ARTICLE IN PRESS Psychoneuroendocrinology (2006) 31, 1009–1018
www.elsevier.com/locate/psyneuen
Suspected non-adherence and weekend versus week day differences in the awakening cortisol response L. Thorna, F. Hucklebridgeb, P. Evansa, A. Clowa, a
Department of Psychology, University of Westminster, 309 Regent Street, London W1B 2UW, UK Department of Human and Health Sciences, University of Westminster, 115 New Cavendish Street, London W1M 8JS, UK
b
Received 16 December 2005; received in revised form 30 May 2006; accepted 30 May 2006
KEYWORDS Human; Saliva; Awakening cortisol response; Adherence; Weekend/weekday differences; Awakening time; Stress
Summary The pronounced rise in cortisol following awakening holds promise as a bio-marker of variables in the psychosocial domain, but its investigation also presents methodological challenges, which we attempted to address in this study. Forty-eight healthy, young students participated and were instructed to collect saliva 0, 15, 30 and 45 min post awakening on two consecutive normal weekdays and two consecutive weekend days (order counterbalanced). Participants’ awakening cortisol response (ACR) profiles differed between the weekdays and weekend with trend analysis revealing a steeper rise on weekdays compared to the weekend. These differences were not accounted for by weekday/weekend differences in awakening time and state stress or by perceived stress over the previous month. Total salivaryfree cortisol secretion (area under the cortisol curve (AUC)) over the 4 study days was negatively correlated with the measure of longer term stress and awakening time. The mean 4-day rise in cortisol (mean increase: MnInc) was also negatively correlated with awakening time. This awakening time effect was not mediated by stress or vice versa, since both were independent predictors of cortisol. In an attempt to address the ubiquitous problem of non-adherence to the requested saliva sampling regime, known to distort the shape of the ACR, suspected non-adherence (SNA) was examined by identifying instances of profiles showing no cortisol rise from the waking sample to either the 15 or 30 min sample post awakening. Analysis controlling for SNA status had no effect upon the observed associations with Perceived Stress Scale (PSS) and awakening time however it abolished the otherwise highly significant flatter profile at weekends. & 2006 Elsevier Ltd. All rights reserved.
Corresponding author. Tel.: +44 20 7911 5000x2174; fax: +44 20 7911 5106.
E-mail address:
[email protected] (A. Clow). 0306-4530/$ - see front matter & 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.psyneuen.2006.05.012
ARTICLE IN PRESS 1010
1. Introduction There is accumulating evidence that the awakening cortisol response (ACR) is a discrete part of the cortisol circadian cycle (Schmidt-Reinwald et al., 1999; Wuest et al., 2000a; Edwards et al., 2001a; Wolf et al., 2005). In healthy adults salivary-free cortisol concentrations increase by between 50% and 160% in the first 30 min immediately post awakening (approximate average increase of 9 nmol/l, range 4–15 nmol/l, see Clow et al., 2004). The ACR, which has been used as an umbrella term to describe both overall salivaryfree cortisol secretory activity as well as measures of dynamic change following awakening, has been shown to have moderate intra-individual stability as demonstrated by correlational analysis (Wuest et al., 2000b; Edwards et al., 2001a). However the complexity of the ACR is noteworthy as it is affected by some situational factors. For example the ACR is enhanced by light exposure (Scheer and Buijs, 1999; Thorn et al., 2004) and according to some reports, earlier awakening time (Edwards et al., 2001b; Kudielka and Kirschbaum, 2003; Federenko et al., 2004). In addition it has been reported that the ACR differs according to day of collection, being attenuated at the weekend compared to weekday (Kunz-Ebrecht et al., 2004; Schlotz et al., 2004). Each of these points deserves careful consideration as an understanding of their mechanism of action will help elucidate the nature of the ACR, as well as reduce possible methodological confounds, which may be responsible for the lack of concurrence in studies examining associations between the ACR and psychosocial variables. For example some studies have shown elevated ACRs to be associated with chronic stress (Schulz et al., 1998; Wuest et al., 2000a) whereas others have found the ACR to be attenuated in those experiencing chronic economic stress (Ranjit et al., 2005). There have been two reports comparing withinsubject differences in the ACR at the weekend relative to weekdays (Kunz-Ebrecht et al., 2004; Schlotz et al., 2004). Both of these studies have demonstrated that the ACR is attenuated at the weekend compared to working weekdays. KunzEbrecht and colleagues also report that 73% of their participants showed a ‘response’ (defined as rising more than 2.49 nmol/l from 0 to 30 min post awakening) on weekdays compared to 48% at the weekend. The differences between the ACR in this study could not however be attributed to self-report stress despite the participants reporting less stress and greater happiness at weekends. Schlotz and colleagues however were able to demonstrate clear
