Abstract
Objective To determine if the degree of baseline fibromyalgia (FM) symptoms in patients with rheumatoid arthritis (RA), as indicated by the Fibromyalgia Survey Questionnaire (FSQ) score, predicts RA disease activity after initiation or change of a disease-modifying antirheumatic drug (DMARD).
Methods One hundred ninety-two participants with active RA were followed for 12 weeks after initiation or change of DMARD therapy. Participants completed the FSQ at the initial visit. The Disease Activity Score in 28 joints using C-reactive protein (DAS28-CRP) was measured at baseline and follow-up to assess RA disease activity. We evaluated the association between baseline FSQ score and follow-up DAS28-CRP. As a secondary analysis, we examined the relationship between the 2 components of the FSQ, the Widespread Pain Index (WPI) and Symptom Severity Scale (SSS), with follow-up DAS28-CRP. Multiple linear regression analyses were performed, adjusting for clinical and demographic variables.
Results In multiple linear regression models, FSQ score was independently associated with elevated DAS28-CRP scores 12 weeks after DMARD initiation (B = 0.04, P = 0.01). In secondary analyses, the WPI was significantly associated with increased follow-up DAS28-CRP scores (B = 0.08, P = 0.001), whereas the SSS was not (B = −0.03, P = 0.43).
Conclusion Higher levels of FM symptoms weakly predicted worse disease activity after treatment. The primary factor that informed the FSQ’s prediction of disease activity was the spatial extent of pain, as measured by the WPI.
Despite continued advances in rheumatoid arthritis (RA) treatment, less than half of patients with RA attain low disease activity within 6 months of disease-modifying antirheumatic drug (DMARD) therapy, and less than a quarter of patients achieve remission.1,2 Knowledge of the factors that affect treatment outcomes in RA is limited.3-6 Addressing this gap in knowledge would be of critical benefit for the management of RA. Better predictors for RA outcomes would mitigate patients’ exposure to potentially toxic drugs, limit the economic burdens linked with trying multiple DMARDs, and reduce the period during which patients experience poorly controlled symptoms.
One possible reason for suboptimal DMARD response is persistent pain. Patients and healthcare providers often interpret pain as a sign of inflammation, and most composite measures of RA disease activity include components affected by pain (eg, tender joint count [TJC] and patient global assessment [PtGA]). However, a subset of patients with RA obtains good inflammatory control but continues to report pain, suggesting noninflammatory processes may affect these patients’ experiences with RA.7-10
A potential mechanism for persistent, noninflammatory pain involves dysregulation of pain processing pathways in the central nervous system (CNS). Quantitative sensory testing (QST) has implicated CNS dysfunction in the experience of pain in patients with RA. Relative to healthy controls, patients with RA were reported to have a more sensitized response to pain with QST, indicative of dysregulated CNS pain processing.11 In another study, greater baseline abnormalities in QST were associated with lower odds of obtaining a good treatment response in RA.12
Despite its value for predicting treatment response, QST is not a practical measure of CNS pain dysregulation in the setting of busy clinical practices. It would therefore be optimal if a patient-reported measure, such as the Fibromyalgia Survey Questionnaire (FSQ), could be used to predict treatment outcomes in the clinical setting. The FSQ is used to diagnose and quantify the severity of fibromyalgia (FM), the prototypical chronic widespread pain condition associated with abnormalities in CNS pain regulation. The FSQ is a continuous, selfreported measure that consists of 2 scales: the Widespread Pain Index (WPI) and the Symptom Severity Scale (SSS).13 To our knowledge, no studies have examined the ability of the FSQ to predict treatment outcomes in patients with active RA who are starting or switching DMARD treatment.
The primary aim of this study was to determine the association between FSQ score prior to DMARD initiation and RA disease activity approximately 12 weeks after treatment with DMARDs. A secondary aim was to investigate if the 2 components of the FSQ, the WPI and the SSS, have varying predictive strength of RA disease activity after treatment.
