Increases in Circulating Cortisol during the COVID-19 Pandemic are Associated with Changes in Perceived Positive and Negative Affect among Adolescents

Catastrophic global events can have reverberating consequences on mental health for extended periods. Major events like natural disasters and war are known to have both acute and chronic impacts on physiological stress responses, as well as perceived psychological distress and mental health (Miller et al., 2007; Stalder et al., 2017). For over a year now, the Coronavirus Disease 2019 (COVID-19) pandemic has spread across the world, killing over 5 million people, damaging economies, overburdening healthcare systems, and physically and psychologically harming individuals (Courtney et al., 2020; Taylor et al., 2021). The United States was hit particularly hard by the pandemic early on, harboring nearly 20% confirmed cases and 15% of global deaths (https://​coronavirus.​jhu.​edu/​). The gravity of the situation during the first wave forced intermittent, sometimes prolonged school closures across the country (Auger et al., 2020), leaving youth disconnected from critical resources (Fisher et al., 2020; Viner et al., 2020).

Notably, a rapidly growing body of research consistently shows that during the pandemic, youth are reporting decreased social support (Poletti & Raballo, 2020), challenges with shifting learning platforms (Petretto et al., 2020), and generalized decreases in mental health (Cost et al., 2021; Courtney et al., 2020; Glynn et al., 2021; Liang et al., 2020; Racine et al., 2021). Multiple reports suggest that globally, lockdowns and the associated outcomes (e.g., reduced access to services, social isolation) are associated with increased depression, anxiety, irritability, and even alcohol and substance use among adolescents (Newlove-Delgado et al., 2021; Panchal et al., 2021; UNICEF Innocenti Research Centre, 2021). This is crucial given that mental health disorders often develop during adolescence (Bitsko et al., 2018; Merikangas et al., 2010), and given that early stressful and traumatic experiences are known to have prolonged consequences for physical and mental health across the lifespan (Conway et al., 2016; Miller et al., 2011).

Studies examining the impact of the pandemic on mental health are critical to informing public health responses and plans to address the mental health needs of youth. However, most studies reporting effects of the pandemic are based on ad hoc surveys distributed after local lockdowns were in place; there is limited longitudinal data from which researchers can determine shifts from baseline in response to the events of this global crisis. One such study explored self-reported psychological wellness among youths prior to and 2 months after lockdowns began, and found generally poorer outcomes during the pandemic that were more severe among female youths who felt socially disconnected from their peer groups (Magson et al., 2021). Another recent study leveraged a longitudinal population-based data set tracking self-reported mental health outcomes among youths in England in 2017 and again in 2020 shortly after the first wave of COVID-19 lockdowns (Newlove-Delgado et al., 2021; Vizard et al., 2020). The study found a significant increase in the prevalence of mental health conditions among adolescents following lockdowns, jumping from 10.8% to 16.0%. This was coupled with reports of reduced sleep quality, loneliness, and fear. These studies are highly informative and suggest a strong shift toward negative mental health outcomes as a result of the pandemic. However, such self-report measures do not directly address the biological impact of stress on an individual. That is, perceived changes in psychological distress do not necessarily imply specific physiological mechanisms underlying changes in mood and behavior (Karlén et al., 2011). Researchers have formally called for longitudinally-measured physiological data to better understand the long-term impacts of the pandemic on mental health (Dantzer et al., 2020; Wade et al., 2020).

One potential target for assessing changes in physiological stress during the pandemic is measuring cortisol, a neuroendocrinological marker of stress (Burnard et al., 2017; Gibb et al., 2020). Research has frequently linked variability in cortisol to aspects of mood, behavior, and even brain morphology in developing youth (Joos et al., 2019; Miller et al., 2007; Russell et al., 2012). Generally, cortisol concentrations tend to increase in response to acute stressors, and basal levels of cortisol can become dysregulated in cases of chronic stress (Miller et al., 2007; Russell et al., 2012; Stalder et al., 2017). Greater increases in cortisol in response to a stressor (i.e., larger increases from basal levels following stress exposure) tend to be associated with greater perceived psychological distress, including increasing negative affect, depressive and anxiety symptoms, and post-traumatic stress (Phan et al., 2020; Stalder et al., 2017; Steudte-Schmiedgen et al., 2015).

