Cortisol.

After each period/task, at the middle of the documentary and at the end of it, the participant was asked to provide saliva samples. These samples reflect activity in the hypothalamus approximately 20 minutes earlier due to the delay in observable cortisol response [71]. Saliva samples were kept in a freezer at −20 degrees Celsius [72] and were collectively sent to the laboratory for analysis. A time-resolved fluorescence immunoassay was implemented to determine the cortisol concentration. Outliers greater than 3 SD above the mean were removed from the analysis due to possible contamination (e.g. blood, medicine).

Heart rate.

Heart rate was measured using a three-lead electrocardiogram and monitored constantly throughout the entire stress procedure. The electrocardiogram was sampled at 512 Hz and stored on a flashcard by means of a portable digital recorder (Vitaport™ System; TEMEC Instruments B.V., Kerkrade, NL). After completion of the recording, physiological data were imported and processed on a Personal Computer using a Vitascore™ software module (TEMEC Instruments BV, Kerkrade, NL). A customized software program calculated the interbeat intervals of the electrocardiogram using R-top detection, resulting in interbeat intervals time series. This time series was inspected to detect and remove artifacts. Heart rate time series were calculated from these interbeat intervals and expressed in beats per minute (bpm); the heart rate time series were subsequently averaged per period during the stress procedure.

Respiratory sinus arrhythmia.

In order to compute an index of RSA, the heart rate time series during the pre-task rest period, the preparation part of the public speaking task, the computer task and the post-task recovery period (i.e. those periods when speaking did not occur) were scrutinized for stationarity. The heart rate time series were subsequently subjected to a discrete Fourier transformation, based on non-equidistant sampling of the R-wave incidences (CARSPAN program, Groningen, The Netherlands; [73], [74]), to yield power spectra of the rhythmic oscillations over a frequency range of 0.02–0.50 Hz, with a resolution of 0.01 Hz. For each period, the power in the high frequency band (0.14–0.5 Hz) of the heart rate time series was calculated as an index of RSA. The data were natural log transformed in order to obtain normal distributions.

For heart rate and PPS, the reactivity measure consisted of the maximum averaged value during the three stressful tasks minus the averaged value of the pre-task rest period. For RSA, we subtracted the minimum averaged value during the tasks from that during the pre-task rest period. Cortisol reactivity was the difference between the maximum value corresponding to cortisol levels during the three stress tasks and the second pre-task value. The first cortisol pre-task value was excluded from the analyses as it was generally higher, thus most likely reflecting anticipatory stress to a greater degree than the second measurement. One extreme outlier in the heart rate difference score was excluded from the analysis.

Individual and developmental factors.

Age, sex and ethnicity were reported by the mother in a general demographics questionnaire. Ethnicity was coded as either of Dutch origin (x = 0) or not of Dutch origin (x = 1). We assessed pubertal stage using self-reported Tanner stages [75] in adolescents only. A Health Questionnaire, developed by Erasmus MC researchers specifically for this study, was used to inquire after general health issues and use of substances. The questionnaire included a parent as well as a self-report version. For all variables the self-report items were used for analysis, except in the cases of missing items, for which the parent report was consulted. In the child sample, menstruation (i.e. are you menstruating yet?, no: x = 0; yes: x = 1) was assessed as a proxy for pubertal development in girls. In the adolescent sample, for free-cycling girls who indicated having a regular cycle (N = 69), menstrual cycle phase was calculated based on the self-reported first day of the last menstrual cycle, coded as either follicular (0–14 days prior to the test session; x = 0) or luteal (15–35 days prior; x = 1). We also assessed OC use (no: x = 0; yes: x = 1) with this questionnaire. Prior to the test session, height and weight were measured in order to calculate BMI. The test assistant obtained the birth weight from a personal record of birth-related variables issued by the hospital upon birth and brought to the test session by the participant. Parent-reported temperament variables were based on the Emotionality, Activity, Sociability scale [76], which consists of 20 items scored on a scale of 1 (not at all) to 5 (very much) and contains four subscales: emotionality (i.e. ‘reacts intensely when upset’), activity (i.e. ‘is very energetic’), sociability (i.e. ‘likes to be with people’) and shyness (i.e. ‘takes a long time to warm up to strangers’). Mother and father reports were averaged. We excluded item 18 (i.e. ‘when alone, child feels isolated’) from the sociability scale in both the child and adolescent samples, and item 7 reversed (i.e. ‘when child moves about, he/she moves slowly’) from the activity scale in the adolescent sample in order to increase reliability. Please see Table 1 for the Cronbach's alphas of the subscales of all questionnaires.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Descriptive statistics of all potential determinants of stress reactivity.

