Indicators of Walking Speed in Rheumatoid Arthritis: Relative Influence of Articular, Psychosocial, and Body Composition Characteristics

Abstract

Objective

To explore the contributions from and interactions between articular swelling and damage, psychosocial factors, and body composition characteristics on walking speed in rheumatoid arthritis (RA).

Methods

RA patients underwent the timed 400 meter long-corridor walk. Demographics, self-reported levels of depressive symptoms and fatigue, RA characteristics, and body composition [using whole-body dual-energy X-ray absorptiometry (DXA), and abdominal and thigh computed tomography (CT)] were assessed and their associations with walking speed explored.

Results

A total of 132 RA patients had data for the 400 meter walk, among whom 107 (81%) completed the full 400 meters. Significant multivariable indicators of slower walking speed were older age, higher depression scores, higher reported pain and fatigue, higher swollen and replaced joint counts, higher cumulative prednisone exposure, non-treatment with disease-modifying anti-rheumatic drugs (DMARDs), and worse body composition. These features accounted for 60% of the modeled variability in walking speed. Among specific articular features, slower walking speed was primarily correlated with large/medium lower-extremity joint involvement. However, these articular features accounted for only 21% of the explainable variability in walking speed. Having any relevant articular characteristics was associated with a 20% lower walking speed among those with worse body composition (p<0.001) compared with only a 6% lower speed among those with better body composition (p-value for interaction=0.010).

Conclusions

Psychosocial factors and body composition are potentially reversible contributors to walking speed in RA. Relative to articular disease activity and damage, non-articular indicators were collectively more potent indicators of an individual's mobility limitations.

INTRODUCTION

Rheumatoid arthritis (RA) is a debilitating systemic inflammatory disease associated with polyarthritis, progressive joint damage, and physical impairment (1). A variety of self-reported assessments of disability in activities of daily living (ADLs) are frequently used in RA clinical practice and trials. However, in the general population, objective measures of performance have been shown to be powerful predictors of disability onset (2-4), and survival (5). In particular, slower walking speed discriminated future risk of adverse events even among those who appeared healthy and high functioning on self-reported measures (6).

The determinants of physical limitation in RA are likely multifactorial, with the greatest contribution presumably from articular characteristics that may be reversible (i.e. joint pain and stiffness from synovitis) and irreversible (i.e. joint deformity from erosion, cartilage loss, joint instability and subluxation). Medium and large joints in addition to overall disease activity contributed to self-reported disability as assessed with the HAQ (7-9). However, there has been little investigation into what specific reversible and irreversible articular characteristics contribute the most to performance limitation in RA.

The assumption that articular features are the largest contributor to mobility limitation in RA may be unfounded, as other characteristics, such as generalized pain, depression, and fatigue, are also potential contributors (10-12). Body composition may also contribute, since higher fat mass and lower lean mass have been associated with higher disability scores in RA (13), and higher thigh fat and lower thigh muscle density were associated with performance limitation in RA patients (14). Fortunately, body composition may be a modifiable risk factor, as RA patients had improved body composition and faster walking speeds after resistance training (15). Whether optimizing reversible contributors (i.e. optimizing body composition) has the ability to lessen the impact of irreversible contributors on physical functioning has not been explored.

Accordingly, we cross-sectionally explored the independent contributions of total joint counts, psychosocial, and body composition characteristics on objective measures of mobility among RA patients enrolled in a longitudinal cohort study. An additional aim was to parse out the individual contributions of articular size, location, and features (i.e. swelling, tenderness, deformity, and replacement) on walking speed, particularly in light of the potential modifying effects of body fat and muscle composition. We hypothesized that psychosocial and body composition characteristics would be at least as strongly associated with mobility as total joint counts. Additionally, we hypothesized that slower walking speed would be influenced more strongly by abnormalities in medium and large lower extremity joints than small joints of the upper extremities, and that the impact of such joint damage and deformity on physical function would be attenuated in those with better body composition independent of contributing psychosocial factors.

METHODS

Study Participants

Participants were enrolled in the Evaluation of Subclinical Cardiovascular Disease and Predictors of Events in Rheumatoid Arthritis (ESCAPE RA) study, a longitudinal cohort study conducted and IRB approved at The Johns Hopkins Hospital. All participants were 45-84 years of age at enrollment, met the American College of Rheumatology 1987 criteria for RA (16), and had no reported history of cardiovascular disease. Details of recruitment and baseline characteristics have been described previously (17). Of the 197 recruited participants, 158 (80%) returned for physical performance testing at the third study visit, an average of 39 ± 4 months post-baseline. Data described in this report are from the third study visit. There were no significant differences in baseline demographic, lifestyle, or RA characteristics between those returning vs. not returning for study visit 3. All participants provided written informed consent prior to study procedures.

