| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
ORIGINAL ARTICLES |
From the Psychophysiology and Stress Research Group (S.E., F.H., A.C., P.E.), Department of Psychology (S.E., A.C., P.E.), and Department of Biomedical Sciences (F.H.), University of Westminster, London, United Kingdom.
Address reprint requests to: Professor Phil Evans, Department of Psychology, 309 Regent Street, London W1R 8AL. Email: evansp{at}westminster.ac.uk
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSNOTESREFERENCES |
|---|
METHODS: Salivary free cortisol concentration was determined in 34 healthy participants (students) at eight time points, synchronized to awakening, on 2 consecutive days. Participants completed a standard questionnaire to assess perceived stress and subsequently kept a daily record of social proximity and symptoms of upper respiratory illness for 2 weeks.
RESULTS: Participants characterized by consistently larger awakening responses went on to report significantly more URI symptoms. Participants with less pronounced diurnal decline (flatter profiles) reported fewer URI symptoms. The two cortisol components were themselves related and interacted such that participants high on an interactive vector reported approximately three times more URI symptoms than other participants. The URI-associated cortisol components (dynamic changes) were not related to perceived stress, but underlying cortisol secretory activity (overall levels) in the first 45 minutes after awakening was. Dynamic components were, however, related to a social proximity measure, which in turn was related to URI symptoms. Proximity and the interactive cortisol vector together explained a substantial (28%) percentage of the variance in URI symptom reports. The cortisol vector independently and significantly explained 12% of the variation; the proximity measure independently and nonsignificantly contributed 6% of the variation.
CONCLUSIONS: URI symptoms were associated with two related dynamic components of the cortisol cycle as determined by synchronization to awakening, whereas stress was related to a measure of overall secretory activity.
Key Words: cortisol, • stress, • upper respiratory illness, • diurnal rhythm.
Abbreviations: AUC = area under the curve;; BSLOPE = regression coefficient;; CPS = Close Proximity Scale;; DAUC = diurnal area under the curve;; HPA = hypothalamic-pituitary-adrenal;; MnInc = mean increase;; NK = natural killer;; NMAC = Nowlis Mood Adjective Checklist;; PSS = Perceived Stress Scale;; Th1, Th2 = T-helper cells;; URI = upper respiratory illness.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSNOTESREFERENCES |
|---|
Several recent studies have explored individual differences in the diurnal cycle of cortisol, in particular the typical negative slope, which is the normal pattern across the day. Abnormal patterns, particularly flatter slopes and smaller differences between earlier and later values across the day, have been shown to characterize those arguably under stress, such as maltreated children (8), and those with poor relationship functioning (9). The same sort of pattern has also been found to be significant in certain clinical disorders, including chronic fatigue syndrome (10) and breast cancer (11). Sephton et al. (12) recently showed that flattened profiles are predictive of earlier mortality in breast cancer patients and are also associated with suppressed antitumor activity of immune natural killer (NK) cells. In stark contrast to these findings that flattened profiles are associated with both pathology and stress-related measures, Smyth et al. (13) reported that during a 2-week study period, individuals with relatively flat cycles were not significantly more stressed and reported fewer upper respiratory illness (URI) symptoms. Methodologies in regard to the assessment of cortisol status have varied somewhat across studies of diurnal factors. Most significantly, few studies have explored the diurnal cycle from and synchronized to awakening or provided concurrent information about the brief but pronounced cortisol response (rise) that immediately follows awakening.
Awakening is a potent stimulus to adrenal cortisol secretory activity. Free cortisol concentration measured in saliva increases some two- to three-fold in the first 30 minutes after awakening (14, 15) . These latter two studies used participants who collected saliva samples in the naturalistic settings of their own homes and while performing routine morning activities. The same pattern of activity has also been demonstrated for total serum cortisol, measured by indwelling catheter, in sleep laboratory studies (16, 17) . The salivary free cortisol response to awakening is largely independent of physical activity, alcohol consumed before sleep, and smoking status (14). It has also been shown to be independent of blood glucose levels at the time of awakening, implying that mobilization of energy reserves in preparation for daytime energy expenditure is not a primary physiological function of this response (18). There is, however, some evidence that the high levels of cortisol during this high-activity period relate to perceived stress, although responses may actually be blunted in those suffering from "burnout" (19).
