Salivary cortisol reactivity to psychological stressors in infancy: A meta-analysis

1 Salivary Cortisol Reactivity to Psychological Stressors in Infancy: A Meta-Analysis Ilmari J. A. Puhakka1,2 & Mikko J. Peltola2 1 Faculty of Education and Culture, Tampere University, Finland 2 Human Information Processing Laboratory, Psychology, Faculty of Social Sciences, Tampere University, Finland CORRESPONDING AUTHOR: Mikko Peltola, Human Information Processing Laboratory, Psychology, Faculty of Social Sciences, 33014 Tampere University, Finland. Tel. +358503186120. E-mail: [email protected] This is the accepted manuscript of the article, which has been published in Psychoneuroendocrinology, 2020, 115, 104603. https://doi.org/10.1016/j.psyneuen.2020.104603 2 Abstract Measurement of salivary cortisol is a practical and non-invasive tool for studying stress reactivity to various types of stressors even in young infants. Whereas studies using physical stressors during the first months of life have found robust cortisol responses to painful stimuli, research with older infants using psychological stressors (e.g., parental separation) has produced mixed findings, limiting our understanding of potential developmental changes in cortisol reactivity across infancy. In the present study, we used meta-analysis to systematically investigate whether psychological stressor paradigms are associated with measurable cortisol responses in infants under 18 months of age and whether the magnitude of the responses is moderated by the type of psychological stressor (i.e., separation, frustration, novelty, or disruption of parental interaction), infant age, and other potential moderators. Across 48 studies (N = 4095, age range: 3-18 months), we found that commonly used psychological stressor paradigms are associated with a small (Hedges’ g = .11) increase in salivary cortisol levels in typically developing infants. Stressor type moderated the effect sizes, and when effect sizes in each category were analyzed separately, only the separation studies were associated with a consistent increase in cortisol following the stressor. Age did not moderate the effect sizes either in the full set of studies or within the separate stressor types. These meta-analytic results indicate that the normative cortisol response to psychological stressors across infancy is small and emphasize the need for standardized stressor paradigms to assess cortisol responses systematically across infancy. Keywords: Cortisol; Infant; Meta-Analysis; Saliva; Stressor 3 1. Introduction Given the central role of stress in health and development, there has been a longstanding interest in charting the typical development and individual differences in stress reactivity (Gunnar & Quevedo, 2007). The methodological advancement of measuring cortisol non-invasively from the saliva has been particularly valuable for quantifying activity of the hypothalamic-pituitary-adrenal (HPA) axis in infants who are unable to subjectively report their stress responses. Salivary cortisol reactivity in infancy has been assessed in various experimental settings including physical stressors (e.g., vaccination or blood draw) and psychological stressors designed to trigger negative emotional responses (e.g., separation from the parent or presentation of novel persons or objects). Salivary cortisol reactivity is measured as the difference between a pre-stressor sample and a post-stressor sample collected typically 20–30 min after the stressor presentation. Research on children’s cortisol reactivity indicates that the magnitude of the salivary cortisol response to stressors appears to decrease across infancy and early childhood, and increase again from middle childhood to adolescence (Gunnar, Talge, & Herrera, 2009). In studies measuring infants’ cortisol responses to inoculations across the first two years, Davis and Granger (2009) observed increases in cortisol in response to inoculation at 2 and 6 months, but not at 12 and 24 months, while Gunnar, Brodersen, Krueger, and Rigatuso (1996) observed increases at 2, 4, and 6 months, but not at 15 months. It has been suggested that by the end of the first year, the infant’s HPA axis becomes buffered as a function of supportive caregiving relationships (see Gunnar & Quevedo, 2007, for a review). The buffered HPA response later in the first year thus represents a functional equivalent of the hyporesponsive period of stress reactivity that has been observed in infant rats approximately during postnatal days 4–14, which functions to protect the immature brain from excess glucocorticoids during sensitive periods of brain development (Rosenfeld, Suchecki, & Levine, 1992). In addition to the increasing influence of caregiver support on cortisol reactivity, a major methodological factor that may be associated with the observed pattern of decreased cortisol 4 reactivity across the first year is that after the first months, a vast majority of studies have used only psychological stressor paradigms to trigger cortisol reactivity. In reviews of the literature, both Gunnar et al. (2009) and Jansen, Beijers, Riksen-Walraven, and de Weerth (2010) noted that while physical stressors generally produced robust increases in cortisol in young infants, studies using psychological stressors in older infants produced more mixed results, with many studies failing to show increases in cortisol at the group level after a psychological stressor. Jansen et al. (2010) averaged effect sizes from multiple studies and estimated that the average effect size of