Hypothalamic–pituitary–adrenal axis activity and upper respiratory tract infection in young children transitioning to primary school
- First Online:
- Cite this article as:
- Turner-Cobb, J.M., Rixon, L. & Jessop, D.S. Psychopharmacology (2011) 214: 309. doi:10.1007/s00213-010-1965-x
- 177 Views
We have previously reported an increase in salivary cortisol in a cohort of 4-year-old children transitioning to primary school. We hypothesised that increased cortisol in response to this acute naturalistic stress in early development may be immunostimulatory and associated with positive health outcomes.
We tested this hypothesis by measuring upper respiratory tract infection (URI) across the first 6 months of school, in relation to salivary cortisol at the end of the second week following school transition
Seventy children supplied morning and evening saliva samples for cortisol assay. Children were psychologically assessed for temperament and behavioural adaptation. Symptoms of URI were recorded in diary form, and variables relating to URI occurrence, duration and severity were assessed.
Children with higher evening cortisol at school transition experienced significantly fewer episodes of URI over the following 6 months. Diurnal cortisol change was negatively correlated with number of illnesses across the 6 months, indicating an association between a greater decline in cortisol across the day and a greater number of colds. URI severity was associated with the greatest resistance to URI infection in children who were less socially isolated and who had a smaller diurnal change in cortisol across the day.
Our results showing that higher cortisol is associated with lower URI may be explained by proposing that increased cortisol in response to the naturalistic stress of school transition may prime the immune system to develop resistance to URI at this critical stage of a child’s development.
The science of psychoneuroimmunology is a potentially powerful tool for elucidating the complex mechanisms which underlie causal associations between stress and illness. Stress is now recognised as a significant contributory factor in both the aetiology and pathophysiology of many diseases (McEwen and Stellar 1993; Chrousos 2000; Straub et al. 2005). The hypothalamic–pituitary–adrenal (HPA) axis, one of the principal pathways which respond to stress, exerts tonic inhibitory control over the immune system. It is well-known that chronic stress and associated elevated cortisol secretion are immunosuppressive, resulting in a resetting of immune system parameters with consequent impairment in the ability of the organism to respond to infection and illness (Glaser and Kiecolt-Glaser 2005; Besedovsky and del Rey 2002). However, immunostimulatory effects of transient increases in cortisol following acute stress are also well documented (McEwen et al. 1997; Dhabhar and McEwen 1997; McEwen 1998; Dhabhar 2009) and may be beneficial in combating disease (Dhabhar et al. 2010) and recovery following surgery (Rosenberger et al. 2009). Relationships between the type of stress, timing of the stressor and selective responses of the immune system and associated health outcomes are poorly understood.
The effects of psychosocial variables and, in particular, stress on health in humans have been widely examined across a number of different acute and chronic health conditions (Lutgendorf and Costanzo 2003). It seems clear from the literature that there is a general positive correlation between levels of perceived stress and the onset and progression of illness. The study of acute upper respiratory infections (URIs) has frequently been used as a model of acute illness in exploring these relationships due to their ease of reporting and relatively brief time course. Both experimental and naturalistic studies of the common cold and flu have linked higher levels of stress with increased susceptibility to acute infection (for a recent review, see Falagas et al. 2010). Research in this area has been much more abundant in relation to adults than to children, probably due to measurement difficulties in young children, not least the complexity of measuring the concept of psychological stress, often difficult for adults to articulate but all the more problematic in studies of children (Jessop and Turner-Cobb 2008). A particularly important period in which to study the interactions between stress and health is during the critical early years of life (Coe and Lubach 2003). However, comparatively little research has examined the link between early life stress and URIs (Turner-Cobb and Steptoe 1998; Ball et al. 2002; Turner-Cobb 2005).