It has been suggested that a primary role of the increase in free cortisol in response to awakening may be to switch the immune system from nighttime cellular (Type 1) domination to daytime humoral (Type 2) domination (18). Petrovsky and Harrison (1) demonstrated the cytokine profile from stimulated whole-blood leukocytes, sampled at intervals during nocturnal sleep, were characterized by T-helper (Th1; Type 1) immune domination, whereas this switched to a Th2 (Type 2) cytokine profile after awakening. Because the degree of this switch correlated with concurrent cortisol levels, these authors argued a role for cortisol in this process. Both in vivo (1) and in vitro (2) studies have demonstrated that cortisol can influence the balance of T-helper cell cytokine production in this way. The balance of an individual’s immune system, with possible implications for infection and disease susceptibility, may therefore be related to the awakening free cortisol response.
We examined the association between the diurnal pattern of cortisol secretory activity, including the awakening response and susceptibility to URI. Using a prospective experimental design, we characterized the diurnal cortisol profile synchronized to awakening in a population of healthy adults. Susceptibility to URI measured by self-report was monitored over a 2-week follow-up period. Because a confounding variable for URI might be incidence of social contact, we also monitored this aspect of social behavior using a proximity measure. In addition, because cortisol has been associated with perceived stress and affect (5), in the same study we examined the associations between perceived stress, mood, and the diurnal cortisol profile.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSNOTESREFERENCES |
|---|
Procedure
The study fell into two main phases. Phase 1 consisted of 2 sampling days during which participants were instructed to collect saliva samples at home on two convenient consecutive days within the same 1-week period of the spring term. On each day samples were collected immediately on awakening and then 15, 30, and 45 minutes after awakening. A further four samples were collected at 3-hour intervals through the day, all synchronized to awakening time. Each sample was collected using the Salivette sampling device (Sarstedt Ltd., Leicester, UK) and frozen as soon as possible. These general procedures have been shown to be reliable in assessing the diurnal cortisol profile, including the awakening response, within an ambulatory setting (20). Participants were instructed to take nothing by mouth, other than water, to avoid postprandial cortisol increases, and to not brush their teeth, to avoid abrasion and possible vascular leakage, until the fourth sample (45 minutes after awakening) had been collected. Participants were otherwise free to go about their normal morning routine. For the remaining samples participants were instructed not to eat, drink, or smoke for at least 30 minutes before collection of samples. A form was provided on which participants were asked to provide details of date and time of sample collection. A badge was also provided as an aide memoir of sampling times.
An initial pack of questionnaires including, for the purposes of the results reported here, the Perceived Stress Scale (PSS; Ref. 21) was given to participants to be completed at any point during the sampling period (phase 1). Participants were instructed to return their frozen samples to the laboratory (using the provided insulated cold pack) together with the initial battery of completed questionnaires.
Phase 2 was a 2-week prospective period during which participants were instructed to complete the following daily questionnaires every evening before bed: a daily URI checklist of symptoms (eg, sore throat, congested nose, sneezing, etc.) that has been used in previous studies of this kind (22), the Nowlis Mood Adjective Checklist (NMAC; Ref. 23), and a Close Proximity Scale (CPS), formulated specifically for the present study. The CPS consisted of three questions: "How many hours approximately have you spent alone today, not in close proximity to anyone (including family, partner etc)?"; How many hours approximately have you spent in busy places, in close proximity to other people (restaurants, pubs and clubs, public transport, cinema, lectures, meetings of more than five people in a room, etc)?"; and "How many different people have you physically met up with today (to qualify as a meeting, it should be more than a wave or good morning in a corridor, ie, you’ve passed some time with the person or small group of persons)? Further explanations for each question were included in the CPS instructions to maximize standardization of responses. To aid compliance a researcher (S.E.) made arrangements to meet the participants at the end of each week to collect completed questionnaires and to hand out new packs for the following week. In the event that participants were unable to meet the researcher, a deposit box was provided for exchange of materials.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSNOTESREFERENCES |
|---|
As background to the results reported here (see Refs. 24 and 25), repeated-measures analyses of variance revealed significant main effects of sample point for awakening samples and diurnal samples on both day 1 and day 2. On both days there was a marked increase in cortisol concentration after awakening, peaking on average at 30 minutes (sample 3). The diurnal pattern (excluding the awakening response) showed a sharp and then smooth linear decline from immediately on awakening until 12 hours after awakening. Thus the general pattern of cortisol for the sampling period was fully in line with what has been reported widely in the literature.
