Parents ’ work demands on next day ’ s cortisol awakening response - the moderating role of family-to-work conflict

Background: Constant availability, overtime and feeling overwhelmed by work can impact employees ’ wellbeing and their biological stress responses. Especially working parents often struggle to balance the demands of their work and family life and were found to be distracted from their work due to family responsibilities. The Family-to-Work Conflict (FWC) indicates the extent to which participating in work is made difficult by family demands. Recent studies have found associations between FWC and biological outcomes such as the Cortisol Awakening Response (CAR), a measure of an individual ’ s Hypothalamic-Pituitary-Adrenal (HPA)-axis activity. This diary study investigates the effect of parental work demands on next day ’ s cortisol response as well as the moderating role of FWC and the mediating role of fatigue. Methods: Over the course of five consecutive days (from Monday to Friday), 168 observations were made on a total of 42 parents. Participants had at least one child and worked a minimum of 20 hours per week. Salivary cortisol samples were obtained immediately, 15 and 30 minutes after awakening each day. Work demands, FWC and fatigue were assessed using standardized questionnaires. Within-person effects were examined using multilevel modeling and mediation analyses. Results: Our results indicate that there are no main effects of work demands on next day ’ s cortisol response. The multilevel analysis revealed that FWC predicts lower wakening cortisol levels and confirmed FWC as an increasing moderator between work demands and next day ’ s HPA-axis activity. Further, work overload was found to increase fatigue, which in turn leads to higher CAR on the following day. This indicates that fatigue mediates the relationship between work demands and CAR. Our findings add to a growing body of research demonstrating further predictors for HPA-axis activity and emphasise the importance of considering family related demands when investigating biological outcomes for working parents.


Introduction
Parenthood and the associated tasks and demands that come with raising a child can affect an employee's well-being in many ways.Particularly in the era of COVID-19 and the resultant irregular hours of care in most children's educational institutions, many working parents have had to adapt to extraordinary demands at home, taking on more responsibility for their children's education or entertainment (Cheng et al., 2021).Moreover, work-related demands such as an unpredictable work schedule (Ananat and Gassman-Pines, 2021) or extended work availability (Dettmers et al., 2016) were found to particularly affect parent's mental health.This inter-role conflict between work and family demands was found to decrease parental well-being (Voydanoff, 2002) and even their job performance (Nohe et al., 2014).Family demands that interfere with work, known as Family-to-Work Conflicts (FWC), can negatively impact mental health.In a sample of 3.659 PhD students, FWC was associated with higher psychological distress and an increased risk of developing a psychiatric disorder (Levecque et al., 2017).In fact, employed parents were observed to take sick leave more often (Brummelhuis et al., 2013) and even show changes in the biological stress response, with one of these changes being the stimulation of the adrenal glands to release cortisol (Zilioli et al., 2016).This diary study investigates the effects of daily work-related demands on next day's cortisol and the moderating role of FWC as well as the mediating role of fatigue in working parents.This paper contributes to the field of parental hormonal adaption in response to work as well as family demands.

Effects of work demands on cortisol levels
The Hypothalamic-Pituitary-Adrenal (HPA) axis is an endocrine system that adapts the organism's ability to adjust homeostasis, for example by regulating glucocorticoids such as cortisol (Tsigos and Chrousos, 2002).It usually follows a diurnal rhythm.As a result, the highest peak in cortisol levels is observed 20-30 minutes after awakening and declines throughout the day.This so-called Cortisol Awakening Response (CAR) is known as an indicator of HPA-axis activity and is described as a biological response to perceived stress (Stalder et al., 2016).
With regard to the effects of work demands on cortisol levels, it was found that stress at work, recovery on the previous evening, sleep quality and quantity all affect next day's energetic state (Dettmers et al., 2016).In terms of the time spent at work, some studies suggest a positive relationship between work hours and the next day's cortisol response (Klumb et al., 2006;Marchand et al., 2013), while on the other hand lower salivary cortisol was found in men working extended hours (Thomas et al., 2009).Interestingly, Parent-Lamarche and Marchand (2018) and Blom et al. (2017), did not find any correlation between the time spent at work and next day's cortisol response.These rather contradictory results may indicate that the association with work hours depends on job type, job properties (e.g., full-time vs. part-time), cortisol measurements, and participant characteristics.Another indicator of work demands is the perception of work overload.Recent studies have shown that work overload predicts higher cortisol responses on the following day (Schlotz et al., 2004) and that students who experienced high overload had a significantly higher CAR than students with low overload (Schulz et al., 1998).Fay and Hüttges (2017) however, only found a marginal effect of work overload on cortisol area under the curve with respect to the ground (AUC G ).The following hypothesis is intended to provide further insight into the effect of work demands on the next day's HPA-axis activity.Hypothesis 1 (H1): Work demands in terms of work hours and overload are associated with next day's cortisol response among working parents.

