Moderation of psychosocial risk factors through dysfunction of the hypothalamic–pituitary–adrenal stress axis in the onset of chronic widespread musculoskeletal pain : Findings of a population-based prospective cohort study

Abstract

Objective

To test the hypothesis that abnormalities in the hypothalamic–pituitary–adrenal (HPA) stress-response system would act as an effect moderator between HPA function and the onset of chronic widespread pain (CWP).

Methods

We conducted a population-based prospective cohort study. Current pain and psychosocial status were ascertained in 11,000 subjects. Of the 768 eligible subjects free of CWP but at future risk based on their psychosocial profile, 463 were randomly selected, and 267 (57.7%) consented to assessment of their HPA axis function. Diurnal function was measured by assessing levels of salivary cortisol in the morning (9:00 AM) and evening (10:00 PM). Serum cortisol levels were measured after an overnight low-dose (0.25 mg) dexamethasone suppression test and a potentially stressful clinical examination. All subjects were followed up 15 months later to identify cases of new-onset CWP.

Results

A total of 241 subjects (94.9%) completed the followup study, and 28 (11.6%) reported the new onset of CWP. High levels of cortisol post-dexamethasone (odds ratio [OR] 3.53, 95% confidence interval [95% CI] 1.17–10.65), low levels in morning saliva (OR 1.43, 95% CI 0.52–3.94), and high levels in evening saliva (OR 2.32, 95% CI 0.64–8.42) were all associated with CWP. These 3 factors were found to be independent and additive predictors of CWP (OR for all 3 factors 8.5, 95% CI 1.5–47.9) in analyses controlling for age, sex, depression, sleep disturbance, recent traumatic life events, and pain status. One or more of these 3 HPA factors identified 26 (92.9%) cases of new-onset CWP.

Conclusion

Among a group of psychologically at-risk subjects, dysfunction of the HPA axis helps to distinguish those who will and will not develop new-onset CWP.

Chronic widespread pain (CWP) affecting the musculoskeletal system, the principal symptom of fibromyalgia, is common, with a 1-month population prevalence of ∼11% (1). It is associated with loss of function and considerable disability, may be associated with increased mortality rates (2), and is a major cause of health care utilization both in primary and secondary care settings. We have previously demonstrated in the prospective Altrincham Pain Study conducted in northwest England (3) that an increased risk of CWP onset is predicted from high levels of psychological distress, other somatic symptoms, and abnormal illness behavior. These factors are indicative of the process of somatization that can be defined as the tendency to express psychological distress as physical symptoms. These results confirmed for the first time that psychosocial factors preceded the onset of CWP, rather than just being a consequence. Although psychosocial factors are strong and robust predictors of CWP, the majority of “at-risk” subjects do not develop symptoms of CWP: only 22% of subjects in the group at highest risk reported new-onset CWP.

We hypothesized that the action of such psychosocial factors would be moderated through a biologic pathway. The hypothalamic–pituitary–adrenal (HPA) stress-response system reacts to both physical and psychological challenges in an adaptive response. The key determinant to a successful adaptive response upon exposure of the subject to a stressor is activation of the HPA axis followed by its termination after the stressor has passed. Upon activation of the HPA axis, increased levels of cortisol lead to the initiation of physiologic and behavioral responses, which ultimately result in allostasis and initiation of adaptive mechanisms, including analgesia. The relevance of these findings to the etiology of chronic pain and fibromyalgia is based on a number of studies.

The clinical features of muscle pain, joint pain, and widespread tenderness that are characteristic of fibromyalgia (as well as additional associated factors, including sleep disturbance and fatigue) are common in other hypocortisolic states (4). Studies of fibromyalgia patients in tertiary care settings have shown reduced basal plasma cortisol levels, reduced 24-hour urinary free cortisol excretion, and a blunted cortisol response to ovine corticotropin-releasing hormone (CRH) (5–7). Crucially, these studies of patients with established symptoms were unable to determine whether the observed HPA axis alterations preceded or were a consequence of having CWP. Neither did they account for the effects of anxiety, depression, life stresses, and sleep disturbance, all of which are associated with HPA axis dysfunction and may explain the observed relationship (8).

The only way to establish the nature of the relationship is to conduct a prospective cohort study in which subjects who are free of CWP but are at risk of developing CWP are identified, their HPA axis function assessed, and their courses are followed over time in order to establish who develops pain. We conducted the first such study to test the hypothesis that among a group of subjects free of CWP, altered HPA function would mediate the relationship between psychosocial risk factors indicative of the process of somatization and the onset of symptoms of CWP. We further hypothesized that this relationship would be independent of the effect of concomitant psychosocial factors that may be confounding the relationship, including depressive symptoms and sleep disturbances.

SUBJECTS AND METHODS

Study design.

We conducted a prospective population-based cohort study. Subjects completed a questionnaire that enquired about aspects of psychosocial status, together with a history of current pain. Those who were free of CWP but were at risk of the future development of CWP based on their psychosocial status (see below) were identified and invited to further participate in a detailed assessment of HPA axis function. Of those who agreed to participate, several subjects had to be excluded (Table 1) because accurate HPA assessment would not have been possible. All subjects were followed up for 15 months after baseline, at which time those reporting the new onset of CWP were identified.