Assessments

Physical Performance and Self-Reported Disability Outcomes

Participants underwent the timed 400-meter (400m) walk test. Non-ambulatory patients were excluded. Participants could opt-out and those who attempted but could not complete 400m were recorded. Participants made 10 laps along a 20m course measured out in a long hallway (40m per lap). Participants were instructed to walk quickly at a maintainable pace, without running. Safety exclusions were as previously described (19). Due to a technical issue, 26 participants who initiated the 400m walk did not have usable data. An average walking speed in meters/second was recorded for all 107 400m walk completers. Self-reported disability was quantified using the Stanford Health Assessment Questionnaire (HAQ) (20), the physical functioning domain of the Short Form 36 Questionnaire (SF36-PF) (21), and by the Valued Life Activates (VLA) questionnaire (22).

Other Assessments

Forty-four joints were assessed for tenderness, swelling, deformity, and replacement by two trained and experienced joint assessors, of which twenty-eight were included in the three item disease activity score with C-reactive protein (DAS-28 CRP) (23). Joints were designated as deformed (yes vs. no deformity) based on typical RA deformities of articular subluxation, impaired range of motion, fixed flexion contracture, and non-reducible deviation, with the exam standardized prior to study initiation. Plain radiographs of the hands and feet were obtained and Sharp van der Heijde Scores (SHS) calculated (24). Current medications and their doses, including disease modifying anti-rheumatic drugs (DMARDs) and prednisone were recorded. Depressive symptoms were assessed with the Center for Epidemiology Studies Depression Scale (CESD) (25). Fatigue was measured using the Functional Assessment of Chronic Illness Therapy (FACIT) fatigue scale (26). Pain was assessed by visual analogue scale (VAS). Patients also underwent chest multi-detector row computed tomography (MDCT), which was assessed for the presence of interstitial lung disease (ILD), as previously described (27).

Height and weight were recorded and body mass index (meters/kg²) was calculated. Abdominal and mid-thigh CT scans were performed on an Aquilion 64-sclice CT scanner, and were analyzed by trained assessors (A.L. for abdominal measures and H.R.K. for thigh measures) using TomoVision sliceOmatic software to quantify total abdominal, subcutaneous, visceral, intramuscular, and thigh fat areas, abdominal muscle area and density, and thigh muscle area and density (14). Whole-body dual-energy X-ray absorptiometry (DXA) scans were performed on a Lunar Prodigy scanner, as described previously (13).

Laboratory Assessments

Serum inflammatory markers were measured from fasting samples. Interleukin-6 (IL-6) was measured by chemiluminescent enzyme immunoassay. CRP was assessed by nephelometry. Rheumatoid factor (RF) was measured by enzyme-linked immunosorbent assay (ELISA), with ≥40 units defining seropositivity. Anti-cyclic-cirtrullinated-peptide (anti-CCP) was assessed by ELISA with ≥60 units defining seropositivity. Shared epitope alleles were determined as previously described (28).

Statistical Analysis

Variable distributions were examined and compared according to subgroups using t-tests for normally distributed continuous variables, the Kruskal-Wallis test for non-normal continuous variables, and the chi-square or Fisher's exact test, as appropriate, for categorical variables. In order to condense the various body composition variables into a single variable reflecting overall body composition, we constructed a propensity score model that included the CT-derived variables abdominal muscle density, abdominal muscle area, abdominal intramuscular fat, abdominal fat area, thigh muscle area, thigh muscle density, and the DXA-derived variables total fat mass index, and appendicular lean mass. Next, associations of patient characteristics with walking speed were explored in univariate linear regression models. Where required, variables were normally transformed. Multivariable prediction models included covariates carried over from univariate models with p≤0.20 (to allow for residual confounding). More parsimonious multivariable models excluded covariates with the weakest associations with the outcome, with the impact of excluding each covariate tested using Akaike's Information Criterion (AIC) for nested models. The adjusted coefficient of variability (R²) was used to estimate the total proportion of the variability in the outcome predicted by the modeled covariates. Checks for normality assumptions and model fit were tested using the Shapiro-Wilk test and graphical plots of studentized post-test residuals. Variance inflation factors were calculated to ensure that covariates with excessive collinearity were not co-modeled.

We next explored the independent associations of articular size, location, and features (i.e. swelling, tenderness, deformity, and replacement) on walking speed by modeling specific joint characteristics in place of total joint counts modeled previously. In additional models, heterogeneity in the association of articular features with performance measures according to strata of body composition (with better body composition defined by being in the top tertile of body composition propensity) was explored by introducing stratum x articular feature interaction terms into linear regression models.

Statistical calculations were performed using Intercooled Stata 12 (StataCorp, College Station, TX). A two-tailed α=0.05 was used.