METHODS
Study sample. Data for this study were from the Central Pain in Rheumatoid Arthritis (CPIRA) cohort.11,12,14,15 CPIRA is a multicenter, prospective observational study consisting of patients with active RA requiring a change or initiation of DMARD therapy. The study was approved by the institutional review boards at each of the 5 participating academic medical centers: Boston University (H-32334), Brigham and Women’s Hospital and Massachusetts General Hospital (2013P000951), Johns Hopkins University (NA_00085841), and University of Michigan (HUM00081289). From January 2014 to July 2017, participants were recruited from the 5 medical centers. Informed consent was obtained from all study participants prior to enrollment.
Inclusion criteria were a diagnosis of RA per the 2010 American College of Rheumatology (ACR)/European Alliance of Associations for Rheumatology (EULAR) criteria,16 and the commencement or change of DMARD because of active RA. Exclusion criteria were as follows: (1) known peripheral neuropathy, (2) severe peripheral vascular disease, (3) Raynaud phenomenon, (4) chronic opioid use, (5) changing dose of central-acting pain medication, (6) corticosteroid therapy > 10 mg prednisone or equivalent during the 24 hours before testing, and (7) NSAID or acetaminophen use 24 hours before testing. Participants with missing baseline or follow-up Disease Activity Score in 28 joints using C-reactive protein (DAS28-CRP) scores were excluded from this longitudinal analysis.
Clinical variables. Baseline clinical variables were evaluated prior to the initiation or switch of DMARD. Follow-up clinical variables were assessed approximately 12 weeks after the initiation of DMARD therapy. The extent of participants’ FM symptoms was determined by the FSQ from the 2010 modified ACR preliminary diagnostic criteria for FM.13 The FSQ is composed of 2 survey-based assessments: the WPI, which is representative of the spatial extent of pain, and the SSS, which is representative of the severity of somatic symptoms, cognitive symptoms, fatigue, and waking unrefreshed. Disease activity was assessed with the DAS28-CRP, a composite score calculated from TJC, swollen joint count (SJC), PtGA, and CRP.17 Blood samples were collected to assess serum CRP and seropositivity, and trained staff members performed standardized joint counts to assess joint tenderness and swelling. PtGA, a patient-reported rating of their state of health over the past 7 days, was provided on a scale of 0 to 10. RA disease duration, defined as time from onset of RA diagnosis, was calculated. Two participants did not have data on RA disease duration, so these values were imputed. The imputed values were the expected disease duration based on the linear regression of RA disease duration on RA symptom duration.
Statistical analyses. The primary outcome measure was follow-up DAS28-CRP score at 12 weeks. The primary predictor was the FSQ score at baseline. The secondary predictors were the WPI and SSS scores, the 2 components of the FSQ. For the primary analysis, univariable regression was performed, followed by multivariable linear regressions that adjusted for demographic variables and additional variables that could confound the relationship between FSQ and follow-up DAS28-CRP (baseline DAS28-CRP, age, sex, BMI, race, study site, seropositivity, RA disease duration, and number of comorbidities). For the secondary analyses, multivariable linear regressions were performed with the WPI and the SSS as the predictor variables, followed by additional multiple linear regressions that controlled for the same clinical and demographic variables included in the primary analysis. Both unstandardized (B) and standardized (β) regression coefficients were calculated. Unstandardized regression coefficients represent the amount of change in the dependent variable due to a change of 1 unit in the predictor variables (ie, FSQ for the primary analysis, or WPI and SSS for the secondary analysis). Standardized regression coefficients enable comparison across predictors by transforming the data such that the variances of the dependent and independent variables are equal to 1. All data analysis was performed on RStudio version 1.4 for macOS (RStudio Team).
RESULTS
Of the 295 participants in the CPIRA study, 57 were excluded from this study because they were lost to follow-up, and another 46 were excluded due to missing baseline data. There were no significant differences in baseline clinical variables between included and excluded participants.