studies using physical stressors to trigger cortisol responses was d = .62 for mild physical stressors and d = 1.10 for painful stimuli. For psychological stressors, results were much more variable with the averaged effect sizes ranging from d = –.15 (novelty paradigms) to d = .37 (separation paradigms). As psychological stressor paradigms provide opportunities to investigate infants’ hormonal responses to stressors that they may also encounter in real life (e.g., separation from the parent), it is important to systematically investigate to what extent psychological stressors presented in a laboratory environment trigger cortisol increases in infants and what factors might affect the magnitude of the responses. In the present study, we used meta-analysis to provide the first quantitative synthesis of studies reporting salivary cortisol responses to psychological stressors in infants between 3 and 18 months of age. Our aim was to analyze whether infants show a measurable cortisol response to psychological stressors and whether the magnitude of the responses is affected by the type of stressor (e.g., separation or novelty), age of the infant, or other potential moderating factors. For individual differences research, it is critical to design paradigms that are able to trigger variable amounts of cortisol reactivity across most infants, rather than relying on the presence or absence of a response (cf. Roos et al., 2017). Therefore, we specifically conducted the meta-analyses on data from normative samples of infants without major risk factors (e.g., preterm birth or parental substance abuse) to derive estimates of cortisol reactivity to psychological stressors in typically developing infants. Furthermore, we focused on studies conducted in the laboratory to 5 obtain a sufficiently standardized set of studies investigating infants’ stress reactivity in a novel but controlled environment. In addition to the reviews of the literature on infants’ salivary cortisol responses to various types of stressors (Gunnar et al., 2009; Jansen et al., 2010), only one meta-analysis has been published previously (Provenzi, Giusti, & Montirosso, 2016). However, Provenzi, Giusti, and Montirosso (2016) focused only on one type of psychological stressor; disruption of parental interaction in the Still-Face paradigm (SFP; Tronick, Als, Adamson, Wise, & Brazelton, 1978), in which the infant’s interaction with the parent is disrupted by the parent becoming unresponsive for a brief period. The analysis indicated that, across 10 studies, interaction disruption in the SFP is associated with a small but statistically significant cortisol increase. As the SFP is typically administered to infants only across a narrow age range of 4–8 months, it is difficult to estimate potential age-related changes in cortisol responses from the SFP data alone. Therefore, in addition to studies using the SFP, we included various other types of psychological stressors administered to infants at various ages, such as separation from the parent, exposure to novel objects or persons, and frustration caused by blocking the infant’s goal. In addition to the influence of stressor type and age on cortisol responses, we considered other factors that may moderate the effect sizes across studies, such as the material of the saliva collection swab (as synthetic swabs have largely replaced the previously used cotton swabs), time of day of the saliva collection, and the “warm-up” time between the infant’s arrival to the laboratory and the baseline saliva collection (i.e., acclimation period), as there are indications that infants’ initial cortisol levels decrease if they are given time to familiarize with the novel laboratory environment (Zmyj, Schneider, & Seehagen, 2017). 6 2. Methods 2.1. Literature Search The literature search was conducted by using PsycINFO and MEDLINE databases. We searched peer-reviewed articles in the English language available by July 18th, 2019, with the keywords cortisol and child* and/or infant* in the title or abstract (the asterisk indicating that the search contained the word or word fragment). The specific search term was ”(cortisol.ti. or cortisol.ab.) and ((infan$ or child$).ti. or (infan$ or child$).ab.)”. The search produced a total of 5,689 articles after removal of duplications. First, titles and abstracts of the articles were screened to exclude studies that were clearly outside the scope of the current analyses (studies with adult or non-human participants, and studies using physical stressors), which left 252 articles for full text screening. The full texts were screened using the following inclusion criteria: the study was conducted in a laboratory setting, a psychological stressor was used, the subjects were under 18 months old human infants who were healthy and did not belong to a high-risk group (pre-term birth, prenatal substance exposure, major postnatal health problem, parental psychiatric disorder), and the study included cortisol measurement from saliva before and after presentation of the stressor. From studies comparing control and risk groups, data from the control group was included in the metaanalysis to best reflect cortisol reactivity in typically developing infants. Figure 1 describes the study selection process. In case of duplicate samples (i.e., the same infant providing data for two or more studies), we chose the largest or most representative study. Articles in which the same infants were assessed at multiple timepoints, the first assessment was included with the exception of one study (Montirosso, Tronick, Morandi, Ciceri, & Borgatti, 2013), where assessments of two separate groups were included (the control group and the assessment of the first stressor exposure of the experimental group). In instances where cortisol means and standard deviations were not reported in the articles, 7 the authors were contacted for the relevant information. For six articles, the required values were obtained from the systematic review by Jansen et al. (2010). Overall, a total of 49 effect sizes from 48 articles were included in this meta-analysis. 