Does psychological stress in patients with clinically suspect arthralgia associate with subclinical inflammation and progression to inflammatory arthritis?

Patients included in the Leiden Clinically Suspect Arthralgia (CSA) cohort between April 2012 and March 2015 were studied. The CSA cohort is a population-based inception cohort that started at the rheumatology outpatient clinic in Leiden, The Netherlands, with the aim of studying the symptomatic phase of RA that precedes clinical arthritis. Inclusion required the presence of arthralgia of small joints for < 1 year which was because of the character of the symptoms, considered as being suspect to progress to RA by a rheumatologist. A detailed description is provided elsewhere [16], but identification of CSA occurred mainly by the clinical expertise of the rheumatologist. Furthermore, CSA was identified at the first visit, before the results of routine laboratory investigations were known. Notably, general practitioners in our region are discouraged to perform anti-citrullinated protein antibodies (ACPA) testing themselves but are encouraged to refer in a case of any suspicion on imminent RA. After inclusion in the CSA cohort, routine visits were performed after 4 months, 1 year and 2 years. At the request of patients (e.g. in case they experienced more symptoms) patients were also seen in between the scheduled visits. All patients were followed for development of clinically apparent arthritis for 2 years. Follow-up in the CSA cohort ended earlier when clinical synovitis had developed, confirmed with joint swelling at physical examination by the treating rheumatologist. CSA patients were not treated with disease-modifying anti-rheumatic drugs (DMARDs) or corticosteroids in the phase of CSA. For the patients that developed clinical synovitis, further data was obtained as they were subsequently included in the Leiden Early Arthritis Clinic (EAC) cohort [22]. Data collected at clinical arthritis onset and 1 year thereafter were used. Patients who were diagnosed with RA were treated in conformance with national guidelines, which consists of early initiation with a DMARD (preferably methotrexate), in case of failure a second conventional DMARD (either switching or adding) and disease activity score (DAS)-steered treatment adjustments. Biologics were allowed if ≥2 conventional DMARDs failed but that this did not occur in the studied period of 1 year after clinical arthritis onset. A flowchart of the study protocol is available in Additional file 1: Figure S1. Written informed consent was obtained from all patients. The study was approved by the local medical ethics committee.

At each visit physical examinations, including 66-swollen and 68-tender joint counts (66-SJC and 68-TJC) were performed and blood samples were taken to measure CRP (positive if ≥5 mg/L); immunoglobulin M-rheumatoid factor (RF) (positive if ≥3.5 IU/mL); and ACPA (anti-CCP2, EliA CCP, Phadia, The Netherlands, positive if ≥7 U/mL). Questionnaires were completed, including the health assessment questionnaire (HAQ)-disability index (DI), 36-item Short Form Health Survey (SF-36) and self-reported wellbeing, pain and fatigue on numerical rating scales ranging from 0 (no complaints) to 10 (extreme complaints). The PSS-10 was added later to the protocol in 2013 and was gathered only at baseline. Further, to measure subclinical joint inflammation, MRI scans of metacarpophalangeal (MCP), wrist and metatarsophalangeal (MTP joints were made of the most affected side (or the dominant side in case of equally severe symptoms) on an MSK Extreme 1.5 T extremity MR-system (GE, \Milwaukee, WI, USA). MRI scans were made of the same side at baseline and at arthritis onset (before the start of disease-modifying drugs including corticosteroids) and 1 year thereafter. Non-steroidal anti-inflammatory drugs (NSAIDs) were stopped 24 h before the MRI scan. The scans were scored for MRI-detected inflammation according to the Outcome Measures in Rheumatology Clinical Trials (OMERACT) RA MRI scoring (RAMRIS) method as described supplementary and previously published [16]. Total inflammation scores consisted of the sum of synovitis, bone marrow edema (BMO) and tenosynovitis scores. All scans were scored by two independent readers and mean scores of both readers were calculated to obtain the total inflammation score (see Additional file 1: Supplementary Methods). The cut-off of MRI positivity was based on healthy controls as described previously [23]. An MRI was considered positive for subclinical inflammation (tenosynovitis, synovitis or BMO) if this was present in < 5% of healthy volunteers (an example is provided in the Additional file 1: Supplementary Methods).