L. Thorn et al. associations between the attenuated weekend ACR and less work overload and fewer worries. These studies clearly indicate that the issue of sampling day is an important consideration in studies of this kind. However, a third study reported no difference in the ACR in women on work or non-work days either during pregnancy, or 9 months post-partum (de Weerth and Buitelaar, 2005) and in a further study there was no difference in ACR between a work day and a day off for victims of bullying at work (Kudielka and Kern, 2004). Most studies examining the ACR are carried out within the domestic setting where the participants are required to follow careful instructions. This type of procedure relies upon participant adherence to protocol in order to obtain accurate data. Use of an electronic monitoring device as an objective measure of adherence to the temporal sequence of daytime saliva sampling has shown timing non-adherence to be in the order of 26–29% for participants who were unaware that they were being monitored (Kudielka et al., 2003; Broderick et al., 2004). Both these studies reported that selfreport adherence cannot be relied upon. However, electronic monitoring takes into account the temporal sequence of sampling but not the exact timing of the ‘waking’ sample in relation to waking time. Other studies have used a self-report measure to exclude participants as non-adherent on the basis of a self-report delay of more than 10 min between awakening and first saliva sample. This strategy identified between 7% and 70% nonadherence (Kunz-Ebrecht et al., 2004; Wright and Steptoe, 2005). Regardless of the strategy used to control for non-adherence studies have demonstrated that non-adherent participants presented with flat ACRs (Kudielka et al., 2003; Broderick et al., 2004; Kunz-Ebrecht et al., 2004; Wright and Steptoe, 2005). Kupper et al. (2005) utilised simultaneous electrocardiogram (ECG) and movement readings to explore actual times of awakening in relation to the shape of the ACR. They found that participants who failed to show any rise in cortisol between their purported waking sample and 30 min later had in fact woken up an average of 42 min (range 10 min–2 h 15 min) earlier than the first sample. By contrast 87% of participants that showed a cortisol response to awakening collected their first saliva sample at the time of waking. Adherence to protocol is clearly a substantial issue and it may be that some discrepancies in the ACR literature can be attributed to inclusion of data derived from participants that did not fully adhere to the experimental protocol. It was the aim of this study to examine the ACR of healthy young adults on 4 days (including 2
ARTICLE IN PRESS Suspected non-adherence and weekend/week day differences in the ACR weekdays and the weekend) and it was hypothesised that there would be weekday/weekend differences. Although such differences have been reported previously, they have not always been found and only one study was able to explain the observed effects. The present study set out to explore whether any such differences were due to one or more of the following mechanisms: (1) changes in state stress at the weekend; (2) later awakening times at the weekends compared to the weekdays; (3) between-subject differences in longer term stress; (4) failure to comply with the sampling protocol on the weekdays (or at the weekend); or (5) a combination of factors. It was also hypothesised that the ACR of healthy young adults would be negatively associated with awakening time, in line with previous findings. As there is no consensus in the literature regarding the relationship between the ACR and measures of stress, the direction of this relationship was not predicted.
2. Methods 2.1. Participants Fifty healthy psychology students were recruited, on the basis that they were not taking medication and had no acute or chronic illness, to participate in the experiment, which took place in England during April and May. The samples of two participants were of insufficient salivary volume to perform assays therefore cortisol data were obtained for 48 participants, eight males, 40 females, mean age(7SD) 20.5(73.9) years, range 18–36 years. Participants received no financial incentive to take part. The ethics committee of the University of Westminster approved the protocol for this study. All participants provided written informed consent.
1011
a questionnaire which included demographic questions and the PSS, and a saliva sampling kit, consisting of four re-sealable plastic bags labelled DAY 1–4 each containing a record sheet, a SACL and four numbered Salivettes (saliva sampling devices, Sarstedt Ltd., Leicester, England).
2.3. Procedure Participants attended a detailed briefing session where they received full verbal and written instructions regarding the procedures of the study. This session included a practice session in the selfcollection of saliva. At the end of this session packs were given out containing instructions, questionnaires and saliva sampling materials. Participants were asked to complete the demographic questions and the PSS at any time during the study period. A repeated measures study was performed on two consecutive normal weekdays and two consecutive weekend days, weekdays versus weekend days were counterbalanced between participants. Participants were instructed to collect saliva immediately on awakening then at 15, 30 and 45 min post awakening on each study day. During the saliva collection period participants were instructed to take nil by mouth other than water, and not to smoke or brush their teeth so as to avoid microvascular leakage. Samples were placed in participants’ home freezer as soon as possible after collection of saliva and transferred to the laboratory in insulated cold packs to be stored at 20 1C until assay. Participants were asked to fill in a record sheet on each day recording awakening time and time of collection of saliva samples. Participants were also asked complete a SACL on each study day after taking the 45 min sample. Other than these instructions participants were asked to follow their normal routine.