Computation of Composite "Profile" Measures
The awakening response can be looked at in two ways: the absolute values in the whole period and the dynamic increase after an initial waking sample, peaking after about 30 minutes. An area under the cortisol curve with reference to zero (AUC) was used to estimate total cortisol secreted during the 45 minutes after awakening. Degree of cortisol response to waking was calculated (see Ref. 26; also Ref. 24) by computing the mean increase of samples 2 to 4 from the initial waking values (MnInc). We have shown that this is in fact a virtually identical measure to calculating an area under the curve with reference to the first awakening sample (r > 0.97). The diurnal samples (samples 1, 5, 6, 7, and 8), which represented cortisol secretion during the 12-hour diurnal period excluding the awakening response, were computed as a further area under the curve for the whole 12-hour day (DAUC). We noted that after a sharp decline from waking values to values 3 hours later, the average profile for the following 9 hours was almost perfectly linear. To investigate individual differences in this linear decline, we calculated regression (slope) coefficients (BSLOPE) from 3 to 12 hours after awakening for each subject on each day.
We previously reported that these measures, for the same participants, were reasonably stable across days (24, 25) . However, general stability overlooks possible inconsistencies with regard to individual difference analyses, so we also assigned participants to groups according to their stability across days, following procedures similar to those of Smyth et al. (13). To do this absolute difference scores between day 1 and day 2 were calculated for each cortisol composite measure. Participants whose differences scores were greater than 1 standard deviation above the difference mean were classified as inconsistent (unstable) on any measure and excluded from further analyses. We thought it was important to exclude such participants. Not only were they relatively small in number, but also their status, if included, would be singularly ambiguous, with one day’s data classifying them totally differently from the other day’s data and the mean data distorting their real status. Remaining participants were then classified as above or below the median on the measure in question. Thus above-the-median participants on AUC (N = 12) had consistently higher waking-period cortisol; on DAUC (N = 12) they had consistently higher cortisol during the 12 hour period (excluding samples 2, 3, and 4); on MnInc (N = 13) they had a consistently larger cortisol response to waking; and on BSLOPE (where the normal pattern is negative slope, ie, decline) they were labeled as having more "flat" profiles (N = 15) as opposed to more "normal" profiles. We have in this instance used the words "normal" and "flat" because the same words were used by Smyth et al. (13) to label similarly derived groupings of participants. However, frequency inspection of our data gave no clear separation of more normal and flatter profile groups in our data; hence the terms normal and flat are used relatively in this study. The mean diurnal profiles by group from +3 hours to +12 hours are plotted in Figure 1.
|
URI Measures in Relation to Cortisol Profile Measures
Missing cortisol data were slightly more common in the waking period measures. Thus total sample sizes slightly vary according to composite under consideration (N = 34 for BSLOPE groupings; N = 33 for DAUC groupings; N = 31 for both AUC and MnInc groupings). Few participants were labeled inconsistent for any measure (six for DAUC and AUC; five for MnInc and BSLOPE).
Differences between the stable above- and below-the-median groups based on each cortisol composite measure (across the 2 sampling days) are presented in Table 1. The dependent variable is the square-root-transformed number of URI symptoms reported in the 2 weeks after cortisol assessment. Raw symptom data were highly positively skewed, with several participants reporting zero symptoms, and the square root transformation significantly normalized the distribution.
|
We examined the sensitivity of these effects, using slightly different analysis strategies. First we looked at straightforward correlational analyses between URI symptoms and continuous cortisol profile measures, with and without exclusion of inconsistent participants. Although coefficients were not statistically significant, directionally they pointed to similar conclusions and were more nearly significant for MnInc than BSLOPE. In contrast further categorical analysis actually reinforced the finding of association between URI vulnerability and cortisol profiles. Thus a tiny fraction (only 6%) of high-negative-slope participants reported zero URI symptoms, whereas the vast majority (94%) reported at least one URI symptom. In contrast 46% of flatter profile participants reported zero symptoms, and 54% reported at least one symptom (p < .019). Interestingly categorical analysis by MnInc groups and URI vulnerability suggested that the significance of this association was maximized by comparing those with most URI symptoms (highest tertile) vs. those with zero or middling counts. None of those categorized as consistently low MnInc participants were classified in the highest symptom tertile (p < .02). These analyses suggest that, for both BSLOPE and MnInc, there are clear associations with URI symptoms but the associations are not entirely linear, and that thresholds for effects also may differ for the two types of cortisol measure.