The psychological stress response was measured by two questionnaires. First, for the main analyses we used the MHI-5 which is a component of the SF-36 questionnaire [19, 24]. Baseline MHI-5 was missing in 18% of the CSA patient and no differences were found in baseline characteristics between patients with an available and missing MHI-5 (Additional file 1: Table S1). The MHI-5 is a brief self-administered questionnaire which includes scales to screen for anxiety and depression [20, 25]. It is well-validated with good psychometric properties for detecting Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) type I axis diagnoses [26], these disorders include clinical mental disorders like anxiety and depression [27]. The MHI-5 has a Cronbach’s α between 0.74 and 0.90 [19, 26]. The items were scored on a six-point frequency rating scale and questions are provided in the Additional file 1: Supplementary Methods. After linear conversion, possible scores on the MHI-5 range from 0 to 100, with lower scores reflecting a higher stress response. As the score has been considered as a dichotomous variable (absence or presence of a psychological stress response), a score ≥ 52 indicated minimal psychological stress (anxiety or depression) and < 52 high psychological stress; here simply called ‘high stress’. The application of this questionnaire to screen for stress by means of depression and anxiety has been examined thoroughly, as illustrated by several studies [25, 26, 28].

Second, to investigate the main results on stress further we analysed the PSS-10. As this questionnaire was added to the protocol at a later time point, PSS-10 data was missing in 44% of CSA patients. No differences were found in baseline characteristics between patients with an available and missing PSS-10 (Additional file 1: Table S2). Patients filled in a Dutch translation which consisted of ten items regarding predictability, controllability and life overload as perceived by the individual during the last month [21]. Questions are provided in the Additional file 1: Supplementary Methods and each item of the questionnaire was rated on a five-point Likert-type scale (0 = never, 1 = almost never, 2 = sometimes, 3 = fairly often, 4 = very often) how they felt with a maximum score of 40. Total scores were calculated after reversing positive items’ scores (questions 4, 5, 7, 8) and then summing up these scores with the negative items’ scores [21]. A higher total score indicates a greater perceived stress response. As there is no predetermined cut-off of high psychological stress, results on this questionnaire were not dichotomized. As reference, the mean score of PSS-10 obtained in 2387 respondents in the United States was 12.1 for males and 13.7 for females [29].

We tested dichotomous outcomes of markers of inflammation (CRP positivity, MRI positivity and arthritis onset). Associations at baseline were tested with logistic regression with MHI-5 as dependent variable and with linear regression with PSS-10 as dependent variable. Longitudinal data of the MHI-5 were assessed with logistic regression with a generalized estimating equation (GEE) model with an unstructured matrix and a logit link function. Longitudinal data on CRP levels and MRI-detected inflammation scores were analysed with a GEE model with an unstructured matrix. In all analyses, (at baseline and follow-up) we corrected for age and gender. Further, as sub-analysis, we repeated analyses in CSA patients that also fulfilled the EULAR definition of arthralgia suspicious for progression to RA. To fulfil the definition patients should have ≥ 3 of these characteristics: joint symptoms of recent onset (duration < 1 year), symptoms of MCP joints, morning stiffness ≥60 min, most severe symptoms in early morning, presence of a first-degree relative with RA, difficulty with making a fist and a positive squeeze test of MCP joints [17]. IBM SPSS v23 (IBM Corp, Armonk, NY, USA) was used. The significance level was set at 0.05. To adjust for multiple testing for the ten different comparisons made at baseline, we applied Bonferroni correction and then p values < 0.005 were considered significant.

Table 1 presents the baseline characteristics of the 241 CSA patients. The mean age was 44 and the majority was female. Forty-five patients developed clinical synovitis after a median follow-up of 17 weeks; 65% of these patients fulfilled the 2010 criteria and 91% started DMARD therapy.

At presentation with CSA, 10% (4/39) of patients with arthritis during follow-up and 12% (24/197) of the total group of CSA patients had a high psychological stress response (‘high stress’) according to the MHI-5 (score < 52). This was not different for patients presenting with or without an elevated CRP (16 versus 11%, p = 0.36), or with or without subclinical MRI-detected inflammation (11 versus 13%, p = 0.56, Fig. 1). Also for the continuous values of CRP and MRI-detected inflammation scores with stress we did not find any significant relationships (p = 0.89 and p = 0.90, respectively). Further, we did not find associations between stress and age, gender, 68-TJC, self-reported pain, or HAQ-DI. Significant associations were observed between stress and self-reported fatigue (odds ratio [OR] = 1.5, (95% confidence interval (95% CI) 1.1; 1.9); p = 0.003) and wellbeing (OR = 1.6, (95% CI 1.3; 2.1); p < 0.001). This means that patients with a high stress response had 1.5 times more severe fatigue and they felt 1.6 times more severely affected in their general wellbeing compared to patients without ‘high stress’.