Table 1. Exclusion and inclusion criteria used in the present study

Exclusion criteria

Type 1 diabetes mellitus

Epilepsy or receiving anticonvulsant therapy

Pregnancy or receiving oral contraceptive (estrogens and/or progesterone) or hormone replacement therapy

Oral or parenteral steroid therapy, including inhaled or nasal-delivery preparations, during the previous 6 months

Acute or serious medical condition (e.g., malignancy, stroke, recent myocardial infarction)

Current or recent participation in another study or clinical trial and under followup for that study or clinical trial

Inclusion criteria

Age 25–65 years

In reasonable health

Approval for the study was obtained from the University of Manchester Ethics Committee and the appropriate local Research Ethics Committees in England. Written informed consent to participate in the study was obtained from all subjects.

Study sample size and sampling frame.

Subjects between the ages of 25 and 65 years were recruited from the population registers of individuals eligible to receive care from 3 primary care physicians in the UK, and the baseline pain and psychological screening questionnaire was sent to them. We have previously reported baseline data gathered from these subjects as part of a larger study of the cross-sectional relationship between HPA function and the presence of pain (9).

Data from the population-based Altrincham Pain Study indicated that of those who were at risk of future CWP based on their psychosocial profiles, ∼22% would develop new-onset CWP (3). Recruitment of 174 community-dwelling subjects at risk of future CWP would allow us to detect a 2-fold increased risk of developing new-onset CWP among subjects in the bottom third of the distribution of baseline post-dexamethasone cortisol levels as compared with those in the middle and top thirds, with 80% power and an error level of 5%. Pilot studies indicated that of the subjects who were mailed a baseline questionnaire, 80% would participate, and 89% of those subjects would be free of CWP and 4% would be at risk of future CWP. Allowing for subjects who did not wish to be contacted further and those who were ineligible to take part in the tests of hormone function, we estimated that we would need to initially mail the questionnaire to 11,000 subjects.

Ascertainment of pain and psychosocial status.

The questionnaire included a blank body manikin on which subjects were asked to indicate the site(s) of any pain lasting at least 24 hours during the previous month. CWP was classified according to the definition in the American College of Rheumatology criteria for fibromyalgia (8). To satisfy these criteria, subjects must have pain in contralateral body quadrants and axial pain, and this pain must have been present for at least 3 months. The questionnaire also included the following items to ascertain psychosocial status, including several scales we have previously demonstrated to be associated with the development of CWP (3).

General Health Questionnaire (GHQ).

The 12-item GHQ was originally developed as a screening instrument for mental disorders in the community (10). It has been widely used in population-based studies as a measure of psychological distress. Each item response is scored as 0 or 1. Total scores range from 0 to 12, with higher scores indicating higher levels of psychological distress.

Illness Attitude Scales (IAS).

The 9 scales of the IAS individually measure worries about health, concern about pain, hypochondriacal beliefs, health habits, bodily preoccupation, fear of dying, disease phobia, treatment experience, and effect of symptoms (11). Based on the scores generated, 2 subscales have been identified: Health Anxiety and Illness Behavior (12).

Hospital Anxiety and Depression Scale (HADS).

The HADS was originally developed for use in patients with physical illnesses (13), but is widely used in population-based studies. The 14 items, each of which is scored on a scale of 0–3, measure degrees of anxiety (7 items) and depression (7 items). The 2 subscale scores range from 0 to 21, with higher scores indicating an increased likelihood that an anxiety or depressive disorder is present.

Sleep Problem Scale.

The Sleep Problem Scale contains 4 items that examine recent problems with sleep (14). Responses are scored from 0 to 5, giving a total score of 0–20. Higher scores indicate increased sleep disturbance.

Classification of subjects at risk of future CWP.

To examine the study hypotheses, all subjects who returned the baseline questionnaire were classified according to their pain reports and psychosocial status. Those who did not have CWP but were at risk for its development were identified. The baseline pain questionnaire included sections to assess aspects of somatization and health-seeking behavior, as measured by the Somatic Symptom Checklist and the Illness Behavior subscale of the IAS, respectively. These 2 scales were chosen because they predicted the onset of new CWP in a previous population-based prospective study (3). For the current study, a score of 4 or more on the Illness Behavior subscale of the IAS, in combination with a score of 1 or more on the Somatic Symptom Checklist (see below) constituted the criteria for characterizing subjects as being at risk of developing CWP. These subjects were eligible to participate further in an assessment of their HPA axis function.

Of the 11,000 subjects who were mailed a baseline questionnaire, 7,784 completed the questionnaire, representing a participation rate of 78% after adjustment for errors in the sampling frame. Of the respondents, a total of 1,450 were at risk of future CWP and agreed to accept further contact by the study team. After exclusions were applied (Table 1), 768 subjects were eligible to participate in the assessment of HPA axis function and formed the sampling frame for the study.

Assessment of HPA axis function.