2.4. Cortisol assay 2.2. Materials Perceived stress over the previous month was measured using the 14-item Perceived Stress Scale (PSS) (Cohen et al., 1983). Items were measured on a five-point scale and scores could range from 0 to 56, with a high score indicating high perceived stress. Cronbachs Alpha for this scale was 0.80. State stress was assessed using the Stress Arousal Checklist (SACL) (Mackay et al., 1978), a 19-item adjective checklist. Possible scores ranged from 0 to 19, with higher scores indicating greater stress. Participants were provided with a study pack containing full standardised written instructions,
Samples were thawed and centrifuged at 3500 rpm for 10 min. Cortisol concentration was determined by Enzyme Linked Immuno-Sorbent Assay developed by Salimetrics LLC (USA). Sensitivity: 0.19 nmol/l (lower limit). Standard range in assay: 0.19–49.0 nmol/l. Correlation of assay with serum: r ¼ 0:960, po 0.0001, n ¼ 19 samples. Intra and inter-assay variations were both below 10%.
2.5. Statistical analysis Cortisol concentrations in this study ranged between 1 and 40 nmol/l and were moderately
ARTICLE IN PRESS 1012
who showed greater SNA frequency at weekends compared to weekdays. In ANOVA analyses reduced degrees of freedom reflect Greenhouse–Geisser correction where the assumption of sphericity was violated. All given values of p are two-tailed.
3. Results 3.1. Weekday/weekend differences and relationships with psychosocial variables A within-subjects ANOVA on the cortisol data revealed no main effect of Period, indicating that there was no mean difference between weekdays and weekend days in total free salivary cortisol levels across the 45 min period. However there was a highly significant interaction between Period and Sample Time (F ð1:8;79:9Þ ¼ 8:652, p ¼ 0:001).Thus the dynamics of participants’ awakening response profiles differed between the weekdays and weekend (see Fig. 1). Trend analysis revealed a significant linear effect such that the weekday awakening response was steeper than that at the weekend (F ð1;45Þ ¼ 12:601, p ¼ 0:001). Subsidiary analysis of possible counterbalancing interactions showed that the observed differences between weekday and weekend responding were not a function of the order of collection (i.e. weekday or weekend first). An alternative analysis (using the composite measure MnInc as the dependant variable) confirmed 18
Cortisol nmol/l
positively skewed. A square root transformation was performed, which normalised sample distributions for inferential analysis purposes. Mean cortisol concentrations shown in figures and tables represent original units, i.e. the square of square root transformed data summaries. A three factor (2 2 4) within-subjects ANOVA was performed on the entire data set, with factors of Period (weekday and weekend), Day of respective period (first and second), and Sample Time (0, 15, 30, and 45 min post awakening), followed by a more detailed trend analysis (permitted by the significance of the omnibus interaction term). Two composite cortisol measures were also computed for each participant: the area under the cortisol curve (AUC) with reference to zero (sample 1+s2+s3+((s4s1)/2)) to give an estimate of overall salivary free cortisol in the first 45 min post awakening and the mean increase (MnInc) from the awakening sample to examine the dynamic aspect of the ACR ((sample 2+s3+s4)/3s1). Differences in state stress, awakening time and these composite cortisol measures between weekdays and weekend were analysed using t-tests; multiple regression analyses were used to predict degree of individual differences in cortisol responding as a function of any weekend/weekday effects. Fourday mean values were calculated for AUC, MnInc, awakening time and state stress. Correlational analyses were performed between these measures, as well as scores on the perceived stress scale and multiple regression analysis applied to predict variance in AUC cortisol. Additional exploration of the data was undertaken to investigate suspected non-adherence (SNA) to protocol. On any one day participants were categorised as being likely adherent or suspected non-adherent based on whether they showed the normal and expected pattern of increased cortisol following awakening, defined as an increase from sample 1 to either sample 2 or sample 3. Difference in SNA between weekdays and weekends was analysed using a sign-test; consistency of SNA across days was analysed by calculation of F-coefficients. The t-tests were performed to examine differences between awakening and final samples for likely adherent and SNA participants on each day. To explore possible effects of SNA, all previously significant effects using the total data set were re-examined on sub-sets of data excluding those categorised as SNA. Effect size estimates were calculated, based on median ‘r’ values, to examine the effect of SNA removal on the relationships between PSS scores, awakening time and composite cortisol measures. The initial ANOVA was repeated omitting those participants
L. Thorn et al.
14
10
Weekday Weekend 6
0
15 30 Minutes after awakening
45
Figure 1 Mean (7SEM) salivary free cortisol concentrations (nmol/l) after awakening on weekdays and at the weekend (N ¼ 48).
