Resilience to Early Family Risk Moderates Daily Stress and Affect on Cortisol Diurnal Slope associations: A 14-Day Ecological Momentary Assessment Study

Objective This study examined whether resilience capacity moderates the impact of daily perceived stress and affect on cortisol diurnal slope among relocated emerging adults. Methods Relocated undergraduates (N=98; aged 18-25 years) were recruited from three groups: Resilient, Vulnerable, and Control. Mixed-effects models were used to test the unique effects of perceived stress, negative affect, and positive affect x group interactions on diurnal cortisol slope across 14 consecutive days. Results The Resilient group did not moderate the associations between daily stress or affect on cortisol diurnal slope. Instead, both the Resilient and Vulnerable groups with early family risk, showed a steeper diurnal slope unique to higher stress and a atter slope unique to higher NA. Conclusions Results suggest that an adverse early family life was signicantly associated with altered cortisol diurnal slope outcomes to stress (i.e., demand) and negative affect (i.e., distress). These associations were not attenuated by current resilience capacity.


Introduction
Cortisol is a stress-sensitive steroid hormone regulated by the hypothalamic-pituitary-adrenal (HPA) axis that prepares the body to mobilize energy in managing threat (Ross et al., 2014). In humans, the cortisol diurnal rhythm is the daily circadian pattern of cortisol secretion peaking around 20 -30 minutes after awakening followed by a decline throughout the day to its nadir 2 to 3 hours after sleep onset (Ross et al., 2014). Cortisol's diurnal slope is derived from this decline from wake cortisol levels to pre-sleep cortisol levels, with a steady decline indicating healthy cortisol regulation (Segerstrom et al., 2014). Flattened cortisol diurnal slopes occur when there is lower wake cortisol levels or sustained elevation of cortisol levels at pre-sleep (Miller et al., 2017). Flattened cortisol diurnal slopes are linked to poorer emotional and physical health such as cardiovascular diseases in a meta-analysis of 80 studies across all ages ranging from children to older adults (Adam et al., 2017).
Chronic stress such as a risky early family background can in uence the development and adult functioning of the HPA-axis (trait-like variation). Cortisol diurnal slope also responds to acute changes in individuals' everyday experiences (state-like variation), notably in response to stress and affect (Ross et al., 2014). Although daily stress and affect co-occur, they are conceptually distinct. Together, stress, negative affect (NA), and positive affect (PA) capture a range of daily experiences including situational demand (stress), negative experiences (NA), and positive experiences (PA) (Selye, 1976 & Hoyt, 2018). Individuals with high resilience capacity, de ned as the trait of bouncing back from hardship (Smith et al., 2008), evaluate challenges as manageable and respond better to stress. However, few studies have examined the role of resilience capacity, stress, and affect with cortisol diurnal slope on a daily basis, with most studies focussing on momentary cortisol levels or reactivity. To our knowledge, no studies have examined whether resilience capacity moderates the links of naturally occurring daily stress and affect with cortisol diurnal slope.

Early Family Risk on Cortisol Diurnal Slope
Research on the impact of early family risk as a type of chronic stress on cortisol diurnal slope has commonly found a attened diurnal slope with elevated cortisol throughout the day among exposed children, attributed to a desensitization of the HPA-axis from over-engagement to recurring stressors (Gunnar & Vazquez, 2001). These studies typically focussed on participants from backgrounds with highly adverse or abusive features, including children raised under institutional neglect (Carlson & Earls, 1997), with childhood trauma (O'Connor et al., 2020), or with Child Protective Services involvement (Bernard et al., 2015). Recent studies included parenting features such as poor parental monitoring (Martin et al., 2014) and maternal neglect (Buhler-Wassmann et al., 2020).