The potential moderating role of FWC
Stress research has emphasised the role of coping in the relationship between workplace stressors and well-being (Lazarus and Folkman, 1984).We assume that FWC may reduce parents' ability to cope with work demands, thereby increasing the negative outcomes.The distraction or interference of work due to family demands was found to be negatively related to job satisfaction (Cardenas et al., 2004) and self-regulation at work (Zhou et al., 2017).Further, high family interference has been found to be associated with lower relaxation and higher cognitive failure at work (Lapierre et al., 2012).Meanwhile, lower rest and recovery were associated with increased cortisol levels (Gustafsson et al., 2008).Zilioli and colleagues (2016) describe higher FWC to predict lower wake-up cortisol values and less healthy diurnal cortisol slopes, whereas Hibel and colleagues (2012) found that mothers with a combination of high job strain and high parenting stress had higher mean morning cortisol levels and steeper increases in CAR on workdays compared to non-workdays.Based on these theoretical considerations and existing empirical evidence we propose Hypothesis 2 (H2): FWC moderates the relationship between work demands and next day's cortisol response among working parents.

The potential mediating role of fatigue
Fatigue, described as a measure of an individual's energetic state, was found to be associated with a variety of psychological and physiological syndromes (Dijk and Swaen, 2003).Investigating 2.502 parents, Watanabe et al. (2017) found that work demands contribute to significantly higher levels of fatigue.In men, work demands were even found to have a greater effect on fatigue than family demands.Jalilian and colleagues (2019) also confirmed that psychological work demands increase general fatigue in a sample of 522 nurses.On the other hand, increased fatigue has been found to be linked to several physiological outcomes.In fact, a meta-analysis revealed fatigue, burnout, and exhaustion to predict lower CAR (Chida and Steptoe, 2009).This negative association of fatigue on the cortisol response was confirmed by Eek et al. (2012) and Kumari et al. (2009), whereas Lindeberg et al. (2008) did not find a significant effect on cortisol.Based on the empiric background we argue that high work demands increase evening fatigue (compare Jalilian et al. 2019 andWatanabe et al. 2017) and that fatigue in turn affects the next day's cortisol response (compare Eek et al. 2012, Chida and Steptoe 2009, and Kumari et al. 2009).Thus, we hypothesise that fatigue acts as a mediator on the relationship between work demands and cortisol.Hypothesis 3 (H3): Fatigue mediates the relationship between work demands and next day's cortisol response among working parents.

Participants
This study was part of a wider research project on working parents, time use, and work-and family-related stress.The original dairy study targeted 140 parents, of whom 130 completed the survey.42 parents of the 140 participants additionally met our criteria and agreed to be part of the cortisol part of our study.Therefore, the study on hand investigates these 42 parents (24 fathers, 18 mothers) aged 25-58 years (M age = 39.75, SD = 7.51) on five consecutive days (Monday through Friday).Participants were recruited by students at the Medical School Hamburg, Germany (n = 29) or the University of Flensburg, Germany (n = 13) and were living in Frankfurt or Hamburg.Participants were working an average of 4.53 hours/day.Inclusion criteria were: 1.) having at least one child aged < 12 years, 2.) working at least 20 hours per week, 3.) no heavy smoking, 4.) no continuous drug intake, 5.) no chronic somatic disease, 6.) no diagnosed psychiatric disorder, 7.) not currently pregnant or nursing, and 8.) no diagnosed insomnia.Mothers and fathers were recruited independently.The participation of mothers and fathers from the same family was denied to avoid dependencies of the data.All participants were offered a small gift, a workshop on balancing family and work demands and a payment of 30€.