ARTICLE IN PRESS Suspected non-adherence and weekend/week day differences in the ACR that average mean increase in cortisol from awakening was significantly greater on weekdays than at the weekend (t ¼ 3.209, df ¼ 47, p ¼ 0.002). Also in line with the ANOVA result of no significant main effect for Period, there were no weekday/weekend differences in overall cortisol levels measured by AUC. Although participants woke up later at weekends (t ¼ 7.901, df ¼ 47, po0.0005) and tended to report less state-stress at weekends (t ¼ 1.964, df ¼ 47, p ¼ 0.056), neither of these variables, or scores on the Perceived Stress Scale, accounted for the observed differences in MnInc when multiple regression analysis was applied. Awakening times were highly inter-correlated across all 4 study days, indicating consistency in patterns of waking up earlier or later, legitimising the calculation of a single mean awakening time for each participant (calculated across all 4 study days; mean (7SD) 0902 h (787 min), range 0602–1137 h). Mean awakening time correlated inversely and highly significantly with overall cortisol levels as measured by the mean 4-day AUC (r ¼ 0.458, p ¼ 0.001) and also with average MnInc (r ¼ 0.314, p ¼ 0.036). Mean cortisol AUC also correlated inversely (r ¼ 0.382, p ¼ 0.011) with perceived stress (PSS scores: mean (7SD) 28.45 (77.08), ranging from 12 to 48) indicating that participants who on average across all study days secreted the most cortisol in the first 45 min following awakening were more likely to have reported less stress for the previous month. However, the ACR as measured by MnInc did not correlate with PSS scores (r ¼ 0.166, p ¼ 0.283). Since awakening time and PSS scores were both predictors of cortisol AUC in univariate analysis, it is important to report that both remained significant independent predictors of cortisol AUC in multiple regression analyses. There was no relationship between cortisol measures and mean state stress (as measured by the SACL, 45 min post awakening on all 4 study days).
1013
exclude the possibility that the reported effects revealed by the total data set could be attributed to non-adherent participants, who are known to present with distorted awakening cortisol profiles. Kupper et al. (2005) using ECG and motility data, reported that a flat or negative awakening response from 0 to 30 min post awakening is, in healthy people, most likely an artefact caused by sampling after the actual awakening response has occurred. In this study cortisol data was available for 15 min as well as 30 min post awakening. Therefore in order to maximise the residual sample of those most likely to have adhered to protocol, the procedure used to categorise participants as being likely adherent or suspected non-adherent was whether they showed the typical increase in cortisol from sample 1 (awakening sample) to either sample 2 or 3 (15 or 30 min post awakening, respectively). The increase did not have to reach any particular threshold, a strategy similar to that employed by Kupper et al. (2005) in removing suspect data. Frequencies of SNA (defined as neither sample 2 or 3 higher than sample 1) on each day are given in Table 1. A simple sign-test analysis revealed significantly more SNA at weekends (z ¼ 2.69; p ¼ 0.007). There was very little consistency across days in who was likely to be SNA and who was not (median degree of correlation ¼ +0.08 (based on F-coefficients for nominal data) and only one of the six possible interday correlations approached significance: F ¼ 0:27, p ¼ 0.07). As SNA was defined by a lack of dynamic (s2/3os1) it should not necessarily be associated with different absolute levels of cortisol at any particular point. However, the rationale for using s2/3os1 as a criterion of non-adherence assumed that the normal dynamic was lacking because the sampling period was shifted later than required by the protocol, as reported by Kupper et al. (2005), leading most likely to higher values for purported awakening samples (nearer peak than they should
3.2. Suspected non-adherence to protocol As participants were asked to record their awakening time and actual time of collection of saliva samples, a measure of self-reported adherence to the protocol was obtained. Only two participants reported that they had taken samples more than 10 min outside the times demanded by the protocol. However, there have been reports that adherence is a major issue in studies of this kind and that self-report adherence cannot be relied upon (Kudielka et al., 2003; Broderick et al., 2004). A strategy was therefore adopted in an attempt to
Table 1 Frequency of suspected non-adherence (SNA) by day. Number of suspected nonadherent participants per day Weekday 1 Weekday 2 Saturday Sunday
8 (17%) 6 (13%) 14 (29%) 16 (33%)
SNA was defined as when on any one sampling day neither samples 2 or 3 were higher than sample 1 (N ¼ 48).