Examination of a novel yet naturalistic life stressor represents an ideal way in which to measure the stress response and psychological predictors of the magnitude of the response in children. One such example is the novel event of starting school for the first time, an unavoidable experience for the majority of children in the western world, and it can be viewed as an opportunity to relate a defined stressful event with acute health outcomes. Previous research has demonstrated that the life event of starting school is a naturalistic social stressor which increases HPA axis activity and is associated with altered immune function (Boyce et al. 1995). We have previously reported that in a cohort of 4-year-old children transitioning to primary school, salivary cortisol was increased in the second week of school (Turner-Cobb et al. 2008). However, there is very little research which examines the potential stressor of school entry on subsequent health outcome in children. One study evaluated susceptibility to the common cold across a longitudinal time period of the first 13 years of life (Ball et al. 2002). Children who attended large-scale day care centres, although observed to have a higher incidence of the common cold at age 2 years, revealed a subsequent protection from the common cold, reporting fewer incidences at ages 6, 8 and 11 years (Ball et al. 2002). However, this study did not assess URI incidence from the perspective of the novel stressor of transitioning to school and did not assess in conjunction with the social experience and temperament of the children or with mediating hormonal responses to such a situation.
The aim of the present research was to examine the link between a naturalistic stress response and immune resistance, using the outcome measure of URI. We hypothesised that greater cortisol reactivity to stress at school entry would be associated with decreased onset, duration and severity of URI over the proceeding 6 months. In addition, we expected subscales of temperament and adaptive behaviour to be associated with measures of URI onset, duration and severity. In addition, in light of our own previous findings (Turner-Cobb et al. 2008), we hypothesised that temperament and behavioural adaptation, specifically internalising social isolation behaviour, would moderate the effect of stress on URI incidence.
Participants and design
Participants comprised 70 children (32 female, 38 male) and one of their parents, from a larger cohort of 105 children recruited for the Transition to School Study, a longitudinal study of the experience of starting school, funded by the Economic and Social Research Council (Turner-Cobb et al. 2008). The 70 participants reported here constitute those remaining in the study at the second assessment point, 4–6 months after baseline sampling, and were followed up over 6 months. All children had attained their fourth birthday prior to the start of school and had entered their first year of formal schooling 2 weeks before the initial data reported here. Seventy-six teachers who agreed to provide a self-report questionnaire regarding their pupils’ behaviour were also recruited.
Ethical approval for the entire study was granted externally by the NHS Local Research Ethics Committee and internally by the Department of Psychology research ethics committee at the University of Bath. Written informed consent was obtained from parents whose children were taking part in the study, and verbal consent was obtained from the children themselves. Permission was sought from the head teacher of each school before approaching the class teachers to obtain written consent.
Demographic, medical and preschool information
Detailed information regarding parent and child demographic, medical and preschool details for this cohort are reported elsewhere (Turner-Cobb et al. 2008). The occurrence of major and minor stressful life events in the 6 months prior to school transition was also recorded.
This was measured using the Child Adaptive Behaviour Inventory, a scale adapted by Cowan and Cowan (1990), designed for use by parents or teachers to describe a child. In this study, it was completed by teachers to assess academic, social and behavioural competence in the classroom. The scale consists of 106 items with a four-point response format ranging from 1 = not at all to 4 = very much. There are 22 lower-order factors which are further divided into six higher-order factors of (1) academic competence (intelligence, creativity and task orientation), (2) social competence (kind, fair, socially rejected, socially skilled and socially perceptive), (3) externalising aggressiveness (anti-social, hostile and oppositional), (4) externalising hyperactivity (distractible, hyperactive and calm), (5) internalising social isolation (socially isolated, introverted and extroverted) and (6) internalising psychological symptoms (depressed, imitative, somaticises and anxious; Johnson et al. 1999). Internal consistency has been reported to range from alpha = 0.66 to 0.90 with a mean of alpha = 0.81 for teachers’ ratings (Cowan et al. 1994). In the current sample, the Cronbach’s alphas for higher-order factors ranged from 0.54 to 0.87 at transition and from 0.48 to 0.91 at follow-up. The alpha scores for the higher-order factors of academic competence, social competence and internalising social isolation achieved alphas of at least 0.76.