Four measures of HPA axis function were selected as being appropriate for use in large samples of community-dwelling subjects: serum cortisol levels in response to an acute stressor (a pain-threshold examination), serum cortisol levels in response to a low-dose dexamethasone suppression test, and salivary cortisol levels obtained in the morning and in the evening to assess diurnal HPA tone. All tests were performed and all samples were collected in the subjects' homes or at the offices of their local general practitioners. The rationale for using these tests and sample collection methods is described in detail in Appendix A. To control for variations in HPA function during different phases of the menstrual cycle, all female subjects were studied during the early follicular phase.

Hormone assays.

Plasma cortisol levels were determined using the Chiron Diagnostics ACS:180 analyzer (Bayer HealthCare Diagnostics, Tarrytown, NY). The assay is a competitive chemiluminescence immunoassay. Cortisol in the sample competes with cortisol labeled with acridinium ester for a limited amount of polyclonal rabbit anticortisol antibody coupled to paramagnetic particles. The cortisol concentration in the sample is inversely proportional to the relative light units detected in the ACS:180 system.

A modified version of the serum cortisol radioimmunoassay method was used to measure salivary cortisol levels. This method depends on competition between cortisol present in the sample or standard and ¹²⁵I-labeled cortisol for a limited number of binding sites on rabbit anticortisol antibody. Separation of the antibody-bound fraction is effected by incubation with donkey anti-rabbit antibody coated to cellulose particles, followed by centrifugation and decantation of the supernatant. The radioactivity in the pellet is then counted. The amount of bound tracer is inversely proportional to the concentration of cortisol present. Before assay, the saliva sample was centrifuged for 5 minutes at 2,500 revolutions per minute, and the supernatant was removed for assay.

Followup phase of study.

Fifteen months after completion of the baseline questionnaire, all subjects were mailed a second questionnaire that included a blank body manikin identical to that in the baseline questionnaire. Subjects were again asked to indicate the site(s) of any pain they had experienced for 1 day or longer during the previous month and to indicate whether that pain had been present for 3 months or longer. New-onset CWP was defined according to the same criteria used at baseline and was assessed independently, with the assessor blinded to the data from the baseline questionnaires and HPA axis function studies. The followup questionnaire included all of the baseline psychosocial questionnaires plus the List of Threatening Life Experiences (15) to examine the role of stressful life events that may have had an impact on symptom reporting. This 12-event inventory explores recent adverse experiences in the subject's personal relationships, employment, illness, and financial and legal issues during the previous 6 months.

Statistical analysis.

The raw data from the cortisol determinations in each of the 4 assessment methods were separately compared between subjects with and those without new-onset CWP at followup. To quantify the strength of the relationship between risk factor exposures and the onset of CWP, we used logistic regression with baseline cortisol levels categorized into tertiles based upon the distribution of cortisol concentrations. To examine the relationship between morning saliva levels and the onset of CWP, the highest tertile was classified as the reference category, while for all other relationships, the lowest tertile was classified as the reference category. The relationship between new-onset CWP versus no CWP and cortisol levels was expressed as the odds ratio (OR), and 95% confidence intervals (95% CIs) were calculated. Baseline psychosocial exposures were then categorized into tertiles based upon the distribution of scores, with the lowest tertile classified as the reference category, and the relationship to cortisol concentrations was examined using logistic regression. To identify factors that independently contributed to the onset of CWP, we constructed a multivariate logistic regression model into which all measures of HPA function were entered. To examine the hypothesis that these relationships would be independent of the effects of any concomitant depression or other psychosocial factors, this analysis was adjusted for psychosocial factors that were found to be associated with cortisol concentrations. All analyses were adjusted for the potential confounding effects of age and sex. Finally, to assess the fit of the model, we examined the odds of developing new-onset CWP according to the number of HPA factors the subjects were “exposed to” at baseline.

RESULTS

Response and participation rates.

Of the 768 subjects who formed the sampling frame for the current study, a random sample of 463 subjects were contacted and asked to undergo HPA axis function assessment; 267 subjects agreed to participate (57.7%), of whom 254 subjects provided complete baseline data. The group of subjects who agreed to participate did not differ significantly from the source population with respect to age (median age 45.1 years [95% CI 43.5–46.7] in the source population and 47.3 years [95% CI 45.9–48.7] in the participants) or sex (proportion female 63.0% [95% CI 59.5–65.4] in the source population and 67.4% [95% CI 61.7–73.2] in the participants). Participation at followup 15 months later was high, with a total of 241 of the original participants (94.9%) providing information on pain status. Subjects who participated were more likely to be older (median age 48.1 years [95% CI 46.7–49.4] in the participants and 36.6 years [95% CI 31.8–41.3] in the nonparticipants; P = 0.0001) and female (47.1 [95% CI 23.0–72.2] of the participants and 68.9 [95% CI 63.0–74.7] of the nonparticipants; P not significant).

Prevalence of new-onset CWP at followup.

Of the 241 subjects who participated, 28 (11.6%) developed new-onset CWP. Subjects with new-onset CWP were older than those who remained free of CWP (median age 51.9 years [95% CI 47.9–55.8] versus 47.6 years [95% CI 46.2–49.0]; P = 0.04), although there was no difference with respect to sex (proportion female 67.9% [95% CI 47.6–84.1] versus 69.0 [95% CI 62.8–75.2]; P = 0.90).

