Sixteen studies (18 articles) met the inclusion criteria. The study sample characteristics are summarized in Table 1. Nine studies included patients with RA 2930343536373839404142 and seven patients with SLE 31333540434445. Two of these studies involved both types of patients 3540. In addition, one study included patients with JIA 46 and one study included a heterogeneous group of patients with inflammatory arthritis (RA, psoriatic arthritis, ankylosing spondylitis and fibrositis) 32. Fifteen studies included healthy participants without systemic inflammatory rheumatic disease as control and one study, patients with osteoarthritis 3637. In three studies a second control group was added, consisting of either patients with myofascial pain 32, patients with sarcoidosis 43, or participants taking corticosteroids 44. Most studies have relatively small sample sizes (N = 10 to 20 per group), with even smaller sample sizes (N <10) in two studies 3540. In addition, patient samples are probably heterogeneous because, for example, strict exclusion criteria regarding comorbidities are lacking in half of all studies 3233353842434446. With regard to diagnostic criteria, one study even included also patients with non-inflammatory arthritis 32. There is also a high variability in pharmacotherapy profiles, but overall these profiles were well-defined for the various classes of drugs (see Table 1), except in two studies 3233. To control for the effects of pharmacotherapy some studies excluded patients using specific medication (for example, corticosteroids 3538394046, opioids 3941, antidepressants 394045, or adrenoceptor antagonists 294045), or the intake of medication was regulated prior to and on the day of testing 3941. In addition, two studies performed subgroup analyses with patients on different medical regimes 4145.

Ten of 16 studies used a psychological stressor, applied for one minute to two hours 293031323637384041434445.

Three studies used exercise stressors, namely, a bicycle ergometer task, which involved cycling at a rate of 50 revolutions per minute against a resistance of 12.5 watt for three minutes 38; or an ergospirometric exercise test on a treadmill to the limit of subject's tolerance for several minutes 3435.

Eight studies assessed sensory stressors, most of which used thermal stimuli, usually experienced as painful. Five studies assessed responses to a cold stressor 3133383946, in which participants either had to put their hand in a bowl filled with (ice) cold water for several minutes 383946, or endure an ice-pack on their hand 31. One study did not describe the cold pressor task 33. Heat pulses 39, an acoustic stress test (several minutes) 4243 and a pupillary light flash test 32 were also used.

Stress-induced self-reported distress was assessed in only two of 16 studies 4145. Subjective distress levels were either measured on a visual analogue scale (VAS) ranging from 0 to 100 41 or on a VAS tension and excitement together with the State Trait Anxiety Inventory (STAI) 45.

Eleven studies assessed ANS responses (see Table 2) 2931323337404142434546, namely, heart rate 293132374145, diastolic and/or systolic blood pressure 2931334145, mean arterial pressure, systemic vascular resistance, plasma volume and cardiac output 3637, and pre-ejection period (PEP) 41. Furthermore, plasma (nor)epinephrine levels 314042444546 and skin conductance 2932 were assessed.

Eight studies assessed neuroendocrine responses (see Table 3) 3034353839414244, namely plasma 303438414244 or serum 3539 cortisol levels and plasma ACTH levels 343841. Two studies measured hormones other than those involved in the HPA axis, namely, plasma growth hormone (GH) and insulin-like growth factor I (IGF-I) levels 34 and serum prolactin levels 35.

Twelve studies assessed immune responses (see Table 4) 303536383940414243444546, namely, the total number of leucocytes 364042434445, and the number of lymphocytes and/or subsets of lymphocytes, including B cells, T cells and NK cells 30354042434445. Two studies assessed NK cell cytotoxicity 4045. Six studies reported on cytokine levels 383940414246, namely, plasma levels of cytokine IL-1β 42, IL-6 394142 and TNF-α 39, ex vivo stimulated mononuclear cell production of IL-2 40, IL-4 3840, IL-6 404146, IL-8 46, IL-10 40, interferon (IFN)-γ 3840 and TNF-α 41, or intracellular concentrations of IL-4, IL-6, IL-10 and IFN-γ 40. Other inflammation markers were also assessed, such as C-reactive protein (CRP) 36, the number of β-adrenoceptors on monocytes 45, β-adrenoceptor sensitivity 45, and plasma soluble IL-2 receptor levels 42.

