Cortisol levels in relation to the wellbeing of children in daycare : a meta-analysis

Faculty of Social and Behavioral Sciences

Graduate School of Child Development and Education

Cortisol levels in relation to the wellbeing of children

in daycare: A Meta-Analysis

Anne Elzerman

10473041

Research Master Child Development and Education Thesis 2

Student number: 10473041

First supervisor: prof. dr. R.G. Fukkink Second supervisor: prof. dr. G. J. Overbeek October 12, 2015

Abstract

In recent years, researchers have increasingly used cortisol as a physiological indicator of stress of young children in daycare. Elevated cortisol levels were repeatedly found. To increase our understanding of elevated cortisol levels of children in daycare, a multi-level meta-analysis was conducted on the relationship between cortisol levels and wellbeing of children in daycare. Higher cortisol levels were significantly related to negative state wellbeing (e.g., crying) with a combined effect size for all 13 studies (n = 1,227, 71 effect sizes) of r = .281. Further, higher cortisol levels were negatively correlated with positive state wellbeing (e.g., smiling) r = -.230. Cortisol levels were not significantly related to negative trait wellbeing (e.g., aggressive temperament) (r = .052) nor with positive trait wellbeing (e.g., sociability) (r = -.013). To facilitate the interpretation of cortisol outcomes in young children, it is advocated to include measures of state wellbeing instead of trait wellbeing in cortisol daycare research. Further, this research contributes to the ongoing debate on the interpretation of elevated cortisol levels of children in daycare. This research indicates that elevated cortisol levels are not only interpretable physiologically but elevated cortisol levels also reflect emotional wellbeing. The results of this meta-analysis further show that children with elevated cortisol levels in daycare feel less comfortable than children with lower cortisol levels. No relationships between cortisol and more stable child characteristics (e.g., child temperament) were found.

Cortisol Levels in Relation to the Wellbeing of Children in Daycare: A Meta-Analysis

For young children, daycare settings can be a challenging and stimulating context. At the same time daycare can also be demanding and may involve emotional arousal, either positively or

negatively (Legendre, 2003). Children in daycare are often confronted with situations that challenge the limits of their socio cognitive skills, which may trigger stress responses (Legendre, 2003). A physiological measure to quantify these stress responses that received much attention in daycare research is cortisol. Cortisol is a hormonal product of the hypothalamic-pituitary-adrenocortical (HPA) system and is usually obtained from saliva samples. The HPA system is activated in reaction to physical or psychological stress and helps the child to adapt to a stressful situation (Albers, 2010; Stratakis & Chrousos, 1995). Therefore, most researchers that evaluated stress levels of children in daycare used this as their primary outcome (see Gunnar, Kryzer, Van Ryzin, & Phillips, 2010; Legendre, 2003; Watamura, Donzella, Kertes, & Gunnar, 2004). Cortisol follows a circadian rhythm (De Weerth, Zijl, & Buitelaar, 2003; Stratakis & Chrousos, 1995). Normally, for instance in a home setting, cortisol levels of children are high in the morning and gradually decrease during the day.

In an early meta-analysis of Vermeer and Van IJzendoorn (2006) it is found that daycare children displayed significantly higher cortisol levels compared to children in a home setting.

Additionally, the cortisol levels of children in daycare tend to increase rather than decrease during the day (Ahnert, Gunnar, Lamb, & Barthel, 2004; Dettling, Parker, Lane, Sebanc, & Gunnar, 2000; Vermeer & Van IJzendoorn, 2006; Watamura, Donzella, Alwin, & Gunnar, 2003; Watamura, Sebanc, & Gunnar, 2002). An increase in cortisol was even found in daycare centers of reasonable to high quality and in family based care, suggesting that this applies to all forms of daycare (Gunnar et al., 2010; Vermeer & Van IJzendoorn, 2006). Although elevated cortisol levels are repeatedly reported, it remains unclear how these elevated cortisol levels are related to child wellbeing.

Interpretation of Elevated Cortisol Levels

et al., 2010; Legendre, 2003), eating and sleeping patterns of the children (Hanrahan, McCarty, Kleiber, Lutgendorf, & Tsalikian, 2006), social challenges of early peergroups (Gunnar & Donzella, 2002; Legendre, 2003) and childcharacteristics (e.g., age, sex) (Oullet-Morin, Tremblay, Boivin, Meaney, Kramer, & Côte, 2010). These studies all reported on the correlations between above mentioned factors and cortisol levels. However, due to methodical differences between the studies it remains unclear what the elevated cortisol levels entail for the wellbeing of children in daycare.