Longitudinally measuring changes in circulating cortisol surrounding uncontrollable or unforeseen events was once nearly impossible because measurements are most commonly taken from saliva or blood (serum), both of which are highly time-sensitive and can rapidly change over the course of just a few minutes. Thus, to accurately measure, one would have to anticipate that the event was coming and take measurements immediately before and after. However, researchers have more recently begun to leverage the strength of measuring hair cortisol concentrations (HCC), as hair samples reliably encapsulate mean cortisol levels over time (Burnard et al., 2017; Lee et al., 2015). More specifically, hair grows at an average rate of 1 cm per month, thus each 1 cm segment of hair neatly catalogues the average concentration of circulating hormones over a one-month period (Short et al., 2016). These measurements are most reliable up to 5 cm from the scalp (Russell et al., 2012), thus researchers are able to retrospectively explore circulating hormones within an individual over an approximately 5 month period of time prior to biosample collection.

Critically, HCC has been repeatedly linked to mental health outcomes in adolescents and adults, though with some mixed findings. For example, a systematic review showed that major depressive disorder is associated with increased HCC, whereas anxiety disorders are generally characterized by decreased HCC, and post-traumatic stress disorder is denoted by an initial increase followed by a dramatic decrease in HCC to below baseline levels (Staufenbiel et al., 2013). Further, individual differences are known to impact the nature of HCC-psychopathology associations. Lehrer et al. (2020) demonstrated a robust link between greater perceived stress and elevated HCC, though this effect was modulated by resilience. Despite the heterogeneity in findings, there is a general consensus in the field that HCC is likely related to mental health outcomes and plays a critical role in the development of emotional neural circuitry, thereby setting the stage for risk and resilience to mental health outcomes across the lifespan (Kamin & Kertes, 2017).

Because HCC is able to historically catalogue physiological stress responses, it is possible to retrospectively and robustly assess changes in average cortisol concentrations surrounding natural, uncontrollable stressful events (Mewes et al., 2017; Steudte-Schmiedgen et al., 2015, 2016). For example, Gao et al. (2014) examined physiological stress following a major earthquake. Survivors exhibited significantly increased HCC shortly after the event, followed by a gradual decrease in HCC which finally normalized after approximately one year. In another study, Dajani and colleagues (2018) studied HCC among war-affected adolescents who were refugees displaced to another country. Youth presented with disparate patterns of cortisol dysregulation based on perceptions of mental health and personal safety. Specifically, HCC was elevated as a function of increasing perceived insecurity and exhibited significant dysregulation (i.e., unreliable change over time) among youth with greater post-traumatic stress disorder symptomatology. From these and other examples, it is clear that HCC is a promising approach for assessing physiological stress responses to uncontrollable events. Moreover, HCC can be linked to perceived changes in psychological well-being, providing scientists a powerful lens for probing the biological underpinnings of psychological distress following major stressors.

The present study focuses on physiological stress and perceived changes in mood among adolescents during the COVID-19 pandemic. We leveraged the historical sensitivity of HCC to assess changes in cortisol from pre-lockdown time periods to post-lockdown time periods. We expected that HCC would be significantly greater during post-lockdown periods relative to pre-lockdown periods. Additionally, we assessed retrospective perceptions of changes in mood, specifically positive and negative affect, prior to- versus since local lockdowns. Based on the extant literature, we hypothesized that youth would report increases in negative affect and decreases in positive affect during the pandemic. Finally, we assessed the extent to which perceived changes in affect and changes in HCC were related to one another. Because there is limited knowledge on the direction of effects, we examined three different associational models: 1) changes in affect and HCC are correlated, 2) changes in affect predict changes in HCC, and 3) changes in HCC predict changes in affect.