When analysing the continuous values of the MHI-5, instead of dichotomized, this showed similar results. We found no association for patients presenting with or without an elevated CRP (p = 0.15), or with or without subclinical MRI-detected inflammation (p = 0.60) and also not between stress and age, gender, 68-TJC and self-reported pain. The association between self-reported fatigue (β = − 2.5 (95% CI -3.4; − 1.5); p < 0.001) and wellbeing (β = − 3.4 (95% CI -4.5; − 2.3); p < 0.001) were also found here and additionally we identified associations between stress and HAQ-DI (β = − 9.2 (95% CI −14.6; − 3.8); p = 0.001). The latter means that per point increase in functional disability patients experienced more severe stress, reflected by a 9.2 lower MHI-5 score (on a range 0–100).

When analysing the outcome arthritis-onset during follow-up, there was no association between the percentage of patients with a high stress response at baseline and the number of patients who did and did not develop clinical arthritis (10 versus 13%, p = 0.68, Fig. 1). Also for the continuous MHI-5 score this relationship was non-significant (p = 0.71).

At the time of arthritis-onset, the percentage of patients with a high psychological stress response increased to 31% (p = 0.025). One year later this had decreased to 8% (p = 0.020, Fig. 2).

Also, inflammation was measured over time in patients that progressed to clinical arthritis. The CRP levels were stable while progressing from CSA to clinical arthritis (mean 9.8 and 9.1 mg/L respectively, p = 0.83) and decreased during the first year of treatment (mean 5.5 mg/L; p = 0.056). Total MRI-detected joint inflammation -scores increased between presentation with CSA and arthritis-onset, though this did not reach statistical significance (mean score 7.1 and 8.5 respectively, p = 0.066). One year after presentation with clinical arthritis the scores had decreased (mean score 6.3).

Then we investigated whether baseline stress associated with the MRI-detected total inflammation score or CRP at arthritis onset and this revealed non-significant relationships (p = 0.11 and p = 0.92).

The percentage of patients with a high psychological stress response among the CSA patients that did not progress to clinical arthritis during 2-years of follow-up (n = 196) was stable and ranged between 7 and 13% (p = 0.42, Fig. 2).

To verify the main findings done on psychological stress present at presentation with CSA, we also analysed stress responses measured by the PSS-10. The mean PSS-10 score in all patients was 13.5 (SD 7.6). Patients that presented with or without an elevated CRP, with or without subclinical MRI-detected inflammation or who did and did not develop arthritis did not have higher stress levels (p = 0.18, p = 0.83 and p = 0.60 respectively, Fig. 1). We observed significant positive associations at baseline between stress and self-reported fatigue (β = 0.94; p < 0.001), wellbeing (β = 1.20; p < 0.001).

Additionally, analyses were repeated in patients that were identified as CSA by their rheumatologist and also fulfilled the EULAR definition for CSA (cut-off ≥ 3 items present), which was fulfilled by 75% of the CSA patients. The main findings were similar here too. Measuring the stress response by MHI-5 revealed that the percentage of patients with ‘high stress’ was not statistically different between patients presenting with or without an elevated CRP (16 versus 12%, p = 0.53), with or without MRI-detected inflammation (14 versus 12%, p = 0.92) or for patients who did and did not develop arthritis (13 versus 13%, p = 0.92). Also when measuring stress levels with the PSS-10 no differences were observed for patients with elevated versus normal CRP, positive or negative MRI-detected inflammation and between patients with and without clinical arthritis during follow-up (p = 0.34, p = 0.91 and p = 0.61, respectively, Additional file 1: Figure S2). Also in this subgroup of patients, association between stress (according to both questionnaires) and wellbeing was statistically significant (p < 0.001). Repeating the longitudinal analyses on stress in the subgroup of patients that fulfilled the EULAR definition revealed similar findings as in the total group; both for the patients that progressed to clinical arthritis as in those that did not progress (Additional file 1: Figure S3).