All studies reported baseline values for the outcome measures, and all, except two 3234, reported a resting period of 3 to 45 minutes. During administration of the experimental stressor, physiological reactivity was either assessed at specific time points (in three studies) 354142, or continuously throughout the stress procedure (only for autonomic measures such as heart rate, blood pressure and/or skin conductance; in four studies) 29323745. All studies reported immediate post stress measurements, except two studies that only assessed stress reactivity at 30 minutes 34 or 60 minutes 35 after cessation of the stressor. Eight of 16 studies have one 29303637384045 or more 394146 additional follow-up measurement points, ranging from 5 to 60 minutes after cessation of the stressor.

ANS function at baseline was not different between patients with rheumatic disorders and controls in most studies. Cardiovascular variables, skin conductance, and catecholamine levels did not differ in nine studies 29313637404142444546, whereas three studies reported significantly higher levels of autonomic activity at baseline (heart rate and pupil area 32; epinephrine levels 46) or lower levels ((nor)epinephrine 31) in patients with rheumatic disorders compared with controls (see Table 2).

Neuroendocrine function at baseline was not significantly different between patients with rheumatic disorders and controls. Five of eight studies reported similar baseline levels of cortisol, ACTH, GH, IGF-I, and prolactin 3435394142. Three studies reported higher levels of cortisol at baseline (in patients with RA) 3038 or lower levels (in patients with SLE) 44 than in controls (see Table 3). Lower cortisol levels at baseline were also observed in one study in which control subjects were taking corticosteroids 44.

Baseline leucocyte counts 364042 were not different between patients with RA and controls. However, two of three studies found patients with RA to have lymphopenia at baseline 3042. Levels of cytokines and other inflammatory factors (IL-2, IL-4, IL-6, IFN-γ, TNF-α, and CRP) were similar in patients with RA and controls in five studies 3638394041. Three studies reported higher baseline levels of IL-6 39 and soluble IL-2 receptor 42 and lower basal levels of IL-10 and IFN-γ 40 in patients with RA compared with controls.

Baseline leucocyte counts were not significantly different between patients with SLE and controls in most (three of four) studies 434445. One study reported baseline leucopenia in patients with SLE 40. All studies reported lower lymphocyte counts (including lower subsets of lymphocytes) in patients with SLE than in controls 404345. Despite this baseline lymphopenia, a higher percentage of B cells (from the total lymphocyte count) was found once 43. Only one study reported similar numbers of helper T and cytotoxic T cells in patients with SLE and controls 35. Cytokine (IL-2, IL-4 and IL-6) levels at baseline were not significantly different between patients with SLE and controls 40. However, cytokine levels of IL-10 and IFN-γ at baseline and the sensitivity of β-adrenoceptors, which are involved in the transduction of autonomic signals to immune cells, were lower in patients with SLE than in controls 45. The one study with immunological data on patients with JIA reported higher cytokine levels (IL-6 and IL-8) than in controls at baseline 46 (see Table 4).

In summary, at baseline autonomic function is similar in patients with RA and SLE and controls, with only limited evidence for heightened autonomic function in (a subgroup of) patients with rheumatic diseases. Neuroendocrine function at baseline is also comparable in patients with RA and SLE and controls, with only three studies reporting altered cortisol levels in patient groups. Specific immune variables, mainly (subsets of) lymphocyte counts, appear to be lower in patients with SLE than in controls.

Self-reported distress was increased significantly from baseline by psychosocial stress in patients with RA 41 and SLE 45 and in controls. This increase was greater in patients with RA 41 than in controls. This was not the case for patients with SLE 45.