https://doi.org/10.1371/journal.pone.0061724.t001

Environmental factors.

Urbanicity was based on the population rate of the reported home city/town of the participant at the time of the test session, coded as rural (x = 0), town (>10,000 inhabitants; x = 1) or urban (>100,000 inhabitants; x = 2; [77]). Population statistics were based on online national archives [78]. Mother-reported SES was based on the higher occupational level of either parent [79] and coded into low (x = 0), average (x = 1) and high (x = 2) SES. Family situation of the child was based on parent reports and categorized as two-parent household (x = 0) and one-parent household (x = 1). An ‘other’ category was available in the questionnaire, but participants who marked this were excluded from the relevant analyses.

Parenting style was assessed with a modified version of the EMBU-C, a child report version of the EMBU (Egna Minnen Beträffande Uppfostran; a Swedish acronym for My Memories of Upbringing; [80]), which measures the child's perception of his/her upbringing. The version we used is to a large extent in accordance with a shorter Dutch version of the EMBU [81] and contains 47 items in total that were answered on a four-point likert scale from 1 (no, never) to 4 (yes, most of the time). Three scales were derived: emotional warmth (i.e. ‘my mother/father accepted me as I was’, rejection (i.e. ‘it was difficult to approach my mother/father’) and overprotection (i.e. ‘I wished my mother/father would worry less about what I was doing’). For each EMBU-C item, participants assessed both the father's and the mother's parenting behaviors. Items for parents' parenting behaviors were averaged in order to achieve a general view of the parenting environment.

Parent-self-reported parenting practices were assessed using the Alabama Parenting Questionnaire (APQ; [82]). The questionnaire consists of 42 items and five subscales, namely involvement (i.e. ‘you ask your child about his/her day in school’), positive parenting (i.e. ‘you praise your child if he/she behaves well’), poor monitoring (i.e. ‘your child is out with friends you do not know’), inconsistent discipline (i.e. ‘you threaten to punish your child and then do not actually punish him/her’) and corporal punishment (i.e. ‘you slap your child when he/she has done something wrong’). Parents rated on a five-point likert scale ranging from 1 (never) to 5 (always) to what extent they displayed the described parenting behavior. For this study, mother and father reports were averaged. Due to insufficient internal reliability, we excluded the poor monitoring scale in both the child and adolescent samples, as well as the corporal punishment scale in the adolescent sample.

Adverse life events were selected from an extensive Life Events Questionnaire (LEQ; [83]) which included both severely and mildly adverse events as well as positive events, and from the post traumatic stress disorder section of the NIMH Diagnostic Interview Schedule Composite (DISC). Eighteen severely adverse events were chosen from these sources, modelled as closely as possible after Lovallo, et al. [66]. Both the DISC interview and the LEQ were completed by the participant and his/her parent. An event was considered an adverse life event if either the parent or participant confirmed that the event was experienced by the participant. For the LEQ, events were only considered an adverse life event if the informant coded the event as ‘unpleasant’ (for the participant). We excluded the entire scale from the child sample due to insufficient internal reliability.

Substance use-related factors.