Several studies reported on the correlation between cortisol and wellbeing of children in daycare. Groeneveld, Vermeer, Van IJzendoorn, and Linting (2010) for example, found a small non-significant correlation of -.03 between cortisol and general wellbeing of children. In contrast, Ahnert et al. (2004) reported a significant correlation of .41 between cortisol levels and the amount of crying and fuzzing of children. Further, Dettling et al. (2000) found a strong correlation of .57 between elevated cortisol levels and negative affectivity of children. The methodical differences and highly diverging findings complicate a straightforward overall interpretation of the available literature on the wellbeing of young children with elevated cortisol levels in daycare.

Wellbeing

The emotional wellbeing of young children can be examined by asking the parents or other caregivers to fill out questionnaires on their perception of the wellbeing of the child. These are

perceptions of longer lasting personality characteristics of children (e.g., temperament), which refer to trait wellbeing. Secondly, the emotional wellbeing of children can be examined by live or video observations. In these observations, it is for instance possible to examine the amount of crying or smiling within a specific environment at a particular moment. This refers to temporary state wellbeing. Since cortisol levels are fluctuating during the day (Hanrahan et al., 2006), we hypothesize that cortisol levels are more related to temporary events (state wellbeing) than to longer lasting wellbeing (trait wellbeing).

In this study, the emotional wellbeing of children is defined as the degree to which children feel safe and relaxed and enjoy the activities they are engaged in (Groeneveld, Vermeer, Van

IJzendoorn, & Linting, 2008; De Kruif et al., 2007). Indicators of wellbeing can be either positive or negative. Positive and negative wellbeing measures are distinguished in order to explore potential differences in the association with cortisol. Indicators of positive wellbeing are for instance smiling, excitement, happiness, sociability and enthusiasm. Indicators of negative wellbeing are for instance crying, fuzzing, fearfulness, anxiety, sadness and aggression.

Central Question in this Study

This is the first meta-analysis in which all available data on the relationship between cortisol and wellbeing was collected and analysed to facilitate a univocal estimate of the strength of the association between elevated cortisol levels and wellbeing of children in daycare. The central question in this study is: What is the association between cortisol levels and the wellbeing of children in

daycare? It is hypothesized that state wellbeing measures show stronger associations with cortisol than trait wellbeing measures because cortisol is fluctuating in reaction to temporary events. Further, it is questioned whether the correlation between cortisol and wellbeing is different for positive and negative wellbeing measures.

Method Selection of Studies

On the 14th of February, 2014 a search was conducted on the databases Web of Science, PsychINFO, Medline, Cochrane Library and ERIC using combined search terms for the key variable (i.e., cortisol level) (“cortisol*, hydrocortisone*, stress hormone*, HPA* or hypothalamic pituitary adrenal*”) and wellbeing variables (“well-being*, wellbeing*, happiness*, happy*, pleasure*, temperament*, excitement*, enjoyment*, self-esteem*, satisfaction*, laugh*, emotional*, arousal*, behavior*, problem*, stress reactions*, stress*, distress*, sadness*, anger*, angry*, aggression*, aggressive behavior*, tantrum*, rage*, anxiety*, anxious*, social anxiety*, social interaction*, peer* or friend*”). Because children in daycare were of interest in this study, descriptions for setting were combined with the above mentioned search terms (“nursery school*, prekindergarten*, pre-kindergarten*, child day care*, day care*, daycare*, child care*, childcare*, early childhood education*, preschool education* or preschool*”) to identify relevant studies that were published until now. The search resulted in 8181 reports.

The broad-ranging search was followed by a selection of studies based on a reading of their titles, abstracts or full content. Studies were only included when they concerned research on the cortisol levels of children measured at the daycare center combined with a

measurement to examine the wellbeing of the child (e.g., with a wellbeing questionnaire or observation). Furthermore, only studies reporting the statistics necessary to obtain the effect sizes were included in this meta-analysis. Studies were excluded when their main focus was on children or parents with psychopathology, medication or on children with disabilities. Of all studies, 13 met the selection criteria (see Figure 1 for the search process).

All studies that were included in the meta-analysis were coded for quality and methodological characteristics. For quality, several home environment characteristics were coded (i.e., general quality of the home environment and sensitivity of the mother).

Furthermore, multiple daycare characteristics were coded as indicators for daycare quality (i.e., type of daycare, adult-child ratio, group size, caregiver quality, general quality of the daycare center and mean turnover rate). Further, child characteristics that might be of

influence on cortisol levels and wellbeing were coded (i.e., mean age of the children, adaption phase, mean time in daycare, amount of space available per child, age group in the overall sample and gender of the children). In addition, the type of wellbeing measure (i.e., physiological, observation or questionnaire) and the way cortisol was collected were coded (i.e., saliva or urine, circadian rhythm, context in which cortisol was obtained, by whom the cortisol was collected and by what the saliva flow was stimulated). Further, a number of methodological characteristics were coded (i.e., publication type, type of design, random assignment or matching, the number of participants at the start and the definitive number of participants). Finally, effect sizes in the studies were coded for the trend of the cortisol levels of the children between at least two time points a day (i.e., positive or negative trend) and the results of the wellbeing measure (i.e., questionnaire or observation).