RESULTS

Among the 158 participants, 132 (84%) had 400m walk data. Of these 132, 16 (12%) did not attempt the walk, 9 (7%) did not complete the entire 400m, and 107 (81%) completed the entire assessment. Among completers, the average walking speed was 0.95±0.18 m/s (range 0.56-1.32 m/s) and 26% walked<0.8 m/s. Patient characteristics according to those who underwent the SPPB and the subgroup that had 400m walk data are summarized in Table 1. On average, participants were middle-aged or older, female, Caucasian, and the majority had some college education. Reported depressive symptoms were generally low. Median RA duration was 12 years and most were seropositive for either RF or anti-CCP. On average, disease activity was in the low to moderate range. For treatment, the majority (87%) were receiving nonbiologic DMARDs, almost half were treated with biologic DMARDs (predominantly TNF inhibitors), and a quarter were treated with prednisone. The sub-group with 400m walk data did not significantly differ from the group without usable 400m walk data with the exception of a slightly lower CES-D score, slightly higher frequency of prednisone use, and a slightly lower frequency of TNF inhibitor use. Compared with those without usable 400m walk data, completers of the 400m walk (n=107) differed significantly only in having lower CES-D and SHS scores.

Table 1.

Participant Characteristics According to Completion Status of the 400 Meter Long-Corridor Walk Test

	Total Group	No 400m Walk Data	400m Walk Sub-group*		400m Walk Completers†
	(n=158)	(n=26)	(n=132)	p-value**	(n=107)	p-value‡
Age, years (Range 47-84)	62 ± 8	64 ± 8	62 ± 8	0.34	61 ± 8	0.10
Male gender, n (%)	58 (37)	10 (38)	48 (36)	0.84	43 (40)	0.87
Caucasian, n (%)	138 (87)	21 (81)	117 (89)	0.27	94 (88)	0.34
Any college, n (%)	123 (78)	18 (69)	105 (80)	0.25	87 (81)	0.18
Exercise, minutes/day	31 (9-76)	23 (4-69)	34 (9-77)	0.46	47 (13-81)	0.20
TV watching, minutes/day	120 (60-180)	120 (77-240)	120 (60-180)	0.43	120 (51-180)	0.21
CES-Depression score	6 (3-10)	9 (6-13)	5 (2-10)	0.005	5 (2-9)	0.001
Ever smoking, n (%)	89 (56)	15 (58)	74 (56)	0.88	60 (56)	0.88
Current smoking, n (%)	13 (8)	3 (12)	10 (8)	0.50	9 (8)	0.62
Any CT-ILD, n (%)	42 (30)	5 (22)	37 (31)	0.37	29 (30)	0.42
BMI, kg/m2	28.3 ± 5.3	27.8 ± 4.5	28.5 ± 5.5	0.48	28.2 ± 5.3	0.83
RA duration, years	12 (7-20)	12 (7-20)	12 (7-20)	0.77	10 (7-18)	0.49
RF or anti-CCP seropositivity, n (%)	117 (75)	18 (69)	103 (79)	0.80	81 (76)	0.27
Any shared epitope alleles, n (%)	110 (71)	18 (69)	92 (71)	0.88	71 (68)	0.88
DAS28-CRP	3.1 (2.3-4.1)	2.9 (1.9-3.5)	3.1 (2.3-4.1)	0.51	2.9 (2.1-3.9)	0.98
CRP, mg/L	2.4 (0.9-5.7)	2.1 (0.6-5.2)	2.4 (0.9-6.2)	0.49	2.0 (0.8-4.8)	0.99
IL-6, pg/mL	4.2 (2.5-9.1)	3.3 (2.0-8.3)	4.2 (2.5-9.8)	0.20	3.8 (2.3-8.0)	0.50
Total SHS	15 (3-48)	18 (8-48)	12 (2-50)	0.20	9 (1-29)	0.035
Pain (100mm VAS)	20 (10-50)	20 (10-50)	20 (10-50)	0.99	20 (10-50)	0.66
AM stiffness, minutes	10 (5-30)	10 (2-30)	15 (5-30)	0.25	10 (3-30)	0.48
FACIT Score	8 (4-17)	7 (5-13)	8 (4-18)	0.55	7 (3-16)	0.92
HAQ (0 – 3)	0.62 (0.12-1.38)	0.56 (0.25-1.38)	0.75 (0.12-1.38)	0.98	0.50 (0-1.12)	0.28
SF-36 Physical Performance Score	65 (45-85)	70 (45-85)	65 (42-88)	0.86	80 (50-90)	0.41
Physical Performance Battery Scr. (0-16)	11 (8-13)	9 (8-12)	11 (7-14)	0.22	11 (9-14)	0.15
Proportion of VLAs Affected	0.30 (0.05-0.71)	0.20 (0.05-0.52)	0.31 (0.05-0.72)	0.39	0.23 (0-0.61)	0.88
Current prednisone, n (%)	39 (25)	3 (12)	36 (27)	0.089	34 (32)	0.19
Cumulative prednisone, grams	4.1 (0.4-11.7)	4.8 (0-9.5)	4.1 (0.4-12.7)	0.66	3.2 (0.4-10.4)	0.96
Current non-biologic DMARDs, n (%)	137 (87)	21 (81)	116 (88)	0.33	92 (86)	0.51
Current biologic DMARDs, n (%)	75 (47)	14 (54)	61 (46)	0.48	51 (48)	0.57
TNF inhibitors, n (%)	57 (36)	14 (54)	43 (33)	0.039	38 (36)	0.086

Data expressed as mean ± standard deviation or median (interquartile range) unless otherwise noted.