The autonomic response to stress was assessed in patients with RA, SLE, JIA and in a heterogeneous group of patients with inflammatory arthritis. Results are summarized in Table 2. In patients with RA, experimental stress increased autonomic activity (heart rate 29363741, blood pressure 293741, systemic vascular resistance 37, pre-ejection period (PEP) 41, plasma volume 36, skin conductance 29, and plasma (nor)epinephrine levels 40), with the increase being similar to that seen in controls in most studies. However, three studies reported either diminished 29 or more pronounced 3741 autonomic responses in (a subgroup of) patients. In patients with SLE, experimental stress increased autonomic activity (heart rate 45, blood pressure 45, and (nor)epinephrine levels 404445), the increase often being similar to that seen in controls 31334045. Only one study observed a diminished autonomic response (heart rate), but only during one specific type of stressor (cold) 31. The one study involving patients with JIA showed experimental stress to increase norepinephrine levels to a similar extent in patients and healthy controls 46. In one study with a heterogeneous group of patients with inflammatory arthritis, experimental stress increased autonomic activity (heart rate, skin conductance, and pupillary constriction) 32, but this increase was smaller (pupillary constriction) or greater (galvanic response) than that of controls.

In summary, patients with rheumatic disorders respond to experimental stress with increased cardiovascular, galvanic and catecholamine responses. Whereas most autonomic responses to stress are similar to those of a control group, there is partial evidence (five studies) for altered stress-induced autonomic responses.

Neuroendocrine reactivity, mainly measured as changes in cortisol levels, was assessed in patients with RA and SLE and in controls (see Table 3). Experimental stress elicited both an increase 3941 and (more often) a decrease 34353842 in cortisol levels in patients with RA. ACTH, which is activated upstream of cortisol, increased in response to stress 3441. In addition to the assessment of HPA axis hormones, one study also reported an increase in GH levels in response to a stressor 34. The neuroendocrine response to stress of patients with RA was not significantly different from that of controls in most studies, but two studies reported that cortisol responses were diminished in patients with RA compared with controls 3538. In one study, experimental stress increased prolactin levels in controls but not in RA patients 35. The effect of stress on neuroendocrine function in patients with SLE was inconsistent, with stress eliciting a decrease 35 or no change 44 in cortisol levels. A lack of responsiveness was also reported in control subjects 3544. Experimental stress did not increase prolactin levels in patients with SLE 35.

In summary, although cortisol and ACTH levels change in response to experimental stress, differences between patients with rheumatic diseases (RA and SLE) and control groups have been reported in only two studies. There is preliminary evidence (one study) that the prolactin response to stress is different in patients than in controls.

Because several immune variables were used in the various studies to assess the effect of experimental stress on immune function, we have classified results on the basis of changes in leucocyte and lymphocyte counts and inflammatory markers (see Table 4).

Five studies investigated a cognitive stressor in patients with RA, SLE, or inflammatory arthritis 29303132363744. In all patient groups, cognitive stress consistently increased autonomic function: heart rate 293237, blood pressure 293637, skin conductance 2932, cardiac output and systemic vascular resistance 37, plasma volume 36, and catecholamines 44. Patients with RA and SLE had different cortisol responses to cognitive stress 3044 (see Table 5). Cognitive stress elicited changes in leucocyte counts, lymphocyte counts, subsets of lymphocytes, and CRP levels in patients with RA 3036, but only changes in leucocyte counts and cytoxic T cell numbers in patients with SLE 44.

A psychosocial stressor was used in three studies of patients with RA and SLE 404145. Psychosocial stress consistently increased autonomic activity (heart rate 4145, blood pressure 41, pre-ejection period 41 and catecholamine levels 4045) in patients with RA and SLE and increased neuroendocrine variables (cortisol and ACTH 41) in patients with RA. Some immune responses were similar in the two patient groups (leucocyte counts, lymphocyte counts, (cytotoxic) T cells, NK cells and NK cell cytoxicity 4045), whereas others were different (helper T cells 40, IL-4 40) (see Table 5). Cytokine levels were not influenced by psychosocial stress, with the exception of an increase in IL-4 in patients with SLE, an increase in TNF-α in patients with RA, and inconsistent findings for IL-6.

The stress of exercise was investigated in two studies with patients with RA and SLE 3435. These studies did not assess autonomic function; however, cortisol levels decreased in both groups of patients 3435 whereas ACTH levels and GH levels increased 34. Exercise stress elicited a different cytotoxic T cell response in patients with RA and SLE 35 (see Table 5).