2.2. Moderators Stressor type was the most relevant categorical moderator in the analysis. A total of 18 different stressors were observed in the set of studies. These stressors were grouped into four categories: 1) Separation, including studies in which the infant is separated temporarily from the parent (e.g., the Strange Situation Procedure, SSP; Ainsworth, Blehar, Waters, & Wall, 1978), 2) Frustration, e.g., studies using the Arm Restraint Procedure (Stifter & Fox, 1990) in which infants are prevented to move their arms for a short period of time, 3) Novelty, in which the infant is presented with a novel situation or object, (e.g., Grunau, Weinberg, & Whitfield, 2004), and 4) Interaction Disruption, which includes stressors utilizing the Still-Face paradigm (SFP; Tronick et al., 1978). The chosen categories are similar to the categories used by Gunnar et al. (2009) and Jansen et al. (2010) for the first three. The fourth category (Interaction Disruption) was added because the usage of this stressor type has increased notably during the last decade (Provenzi, Giusti, & Montirosso, 2016). The Interaction Disruption studies included in the current analyses used either the 3or 5-episodic SFP. The traditional 3-episodic stressor by Tronick et al. (1978) consists of a play, still-face, and reunion episodes (A-B-A), whereas the modified 5-episodic version includes additional still-face and reunion episodes (A-B-A-B-A; Haley & Stansbury, 2003). In addition to stressor type, the material of the saliva collection instrument (cotton vs. synthetic) and the time of day when the study was conducted (between 8–12 or 12–18) were added as categorical moderators. Participants’ mean age (in months) was the most relevant continuous moderator in the analysis. The other continuous moderator was the duration of the acclimation period (in minutes). Descriptive information about the studies included in this meta-analysis is presented in Table 1. 8 2.3. Meta-Analytic Procedures The meta-analyses were performed using the Comprehensive Meta-Analysis program (CMA, version 3.0; Borenstein, Rothstein, & Cohen, 2013). Effect sizes (Hedges’ g) were calculated for each study as a standardized mean difference between cortisol levels before and after presentation of stressor. Increase in cortisol level was defined as a positive effect size. Hedges’ g was chosen as the effect size because Cohen’s dtends to overestimate the population effect size especially when the sample size is smaller than 50 (Cumming, 2012; Lakens, 2013). In terms of interpretation, the effect sizes expressed as Hedges’ gand Cohen’s dare directly comparable. Means (M), standard deviations (SD) and correlations (r) were primarily used in calculating effect sizes. In cases where these values were not reported, effect sizes were calculated using the mean difference (one study) or the correlation coefficient (one study). Correlations between the cortisol assessment points were seldom reported in the studies, thus we estimated the correlation to be r= .60 based on the studies which reported the coefficient. In the current set of studies, cortisol was measured 1–2 times before the presentation of stressor and 1–3 times after the stressor. Cortisol assessment points at the baseline and 20–30 minutes after the stressor were chosen to obtain a comparable set of studies reflecting the typical temporal pattern of cortisol reactivity (Gunnar et al., 2009). First, the main analysis including all studies produced a combined weighted effect size (Hedges’ g) and its 95% confidence interval (95% CI). Statistical testing for heterogeneity of the effect sizes was carried out using Q-tests. The main analysis was carried out using random-effects models because of the variation in study practices (Borenstein, Hedges, Higgins, & Rothstein, 2009; Field & Gillett, 2010). Following this model selection, we wanted to examine variance of the true effects in addition to the combined effect size. Estimates of variance (T2) and standard deviation (T) of the true effect sizes, and proportion of true variance from observed variance (I2) were thus calculated. Finally, we calculated a 95% prediction interval (95% PI) for the combined weighted 9 effect size, which is the estimated distribution of the true effect sizes around the combined weighted effect size. In addition to the p-value of the Q-test, the certainty of this heterogeneity estimate was examined by calculating 95% confidence intervals for I2. Combined weighted effect sizes were calculated also for the subgroups of the categorical moderators. Differences between subgroups were examined with Q-tests and these analyses were conducted using mixed-effect models, assuming no common among study variance across subgroups (Borenstein et al., 2009). Proportion of true variance of the observed variance (I2) was calculated for each subgroup. Finally, we calculated 95% prediction intervals for the combined weighted effect sizes of each subgroup. Meta-regression was used to examine