The greatest strength of this study was the measurement of HCC prior to and after local lockdowns began. By using HCC, we were able to reliably index shifts in circulating cortisol levels among youth surrounding a pivotal, high-stress period of the COVID-19 pandemic. As expected, we saw robust increases in HCC following local pandemic-related lockdowns. Moreover, the degree of change in circulating cortisol was dependent on baseline levels, such that youth who had higher HCC prior to lockdowns tended to show smaller increases in circulating cortisol than youth with lower baseline HCC. Our findings were well-aligned with the extant literature showing increases in cortisol following a major stressful event (e.g., Dajani et al., 2018; Gao et al., 2014). Additionally, our data corroborate previous findings suggesting that individuals with lower relative baseline cortisol levels tend to show greater increases following a stressor (Joos et al., 2019; Steudte-Schmiedgen et al., 2015).

These data uniquely contribute to our knowledge of pandemic-related stress from a neuroendocrinological approach. Notably, physiological stress responses are markedly different depending on the nature of the stressor, and researchers have yet to characterize shifts in circulating cortisol levels surrounding a global pandemic. Meta-analyses and large-scale reviews of physiological stress responses have shown that stressful events which threaten the physical self (e.g., bodily harm/war) versus the social self (e.g., divorce) have very different presentations (Miller et al., 2007; Russell et al., 2012; Stalder et al., 2017). Additionally, uncontrollability, unpredictability, chronicity, and perceived trauma and psychological distress each contribute to physiological responses. The COVID-19 pandemic is unique relative to many other studied catastrophic events given the multiplicative types of stressors it presents. For instance, there are threats to the physical self (e.g., falling ill) and to the social self (e.g., loss of friendships/social supports), unpredictability and uncertainty surrounding the progression and duration of the global spread (e.g., poor government organization and communication), and psychological distress and disruptions to daily life due to lockdowns (Courtney et al., 2020). This is compounded by the potential for other forms of harm, such as reduced access to healthcare and social services, and increased risk of child abuse in the home going unnoticed (Fisher et al., 2020; Park et al., 2020; Patrick et al., 2020).

In the current study, we also explored the extent to which shifts in HCC and affect were related to one another. Changes in cortisol and affect were all well correlated, which was in accordance with prior literature (e.g., (Peeters et al., 2003). We found that changes in HCC robustly predicted increases in negative affect and decreases in positive affect, though affect did not significantly predict physiological stress responses. These data suggest that physiological stress responses may, at least in part, drive changes in affect, though this requires further investigation. If these findings are replicable in larger, more sociodemographically diverse samples, they could allude to potential avenues for intervention or prevention of negative mental health outcomes. Namely, pharmacologic regimens that help regulate or balance hormone levels may tamp subsequent maladaptive psychological impacts following major stressful events. Others have similarly suggested that dysregulated levels of circulating cortisol over prolonged periods may cause new or worsening psychological symptoms, even in otherwise “healthy” individuals (Staufenbiel et al., 2013).

One potential avenue by which elevated cortisol may contribute to increasing psychological symptoms is through its impacts on brain structure and function. Ulrich-Lai and Herman (2009) detailed the manner in which limbic regions of the brain, which span cortical and subcortical areas, contribute to the homeostatic balance in neuroendocrine and autonomic functioning in day-to-day life. However, under times of extreme or prolonged stress, cortisol secretions become dysregulated and actually cause damage to these neural structures via dendritic modifications and atypical expression of neurotransmitter receptors. These fundamental shifts in physiology affect subsequent neural functioning and in turn, psychological well-being, including anxious and depressive symptomatology (Gillespie & Nemeroff, 2005; Ulrich-Lai & Herman, 2009). It is possible that, in the case of the COVID-19 pandemic, the prolonged stress of local lockdowns and fear for the health of oneself and loved ones may be gradually affecting neural structure and function in these youth, thereby leading to poorer psychological wellbeing. However, further work is needed to determine the extent to which brain structure and function may have been altered during the current pandemic. This would be an excellent area of research for those with access to large longitudinal datasets with pre-pandemic baseline measures available.