Sensory stressors were used in six studies involving patients with SLE and JIA 313233394346. Sensory stress increased catecholamine levels 46 and cortisol levels 39. It also increased leucocyte counts, changed percentages of certain subsets of lymphocytes 43, and increased levels of IL-6 3946 and TNF-α 39 (see Table 5).

In summary, although few data are available about the effects of specific types of stress, results suggest that psychosocial, cognitive and sensory stress consistently increase autonomic activity. Changes in neuroendocrine variables in response to psychosocial, cognitive and sensory stressors are observed in patients with RA only. Psychosocial stress seems to elicit the strongest immune response in all patient groups studied.

Although previous studies that used regular autonomic function tests showed autonomic reactivity to be altered in different populations of patients with rheumatic diseases compared with controls 247484950, we failed to find consistent evidence for these alterations in studies investigating psychological, exercise, or sensory stress. The autonomic function of patients with rheumatic disorders was not only similar to that of controls at baseline but also after experimental stress, with cardiovascular, galvanic and catecholaminergic responses largely comparable to those of controls. Only a minority of studies reported autonomic function to be altered in these patients (as compared with controls) 29313237. Studies have shown that alterations in autonomic function in response to stress are correlated with disease severity 36374851, which suggests that only subgroups of patients with severe disease show distinct alterations in autonomic function. Changes in autonomic function might also be linked to physical deconditioning, vascular inflammation and accelerated atherosclerosis 52, which could explain the inconsistencies in the literature on this subject. It should also be noted that alterations in autonomic function might reflect a change in sympathetic or parasympathetic activity or in the balance of the two systems 5354. Autonomic measures used in experimental stress paradigms (for example, the heart rate response) do not always differentiate between these systems, in contrast to regular autonomic function tests 55, which specifically aim to distinguish between sympathetic dysfunction and parasympathetic neuropathy 495156.

The expected increase in neuroendocrine variables in response to a stressor was only observed in two studies with patients with RA 3941. This increase is consistent with previous studies reporting significant and consistent endocrine responses to psychosocial stress both in healthy subjects and in various patient populations without rheumatic disorders 24254157585960606162. The change in cortisol levels is consistent with the endocrine response to real-life anticipation stress before hip or knee surgery in patients with RA 6364. However, most studies evaluated in this review reported a decrease in cortisol levels after stress, both in patients and controls, which might reflect the normal diurnal rhythmic decline in cortisol levels instead of a stress response 34353842. All studies, except the two reporting an increase in cortisol levels, assessed patients in the early morning hours, when cortisol levels are known to decrease sharply.

Fuelling the ongoing debate on whether or not HPA axis responses are diminished in patients with rheumatic disease compared with healthy subjects, we did not find consistent changes in HPA axis function in response to psychological and exercise stress in patients with rheumatic diseases. Indeed, only two studies showed a different neuroendocrine response to stress in patients and controls 3538. Previous studies involving pharmacological stimulation of the HPA axis reported either similar or slightly lower responses in patients with rheumatic diseases compared with controls 2728. However, results should be interpreted with caution because stress manipulation may have been ineffective. Variations across studies in disease duration and activity, age, sex and sampling time might contribute to the variability in HPA axis responses in patients with rheumatic diseases. Moreover, attention should be paid to the role of pharmacotherapy (for example, corticosteroids), which could influence the endocrine response to stress 65.