Two independent coders coded the studies to determine the inter-coder reliability of the variables of most interest (i.e., state or trait, wellbeing or temperament, positive or negative wellbeing measure). Cohen’s Kappa for the variable state was .93, for the variable trait it was .90, for the variable wellbeing it was .89, for the variable temperament it was .87, for the variable positive wellbeing it was .89, and for the variable negative wellbeing it was .85. These results indicate excellent inter-coder agreement according to Landis and Koch (1977).

Analysis

Most of the studies that were identified in the search provided multiple effect sizes reflecting different wellbeing measures. To account for nested data at least two levels are required in the analyses (i.e., study level and effect size level). However, some of the studies provided several effect sizes for one wellbeing measure. Therefore, a third level was required because of the multiple ways of measuring one construct were nested as well (i.e., wellbeing level). A three-level multilevel approach was used to analyse the data in a structured manner. The first level was the study level, the second level was the wellbeing level, the third level was the effect size level. The specification and testing of models was conducted with MlwiN, using restricted maximum-likelihood estimation (Bryk & Raudenbusch, 2002; Hox, 2010).

Effect sizes were calculated directly from the reported correlations (r). All correlations were transformed using the Fisher’s z transformation to correct for skewness in the sampling distribution of r (Borenstein, Hedges, Higgins, & Rothstein, 2009; Field, 2001). All analyses were performed using Fisher’s z. After analyses the outcomes for the models of most interest were transformed back into correlations (r) for interpretation purposes. For inherently

negative variables, effect sizes were recoded by changing the sign. To examine potential publication bias, a funnel plot was evaluated and an Egger’s test for asymmetry was performed.

Results

Thirteen papers were found in which children’s cortisol levels in childcare were associated with at least one wellbeing measure (see Table 1). Nine studies were conducted in the USA and four studies were conducted in European countries. In total 1021 children were involved in the studies. The mean age of the children was 36.3 months ranging from 2 to 106 months. In total 71 effect sizes were obtained from the studies. Fourteen effect sizes were obtained for state wellbeing while 57 effect sizes were obtained for trait wellbeing (see Table 2). Most effect sizes were reported for positive trait wellbeing measures in relation to cortisol levels of children. Visual inspection of the funnel plot did not result in suspicion of

publication bias (see Figure 2). Further, Egger’s regression test showed no significant asymmetry (t = 0.38, df = 70, p = .70) suggesting that there is no publication bias present in the sample.

Baseline model

The baseline model showed that there was a significant overall relation between cortisol and wellbeing (z = .076; SE = .033, p < .05). The z-score was calculated on the basis of the regression coefficient and intercept as listed in Table 3. This aggregated effect size may not give a valid indication of the true relationship between cortisol and wellbeing because studies on state wellbeing and trait wellbeing were analysed together. The variance at the first level in the baseline model was significant (z = .019; SE = .008, p < .05). This can be

interpreted as the studies being different in their findings on the relation between cortisol and the wellbeing of children in daycare. Further, the variance on the second level in the baseline model was non-significant (z = .001; SE = .002, p < .05). This indicates that between the different wellbeing measures the association between cortisol and wellbeing does not vary to a significant extent. The baseline model fits the data quite well and there is little variance left

to be explained. Statistically, it may therefore not make sense to add predictors to the model. However, in the baseline model, studies that measured wellbeing as a state or trait and studies that examined negative or positive wellbeing measures were analysed together as if they were not different in their interpretation. This might blur the picture of the baseline model. Based on these theoretical considerations and the hypothesis posted in this study, several parameters were nevertheless inspected.

State Wellbeing versus Trait Wellbeing

State wellbeing, both positive and negative, was significantly correlated with cortisol levels of children in daycare (z = .221, SE = .056, p < .05). Trait wellbeing was not

significantly correlated with cortisol levels of children in daycare (z = .032, SE = .030, p > .05). See Model 1 and Model 2 in Table 3 for the regression coefficients and intercepts that were used to calculate the z-scores.

Positive Wellbeing versus Negative Wellbeing

Positive wellbeing measures, for both state and trait wellbeing, were not significantly correlated with cortisol levels of children in daycare (z = .062, SE = .045, p > .05). Negative wellbeing measures were significantly correlated with cortisol levels of children in daycare (z = .099, SE = .047, p < .05). See Model 3 and Model 4 in Table 3 for the regression

coefficients and intercepts that were used to calculate the z-scores.