effects of the continuous moderators on the effect sizes. Effects of age and duration of the acclimation period were examined for all effect sizes and additionally for the stressor type subgroups. Using the CMA program, we calculated regression coefficients (b) which were tested for statistical significance (Q) and calculated the proportion of total between-study variance explained by the model (R2). Potential outliers were screened by calculating Fisher’s Z-values for each effect size and standardizing them. No outliers (standardized Z-value >|3.29|; Tabachnick & Fidell, 2001) were observed. We examined potential publication bias graphically by observing funnel plots of the effect sizes against their standard errors. Additionally, we calculated Egger’s t-test (Egger, Davey Smith, Schneider, & Minder, 1997) to examine statistical asymmetry in the funnel plot and used Rosenthal’s fail-safe method (Rosenthal, 1991) to estimate the number of missing studies that would be required to nullify the effect. Finally, we used p-curve-analysis (Simonsohn, Nelson, & Simmons, 2014) to test whether the distribution of significant p-values was right-skewed towards smaller p-values, thus indicating that the published results have evidential value. 16 cortisol responses to stressors were larger when the infants were given more time to acclimatize to the novel laboratory environment. As information about the length of the acclimation period was provided by less than half of the studies, the results should be interpreted with caution. Nevertheless, as this association suggests that initially high cortisol levels directly upon entering the laboratory may mask some of the effects triggered by the stressor presentation, in future research it will be important to strive for more standardized laboratory protocols including fixed and sufficiently documented acclimation periods. Additional studies on the effects of varying acclimation periods on cortisol levels are also needed to ascertain what amount of time, on average, is enough for cortisol levels to return to a regular level after arrival to the laboratory. Effect sizes were not affected by the time of day of the saliva collection in the main analysis nor when analyzing stressor types separately. Even though in our analysis the included cortisol measurement points were relatively close to each other (the gap between preand post-stressor measurements was 17–56 min), these results are consistent with studies on the development of the circadian rhythm in childhood (Gunnar & Donzella, 2002) in which differences in cortisol levels between the morning and mid-afternoon have not been observed in children until 4 years of age. Regarding the influence of different saliva collection instruments, our analysis did not indicate that using cotton-based instruments results in systematically lower cortisol levels than when using synthetical saliva swabs, although such differences have been observed in previous studies with adults (e.g., Gröschl & Rauh, 2006). It should be noted that we focused on cortisol reactivity and did not compare cortisol levels in single measurement points between studies. Therefore, it is possible that differences in cortisol levels caused by collection material are masked because cortisol reactivity accounts only for the difference between cortisol measurements within a study and not between studies using cotton or synthetic instruments. Nevertheless, given that direct comparisons have indicated differences between synthetic and cotton-based instruments (Gröschl & 17 Rauh, 2006), we recommend systematically using synthetic swabs in studies assessing infants’ cortisol. In the future, it will also be important to investigate with sufficiently large samples whether there are sex differences in infants’ cortisol reactivity, because results from previous studies have been mixed (e.g., Davis & Emory, 1995; Kirschbaum, Wüst, & Hellhammer, 1992; Tout, Haan, Campbell, & Gunnar, 2008). In the present meta-analysis, we were able to obtain only information about the sex distributions of some of the studies, thus analyzing sex differences in cortisol reactivity in a reliable way was not possible. Another potentially relevant moderator we were not able to examine for the same reasons as with sex was race/ethnicity. A recent study by Dismukes et al. (2018) indicated that race may moderate cortisol reactivity in 12-month-old infants in the SFP. This result in addition to results from adults (e.g., Chong, Uhart, McCaul, Johnson, & Wand, 2008; Wilcox, Bopp, Wilson, Fulk, & Hand, 2005) underlines the need for more research on potential race differences on infants’ cortisol reactivity. 4.4. Conclusions Our main analysis found that commonly used psychological stressor paradigms are associated with a small increase in salivary cortisol concentrations in typically developing infants. Investigation of a large number of studies also highlighted the large heterogeneity in the magnitude of the effects and also in the procedures of the included studies. Age, time of day, and saliva collection instrument material did not significantly moderate the effect sizes in the full set of studies or within the separate stressor types, but the duration of the acclimation period was found to be associated with the magnitude of cortisol reactivity. Although a clear association between infant age and cortisol reactivity was not found, the most consistent cortisol responses were observed in the separation studies, which were also administered to infants who were older than infants in the