Impact of Daily Acute Stress and Affect on Cortisol Diurnal Slope
Acute stress differs from chronic stress in the impact on cortisol diurnal slope but is rarely uniquely studied (Klein et al., 2016). Yet, the daily variations in cortisol to momentary changes are important to study as they exert downstream biological processes relevant to disease (Ross et al., 2014). A naturalistic daily study showed cortisol diurnal slopes are steepened due to higher wake cortisol levels among older adults who reported higher frequency of stressors (Stawski et al., 2013). Overall, there is a lack of daily studies examining the relationship between acute stress and cortisol diurnal slope, with most examining momentary cortisol levels and nding increases during the anticipation and experience of stressors (Smyth et al., 1998;van Eck et al., 1996), and to higher perceived stress (Sladek et al., 2016). Higher momentary and wake cortisol levels with higher stress may be explained by the higher engagement of HPA-axis to meet the demands of the situation (Wong et al., 2012). This adaptive response of diurnal cortisol activation if sustained can result in long-term physical and mental health issues (Miller et al., 2007).
Beyond stress, NA and PA in uence cortisol outcomes. Emotional responses to stressors, rather than chronic or daily stressors, predicted higher cortisol secretion levels in some (Jacobs et (Doane & Adam, 2010). However, affect may be in uenced by antecedent stressors (Stawski et al., 2013). Hence, separate assessment of affect and stress and simultaneous examination in one model is necessary to distinguish their unique in uences on cortisol diurnal slope.
Higher NA levels are associated with atter cortisol diurnal slope among adolescents (DeSantis et al., 2007;Hoyt et al., 2015) and among adults (Adam et al., 2006). The experience of sadness manifested in social withdrawal and behavioural inactivity may result in lower wake cortisol levels or the experience of anger and tension may result in higher pre-sleep cortisol levels. The few daily studies that examined PA and cortisol diurnal slope found steeper diurnal slopes with higher PA among midlife healthy adults (Miller et al., 2016) and high school students (Hoyt et al., 2015), but no impact of PA on wake cortisol levels. The steeper decline in cortisol slope in these studies not driven by a higher wake cortisol level was interpreted as indicative of healthy functioning.

Resilience Capacity on Cortisol Diurnal Slope
To our knowledge, no studies have investigated if psychological resilience moderates daily stress/affect associations with cortisol diurnal slope. However, there are ndings suggestive of the protective role of resilience or resilience-related constructs on cortisol metrics. Adults with higher psychological resilience had a weaker association between perceived stress and hair cortisol levels over 3 months compared to those with low resilience (Lehrer et al., 2020). High-risk individuals with high emotional regulation show minimal dysregulation in average basal cortisol across time (Kliewer et al., 2009). Among children of parents with HIV, resilience was associated with steeper cortisol diurnal slopes via less experienced stigma (Chi et al., 2015).

Aims and Hypotheses
The relations between resilience capacity, momentary stress, and affect with daily diurnal cortisol slopes remains underexplored. Further, resilience capacity should be most helpful in the presence of risk, which often is not captured in resilience studies (Colich et al., 2020). Using retrospective reports of family risk as an index of prior adversity, our study recruited participants from three groups: (1) Resilient, including individuals who originated from risky families and reported high current resilience capacity, (2) Vulnerable, including individuals from risky families but with low reported current resilience capacity, and (3) Control, including individuals with low early family risk and average current symptoms of anxiety and depression. Participants were repeatedly assessed in their daily stress, affect, and cortisol at speci c timepoints using ecological momentary assessments (EMA). All participants were emerging adults who relocated for tertiary studies so that their daily experiences represent a known stressful transition period (Arnett, 2000).
We hypothesized that 1a) across the sample, a steeper, negative cortisol diurnal slope would be associated with higher perceived stress, higher PA, and lower NA, 1b) on average, the Resilient group would demonstrate a similar diurnal cortisol slope as the Control group whereas the Vulnerable group would demonstrate a atter diurnal slope than the Resilient or the Control groups. Our study further explored whether the relation between stress and affect on diurnal cortisol slope is moderated in the Resilient group compared to the similarly high-risk Vulnerable group. Speci cally, compared to the Vulnerable group, the Resilient group individuals were expected to have a weaker association between 2a) stress and diurnal cortisol slope, b) NA and diurnal cortisol slope, and c) PA and diurnal cortisol slope.   Table 2 and 3). A-priori power analysis conducted through G*Power (Faul et al., 2007) indicated that with α=0.05, 10 total predictors, and testing two predictors at once (appropriate as we have three groups), 75 participants, assuming a 75% compliance rate of 2 daily cortisol assessments across 14 days and intraclass correlation coe cients (ICCs) of 0.20 or 0.40 for cortisol provides 175 -315 effective independent observations which provides 80% power to detect a medium effect size (Cohen's f 2 = .15) or a small-to-medium effect size (Cohen's f 2 = .05, roughly equivalent to a Pearson's r = 0.20). More participants were recruited to allow for attrition and other aims and outcomes from the broader study. Participants collected saliva samples using a synthetic cotton roll in Salivette tubes (SARSTEDT, Australia) immediately after waking and just before bedtime for 14-consecutive days. Participants recorded the date and time of saliva collection on the tube label, then attached a photo of the label to their morning and pre-sleep surveys. Recorded collection time was validated against the digital photo timestamp and self-reported wake/sleep times to monitor compliance. Participants were given a set of saliva collection compliance behaviour instructions (Supplementary section). They were encouraged to accurately report violations of these instructions, if any, in their pre-sleep surveys. Samples were removed if they did not meet collection criteria. Samples were stored in subjects' home freezers until transport to the lab, where uncentrifuged samples were kept at either -25 or -80°C freezers until analysis. The samples were then sent to the University of Dresden for assay analyses.