^*400m Walk Sub-group includes completers and those who did not attempt or complete the walk. See table 2 for characteristics according to subgroups of this group.

^**p-values represent comparisons between the groups with vs. without 400m walk data

^†Group excludes participants who could not initiate or complete the 400m walk test

^‡p-values represent the comparison with the group without 400m walk data

TV=television; CES=Center for Epidemiologic Studies; CT=computed tomography; ILD=interstitial lung disease; RA=rheumatoid arthritis; RF=rheumatoid factor; CCP=cyclic citrullinated peptide; DAS=disease activity score; CRP=C-reactive protein; IL=interleukin; SHS=Sharp-van der Heidje score; VAS=visual analogue scale; HAQ=Health Assessment Questionnaire; SF=Short Form; m/s=meters per second; VLA=Valued Life Activities; DMARD=disease modifying anti-rheumatic drug; TNF=tumor necrosis factor; BMI=body mass index

Body composition characteristics according to tertiles of the calculated body composition propensity score are summarized in Supplemental Table 1. The score discriminated each group on all the body composition variables, such that those in the highest tertile had significantly higher levels for all muscle measures and lower levels for all adipose measures (including those not specifically included in the model) compared with the lowest tertile.

Indicators of Not Attempting or Completing the 400m Long Corridor Walk

Patient characteristics according to completion status of the 400m walk are summarized in Table 2. Compared with those in the fastest tertile (0.99-1.32 m/s), those not attempting/completing were significantly older, more likely to be female, reported fewer minutes of intentional exercise, and more minutes of TV watching. The median CES-D score was more than 3-fold higher among those not attempting/completing the 400m walk compared with those in the fastest tertile (p<0.001). Those not attempting/competing the 400m walk had a significantly longer median RA duration (by more than 10 years) and higher levels of RA disease activity (as manifested by higher DAS28 scores, CRP, and IL-6), disease severity (as manifested by higher SHS scores), and symptoms (as manifested by higher reported pain scores, more minutes of AM stiffness, and higher FACIT fatigue scores). In addition, self-reported disability for all assessment tools was significantly higher among those not attempting/completing the walk compared with those in the fastest tertile, and the group also performed more poorly on the other objective physical performance measures tested. Among RA therapies, those not attempting/completing the 400m walk were more frequently treated with prednisone compared with the fastest tertile; however, other RA treatments did not differ. There was a striking difference in the propensity for better body composition between those not attempting/completing the 400m walk compared with the fastest tertile, as the median propensity for better body composition was only 0.10 in the former group vs. 0.76 in the later (p<0.001).

Table 2.

Characteristics of the 132 Participants with 400 Meter Walk Data, According to 400 Meter Walk Speed