Clinical characteristics of the included participants are found in Table 1. Among the 192 participants in this analysis, the majority were women (83.9%) and White (76%). The mean (SD) of age and RA disease duration were 55.2 (14.4) and 10.4 (12.5) years, respectively. The average baseline scores for the FSQ, WPI, and SSS were 11.1, 5.8, and 5.3, respectively. The observed range of baseline FSQ scores was 0 to 28, whereas the potential range was 0 to 31. The observed range was identical to the potential range for the WPI (0-19) and the SSS (0-12), respectively. Mean (SD) baseline DAS28-CRP was 4.3 (1.3), which is within range of moderate disease activity (DAS28-CRP > 3.2 and ≤ 5.1).18 Mean (SD) follow-up DAS28-CRP was 3.3 (1.3), decreasing by 1.0 (1.1) from mean baseline DAS28-CRP. TJC decreased by 4.5 (6.8) joints, SJC decreased by 2.6 (5.0) joints, PtGA decreased by 1.4 (2.5), and high-sensitivity CRP decreased by 2.7 (11.4) mg/L.
Baseline clinical characteristics of participants.
We hypothesized that a higher FSQ score prior to DMARD initiation or switch would predict a higher DAS28-CRP score at follow-up. Univariable linear regression (Table 2) demonstrated that participants’ baseline FSQ score significantly predicted follow-up DAS28-CRP score (B 0.09, 95% CI 0.06 to 0.12; β 0.39, 95% Cl, 0.26 to 0.52, P < 0.001). The fit of the model, however, was limited, with an adjusted R2 of 0.15. When adjusted for baseline DAS28-CRP and other demographic characteristics, baseline FSQ score likewise positively predicted DAS28-CRP score after 12 weeks of DMARD treatment, though with a lower regression coefficient (B 0.04, 95% CI 0.01 to 0.07; β 0.16, 95% Cl 0.03 to 0.29, P = 0.01). The adjusted R2 of this model was 0.42. The addition of further covariates to model 3 did not appreciably change the regression coefficient or CIs (B 0.04, 95% CI 0.01 to 0.07; β 0.17, 95% Cl 0.04 to 0.30, P = 0.01).
Linear regression models for the association between baseline FSQ and DAS28-CRP after 12 weeks of DMARD therapy (N = 192).
We hypothesized that, upon splitting the FSQ into its component scales (WPI and SSS) both the WPI and the SSS would positively predict follow-up DAS28-CRP. Multivariable linear regression (Table 3) demonstrated, however, that the WPI was the only component of the FSQ score that significantly predicted follow-up DAS28-CRP (B 0.08, 95% CI 0.03 to 0.12; β 0.23, 95% Cl 0.09 to 0.37, P = 0.001 for model 3). Conversely, the SSS did not predict follow-up DAS28-CRP (B −0.03, 95% CI −0.10 to 0.04, β −0.06, 95% Cl −0.19 to 0.08, P = 0.43 for model 3). In comparing standardized coefficients of WPI and FSQ across equivalent models, the standardized coefficient of WPI was greater (WPI: β 0.23 vs FSQ: β 0.17 for model 3, respectively).
Multivariable linear regression model for the association between the components of FSQ (the WPI and SSS), at baseline and DAS28-CRP after 12 weeks of DMARD therapy (N = 192).
DISCUSSION
FM symptom severity, defined by the FSQ score, weakly predicts RA disease activity levels after patients initiate or switch DMARD therapy. Upon separating the FSQ into its 2 components, the WPI and the SSS, the WPI predicted RA disease activity, but the SSS did not. While the standardized coefficients for the SSS were close to zero with nonsignificant P values, the standardized coefficients for the WPI were statistically significant and greater than the standardized coefficients for the FSQ in equivalent models, providing evidence that the WPI is the major component of the FSQ that predicts RA disease activity. To our knowledge, this is the first longitudinal analysis to demonstrate that FSQ scores independently predict RA disease activity after treatment, and that the WPI is the primary source of the FSQ’s predictive value for DAS28-CRP score after DMARD treatment.