Influence of baseline HPA axis function on the risk of new-onset CWP.

Baseline post–physical examination cortisol levels.

Baseline post–physical examination cortisol levels ranged from 73 nmoles/liter to 1,102 nmoles/liter. Although, as shown in Figure 1A, subjects reporting new CWP at followup had higher baseline median values (median 337.5 [95% CI 235–368]) than did those not reporting new CWP (median 304 [95% CI 287–322]), this difference was not statistically significant (P = 0.62 by Mann-Whitney U test). Results from the logistic regression model (Table 2) demonstrated that compared with subjects who scored in the lowest third of the cortisol distribution, those in the middle and highest thirds did not have an increased risk of developing CWP.

Figure 1. Baseline hormonal factors stratified according to chronic widespread pain status at followup. Hypothalamic–pituitary–adrenal axis function was determined by measuring A, serum cortisol levels in response to an acute stressor (a pain-threshold examination), B, serum cortisol levels in response to a low-dose dexamethasone suppression test, C, salivary cortisol levels obtained in the morning, and D, salivary cortisol levels obtained in the evening. Data are shown as box and whisker plots. Lines inside the boxes represent the median. Boxes represent the interquartile range. Whiskers represent the highest and lowest values. Solid circles represent outliers.

Table 2. Relationship between baseline hormonal factors and the onset of CWP*

Cortisol concentration	No. of subjects	Final pain status		Odds ratio (95% CI)†
		No. (%) with CWP	No. (%) without CWP
* Numbers of subjects do not total 241 because of exclusion of subjects with missing or incomplete data. CWP = chronic widespread pain; 95% CI = 95% confidence interval. † Adjusted for age and sex.
Postexamination, nmoles/liter
73–260	80	9 (11.3)	71 (88.8)	1 (reference)
261–365	83	9 (10.9)	74 (89.2)	0.87 (0.31–2.40)
366–1,102	77	10 (13.0)	67 (87.0)	1.05 (0.39–2.80)
Post-dexamethasone, nmoles/liter
30–179	78	5 (6.4)	73 (93.6)	1 (reference)
183–324	80	10 (12.5)	70 (87.5)	2.87 (0.89–9.22)
326–810	80	13 (16.3)	67 (83.8)	3.53 (1.17–10.65)
Salivary, nmoles/liter
Morning
8–141	62	7 (11.3)	55 (88.7)	1 (reference)
4–7	95	9 (9.5)	86 (90.5)	0.94 (0.32–2.70)
<1–3	82	12 (14.6)	70 (85.4)	1.43 (0.52–3.94)
Evening
<1	124	116 (93.6)	8 (6.4)	1 (reference)
1–2	88	72 (81.8)	16 (18.2)	3.39 (1.37–8.42)
3–18	27	23 (85.2)	4 (14.8)	2.32 (0.64–8.42)

Baseline post-dexamethasone cortisol levels.

Baseline post-dexamethasone cortisol levels ranged from 30 nmoles/liter to 810 nmoles/liter. Median levels were higher in subjects with new-onset CWP (287.5 nmoles/liter [95% CI 219–411]) as compared with those in subjects who had not developed CWP at followup (223.0 nmoles/liter [95% CI 202–261]) (Figure 1B), although this difference was not statistically significant (P = 0.11 by Mann-Whitney U test). Nevertheless, after adjusting for the effects of age and sex, post-dexamethasone cortisol levels were found to be strong predictors of CWP onset: subjects who scored in the middle third had an almost 3-fold increased risk as compared with those who scored in the lowest third (OR 2.87, 95% CI 0.89–9.22), whereas subjects who scored in the highest third of cortisol distribution had a >3-fold increased risk (OR 3.53, 95% CI 1.17–10.65).

Baseline morning and evening salivary cortisol levels.

Compared with serum cortisol levels, the range of both morning and evening salivary levels was narrow (<1–141 nmoles/liter and <1–18 nmoles/liter, respectively). When contrasted with those who did not report new-onset CWP, subjects with CWP had lower median morning levels of cortisol (4.5 nmoles/liter [95% CI 3–7] versus 5 nmoles/liter [95% CI 4–5.5]) (Figure 1C) and higher median evening levels of cortisol (1 nmole/liter [95% CI 0.7–1] versus 0.5 nmoles/liter [95% CI 0.5–1]) (Figure 1D). Logistic regression analysis revealed that both morning and evening salivary cortisol levels were associated with an increased risk of the new onset of CWP. Thus, subjects whose morning saliva values were in the lowest third were nearly 1.5 times more likely to report symptoms, while those whose evening saliva values were in the middle and highest thirds were more than twice as likely to report symptoms.

Influence of baseline psychosocial factors and pain status on the risk of new-onset CWP.

The relationship between psychosocial factors measured at baseline and the new onset of CWP among this psychologically distressed group of subjects was then examined (Table 3). High levels of psychological distress, as indicated by a higher GHQ score, higher HADS depression subscale score, higher levels of sleep disturbance, having experienced 2 or more threatening life events in the 6 months prior to the followup survey, and higher Illness Behavior subscale score, were all associated with a moderately, although not significantly, increased risk of CWP. In addition, compared with subjects who were pain-free at baseline, those who reported regional pain at baseline had a 3-fold increased risk of new-onset CWP at followup.