Procedure
The diary study was conducted over five consecutive working days from April to August 2018.Before data collection, participants were instructed regarding the conduction and storage of their own saliva samples and when exactly the questionnaires had to be answered.Psychological data was collected using a brief online questionnaire at four different times throughout the day: upon awakening, at the start of work, after work, and before bedtime.To guarantee timely completion of the short questionnaires, reminders were sent to participants via popups on their mobile devices.Additionally, we recorded the exact times at which the surveys were completed to ensure compliance with the measurement schedule.After the five workdays, participants completed a general survey that assessed information regarding their working arrangements, as well as family and sociodemographic data.The Ethics Committee of the Medical School Hamburg, Germany (MSH-2017/03) approved this study, which was conducted in accordance with the Helsinki Declaration.

Measurements
Salivary cortisol samples (single determination) were collected with salivettes by chewing a cotton roll for two minutes, provided with each salivary tube.Before the sample was taken, participants were asked to wait at least 10 min after smoking, consuming any food or drink, or D.P. Akko and J. Dettmers teeth-brushing.Saliva was taken on Monday through Friday at awakening (C1), +30 min after awakening (C3) as the peak in cortisol should be reached by then, and +15 min after awakening (C2) in order to establish the same time interval between C1 and C3 (compare Stalder et al., 2016).During the collection of the three morning samples, participants were asked not to brush their teeth, not to drink (except water) and not to eat anything, to state the date and the exact time of sampling on the salivettes and to store the samples in their freezers.After being collected by the research staff, the samples were stored at − 20 • C and analysed using a luminescence assay (IBL International).Hormonal analysis was carried out by Prof. Dr. Lars Schwabe, Cognitive Psychology, University of Hamburg, Germany.Intra-and inter-assay coefficients of variation were both < 12 %.To ensure compliance with the measurement plan, participants were instructed to write the exact time of measurement on the respective saliva samples.Additionally, participants were given the opportunity to report any non-compliance with the measurement, such as taking the first cortisol sample too late, without fear of consequences.Hence, observations which were reported to be inaccurate were named into NA.This only affected two observations for C3.The total number of observations was 168 for C1, 168 for C2 and 166 for C3.The three salivary cortisol samples were used to calculate the area under the curve with respect to increase (AUC I ) indicative of the dynamic and change in cortisol over time (Fekedulegn et al., 2018;Pruessner et al., 2003).The AUC I uses the first cortisol value (C1) as a reference point and was calculated using the following formula according to Pruessner et al. (2003): The time intervals are similar, and therefore, time points are excluded from the calculation.The time distance equals 1 and total t i is 2. The formula for individual AUC I for each day was applied as ( Since the time distance is set to 1, this formula does not represent the true area under the curve.Consequently, the resulting values must be regarded as a linear transformation of the area under the curve (Pruessner et al., 2003).Further, C1 was included in our models, indicating the starting point of the CAR period and known to be inversely correlated with the CAR (Stalder et al., 2016).
Work demands were assessed by daily overload immediately after work and daily work hours.Overload was measured Monday to Friday with three items from the Perceived Stress Questionnaire (Levenstein et al., 1993) after work.Sample items included "Today you had too many things to do," "Today you had too many decisions to make" and "Today you felt loaded down with responsibility."The scale was found to be acceptably reliable (3 items; α =.72).Work hours were assessed from Monday to Friday after work.
FWC consisted of three items "Today the demands of my family or spouse or partner interfere with work-related activities," "Today I had to put off doing things at work because of demands on my time at home" and "Today my home life interfered with my responsibilities at work such as getting to work on time, accomplishing daily tasks, and working overtime" based on (Netemeyer et al., 1996).The reliability can be described as good (3 items; α =.86).
FBB (Fatigue Before going to Bed) was measured immediately before going to bed from Monday to Friday.Reliability was found to be acceptable (α =.71) for the three items "I feel rested," "I feel tired" and "I am filled with energy."