State and Positive Wellbeing versus State and Negative Wellbeing

Positive wellbeing measured as a state (e.g., observed smiling, peer play, positive engagement) was significantly correlated with cortisol levels of children in daycare (z = .234,

SE = .100, p < .05). The corresponding Pearson correlation was r = -.230 indicating that the

wellbeing measured as a state (e.g., observed crying, fuzzing, anxiety, stress) was also significantly correlated with cortisol levels of children in daycare (z = .289, SE = .095, p < .05). The corresponding Pearson correlation was r = .281 indicating that the higher the cortisol levels, the more negative wellbeing was observed (see Table 4). The strongest predictor for the relationship between cortisol and wellbeing appeared to be negative state wellbeing. See Model 5 and Model 6 in Table 3 for the regression coefficients and intercepts that were used to calculate the z-scores.

Trait and Positive Wellbeing versus Trait and Negative Wellbeing

Positive wellbeing measured as a trait (e.g., questionnaire reported pleasure, positive skills, interest, positive affect) was not significantly correlated with cortisol levels of children in daycare (z = .013, SE = .052, p > .05). The corresponding Pearson correlation was r = -.013 (see Table 4). Negative wellbeing measured as a trait (e.g., questionnaire reported aggression, negative affectivity, sadness, anxiety, negative social behaviour, anger) was not significantly correlated with cortisol levels of children in daycare (z = .052, SE = .054, p > .05). The corresponding Pearson correlation was r = .052 (see Table 4). Both positive and negative trait wellbeing measures were not significantly related to cortisol levels of children in daycare. See Model 7 and Model 8 in Table 3 for the regression coefficients and intercepts that were used to calculate the z-scores.

Overall Wellbeing

Three studies reported on the overall wellbeing of children in daycare. Watamura et al. (2003) used the Infant Behavior Questionnaire (IBQ) and the Toddler Behavior Questionnaire (TBQ) to assess the wellbeing of infants and toddlers in daycare. On a seven point Likert scale (1= not present, 7= always present) they found a score of 3.56 on social fear, 3.79 on anger and 4.23 on positive affect in infants. For toddlers they found a score of 2.79 on social

fear, 3.38 on anger and 5.23 on positive affect. In contrast, Groeneveld et al. (2010) used video observations to assess the wellbeing of children in daycare. On a seven point Likert scale (1= very low wellbeing, 7= very high wellbeing) they found a wellbeing score of 4.27. In another study the same researchers found a wellbeing score of 4.60 (Groeneveld, Vermeer, Van IJzendoorn, & Linting, 2012). A neutral score (i.e., score = 4.0) means that no real signs of discomfort nor wellbeing are observed (De Kruif et al., 2007). Further, Ahnert et al. (2004) reported that children in daycare were observed for 30 minutes and that the children were crying and fuzzing 13% of the time.

Discussion

This meta-analysis was aimed at strengthening the understanding of elevated cortisol levels of children in daycare in relation to their wellbeing. The main finding in the study is that elevated cortisol levels of children in daycare are significantly related to lower wellbeing of children. A secondary finding in the study is that the correlation between cortisol and wellbeing is only related to temporary state wellbeing measures. The strongest correlation found in this study is the one between cortisol and negative state wellbeing (r = .289). This correlation indicates that higher cortisol levels are positively related to observed negative wellbeing in daycare. The correlation between cortisol and positive state wellbeing is -.230. This correlation indicates that higher cortisol levels are related to lower state wellbeing in daycare. Neither the correlation between cortisol and positive trait wellbeing nor the correlation between cortisol and negative trait wellbeing are significant indicating that cortisol is not significantly related to trait wellbeing of children. To summarize, cortisol levels of children in daycare are related to both positive and negative state wellbeing but not to trait wellbeing.

The results of this study are of interest to theory as well as practice. Earlier research has repeatedly shown that cortisol levels of children in daycare are elevated compared to their cortisol levels in the home situation (Ahnert et al., 2004; Dettling et al., 2000, Vermeer & Van IJzendoorn, 2006; Watamura et al., 2003; Watamura et al., 2002). This meta-analysis contributes to the knowledge on the interpretation of cortisol levels of children in daycare. Children with elevated cortisol levels in daycare feel less comfortable than children with lower cortisol levels. Further, this meta-analysis shows a difference in the strength of the association between cortisol and wellbeing for state and trait.