	Did not attempt/Could not complete (X) (n = 25)	Slowest Tertile (T1) [0.56-0.89 m/s] (n=36)	Middle Tertile (T2) [0.90-0.98 m/s] (n=36)	Fastest Tertile (T3) [0.99-1.32 m/s] (n=35)	T3 vs. X p-value	T3 vs. T1 p-value
Age, years (Range 47-84)	67 ± 8	64 ± 9	61 ± 7	58 ± 7	<0.001	0.002
Male gender, n (%)	5 (20)	11 (31)	15 (42)	17 (49)	0.024	0.12
Caucasian, n (%)	23 (92)	31 (86)	30 (83)	33 (94)	0.99	0.43
Any college, n (%)	18 (72)	28 (78)	30 (83)	29(83)	0.31	0.59
Exercise, minutes/day	13 (4-26)	32 (6-74)	63 (24-99)	43 (9-77)	0.018	0.44
TV watching, minutes/day	180 (86-240)	120 (51-180)	120 (64-180)	69 (36-120)	0.003	0.18
CES-Depression score	10 (6-15)	5.5 (4-10.5)	4 (2-6.5)	3 (1-6)	<0.001	0.005
Ever smoking, n (%)	14 (56)	23 (64)	20 (56)	17 (49)	0.57	0.19
Current smoking, n (%)	1 (4)	4 (11)	3 (8)	2 (6)	0.99	0.67
CT-ILD Score (range 0-32), units	0 (0-2)	0 (0-2)	0 (0-2)	0 (0-0)	0.29	0.15
RA duration, years	20 (7-24)	11 (9-18)	13 (7-19)	9 (6-16)	0.027	0.092
RF or anti-cCP seropositivity, n (%)	22 (92)	27 (75)	29 (81)	25 (71)	0.098	0.73
Any shared epitope alleles, n (%)	21 (84)	27 (79)	21 (58)	23 (66)	0.15	0.20
DAS28-CRP	3.8 (3.0-4.6)	3.6 (2.9-4.5)	2.6 (2.1-3.3)	2.4 (1.9-3.1)	<0.001	<0.001
CRP, mg/L	7.3 (2.4-9.3)	2.4 (0.9-5.3)	2.5 (0.9-4.6)	1.6 (0.5-4.9)	<0.001	0.24
IL-6, pg/mL	10.1 (3.8-17.4)	6.9 (2.5-10.4)	3.2 (2.2-8.7)	3.0 (2.5-4.8)	<0.001	0.038
Total SHS	66 (22-132)	19 (4-59)	11 (2-30)	4 (0-13)	<0.001	0.004
Pain (100mm VAS)	40 (20-60)	30 (20-70)	15 (10-35)	10 (0-40)	0.003	0.002
AM stiffness, minutes	30 (8-105)	15 (5-45)	10 (2-30)	10 (2-30)	0.002	0.03
FACIT Score	16 (9-19)	14 (7-22)	5.5 (4-10)	4 (2-11)	<0.001	<0.001
HAQ (0 – 3)	1.5 (1.1-1.8)	1.3 (0.8-1.6)	0.2 (0-0.9)	0.1 (0-0.5)	<0.001	<0.001
SF-36 Physical Performance Score	35 (25-45)	45 (33-65)	80 (65-88)	90 (80-95)	<0.001	<0.001
Short Physical Perf. Battery Scr. (0-16)	5 (3-9)	8 (7-10)	12.5 (11-14)	13 (12-15)	<0.001	<0.001
6 meter paced walk speed, m/s	0.97 (0.75-1.04) [n=19]	0.95 (0.83-1.07)	1.12 (1.00-1.22)	1.23 (1.06-1.36)	<0.001	<0.001
6 meter fast walk speed, m/s	1.17 (1.11-1.39) [n=19]	1.30 (1.12-1.43)	1.61 (1.47-1.73)	1.71 (1.56-1.84)	<0.001	<0.001
Proportion of VLAs Affected	0.79(0.59-1.0)	0.59 (0.27-0.75)	0.22 (0.03-0.50)	0 (0-0.21)	<0.001	<0.001
Proportion of Obligatory VLAs	0.80 (0.40-1.0)	0.40 (0.20-0.80)	0 (0-0.20)	0 (0-0)	<0.001	<0.001
Proportion of Committed VLAs	1.0 (0.75-1.0)	0.75 (0.45-1.0)	0.33 (0-0.66)	0 (0-0.33)	<0.001	<0.001
Proportion of Discretionary VLAs	0.80 (0.38-1.0)	0.66 (0.21-0.75)	0.21 (0-0.50)	0 (0-0.14)	<0.001	<0.001
Current prednisone, n (%)	11 (44)	12 (33)	7 (19)	6 (17)	0.023	0.12
Current non-biologic DMARDs, n (%)	24 (96)	32 (89)	31 (86)	29 (83)	0.22	0.51
Current biologic DMARDs, n (%)	10 (40)	21 (58)	14 (39)	16 (46)	0.66	0.29
TNF inhibitors, n (%)	5 (20)	17 (47)	10 (28)	11 (31)	0.32	0.17
Body composition propensity (0-1)	0.10 (0.01-0.40)	0.40 (0.07-0.57)	0.72 (0.45-0.87)	0.76 (0.49-0.93)	<0.001	<0.001

Data expressed as mean ± standard deviation or median (interquartile range) unless otherwise noted

See Table 1 for abbreviations

Univariate Indicators of Completing the 400m Walk Test

Among those who completed the 400m walk, there were univariate differences between those with slower average walking speeds (slowest tertile speed=0.56-0.89 m/s) and those with the fastest speed, albeit the differences were not as striking as the comparisons with the group that did not attempt or could not complete the walk (Table 2). Compared with the fastest tertile, those in the slowest tertile were, on average, 6 years older, and reported more depressive symptoms. Among RA characteristics, DAS28, IL-6, SHS, reported pain, stiffness, FACIT fatigue, and all subjective and objective disability measures were significantly higher for the 1^st vs. 3^rd tertile of walking speed; however, RA treatments did not differ between the groups. The median propensity for better body composition was also lower for the slowest vs. fastest tertiles of walk speed (0.40 vs. 0.76, respectively; p<0.001).

Multivariable Indicators of 400m Walking Speed

Multivariable indicators of 400m walking speed are summarized in Table 3. Because of robust associations between univariate indicators (correlation matrix provided as a Supplemental Table), many were not retained in multivariable models. Nine indicators (8 negative and one positive) were found to contribute significantly to model fit. Among the negative indicators, higher age, higher CESD score, higher swollen and replaced joint counts, higher reported pain and fatigue, higher cumulative prednisone exposure, and non-treatment with DMARDs were all associated independently with slower 400m walking speed. The largest negative indicator was higher age, which accounted for 27% of the explainable variability in 400m walking speed. Articular signs were the next strongest indicator, with replaced joints accounting for double the explainable variability of swollen joints (8.7% vs. 4.4%). The only positive indicator was the propensity for better body composition. Together, these indicators accounted for nearly 60% of the total variability in walking speed (adjusted R²=0.594; model F-test<0.001). In the final multivariable model, gender, the presence of ILD, RA duration, autoantibody status, circulating inflammatory marker levels, and SHS scores did not independently associate with walking speed. For the final model, a plot of predicted values on studentized residuals revealed random scatter around zero with no evidence of leveraging outliers (data not shown). The Shapiro-Wilk test p-value was 0.57, indicating that normality assumptions of linear regression were met.