Control variables
Regarding trait-like covariates, we defined strict inclusion criteria (see "participants"), excluding e.g., heavy smoking or drinking.Furthermore, the study analyzed a relatively homogeneous age group (range = 25-58) and targeted a relatively similar number of fathers and mothers.According to Stalder et al. (2016), the state factors are even more influential on the daily CAR than stable trait-like influences.Therefore, in this diary study we have added the following state covariates to our models in accordance with the guidelines of Stalder et al. (2016): 1.) Sleep hours representing the total duration of sleep from going to bed to awakening, 2.) sleep quality, representing the subjective quality of the previous night, rated from 1 "bad" to 4 "good", 3.) Upcoming events, indicative for the anticipation of the day ahead, operationalized via a self-constructed questionnaire including the items "Do you have an extraordinary deadline to meet tomorrow?","Do you expect to have exceptionally little time tomorrow?"and "Are you expecting a particularly stressful experience tomorrow?"All items had the characteristics "yes" and "no".

Statistical analysis 2.5.1. Data preparation
The final sample consisted of 166 observations (including outliers).Cortisol outliers were excluded for dependent variables by using the 1.5 interquartile ranges (C1 = 6 outliers, CAR = 3 outliers).After this procedure, CAR was normally distributed while C1 was not (p = 0.018).Because C1 had a total of 160 observations, and to better interpret the outcome measures, no further log-transformation was applied.The independent variables FBB and work overload can be described as normally distributed, whereas work hours and FWC were not.Given the scientific discussion about the implications of Log-transformations (Feng et al., 2014) and the relative stability of MLM-related estimates (Paccagnella, 2011), work hours and FWC were used in their non-normal structure.Person mean-centering in line with Hamaker and Grasman (2015) was applied for work hours, overload, FWC and FBB, respectively.

Multilevel analysis
The MLM was nested for weekday (Monday to Friday; level 1) and participant (level 2) and was separated into three models for each cortisol indicator (C1 and CAR): Model 1 includes the control variables, Model 2 adds work hours and overload after work, and Model 3 adds FWC as well as interaction effects of FWC with work-related indicators.The dependent variables for all models are C1 (Table 2) and CAR (Table 3).The ICC(1) value for C1 was 0.60, indicating that 60 % of the variance was attributed to between-person and 40 % to the withinperson differences.In terms of the CAR, 29 % of the variance was attributed to between-and 61 % to the within-person differences (ICC (1) = 0.29).All multilevel models were controlled for sleep duration, sleep quality, and anticipation for the next day, as described under "Control variables".Due to cortisol's multiple predictors, we also tested the effect of sex and age on next day's cortisol response (C1 and CAR) in separate models with day of measurement and participant as random effects.None of the variables significantly affected next day's CAR.Therefore, sex and age were not added in the final models.RStudio, Version 2022.02.3, was used for statistical analysis.

Mediation analysis
To investigate H3, mediation models were implemented with work demands (work hours and overload) as independent variables, cortisol response (C1 and CAR) as the dependent variable, and FBB as the potential mediator.Within-person effects were modelled using MPlus 6.12 (Muthén and Muthén, 2017).

Descriptive statistics
Means (M), standard deviation (SD), and intercorrelations for the study variables are shown in Table 1.Women in particular were found to have lower sleep quality, work less hours, and to have higher FBB.Higher age was found to be linked to decreased work hours and lower D.P. Akko and J. Dettmers CAR; whereas overload was associated to both higher FWC and FBB.