For parents, caregivers and daycare providers it is important to be aware of the relation between elevated cortisol levels and negative wellbeing of children. Center-based care is characterized by several stressors like separation from the parents, confrontation with an unfamiliar setting, different routines than at home, different caregivers, larger numbers of peers than in the home situation and higher levels of noise than at home (Beijers, Riksen-Walraven, Putnam, De Jong, De Weerth, 2013). Further, cortisol levels of children in daycare are higher (Vermeer & Van IJzendoorn, 2006) and are related to more negative feelings and expressions of negative wellbeing in children, as this

meta-environment where children are challenged and experience new things (Albers, Riksen-Walraven, & De Weerth, 2010; Gubbels, Van Kann, & Jansen, 2012), the positive arousal that these experiences might entail does not explain the elevated cortisol levels that are repeatedly found in children in daycare. This meta-analysis suggests that a negative interpretation of elevated cortisol levels of the children in daycare is more probable.

This research shows that elevated cortisol levels are significantly related to lower state wellbeing. The conclusion could be drawn that the decreased wellbeing is probably not very harmful since it is only temporary. This study does not provide evidence on whether elevated cortisol levels and temporary negative wellbeing have long-term effects for children. It is known however, that frequent neurobiological stress responses increase the risk of physical and mental health problems (Gunnar & Quevedo, 2007; McEwen, 2011). There are several studies reporting on long-term effects of daycare attendance. Infants scoring high on negativity who were attending daycare at the age of three months, showed more mother-rated internalizing and externalizing problems 27 months later (Beijers et al., 2013). Further, Vandell, Belsky, Burchinal, Steinberg, and Vandergrift (2010) report that more hours in daycare predicted greater risk taking and impulsivity at age 15. Another study detected an association between more hours in daycare and lower awakening cortisol levels at age 15 (Roisman et al., 2009). Future research is needed, however, to examine whether there is a direct relation between elevated cortisol levels at daycare and both short- and long-term child outcomes. Nevertheless, parents might consider to change their non-parental care plans because they do not want to expose their child to potentially elevated cortisol levels at daycare (Vermeer & Van IJzendoorn, 2006) and the associated decreased wellbeing as found in this study.

Although the results of this meta-analysis indicate that higher cortisol levels are related to lower wellbeing, an overall degree of wellbeing (e.g., is the wellbeing of the child below a critical boundary?) is not explicitly specified in this study. This is caused by the fact that the researchers examined wellbeing in different ways. Only few researchers reported overall wellbeing scores of children (see Watamura et al., 2003; Groeneveld et al., 2010; Groeneveld et al., 2012). Most

researchers only presented the correlation between cortisol and wellbeing or the report included only a regression analysis without providing mean wellbeing scores. Further, the limited number of studies

reporting on the overall wellbeing of children also complicates a straightforward interpretation on the overall wellbeing of the children in the studies included in this meta-analysis. However, to further explore the degree of negative wellbeing of children with elevated cortisol levels in daycare, future research is needed. Nevertheless, none of the studies analysed in this meta-analysis reported excellent overall wellbeing of the children. It can be debated what wellbeing-standards are acceptable or favourable for infants and toddlers in our society.

In this meta-analysis the association between cortisol and state wellbeing is found to be small to moderate. However, the distribution of studies reporting on state versus trait wellbeing is skewed. This indicates that most studies included in this meta-analysis focussed on the relation between elevated cortisol levels of children in daycare and trait wellbeing. Only 14 effect sizes were obtained for state wellbeing compared to four times as many effect sizes for trait wellbeing (i.e., 57 effect sizes). Nevertheless, despite the small power, significant associations between cortisol and state wellbeing of children were found.

Although, quality of care (Geoffroy et al., 2006; Groeneveld et al., 2010; Gunnar et al., 2010; Legendre, 2003), eating and sleeping patterns of the children (Hanrahan et al., 2006), social challenges of early peergroups (Gunnar & Donzella, 2002; Legendre, 2003) and childcharacteristics (e.g., age, sex) (Oullet-Morin et al., 2010) are related to cortisol levels of children, these factors were not included in this meta-analysis because the analysed studies did not always provide the necessary figures. Nevertheless, it would have been interesting to examine whether these factors also influence the relation between cortisol and wellbeing. In addition, no variables could be added to the model since there was no variance left to be explained due to lack of power. Future research should explore the influence of the above mentioned factors on the relation between elevated cortisol levels and wellbeing of children in daycare to determine potential differences in susceptibility to lower wellbeing for example in younger children or children with a difficult temperament (Beijers et al., 2013).

To our knowledge, this meta-analysis is the first providing an estimate of the association between cortisol and wellbeing. Previous research has shown that cortisol and wellbeing are related but until now the average strength of this association remained unknown. Further, this is the first

meta-cortisol levels of children. In addition, by performing a meta-analysis we were able to analyse data from a large total sample of 1021 children. By using a three-level analysis model that corrected for dependency between correlations reported within and across studies, the chance of a type I error (i.e., finding a false positive finding) was reduced (Van Noortgate, López-López, Marín-Martínez, & Sánchez-Meca, 2013).