other stressor categories. While this finding appears inconsistent with studies showing dampened cortisol 18 responses to physical stressors in older infants (Davis & Granger, 2009; Gunnar et al., 1996), determining the age-related changes in cortisol reactivity to psychological stressors reliably requires that comparable psychological stressors are applied in a similar manner to a wide age range of infants. Our recommendation would be to use age-appropriate variations of parental separation as the most feasible psychological stressor across infancy. In conclusion, a somewhat concerning overall outcome of this analysis was that despite the large number of studies, cortisol reactivity to psychological stressor paradigms in infancy appears to be small and inconsistent. This is problematic for research aiming to characterize individual differences in stress reactivity to naturally occurring stressors. Without a clear understanding of cortisol reactivity to psychological stressors in typically developing infants, determining what constitutes an endocrine marker of risk for the development of the stress systems remains very challenging. The contribution of this meta-analysis is to help to elucidate how efficiently the currently used experimental paradigms trigger cortisol reactivity in typically developing infants, which will be important for attempts to design reliable stressor paradigms for research on individual differences in infants’ cortisol reactivity to stress. 19 Acknowledgments This research was supported by a grant from the Academy of Finland (#307657) to M.J.P. Conflicts of interest: none. 20 References (Studies included in the meta-analysis are marked with an asterisk) Ainsworth, M. D. S., Blehar, M. C., Waters, E., & Wall, S. (1978). Patterns of attachment: A psychological study of the strange situation. Hillsdale, NJ: Erlbaum. *Atkinson, L., Gonzalez, A., Kashy, D. A., Santo Basile, V., Masellis, M., Pereira, J., … Levitan, R. (2013). Maternal sensitivity and infant and mother adrenocortical function across challenges. Psychoneuroendocrinology,38(12), 2943–2951. https://doi.org/10.1016/J.PSYNEUEN.2013.08.001 *Azar, R., Paquette, D., Zoccolillo, M., Baltzer, F., & Tremblay, R. E. (2007). The association of major depression, conduct disorder, and maternal overcontrol with a failure to show a cortisol buffered response in 4-month-old infants of teenage mothers. Biological Psychiatry,62(6), 573–579. https://doi.org/10.1016/j.biopsych.2006.11.009 *Beijers, R., Riksen-Walraven, M., Sebesta, K., & de Weerth, C. (2017). Differential associations between behavioral and cortisol responses to a stressor in securely versus insecurely attached infants. Behavioural Brain Research,325, 147–155. https://doi.org/10.1016/j.bbr.2016.10.008 *Bernard, K., & Dozier, M. (2010). Examining infants’ cortisol responses to laboratory tasks among children varying in attachment disorganization: Stress reactivity or return to baseline? Developmental Psychology,46(6), 1771–1778. https://doi.org/10.1037/a0020660 *Bernard, N. K., Kashy, D. A., Levendosky, A. A., Bogat, G. A., & Lonstein, J. S. (2017). Do different data analytic approaches generate discrepant findings when measuring mother-infant HPA axis attunement? Developmental Psychobiology,59(2), 174–184. https://doi.org/10.1002/dev.21474 Borenstein, M., Hedges, L. V., Higgins, J. P. T., & Rothstein, H. R. (2009). Introduction to MetaAnalysis. Chichester, UK: John Wiley & Sons, Ltd. https://doi.org/10.1002/9780470743386 Borenstein, M., Rothstein, D., & Cohen, J. (2013). Comprehensive meta-analysis: A computer 21 program for research synthesis. Englewood, NJ: Biostat. *Bosquet Enlow, M., King, L., Schreier, H. M., Howard, J. M., Rosenfield, D., Ritz, T., & Wright, R. J. (2014). Maternal sensitivity and infant autonomic and endocrine stress responses. Early Human Development,90(7), 377–385. https://doi.org/10.1016/j.earlhumdev.2014.04.007 Chong, R. Y., Uhart, M., McCaul, M. E., Johnson, E., & Wand, G. S. (2008). Whites have a more robust hypothalamic-pituitary-adrenal axis response to a psychological stressor than blacks. Psychoneuroendocrinology,33(2), 246–254. https://doi.org/10.1016/j.psyneuen.2007.10.014 *Conradt, E., Fei, M., LaGasse, L., Tronick, E., Guerin, D., Gorman, D., … Lester, B. M. (2015). Prenatal predictors of infant self-regulation: The contributions of placental DNA methylation of NR3C1 and neuroendocrine activity. Frontiers in Behavioral Neuroscience,9, 130. https://doi.org/10.3389/fnbeh.2015.00130 Cumming, G. (2012). Understanding the new statistics: Effect sizes, confidence intervals, and metaanalysis. New York, NY: Routledge. Davis, E. P., & Granger, D. A. (2009). Developmental differences in infant salivary alpha-amylase and cortisol responses to stress. Psychoneuroendocrinology,34(6), 795–804. https://doi.org/10.1016/J.PSYNEUEN.2009.02.001 Davis, M., & Emory, E. (1995). Sex differences in neonatal stress reactivity. Child Development, 66(1), 14–27. https://doi.org/10.1111/j.1467-8624.1995.tb00852.x Dickerson, S. S., & Kemeny, M. E. (2004). Acute stressors and cortisol responses: A theoretical integration and synthesis of laboratory research. Psychological Bulletin,130(3), 355–391. https://doi.org/10.1037/0033-2909.130.3.355 *DiCorcia, J. A., Snidman, N. C., & Tronick, E. (2018). Here today, gone tomorrow? Changes in 4month-olds’ physiologic and behavioral responses do not indicate memory for a social stressor. Frontiers in Psychology,9, 128. https://doi.org/10.3389/fpsyg.2018.00128 *Diego, M. A., Field, T., Jones, N. A., Hernandez-Reif, M., Cullen, C., Schanberg, S., & Kuhn, C. 22 (2004). EEG responses to mock facial expressions by infants of depressed mothers. Infant Behavior & Development,27(2), 150–162. https://doi.org/10.1016/j.infbeh.2003.10.001 *Dismukes, A., Shirtcliff, E., Jones, C. W., Zeanah, C., Theall, K., & Drury, S. (2018). The development of the cortisol response to dyadic stressors in Black and White infants. Development and Psychopathology,30(5), 1995–2008. https://doi.org/10.1017/S0954579418001232 Egger, M., Davey Smith, G., Schneider, M., & Minder, C. (1997). Bias in meta-analysis detected by a simple, graphical test. British Medical Journal,315(7109), 629–634. https://doi.org/10.1136/bmj.315.7109.629 *Eiden, R. D., Molnar, D. S., Granger, D. A., Colder, C. R., Schuetze, P., & Huestis, M. A. (2015). Prenatal tobacco exposure and infant stress reactivity: Role of child sex and maternal behavior. Developmental Psychobiology,57(2), 212–225. https://doi.org/10.1002/dev.21284 *Eiden, R. D., Veira, Y., & Granger, D. A. (2009). Prenatal cocaine exposure and infant cortisol reactivity. Child Development,80(2), 528–543. https://doi.org/10.1111/j.14678624.2009.01277.x *Erickson, S. J., MacLean, P., Qualls, C., & Lowe, J. R. (2013). Differential associations between infant affective and cortisol responses during the Still Face paradigm among infants born very low birth weight versus full-term. Infant Behavior and Development,36(3), 359–368. https://doi.org/10.1016/J.INFBEH.2013.03.001 *Feldman, R., Singer, M., & Zagoory, O. (2010). Touch attenuates infants’ physiological reactivity to stress. Developmental Science,13(2), 271–278. https://doi.org/10.1111/j.14677687.2009.00890.x Field, A. P., & Gillett, R. (2010). How to do a meta-analysis. British Journal of Mathematical and Statistical Psychology,63(3), 665–694. https://doi.org/10.1348/000711010X502733 *Frigerio, A., Ceppi, E., Rusconi, M., Giorda, R., Raggi, M. E., & Fearon, P. (2009). The role 23 played by the interaction between genetic factors and attachment in the stress response in infancy. Journal of Child Psychology and Psychiatry,50(12), 1513–1522. https://doi.org/10.1111/j.1469-7610.2009.02126.x *Goldberg, S., Levitan, R., Leung, E., Masellis, M., Basile, V. S., Nemeroff, C. B., & Atkinson, L. (2003). Cortisol concentrations in 12to 18-Month-Old infants: Stability over time, location, and stressor. Biological Psychiatry,54(7), 719–726. https://doi.org/10.1016/S00063223(03)00010-6 Goldsmith, H. H., & Rothbart, M. K. (1999). Laboratory Temperament Assessment Battery (LabTAB), Version 3.1. Wisconsin Temperament and Twin Center, University of Wisconsin, Madison. *Grant, K.-A., McMahon, C., Austin, M.-P., Reilly, N., Leader, L., & Ali, S. (2009). Maternal prenatal anxiety, postnatal caregiving and infants’ cortisol responses to the still-face procedure. Developmental Psychobiology,51(8), 625–637. https://doi.org/10.1002/dev.20397 Gröschl, M., & Rauh, M. (2006). Influence of commercial collection devices for saliva on the reliability of salivary steroids analysis. Steroids,71(13–14), 1097–1100. https://doi.org/10.1016/J.STEROIDS.2006.09.007 *Grunau, R. E., Weinberg, J., & Whitfield, M. F. (2004). Neonatal procedural pain and preterm infant cortisol response to novelty at 8 months. Pediatrics,114(1), e77–e84. https://doi.org/10.1542/peds.114.1.e77 Gunnar, M., & Quevedo, K. (2007). The neurobiology of stress and development. Annual Review of Psychology,58, 145–173. https://doi.org/10.1146/annurev.psych.58.110405.085605 Gunnar, M. R., Brodersen, L., Krueger, K., & Rigatuso, J. (1996). Dampening of adrenocortical responses during infancy: Normative changes and individual differences. Child Development, 67(3), 877–889. https://doi.org/10.1111/j.1467-8624.1996.tb01770.x Gunnar, M. R., & Donzella, B. (2002). Social regulation of the cortisol levels in early human 24 development. Psychoneuroendocrinology,27(1–2), 199–220. https://doi.org/10.1016/S03064530(01)00045-2 *Gunnar, M. R., Larson, M. C., Hertsgaard, L., Harris, M. L., & Brodersen, L. (1992). The stressfulness of separation among nine-month-old infants: Effects of social context variables and infant temperament. Child Development,63(2), 290–303. https://doi.org/10.1111/j.14678624.1992.tb01627.x *Gunnar, M. R., Mangelsdorf, S., Larson, M., & Hertsgaard, L. (1989). Attachment, temperament, and adrenocortical activity in infancy: A study of psychoendocrine regulation. Developmental Psychology,25(3), 355–363. https://doi.org/10.1037/0012-1649.25.3.355 *Gunnar, M. R., & Nelson, C. A. (1994). Event-related potentials in year-old infants: Relations with emotionality and cortisol. Child Development,65(1), 80. https://doi.org/10.1111/j.14678624.1994.tb00736.x Gunnar, M. R., Talge, N. M., & Herrera, A. (2009). Stressor paradigms in developmental studies: What does and does not work to produce mean increases in salivary cortisol. Psychoneuroendocrinology,34(7), 953–967. https://doi.org/10.1016/j.psyneuen.2009.02.010 *Haley, D. W. (2011). Relationship disruption stress in human infants: A validation study with experimental and control groups. Stress,14(5), 530–536. https://doi.org/10.3109/10253890.2011.560308 Haley, D. W., & Stansbury, K. (2003). Infant stress and parent responsiveness: Regulation of physiology and behavior during Still-Face and reunion. Child Development,74(5), 1534–1546. https://doi.org/10.1111/1467-8624.00621 *Hibel, L. C., Trumbell, J. M., Valentino, K., & Buhler-Wassmann, A. C. (2018). Ecologically salient stressors and supports and the coordination of cortisol and salivary alpha-amylase in mothers and infants. Physiology & Behavior,195, 48–57. https://doi.org/10.1016/j.physbeh.2018.07.024 25 *Holochwost, S. J., Gariépy, J.