Cortisol pattern
Regarding the daily cortisol measurements, a significant increase was found between 0 and 15 min after awakening at all weekdays (all ps <.01).From 5-30 min after awakening, no significant changes in cortisol were found for any of the weekdays.One-way ANOVA showed no significant differences between weekdays and the three cortisol time points (all ps >.05).The descriptive distribution is shown in Fig. 1.

Testing within-person main effects
We hypothesized that daily work demands, measured by work hours and overload after work are related to next day's cortisol response (H1).
To test this, we ran a model in which daily C1 and CAR were regressed on daily previous day work hours and overload as well as control variables.However, looking at the direct effect, no significant effects were found for work hours nor overload on C1 (Table 2, Model 2) and CAR (Table 3, Model 2).Thus, H1 must be rejected.The control variables had no significant effect on next day's cortisol response.

Testing within-person main and moderation effects of FWC
Further, we proposed that daily FWC moderates the relationship between daily work demands and next day's cortisol response (H2).To test this hypothesis, we extended the previous Model 2 by the following variables: FWC, FWC*work hours interaction term and FWC*overload interaction term.In this model, FWC significantly decreased next day's C1 (p =.008) FWC*work hours predicted negatively (p =.014) and FWC*overload positively (p =.028) next day's C1 (Table 2, Model 3).These effects could not be found for next day's CAR (Table 3, Model 3).Therefore, H2 can only be accepted regarding cortisol immediately after awakening (C1).

Testing within-person mediation effects of FBB
In H3, we assumed that daily FBB mediates the relationship between work demands and next day's cortisol response.In fact, mediation models revealed a significant effect for Path a (indirect effect = 0.17

Discussion
The aims of the current study were to investigate the effect of work demands on next day's cortisol response and to gain a better understanding of the role of family demands as well as fatigue in this context.

Work demands on cortisol response
Contrary to H1, no direct effect of day-to-day work hours nor overload on next day's cortisol response could be observed (see Tables 2 and  3).Therefore, we argue that work demands are not always predictive for next day's cortisol in parents working at least 20 hours per week.Regarding the rejected effects of daily work hours on cortisol, Klumb and colleagues (2006) found positive associations between work hours  and cortisol levels, with each hour of paid work increasing outcome cortisol levels by 64 nmol/l.Furthermore, the amount of overtime at work was found to increase morning salivary cortisol in 210 working women (Lundberg and Hellström, 2002), whereas higher salivary cortisol levels were found in individuals with high job strain, independently of age or sex (Steptoe et al., 1998).However, these studies did not target parents who work part-time and may be engaged in additional activities that could be considered demands, such as aiding their children with school homework, cleaning, or cooking for the family.Therefore, it is uncertain whether the criteria of working at least 20 hours per week in the study on hand is sufficient to significantly increase next day's cortisol outcomes.
Regarding the effect of overload on next day's cortisol response, we did not find any significant results.Conversely, Schlotz and colleagues (2004) stated that participants with higher levels of chronic work overload and worrying show stronger increases and higher mean levels of cortisol after awakening.In another study investigating 100 subjects with a mean age of 25.7 years, it was found that individuals with chronic work overload had elevated salivary cortisol after awakening as well (Schulz et al., 1998).However, in this study, we focused on daily or acute work overload and within-person effects, while the latter two studies targeted chronic work demands at a between-person level.In addition, the rejected H1 could potentially be valid as perceived stress is not necessarily associated with a biological stress response.In fact, a recent review suggests that the intensity of workload plays a significant role in moderating the relationship between biological and psychological stress indicators.Relationships between self-reported stress scales and hair cortisol were only observed in high-workload samples with an average of 43.7 hours per week (Meij et al., 2018).These findings, as well as our rejected H1, seem to underline that the intensity of a stressor is crucial to observe changes in next day's HPA-axis activity.Moreover, it has been suggested that the majority of employees are capable of managing and coping with the stressors they encounter in the workplace, with some even able to recover from them.This can make it difficult to observe biological changes (Eek et al., 2012).