Our results are consistent with previous studies showing that, among patients with RA, a comorbid FM diagnosis is associated with higher RA disease activity.19-23 In particular, previous studies have shown that the patient-reported components of RA disease activity scores (ie, TJC and PtGA) are elevated for patients who have both RA and FM relative to patients with RA who do not have concomitant FM. Conversely, the clinician-reported and laboratory components of RA disease activity scores (ie, SJC and CRP/erythrocyte sedimentation rate) are typically equivalent among patients with RA with and without FM.20-23 Likewise, Durán et al found that patients with RA and concomitant FM are less likely to reach DAS28-CRP remission upon treatment, with persistently elevated TJC—the sole patient-reported variable of DAS28-CRP3—preventing DAS28-CRP remission.22 Andersson et al similarly showed that over a 5-year period, persistently elevated TJCs and PtGAs led to higher DAS28-CRP scores among patients with RA who have chronic widespread pain, a diagnosis on the spectrum of FM.24
Our study builds upon these studies by using the FSQ, a continuous measure of the severity of FM symptoms, rather than the binary outcome of physician-diagnosed FM. Previous studies suggest that FM may be more appropriately conceived as a condition that spans a continuum of severity, rather than as a discrete, binary diagnosis.25-28 Wolfe et al reported in a cross-sectional analysis that with increasing FSQ scores, the severity of common clinical indicators of RA disease activity likewise increases, though the relationship is stronger with patient-reported rather than clinician-reported variables.29 In terms of the relationship between FSQ and long-term outcomes, Kim et al showed that, among patients with RA, increased FSQ scores at baseline are independently predictive of worse Multidimensional Health Assessment Questionnaire scores, a measure of functional status, 2 years later.30 Our study contributes additional information by demonstrating that baseline FSQ scores predict DAS28-CRP scores approximately 12 weeks after initiating or changing DMARD treatment.
This study also builds upon previous analyses from the CPIRA study, in which we evaluated the ability of QST to predict treatment response.12 Inefficient conditioned pain modulation (CPM), a measure of endogenous analgesia, was associated with significantly lower odds of good treatment response. These results are consistent with findings from our current study because abnormalities in endogenous analgesia have been implicated in the pathogenesis of FM.31,32 In subsequent analyses of CPIRA data, however, we found no association between FSQ and CPM, and only weak correlations between FSQ and other QST measures.33 These results suggest that the FSQ likely reflects different aspects of the FM experience than the abnormalities in endogenous analgesia assessed by CPM.
The predictive ability of FSQ does not seem limited to RA. A few studies have also reported that the FSQ predicts pain-related outcomes after surgery. A study by Cheng et al indicated that greater FSQ scores are predictive of worse quality of recovery from shoulder arthroscopy.34 Other studies revealed higher FSQ scores to be associated with postoperative opioid use after a hysterectomy and lower-extremity joint arthroplasty.35,36 Taken together, these studies suggest that the pathway linking the FSQ to treatment outcomes is likely related to its ability to predict individuals’ pain experience, as opposed to specific disease mechanisms.
In addition to analyzing the FSQ as a whole, this study was the first to examine associations between the 2 scales that compose the FSQ (ie, the WPI and the SSS) and RA treatment outcomes. The WPI was the main source of the FSQ’s predictive value for DAS28-CRP, whereas the SSS did not contribute to FSQ’s prediction of RA disease activity after treatment. One potential implication is that the WPI better captures CNS regulatory processes that do not respond to the antiinflammatory effects of DMARDs. Alternatively, we must also consider the possibility that the WPI and DAS28-CRP do not have sufficient specificity to distinguish between abnormalities in CNS regulation associated with noninflammatory muscle pain vs inflammatory joint pain. For example, patients may indicate that they have pain in the forearm on the WPI, when the pain is actually due to inflammation at the wrist. Conversely, patients may be tender at joint sites because of widespread pain sensitivity due to FM, rather than joint inflammation. Further investigation is warranted to determine if there is overlap or conflation between joint and nonjoint pain sites in patients with RA and, if so, whether the WPI and/or DAS28-CRP may warrant modification.