Table 3. Relationship between baseline scores for assessments of exposure to psychosocial factors, pain status, and the onset of CWP*

Assessment, range of scores	No. of subjects	Final pain status		Odds ratio (95% CI)†
		No. (%) with CWP	No. (%) without CWP
* Numbers of subjects for some categories do not total 241 because of exclusion of subjects with missing or incomplete data. CWP = chronic widespread pain; 95% CI = 95% confidence interval; GHQ = General Health Questionnaire; HADS = Hospital Anxiety and Depression Scale; IAS = Illness Attitude Scales. † Adjusted for age and sex.
GHQ
0	107	12 (11.2)	95 (88.8)	1 (reference)
1–4	70	5 (7.1)	65 (92.9)	0.63 (0.21–1.89)
5–12	61	10 (16.4)	51 (83.6)	1.68 (0.67–4.22)
HADS depression subscale
0–2	99	11 (11.1)	71 (88.9)	1 (reference)
3–5	73	11 (15.1)	65 (84.9)	1.59 (0.63–4.00)
6–21	68	5 (7.4)	69 (92.3)	0.63 (0.21–1.92)
HADS anxiety subscale
0–4	71	9 (12.7)	62 (87.3)	1 (reference)
5–8	102	9 (8.8)	93 (91.2)	0.73 (0.27–1.99)
9–21	67	9 (13.4)	58 (86.6)	1.09 (0.40–2.95)
Sleep Problem Scale
0–4	76	7 (9.2)	69 (90.8)	1 (reference)
5–10	92	12 (13.0)	80 (87.0)	1.42 (0.52–3.87)
11–20	67	9 (13.4)	58 (86.6)	1.46 (0.51–4.23)
List of Threatening Life Experiences
0	103	10 (9.7)	93 (90.3)	1 (reference)
1	64	6 (9.4)	58 (90.6)	0.86 (0.29–2.54)
2–6	72	12 (16.7)	60 (82.3)	1.69 (0.68–4.22)
IAS Illness Behavior subscale
5	94	8 (8.5)	86 (91.5)	1 (reference)
6–8	74	14 (18.9)	60 (81.1)	2.38 (0.93–6.13)
9–18	71	6 (8.5)	65 (91.5)	0.85 (0.28–2.62)
IAS Health Anxiety subscale
0–7	83	10 (12.1)	73 (87.9)	1 (reference)
8–13	77	5 (5.0)	72 (93.5)	0.51 (0.17–1.59)
14–39	74	12 (16.2)	62 (83.8)	1.31 (0.52–3.31)
Somatic Symptom Checklist
2	146	16 (10.9)	130 (89.1)	1 (reference)
3–5	67	7 (10.5)	60 (89.5)	0.96 (0.37–2.48)
Pain status
No pain	60	3 (5.0)	57 (95.0)	1 (reference)
Regional pain	181	25 (13.8)	156 (86.2)	2.67 (0.77–9.30)

Since it is likely that a number of these psychosocial measures would be associated with each other, we examined the strength of correlation. For factors that were highly correlated (r ≥ 0.5), 1 assessment was excluded from further analysis. Thus, the GHQ score (correlated with the HADS depression subscale score [0.58] and HADS anxiety subscale sore [0.54]) and the HADS anxiety subscale score (correlated with the IAS Health Anxiety subscale score [0.50]) were excluded from further analysis. The remaining variables were then considered as potential confounders for subsequent analyses.

Independent predictors of new-onset CWP.

To identify measures of HPA function that were independently associated with an increased risk of developing CWP, we constructed a multivariate logistic regression model (Table 4). Of the baseline hormonal factors, higher post-dexamethasone serum cortisol, lower morning salivary cortisol, and higher evening salivary cortisol values were all found to be independent predictors, after adjusting for the effects of the concomitant psychosocial factors.

Table 4. Independent hypothalamic–pituitary–adrenal axis function predictors of new-onset CWP, by multivariate logistic regression model*

Cortisol concentration	No. of subjects	Final pain status		Odds ratio (95% CI)†
		No. (%) with CWP	No. (%) without CWP
* Numbers of subjects do not total 241 because of exclusion of subjects with missing or incomplete data. CWP = chronic widespread pain; 95% CI = 95% confidence interval. † Adjusted for age, sex, baseline pain status, life events, illness behavior, and Hospital Anxiety and Depression Scale depression subscale score.
Post-dexamethasone, nmoles/liter
30–179	77	5 (6.6)	71 (93.4)	1 (reference)
183–324	75	10 (13.5)	64 (86.5)	2.82 (0.85–9.42)
326–810	75	10 (13.5)	64 (86.5)	3.82 (1.19–12.22)
Salivary, nmoles/liter
Morning
8–141	63	7 (11.3)	55 (88.7)	1 (reference)
4–7	87	9 (9.5)	86 (90.5)	1.14 (0.38–3.46)
<1–3	80	12 (14.6)	70 (85.4)	2.18 (0.74–6.45)
Evening
<1	124	116 (93.6)	8 (6.4)	1 (reference)
1–2	88	72 (81.8)	16 (18.2)	3.52 (1.38–8.97)
3–80	27	23 (85.2)	4 (14.8)	2.01 (0.53–7.59)