The moderating role of FWC
FWC itself did not show a significant effect on next day's CAR, measured by cortisol AUC I .Therefore, we argue that in working parents the family interference with work has no effect on the dynamic changes of the HPA-axis activity.On the other hand, we found FWC to predict lower cortisol levels immediately at awakening (C1).These results suggest that family demands interfering with work, lead to a decrease in cortisol levels upon waking the next day, while the CAR remains unaffected.This is in line with the results of Zilioli et al. (2016), who found that greater negative family-to-work spill over was associated with lower cortisol levels at awakening but did not have a significant impact on the CAR.
In line with H2, the FWC*work hours and FWC*overload interactions were found to predict next day's C1.Interestingly, the FWC*work hours interaction, was found to negatively predict next day's wakening cortisol, whereas the FWC*overload interaction predicted higher C1 at the following day.These findings suggest, that the interaction between FWC and work demands play an important role in this relationship between work demands on HPA-axis activity.More specifically, FWC in combination with work overload was found to contribute to higher cortisol levels on the next day.These results are in line with the findings of Hibel and colleagues (2012) who reported parents with the combination of job strain and parenting-related stress to have increased AM (mean of cortisol at awakening and 30 min post awakening) cortisol levels.The negative effect of the FWC*work hours interaction is surprising and might also be varying between internal and external FWC (Krisor et al., 2015).Another explanation might be that the participants who work a considerable amount of hours might be less involved in daily family duties and therefore do not experience as much stress when they are torn between two roles.As described above, mean working time in the present sample was 4.53 hours / day (SD = 1.74) and work hours were not normally distributed.Due to this critical data situation, work hour-related results must be interpreted carefully.
Interestingly, the moderating role of FWC could only be observed for cortisol immediately after awakening (C1), while there were no significant effects for the CAR (see Table 3).Nevertheless, this study suggests that family demands that interfere with work act as a significant moderator for potential effects between work demands and HPA-axis activity.This should be taken into account when investigating parental samples in the future.

The mediating role of fatigue
While the direct effect of FBB on next day's CAR (Path c) was not significant, Path a (0.17*) and Path b (1.88*) of the mediation model showed a significant indirect effect (see Fig. 2-II).These findings confirm H3 and make FBB a mediator for work demands on next day's CAR.Therefore, we argue that work demands, such as overload, increase fatigue and thereby exacerbate next day's HPA-axis activity.The findings of Path a are in line with the findings of Watanabe et al. (2017) and Jalilian et al. (2019), who reported work demands to increase fatigue in.While the increasing effect of fatigue on next day's CAR (Path b) has not yet been shown in a parental sample, Dahlgren et al. (2009) found exhaustion to elevate next day's CAR.Other studies looking at employees (not just parents) found that fatigue decreased cortisol responses (Eek et al., 2012;Kumari et al., 2009).Additionally, prior-day fatigue was found to predict lower wakeup cortisol levels in older adults (Adam et al., 2006).Regarding C1, only Path a of the mediation model was significant, whereas Path b was not (see Fig. 2-I.).Thus, we found that FBB is more strongly related to CAR outcomes, whereas FWC appears rather related to cortisol levels immediately after awakening.This is important to note, given that scientific publications have found an inverse relationship between the first sample taken upon awakening (C1) and the cortisol increase after awakening (CAR) (Stalder et al., 2016).