We assessed use of substances with the above-mentioned Health Questionnaire. Caffeine use was assessed with questions on ever use of cola and coffee. Similarly, ever use of alcohol (at least one glass) and cannabis were coded as yes (x = 1) or no (x = 0). Tobacco use was coded dichotomously as those who have never smoked, have smoked one or two cigarettes ever and currently smoke once in a while (x = 0); and those who smoke every day (x = 1). Those who had smoked in the past but had quit were excluded from analyses on tobacco use (N = 10). Cola-use was used the analyses in both child and adolescent samples, all other substance use items were used in the adolescent sample only.

Cortisol reactivity.

In children, sex, age, and parenting styles emotional warmth and rejection significantly predicted cortisol reactivity in the exploratory linear regression analyses. In the confirmatory, final model, individuals with male sex, who perceived less emotional warmth from their parents portrayed lower cortisol reactivity to stress. In adolescents, sex, oral contraceptive use, emotionality, sociability, age, urbanicity, parenting styles emotional warmth, involvement and inconsistent discipline, and tobacco use significantly predicted cortisol reactivity independently. In the final model, individuals who scored higher in emotionality and sociability, who lived in more urban areas, and whose parents showed less involvement portrayed blunted cortisol reactivity. Because oral contraceptive use was a significant factor in the first step, we reran the final model in girls only, and oral contraceptive use did not remain a significant determinant. In sum, cortisol reactivity was determined by sex and perceived parental emotional warmth in children, and emotionality, sociability, urbanicity, and parental involvement in adolescents.

Heart rate reactivity.

In children, sex, BMI, menstruation, urbanicity, SES, parenting style positive parenting, and cola use were the initial significant predictors of heart rate reactivity. The final model showed that having a lower SES and a higher urbanicity score predicted lower heart rate reactivity to stress. In the final model in girls only, menstruation did not remain a significant predictor. In adolescents, sex, ethnicity, BMI, oral contraceptive use, sociability, urbanicity, SES, family situation, parenting styles emotional warmth and involvement, and tobacco and cannabis use were initially significant predictors of heart rate reactivity. In the final model, individuals with male sex, who were more sociable, lived in more urban areas, had a lower SES, whose parents were less involved, and who were more likely to use tobacco daily portrayed blunted heart rate reactivity. In girls only, oral contraceptive use did not remain a significant predictor. In sum, heart rate reactivity was determined by urbanicity and SES in children, and sex, sociability, urbanicity, SES, parental involvement and tobacco use in adolescents.

Respiratory sinus arrhythmia reactivity.

In children, BMI, age, and SES significantly predicted RSA reactivity in the exploratory phase. In the final model, older age and lower SES were related to less pronounced RSA reactivity (i.e. RSA decreased less in response to the tasks). In adolescents, oral contraceptive use, activity, sociability, and cola use significantly predicted RSA reactivity. In the final model, individuals who were more active portrayed less pronounced RSA reactivity. Oral contraceptive use did not remain significant in the final model in girls only. In sum, RSA reactivity was determined by age and SES in children, and activity level in adolescents.

Perceived physiological stress reactivity.

In children, shyness, activity, sociability, age, family situation, and parental rejection significantly predicted PPS reactivity initially. In the final model, a lower level of shyness, younger age and lower perceived parental rejection were related to lower PPS reactivity. In adolescents, parenting styles overprotection and inconsistent discipline and tobacco use were significantly related to PPS reactivity in the exploratory phase. In the final model, lower perceived parental overprotection, higher parent-reported inconsistent discipline, and a higher likelihood of daily tobacco use were related to lower PPS reactivity. In sum, PPS reactivity was determined by shyness, age and perceived parental rejection in children, and perceived parental overprotection, parental inconsistent discipline and tobacco use in adolescents.

Individual factors.