The risk of new-onset CWP was then examined in relation to these factors acting in combination. For this analysis, we collapsed the categories of cortisol concentrations where an increased risk that was similar in size of effect was observed. This allowed increased power to identify relationships between the cortisol measures and outcome. Thus, subjects in the lowest third of post-dexamethasone serum cortisol and evening salivary cortisol levels were categorized as “low,” and those in the middle or highest third were categorized as “high.” Subjects in the highest and middle thirds for morning salivary cortisol levels were categorized as “low,” while those in the lowest third were categorized as “high.” As shown in Table 5, compared with subjects who had none of these factors, those who had 2 of these 3 factors were twice as likely to develop new-onset CWP, while those who had all 3 factors were almost 9 times more likely to develop new-onset CWP. Indeed, 1 or more of these HPA factors identified 26 cases of new-onset CWP (92.9%).

Table 5. Prevalence and odds of developing CWP, based on post-dexamethasone, evening salivary, and morning salivary cortisol levels*

No. of risk factors	No. of subjects with risk factor(s)	No. (%) with CWP	No. (%) without CWP	Odds ratio (95% CI)†
* Risk factors were as follows: a post-dexamethasone cortisol value ≥183 nmoles/liter, an evening salivary cortisol value ≥1 nmoles/liter, and a morning salivary cortisol value of ≤2 nmoles/liter. CWP = chronic widespread pain; 95% CI = 95% confidence interval. † Adjusted for age and sex.
None	27	2 (7.4)	25 (92.6)	1 (reference)
1	90	5 (5.6)	85 (94.4)	0.8 (0.1–4.2)
2	95	13 (13.7)	82 (86.3)	2.3 (0.5–11.2)
3	24	8 (33.3)	16 (66.7)	8.5 (1.5–47.9)

DISCUSSION

In the current study, we demonstrated for the first time that in persons who were free of CWP at baseline but were at risk of the future development of CWP, alterations in HPA axis function were associated with an increased risk of new-onset CWP. Subjects with higher post-dexamethasone serum cortisol levels, which indicate a failure to suppress the HPA axis, and lower morning and higher evening salivary cortisol levels, which indicate a blunting of the diurnal rhythm, were at increased risk of new-onset CWP. These relationships were independent of the effect of coexistent depressive symptoms and other psychosocial factors or sleep disturbance that may affect the function of the HPA axis. High scores on the Illness Behavior subscale of the IAS (OR 2.1, 95% CI 0.7–6.2) and reporting recent Threatening Life Experiences (OR 2.1, 95% CI 0.7–6.3) remained independent predictors of symptom onset. Similarly, these relationships were not simply a reflection of having some pain, since the relationship persisted after adjustment for pain status at the time of the baseline survey. These data therefore suggest that chronic, and presumed unexplained, pain has both psychological and physical antecedents.

This study was a population-based prospective study that was free of a number of biases inherent in clinic-based cross-sectional studies. A very small proportion (7%) of subjects who participated in the baseline assessment of HPA axis function did not take part in the followup study 15 months later. Nevertheless, we assessed the possible impact of this nonparticipation by examining baseline levels of cortisol, psychosocial factors, pain status, and tender point count in the participants as compared with the nonparticipants. We found no significant differences in any of these baseline measures (data not shown). HPA function is difficult to assess in the community, and a number of issues related to our measurements may have influenced our results. First, for the measures we chose, a number of subjects had to be excluded because of extraneous factors that might influence the results. There may have been other factors contributing to errors in assessing the HPA axis, although they should have acted to attenuate any association we found.

Second, we used a low dose of dexamethasone. There is concordance between different doses of dexamethasone, but the low dose results in measurable cortisol in the great majority of subjects, and thus allows for sensible analysis. Higher doses result in undetectable cortisol levels in most measurements (16). We have previously shown that results of the low-dose dexamethasone suppression test are associated with genetic markers on the glucocorticoid receptor gene (17).

Third, the measurements of salivary cortisol may be subject to differences due to timing of sample collection. However, the evening cortisol plateau is stable over many hours and is unlikely to be influenced by small differences in the timing of the sample. The morning measurements are more vulnerable to such effects, but there is nothing to suggest that those who did and those who did not develop new-onset CWP would sample at different times; thus, the net effect would be to increase signal noise, rather than to induce an effect. In addition, the increase in evening levels of salivary cortisol would be predicted to be accompanied by a decrease in morning levels, which is what we observed.

Fourth, as a consequence of our obtaining only 1 measurement 20 minutes after the examination, we may not have observed a relationship between post–tender point examination cortisol concentrations and new-onset CWP. Thus, we may not have adequately captured the response profile.

Fifth, we did not exclude subjects on the basis of major endocrine disorders, including Addison's disease or Cushing's syndrome. Both are very rare, and it is highly unlikely, based on their reported incidence in the population, that any subjects had these disorders. It is important that none of the subjects had unequivocally high cortisol levels so as to make the diagnosis of Cushing's syndrome appear to be likely. The distribution of post-dexamethasone cortisol concentrations in this study population was similar to that reported in the general population in our previous study (17).