There was considerable overlap between participants categorized by BSLOPE and MnInc, although a straight correlation between the two (r = -0.20) was not significant. Nearly 80% of participants were categorized as either high MnInc and more normal slope or as low MnInc and flatter slope (p < .004). Inspection of the data suggested that the two cortisol profiles interacted such that URI symptom reports were a characteristic of those high on MnInc and with high negative slope. Although correlational analysis, as reported above, did not yield significant coefficients between URI symptoms and MnInc and BSLOPE taken separately, there was a strong and clearly significant relationship between an interactive vector and URI symptoms (r = 0.46, p < .01). For illustration purposes only, we have presented in Figure 2 means for those whose product vectors indicate flattest slopes and/or lowest MnInc (N = 8), those who are intermediate (N = 10), and those whose products indicate steepest slopes and/or highest MnInc (10). The groups are not exact tertiles but reflect discrete gaps in the distribution of the product vector. It can be seen that the effect is explained solely by the greater incidence of URI symptoms in the last group.
|
Mean differences by cortisol profile groups were examined for perceived stress (PSS) at the time of cortisol assessment and average daily reported positive and negative moods during the prospective period. No significant differences emerged on either measure between more normal and flatter BSLOPE groups; nor were any effects evident for MnInc. Thus neither perceived stress nor mood were significantly associated with the URI-associated cortisol profiles. Higher AUC participants did, however, report significantly more stress (p < .05) on the PSS measure (mean = 26.67, SD = 5.86) than lower AUC participants (mean = 22.29, SD = 5.74). This effect was not apparent for groupings based on DAUC, suggesting the possible importance of absolute waking period values of cortisol for identifying individual differences in stress experience.
Proximity was looked at in terms of mean hours alone per day, mean hours in busy places per day, and mean number of persons closely encountered. The latter two measures were significantly related to number of URI symptoms. Thus those seeing more persons (r = 0.37, p < .01) and those spending more time in busy places (r = 0.31, p < .025) reported more URI symptoms. Mean hours spent alone was not predictive of URI symptoms.
In terms of cortisol measures, means by groupings were examined for the two proximity measures that were significantly related to URI symptoms (seeing more people and spending more time in busy places). Results are presented in Table 2.
|
Finally the two major and clear predictors of URI symptoms (the proximity measure of time in busy places and the interactive vector, representing the combined effects of MnInc and BSLOPE) were both entered in a multiple regression. Together both variables account for approximately 28% of the variance in URI symptoms. Independently of proximity, the combined cortisol vector significantly increases explained variance by 12% (p < .05), whereas independently of cortisol, proximity increases explained variance by 6% (p < .15, NS).
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSNOTESREFERENCES |
|---|
The other cortisol profile measure associated with URI symptoms in our study was also a dynamic rather than absolute measure, namely the degree of diurnal linear decline in cortisol from 3 hours to 12 hours after waking. Although the two URI-associated measures are independent in the sense that MnInc and BSLOPE used nonoverlapping sample points in their computation, there was some indication that they were empirically related, especially when categorical analysis was used: Those consistently showing high awakening responses also showed consistently more normal sharp decline over the rest of the day. As far as slope is concerned, our results are similar to those of Smyth et al. (13) despite some differences in approach. In deriving slope coefficients, Smyth et al. took clock-timed rather than waking-related samples between 8:00 and 22:00 hours in healthy individuals. Like us, however, they assessed cortisol status on 2 consecutive days and distinguished between consistently normal and consistently flat groupings. The percentage of "inconsistent" participants was somewhat greater and the percentage of flat profile participants smaller in their study; also there seemed to be a qualitative difference between groups that was not apparent in our data, where the frequency distribution of slope coefficients was reasonably smooth. Smyth et al. (13) did not report on awakening responses or morning values tied specifically to individual waking times, but it seems from their presentation that flat profile participants did have lower values at 8:00 AM despite no over all significant difference in total mean levels between groups over days. We could therefore speculate that lower 8:00 AM values might reflect, partly at least, less waking-stimulated rises in some of their participants.
In all events (and taken together interactively), MnInc and BSLOPE explained more than a quarter of the total variation in URI symptom reporting in our study. This is not a small effect size and leads us to speculate about the reasons for it. Much has been made of links between stress and vulnerability to infectious illness perhaps being mediated by neuroendocrine and immunological processes. This study provides no clear or simple illumination in this regard: Aspects of the cortisol profile that related to stress, notably mean levels in the post waking period, were not significantly related to URI symptoms; equally the cortisol measures that were related to URI were not related to stress.
We speculated in the Introduction that links between cortisol profiles and URI might involve the role of diurnal cortisol rhythms in modulating the immune system. In particular, one function of the immediate rise in cortisol after awakening might be to shift the immune system away from typical nocturnally dominant Th1 mode toward typical daytime dominant Th2 mode. Although absolute (higher) levels of cortisol might be expected prima facie to favor more Th2 activity, it must be remembered that the effects of cortisol will depend crucially on prevailing receptor sensitivity. Thus individual differences in immunoregulatory effects may in turn be more sensitive to degrees of short-term dynamic changes within individuals. This may be one reason why in general terms it was the dynamic rather than the absolute measures that were associated with URI symptoms. But what of the direction of findings?
