Correlates of physical activity in adults with spondyloarthritis and rheumatoid arthritis: a systematic review

A systematic search was conducted using a tailored search strategy in the following databases: PubMed (inc. Medline), Web of Science, Embase, APA PsycNET, and Scopus. The search strategy was developed based on a preliminary search in PubMed to identify relevant terminology, search terms employed by similar reviews into correlates of PA in inflammatory arthritis, and discussions with a subject librarian and the research team. The databases were searched from 2004 to June 2019, to develop and further a review previously conducted by Larkin and Kennedy [24]. An example of the search strategy used is shown below. The search was restricted to articles with human participants and published in English. Additional eligible studies the author became aware of were included.

Quantitative observational studies and experimental studies that explored correlates of PA in adults (≥ 18 years) with SpA and/or RA were included. Intervention studies were eligible for inclusion if baseline data were reported. Study participants must have had a diagnosis of RA using the American College of Rheumatology (ACR) /EULAR criteria [26]. Due to the limited number of studies, those with SpA were included if they received a doctor diagnosis, but not if the diagnosis was self-reported. SpA diagnoses that were eligible included: axial spondyloarthritis (axSpA); non-radiographic spondyloarthritis (nr-axSpA); ankylosing spondylitis (AS); psoriatic arthritis (PsA); enteropathic arthritis (EnA); reactive arthritis (ReA); and undifferentiated spondyloarthritis (uSpA). Studies utilising self-reported (questionnaires) and objective measures (accelerometers, pedometers, calorimetry, doubly labelled water, and heart rate monitors) of PA were included. For consistency with other reviews [24], PA was defined in this work as ‘any bodily movement produced by skeletal muscles that results in energy expenditure’ [27]. Exercise is a subcategory of PA, that is distinct in definition by its ‘planned, structured and repetitive’ nature, with an ‘objective to improve or maintain physical fitness component(s)’ [27]. Cross-sectional studies where PA was not the dependent variable were included.

Articles were excluded if they were not published between 2004 and June 2019 and if the study participants were aged younger than 18 years. Qualitative studies and studies that exclusively explored functional or physiological variables of PA in SpA and RA were excluded. Studies that included multiple forms of arthritis but did not analyse the data separately were excluded. Studies that reported on physical inactivity were included, but those reporting exclusively on sedentary behaviour were excluded as sedentary behaviour is independent from physical (in)activity [28]. Longitudinal studies in which PA was not the dependent variable were excluded, unless baseline data were reported. Abstracts, conference proceedings, and grey literature were excluded.

The first author conducted the database searches and imported them to the reference manager software Endnote, whereby duplicates were removed. The abstracts were then imported into the review management software Covidence [29] and further duplicates removed. The first author screened all titles and abstracts independently against the selection criteria, with a second author independently screening 40% to determine eligibility. The same method was used for the review and selection of full text articles. During these two stages a consensus approach was used to resolve any disputes. All disputes were resolved without the need for consultation with a third author. Data extraction was conducted by the first author. Data such as participant characteristics, study design, measure of PA, variables analysed, statistical test, and statistical output were extracted.

Selected articles fulfilling the inclusion criteria were assessed for methodological quality using The National Institutes of Health (NIH) Quality Assessment Tools for Observational Cohort and Cross-sectional Studies and Case–Control studies [30]. These assessment tools are recommended and deemed acceptable [31]. The assessment tools appraise studies based on the following concepts: objectives; population and sample size; timeframe; outcome measure; exposure variables and measures; attrition; blinding; and analysis. Each of the questions were scored as yes, no, cannot determine, not applicable, or not reported as appropriate. A final quality rating of good, fair, or poor is afforded to each study by the reviewer. The lead author independently assessed the quality of studies, with a following two authors independently assessing 50% each. Any disputes were resolved via a consensus method between the authors, and further consultation from a third author was not warranted.

Following discussions with the research team, meta-analysis was not conducted due to the nature of included studies (e.g., type of PA and statistics reported). Rather, the coding framework employed by Sallis et al. [32] was used to pool and classify the associations between PA parameters and potential correlates into largely positive/negative ( ±), indeterminant / inconsistent (?) or no association (0). Following this framework, univariate tests were reported, even if multivariate tests were presented. The results will be presented using this approach for consistency across studies. Yet, an associations table utilizing the most adjusted or final statistical model for each variable can be seen in Online Resources 1 and 2. Conceptually similar variables were grouped together, for example, “high-density lipoprotein” (HDL) included, HDL, HDL particle, small particle, and large particle concentration. To be consistent with previous reviews [24], the correlates identified were categorised into the following domains: sociodemographic, physical, psychological, social, and environmental.