Most studies reported that total leucocyte counts and specific subsets of lymphocytes (not total number per se) increased after stress in patients with rheumatic diseases. The most consistent finding was an increase in the number of NK cells, as reported in earlier studies showing that the most robust effect of acute time-limited stressors in healthy subjects is a marked increase in the number of NK cells, which suggest an activation of innate immunity 15. However, NK cell cytotoxicity after stress was lower in patients than in controls, possibly due to permanent activation as a consequence of the inflammatory disease 4045. The change in cytokine levels elicited by stress in patients with rheumatic disorders was small or inconsistent, for example, for the most frequently measured cytokine IL-6 404146. Previous semi-experimental studies with patients with RA reported ambiguous results, reporting either significantly increased 63 or decreased levels of IL-6 64 during anticipation of planned knee or hip arthroplasty. In addition, in a study evaluating responses to cryotherapy, patients with RA not treated with glucocorticoids responded with an increase in IL-6 levels 65. A recent meta-analysis showed IL-6 levels to be consistently increased after psychosocial stress in healthy individuals 16. The contradictory findings regarding cytokine responses in patients with rheumatic diseases might be due to methodological caveats, such as failed stress manipulation or small sample sizes (for example, in Jacobs et al. (2001)). However, it is also possible that cytokines do not effectively regulate immune function systemically, but instead near the effector site 40. Future studies involving patients with rheumatic diseases should measure local immune variables in samples obtained from joint tissue or synovial fluid 6667. Another explanation is that patients or subgroups of patients might have different cytokine responses (for example, TNF-α) because of differences in disease severity 68 or pharmacotherapy 4165. In general, as could be expected for immune-mediated diseases, the immune response to stress was different in patients and controls, and also differed between the various types of inflammatory rheumatic diseases.

Only the studies of psychosocial stress evaluated all three physiological stress response systems (autonomic, neuroendocrine and immune) and, moreover, consistently reported stress-induced increases at all three levels. This is in line with previous studies showing significant and consistent autonomic, neuroendocrine, and immune responses to psychosocial stress in both healthy subjects and patient groups 25415758596060616269. In contrast, cognitive stressors did not consistently alter neuroendocrine variables, and immune responses were small in patients with SLE. The stress of exercise was found not to activate the HPA axis in patients with rheumatic diseases, in contrast to what has previously been observed in healthy subjects 70, although studies suggest that neuroendocrine changes are only observed after exercise of longer duration 71. Data on immune responses were very limited. The studies that assessed sensory stress reported stress-induced increases in autonomic, neuroendocrine, and immune function, but data were limited. Because most sensory stressors are also frequently applied to induce pain 7273, future studies of stress should take into account whether a stressor is pain-inducing in order to unravel the relationship between stressors, pain and physiological responses 74. Overall, more research into the effects of psychosocial stress on physiological function is needed, especially because these stress paradigms seem to generate the strongest response at all three physiological levels and probably have the greatest ecological validity. Importantly, insight in physiological responses to acute stressors cannot automatically be translated to the field of daily stressors, major stress events or even chronic stressors 575. For example, studies in healthy and other disease populations suggest that chronic stress might induce hypocortisolism 676 and global immunosuppression of both innate and specific immunity 15. Overall, there is a lack of studies examining the association between physiological parameters and chronic stressors in rheumatic populations 2. Studies investigating stressors in daily life reported an association between stressors, mainly of an interpersonal nature, and concentrations of cytokines and lymphocytes (for example, IL-6, soluble IL-2 receptors and T cell numbers) 777879. However, results are inconclusive 80 and future studies are needed to elucidate how acute and chronic stressors differentially affect inflammatory rheumatic diseases.

In contrast to autonomic and neuroendocrine responses, the immune response elicited seemed to be specific to the type of inflammatory rheumatic disorder involved, and this was especially evident for psychosocial stress. Patients with SLE often have baseline lymphopenia and this review showed that stress led to diminished cell mobilization and consequently less pronounced stress-induced changes in lymphocyte (sub)populations in these patients compared with controls and patients with RA. In contrast, patients with RA did not consistently show these alterations, but only alterations in certain subpopulations of lymphocytes. Furthermore, the cytokine response to stress might be distinct for different rheumatic populations. The effect of pharmacotherapy (for example, corticosteroids and biologicals) on neuroendocrine and immune responses to stress in these populations is unclear. Only two of 16 studies conducted subgroup analyses to reveal possible effects of pharmacotherapy on response patterns. Motivala et al. (2008) found significant stress-induced increases in TNF-α only in patients with RA not treated with TNF-α antagonists (and no effect of corticosteroid treatment), and Pawlak et al. (1999) reported no significant difference in physiological variables between patients with SLE treated with corticosteroids and/or disease-modifying antirheumatic drugs (DMARDs) and non-treated patients 45. Four other studies also suggest no effect of pharmacotherapy (more specifically, DMARDs 38, nonsteroidal anti-inflammatory drugs (NSAIDs) 3438, and corticosteroids 344344) on stress reactivity patterns based on descriptive data. The other studies do not mention possible effects of pharmacotherapy 293031323335363739404246.