In conclusion, the results of this meta-analysis show that children with elevated cortisol levels feel less comfortable than children with lower cortisol levels. This relation is only significant for state wellbeing. For future research on the relationship between cortisol levels of children in daycare and their wellbeing within the daycare environment, researchers should focus on state wellbeing measures instead of trait wellbeing measures. Trait wellbeing measures could merely be indicative for the general wellbeing or the temperament of the child. Further, the results of this meta-analysis contribute to the debate on whether we should expose young children to the stressors daycare might entail.

References1

*Ahnert, L., Gunnar, M., Lamb, M., & Barthel, M. (2004). Transition to child care: Associations with infant-mother attachment, infant negative emotion, and cortisol elevations. Child Development, 75(3), 639-650. doi: 10.1111/j.1467-8624.2004. 00698.x

Albers, E. (2010). The challenge of childcare for very Young infants (Doctoral dissertation). Radboud Universiteit, Nijmegen.

Albers, E. M., Riksen-Walraven, J. M., & De Weerth, C. (2010). Developmental stimulation in child care centers contributes to young infants’ cognitive development. Infant

Behavior and Development, 33(4), 401-408. doi: 10.1016/j.infbeh.2010.04.004 *Alwin, J. (2006). Salivary cortisol in infants at daycare and at home. Dissertation Abstracts

International: Section B: The Sciences and Engineering, 66(9-B), 5119.

Beijers, R., Riksen-Walraven, J. M. A., Putnam, S., De Jong, M., De Weerth, C. (2013). Early Non-parental Care and Toddler Behaviour Problems: Links with Temperamental Negative Affectivity and Inhibitory Control. Early Childhood Research Quarterly, 28(4), 714-722. doi: 10.1016/j.ecresq.2013.06.002

Borenstein, M., Hedges, L. V., Higgins, J. P. T., & Rothstein, H. R. (2009). Introduction to

meta-analysis. Chichester, UK: Wiley.

Bryk, A. S., & Raudenbusch, S. W. (2002). Hierarchical linear models: Applications and

data analysis methods. Thousand Oaks, CA: Sage Publications.

De Kruif, R. E. L., Vermeer, R. G., Fukkink, R. G., Riksen-Walraven. J. M. A., Tavecchio, L. W. C., Van IJzendoorn, M. H., & Van Zeijl, J. (2007). De nationale studie

pedagogische kwaliteit kinderopvang: Eindrapport project 0 en 1. [The national study

on pedagogic quality of childcare: Final report project 0 and 1.]. Amsterdam: NCKO. Retrieved from http://dare.uva.nl/document/2/53540

De Weerth, C., Zijl, R. H., & Buitelaar, J. K. (2003). Development of cortisol circadian rhythm in infancy. Early Human Development, 73(1-2), 39-52. doi: 10.1016/S0378-3782(03)00074-4

*Dettling, A. C., Parker, S. W., Lane, S., Sebanc, A., & Gunnar, M. R. (2000). Quality of care and temperament determine changes in cortisol concentrations over the day for young children in childcare. Psychoneuroendocrinology, 25(8), 819-836. doi: 10.1016 /S0306-4530(00)00028-7

Field, A. P. (2001). Meta-analysis of correlation coefficients: a Monte Carlo comparison of fixed-and random-effects methods. Psychological methods, 6, 161-180. doi: 10.1037/ 1082-989X.6.2.161

Geoffroy, M., Côté, S.M., Parent, S., & Séguin, J.R. (2006). Daycare Attendance, Stress, and Mental Health. Canadian journal of psychiatry, 21(9), 607-615. Retrieved from http://web.a.ebscohost.com.proxy.uba.uva.nl:2048/ehost/detail/detail?sid=76ae8405-ca92-4f48-861c-76a8c7b3368a%40sessionmgr4003&crlhashurl=login.aspx%253 fdirect%253dtrue%2526scope%253dsite%2526db%253daph%2526AN%253d224624 02%2526msid%253d201344943&hid=4201&vid=0&bdata=JnNpdGU9ZWhvc3QtbG l2ZQ%3d%3d#AN=22462402&db=aph

Groeneveld, M. G., Vermeer, H. J., Van IJzendoorn, M. H., & Linting, M. (2008). Welbevinden en stress van kinderen in de kinderopvang: een vergelijking van kinderdagverblijven en gastoudergezinnen. [Wellbeing and stress of children in childcare: a comparison of center based and family based daycare.]. Leiden:

Universiteit Leiden. Retrieved from http://media.leidenuniv.nl/legacy/Welbevinden %20en%20stress%20van%20 kinderen%20in%20de%20kinderopvang.pdf

*Groeneveld, M. G., Vermeer, H. J., Van IJzendoorn, M. H., & Linting, M. (2010). Children's wellbeing and cortisol levels in home-based and center-based childcare. Early

Childhood Research Quarterly, 25(4), 502-514. doi: 10.1016/j.ecresq.2009.12.004 *Groeneveld, M. G., Vermeer, H. J., Van IJzendoorn, M. H., & Linting, M. (2012).