The researcher explained to parents how to assist their child in collecting the saliva samples. Parents were asked that their children refrain from eating, drinking, brushing teeth or clearing their throat prior to sample collection at both sampling times. Although we did not use an electronic means of assessing wakeup, the sampling protocol was the same as that in our previous study of children of a similar age (Chryssanthopoulou et al. 2005) in which we achieved a high level of compliance. Several steps were taken to control for non-compliance with the awakening sample, including the use of a saliva sampling booklet given to all parents where they were able to document any difficulties with sampling such as any delay between the child waking and the sampling time. Furthermore, the researcher was in contact with the parents by phone and able to check accuracy of sampling. Parents were asked to avoid any interference from factors such as eating, sleeping or aerobic activity with respect to the early evening (5.30 p.m.) sample. Parents were asked to collect the sample at this time as it represented a period when most children would be comparatively restful. This was viewed as a time when they would have returned from school or any after-school activity but prior to eating dinner (or at least an hour after eating). Parents were instructed to take the sample at a time that met these criteria. Saliva was collected using plastic Salivette devices (Sarstedt, Germany). Parents were instructed to collect saliva samples over 2 days at the end of the second week of school. The 2-day protocol for each time point was followed as recommended in order to obtain reliability and validity of assessment (Clow et al. 2004). Saliva samples were taken on average just after 7 a.m. for the awakening sample (7.10 a.m. ± 70 min) and around 5.30 p.m. for the evening sample (5.40 p.m. ± 82 min). Saliva samples were stored at −20°C prior to analysis.
Upper respiratory infection
A daily checklist of symptoms was completed by participants in the form of a pencil and paper health diary. This consisted of a daily checklist of 14 cold and flu symptoms and questions regarding the severity of the illness and utilisation of healthcare resources. The minimum criterion of at least one symptom lasting for 48 h or at least two symptoms lasting for 24 h was used to determine the occurrence of a cold episode from the diary reports (for full details, see Turner-Cobb and Steptoe 1996, 1998). Health diaries were completed for each illness episode and at the end of 6 months regardless of the presence or absence of a URI. Each diary lasted for 1 month and was returned by post. A number of variables pertaining to onset, duration and severity were generated from the diaries using a standardised scoring procedure. Measures of URI onset included the total number of URIs over the 6 months of the study, whether or not the URI onset occurred during the first 2 weeks of term and whether or not it occurred during a school holiday period, compared to other times in the academic year. Duration of URIs was assessed by calculating the mean number of days for a URI episode over the 6-month period and the longest URI episode that occurred during that time. Severity of URIs was assessed by calculating the mean number of symptoms across episodes that occurred over the 6-month period.
This study utilised a prospective cohort design, and the data reported here start with the second assessment of this longitudinal study. Participants were provided with study materials (questionnaires, saliva assessment kits and URI symptom diaries) via in-person delivery from the research officer to participant homes. Questionnaires for teachers were mailed to their school address. Saliva samples and questionnaires were completed and collected at the end of the second week of school. Over the following 6 months, all URIs in the children were recorded. To encourage compliance with the protocol, parents were contacted by telephone at regular intervals, approximately monthly, as a study reminder, to assess whether a URI episode had occurred and to encourage return of health diaries immediately after an illness episode had occurred. To maintain compliance, in the absence of a URI, parents were still asked to return a health diary at the end of each month, and a new one was sent.
Cortisol values from samples collected on consecutive days were meaned, and the mean value was used for subsequent analysis. Cortisol values greater than 40 ng/ml were considered outliers and were excluded from analysis. On this basis, two children were excluded: one child with consistently high morning values over both days and one child with high evening values on both days as well. The remaining cortisol values had a positively skewed distribution with high kurtosis which required logarithmic transformations (log10) prior to subsequent analysis.
Four measurements of cortisol were computed from the morning and evening samples: (1) awakening cortisol was intended to reflect cortisol production immediately upon awakening prior to breakfast or brushing teeth. To ensure that the time of sampling was not related to the cortisol value, the difference between time of waking and time of measure was correlated with the morning cortisol; (2) the evening measure was taken at a time when children would have returned home from school and be adapting from the school to home environment; (3) diurnal change was computed by subtracting awakening from evening values (thus, more negative values reflect a steeper decline from morning to evening) as applied in previous adult and child studies (Chryssanthopoulou et al. 2005; Sephton et al. 2000; Watamura et al. 2003) and (4) mean cortisol was measured averaging the morning and evening values for each time point.