In our study we found that sex was related to stress reactivity, with girls showing higher heart rate and cortisol reactivity, which is consistent with earlier findings pertaining to heart rate, however sex differences in cortisol reactivity are generally not found in children [31], [32]. Oral contraceptive use and menstrual cycle phase were unrelated to stress reactivity. These factors were shown to be of influence in adults (i.e. [35], [37]), but not in adolescents [34], perhaps because the menstrual cycle of girls is less stable, or a different ratio of sex hormones contribute differently to cortisol reactivity [34]. PPS reactivity was found in our study not to vary according to sex which is consistent with previous studies [33], [34].

In line with theoretical propositions [46], [47], [48], we found effects of activity level, emotionality, shyness and sociability on cortisol, heart rate, RSA and PPS reactivity. Specifically, in children, decreased shyness was related to lower PPS reactivity. In adolescents, a higher sociability score was related to lower cortisol and heart rate reactivity. A higher emotionality score was also related to lower cortisol reactivity, and a higher activity level was related to lower RSA reactivity. Previous research would suggest a strong relation between temperament measures and physiological stress reactivity [88], which we found, interestingly, for adolescents but not children, as only shyness was related to PPS reactivity in children. Possibly, in children, the effects of temperament on stress responses were outweighed by other determinants included in the models. This underlines the importance of examining multiple determinants of physiological reactivity in a single study. Previous studies in young children and adults have also reported significant effects of temperament on physiological stress reactivity [89], [90]. It is unclear why these relations should be of greater importance in adolescence as opposed to late childhood, and this warrants further research. The other individual factors that were examined in this study (i.e. ethnicity, BMI and birth weight) were not related to any of the physiological or PPS responses.

Developmental factors.

Because of the large sample in the present study, we were able to examine determinants of stress reactivity in children and adolescents separately. The physiological stress systems are proposed to undergo developmental changes between childhood and adolescence [16], and indeed, our data showed distinct determinants influencing stress reactivity in childhood versus adolescence. Only the influence of urbanicity and SES on heart rate reactivity was uniform in both samples. In general, pertaining to physiological stress alone, the physiological reactivity indices (heart rate with cortisol and RSA) were less strongly correlated in adolescents as compared to children. Also, the manipulation check showed that the stress response patterns were as expected in children, but not in adolescents. Pertaining to the determinants, we observed that temperament seemed to be a more important influence in adolescents than in children. Age remained a significant factor in children, but not in adolescents. We were particularly interested in pubertal development, as some previous studies suggested this might be important (e.g. [33]), but this did not remain a significant factor in adolescents our study. In children, we examined menstruation in girls as a proxy for pubertal development, but this was also not related to stress reactivity. We observed significant effects of parenting in both children and adolescents, albeit different parenting styles. In sum, the differential response patterns and differential determinants of stress reactivity in children and adolescents found in our study confirm previous propositions of developmental changes in physiological stress reactivity in humans during adolescence [16], [51].

Environmental factors.

Environmental determinants of the stress response have been neither frequently examined nor controlled for in past studies. This study showed that such factors may, however, be crucial variables, accounting for significant variability in all stress responses. Initial hypotheses in the literature proposed that (early life) adversities are associated with physiological hyper-reactivity to stress (e.g. [55]). However, there also seems to be evidence in animal and human research that early environmental adversity is linked to subsequent blunted physiological stress responses (i.e. [91], [92]). The findings from this study generally support this latter hypothesis, as we found that a lower SES and living in a more urban area were related to blunted heart rate reactivity in children and adolescents, blunted RSA reactivity in children and blunted cortisol reactivity in adolescents, confirming earlier findings [40]. As lower SES environments and city-dwelling are viewed as a more demanding and socially stressful, these could be considered more adverse environments to grow up in, thus blunting stress responses. Alternatively, given that physiological stress systems are partially genetically determined, it is possible that (the parents of) individuals who are inherently hypo-aroused, move to, for example, more populated areas or attain a lower SES.