Finally, due to the age range of subjects included in the study, it was possible that a number of postmenopausal women were recruited, and their menopause status may have confounded the relationship between HPA function and new-onset CWP. Indeed this was the case. Of the 166 female subjects, 89 (53.6%) reported having experienced menopause. However, menopause status was not associated with the onset of new CWP (OR 0.99 [95% CI 0.3–3.6], adjusted for age) and did not attenuate the relationship between HPA function and new-onset CWP.

It is important to consider whether there is a coherent biologic mechanism linking HPA axis function, psychological distress, and CWP. On activation of the HPA axis, a cascade of cause and effect begins with the release of CRH from the hypothalamus into the hypothalamic–hypophyseal portal system. Elevated levels of CRH affect the release of adrenocorticotropic hormone (ACTH) from the anterior pituitary gland, which, in turn, stimulates the adrenal glands to release cortisol. Variability in this neuroendocrine response may result from a number of factors, including stressors experienced during childhood and adulthood as well as altered central nervous system availability of certain neurotransmitters. The latter condition includes both lower levels of serotonin, a neurotransmitter involved in pain modulation and regulation of the HPA axis (18), and higher levels of substance P, which acts to convey information about the severity, duration, and location of pain (19). In particular, serotonin is associated with the stimulation of CRH and ACTH during stress. CRH has been shown to act at all levels of the neuraxis to produce analgesia. Crofford (20) hypothesized that such variability in neuroendocrine responses confers a vulnerability to stress-related syndromes that may be precipitated by “trigger” events, including high levels of psychological distress.

While there are no comparable population-based studies available, the present findings are supported by previous reports of HPA axis dysfunction among clinic patients with fibromyalgia. Such patients have been shown to display a number of abnormalities of HPA function, including abnormal suppression of the HPA axis after administration of dexamethasone (21, 22), reduced 24-hour free cortisol excretion but loss of the circadian fluctuation in glucocorticoid levels, with elevated levels during the circadian trough (23), exaggerated ACTH but blunted cortisol response to exogenous administration of human CRH and to insulin-induced hypoglycemia (7), and loss of circadian fluctuation of plasma cortisol levels because of elevated evening levels, with reduced overall 24-hour excretion of urinary free cortisol (5, 24). Griep et al (6) demonstrated that compared with sedentary and pain-free control subjects, patients with fibromyalgia displayed dysfunction of the HPA axis, whereas patients with less widespread pain (noninflammatory low back pain) displayed similar dysfunction, but it was less marked as compared with that in the fibromyalgia group.

Conclusions from these studies support the concept of disturbed HPA axis function in subjects with chronic pain, and specifically, with fibromyalgia. The focus would appear to be functional deficits in CRH release from the hypothalamus. Enhanced pituitary ACTH responses to ovine CRH, however, may be due to up-regulation of pituitary CRH receptors in response to chronically reduced central CRH release. Central CRH tone itself therefore appears to lie at the heart of the interrelationships between stress stimuli and response adaptation. The current results further elucidate their temporal relationships, having determined that HPA dysfunction precedes the onset of symptoms.

Our hypothesis was that the influence of psychosocial factors known to increase the risk of the onset of CWP would be moderated through the presence of altered HPA axis function. One could argue that among a group of distressed subjects, those with altered HPA axis function are even more distressed, that is, we have further refined the “distressed phenotype.” Assuming that to be true, it would not be surprising to observe an increased risk of CWP onset in persons with markers of distress plus altered HPA function, the latter being an indicator of higher levels of distress. However, our analysis would indicate that this is not the case. Thus, although the association between HPA axis dysfunction and the subsequent development of CWP was attenuated after adjustment for psychosocial factors, the relationship persisted.

The other explanation is that the observed HPA dysfunction is a consequence of being distressed and is therefore a path variable to the development of CWP. Although that is a possibility, it is likely that alterations of the HPA axis take place before the “trigger” event that is associated with the onset of CWP occurs. Evidence indicates that HPA function is likely to be compromised in early childhood (25) by adverse experiences, including childhood abuse and parental loss (26), and that prenatal factors may also play a role (27). Polymorphisms of genes related to serotonin metabolism have also been studied in relation to fibromyalgia. The serotonin transporter gene (28) and the serotonin receptor gene (28) have both been implicated, although these findings are equivocal (29, 30).

We would argue that the current data support our hypothesis that psychosocial factors are moderated through the presence of HPA axis dysfunction, leading to the onset of CWP. We propose the following model of the onset of CWP (see Figure 2): stressful events in early childhood (25), genetic factors (28), decreased serotonin and increased substance P availability (31), and genetic factors that directly influence the functioning of the HPA axis. The subsequent HPA axis dysfunction confers an increased vulnerability to the development of physical symptoms (20). Subsequent “trigger events,” including psychosocial factors such as high levels of psychological distress, are moderated through the HPA axis and lead to, among other symptoms, CWP. We have examined the role of HPA function in symptom onset. Future studies would further explore this model by, for example, examining the role of early life events, genetic factors, and other neuroendocrine markers of stress response with HPA function and symptom onset. We have recently demonstrated that CWP shares common associated factors with other disorders thought to be functional in origin, including chronic orofacial pain, irritable bowel syndrome, and chronic fatigue (32), and the current findings may suggest a true pathogenic role of HPA changes in other functional disorders, including chronic fatigue syndrome.

