.
2014; Hale III et al.
2009). In addition, we found that the development of interpretation and memory biases, matched the class trajectories of anxiety and depression. These findings provide insight into the potential risk and protective factors that may contribute to levels of anxiety and depressive symptoms in adolescence. The four distinct class trajectories identified and their associated cognitive biases are discussed below.
The ‘Low symptoms’ group displayed consistently low levels of anxiety and depression, with a slight increase in depression over time. This group are perhaps representative of the non-clinical proportion of adolescents assessed in the study, indicating a healthy pathway of development for adolescents with low-risk of developing psychopathology. Given the age range of our sample, this pattern is consistent with evidence in the literature, which suggests that the onset of anxiety often occurs during childhood, while depression tends to develop during adolescence (Hankin et al.
1998; Merikangas et al.
2010; Roza et al.
2003). The ‘Low symptoms’ group displayed a small decrease in interpretation bias and a small increase in memory bias over time. The increase in negative memory bias is perhaps linked to the slight increase in depressive symptoms, suggesting that negative memory bias may be more closely associated with depression than anxiety. This is in line with previous research, which shows that depressive symptoms are associated with enhanced recall for negative compared to positive self-referent information (Lau and Waters
2016; Platt et al.
2016). However, compared to the other groups, adolescents with ‘Low symptoms’ showed lower levels of negative interpretation bias and negative memory bias, suggesting that their bias towards positive, as opposed to negative processing, may act as protective mechanisms against the development of anxiety and depression.
Twenty-five percent of the sample showed elevated symptoms of anxiety and depression at some point, which supports previous epidemiological evidence (NHS Digital,
2017). The ‘Decreasing anxiety symptoms’ group, showed moderate but decreasing levels of anxiety and stable low depressive symptoms. This developmental trajectory indicates that anxiety was more severe during early adolescence, but as anxiety levels decreased over time, symptoms became level with depression. However, there was a non-significant increase in depressive symptoms in this group, which matched the magnitude of increase in the ‘Low symptoms’ group. This could be attributed to low power, due to the small number of participants in the ‘Decreasing anxiety symptoms’ group. This trajectory is consistent with Pathway one outlined in the Multiple Pathways Model (Cummings et al.
2014), which describes youth that have a predisposition for anxiety that later becomes comorbid with depressive symptoms. Furthermore, adolescents in the ‘Decreasing anxiety symptoms’ group showed a decrease in interpretation bias, which suggests that less negative (or more positive) interpretations of ambiguous scenarios are protective factors associated with decreasing anxiety levels. This is consistent with a recent meta-analysis that showed a robust association between high anxiety and negative interpretation bias in children and adolescents (Stuijfzand et al.
2017). In contrast, negative memory bias increased over time in the ‘Decreasing anxiety symptoms’ group, which was a consistent finding across the ‘Low symptoms’ and ‘Comorbid increasing symptoms’ groups, suggesting that this is an adolescent-typical effect.
The ‘Comorbid increasing symptoms’ and ‘Comorbid decreasing symptoms’ groups provide support for Pathway two of the Multiple Pathways Model, where anxiety and depressive symptoms co-develop simultaneously. Cognitive biases may reflect transdiagnostic risk factors in this pathway (Cummings et al.
2014). In our sample, the ‘Comorbid increasing symptoms’ group showed increasing levels of anxiety and depression over time, as well as increasing negative interpretation and negative memory biases. Whereas the ‘Comorbid decreasing symptoms’ group, showed initially high but decreasing levels of anxiety and depression, and decreasing negative interpretation and negative memory biases over time. Therefore, the cognitive bias pathways matched the direction of symptoms, indicating that biases are key risk factors associated with anxiety and depressive symptoms. Overall, these two groups showed the most elevated symptoms of anxiety and depression and displayed more negative interpretation and memory bias, compared to the other groups. This suggests that anxiety and depression share a high degree of overlap and common risk and protective mechanisms that may worsen or improve overall adjustment in adolescence. Thus, our study provided evidence for Pathway one and two outlined in the Multiple Pathways Model. We found no evidence for Pathway three, which is characterised by a diathesis for depression and later comorbid anxiety. However, we did not expect to observe evidence for Pathway three, which is thought to distinguish older adolescents and adults.
The development of cognitive biases in relation to anxiety and depression is a novel aspect of the present study. We found that the development of interpretation and memory biases corresponded to the anxiety and depressive symptom trajectories in the four distinct groups, suggesting that they are closely related. Increasing negative memory bias over time was present in all groups, apart from the ‘Comorbid decreasing symptoms’ group, and therefore may reflect a normative effect. Adolescence is an important phase of developing a sense of identity and studies have shown that negative self-evaluations are highly prevalent amongst depressed adolescents (Orchard et al.