For studies utilizing multiple measures of PA (i.e., subjective, objective, or multiple subjective), correlations from all PA measures were included. Both categorical (e.g., active vs. inactive, meeting PA recommendations, health-enhancing PA, exercising ≥ 2/week) and continuous PA parameters (e.g., total PA min, step count, time spent in PA intensities, metabolic equivalent of task, physical activity energy expenditure (PAEE), accelerometer counts) were eligible for inclusion. Continuous parameters of PA are reported above categorical parameters of PA if the same study reports both. However, if a variable was analysed as a potential correlate of a categorical PA parameter, but not of a continuous PA parameter, this was included. Up to two PA parameters were included for each study (indicators of total PA considered first), apart from if multiple measures of PA were used, so that one study population is not overrepresented. If more than two types of PA intensity were reported in a study (i.e., light PA (LPA), moderate-intensity PA (MI-PA), vigorous-intensity PA (VI-PA)), the two most intense categories were included. For accelerometer data, PAEE was considered over total energy expenditure (TEE).

Fifty-one published articles (RA = 40; SpA = 11) were included with the full screening process provided in Fig. 1. The predominant reasons for exclusion were that the studies did not report correlates of PA (n = 30) and that the study included mixed diagnoses, non-specific arthritides or did not analyse RA and SpA data separately (n = 19). A description of each study and the overall methodological quality can be seen in Tables 1 and 2. The number of RA participants ranged from 20 to 3,152, whereas the number of SpA participants ranged from 24 to 2,167. Three RA studies utilized participants from the Swedish Rheumatology Quality Registers [35‐37], and a further two used participants from the Swedish RA register [40, 41] . Two RA studies contained the same patients from 3 hospitals in The Netherlands [49, 53]. Two other RA studies used participants and cross-sectional data from the same randomized clinical trial investigating the effectiveness of a PA promotion intervention in the USA [38, 54]. One randomized intervention study [63] investigated the effects of a 12-week low-impact aerobic exercise programme on disease and psychological variables, and aerobic fitness, in those with RA. The randomized intervention study compared a class exercise treatment group, a home exercise treatment group (via videotape), and a control group [63]. Two SpA studies included patients from the Skåne Healthcare Register [78, 79]. Most RA studies had a larger proportion of female participants than male, and four RA studies included only female participants [39, 64, 65, 67]. Twelve RA studies were conducted in the United States and twenty-five were conducted in Europe. Ten of the eleven SpA studies were conducted in Europe. The pooled associations and correlates of PA in those with RA and SpA are displayed in Tables 3 and 4. The results are coded in terms of the specific variable and not necessarily the statistical direction derived from the measure of the variable. Online resources 3 and 4 show the positive and negative statistics for each variable related to PA in all included studies.

Bild vergrößern

Self-reported measures of PA were used in twenty-seven RA studies. Twenty-one RA studies used known self-report questionnaires, whereas six studies applied custom questionnaires or interviews. Objective measures of PA were used in fourteen RA studies. Thirteen RA studies used an accelerometer, and one used a Liferecorder. Nine SpA studies used self-reported measures of PA, whereby, five SpA studies used validated questionnaires, and a further four used customized scales. Objective measures of PA were used in four SpA studies (i.e., accelerometers).

The current systematic review extends and updates Larkin and Kennedy’s [24] review that was limited to ten studies with mainly cross-sectional designs (8/10) and only one study utilizing objective measures of PA. The pooled results of the RA studies in this review support those presented by Larkin and Kennedy [24]. Motivation, self-efficacy, and previous levels of PA were positively associated with PA parameters. Fatigue and plasma glucose were negatively related with PA. HDL was consistently positively related with PA parameters. However, the positive association of health perception with PA was indeterminant in this review. Additional positive associations identified via this latest review include, any use of alcohol, gait speed, lean fat/mass, apolipoprotein A1 and subjective vitality. Additional negative associations include functional inability, stiffness, abdominal obesity, body fat, hypertension, blood pressure, metabolic syndrome, nutritional complications, insulin resistance, RA-related joint surgery, and hospital admissions/length. The most frequently studied variables showing a lack of consistent findings include age, gender, educational level, BMI, disease activity, exercise beliefs, health perception, and health-related quality of life.