Stress, cortisol and well-being of caregivers and children in home-based child care: A case for differential susceptibility. Child: Care, Health and Development, 38(2), 251- 260. doi: 10.1111/j.1365-2214.2010.01194.x

Gubbels, J. S., Van Kann, D. H. H., & Jansen, M. W. J. (2012). Play Equipment, Physical Activity Opportunities, and Children’s Activity Levels at Childcare. Journal of

Environmental and Public Health, 2012, 1-8. doi: 10.1155/2012/326520

Gunnar, M. R., & Donzella, B. (2002). Social regulation of the cortisol levels in early human development. Psychoneuroendocrinology, 27(1), 199-220. doi: 10.1016/S0306-4530 (01)00045-2

Gunnar, M. R., Kryzer, E., Van Ryzin, M. J., & Phillips, D. A. (2010). The Rise in Cortisol in Family Daycare: Associations With Aspects of Care Quality, Child Behavior, and Child Sex. Child Development, 81(3), 851-869. doi: 10.1111/j.1467-8624.2010 .01438.x

Gunnar, M. R., & Quevedo, K. (2007). The Neurobiology of Stress and Development. Annual

Review of Psychology, 58, 145-173. Retrieved from http://www.annualreviews.org.

proxy.uba.uva.nl:2048/doi/abs/10.1146/annurev.psych.58.110405.085605

Hanrahan, K., McCarty, A. M., Kleiber, C., Lutgendorf, S., & Tsalikian, E. (2006). Strategies for salivary cortisol collection and analysis in research with children. Applied Nursing

Research, 19(2), 95-101. doi: 10.1016/j.apnr.2006.02.001

Hox, J. J. (2010). Multilevel analysis. Techniques and applications. 2nd Edition. New York: Routledge.

Landis, J. R., & Koch, G. G. (1977). The measurement of observer agreement for categorical data. Biometrics, 33, 159-174. doi: 10.2307/2529310

Legendre, A. (2003). Environmental features influencing toddlers’ bioemotional reactions in day care centers. Environment and Behavior, 35(4), 523-549. doi: 10.1177/00139 16503035004005

McEwen, B. S. (2011). Effects of stress on the developing brain. Cerebrum. Retrieved from http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3574783/

Oullet-Morin, I., Tremblay, R. E., Boivin, M., Meaney, M., Kramer, M., & Côté, S. M. (2010). Diurnal cortisol secretion at home and in child care: a prospective study of 2- year-old toddlers. Journal of Child Psychology and Psychiatry, 51(3), 295-303. doi: 10.1111/j.1469-7610.2009.02167.x

*Reunamo, J., Sajaniemi, N., Suhonen, E., & Kontu, E. (2012). Cortisol levels and childrens’s orientation in day care. Early Child Development and Care, 182(3-4), 363-381. doi: 10.1080/03004430.2011.646727

Roisman, G. I., Susman, E., Barnett-Walker, K., Booth-La-Force, C., Owen, M. T., Belsky, J., et al. (2009). Early family and child-care antecedents of awakening cortisol levels in adolescence. Child Development, 80(3), 907-920. doi: 10.1111/j.1467-8624.2009 .01305.x

Stratakis, C. A., & Chrousos, G. P. (1995). Neuroendocrinology and pathophysiology of the stress system. In: Chrousos, G. P., McCarty, R., Pacak, K., Cizza, G., Sternberg, E., Gold, P. W., Kvetsnansky, R. (Eds.), Stress: Basic Mechanims and Clinical

Implications. N. Y. Acad. Sci., New York, pp. 1-18.