2019). Thus, as adolescents get older they may become more sensitive to negative self-perceptions and recalling negative thoughts. Furthermore, all groups showed decreasing social and non-social interpretation bias over time, except for the ‘Comorbid increasing symptoms’ group. Together, this pattern suggests that negative interpretation bias is more prominent in younger adolescence, while negative memory bias is more characteristic of mid-adolescence.
Social bias was predominantly more negative than any other bias, indicating that negative interpretations of social scenarios were relatively high across all groups. This is perhaps related to adolescence in general, as this developmental period is characterised by significant neurodevelopmental changes and heightened sensitivity, in particular to the social environment and peers (Fuhrmann et al.
2015; Nelson et al.
2016). Heightened levels of social bias in our sample are therefore likely to reflect changes in information-processing, as adolescents become more sensitive to social input from peers and are more vulnerable to negative social situations. Negative social bias was highest in the ‘Comorbid decreasing symptoms’ group at W1 and this decreased over time. In the ‘Comorbid increasing symptoms’ group, social bias became more negative across waves, reflecting the only increase across groups. Furthermore, in this at-risk group, social bias was markedly higher than non-social bias and memory bias across all three waves. This pattern indicates that negative social biases may play a role in increasing comorbid anxiety and depression symptom trajectories and are perhaps potential targets for early interventions.
One challenge for early intervention approaches is being able to identify adolescents who are the most at-risk (e.g., the ‘Comorbid increasing symptoms’ group) from those who may show a natural decrease in symptom trajectories over time (e.g., the ‘Comorbid decreasing symptoms’ group). It is important to note that because we only assessed three time points we cannot say with any certainty that those in the ‘Comorbid decreasing symptoms’ group would continue to show a decrease in symptoms. Previous research using more time points, suggests that some at-risk adolescents display high and fluctuating symptoms (Shore et al.
2018). Therefore, it is possible that this seemingly improving group would show a spike in symptoms at later stages of adolescence, particularly as their symptoms and negative biases remained at a similar level to the ‘Comorbid increasing symptoms’ group. Further investigation of other factors such as SES, gender, age, school type, peers, friendships, social support, family environment, or genetics, in addition to cognitive biases, may help to identify adolescents who are at greatest risk of comorbid increasing anxiety and depression trajectories (Field and Lester
2010; Van Harmelen et al.
2017), and thus benefit the most from early interventions targeting cognitive biases.
We found evidence in our sample that being female and having lower SES were risk factors for elevated symptoms of anxiety and depression, which is consistent with previous research (Shore et al.
2018; Van Oort et al.
2009). Females were more likely to be in the ‘Decreasing anxiety symptoms’, ‘Comorbid increasing symptoms’, or ‘Comorbid decreasing symptoms’ groups than the ‘Low symptoms’ group. Furthermore, adolescents with higher SES were more likely to be in the ‘Low symptoms’ and ‘Decreasing anxiety symptoms’ groups, suggesting lower risk to psychopathology over time, perhaps due to greater access to resources or certain environmental or familial advantages. The ‘Comorbid increasing symptoms’ group showed the lowest level of SES, therefore lack of resources or familial disadvantage may have contributed to this poor trajectory. This may also reflect a gender interaction, as previous research has found that the greatest negative impact of low SES was found in older adolescent girls (Patalay and Fitzsimons
2018). This highlights the importance of taking into consideration demographic characteristics as well as social, biological, or cognitive factors that might influence risk and resilience for psychopathology, which may be important for identifying adolescents and developing more targeted interventions.
Given that symptoms of anxiety and depression often persist beyond childhood, through adolescence and into adulthood, prevention and early intervention programmes are key. The findings of the present study identify potential cognitive biases that may be useful to target in the development of anxiety and depressive symptoms in early to mid-adolescence. However, these results should be interpreted with caution and future research directly targeting these cognitive mechanisms in intervention studies and randomized controlled trials in adolescents is needed, particularly in clinical and at-risk youths. Further, as our sample included healthy adolescents with elevated symptoms, rather than a clinical sample, we can only make inferences about the early development of anxious and depressive symptoms and the suggestion that cognitive biases may exacerbate psychopathology (Fox and Beevers
2016). Nonetheless, this study has the potential to inform future research on the trajectories of anxiety and depressive symptoms in early to mid-adolescence in a normative sample and associations with the development of interpretation and memory biases.