-L., Mills-Koonce, W. R., Propper, C. B., Kolacz, J., & Granger, D. A. (2017). Individual differences in the activity of the hypothalamic pituitary adrenal axis: Relations to age and cumulative risk in early childhood. Psychoneuroendocrinology,81, 36– 45. https://doi.org/10.1016/j.psyneuen.2017.03.023 *Huot, R. L., Brennan, P. A., Stowe, Z. N., Plotsky, P. M., & Walker, E. F. (2004). Negative affect in offspring of depressed mothers is predicted by infant cortisol levels at 6 months and maternal depression during pregnancy, but not postpartum. Annals of the New York Academy of Sciences,1032(1), 234–236. https://doi.org/10.1196/annals.1314.028 Jansen, J., Beijers, R., Riksen-Walraven, M., & de Weerth, C. (2010). Cortisol reactivity in young infants. Psychoneuroendocrinology,35(3), 329–338. https://doi.org/10.1016/j.psyneuen.2009.07.008 Kirschbaum, C., Wüst, S., & Hellhammer, D. (1992). Consistent sex differences in cortisol responses to psychological stress. Psychosomatic Medicine,54(6), 648–657. https://doi.org/10.1097/00006842-199211000-00004 *Kuo, P. X., Saini, E. K., Tengelitsch, E., & Volling, B. L. (2019). Is one secure attachment enough? Infant cortisol reactivity and the security of infant-mother and infant-father attachments at the end of the first year. Attachment & Human Development,21(5), 426–444. https://doi.org/10.1080/14616734.2019.1582595 Lakens, D. (2013). Calculating and reporting effect sizes to facilitate cumulative science: A practical primer for t-tests and ANOVAs. Frontiers in Psychology,4:863. https://doi.org/10.3389/fpsyg.2013.00863 *Larson, M. C., Gunnar, M. R., & Hertsgaard, L. (1991). The effects of morning naps, car trips, and maternal separation on adrenocortical activity in human infants. Child Development,62(2), 362–372. https://doi.org/10.1111/j.1467-8624.1991.tb01537.x *Laurent, H. K., Duncan, L. G., Lightcap, A., & Khan, F. (2017). Mindful parenting predicts 32 TABLE 1. (continued) Study + year NAge (months) Gender (% boys) Stressor Stressor type Duration of acclimation period (min) Material of the saliva collection instrument Time of study Provenzi et al. (2016) 53 3.31 50.9 Modified SFP a SF Not reported Cotton 8–12 Reck et al. (2013) 49 4.19 64.4 SFP SF Not reported Cotton Not reported Schuetze et al. (2008) 40 7.00 Not reported Lab-TAB b Frustration 5.0 Cotton 8–18 Seehagen et al. (2015) 12 14.92 46.2 SSP g Separation 40.0 Synthetic 8–12 Spangler & Grossmann (1993) 32 12.10 58.5 SSP Separation Not reported Cotton 8–12 Spangler & Schieche (1998) 106 12.40 Not reported SSP Separation Not reported Cotton 8–12 Thomas et al. (2017) 272 6.00 52.2 Lab-TAB f Frustration Not reported Synthetic Not reported van Bakel & Riksen-… (2004) 85 15.10 52.0 SSP g Separation 46.0 Cotton 8–12 Waters et al. (2013) 152 12.80 Not reported Emotional setting Novelty 25.0 Synthetic 12–18 Weisman et al. (2013) 35 5.00 48.6 SFP SF 45.0 Synthetic 12–18 N = sample size; 1 = control group; 2 = experimental group (1. exposure); SSP = Strange Situation Procedure (Ainsworth et al., 1978); ARP = Arm Restraint Procedure (Stifter & Fox, 1990); SFP = traditional 3-episodic Still-Face paradigm (Tronick et al., 1978). a = 5-episodic SFP (Haley & Stansbury, 2003). b = stressor: puppet-show and the ARP, in which the mother prevents the infant’s play (Lab-TAB; Goldsmith & Rothbart, 1999). c = 3-episodic SFP, which includes an additional one-minute separation from the mother. d = stressor: 90 s still-face and the ARP, in which the mother prevents the infant’s play (Lab-TAB; Goldsmith & Rothbart, 1999). e = stressor: car seat restraint, 3 loud noises and the ARP. f = stressor: toy removal, toy barrier and the ARP, in which the mother prevents the infant’s play (Lab-TAB; Goldsmith & Rothbart, 1999). g = modified SSP: interaction with a robot included. 33 k = number of effect sizes; N = total sample size; g = effect size (Hedges’ g); SE = standard error of effect size; 95% CI (g) = 95% confidence interval around effect size; T2 = estimated true variance of effect sizes; T = estimated true standard deviation of effect sizes; I2 = proportion of true variance; 95% CI (I2) = 95% confidence (”uncertainty”) interval for proportion of true variance; 95% PI (g) = 95% prediction interval around effect size; Q-W = a statistic testing for the homogeneity within a subgroup; Q-B = a statistic testing for the significance of the contrast between subgroups; 3-ep = 3-episodic SFP; 5-ep = 5-episodic SFP. *p< .05. **p< .01. ***p< .001. TABLE 2. Categorical moderators k N5g.SE 95% CI (g)T2T I295% CI (I2) 95% PI (g)Q-W*Q-B* Stressor type 8.13* Separation 18 1752*.15*** .05 –.01–.24 .02 .15-71.21% 51.22–80.96 –.19–.49 59.06*** Frustration 10 1088*.00*** .04 –.08–.07 .01 .08-47.81% –9.05–71.36 –.21–.21 17.25*** Novelty 5 340*.12*** .08 –.04–.28 .02 .13-57.07% –15.74–84.08 –.32–.56 9.32*** SF 15 915*.16*** .08 –.00–.32 .08 .28-83.62% 74.31–89.56 –.46–.78 85.48*** 1.55* 3-ep 7 552*.06*** .09 –.12–.23 .04 .20-77.14% 52.26–89.05 –.51–.63 26.24*** 5-ep 8 363*.25*** .13 –.00–.50 .11 .33-83.61% 69.27–91.26 –.60–1.11 42.72*** Material .03* Cotton 27 1571*.12*** .05 –.03–.22 .04 .20-71.25% 56.16–79.66 −.30−.55 90.43*** Synthetic 19 2327*.11* .04 –.03–.20 .03 .16-80.19% 68.34–86.19 −.25−.47 90.86*** Time of study .33* Morning 19 1356*.19** .06 –.10–.31 .05 .23-79.99% 68.00–86.06 −.32−.69 89.97*** Afternoon 12 1413*.14** .06 –.03–.25 .02 .16-74.89% –52.24–84.29 –.23–.51 43.81*** 34 