Table 3.

Univariate and Multivariable Indicators of 400 Meter Walk Speed [400 m Walk Completers Only (n=107)]

	Univariate Crude Models		Multivariable Extended Model		Multivariable Reduced Model
Covariate	β	p	β	p	β	p	% total R2
Age, per year	−0.007	0.001	−0.009	<0.001	−0.009	<0.001	27.2%
Male vs. female	0.077	0.024	−0.033	0.22
Caucasian vs. other	0.050	0.34
Any college vs. none	0.014	0.75
Exercise, per daily 30 minutes	0.013	0.059	0.008	0.12
TV watching, per daily 30 minutes	−0.013	0.030	−0.005	0.29
Square root CES-D score, per unit	−0.056	<0.001	−0.022	0.084	−0.024	0.039	2.4%
Ever smoking vs. never	−0.037	0.28
Current smoking vs. non-smoking	−0.057	0.36
Square root CT-ILD Score	−0.039	0.062	0.013	0.46
RA duration, per year	−0.002	0.29
RF or anti-CCP seropositivity	−0.003	0.93
Any shared epitope alleles vs. none	−0.031	0.39
Square root DAS28-CRP	−0.185	<0.001
Square root swollen joints (0-44)	−0.054	<0.001	−0.030	0.037	−0.028	0.009	4.4%
Square root tender joint (0-44)	−0.031	0.002	0.011	0.29
Square root deformed joints	−0.046	0.001	−0.001	0.97
Any replaced joints	−0.142	<0.001	−0.076	0.015	−0.099	<0.001	8.7%
log CRP, per log mg/L	−0.018	0.16
log IL-6, per log pg/mL	−0.039	0.022	−0.013	0.35*
Square root Total SHS	−0.015	0.002	−0.004	0.42
Square root Pain (100 mm VAS)	−0.027	<0.001	−0.011	0.12	−0.011	0.046	2.2%
AM stiffness, per 10 minutes	−0.017	0.009	0.000	0.99
Square root FACIT Score	−0.050	<0.001	−0.032	0.013	−0.025	0.018	3.4%
Body composition propensity, per 10%	0.024	<0.001	0.009	0.046	0.008	0.037	1.2%
Current prednisone vs. none	−0.086	0.031	0.028	0.49
Square root cumulative prednisone, per gram	−0.020	0.018	−0.017	0.045	−0.010	0.093	1.3%
No DMARDs	−0.097	0.16	−0.158	0.002	−0.147	0.002	6.4%
Adjusted R2	N/A		0.570		0.594

β-coefficients represent the difference in 400m walking speed associated with a one unit change in the covariate. Univariate crude models are from individual linear regression models including only the covariate of interest. Multivariable models simultaneously include all of the covariates listed in the model. Adjusted R² is the proportion of the variability in the outcome explained by the covariates included in the model. % total R² is the proportion of the adjusted R² for the model lost when the covariate is excluded from the multivariable model. See Table 1 for abbreviations.

^*CRP not included in the extended model due to potential collinearity with IL-6. In an alternate extended model, CRP was not associated with walk speed when substituted for Il-6.

Articular Characteristics as Indicators of Walking Speed

Given the independent contribution of swollen and replaced joints to walking speed, we next explored whether specific articular characteristics (i.e. upper extremity vs. lower extremity, large joint vs. small joint, etc...) would correlate with 400m walking speed (Table 4). While many characteristics were inversely associated with walking speed in univariate models, only 6 (the presence of two or more swollen upper extremity large or medium joints, bilateral knee swelling, any hip tenderness, any ankle deformity, any hip replacement, and bilateral knee replacement) remained significantly and independently associated in multivariable modeling that also included adjustment for the other indicators identified in Table 4. Together, the articular characteristics and other indicators (i.e. age, pain, CES-D, FACIT-fatigue, cumulative prednisone, and non-treatment with DMARDs) accounted for nearly 61% of the total model variability for walking speed (R²=0.606; model F-test<0.001), with the articular features accounting for 21% of the explainable variability of the final multivariable model. At least one of these articular characteristics was present in 36% of the group who completed the 400m walk, but only 11 (10%) had two or more present. However, the impact of having two or more of these articular characteristics was not significantly different from having only one (Figure 1), and translated to an approximately 20% lower walking speed if any characteristic was present.

Table 4.