Measures
Salivary Cortisol Diurnal Slope. Our study derived diurnal slopes from 2 cortisol samples on each of the 14 days, based on Segerstrom and colleagues' (2014) nding that diurnal slopes measured with this 2sample approach (one at wake time and the other 9pm) correlated .97 and .99 with diurnal slopes measured using 4 and 3 samples, and was better able to detect between-and within-person differences compared to more samples across fewer days. Raw cortisol values were natural log transformed to reduce skewness.
Stress. Perceived stress was measured using a single item "Since the previous survey, how stressful has your day been?" rated from 0 (Not at all stressful) to 10 (Very stressful).
Covariates. Covariates were selected a priori based on literature. Baseline covariates included age (years) and sex (male/female) (

Analysis
Data were analyzed using R version 3.6.3 (R Core Team, 2019). Mixed models were estimated using lme4 v1.1-23 with restricted maximum likelihood. lmerTest v3.1-2 was used for degree of freedom and signi cance testing. All mixed models included a random intercept by participant to address nonindependence and all covariates. Signi cance was set at α = .05, two-tailed. Visual model diagnostics were checked to evaluate assumptions. All assumptions were met. Simple slopes were calculated for any signi cant interactions using high and low levels of perceived stress (4.24, 0.45), NA (2.06, 1.04) and PA (1.57, 3.64), determined based on the 90% and 10% percentiles in our sample.
Hypothesis 1: Main Effects of Perceived Stress, NA, PA, and Group on Cortisol Diurnal Slope (Two-way Interactions). This model tested whether between-person differences in daily perceived stress, NA, PA, and group were associated with cortisol diurnal slope via two-way interactions between daily stress/affect variables and time [wake or pre-sleep].

Hypothesis 2: Moderation of Group on Daily Stress/Affect-Cortisol Diurnal Slope associations (Three-way
Interactions). This model tested the between-group differences in interactions with daily stress, NA and PA on cortisol diurnal slope (i.e., three-way interactions between groups [Resilient; Vulnerable; Control], daily stress/affect, and time [wake or pre-sleep] on cortisol levels). Table 1 shows the descriptive statistics and Supplemental Table 1 shows bivariate correlations. Most participants were female, of Asian descent, and were international students who spent less than a year in Melbourne. Participants exhibited the expected diurnal cortisol pro le with average wake and pre-sleep cortisol of 11.32 and 1.29 nmol/l, which are comparable to healthy subjects in population-based studies (Shin et al., 2011). None of the demographic variables or covariates were signi cantly associated with average cortisol levels except for age (r = -0.28, p = .005) and time spent in Melbourne (r = -0.22, p = .029). Participants reported average levels of daily stress, NA, and PA that were comparable to other young adult samples (Merz & Roesch, 2011;Ong et al., 2016). The Resilient and Control groups did not meaningfully differ (Chen et al., 2018) on anxiety and depression symptoms, both showing lower symptoms than the Vulnerable group (Table 1).