ARTICLE IN PRESS 1014
L. Thorn et al.
Table 2 Mean salivary-free cortisol levels (nmol/l) and t-test comparisons between likely adherent and suspected non-adherent (SNA) participants for s1 (awakening sample), and s4 (45 min sample), (N ¼ 48). Awakening sample (s1)
Diff.
Adherent (nmol/l)
SNA (nmol/l)
Weekday 1 Weekday 2 Saturday Sunday
9.0 7.3 10.7 7.9
17.0 11.7 12.0 11.3
*
***
po.05; **po.01;
+45 min sample (s4)
ns * ns *
Diff.
Adherent (nmol/l)
SNA (nmol/l)
14.5 16.2 14.3 14.4
10.5 10.0 7.1 8.4
* * *** **
po.0001; ns ¼ not significant.
Table 3 Effect sizes (calculated as the median r-value of four correlations computed separately for each day) for the relationships between composite cortisol measures, PSS and awakening time for the total sample and when controlling for SNA participants.
PSS & AUC cortisol Awakening time and AUC cortisol Awakening time and MnInc cortisol
SNA removed
0.24 0.22 0.17
0.33 0.23 0.13
18
Cortisol nmol/l
be), and lower 45 min samples (more post-peak than they should be). Table 2 presents evidence supporting this. On all 4 days, SNA individuals have higher cortisol for the purported awakening sample and lower for the 45 min sample than do likely adherent individuals. The t-test analyses illustrate that six out of the eight SNA versus adherent comparisons were statistically significant, including all four comparisons at 45 min.
All data
14
10
3.3. SNA effects on PSS scores, awakening time and weekend/weekday differences In order to provide a check on the possible influence of SNA on the findings reported earlier for all participants, further analyses were performed excluding the SNA participants on each day. Table 3 illustrates effect size estimates (based on median ‘r’ values) for the relationships between PSS and cortisol AUC and between awakening time and both cortisol AUC and MnInc for the full-data set and excluding SNA participants on each day. This indicates that the effect sizes, previously reported as statistically significant, were minimally affected by restriction of the data-set. (Significance values were not relevant as they would have been affected by the reduction in sample size attendant on the analysis of fewer data.) The resilience of effect was not however shown in regard to the weekday versus weekend differ-
Weekday Weekend 6
0
15 30 Minutes after awakening
45
Figure 2 Mean (7SEM) salivary-free cortisol concentrations (nmol/l) after awakening on weekdays and at the weekend when participants showing more SNA at the weekend were excluded from the analysis (N ¼ 27).
ences. Figure 2 illustrates a repeat of the initial within-subjects ANOVA, using a simple exclusion criterion: omitting those participants ðn ¼ 21Þ who showed greater SNA frequency at weekends compared to weekdays, and thus removing the possibility that an increase in SNA could explain the
ARTICLE IN PRESS Suspected non-adherence and weekend/week day differences in the ACR previously reported difference in ACR. However ANOVA revealed that the weekday/weekend sample interaction reported for the total data set was no longer evident with an F-value approaching unity (F ð1:7;41:0Þ ¼ 1:069, p ¼ 0.344). Figure 2 illustrates the degree to which the original effect was abolished after removing SNA data.