Univariate and Multivariable Joint Exam Indicators of 400 Meter Walk Speed [400 m Walk Completers Only (n=107)]

	Univariate Crude Models		Multivariable Adjusted Model*
	β	p	β	p
Regional Swollen Joints
Number of swollen upper extremity large/medium joints
None (n=84)	referent	--
1 (n=14)	−0.036	0.47
2 or more (n=8)	−0.182	0.005	−0.090	0.037
Number of swollen upper extremity small joints
None (n=42)	referent	--
1-7 (n=58)	−0.029	0.41
8 or more (n=7)	−0.20	0.004
Knee swelling
No knee swelling (n=75)	referent	---
Unilateral (n=25)	−0.020	0.63
Bilateral (n=7)	−0.118	0.090	−0.089	0.042
Ankles/Tarsi swelling
No ankle/tarsi swelling (n=69)	referent	---
Any ankle/tarsi swelling (n=38)	−0.080	0.024
Number of swollen lower extremity small joints, per joint	−0.016	0.13
Regional Tender Joints
Num. of tender upper extremity large/med joints, per joint	−0.043	0.002
Number of tender upper extremity small joints, per joint	−0.011	0.005
Hip tenderness
No hip tenderness (n=99)	referent	---
Any hip tenderness (n=8)	−0.165	0.010	−0.093	0.035
Knee tenderness
No knee tenderness (n=60)	referent	---
Any knee tenderness (n=46)	−0.080	0.019
Ankles/Tarsi tenderness
No ankle/tarsi tenderness (n=58)	referent	---
Unilateral (n=19)	−0.043	0.34
Bilateral (n=30)	−0.106	0.007
Number of tender lower extremity small joints, per joint	−0.006	0.20
Regional Deformed Joints
Num. deformed upper extremity large/med joints, per joint	−0.036	0.004
Number of deformed upper extremity small joints, per joint	−0.004	0.41
Hip deformity
No hip deformity (n=103)	referent	---
Any hip deformity (n=4)	−0.112	0.21
Knee deformity
No knee deformity (n=104)	referent	---
Any knee deformity (n=3)	−0.157	0.13
Ankles/Tarsi deformity
No ankle/tarsus deformity (n=95)	referent	---
Any ankle/tarsus deformity (n=12)	−0.132	0.013	−0.099	0.049
Number of deformed lower extremity small joints, per joint	−0.021	0.017
Regional Replaced Joints
Number of replaced upper extremity large/medium joints
None (n=97)	referent	--
Any (n=10)	−0.144	0.013
Number of replaced upper extremity small joints, per joint	−0.012	0.22
Hip replacement
No hip replacement (n=101)	referent	---
Any replacement (n=6)	−0.125	0.089	−0.111	0.044
Knee replacement
No knee replacement (n=92)	referent	---
Unilateral (n=4)	−0.080	0.36
Bilateral (n=11)	−0.182	0.001	−0.109	0.008
Ankle replacement
No ankle replacement (n=104)	referent	---
Unilateral (n=3)	0.008	0.94
Total Adjusted R2			0.606

^*Adjusted for age, CES-D score, pain score, FACIT fatigue score, body composition propensity, cumulative prednisone, and lack of DMARD treatment

β-coefficients represent the difference in 400m walking speed associated with a one unit change in the covariate. Univariate crude models are from individual linear regression models including only the co₂ variate of interest. Multivariable models simultaneously include all of the covariates listed in the model. Adjusted R² is the proportion of the variability in the outcome explained by the covariates included in the model.

Figure 1. Average 400 meter walking speed according to the presence of relevant articular characteristics.

The presence of any of the following relevant articular examination features (two or more swollen upper extremity large or medium joints, bilateral knee swelling, any hip tenderness, any ankle deformity, any hip replacement, and bilateral knee replacement) conferred a 20% reduction in average walk speed, with the impact of two or more joints not significantly different from having only one relevant articular feature present, even after adjusting for confounders (age, CES-D score, FACIT fatigue score, body composition propensity, cumulative prednisone, and lack of DMARD treatment).

Interaction of Body Composition and Articular Features on Walking Speed

We next explored whether strata of certain patient characteristics [gender, age greater than the median, depression (i.e. CES-D>16 units), fatigue dichotomized at the median, and better body composition (i.e. being in the top tertile of body composition propensity)] modified the associations of the specific articular features identified as indicators of 400m walking speed. Men and women with no relevant articular characteristics had the same adjusted mean walking speeds (0.97 m/s). However, among those with any relevant articular characteristics, women demonstrated a slower walking speed compared with men (adjusted mean walking speed=0.79 vs. 0.87 m/s, respectively; p-value for interaction=0.088, data not shown).

Better body composition was a robust modifier of the association of articular features with walking speed (Figure 2). Having any relevant articular characteristics was associated with a 20% lower walking speed among those in the first and second tertiles of body composition propensity (p<0.001) compared with only a 6% lower speed among those in the highest tertile of body composition propensity (p=0.015; p-value for interaction=0.010, Figure 2). Age, fatigue, and depression did not modify the association of articular characteristics with 400m walking speed (data not shown).

Figure 2. Average 400 meter walking speed according to the presence of relevant articular characteristics and strata of body composition propensity.