Cleaning of URI variables involved the exclusion of outliers (z scores less than −3 and or greater than +3 standard deviations), resulting in two outliers based on URI duration. Logarithmic transformations were applied to normalise URI distributions as necessary. Bivariate correlation (Pearson’s r and Spearman’s rho) analyses were used to assess relationships between demographic, preschool and life event data with the dependent URI variables. As significant relationships did not emerge with URI variables, they were not subsequently employed as control factors in analysis. Frequency analyses of URI occurrence across the 6 months of the study were explored using Cochran’s Q test and McNemar test for post hoc comparisons. To examine the relationships between salivary cortisol reactivity at school transition, the role of psychological variables and the effect on URI occurrence, duration and severity, correlation analyses (Pearson and Spearman as appropriate) were used. Hierarchical regression analyses were employed to examine the interaction terms of the psychological factors (temperament and adaptation) with cortisol reactivity at time 2. Time 1 cortisol was used as a control variable on the first step in each hierarchical regression analysis, with independent predictors entered on the second step and the interaction term entered on the third step to test for moderation effects. Variables were centred prior to computing the interaction term in order to avoid multi-collinearity (Howell 2002).
Details of acute URI variables (n = 70)
Number of URIs over 6 months of study
Participants with URI onset during school holidays
Participants with URI during term time
Participants with URI onset during first 2 weeks of school term
Mean number of days URI episode lasted
Longest URI mean episode (days)
Mean number of symptoms across episodes
Largest mean number of URI symptoms for any one episode
Cortisol measurements in the morning and evening for all children were 9.63 ± 6.41 and 1.30 ± 1.52 ng/ml, respectively (mean ± SD) as reported in Turner-Cobb et al. (2008). Overall, children who had a greater evening cortisol reactivity at school transition were found to experience significantly fewer episodes of URI over the following 6 months (rs = −0.318; p = 0.004). Diurnal cortisol change was negatively correlated with number of illnesses across the 6 months (rs = −0.231; p = 0.043), indicating an association between a greater decline in cortisol across the day and a greater number of colds. There was no significant difference in URI occurrence for morning cortisol.
Evening cortisol at transition was significantly associated with URI onset during school holidays (rs = −0.330, p = 0.009) compared to the lower correlation of URI onset during the school term (rs = −0.299, p = 0.019). Similarly, greater diurnal cortisol change at transition was associated with higher URI onset during school holidays (rs = −0.273, df = 59, p = 0.033) compared to a non-significant effect for term time. Cortisol variables were not significantly associated with mean duration or severity of URIs over the study period as main effects.
Correlations between temperament/adaptive behaviour and URIs
Correlational analyses examining the relationship between higher-order subscales of temperament and onset, duration and severity of URIs found no significant relationships. For adaptive behaviour, significant effects were found for four of the higher-order subscales: academic competence, social competence, externalising hyperactivity and internalising social isolation. Academic competence was negatively associated with cold onset during the school holidays (r = −0.348, p = 0.012), social competence was negatively associated with number of URIs over the 6 months (r = −0.394, p = 0.004), externalising hyperactivity was negatively associated with mean number of symptoms (r = −0.289, p = 0.04) and internalising social isolation was positively associated with having a URI in the first 2 weeks of starting school (r = 0.317, p = 0.023).
These results demonstrate that children with higher evening cortisol on school entry had fewer episodes of URI over the following 6 months. No significant correlation was observed between awakening morning cortisol and onset of URI. Those children with a steeper diurnal cortisol slope throughout the day also experienced significantly more episodes of URI. A number of studies have reported a positive correlation between stress and increased susceptibility to the common cold and URI (reviewed in Falagas et al. 2010). However, we believe that this is the first study to demonstrate a positive correlation between the naturalistic developmental stress of school transition and immune resistance, as measured by decreased incidence of URI. We acknowledge that the child’s URIs were reported by the parents and that not having symptom verification or viral identification is a possible limitation of the study. However, parents were given clear instructions for the use of the symptom diary, and strict criteria were applied to the reported symptoms to determine a cold episode.