Practical implications
Our study indicates that the double burden of work demands and FWC significantly impact the biological stress response.It further shows that elevated work demands cause fatigue, which in turn leads to an increase next day's CAR.This seems even more alarming since parental cortisol patterns in some cases were found to be linked to those of their children (Stenius et al., 2008).Thus, additional preventive measures against workplace stress should be developed to mitigate such overload and to enhance resources that can be beneficial to the daily life of parents.According to our results, programs aimed at preventing high evening fatigue might be particularly helpful.Although it may not be feasible to entirely shield working parents from the difficulty of balancing work and family responsibilities, providing greater flexibility in the workplace could aid in managing familial obligations (Dettmers et al., 2020).For us, this also includes avoiding overtime work to enable working parents to recover from demanding work events.Furthermore, parents with a combination of high work and family demands are likely to have a particularly high stress response (Hibel et al., 2012).Indeed, it has been observed that one's own perceived stress can also have an effect on the stress of one's spouse (Watanabe et al., 2017).In order to avoid excessive physiological stress responses, it therefore seems reasonable to share stressful family demands (which may lead to higher FWC) between both parents or other members of the social network, if circumstances allow.Further, our findings indicate that when examining parental work demands, it is crucial to take family-related stress and signs of exhaustion, such as fatigue, into account.Additionally, more research is needed to improve our understanding of the biological stress response.

Limitations
This study has some limitations that must be noted to help guide future research.First, the outcomes of the moderating variable FWC were positively skewed.As a result, a substantial portion of the sample reported having no FWC at all (M = 1.79,SD = 0,97).This may be explained by the fact that the sample consists of parents who work at least 20 hours per week.For future studies, it may be worthwhile to investigate parents who work more hours and are more prone to being distracted by their familial responsibilities.This could include full-time working parents who are currently working from home.Second, this study only examines one individual within a family system and does not particularly consider job characteristics such as flexible work hours or availability.Thirdly, the average levels of cortisol were lower in all measurements (C1, C2, C3) compared to previous studies involving individuals in the same age range.A visual analysis of Fig. 1 suggests lower levels of cortisol nmol/l on Wednesday and Thursday.Nevertheless, there were no significant differences in cortisol levels between the different weekdays across any of the three time points.Fourthly, it should be noted that some of the guidelines outlined by Stalder et al. (2016) were not adequately considered.These include the lack of physiological and objective determination of wake time (e.g. by pulse measurement), as well as the lack of assessment of ethnicity and socioeconomic status, both of which may influence cortisol outcomes (Stalder et al., 2016).Our results should be interpreted accordingly.

Conclusion
This study confirmed that FWC moderates the relationship between work demands and next day's HPA-axis activity.Moreover, fatigue was identified as a mediator, as work overload increased fatigue, which in turn increased next day's CAR.Interestingly, no direct effect of work demands on next day's cortisol response was observed, and some outcomes demonstrated discrepancies depending on various cortisol computations.Our findings add to the growing body of research describing additional predictors of HPA-axis activity and highlight the importance of considering family demands when studying biological stress outcomes in working parents.

Fig. 1 .
Fig. 1.Mean wakening cortisol levels of working parents on weekdays.The xaxis shows the time points of saliva sampling, and the y-axis shows the average raw cortisol levels for all five days.The error bars indicate the standard deviation of the mean.

Fig. 2 .
Fig. 2. Coefficients and standard errors (in parentheses) for the mediation models.Coefficients and standard errors (in parentheses) for the mediation models with C1 (Model I.) and CAR (Model II.) as the outcome variables.CAR = Cortisol Awakening Response, C1 = wakening cortisol level, FBB = Fatigue Before going to Bed. * indicates p <.05.

Table 1
Descriptive Statistics and Correlations for Study Variables.Note.Within-person correlations including all observations (n individual level = 42, n day level = 168).M and SD are used to represent mean and standard deviation, respectively.CAR = Cortisol Awakening Response, C1 = wakening cortisol levels, FBB = Fatigue Before going to Bed, FWC = Family-to-Work Conflict.For sex, 1 = male and 2 = female.Two-sided Pearson correlation analysis.

Table 2
Multilevel model for next day's wakening cortisol (C1) as the dependent variable.Fatigue Before going to Bed, FWC = Family-to-Work Conflict.FWC-related p-values are one-tailed, all other p values are two-tailed.

Table 3
Multilevel model for next day's Cortisol Awakening Response (CAR) as the dependent variable.Fatigue Before going to Bed, FWC = Family-to-Work Conflict.