4. Discussion The findings of this study revealed that in relation to levels and dynamics of salivary cortisol following awakening, there may be complex interplay between several variables: levels of perceived stress, awakening time, whether sampling takes place on weekdays or weekend days, and, not least, the extent to which the exact timing of self-administered saliva collections can be trusted. The primary finding of this study was a weekday/ weekend difference in the cortisol response to awakening. Although there was no weekday/weekend effect on total free salivary cortisol secreted, the dynamic rise in cortisol following awakening was greater on weekdays than at weekends. This was consistent with some recent reports (KunzEbrecht et al., 2004; Schlotz et al., 2004), although differences in the ACR between work and non-work days have not always been described (Kudielka and Kern, 2004; de Weerth and Buitelaar, 2005). Measures of state self-report stress tended to differ from weekday to weekend with participants reporting feeling less stressed at weekends (despite being students not in full-time paid employment). Participants also reported later waking times at the weekend. Differences in these variables could potentially account for the weekday/weekend effect on the ACR as awakening time has been shown to correlate negatively with the ACR (Edwards et al., 2001b; Kudielka and Kirschbaum, 2003; Federenko et al., 2004) and positive correlations between the ACR and self-report stress have been reported (Schulz et al., 1998; Wuest et al., 2000a). However, neither of these variables explained the weekday/weekend difference in the ACR reported here. Additionally, in this study, the weekend/weekday differences in the ACR were not associated with reports of more long-term stress as found by Schlotz et al. (2004). Although the weekend/weekday differences in the ACR could not be accounted for by either measure of stress or awakening time, negative relationships between composite measures of cortisol and both awakening time and scores on the PSS were observed. Higher MnInc and AUC cortisol
1015
levels averaged across all 4 study days were associated with earlier mean reported awakening times across the 4 days, in line with previous reports (Edwards et al., 2001b; Kudielka and Kirschbaum, 2003; Federenko et al., 2004). More recently it has been suggested that the effect for time of awakening can be explained by confounds with stress (Williams et al., 2005), however in the present study both PSS scores and awakening time were independent predictors of cortisol levels indicating that in this population, effects of the one were not significantly mediated by the other. Scores on the perceived stress scale, which measures how unpredictable, uncontrollable and overloaded respondents have found their lives over the last month, were also negatively associated with overall cortisol levels (but not with the dynamic measure). Participants who on average across all study days secreted the most cortisol in the first 45 min following awakening were more likely to have reported less stress for the previous month. Previously, an elevated rise in cortisol secretion following awakening has been found to be associated with higher levels of chronic stress in both males and females (Schulz et al., 1998; Wuest et al., 2000a). However the results of the present study are consistent with a study showing attenuation of mean cortisol secretion (as well as the dynamic of the rise) in women undergoing chronic economic stress (Ranjit et al., 2005). In part our results may therefore be attributable to a high percentage of female participants as it has also recently been shown that males and females differ in the association between trait negative affect (NA) and the ACR: females with high NA showed a reduced cortisol rise following awakening compared to females with low trait NA whereas males with high trait NA had a higher awakening rise in cortisol (Polk et al., 2005). In the present study the unequal number of male and female participants rendered no comparison between the sexes possible. Accurate measurement of the ACR relies upon participants collecting saliva samples at strict and frequent time intervals within the first 45 min after waking up. Previous research using electronic timing devices has shown that non-adherence to saliva sampling regimes is ubiquitous particularly when participants are unaware that they are being monitored (Kudielka et al., 2003; Broderick et al., 2004). In this study, we adopted a way of detecting potentially non-adherent participants that did not rely upon self-report data and that has been validated with cross reference to ECG and motility recording data (Kupper et al., 2005). Kupper and co-workers showed that in participants that failed
ARTICLE IN PRESS 1016 to show a rise in cortisol following awakening the actual waking time was on average 42 min earlier than the first purported waking sample (range 10 min–2 h 15 min). To be deemed adherent on any day in the present study participants had to show a rise (no matter how small) in cortisol concentration from sample 1 (immediately post awakening) to sample 2 (15 min post awakening) or sample 3 (30 min post awakening). Therefore SNA was defined, for any one day, as where the second or third sample was less than or equal to the first. It is important to emphasise that the exclusion criterion was not over-strict, in that it removed only data with no evidence of any rise whatsoever in the first 30 min. Although adherence to protocol was not objectively measured in this study, there were several factors which indicated that apparent non-responding was in fact due to non-adherence rather than intrinsically aberrant ACRs. Abnormal ACRs are associated with various pathologies both psychological and physical (Kudielka and Kirschbaum, 2003; Buchanan et al., 2004; Wolf et al., 2005). However, all participants in the present study were young, healthy and medication free. In a healthy population the ACR has a moderate to high within-subject stability (Wuest et al., 2000b; Edwards et al., 2001a) and appears to be a robust finding (Clow et al., 2004). Only one of our participants failed to show a cortisol response to awakening on any of the 4 study days. Furthermore, there was very little consistency across days in who was SNA suggesting that influences on