This review did not find any of the sociodemographic variables to be consistently related to PA in those with RA. Some evidence that any use of alcohol (100%, n = 1/1) was positively related to PA was reported, however, this was investigated in one study, and was unrelated to PA when the results where pooled based on the most adjusted or final models for each variable. Age and female gender were inconsistently related to PA; however, the associations were predominantly negative (age, 53%, n = 16/30; female gender, 47%, n = 9/19) and non-significant. The trend towards a negative relationship is somewhat consistent with findings in the general population of older adults, whereby higher age and female sex are associated with a decreased likelihood of persistent PA over ten years [86]. The reason for inconsistent or indeterminant pooled associations between sociodemographic variables and PA is likely to be multifaceted. For example, the age ranges between samples varies considerably, with studies utilizing both narrow (i.e., 40–70 years) [63] and wide (i.e., 19–90 years) [40] age ranges. Further, a gender imbalance is observed in many studies, with a female dominance of participants. Assessment of PA may also be a factor, as when pooling only associations of objective measures of PA, female gender was consistently and negatively related to PA (80%, n = 4/5). A possible reason for the negative association between female gender and PA could be that females report greater disease activity and physical function, compared to males with RA [87]. Race/ethnicity was consistently unrelated to PA parameters, yet, caution should be taken in the interpretation of this result, as racial or ethnic minorities were not proportionately represented in some studies. For example, one study reported that 71% of participants were Caucasian, whereas 2% were Asian, and 22% African American [47]. In another study of 573 individuals with RA, 94% of the participants were Caucasian [50].Of the physical variables, functional disability was the most consistent and negative correlate of PA parameters (60%, n = 18/30). Although the findings were more inconsistent when considering the most adjusted or final model statistics, this finding corroborates recent reports linking worse physical functioning to lower levels of PA in those with early RA [88] and PA maintenance over a 7-year period [89]. Further, fatigue was negatively related to PA (61%, n = 11/18), yet the results may be distorted due to one study [55] utilizing the five separate domains of the MFI-20. Univariately, Løppenthin et al. [55] found four of the five domains of fatigue to be inversely associated with regular PA. However, in the multivariate model, only reduced activity due to fatigue and physical fatigue remained significant. Predominantly, our findings support recent longitudinal studies showing a negative relationship between fatigue and PA [89, 90]. Interestingly, disease activity was inconsistently related to PA, however, almost half of the associations were negative (42%, n = 13/31). Disease activity was unrelated to PA when pooling the results based on the most adjusted statistics or when considering studies utilizing objective measures of PA. One possible reason for the inconsistency in results might relate to whether participants were experiencing a flare when PA was measured. Further, the relationship between disease activity and PA might be dependent on the contribution of other variables in explaining PA. For example, recent reports have shown no association between PA and disease activity, when controlling for age and fatigue [91]. BMI and obesity have been shown to negatively associate with PA in the general population [86] and those with hip osteoarthritis [92], yet our findings suggest this relationship is more inconsistent in people living with RA (38%, n = 8/21). This inconsistency in findings is not surprising considering the complex nature of energy expenditure, which is largely determined by body composition and body weight [93]. In support of a previous review [24], RA duration and pain was unrelated to PA parameters. The most consistent physiological variable positively related to PA parameters was HDL (67%, n = 6/9). Other physiological variables, especially those detecting inflammation (i.e., ESR and CRP), were unrelated to PA.

Self-efficacy was one of the most consistent psychological variables positively related with PA (70%, n = 7/10). Greater self-efficacy has been positively associated with PA in healthy populations [94] and women with other long-term conditions such as fibromyalgia [95]. This review adds to the evidence supporting the association between self-efficacy and PA previously reported by Larkin and Kennedy [24], with the addition of six more studies (seven associations) in the RA population. Inconsistent findings were reported for exercise beliefs and expectations, health perception (including global health/assessment), and health-related quality of life. Motivation was the most consistent psychological variable positively related to PA (83%, n = 5/6) in this review. This positive relationship is supported by studies highlighting the importance of autonomous motivation in predicting PA participation levels in 207 RA members of the National Rheumatoid Arthritis Society [96] and in increased engagement in moderate-to-vigorous PA (MVPA) during a randomized controlled trial [97]. Two studies investigating motivation regulation style (on the self-determination theory continuum), showed an autonomous regulatory style to be associated with higher PA, and a more controlling regulatory style to relate to lower PA. The importance of autonomous regulations in fostering PA behaviour is supported by the findings of a systematic review exploring the relationship between self-determination theory constructs (e.g., autonomous vs. controlled motivation) and PA [85].

Social variables in this review were limited to social support and no significant relationship was found. This non-significant relationship was surprising as research conducted with people living with knee and hip osteoarthritis, has reported that perceptions of autonomy support from their spouse regarding PA to be positively related to PA, while perceived pressure from a female spouse (to a male partner), has been shown to be negatively related to PA [92]. In the current review, positive (n = 2), negative (n = 1), and non-significant relationships (n = 6) were found between social support and PA parameters. The variability in results may be due to the differences in assessment methods (Social Support for Exercise Behaviors Scale, 2 studies; Important Other Climate Questionnaire, 1 study; Health Care Climate Questionnaire, 1 study; Medical Outcomes Study Social Support Survey, 2 studies), type of PA parameter (e.g., total PA, trajectory of PA, health-enhancing PA or LPA), or the type of individual assessed for social support (e.g., spouse, friend, rheumatologist). Further research is needed to explore this variability and the factors that affect the relationship between social correlates and PA behaviour.