Figure 2. Hypothesized model of the relationship between psychological distress, hypothalamic–pituitary–adrenal (HPA) axis dysfunction, and chronic widespread pain. Under this model, adverse early childhood factors, genetic factors, and associated neurotransmitter alterations, as well as other factors, are associated with HPA axis dysfunction. Psychological distress in adulthood is moderated through the dysfunctional HPA axis, leading to the onset of symptoms including chronic widespread pain. CRH = corticotropin-releasing hormone; ACTH = adrenocorticotropic hormone.

In summary, therefore, among a group of psychologically at-risk subjects, aspects of the HPA axis function help to distinguish those who will from those who will not develop new-onset CWP. Symptoms often labeled as “unexplained,” including CWP, are recognized as having a major underlying psychological component. The findings of this study do not refute this, but provide evidence that this relationship can have a definite neurobiologic basis.

AUTHOR CONTRIBUTIONS

All authors had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Study design. Drs. McBeth, Silman, Gupta, Ray, Morriss, Dickens, and Macfarlane.

Acquisition of data. Drs. McBeth, Gupta, Chiu, Ray, Ms King, and Dr. Macfarlane.

Analysis and interpretation of data. Drs. McBeth, Silman, Gupta, Chiu, Ray, Morriss, Dickens, and Macfarlane.

Manuscript preparation. Drs. McBeth, Silman, Chiu, Ray, Morriss, Dickens, Ms King, and Dr. Macfarlane.

Statistical analysis. Drs. McBeth, Gupta, and Macfarlane.

ROLE OF THE STUDY SPONSOR

The study sponsor had no role in study design; in the collection, analysis, and interpretation of data; in the writing of the report; or in the decision to submit the manuscript for publication.

Acknowledgements

The authors are grateful for the participation and help of the doctors, staff, and patients of the 3 general practices in greater Manchester, and to Joanne Bradley and Karen Schafheutle for survey administration.

APPENDIX A

HPA FUNCTION TESTS AND SAMPLE COLLECTION METHODS

Determination of serum cortisol levels after pain-threshold examination.

Although there is no precedent for the collection and interpretation of cortisol levels following a pain-threshold examination, we hypothesized that serum cortisol levels collected in this manner would represent the HPA axis response to an acute stressor. Pain threshold was measured using a Fischer pressure threshold meter (1) at 8 body sites bilaterally: the anterior border of the tibia, the medial fat pad of the knee, the gluteus maximus, the medial supraspinatus, the deltoid, the lateral epicondyle, the dorsum of the forearm, and the trapezius. The pressure threshold meter is a force gauge calibrated in kg/cm², with a range of 0–20 kg. Pressure was applied at the predefined sites at an increasing rate of 1 kg/second. Subjects were asked to indicate the point at which a pressure sensation became a painful sensation by saying “now”; at that point, the pressure that had been applied was read from the meter and recorded. The examination lasted a total of 10 minutes, and a blood sample was taken 30 minutes after the start of the procedure.

Overnight low-dose dexamethasone suppression test.

Lower levels of ACTH are released by the HPA axis in response to increased circulating blood levels of glucocorticoids, which results in lower levels of cortisol production by the adrenal gland. The integrity of this feedback mechanism can be tested by administering a low dose of dexamethasone, a potent glucocorticoid, in the evening and judging the level of ACTH secretion in the morning by analysis of plasma cortisol levels (2). The morning level of plasma cortisol in normal subjects will be <140 nmoles/liter (2). In the current study, we used 0.25 mg of dexamethasone (3, 4). There is concordance between different doses of dexamethasone, but the low dose results in measurable cortisol in the great majority of subjects and so allows for a sensible analysis. Higher doses result in undetectable levels of cortisol in most measurements (5). This is a standard research investigation. Dexamethasone was administered at 10:00 PM (4), and a blood sample was taken by a research nurse at 8:00 AM.

Diurnal salivary cortisol levels.

Salivary cortisol determination affords a convenient method of assessment that provides a valid and reliable correlate of serum/plasma free diurnal cortisol levels (6, 7). Salivary cortisol analysis has several advantages over blood cortisol analysis. Since it is noninvasive, and therefore sampling is stress-free, it is easy to administer, which is especially advantageous in studies of large numbers of subjects, and it minimizes disruption of participants' lives. Subjects were supplied with a standardized protocol detailing the sample collection schedule. Saliva samples were obtained by placing a swab in the mouth for 90 seconds to soak up saliva; 1 swab was obtained in the evening (at 10:00 PM) and 1 swab was obtained the following morning (between 8:00 and 9:00 AM). The swabs were immediately placed in the freezer until they were collected by a research nurse. The research nurse stressed the importance of the timing of the samples, and when the research nurse collected the swabs, the subjects were asked to confirm the timing of their samples.