adherence in this study were largely circumstantial rather than participant characteristics: SNA was a day-level rather than a person-level variable. This is illustrated by the finding that SNA was greater at weekend. Finally, the results indicated that the normal dynamic was lacking because the sampling period was shifted later than required by the protocol. The pattern for those deemed suspected non-adherent was of higher cortisol concentrations in samples collected purportedly on awakening and lower cortisol concentrations in final samples (purportedly 45 min after awakening). This would be the expected pattern if sampling had taken place later in a period normally characterised by a substantive rise followed by a fall. This is in line with previous findings of flatter or negative ACRs for nonadherent participants (Kudielka et al., 2003; Broderick et al., 2004; Kunz-Ebrecht et al., 2004; Wright and Steptoe, 2005). In order to provide a check on the possible influence of SNA on the findings reported for all participants, further analyses were performed taking SNA into account to examine whether the
L. Thorn et al. findings were resilient to its exclusion. Excluding the SNA data across each day did not diminish the effect sizes for the significant associations between mean awakening time, PSS scores and composite measures of cortisol providing corroborative evidence for the robustness of the effects. Therefore it seems unlikely that the association found in the total group could be explained by ‘noise’ from suspected non-adherent participants. However, analysis taking account of SNA status abolished the previously found weekday/weekend difference in the ACR. Others have speculated that reduced stress may account for the reduced ACR at weekends (Kunz-Ebrecht et al., 2004). In this study, even though participants woke up later and tended to report less stress at weekends, when SNA was taken into account there was no weekend/weekday difference. This finding is in contrast to Schlotz et al. (2004) who were able to demonstrate an attenuated weekend ACR, which was clearly associated with reduced work overload and fewer worries. The participants in the current study were all undergraduate students, not in full time employment and apart from levels of state stress and awakening time, lifestyle differences between the weekend and weekdays were not recorded. Many students hold weekend jobs and it is possible that the likely adherent participants were those about to initiate a day of work or even a day of hard study. Thus, minimal differences in lifestyle may have contributed to the failure to find differences after taking account of SNA. This has implications for best practice guidelines when studying the ACR across days on which there may be lifestyle differences. It may not be enough to simply record differences in levels of stress; active versus nonactive days or non-working versus work days should be considered. Although using the latter criteria studies have found no difference in the ACR (Kudielka and Kern, 2004; de Weerth and Buitelaar, 2005). That SNA explained an apparent difference between weekday and weekend responding in this study is a finding which needs to be treated with caution until more controlled research is done to examine actual adherence effects. However, these findings do suggest implications for best-practice guidelines for measurement of the ACR. The issue of non-adherence needs to be addressed in all future publications on the ACR. Self-report adherence may not be a guide to actual adherence (Kudielka et al., 2003; Broderick et al., 2004). In this study, only two participants reported that they had taken samples more than 10 min outside the times demanded by the protocol. Ideally, future studies need to objectively assess temporal
ARTICLE IN PRESS Suspected non-adherence and weekend/week day differences in the ACR sequence of sampling, for example using an electronic monitoring device (Kudielka et al., 2003) and awakening time using ECG and/or motility recording of the night and early morning hours (Kupper et al., 2005). Further, as participants appear to be less adherent at the weekend, the ACR may best be measured during the normal working week. The strategy used in this study could be used in retrospective re-analyses of data from healthy participants collected before adherence issues came to the fore. However, we would strongly urge caution in using it in different populations where it has been shown that the ACR may be inherently abnormal, e.g. in cardiovascular disease, psychiatric disorder, and hippocampal damage (Kudielka and Kirschbaum, 2003; Buchanan et al., 2004; Wolf et al., 2005). In such populations the strategy employed here would exclude adherent participants with aberrant ACRs. In summary, an apparent weekend/weekday difference in the dynamic of the ACR, which could not be accounted for by reduced stress or later awakening times, was found to be confounded by SNA to the sampling regime. SNA was examined by identifying instances of profiles showing no cortisol rise from awakening to 15 or 30 min post awakening and was found largely to be driven by situational factors (day of participation rather than participant characteristics). Analysis controlling for SNA status abolished the otherwise highly significant flatter profile at weekends, but not reported associations with perceived stress scores and awakening time. This finding has implications for future studies involving the measurement of the ACR as reported here, depending on day, between 17% and 33% of healthy adult participants may not have adhered to protocol. Data from suspected non-adherent participants could confound results leading to potential misinterpretation of findings.
References Broderick, J.E., Arnold, D., Kudielka, B.M., Kirschbaum, C., 2004. Salivary cortisol sampling compliance: comparison of patients and healthy volunteers. Psychoneuroendocrinology 29, 636–650. Buchanan, T.W., Kern, S., Allen, J.S., Tranel, D., Kirschbaum, C., 2004. Circadian regulation of cortisol after hippocampal damage in humans. Biol. Psychiat. 56, 651–656. Clow, A., Thorn, L., Evans, P., Hucklebridge, F., 2004. The awakening cortisol response: methodological issues and significance. Stress 7, 29–37. Cohen, S., Kamarck, T., Mermelstein, R., 1983. A global measure of perceived stress. J. Health Soc. Behav. 24, 385–396.