Compliance
The 95 participants provided 4,333 surveys (74% completion rate) and 2,345 viable saliva samples (85% compliance rate) across 14 days, with no signi cant differences across groups (p = .92 and p = .39 respectively). In total, there were 71 (5.5%) saliva samples with reported violations of compliance behaviour, controlled for in analyses.
Hypothesis 1A: Steeper, negative cortisol diurnal slope will be associated with higher perceived stress, higher PA, and lower NA There was a signi cant interaction effect for PA with cortisol diurnal slope (b [95% CI] = -0.09 [-0.17, -0.00], p = .046) where higher levels of PA predicted a steeper, negative diurnal cortisol slope. No signi cant interaction was found for perceived stress (p = .23) or NA (p = .065) with cortisol diurnal slope ( Table 2).
Hypothesis 1B: On average, the Resilient group will demonstrate a similar negative diurnal cortisol slope as the Control group whereas the Vulnerable group will demonstrate a atter diurnal slope than the Resilient or the Control groups.
There were no signi cant differences in cortisol diurnal slope between all three groups (all p ≥ 0.14, Table  2).
Hypothesis 2A: Compared to the Vulnerable group, the Resilient group will have a weaker association between stress and cortisol diurnal slope.
Contrary to prediction, the Resilient and Vulnerable groups did not differ in their interactions with stress on cortisol diurnal slope, indicated by the absence of signi cant three-way interaction (p = .10) ( Table 3).
Compared to the Control group, both the Resilient (p < .001) and Vulnerable (p = .002) groups showed a stronger association between high stress and steeper diurnal slope (Table 3, Figure 2A). Overall, groups signi cantly interacted with stress and cortisol diurnal slope at a small effect size f 2 = 0.009, p < .001.
The steeper decline for the Resilient group at high stress was accompanied by a non-signi cant higher wake cortisol level at high stress compared to low stress (1.87, p = .12), which was also signi cantly higher than the Control group (p = .03). The steeper decline for the Vulnerable group was accompanied by a non-signi cant lower pre-sleep cortisol level at high stress (-0.44, p = .11). In contrast, the Control group showed a signi cant atter diurnal slope with high stress (p = .006) accompanied by lower wake cortisol levels (-0.87, p = .09).
Hypothesis 2B: Compared to the Vulnerable group, the Resilient group will have a weaker association between NA and cortisol diurnal slope.
Contrary to prediction, the Resilient group showed a signi cantly larger interaction effect with NA levels on cortisol diurnal slope compared to the Vulnerable (p = .029) and Control (p = .002) groups, indicated by signi cant three-way interactions (Table 3). Compared to both the Vulnerable and Control group, the Resilient group showed a signi cantly atter diurnal slope at high NA (Table 3; Figure 2B). The Vulnerable group also showed a signi cantly larger interaction effect (p = .020) and a atter diurnal slope at high NA (p = .033) compared to the Control group. Groups signi cantly interacted with NA and cortisol diurnal slope at a small effect size (f 2 = 0.006, p = .003).
The atter slope of the Resilient group was accompanied by both non-signi cant lower wake cortisol (-1.38, p = .09) and higher pre-sleep cortisol levels at high NA (0.20, p = .78) while the attened slope of the Vulnerable group was accompanied by signi cant higher pre-sleep cortisol levels (0.39, p = .04). In contrast, the Control group showed no signi cant changes in cortisol diurnal slope by NA (p = .11).
Hypothesis 2C: Compared to the Vulnerable group, the Resilient group will have a weaker association between PA and cortisol diurnal slope.
Contrary to prediction, the Resilient and Vulnerable groups did not differ in their interactions with PA levels on cortisol diurnal slope, shown by the absence of three-way interaction (p = .87) ( Table 3). Both the Resilient and Vulnerable groups showed no signi cant changes in cortisol diurnal slope to PA (Table   3; Figure 2C). The Control group showed a signi cant steeper negative diurnal slope at high PA (p = .003), although not signi cantly different than either the Vulnerable (p = .061) nor Resilient (p = .30) groups.