Without a doubt, the COVID-19 pandemic is a unique phenomenon that warrants continued investigation of prolonged stress responses and changes over time. Presently, these data provide a window into the underlying physiological stress responses surrounding adolescents’ experiences early in the pandemic. Other recently published works have detailed prolonged self-reported psychological outcomes among youths as the pandemic has continued through additional waves. For instance, Ravens-Sieberer et al. (2021) assessed mental health among children and adolescents in Germany during the first wave of the pandemic, and again almost one year into the pandemic. The study found generally increased reports of depressive and psychosomatic symptoms at the second measurement point, though risk (e.g., social disadvantage) and resilience factors (e.g., positive family environment) certainly moderated these outcomes. Further investigation should follow up with youths and assess the degree to which prolonged neuroendocrinological dysregulation is associated with these mental health outcomes.

Our data are consistent with a growing body of literature reporting robust increases in psychological distress and negative mood during the pandemic in a multitude of populations (Duan et al., 2020; Hyland et al., 2020; Marques et al., 2020; Rogers et al., 2021). It is rather intuitive that negative affect would have significantly increased given the circumstances of the lockdowns. The pandemic forced municipalities across the country to close schools and businesses, restrict access to public areas like parks, and recommend that individuals confine interpersonal interactions to members of their own households (Coyne et al., 2020; Fisher et al., 2020). These burdens pervasively impacted youth in the present study as evidenced by reported levels of recent concern. An overwhelming majority of youth indicated that at least one pandemic-related stressor had concerned them in the past week, with an average of roughly seven-to-eight endorsed stressors per participant. Stress burden is known to concomitantly increase with psychological distress and negative mood (Mohler-Kuo et al., 2021; Park et al., 2020; Rajkumar, 2020; Taylor, 2019).

Contrary to our hypothesis, we did not see changes in positive affect in the sample overall. Rather, we found that significant decreases in positive affect were specific to youth who reported a greater degree of recent concern over pandemic-related stressors. This finding critically identifies youth who may be more at risk of suffering negative mental health outcomes due to the pandemic. Positive affect has long been associated with resilience in the face of adversity; those who report greater positive affect tend to have better mental and physical health outcomes, even in light of major stressors like life-changing medical diagnoses (Lord et al., 2015; Murphy et al., 2017). Thus, it is possible that youth in the current study who experienced greater concern and greater decreases in positive affect may be less resilient to, and more affected by, the stress induced by the COVID-19 pandemic. These youth may be prone to anxious or depressive moods, and thus may benefit from targeted services that provide more structured support to reduce concerns. For example, safe access to social support networks and improved approaches to online or hybrid learning could help alleviate concerns with maintaining friendships and returning to school (e.g., Magson et al., 2021).

The present study is not without limitations. First, affect was measured retrospectively, whereas HCC was assessed longitudinally making it difficult to equivalently compare temporal relationships. Ideally, researchers would have baseline measures of typical affect prior to the onset of a stressful event; however, this is a common shortcoming of research focused on the impacts of unpredictable catastrophic events. Large-scale studies like the Adolescent Brain Cognitive Development (ABCD) study (Casey et al., 2018), which collect longitudinal measurements of emotional, mental, and physical health, could be critical in affording the field solid baseline measurements of individuals in the event of unpredictable major stressors. Second, the sample in the current study was relatively small and sociodemographically homogenous, and may not be representative of the greater diversity of youth across the country. Further efforts should be taken to study socioeconomically and demographically diverse individuals across multiple geographic regions to better characterize the effects of the pandemic. In particular, capturing variability in HCC among males, for whom our sample was very limited due to the long hair length requirements, would be beneficial. In addition, testing our proposed latent change score models in larger samples would be ideal. It is a common misconception that such models require large (N > 200) samples to estimate properly (Iacobucci, 2010; Kim, 2005; Marsh et al., 2004; Wolf et al., 2013). That said, testing the models in larger and more diverse samples would address the generalizability of our findings to the broader public. Finally, for our hair samples, we relied on the notion that hair grows an average of 1 cm per month. Hair growth rates may vary from person-to-person (Burnard et al., 2017), and our presumed timeframes captured by the different hair segments may vary slightly between individuals.