1017
de Weerth, C., Buitelaar, J.K., 2005. Cortisol awakening response in pregnant women. Psychoneuroendocrinology 30, 902–907. Edwards, S., Clow, A., Evans, P., Hucklebridge, F., 2001a. Exploration of the awakening cortisol response in relation to diurnal cortisol secretory activity. Life Sci. 68, 2093–2103. Edwards, S., Evans, P., Hucklebridge, F., Clow, A., 2001b. Association between time of awakening and diurnal cortisol secretory activity. Psychoneuroendocrinology 26, 613–622. Federenko, I., Wust, S., Hellhammer, D.H., Dechoux, R., Kumsta, R., Kirschbaum, C., 2004. Free cortisol awakening responses are influenced by awakening time. Psychoneuroendocrinology 29, 174–184. Kudielka, B.M., Broderick, J.E., Kirschbaum, C., 2003. Compliance with saliva sampling protocols: electronic monitoring reveals invalid cortisol daytime profiles in noncompliant subjects. Psychosom. Med. 65, 313–319. Kudielka, B.M., Kern, S., 2004. Cortisol day profiles in victims of mobbing (bullying at the work place): preliminary results of a first psychobiological field study. J. Psychosom. Res. 56, 149–150. Kudielka, B.M., Kirschbaum, C., 2003. Awakening cortisol responses are influenced by health status and awakening time but not by menstrual cycle phase. Psychoneuroendocrinology 28, 35–47. Kunz-Ebrecht, S.R., Kirschbaum, C., Marmot, M., Steptoe, A., 2004. Differences in cortisol awakening response on work days and weekends in women and men from the Whitehall II cohort. Psychoneuroendocrinology 29, 516–528. Kupper, N., de Geus, E.J.C., van den Berg, M., Kirschbaum, C., Boomsma, D.I., Willemsen, G., 2005. Familial influences on basal salivary cortisol in an adult population. Psychoneuroendocrinology 30, 857–868. Mackay, C., Cox, T., Burrows, G., Lazzerini, T., 1978. An inventory for the measurement of self-reported stress and arousal. Br. J. Soc. Clin. Psychol. 17, 283–284. Polk, D.E., Cohen, S., Doyle, W.J., Skoner, D.P., Kirschbaum, C., 2005. State and trait affect as predictors of salivary cortisol in healthy adults. Psychoneuroendocrinology 30, 261–272. Ranjit, N., Young, E.A., Kaplan, G.A., 2005. Material hardship alters the diurnal rhythm of salivary cortisol. Int. J. Epidemiol. 34, 1138–1143. Scheer, F., Buijs, R.M., 1999. Light affects morning salivary cortisol in humans. J. Clin. Endocrinol. Metab. 84, 3395–3398. Schlotz, W., Hellhammer, J., Schulz, P., Stone, A.A., 2004. Perceived work overload and chronic worrying predict weekend–weekday differences in the cortisol awakening response. Psychosom. Med. 66, 207–214. Schmidt-Reinwald, A., Pruessner, J.C., Hellhammer, D.H., Federenko, I., Rohleder, N., Schurmeyer, T.H., Kirschbaum, C., 1999. The cortisol response to awakening in relation to different challenge tests and a 12-hour cortisol rhythm. Life Sci. 64, 1653–1660. Schulz, P., Kirschbaum, C., Prussner, J., Hellhammer, D., 1998. Increased free cortisol secretion after awakening in chronically stressed individuals due to work overload. Stress Med. 14, 91–97. Thorn, L., Hucklebridge, F., Esgate, A., Evans, P., Clow, A., 2004. The effect of dawn simulation on the cortisol response to awakening in healthy participants. Psychoneuroendocrinology 29, 925–930. Williams, E., Magid, K., Steptoe, A., 2005. The impact of time of waking and concurrent subjective stress on the cortisol response to awakening. Psychoneuroendocrinology 30, 139–148.
ARTICLE IN PRESS 1018 Wolf, O.T., Fujiwara, E., Luwinski, G., Kirschbaum, C., Markowitsch, H.J., 2005. No morning cortisol response in patients with severe global amnesia. Psychoneuroendocrinology 30, 101–105. Wright, C.E., Steptoe, A., 2005. Subjective socioeconomic position, gender and cortisol responses to waking in an elderly population. Psychoneuroendocrinology 30, 582–590.
L. Thorn et al. Wuest, S., Federenko, I., Hellhammer, D.H., Kirschbaum, C., 2000a. Genetic factors, perceived chronic stress, and the free cortisol response to awakening. Psychoneuroendocrinology 25, 707–720. Wuest, S., Wolf, J., Hellhammer, D.H., Federenko, I., Schommer, N., Kirschbaum, C., 2000b. The cortisol response to awakening—normal values and confounds. Noise Health 7, 79–88.