An explanation for this positive correlation between increased cortisol and lower incidence of URI may lie in the well-documented immunostimulatory effects of low doses of glucocorticoids in vitro (Wiegers et al. 1994) and in transient increases in glucocorticoids following acute stress in vivo (McEwen et al. 1997; Dhabhar and McEwen 1997; McEwen 1998; Dhabhar 2009). It has been recently reported that healthy adults infused with a low dose of cortisol for 6 h exhibited an enhanced IL-6 response to stress compared to the group infused with the higher dose (Yeager et al. 2009). The elevated levels of evening cortisol in the children in our study may be sufficient to prime the immune system to elicit a protective effect on URI. Furthermore, recent evidence has been found to support the notion that an acute stressor can increase resistance to tumour development in mice (Dhabhar et al. 2010). Our data are consistent with a protective effect of acute stress on health in humans. Long-term effects of an acute stressor on HPA axis activity and immune function are well documented (Tilders et al. 1999; Valles et al. 2003; Armario et al. 2004; Harbuz et al. 2002; Richards et al. 2006). Timing of glucocorticoid administration is also important in mediating the effects of lipopolysaccharide, a bacterial compound released during infection which induces an acute stress response (Barber et al. 1993). It is possible that the stress of transition exerts complex underlying effects on hormone and immune interactions over time. It will be very important to elucidate the mechanisms and timing which underlie the apparent positive effects of acute stress on health outcomes in children as observed in our study.
We are aware that the difficulties of defining stress in very young children are highly problematic in terms of drawing causal relationships with health profiles. A meta-analysis of the literature on the relationship between psychosocial stressors and acute respiratory tract infections has been recently published (Falagas et al. 2010). From these largely adult population studies, the authors concluded that individuals with higher levels of self-reported ‘negative’ stress have higher incidence of URI compared to those exhibiting ‘positive’ traits of extraversion, assertiveness, a better sense of well-being, optimism and a more positive outlook on life who are less susceptible to URI. The majority of these studies utilised subjective self-evaluation as a measure of stress rather than cortisol. Very few studies of this nature on young children of preschool age have been published, and it is notable that much of the literature associating stress with increased URI has focussed on studies on adults and older children who may be more capable of self-reporting perceived stress. Subjective evaluation of stress is very difficult to obtain in young children, particularly of preschool age, due to the limitations in vocabulary required to comprehend the questions and articulate responses. Consequently, stress in young children may more reliably be determined objectively by measurements of HPA axis activity through salivary cortisol (Jessop and Turner-Cobb 2008). This provides a measurement of physiological stress from which psychological stress is often assumed. However, we have insufficient data to conclude that cortisol is a reliable marker of perceived subjective stress in children. Moreover, it has recently been suggested that the relationship between physiological and psychological stress is far from clear (Marshall 2009). It may be that the subjective perception of stress as a negative concept and the physiological response as expressed by HPA axis activity may exert quite different influences on the immune system and predisposition to illness. We have previously reported an increase in salivary cortisol in young children prior to attending school for the first time (Turner-Cobb et al. 2008), which we interpreted as an anticipatory effect of stress having a positive effect on behaviour in the novel environment of school. In other words, it is possible that elevated cortisol in the children may be a consequence of novelty or excitement at the new social challenge of school transition, and this ‘positive’ stress may have its own influence on the immune system and incidence of URI. Indeed, school transition was not conceptualised as a risk factor, merely a naturalistic opportunity to observe early physiological responses to challenge. How the responses to this challenge are translated into subsequent immune adaptation and health outcomes in children are important questions for future research.