No environmental correlates were investigated in the RA studies. Of the other variables, and in concordance with two other reviews [23, 24], previous levels of PA were shown to be positively associated with PA parameters (100%, n = 3/3).

In SpA studies, positive associations with PA were found for components of spinal mobility (Modified Schober test and cervical rotation), maximum volume of oxygen (VO2max), quality of life, motivation, previous PA and seasonal variation. However, in a model predicting PA after two years, spinal immobility (total BASMI) was not significant [74]. Negative associations with PA were found for functional disability, having AS, being amotivated, and age at onset of anti-TNFα therapy. Indeterminant associations were found for marital status, smoking, disease activity, and PsA subtype. All other variables were not significantly associated with PA parameters.

None of the sociodemographic variables were positively or negatively related to PA in those with SpA. Gender and age were unrelated to PA in this review. However, total self-reported PA, in 24 AS patients (aged 48 ± 11 [83]) and meeting the WHO MI-PA recommendations in 2,167 SpA patients (aged 55 ± 14 [78]) were positively associated with age in two studies (29%, n = 2/7). Marital status and smoking were inconsistently related to PA and employment status, education and children raised under 12 years old were unrelated in a limited number of studies.

Of the physical variables, functional disability was frequently investigated, with studies consistently supporting a negative correlation with PA parameters (62%, 8/13). This is supported by recent findings from a prospective cohort study which found that time spent in MVPA was associated with better function [98]. Surprisingly, spinal immobility, measured via total BASMI score was only reported in one study [74]. Here, a non-significant association with PA was reported. However, Arends et al. [73] reported on the relationship between PA and the separate domains of spinal mobility, and together, the direction of associations indicates that higher PA is related to better spinal mobility. This association has also been supported in relation to time spent in MVPA [98], yet the evidence from exercise programmes is more conflicting and lacks clarity [99]. Longitudinal studies are needed to help clarify the relationship between PA and spinal mobility, and the types of PA most related.

Mixed and inconsistent associations were found between PA and disease activity, with studies reporting positive (n = 3), negative (n = 8) and non-significant (n = 8) associations. Such disparity in findings is commensurate with a review investigating the correlates of PA in people with arthritis [23]. The inconsistency in results may partially be due to the intensity of PA measured, for example, one study showed that higher disease activity (measured via the BASDAI) was associated with greater LPA in adults with AS aged ≥ 52 years, but not associated with MVPA in AS patients across all ages [82]. A further explanation for the inconsistency in results may be that some individuals with SpA may participate in PA as a reaction to disease activity, in comparison to those who may start PA engagement earlier as a preventative measure against disease activity. Interestingly, only one study investigated whether SpA subtype was associated with PA [78]. In multivariate analyses, uSpA and IBD-related arthritis were not significantly associated with meeting the MI-PA or VI-PA recommendations, however, PsA was negatively related to meeting VI-PA recommendations. One possible explanation is that individuals with PsA typically experience more inflammation of the peripheral joints than other SpA subtypes [100] and have separate clinical presentations. Due to the limited number of studies, more research is needed to determine potential differences between SpA subtype and PA parameters.

Of the psychological variables, a self-reported better quality of life was associated with higher PA (83%, n = 5/6) in those with SpA. This association is supported in an online survey of 445 Danish individuals with SpA [101] and those with hip osteoarthritis [92]. Motivation to exercise was investigated in one study [75], whereby intrinsic motivation was positively related to PA levels. Using a path analysis, autonomous forms of motivation (i.e., intrinsic motivation and identified regulation) were positively associated to PA, whereas amotivation was negatively associated. Regulation style has also been found to be a significant determinant of PA in those with RA [49]. The importance of motivation is further highlighted in a behaviour change intervention for those with axial spondyloarthritis. Those undertaking the intervention aimed at supporting PA motivation had improved health-enhancing PA and spinal mobility [102]. Despite the apparent importance of motivation on PA behaviour [85], little research has been conducted in those with SpA. Future research should focus on longitudinal study designs or the utilization of ecological momentary assessment methods [103] to yield insight into the causal and fluctuating relationships between PA motivation and PA behaviour.

Previous levels of PA were found to positively correlate to PA [74]. Based on retrospective data, one study reported that individuals who started anti-TNFα therapy earlier had greater gains in PA [81]. Seasonal variation (summer) was positively associated with PA energy expenditure in one study [77].