Our study has many strengths. First, we assessed the severity of FM symptoms as a continuous variable rather than simply the binary outcome of physician-diagnosed FM. There was a wide range of baseline FSQ scores (0-28 out of a possible range of 0-31), enabling assessment of a per-unit increase across a sufficient range of FSQ. Second, the FSQ is based on the 2010 modified ACR diagnostic criteria for FM, which is a validated measure for diagnosing FM.13 The use of a validated, standardized assessment is important because FM is frequently overdiagnosed when the diagnosis is based primarily on physician gestalt.37 Third, our study goes beyond examining the FSQ’s association with RA disease activity by investigating the distinctive relationships of the 2 components of the FSQ with RA disease activity. Last, our study incorporated multivariable analyses, enabling adjustment for confounders. Unlike many prior analyses, we included baseline DAS28-CRP as a covariate. This decision was based on previous literature indicating that (1) FM is cross-sectionally associated with elevated DAS28-CRP,20,22,23 and (2) higher DAS28 scores at baseline are predictive of elevated DAS28 scores in the future.3,38,39 Thus, excluding baseline DAS28-CRP scores may lead to an overestimation of FSQ’s relationship to treatment outcomes.
The study also has several limitations. First, the regression coefficient for the association between baseline FSQ and follow-up DAS28-CRP was small, with a 1-unit difference in FSQ predicting a 0.04 difference in follow-up DAS28-CRP. Second, the statistical models have a notable amount of unexplained variation, with R2 = 0.41 for the most robust model in the primary analysis. The degree of unexplained variation is consistent with the absence of strong predictors for RA treatment outcomes in the literature.3-6 In addition, composite measures, such as the FSQ and DAS28-CRP, likely capture signs and symptoms from heterogenous pathologies, contributing to variance. We also did not control for the potential initiation of nonpharmacologic treatments, such as mental health care or exercise, which have beneficial effects on chronic pain outcomes that may be partially captured in the TJC and PtGA components of the DAS28-CRP. Further, patients received heterogenous RA therapies with differing mechanisms of action. Last, our study included only 2 timepoints over a 12-week period. A study of longer duration with more timepoints would provide added confidence in the observed changes.
The findings of this study have notable clinical implications. First, this study provides further evidence that noninflammatory processes affect DAS28-CRP, a tool designed to assess the activity of an inflammatory disease. Given the influence of FM symptoms on RA disease activity as measured by DAS28-CRP, accounting for the effect of FM symptoms on RA treatment outcomes should be considered. Second, the easily administered FSQ can contribute to a more informed prognosis of patients’ RA. The results, moreover, suggest that specific FM symptoms have prognostic implications. A patient with high FSQ scores due to widespread pain may be more likely to have high RA disease activity after DMARD initiation than a patient with high FSQ scores due to fatigue and cognitive symptoms. These patients may benefit from adjunctive pharmacologic and nonpharmacologic therapies to address FM symptoms, such as tricyclic antidepressants, serotonin and norepinephrine reuptake inhibitors, exercise, and cognitive behavioral therapy. Healthcare providers should consider using the FSQ, in particular the WPI component of the FSQ, when developing personalized care plans for patients with RA.
Footnotes
The CPIRA study was supported by the National Institutes of Health (NIH)/National Institute of Arthritis and Musculoskeletal and Skin Diseases (NIAMS) R01 AR064850. YCL, LNM, and JS were supported by NIH/NIAMS P30 AR072579. COB receives support from the NIH (AR070254).
YCL has received research support from Pfizer, consulted for Sanofi Genzyme (< $10,000), and has stock in Cigna. MBB has received grant funding from the Rheumatology Research Foundation, clinical trial support from Genentech, and honoraria from the American Board of Internal Medicine and Merck Manual; she is an associate editor for PracticeUpdate and American College of Rheumatology/Association of Rheumatology Professionals Advanced Rheumatology Course. The remaining authors declare no conflicts of interest relevant to this article.
- Accepted for publication October 17, 2022.
- Copyright © 2023 by the Journal of Rheumatology