It is worthy of comment that we observed a positive correlation between extroversion/surgency and increased cortisol in this cohort of children on their first experience of starting school (Turner-Cobb et al. 2008), consistent with that previously reported in children at the start of a new school year (Davis et al. 1999). The relationship between extroversion, HPA axis activity and health outcomes may have important implications for the social hierarchy and leadership roles which children establish at this critical phase of personal development.
It is of interest that, while we observed that elevated evening levels of cortisol and a smaller change in cortisol across the day were associated with fewer URIs over the whole of the 6-month period, this effect was reversed during the school holidays. Therefore, resistance to infection was the poorest during holiday time. One speculative explanation for this unexpected direction of effects is that it is possible that holiday periods may have been confounded with greater stress due to family difficulties or travel or other disruption to routine, which may have raised cortisol levels during these times to a point of becoming immunosuppressant. We did not measure cortisol during holiday time since this was not part of the original study design. More frequent cortisol measurements in future studies would address the issue as to whether changes in cortisol are associated with the onset of URI during holiday periods.
There is a well-documented literature on the association between negative life events as stressors and increased susceptibility to infection in adults (Cohen et al. 1993, 2002; Hamrick et al. 2002; Evans and Edgerton 1991) and adolescents (Stone et al. 1992; Haavet et al. 2004; Lien et al. 2007). We could find no significant association between stressful life events in our children prior to school transition and URI. This suggests that the predictive effect of cortisol on URI is associated with the acute stress of starting school rather than more long-term chronic life event stress occurring prior to school. However, given that only about 25% of the children in our study reported a life event in the 6 months prior to starting school, we cannot rule out the possibility that ongoing events may have exerted an effect. These effects could be examined in future research with a more substantial longitudinal design and a larger cohort with greater socioeconomic diversity.
We are aware that we are basing our observations on a single time point of cortisol measurement. Although this is a valid method for determining an acute response to the naturalistic stress of transition, it can provide little information about the ongoing interrelationship between the HPA axis, immune system and health outcomes for the first year of school. It is possible that, in contrast to the acute response to school transition, the HPA axis in some children may be chronically activated for some time after starting school, which may exert a quite different effect on immune function and health outcome. To determine the relationship between chronic HPA axis activity and URI, further studies measuring cortisol at more frequent time points following transition to school are required. It may also be instructive in this context to investigate a range of steroids which respond to stress, such as dehydroepiandrosterone which has immunostimulatory effects (Hazeldine et al. 2010) or by using additional methods of cortisol assessment, such as a retrospective assessment of cortisol measured in hair samples (Gow et al. 2010), to assess more enduring chronic stress states.
With respect to the influence of temperament, we found no significant effects for the higher-order factors. It is possible that these factors as measured were not sufficiently sensitive to detect an effect with the respiratory symptoms. Similarly, for the adaptive behaviours, only a few effects emerged with URI symptoms. Given the number of URI-dependent variables, we are cautious about over-interpreting these effects in respect to type 1 error. However, the effect observed for internalising social isolation moderating diurnal change in cortisol is worth noting, particularly given the effects previously reported for this factor (Turner-Cobb et al. 2008). Here, we found the greatest resistance to URI infection to be in children who are less socially isolated and who have a smaller diurnal change in cortisol across the day. This adds to the main effects reported for cortisol and URI variables as social isolation and may indicate a smaller exposure to URI infection through reduced social contact but increased resistance only taking effect when the HPA axis diurnal variability is small. This interaction of psychological and physiological adaptation offers potential for future research.
In conclusion, we have observed that higher evening cortisol levels in children in transition to school are associated with relative immune resistance as measured by fewer URI. There is a clear and largely unmet need for further studies into childhood responses to naturalistic stressors and the impact of early life critical periods on immune function and health. Our data perhaps may stimulate a re-evaluation of the association between naturalistic stressors of novel environments such as school entry and consequent immune adaptation with implications for health outcomes.
This research was funded by a grant from the Economic and Social Research Council (ESRC), UK (#RES-000-23-0141) awarded to the principal investigator, Julie M. Turner-Cobb. The authors would like to express their gratitude to the children, their parents, teachers and schools who took part in this study. The authors have no conflicts of interest with any aspect of this study or its funders.