Serum cortisol is a widely accepted indicator of energy mobilisation and hence a useful indicator of stress [1]. It has been discovered by Kirschbaum and Hellhammer and confirmed by other groups that variations in saliva concentration reliably reflect variations in the serum concentration of free cortisol [2, 3]. Several studies of saliva cortisol variations during the normal round of life and in relation to stressful experiences in several normal groups have been published. However, few studies of representative working population groups have been studied. In the present study the relationship between one of the more widely used theoretical job stress models, the demand control model, and circadian variations in saliva cortisol has been explored.

Previous studies have shown that there are elevated levels of saliva cortisol during the early morning hours in subjects with high demands and low decision latitude (control). The findings have not been consistent however [4–8]. One of the reasons for this could be that subjects with marked disturbances of the capacity to regulate cortisol excretion could distort such a relationship. For instance, studies of serum cortisol variations in patients with severe long lasting psychiatric depression have shown that these subjects are frequently unable to lower their cortisol concentration in the evening [9]. This has also corresponded to inability in a large proportion of these subjects to lower serum cortisol during the dexamethasone test [9]. Subjects with this inability to lower cortisol ("high curves") are less likely than others to respond in the normal way to stressors in everyday life. Similarly it has been discovered that subjects with pronounced symptoms of exhaustion such as the chronic fatigue syndrome are unable to raise their cortisol level in challenging situations and they also show very small circadian variation ("low flat curves") [10, 11]. These physiological processes may be behind some of the contradictory findings. It could be that the proportion of subjects who have been exposed for a very long time to adverse conditions – in ways which may influence the regulatory capacity – has been different in different study samples. In the present study a representative sample of the actively working population in Stockholm was studied. This means that serious depression and exhaustion making active work impossible could not disturb our findings. Another reason behind inconsistent findings may be differences in the prevalence of potential confounders – such as smoking habits, alcohol consumption, body mass index and medication – between study samples. In the present study it was possible to explore the importance of such factors to possible relationships between the demand control model and saliva cortisol concentration medication.

There were no differences between men and women with regard to when they produced their saliva samples. The second, third and fourth saliva sampling times were on average given for men and women respectively at 38 min (SD = 28 min) and 34 min (SD = 16 min) after awakening, 5h43 min (SD = 1h24 min) and 5h38 min (SD = 1h13 min) after awakening and 15h57 min (SD = 1h29 min) and 15h44 min (SD = 1h15 min) after awakening. When subdividing subjects into different demand-control categories the pattern did not change markedly neither for men or women.

The figure depicts the median cortisol values among women at the four time points for each demand control group. It shows a consistent difference between the low strain group and the other three groups both at awakening and at half an hour later. There are no differences between the active, passive and job strain groups (see figure 1).

Medication had surprisingly small effects on the median saliva concentration in the different groups. Accordingly, exclusion of subjects with medication that was judged to be particularly significant to cortisol excretion did not have any effects on the median saliva concentration of cortisol in the four categories. However, it could be that some of the more prevalent medications such as antidepressants restore the capacity to respond with elevated excretion of cortisol as a response to job stress. Similarly, analyses including medication as a possible confounder had very small effects on the results.

No significant demand-control*time point interactions were seen. There were, however, significant differences between the female groups in the separate analysis half an hour after awakening. Since there were no significant main or interaction effects in analysis of variance this finding should be interpreted with caution. Half an hour after awakening is probably the period when anticipation of a stressful day may exert strong effects on mobilisation of energy. Accordingly, despite the relative statistical weakness of these findings they are probably not random findings. It should also be pointed out, however, that first of all that there tends to be a difference between the low strain group and the others already at awakening and secondly that the rise in saliva cortisol concentration between awakening and half an hour later that has been discussed extensively in the literature did not differ between the demand control groups [2, 21]. This could speak against the interpretation that we are dealing with an anticipation effect. One factor that could be of importance to this negative finding is the definition of "awakening". Whether subjects physically rise to their feet or not has been shown to be unimportant [25] but we may not have defined the point of "awakening" in a sufficiently standardised way in this study. Hence the true time lag between "awakening" and half an hour later may have varied considerably between subjects.