Van den Noortgate, W., López-López, J. A., Marín-Martínez, F., & Sánchez-Meca, J. (2013). Three-level meta-analysis of dependent effect sizes. Behavior research methods, 45, 576-594. doi: 10.3758/s13428-012-0261-6

Vandell, D. L., Belsky, J., Burchinal, M., Steinberg, L., & Vandergrift, N. (2010). Do Effects of Early Child Care Extend to Age 15 Years? Results from the NICHD Study of Early Child Care and Youth Development. Child Development, 81(3), 737-756. doi:

10.1111/j.1467-8624.2010.01431.x

Vermeer, H. J., & Van IJzendoorn, M. H. (2006). Children’s elevated cortisol levels at daycare: A review and meta-analysis. Early Childhood Research Quarterly, 21(3), 390-401. doi: 10.1016/j.ecresq.2006.07.004

*Watamura, S. E., Donzella, B., Alwin, J., & Gunnar, M. R. (2003). Morning-to-Afternoon Increases in Cortisol Concentrations for Infants and Toddlers at Child Care: Age Differences and Behavioral Correlates. Child Development, 74(4), 1006-1020. doi: 10.1111/1467-8624.00583

Watamura, S. E., Donzella, B., Kertes, D. A., & Gunnar, M. R. (2004). Developmental changes in baseline cortisol activity in early childhood: Relations with napping and effortful control. Developmental Psychobiology, 45(3), 125-133. doi: 10.1002/ dev.20026

Watamura, S. E., Sebanc, A. M., & Gunnar, M. R. (2002). Rising cortisol at childcare: Relations with nap, rest and temperament. Developmental Psychobiology, 40(1), 33- 42. doi: 10.1002/dev.10011

Table 1

Characteristics of Included Studies.

Study Continent

of the study

Sample size (n)

Age (Months) Number of provided effect sizes State/ Trait M Range

Ahnert et al. (2004) Europe 56 15 11-20 2 State

Alwin (2005) USA 25 10 3-18 5 Both

Badanes (2010) USA 166 48 24-65 14 Trait

Badanes et al. (2012) USA 110 48 24-65 3 Trait

Dettling et al. (1999) USA 70 n.a. 39-106 7 Trait

Dettling et al. (2000) USA 21 n.a. 40-69 4 Trait

Groeneveld et al. (2010) Europe 116 31 20-40 2 State

Groeneveld et al. (2012) Europe 44 29 20-40 6 Both

Gunnar et al. (2011) USA 107 45 36-54 2 Both

Lane (2012) USA 20 n.a 12-24 3 Trait

Lisonbee et al. (2008) USA 191 53 43-67 6 Trait

Reunamo et al. (2012) Europe 55 58 36-84 4 Trait

Watamura et al. (2003) USA 40 26 2-38 14 Both

Total 1021 36.3 2-106 71

Table 2

Distribution of obtained effect sizes

Negative wellbeing Positive wellbeing Total

State 5 9 14

Trait 22 35 57

Table 3

Fixed effects and Random effects for all analysed models (Fisher’s z scores).

Parameter Baseline Model Model 1 State Model 2 Trait Model 3 Positive Model 4 Negative Model 5 State*Positive Model 6 State*Negative Model 7 Trait*Positive Model 8 Trait*Negative Fixed Effects Intercept .076 (.033) .033 (.030) .220 (.056) .098 (.047) .063 (.045) .053 (.034) .053 (.031) .113 (.040) .092 (.043) State - .188 (.064) - - - - Trait - - -.188 (.064) - - - - Positive - - - -.036 (.065) - - - - - Negative - - - - .036 (.065) - - - - State*Positive - - - .181 (.093) - - - State*Negative - - - .236 (.100) - - Trait*Positive - - - -.100 (.065) - Trait*Negative - - - -.040 (.069) Random Effects First Level .019 (.008) .010 (.005) .010 (.005) .018 (.008) .018 (.008) .016 (.007) .013 (.006) .017 (.007) .020 (.008) Second Level .001 (.002) .001 (.002) .001 (.002) .001 (.002) .001 (.002) .001 (.034) .000 (.002) .001 (.002) .001 (.002) Third Level 1.000 (.000) 1.000 (.000) 1.000 (.000) 1.000 (.000) 1.000 (.000) 1.000 (.000) 1.000 (.000) 1.000 (.000) 1.000 (.000) -2*log likelihood -41.705 -49.150 -49.150 -43.600 -43.600 -45.443 -46.961 -44.068 -42.012

Table 4

Cortisol levels in correlation with positive state wellbeing, negative state wellbeing, positive trait wellbeing and negative trait wellbeing only.

Positive wellbeing Negative wellbeing State r = -.230* r = .281*

Trait r = -.013 r = .052 Note. *p < .05

Figure 1. Flowchart of Literature Search and Screening Literature search

8181 papers yielded

Duplicates removed 6011 papers yielded

Titles + Abstracts reviewed 643 papers selected

Papers read 28 papers

13 studies included in meta-analysis

5368 papers excluded (not English language, not human,

not on young children)

615 papers excluded (studies on psychopathology, medication, children with disabilities)

15 papers excluded (no usable statistics reported)

Figure 2. Funnel plot with on the X-axis the effect sizes and on the Y-axis the number of