Parenting styles index the family environment of youth and in this way may influence the developing stress systems. Protective parenting factors were related to increased cortisol, heart rate and PPS reactivity in the present study. This is consistent with earlier findings of less optimal parenting behaviors being related to HPA hypo-reactivity [62]. In our study, less emotional warmth was related to blunted cortisol reactivity in children, less involvement was related to blunted cortisol and heart rate reactivity, and less overprotection was related to blunted PPS reactivity in adolescents. Comparably, a higher degree of inconsistent discipline was related to blunted PPS reactivity in adolescents. The less optimal perceived parental rejection was positively related to PPS reactivity in children, which does not fit with this pattern. However, this pertains to perceived parental rejection, and likewise, perceived physiological stress. That this relation is positive, may indicate that some children are more sensitive to external influences, and for this reason reported a higher degree of both parental rejection and PPS during the psychosocial stress procedure.

(Early) adversity may also be indexed by number of adverse life events. Previous studies found that having experienced more adverse life events was related to blunted cortisol and heart rate reactivity in boys [42], [64] and young adults [65], [66]. In contrast to these findings, we observed no relation between adverse life events and physiological stress responses in adolescents. Internal reliability was, however, quite low in the adolescent sample, which could account for these findings.

Substance use-related factors.

In our study, in the adolescent sample, we found tobacco use to be associated with blunted cortisol and PPS reactivity. The former is consistent with previous research [67], [93]. Previous studies also found alcohol and cannabis use to be related to blunted cortisol and heart rate reactivity [67], [94], though these relations were not evident in the present study. An effect of tobacco use on physiological reactivity was expected, yet the finding of its effect on PPS seems less intuitive, and has not been previously examined to our knowledge. It is not entirely clear why we observed this relation in the present study, and further research will be needed to confirm or disprove this finding.

The results of this study should be considered in light of some limitations. First, the study was cross-sectional, therefore no conclusions can be drawn as to the direction of influence of the determinants and the stress reactivity variables. Second, information on pubertal stage was only available for the adolescent sample. We were able to assess puberty-related changes in girls in the child sample with menstruation, however, it would have been preferable to have information on pubertal stage in children as well. Developmental differences within the child and adolescent samples are likely to have still been influential, as the age ranges within these samples were still quite large. In order to fully examine developmental differences in relation to stress reactivity, it is necessary to have a sufficiently large sample to divide it into groups of smaller age ranges. Third, SES was operationalized based on the higher occupational level of either parent and coded into only three levels. Furthermore, only a small percentage of participants had a low SES background. A more comprehensive definition of SES, including financial, educational as well as occupational information is necessary to fully examine the relation of SES with stress response indices. Fourth, though our PPS questionnaire has been used in previous studies, it has not yet been formally validated. Fifth, our measures of stress reactivity were based on difference scores. Such measures are still widely used and clearly superior to single measurements, however, more comprehensive measures that utilize all of the measurement points, such as area under the curve estimates or growth curve modeling, are more sophisticated than difference scores [95] and preferable in stress reactivity research.

Current research has focused increasingly on physiological stress reactivity as a vulnerability marker for (mental) health problems in children and adolescents. However, such studies are inconsistent in the inclusion of potential covariates that may influence the developing stress systems. Determinant factors of stress reactivity were outlined in several theories (e.g. [5], [6], [7]), and although several reviews of such determinants are available (i.e. [8], [9], [10], [11]), our sample of children and adolescents from the general population provided an exceptional opportunity to systematically test such factors, also as reports in the literature pertaining to many of these factors are either lacking or inconclusive. In children and adolescents, individual, developmental, environmental and substance use-related factors influenced stress reactivity measures. Furthermore, we investigated the hypothesis that PPS reactivity was positively related to cortisol, heart rate and RSA reactivity, but found that, for the most part, this did not hold in our samples, in contrast to a recent study in adolescents [14]. In sum, our study showed that it is imperative that future studies take into consideration determinants of stress reactivity that may account for found relations. This study provides an indication of which determinants should be considered in children and adolescents.