Another puzzling finding in this study was that the job strain, active and passive groups could not be differentiated from one another. We have no adequate explanation of this. According to the main hypothesis we should have found higher levels in the job strain group than in the active and the passive groups. All we can say is that there was no interaction between demand and control in the expected direction. Women reporting neither high demands nor low control had lower saliva cortisol than others half an hour after awakening.

There is a rapidly growing literature which relates both the demand control support model and the effort reward imbalance model to cortisol regulation. Steptoe et al have studied variations in saliva cortisol over the day in relation to overcommitment and the external part of the effort reward model [26]. That study which was based upon contrasting samples from the Whitehall II study showed that men who had high scores on the overcommitment scale had on average 22% higher saliva cortisol concentrations than men who had low scores. Comparisons between these groups also showed that the rise in saliva cortisol concentration from awakening to half an hour later was higher in overcommitted than in non-overcommitted men. No such findings were made in women. There is clearly a relationship between overcommitment and high demands. The external part of the effort reward score (a bad score indicates low reward for high effort) was not related to saliva cortisol levels neither in men nor in women in that study. Steptoe et al have also shown that teachers (men or women) with job strain have higher saliva cortisol levels at 800 and 830 in the morning than other teachers [7]. These associations were particularly strong in subjects who reported a high level of "anger out", tendency to react with openly expressed anger in stressful situations. A study of Japanese female health care workers showed higher urinary catecholamine output in those with self-reported job strain than in others. Saliva cortisol levels, on the other hand, were consistently lower in the job strain group than in the others [4].

Accordingly previous literature has shown that cortisol regulation is influenced by job stress. However the Japanese study shows that in some occupations early stages of physiological exhaustion leading to small differences between morning and evening cortisol excretion could be more prevalent in the job strain category than in the other categories [4]. The literature so far has accordingly produced inconsistent findings and our own study does not seem to provide any final answer. We only find weak support for the general demand control hypothesis. Women expecting neither high demands nor low control at work during their working day have a lower saliva cortisol level half an hour after awakening (perhaps when they have started to prepare themselves mentally for the upcoming conditions).

It is quite possible that there is a true difference between men and women. Kunz-Ebrecht et al showed that women seemed to react with more morning saliva cortisol elevation to the expectation of a stressful work day than men [21]. This could be one indication that women are more sensitive to work stress with regard to cortisol excretion than men but this has not been explored systematically in the literature. That the findings for men are non-significant in the present study might possibly also be due to a limited number of men.

The design of this study was based on selection of subjects with similar numbers from two groups. The groups consisted of those having low or high well-being, respectively, according to the WHO (Ten) Well-Being Index in the questionnaire in the second phase. It could be argued that a selection of subjects should have been done to secure similar numbers in the four different demand-control categories instead. Despite all, well-being was not related to cortisol and thus did not change the relation between job strain and cortisol. We also performed some analyses where the relations between job strain and cortisol were examined separately among those with high and low well-being. These analyses showed very similar patterns in the two well being groups. Low strain showed the lowest means in both groups while higher and comparable cortisol levels were found in the other three job strain categories in both groups.

The circadian variation in saliva cortisol concentration was studied in a representative sample of actively working people in the greater Stockholm area. The study design allowed adjustment for a number of potential confounders such as mental state, medication, body mass index, smoking and alcohol consumption none of which had any significant impact on our findings. The results indicated that the saliva cortisol concentration was lower half an hour after awakening in women who reported neither high demands nor low decision latitude than in other working subjects. This association was only found in women and not in men, however. There was no difference between the three categories active, passive and high job strain. The findings only give weak support to the hypothesis that the demand control model is associated with saliva cortisol concentration.