The present study has a number of strengths. First, we used a longitudinal design to assess adolescent developmental psychopathology across multiple time points, within a large normative sample. This provides valuable insights into the symptom trajectories of anxiety and depression in healthy adolescents and the development of psychopathology in the early stages. Second, to the best of our knowledge, this study was the first to investigate cognitive biases longitudinally. This novel aspect of the study sheds light on the development of interpretation bias and memory bias during adolescence and highlights associations between cognitive biases and trajectories of anxiety and depressive symptoms. Third, we used a person-oriented approach (i.e., GMM), which allowed us to identify distinct subgroups of adolescents with varying levels and rates of change in anxiety and depression. Finally, we were able to retain a large sample across three waves and used a wide range of behavioural and self-report measures. Therefore, the variability within our sample allowed us to investigate developmental trajectories of anxiety and depression, and advance current knowledge of the cognitive factors associated with adolescent psychopathology.
However, there are several limitations to the study worth noting. One limitation is that there may be other biological, social, or cognitive factors (e.g., genetics, temperament, executive functions, peers, friendships, social support, family environment or parental behaviours) associated with developmental trajectories of anxiety and depression in adolescence (Field and Lester
2010; Van Harmelen et al.
2017). In the present study, we focused on cognitive biases, based on previous literature that highlights the importance of attention, interpretation, and memory biases in youth (Lau and Waters
2016). We were unable to include attention bias, which may be important, in our analysis due to low internal consistency and poor psychometric properties of the Dot-probe task (Booth et al.
2019). It is important for future research to examine multiple cognitive biases, such as the development of attention, interpretation, and memory bias longitudinally, in order to assess cognitive models that emphasise the importance of these information-processing biases in anxiety and depression in greater detail.
Another limitation is that we did not examine the trajectories of specific anxiety disorder symptoms (i.e., generalised anxiety, social anxiety, separation anxiety etc.), which may have influenced the results. Less research has examined the developmental trajectory of anxiety, yet it is likely that different anxiety subtypes show different developmental pathways (Cummings et al.
2014; Hale III et al.
2009; Van Oort et al.
2009). However, we were unable to assess anxiety sub-types, due to not enough power to investigate numerous models. In addition, we used the short version of the RCADS questionnaire, which is designed primarily to assess anxiety symptoms as a whole. Future studies in adolescence could test larger samples and use the long version of the RCADS to disentangle the trajectories of anxiety sub-types further.
Finally, growth mixture modeling has received some criticism in the literature (Petersen et al.
2019). One criticism to this approach is that growth mixture models are essentially clustering procedures yielding sample specific results. However, the class trajectories identified in our sample are in line with previous studies that have investigated the developmental pathways of anxiety and depression in adolescents (Cummings et al.
2014), supporting the validity of our findings. More large-scale longitudinal studies in normative samples are needed to replicate our findings. Another problem with this approach is that growth mixture models are not the preferred method to study temporal relations. Whilst these relations were not the primary focus of the current study, prior work suggests that cognitive biases play a key role in the development and maintenance of internalising disorders (Lau and Waters
2016; Mathews and MacLeod
1994; Muris and Field
2008). However, these relationships have not been investigated in adolescents longitudinally, which would be an important direction for future research. In addition, further research using analytical approaches such as cross-lagged panel models or experimental designs such as Cognitive Bias Modification (CBM) studies or randomized controlled trials that target interpretation or memory biases in adolescents would provide a deeper insight into temporal relations and directionality (Lau and Pile
2015). Finally, future longitudinal research should investigate the developmental period of younger children or later adolescence to provide a more holistic picture of when cognitive biases are likely to develop and how they are associated with anxiety and depression over the lifespan (Field and Lester
2010).
In summary, this study investigated the development of anxiety and depressive symptom trajectories in adolescence and the co-occurring development of cognitive biases. We found evidence for four distinct developmental classes of anxiety and depression and demonstrated that interpretation and memory biases are risk and protective factors associated with symptom trajectories. This novel study sheds light on the longitudinal development of social interpretation bias, non-social interpretation bias, and memory bias across adolescence. Negative social interpretation bias was particularly high in our sample, which may reflect high sensitivity to peers or the social environment, behaviours typically observed in this age group. Additional longitudinal research investigating cognitive biases and further replication of this approach with larger samples is required to validate the class trajectories identified in our sample. The current study advances our understanding of the developmental trajectories of psychopathology in early to mid-adolescence and has the potential to inform future research on potential cognitive mechanisms to target for prevention and early interventions.