Discussion
We found no overall changes in cortisol diurnal slope to higher stress and NA nor overall group differences in basal cortisol diurnal slope. When cortisol diurnal slope changes to stress and affect were examined by group, signi cant differences emerged. The Resilient group did not show the expected moderation effect on cortisol diurnal slope compared to their Vulnerable counterparts. Instead, the Resilient and Vulnerable groups both demonstrated a steeper diurnal cortisol slope with high stress compared to the Control group. Additionally, the Resilient group showed a atter diurnal slope with high NA compared to the Control or Vulnerable groups. Hypothesis 1 was partially supported. Higher PA was associated with a steeper, negative cortisol diurnal slope, consistent with previous research (Miller et al., 2016;Hoyt et al., 2015). There were no signi cant associations between daily stress and NA with cortisol diurnal slope. The absence of a signi cant association for daily stress is consistent with similar null ndings in other studies that controlled for affect as our study did (Jacobs et al., 2007;Smyth et al., 1998;van Eck et al., 1996). Other studies also found null associations between NA with cortisol diurnal slopes or associations in the opposite direction to our hypothesis, requiring the consideration of arousal levels of NA or the interpretation of NA as harmful or not in measuring these associations (Miller et al., 2016;Park et al., 2020). Differences in our ndings compared to existing literature also may be attributable to our collection of 2 saliva samples daily for 14 days, which better characterizes between-subject differences on cortisol slope (Segerstrom et al., 2014). Contrary to our prediction, the Vulnerable group did not exhibit an overall atter cortisol diurnal slope than the Resilient or Control groups, perhaps because their early family risk was not at the severity or chronicity of abuse or maltreatment where atter cortisol diurnal slopes previously have been shown. At high NA levels, the Resilient group showed a signi cantly atter cortisol diurnal slope compared to both the Vulnerable and Control groups. Our assumption was that the better psychological outcomes found among individuals with high resilience capacity compared to those with low resilience capacity (Werner, 1993) would translate to healthier physiological outcomes in adverse situations. The atter cortisol diurnal slope among the Resilient group compared to the similarly high-risk Vulnerable group may be interpreted as a heightened physiological sensitivity to emotional distress, re ecting a physiological cost for psychological adjustment (Brody et al., 2013). These speculations need further exploration.
Nevertheless, both the Resilient and Vulnerable groups showed atter diurnal slope to high NA, which is indicative of the allostatic load of higher family risks resulting in heightened vigilance to threat (i.e., negative experiences) and therefore greater physiological reaction (Dmitrieva et al., 2013)  The non-signi cant associations found for both the Vulnerable and the Resilient groups between PA and cortisol diurnal slope are not unique (Van Eck et al., 1996). However, the steeper cortisol diurnal slope at high PA for the Control group suggests that the riskier family backgrounds of the Vulnerable and Resilient groups may predispose them to less physiological responsivity to PA. It is also possible that the Control group experienced more high arousal PA linked to steeper cortisol diurnal slope while the Vulnerable and Resilient groups had more low arousal PA linked to no signi cant changes in cortisol diurnal slope (Hoyt et al., 2015). Nevertheless, these ndings collectively suggest a unique pathway between resilience capacity and cortisol diurnal slope responses to negative vs positive emotions (Sin et al., 2015). Further, there may be a difference in physiological sensitivity to affective valence between those from high and low risk families, such that the high risk Vulnerable and Resilient groups were more physiologically sensitive to demands and distress whereas the low risk Control group was more responsive to positive emotions.
Overall, the mostly similar cortisol diurnal patterns between the Vulnerable and the Resilient groups suggest that their common riskier early family backgrounds may be impacting on physiology more than current resilience capacity. This conjecture nds a basis in studies that proposed the programming of HPA-axis at sensitive early periods of development (Koss & Gunnar, 2018). Additionally, the different cortisol diurnal slope changes to stress (i.e., demand) compared to NA (i.e., distress) support the need to look at these variables separately in future studies. Despite the groups not signi cantly differing in diurnal slope on average, meaningful differences emerged by stress and affect with the Control group showing opposite trends in cortisol diurnal slope responses when compared to their high risk Vulnerable and Resilient counterparts. These ndings suggest that some of the con icting results in cortisol research may be attributable to the lack of differentiation by resilience and family risk. Future cortisol studies may bene t from differentiating individuals by early risks and resilience capacity (Cicchetti & Rogosch, 2007).