TABLE 3. Meta-regression statistics for the continuous moderators in the full set of studies and separately by stressor type Age (months) k-N b SE 95% CI Q R2* Total 48 4095 –.00 .01 –.01–.02 .21** .00a Separation 18 1752 –.02 .02 –.06–.02 1.03** .00 a Frustration 10 1360 –.01 .02 –.02–.05 .60** .00 a Novelty 5 340 –.02 .02 –.03–.06 .55** .00 a SF 15 915 –.06 .07 –.19–.07 .79** .00 a Duration of acclimation period (min) k-N b SE 95% CI Q R2* Total 22 1880 –.01 .00 –.00–.01 –6.36**.29a Separation 7 619 –.02 .01 –.00–.03 7.53** .38 a Frustration 5 559 –.01 .00 –.00–.01 5.72**N/A a Novelty 4 318 –.01 .00 –.00–.02 8.62** N/A a SF 6 384 –.01 .00 –.02–.00 3.68** N/A a k = number of effect sizes; N = sample size; b = regression coefficient; SE = standard error of coefficient; 95% CI = 95% confidence interval around the coefficient; Q = statistic testing that the coefficient is zero; R2 = proportion of total between-study variance explained by the moderator; a = the model’s goodness-of-fit test is non-significant and, therefore, the R2 values are not valid. *p< .05. **p< .01 35 FIGURE 1. Flowchart of the study selection process. Database search PsycINFO N= 2452 N= 1723 duplicates removed N= 5437 excluded N= 252 articles in full-text screening N= 205 excluded •Inclusion criteria not met •Duplicate samples N= 48 effect sizes included in the meta-analysis N = 5689 Articles screened on basis of title and abstract for inclusion criteria Database search MEDLINE N= 4960 N= 1 study included from reference lists 36 FIGURE 2. Effect sizes and forest plot of studies included in meta-analysis grouped by stressor type Separation Study + year g p N Age Hedges’ g, 95% CI and 95% PI Seehagen et al. (2015) .933 .002 12 14.92 Larson et al. (1991) .659 .000 27 9.50 van Bakel & Riksen-... (2004) .570 .000 85 15.10 Beijers et al. (2017) .368 .000 183 12.39 Gunnar et al. (1989) .345 .063 51 13.00 Gunnar et al. (1992) .239 .098 38 9.10 Spangler & Grossmann (1993) .204 .192 32 12.10 Goldberg et al. (2003) .195 .248 27 14.10 Luijk et al. (2010) .138 .007 310 14.70 Bernard & Dozier (2010) .119 .470 28 15.20 Atkinson et al. (2013) .097 .072 277 16.95 Kuo et al. (2019) .006 .929 144 12.00 O’Connor et al. (2013) .000 1.000 108 16.77 Spangler & Schieche (1998) .000 1.000 106 12.40 Martinez-Torteya et al. (2018) .000 1.000 92 11.86 Nachmias et al. (1996) −.011 .916 77 18.00 Gunnar & Nelson (1994) −.089 .480 49 11.90 Frigerio et al. (2009) −.301 .128 106 14.60 Total .149 .001 1752 13.59 −1.00 −.50 .00 .50 Frustration Study + year g p N Age Hedges’ g, 95% CI and 95% PI Hibel et al. (2018) .213 .014 107 5.90 Bernard et al. (2017) .118 .100 155 11.80 Thomas et al. (2017) .037 .466 272 6.00 Lewis et al. (2006) .030 .801 56 4.00 Huot et al. (2004) .017 .871 71 6.00 Eiden et al. (2015) −.017 .873 69 9.00 Holochwost et al. (2017) −.095 .260 111 6.00 Azar et al. (2007) −.100 .208 126 4.38 Eiden et al. (2009) −.165 .124 81 7.00 Schuetze et al. (2008) −.190 .099 40 7.00 Total −.004 .914 1088 6.71 −1.00 −.50 .00 .50 (figure continues on the next page) 1.00 1.00 37 FIGURE 2. (continued) Novelty Study + year g p N Age Hedges’ g, 95% CI and 95% PI Waters et al. (2013) .287 .000 152 12.80 Grunau et al. (2004) .220 .236 22 8.00 Oberlander et al. (2008) .154 .242 45 3.00 Laurent et al. (2016) −.024 .787 98 10.01 Diego et al. (2004) −.101 .576 23 4.12 Total .121 .141 340 7.59 −1.00 −.50 .00 .50 Still-Face Study + year g p N Age Hedges’ g, 95% CI and 95% PI Haley (2011) 1.124 .000 16 5.57 Feldman et al. (2010) .737 .000 33 5.90 Conradt et al. (2015) .639 .000 128 4.39 Montirosso et al. (2013) 2 .296 .045 37 4.00 Provenzi et al. (2016) .209 .088 53 3.31 Montirosso et al. (2013) 1 .199 .170 37 4.46 Dismukes et al. (2018) .118 .057 207 4.00 Lewis & Ramsay (2005) .103 .288 84 5.90 DiCorcia et al. (2018) .084 .580 33 3.68 Laurent et al. (2017) −.027 .811 73 6.00 Weisman et al. (2013) −.108 .643 35 5.00 Grant et al. (2009) −.112 .286 71 7.20 Bosquet Enlow et al. (2014) −.113 .528 35 6.39 Reck et al. (2013) −.252 .049 49 4.19 Erickson et al. (2013) −.274 .129 24 7.03 Total .157 .051 915 5.13 −1.00 −.50 .00 .50 Total (combined) .109 .000 4095 8.89 −1.00 −.50 .00 .50 g = effect size (Hedges’ g); 95% CI = 95% confidence interval around the effect size; 95% PI = 95% prediction interval around the effect size; 1 = control group; 2 = experimental group (first exposure). Age is expressed in months. Effect sizes are arranged from largest to smallest from top to bottom. Size of the square represents the weight of the individual study within the stressor category (larger size equals larger weight) and the horizontal line the 95% confidence interval. Width of the diamond shape in the total effect sizes represents a 95% confidence interval and the horizontal line the 95% prediction interval. 1.00 1.00 1.00 38 FIGURE 3. Association between the duration of the acclimation period and cortisol reactivity. Each circle represents a study, with a larger radius signifying larger weight (i.e., a smaller standard error). Two lines around the regression line denote the 95% confidence interval for the regression line. 1.40 1.20 010 0.60 0.40 20 30 40 50 60 70 0 . 80 –0.40 –0.60 Hedges’ g 1.00 0.20 0 . 00 –0.20 Acclimation period (min) 39 FIGURE 4. Funnel plot of effect sizes (Hedges’ g) and standard errors (SE) of studies included in the main analysis. The diamond shape represents the total mean effect size. − 2.0 − 1.5 0.5 − 0.5 − 1.0 0 1.0 1.5 2.0 Hedges’ g 0.1 0.2 0.3 0.4 0.0 Standard error (SE) 40 FIGURE 5. p-curve obtained from https://www.p-curve.com/ (version 4.06)