Clearly, patient samples are often too small to draw conclusions. More studies comparing different inflammatory diseases and healthy participant are needed to address the question whether the physiological response to stress is disease-specific, knowledge which might facilitate our understanding of factors contributing to the maintenance and exacerbation of rheumatic disorders.

The heterogeneity of outcome measures, especially with regard to immune variables, makes it difficult to compare studies of the stress response. Moreover, stressors that activate the physiological stress system via the hypothalamus and subsequent down-stream cascades might induce alterations in hormones, peptides and cytokines other than discussed and assessed in this review (for examples, dehydroepiandosterone (sulfate) (DHEA-S), neuropeptide Y (NPY), substance P). Previous studies suggest that alterations in the stress response in patients with rheumatic disorders might be located in the interaction between the two stress axes (HPA and ANS) and the immune system, more specifically in the number or signalling capacity of β-adrenoceptors 45486681 or glucocorticoid receptors 828384 on lymphocytes.

Although the HPA axis, ANS, and immune system are thought to function in an interdependent fashion, only four of 16 studies in this review investigated outcome variables that evaluate all three stress response systems. Cardiovascular reactivity during a mental stress task has been shown to be associated with the subsequent cortisol response 25, and cardiovascular responses have been associated with post stress levels of cytokines such as IL-6 and TNF-α 85. In addition, HPA axis variables are correlated with immune responses after stress 53. Future studies should try to integrate the responses of different systems by simultaneously assessing (para)sympathetic responses, neuroendocrine variables, and immune factors, to increase our knowledge of the coordination and possible dysregulation of these systems 59.

In addition to assessing the physiological response to experimental real-life stressors, it is important to consider other key elements of the stress response, such as the individual's appraisal of the threat of the event, perceived distress, coping behaviour and personality characteristics 4. As the appraisal of threat might partly determine the biological stress response of an individual, self-reported measures of distress appear to be a simple and adequate manipulation check in studies of psychosocial stress. Furthermore, acute stress responses are known to be correlated with individual differences in psychological factors, such as coping styles 86, trait anxiety 8788, depressed mood 598990, perfectionism 91, social inhibition 92 and anticipatory cognitive appraisal 93949596. However, none of the studies, except one 74, assessed individual psychological characteristics. In addition, only two studies in this review assessed stress-induced self-reported distress 4142 and only three studies 36373941 explicitly excluded affective disorders. Although half of all studies excluded patients with (severe or chronic) comorbidity, these were mainly physical in nature. We cannot rule out the possibility that patients with psychiatric disorders are included in these studies, obscuring the major research question whether inflammatory rheumatic diseases are related to a specific physiological stress responses profile. Future studies of stress responses in patients with rheumatic diseases should include individual psychological characteristics (for example, personality, coping styles) and affective responses to help identify whether patients differ with regard to psychophysiological responses to stress.

Some consideration should be given to when outcomes are measured. Almost all the studies included at least one immediate post stress measurement, but only half of the studies included an additional measurement (mostly 30 or 60 minutes after cessation of the stressor), with only two studies including more than two follow-up measurements. Cytokine responses may be delayed for minutes or hours relative to ANS and HPA responses, because cytokines are produced by activated lymphocytes and macrophages 979899. This makes it difficult to detect the peak immune response when collecting only one or two samples, for example for the increase in IL-6 and IL-1 receptor agonists following psychosocial stress 85.

Additionally, in accordance with the allostatic load concept 100 the ability and time taken by the activated stress system to return to baseline (recovery period) might be an important factor in the link between stress and disease 99101102103. Moreover, repeated mental stress has been linked to a lack of habituation of the cortisol response 89104 and plasma interleukin-6 99 in subgroups of individuals. Thus future studies should have more frequent evaluation time points and investigate individual differences in the return-to-baseline values and in habituation patterns after repeated stress.