Limitation and Strengths
This study had limitations. Menstrual cycle phase among females (Kudielka et al., 2009) was not controlled. Despite carefully controlling for covariates that may impact stress and affect, we cannot rule out the in uence of unmeasured aspects of the current psychosocial environment on cortisol outcomes such as quality of social contacts (Adam & Gunnar, 2001). Moreover, cortisol levels only index part of physiological health (Juster et al., 2011) and are not a pure estimate of HPA-axis functioning (Hellhammer et al., 2009). As with most daily studies, missing data was inevitable. Although the participants were grouped into high and low-risk families, the overall family risk is still low. Further, the sample size of each group varied, and family risk was measured retrospectively. That said, ndings suggest that the appraisal of early family life rather than objective occurrences impact outcomes more (Kaufman et al., 1997). Our ndings are not generalizable to different age group individuals nor those from abusive, high-risk families.
This study also had strengths. The compliance rate in our study was higher than the 44-66% rate of other similarly designed daily studies > 7 days (Heron et al., 2017). Also, we cross-checked reported saliva collection time against survey completion time and self-reported wake and sleep time, allowing higher con dence in our cortisol data accuracies (Kudielka et al., 2003). Reviews on salivary cortisol collection have emphasized maximizing compliance while maintaining low participant burden as priorities (Adam & Kumari, 2009). We collected two saliva samples daily for 14 days, which should better characterize between-subject differences on cortisol slope (Segerstrom et al., 2014). Most prior studies collect more saliva samples across fewer days, some only on a single day (Van den Bergh et al., 2008). The EMA design helped reduce retrospective recall bias. Given that the de nition of resilience necessitates the presence of adversity, we measured the resilience of our participants against a context of early family life risk and current transitory stress by studying emerging adults at a developmental transition who moved at least interstate often internationally for tertiary studies. Most studies measure resilience based on selfreport without the reporting of risks (Colich et al., 2020). We also categorized participants from low-risk families into the Control group as demographically similar comparisons. However, our results could be even more con dently stated if we had an additional comparison group with low family risk and high emotional distress.

Conclusions, Future Directions, And Implications
Our daily study showed that early family life, even at the lower spectrum of risk, exerted signi cant differences in cortisol diurnal slope outcomes when experiencing high stress and distress that were not attenuated by current resilience capacity. These modest but daily physiological differences may accumulatively amount to substantial health impact (

Declarations AUTHOR CONTRIBUTIONS AND ACKNOWLEDGEMENTS:
Material preparation, data collection, and analysis were performed by NT, YY, and JW. The rst draft of the manuscript was written by NT and all authors commented on previous versions of the manuscript. All authors read and approved the nal manuscript. We thank the participants who volunteered their time towards this study, and everyone who assisted with data collection.

CONFLICTS OF INTEREST:
All authors declare no con ict of interest. The funders had no role in the study design, data collection, analysis, interpretation, or presentation of results.

FUNDING:
Wiley (1178487) and Bei (1140299) were supported by NHMRC fellowships. DATA AVAILABILITY: Data will be made available on reasonable request and is planned for public sharing as redacted dataset in the future.   Note. Bold highlights p < .05. Estimates are unstandardized regression coe cients for interactions or simple slopes, followed by p-values, and 95% con dence intervals. All estimates are from linear mixed models with all covariates included. Note. Bold highlights p < .05. Variable in the table varies by column and is either stress, negative affect, or positive affect. Estimates are unstandardized regression coe cients for interactions or simple slopes, followed by p-values, and 95% con dence intervals. All estimates are from linear mixed models with all covariates included. Figure 1 Study Procedure and Survey Completion. Note. All participants started on Mondays and ended on Mondays, providing 10 weekdays and 4 weekend days. All surveys were closed outside their respective time windows to prevent retrospective reports. Stress and affect survey questions were identical over time to reduce participant burden, with some additional questions (not explored in this study) at the pre-sleep survey, resulting in longer completion time. Median completion time was preferred over mean time as it was possible that participants partially attempted surveys and completed them at a later time, leading to arti cially longer completion duration. Participants received hourly push noti cations (an average of 4 prompts), with backup SMS or automatic calls if surveys were not completed within time window.

Figures
Surveys were completed via MetricWire mobile application on Android or iOS operating system.
Participants attended a 1-hour orientation session before starting the daily study, where they were trained in data collection protocol and provided with instruction manuals for completing surveys and saliva collection. *Participants who joined the study during COVID-19 restrictions were given the instructions through a recorded video and supported through teleconference platform Zoom. Subjects could reach a member of the research team by telephone if they had questions or problems during the sampling period.
This period is expected to be a moderate stress period where university students are addressing daily hassles during an academic period of completing assessments or sitting for mid-semester examinations. All daily studies were timed to start during the academic semester and to avoid starting during holidays or break periods.