Any relevant articular features includes any of the following: bilateral knee swelling, bilateral ankle deformity, any deformed lower extremity small joints, and bilateral knee replacement. Best body composition was defined as being in the top tertile of body composition propensity. The effect of having any relevant articular features on walking speed was smaller (and non-significant) among the group in the better body composition stratum (p-value for interaction<0.05). Associations depicted were adjusted for Adjusted for age, CES-D score, pain score, FACIT fatigue score, cumulative prednisone, and lack of DMARD treatment.

DISCUSSION

To our knowledge, our report is the largest and most comprehensive study of the indicators of walking speed in RA. We detected several robust independent contributors including age, psychosocial characteristics (i.e. depression, fatigue, pain), articular features, and body composition, which together accounted for nearly 61% of the variation in walking speed. Interestingly, articular features accounted for only a small percentage of the variation in walking speed. Articular characteristics affecting walking speed were primarily large lower extremity joints, but large and medium upper extremity joint swelling was also a notable independent indicator. Faster walking speed was not associated with any small joint signs, radiographic joint damage in the hands and feet, or RA duration in multivariable models.

Only a handful of prior studies have explored walking speed in RA (29-32), and studies of the associations between articular features and self-reported disability have yielded divergent results (7-9, 33-34). Radiographic joint damage was correlated with reported disability in late but not early RA (33), whereas disease activity was shown to contribute to disability in early but not late RA (34). Importantly, these studies did not account for many of the important psychosocial and body composition characteristics that we found to influence walking speed, perhaps explaining the discrepancy with our study. Prior longitudinal studies tracking the progression of physical function limitation in elderly individuals have also identified prominent effects of depression and fatigue on performance (35-36). However, it is notable that these declines were noted in populations older than ours by a decade or more. In populations of comparable age with chronic conditions affecting mobility (i.e. lower extremity osteoarthritis), depression and fatigue were also prominent contributors to objective physical impairment (37-38), suggesting that our findings are not necessarily RA-specific. These findings may suggest that interventions to improve depression and fatigue in RA patients may also improve functional limitations to a degree equal or greater than improving RA disease activity.

Although total joint counts of swelling and tenderness are frequently used to measure articular involvement in RA, specific joint characteristics may contribute differently to mobility and performance than others. It is not surprising that joints involved in stance, gait, and initiating movements demonstrated the strongest associations with mobility limitation in our cohort. We also observed that hip and knee replacements were associated with slower walking speed. Based on prior research on knee arthroplasty, one would expect impaired walking speed in the early recovery process (39) but with improvement by six months postoperatively (40); however, our data may indicate persistent long-term impairment in RA patients. Both limited hip and knee flexion-and-extension have been associated with greater self-reported disability (41), which may be mechanistically linked to shorter stride length, poor gait stability, and consequently limited mobility.

Interestingly, we observed that the association of articular features with performance measures was minimal among the patients with better body composition, perhaps suggesting that lower total and regional adipose, higher muscle mass, and higher muscle density may be able to counteract the mobility limiting effects of joint pain and deformity. While these data may suggest that improving body composition could improve mobility and reduce disability, our study is limited by its cross-sectional design with no ability to infer causality. Interestingly, we did not observe the same ability of lower levels of reported depressive symptoms or fatigue to modify the association of joint activity and damage on performance, suggesting an independence of effects.

Our study has strengths and weaknesses. A strength was the extensive collection of patient characteristics, including a detailed joint exam, several measures of body composition, and numerous potential confounding variables, such as depression, fatigue and medication use. The body composition propensity score was helpful because it reduced a large number of body composition parameters into one descriptor that correlated well with each independent measurement. The score also restricted the ability to assess the contributions of individual body composition elements, which may identify particular targets for intervention. While beyond the scope of this analysis, investigation into the correlation between individual body composition parameters and performance is currently underway. Finally, physical performance data from an earlier time point could have determined if declining walking speed is a predictor of the onset of disability in ADLs in this cohort; however, these were only incorporated into the study for the final study visit.

In summary, we identified both irreversible and potentially reversible causes of mobility limitation in RA patients, including specific joint characteristics and poor body composition, and demonstrated that better body composition negated much of the detrimental impact of symptomatic joints on physical impairment. Lifestyle modifications to optimize body composition may be an important adjunct to intensive medical management with DMARDs, as recent investigations have shown that through structured physical training RA patients are able to alter body composition (15), decrease disability, and improve physical performance (42-43).

Significance and Innovations.

• - Walking speed in individuals with rheumatoid arthritis is influenced by potentially reversible characteristics such as psychosocial factors, body composition, and joint symptoms.

• - Collectively, non-articular characteristics including age, psychosocial factors, and medical therapies, had a greater effect on walking speed than articular characteristics.

• - Of the involved joints, large and medium lower extremity joints most significantly effected walking speed.

• - The negative impact of joint involvement on walking speed was minimized in patients with better overall body composition (i.e. higher amount and density of muscle, and lower amount of adipose).