The association between rheumatoid arthritis and reduced estimated cardiorespiratory fitness is mediated by physical symptoms and negative emotions: a cross-sectional study

The present study is part of a larger study, FysKond2, investigating the relationship between different patient-reported outcome measures and PA in patients with inflammatory joint diseases. Results from a subgroup of the participants who performed an optional 6-minute walk test have previously been published [20]. The present study included an RA group and a control group. To have large enough sample size to perform structural equation modelling (SEM), we aimed for at least 200 participants in each group.

Persons with an RA diagnosis according to the American College of Rheumatology/EULAR 2010 criteria were recruited at the Rheumatology Outpatient Clinics at Levanger Hospital and St. Olavs hospital, Trondheim University Hospital, Norway, in 2019 and 2021 [3]. All patients received an information letter about the study before they were contacted. At St. Olavs hospital, patients with well-controlled disease may be transferred to a patient-centred follow-up programme with regular controls at their general practitioner. To recruit a representative sample of RA patients, two recruitment strategies were used. We contacted a random selection of those in the patient-centred follow-up programme as well as patients with scheduled physical or digital appointments at the Rheumatology Outpatient Clinics during the recruitment periods, see Fig. 1. Persons with physical appointments were recruited during their visits. Those without a physical appointment were contacted by telephone and the questionnaires and a return envelope were sent to the participants by mail.

Blood donors were chosen as the control group because they are relatively easy to recruit and consist of persons of different ages and backgrounds. They fulfilled the national criteria for blood donors and were recruited in 2019 during an appointment at the Blood Bank at St. Olavs hospital [21].

All participants filled in questionnaires regarding PA, psychological factors, and physical symptoms, as further described below. Questions about frequency, duration, and intensity of performed PA were used to evaluate fulfilment of the aerobic PA recommendations and to calculate a PA index [16, 22]. Data regarding the rheumatic disease, use of anti-rheumatic medication, and comorbidities were collected from hospital records for the participants with RA. Due to the recruitment strategy explained above, many patients had not recently seen a rheumatologist and did not have a recent disease activity score of 28 joints (DAS28).

Non-exercise models developed in the same region of Norway were applied to calculate eCRF. An RA-specific model was used for the participants with RA and a model developed in healthy individuals was used for the controls [22, 23]. Both models include data on sex, age, body mass index, resting heart rate, and the PA index, and the RA model further includes smoking status and the patient global assessment [23, 24].

Degree of morning stiffness, intensity of joint pain in the past 6 months, and degree of pain in the neck, back, or hips were rated on a Likert scale from 0 to 10. A higher score indicates more severe symptoms.

The 10-item version of Cohen’s perceived stress scale was applied to assess level of stress [25]. The items include questions about the ability to handle stressors and emotional response to stressors in the past month. The scale ranges from 0 to 40, with higher scores indicating more perceived stress. The Hospital Anxiety and Depression Scale’s Depression score (HADS-D) was applied to assess the level of depressive symptoms [26]. The items include questions about mood, enjoyment etc. in the past week. HADS-D is based on 7 items and has a possible range of 0-21, with higher scores implying more depressive symptoms.

The participants with RA responded to the patient global assessment on a 100 mm visual analogue scale with the phrasing “Please consider the activity of your rheumatic disease in the past week. When considering all the symptoms, how do you think your state is?” Self-reported function in the past week was assessed with the modified Stanford Health Assessment Questionnaire (mHAQ) in the RA patients [27]. The mHAQ score has a possible range from 0.00 to 3.00, with higher scores indicating worse physical function. The RA patients also rated the degree of joint tenderness or swelling on a Likert scale from 0-10, with a higher score indicating more symptoms.

For participants who were recruited during a physical visit, resting heart rate was measured after the participant had been sitting for at least 5 minutes. Participants who mailed in their questionnaire were encouraged to count their resting heart rate after sitting for 10 minutes.

Statistical analyses were performed using Stata (v16, StataCorp, College Station, TX, USA). P-values <0.05 were considered statistically significant. Normality was assessed with histograms and the Shapiro-Wilk test. Continuous data are presented as median with interquartile range or as mean with standard deviation as appropriate. Categorical data are presented as number with percentage. Comparisons of descriptive data were performed with the Mann-Whitney U test for non-normally distributed continuous data, the t-test for normally distributed continuous data, and the chi-square test for categorical data.

