Arthritis & Rheumatism, Volume 62,
November 2010 Abstract Supplement

Abstracts of the American College of
Rheumatology/Association of Rheumatology Health Professionals
Annual Scientific Meeting
Atlanta, Georgia November 6-11, 2010.

Associations between Composite Measures of Multi-Joint Osteoarthritis, Gait Speed, and Health Assessment Questionnaire Scores: The Johnston County Osteoarthritis Project.

Nelson6,  Amanda E., DeVellis5,  Robert F., Renner5,  Jordan B., Purser3,  Jama, Schwartz5,  Todd A., Conaghan1,  Philip G., Kraus2,  Virginia Byers

Chapel Allerton Hospital, Leeds, United Kingdom
Duke Univ Med Ctr, Durham, Durham, NC
Duke Univ Med Ctr, Durham, NC
Thurston Arthritis Research Center, Univ of North Carolina, Chapel Hill, Chapel Hill, NC
Thurston Arthritis Research Center, Univ of North Carolina, Chapel Hill, NC
UNC School of Medicine, Chapel Hill, NC


As there is no widely accepted definition of multi-joint osteoarthritis (OA), it is difficult to quantify the effect of multiple joint involvement on OA outcomes. We used factor analysis to create composite variables reflecting multi-joint radiographic OA, and determined associations between these variables and systemic outcomes (gait speed and health assessment questionnaire [HAQ] scores).


Data were from a cross-sectional sample of the Johnston County OA Project, including individuals with multi-joint radiographs, HAQ, and gait speed (n=1350, 66% women, 33% African Americans, mean age 67 ± 10 years, mean body mass index [BMI] 31 ± 6 kg/m2). HAQ scores were categorized as 0, 0 to < 1, or >= 1. Mean seconds to complete an 8-foot walk were converted to gait speed in m/s. Radiographs of the bilateral hands, tibiofemoral joints (TFJ), and hips were read for Kellgren-Lawrence grade (KL 0–4) at each joint. Lateral lumbosacral spine (LS) films were read for osteophytes (OST 0–3) and disc space narrowing (DN 0–3) at 5 levels, and patellofemoral joint (PFJ) radiographs were read for OST (0–3), using the Burnett atlas. During factor analysis using these scores, the hip and LS variables did not load onto factors and were assessed separately (Hip OA if KL >= 2 at either hip, LS OA if OST and DN >= 1 at a single level). Linear and partial proportional odds regression models for gait speed and HAQ, respectively, were used to determine associations between OA variables and outcomes, adjusting for age, BMI, gender, and race.


Factor analysis produced 3 factors. The 1st consisted of the 1st interphalangeal (IP) joints, distal IP joints 2–5, and proximal IP joints 2–5 (IP factor, a=0.96); the 2nd included the metacarpophalangeal joints 2–5 (MCP factor, a=0.81); the 3rd included TFJs and PFJs (Knee factor, a=0.87). The mean gait speed in the sample was 0.7 ± 0.2 m/s. Proportions by HAQ categories were 30% (0), 24% (> 0 to < 1) and 46% (>= 1). Gait speed was negatively associated with hip OA and the 3 factors, but the associations were greatly attenuated after adjustment (Table). For HAQ, a 1-SD increase in either the IP or knee factor resulted in 18% and 16% increased odds, respectively, of being in a higher HAQ category after adjustment (IP factor cumulative odds ratio [cOR]=1.18 [95% CI 1.03–1.36]; knee factor cOR 1.16 [95% CI 1.03–1.30]).

Association between composite factors representing multi-joint OA and functional outcomes.

 HAQaGait speedb
OA variablecUnadjusteddAdjustedeUnadjusteddAdjustede
 cOR95% CIcOR95% CIBeta95% CIBeta95% CI
IP factor1.451.29, 1.641.181.03, 1.36-0.025-0.038, -0.012-0.003-0.016, 0.010
MCP factor0.940.84, 1.061.15f0.96, 1.35-0.014-0.026, -0.001-0.006-0.018, 0.006
Knee factor1.241.11, 1.391.161.03, 1.30-0.030-0.043, -0.018-0.010-0.022, 0.002
Hip OA1.230.98, 1.531.120.89, 1.40-0.044-0.070, -0.019-0.015-0.038, 0.009
LS OA1.110.90, 1.371.020.82, 1.27-0.016-0.041, 0.006-0.001-0.024, 0.022
HAQ: Health Assessment Questionnaire; cOR: cumulative odds ratio; CI: confidence interval; DIP: distal interphalangeal joint; PIP: proximal IP joint; MCP: metacarpophalangeal joint; TFJ: bibotemoral joint; PFJ: patellofemoral joint; Hip OA-KL grade 82 in either hip; LS OA-Lumbosacral spine OA, defined as an osteophyte 81 and disc narrowing at the same level.
a.Partial proportional odds regression model
b.Linear regression model
c.Factor scores are the standardized average of the radiographic scores of the included joints (average/std deviation)
d.Unadjusted results are for aI OA variables only.
e.Adjusted for age, BMI, gender, and race, with all OA variables in the model.
f.Factor 2 does not meet proportional odds assumption for categories of HAQ: the given OR is for the lowest category of HAQ (0) vs. the higher 2, for the lower 2 vs. the highest, OR 0.93 (95% CI: 0.81, 1.05)


Composite factors representing multi-joint OA involvement were associated with reduced gait speed and higher HAQ scores, although the estimates were attenuated by covariates, particularly age and BMI. Such composite factors can represent multiple joints in a single variable, reducing dimensionality and allowing more precise estimation of the effects of multi-joint OA involvement in models.

To cite this abstract, please use the following information:
Nelson, Amanda E., DeVellis, Robert F., Renner, Jordan B., Purser, Jama, Schwartz, Todd A., Conaghan, Philip G., et al; Associations between Composite Measures of Multi-Joint Osteoarthritis, Gait Speed, and Health Assessment Questionnaire Scores: The Johnston County Osteoarthritis Project. [abstract]. Arthritis Rheum 2010;62 Suppl 10 :181
DOI: 10.1002/art.27950

Abstract Supplement

Meeting Menu