There is certainly evidence that more dominant Th1 bias through associated cytokine activity may be associated with greater severity and therefore awareness of sickness symptoms (1, 28) . This is still perfectly consistent with the view that efficient Th1 functioning is crucial to cellular immune system defense against viral infection and is selectively compromised by stress and HPA hyperactivity (29). One plausible interpretation of our results and those of Smyth et al. (13) might therefore be that greater awareness of (but not necessarily vulnerability to) URI might be a feature of Th1-dominant individuals. Such individuals may be characterized by the need for a more vigorous but very short-term boost in cortisol after awakening to drive a normal Th2 shift, but thereafter their characteristic Th1 bias may be evidenced by steeper-than-normal declines in cortisol over the daytime period. This steep decline in a powerfully anti-inflammatory hormone (relative to their own starting values) may result in greater awareness of what are in effect inflammatory symptoms. Although speculative, this interpretation does at least help to reconcile what certainly seems otherwise to be a paradox in the results of Smyth et al. (13) and the replication and extension of those results reported here. The paradox quite simply is that for the most part normal dynamic responses (awakening surges and thereafter diurnal declines) have tended to be associated with good health rather than poor health, in respect to both physical and psychological dimensions.
Notwithstanding such arguments, the findings of significant associations between our index of proximity and both cortisol profile and URI symptoms may suggest that cortisol profiles could be a marker for certain behavior (such as social interaction) that in turn involves greater exposure to infectious agents. If so, it is likely to be a broader marker than our simple proximity items questionnaire since cortisol profile remained independently significant in the multiple regression. It is important to remember, however, that in this study and that of Smyth et al. (13), ill health was assessed by symptom self-reports, and infectious status was not assessed. We have argued in previous work involving similar measurement of URI (22) that self-reports of such illness have generally good validity and are highly correlated with independent clinical assessments of incidence and severity (30). Nevertheless, it may still be the case that subjective differences in symptom perception may have played a role in determining the triangular relationships between URI, cortisol profiles, and proximity. For example, it could be argued that more emotionally flat or repressed individuals might be marked by a blunted awareness of symptoms, fewer social contacts, and a more flattened cortisol profile.
All of our discussion so far would suggest that an excellent research strategy for future investigation would be to study diurnal and awakening profiles in experimental viral exposure studies, where exposure to viruses and infectious status is by the nature of the study properly controlled. If our speculations have any validity, it may well be that the results of such a study would be in stark contrast to those reported here and by Smyth et al. (13), and it may even indicate with respect to actual vulnerability to infection that flat profiles and diminished early morning rises are associated with poorer response to viral challenge.
One further issue needs some mention. It is theoretically possible that the results reported here and by Smyth et al. (13) may partly reflect differences in cortisol profiles consequent on recent but unreported prior illness episodes, which might of course influence the probability of reporting further symptoms in the prospective period. We cannot speak to this decisively, but we did have access to information from a database that recorded answers to questions about general health in the month before the present study period. In particular we looked at responses to a question about whether the participant had had an episode of cold or flu. What we can report is that responses to this question were not predictive of either subsequent cortisol profiles or prospective URI measures.
Although stress is not the primary focus of the results reported here, it is interesting to contrast the negative results reported here and by Smyth et al. (13) with studies that have linked stress and pathology to flatter diurnal profiles. Clearly there are several issues that make it difficult to find any simple order in the overall body of findings. There is the usual issue of specificity of stress measures (eg, perceived stress vs. measures of relationship functioning; Ref. 9). Equally clearly, we cannot easily generalize from studies involving particular clinical groups to questions of differences among nonclinical participants. There are also issues of how, how often, and over what time course cortisol is sampled, whether samples are pegged to clock-time or awakening time, and indeed whether the awakening response itself is considered. However, one other point that needs to be emphasized is that similar slope statistics, like any summary measure, can be the result of very different profiles. Average profiles by group (such as are shown in Fig. 1) or similar information may be gleaned in several studies but may be quite limited. We can say that morning differences rather than later differences characterize the slope groups in our study. However, we probably need to know more detail before we can expect to find order across studies. The problem is that without prohibitively large samples, it is difficult in a single study to investigate the very different subgroups that might all be characterized, for example, by flatter slope. All of these considerations serve to emphasize the complexities of research in this area. The results reported here arouse a particular interest because they are partially replicative but also in a wider context paradoxical. We have sought to offer interpretations that are plausible in terms of reconciling paradox, but more research is needed to turn speculation into fact.
|
CONCLUSIONS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSNOTESREFERENCES |
|---|
|
NOTES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSNOTESREFERENCES |
|---|
Received for publication August 14, 2001.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSNOTESREFERENCES |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