SEM was applied to evaluate the relationship between RA status and eCRF, including different potential mediators. SEM is a statistical method that tests whether a hypothesized model fits the observed data, and is suitable for mediation analysis, complex models, and models with latent variables. Mediation analysis tests whether some of the effect of the independent variable on the dependent variable is mediated by an intermediary variable (a mediator), meaning that the independent variable influences the mediator which in turn influences the dependent variable. A latent variable is not measured directly, but rather derived from numerous observed variables representing a common concept. SEM can include participants with missing data in the analysis, utilizing the available data. Furthermore, several dependent variables may be included in the same SEM model, and variables may correlate. In SEM several fit indices are used when evaluating whether the observed data fit with the proposed model. A model with good fit should have a non-significant chi-square test, root mean square error of approximation (RMSEA) <0.10, comparative fit index (CFI) ≥0.90, and Tucker-Lewis Index (TLI) ≥0.90 [28]. Several hypothesized models might fit the data adequately. Further details regarding SEM are provided in Online Resource Text 1. Both the unstandardized and standardized path coefficients were calculated. Following standardization, all coefficients in the model are given on the same scale, namely standard deviations for each included variable.

All SEM models were adjusted for sex and age and are presented in Figs. 2 and 3. Model 1 investigated the direct effect of RA status on eCRF. Model 2a builds on Model 1 by adding the latent variable “physical symptoms” as a possible mediator of the relationship between RA status and eCRF. The physical symptoms latent variable was derived from three observed variables: morning stiffness, joint pain in the past 6 months, and pain in neck, back, or hips. Model 2b added a different possible mediator of the relationship between RA status and eCRF to Model 1, namely “negative emotions”. The negative emotions latent variable was derived from the observed variables depressive symptoms and perceived stress. Model 3 combined Model 2a and 2b to explore how including both physical symptoms and negative emotions as mediators of the relationship between RA status and eCRF in the same model would impact the mediating role of each latent variable. We did not adjust for smoking habits because of the complex relationship between smoking, mental health, and physical fitness. For example, smoking is a risk factor both for developing RA and reduced eCRF but might also be a mediator of the relationship between negative emotions and eCRF.

The Regional Committee for Medical and Health Research Ethics approved the project (#23420). The project was performed in accordance with the Helsinki declaration. All participants provided informed consent.

The RA patients had comparable CRF estimates to previous CRF results in Norwegian RA patients, though higher than results in British RA patients [7, 9, 15]. The differences may be explained by recruitment strategy, recruitment year, regional and cultural factors related to PA habits, disease activity, and comorbidities. The mean difference in eCRF between RA patients and controls was smaller than expected, but in the expected direction and statistically significant. In a large Norwegian population-based cohort, the age-adjusted mean eCRF in RA patients was 3.2-5.0 mL/kg/min lower in women and 1.8-4.0 mL/kg/min lower in men compared to controls [10]. Perhaps the fitness level among Norwegian RA patients has improved over the past decade with better medical treatment. Another explanation may be participation bias in the present study. Inactive patients are less likely to participate in a study about PA especially if it involves a physical test, and perhaps the invitation to the optional 6-minute walk test for a subgroup of the RA patients made some more reluctant to participate. However, many participants with low PA levels agreed to participate as only 28% of RA patients fulfilled the aerobic PA recommendations. The eCRF results among the controls were comparable to CRF results published in healthy individuals of similar mean age in the same region of Norway [29].

The standardized coefficients for the effect of RA status on eCRF might appear small. However, the size of standardized coefficients depend on the underlying distribution and relative importance of each variable. This association is still of clinical interest as even small improvements in CRF have positive effects on long-term health [1].

As the RA participants in the present study had high self-reported function and relatively high eCRF estimates, they might not be representative for all RA patients or RA patients in other countries. However, we expect that the role of mediating variables would be even more prominent in a population with larger differences in CRF between RA patients and controls.

Physical symptoms, represented by a latent variable based on the observed variables morning stiffness, joint pain, and pain in neck, back, or hips, explained 37% of the association between RA status and eCRF in the final model. Both pain and stiffness have been reported as important barriers to PA by RA patients [30]. Stiffness has been associated with PA habits in one study, whereas the association between pain and PA habits appears to be non-significant in most studies of RA patients [31, 32]. In patients with fibromyalgia more pain has been associated with lower fitness [33]. We may speculate that pain mainly affects exercise intensity, as our RA patients reported lower PA intensity than controls and as PA at low intensity would lead to lower CRF compared to PA at higher intensity. Regardless, exercise may lead to some reduction in pain among persons with RA. Thus, exercise can affect pain and pain can affect exercise habits [14, 30]. As the physical symptoms variable was based on three distinct symptoms, the results are not directly comparable to previous studies. The symptoms included were chosen as they are closely related to each other and therefore fit as indicators of the same latent variable. Furthermore, musculoskeletal symptoms are common both in RA patients and in healthy individuals.

Pain and morning stiffness may be prominent symptoms even when the patient is in clinical remission. EULAR recommends a patient-centred approach to pain management that involves proper assessment, a personalized management plan, and patient education [34]. Treatment of pain may involve non-pharmacological options such as PA, cognitive-behavioural therapy, and weight management, and if necessarily also pharmacological options. Some patients fear that PA can lead to disease exacerbation and joint damage [30]. This concern should be addressed with patient education, clear exercise advice, and possibly referral to a physiotherapist or other interventions that can further increase the patients’ knowledge and confidence regarding PA and exercise [16, 35].

Negative emotions, represented by a latent variable combining depressive symptoms and perceived stress, explained 37% of the association between RA status and eCRF in the final model. The relationship between RA, depression, stress, and CRF is complex. RA itself increases the risk of depression and some types of stress, however both depression and stress may increase the risk of developing RA [17, 18, 36, 37]. Both depression and stress have been associated with a reduction in PA over time, and stress has been associated with lower self-reported fitness levels [19, 38, 39]. However, high CRF protects against developing depressive symptoms [2].

Depression has several detrimental effects in RA patients, including lower chance of remission, lower adherence to treatment, higher pain levels, lower quality of life, and even increased mortality rates [17]. Stress in RA patients is associated with more pain, anxiety, and reduced quality of life [40, 41]. Active behaviour coping strategies might buffer some of the impact of perceived stress on depressive symptoms in RA patients [41]. EULAR recommends regular assessment of mental health in persons with inflammatory joint diseases and initiation of interventions if necessary, as better mental health can lead to better self-management [35]. The present study demonstrated that negative emotions are associated with lower eCRF. We speculate that improvements in mental health, particularly if the individual has high levels of depressive symptoms and perceived stress, could lead to increased PA engagement, which again may increase CRF levels. Healthy lifestyle choices such as a healthy diet, regular PA, enough sleep, strong social connections, and avoidance of risky substance intake, are beneficial for mental health [42]. Persons with clinical depression might need psychological and pharmacological interventions as well. Spending time in nature and mind-body practices are other strategies that may alleviate stress [42].

Other factors than physical symptoms and negative emotions probably also mediate the relationship between RA and low CRF. Potential mediators might include fatigue, comorbidities, lung function, body composition, inflammation, exercise habits, anxiety, motivation, self-efficacy, coping strategies, and sarcopenia [11, 32, 43, 44]. We focused on factors that are common both in RA patients and the general population, though generally more prevalent in RA patients. Including too many potentially mediating factors will lead to a very complex model and is only feasible in very large studies. RA-specific variables such as disease duration, seropositivity, tender and swollen joint counts, or disease activity, could not be included as potential mediators due to collinearity with RA status. For example, we would not expect the controls to have swollen joints, so including swollen joint count as part of the physical symptoms latent variable would lead to collinearity with RA. Although the presence of RA in the present study was associated with lower eCRF, other factors like sex, age, PA habits, and genetics probably have a larger impact on an individual’s CRF than RA status [11, 12].

A strength of this study is that we included a relatively large sample of RA patients and controls recruited from the same geographical region. Even if the blood donors are a healthy sample of the population with few comorbidities and might not be directly comparable to persons with RA, both groups consisted of persons with a wide age range with varying levels of PA, and the analyses were adjusted for age and sex. The models had good fit. Nevertheless, there are likely other SEM models with acceptable fit, especially since several of the factors in the model may act bidirectionally.

Missing data for eCRF in 27% of the RA patients is a weakness. This was due to changes in the recruitment process to also include mail-only participation during the COVID-19 pandemic, which resulted in missing data for resting heart rate and consequently eCRF. However, all participants were included in the SEM models with their available data, which is a recognized strategy to deal with missing data that generally introduces less bias than for example listwise deletion [45]. Using non-exercise models to calculate eCRF can lead to both over- and underestimation of physical fitness. A cardiopulmonary exercise test would have given more accurate CRF results. Nevertheless, we did not have the resources to perform such testing. Moreover, by not including such testing less effort was required to participate, which probably resulted in a more representative and larger sample. Physical symptoms, depressive symptoms, and perceived stress are subjective variables and there is inevitably some uncertainty with these measures. HADS-D and Cohen’s perceived stress scale were chosen to assess the negative emotions because they are well-known, validated, and frequently used instruments [25, 26]. Due to the cross-sectional design, evaluation of causality was not possible.