s / Osteoarthritis and Cartilage 21 (2013) S63–S312 S263 a capsule per randomization schedule. After 2 hours subjects began a series of two additional SP walks followed by a high paced (HP) walk, at 30% faster pace than their SP walk. Prior to (t 1⁄4 0) and every 3 minutes after each walk initiation the subjects were queried for their current KOA PI score. Primary endpoint time weighted average pain scores on the treadmill (TWAP), used to estimate differences in average pain scores (APS) between P and N treatments. Secondary endpoints: included changes in WOMAC 3.1 VAS scores, Investigator Global Assessment of Disease Status (IGADS) and adverse events (AEs). Other assessments included surface joint temperature and algometry. Statistical analyses included a mixed effect model with fixed effects for period and treatment and random effect for subject with carryover investigated in a sensitivity analysis. For the analysis of WOMAC scores the pre-treatment baselines recorded in each period were included as two covariates representing the individual's average baseline and the difference from this average. Results: In randomized KOA subjects (N1⁄423), there were significant treatment effects during naproxen treatment vs placebo treatment at t 1⁄4 0 and during the treadmill walk, noted in APS by TWAP andWOMAC. The most sensitive endpoints, were t1⁄40 values (N vs P: APS -0.93(CI: -1.52, -0.34), p1⁄40.004, on 3rd SP walk. Increased effect size by TWAP was seen in the HP walks (N vs P: -1.02; (CI: -1.61, -0.43), p1⁄40.002). WOMAC Pain scores at the start of the study were 53.4 (sd 15.8). Total WOMAC and all subscale scores showed significant differences between naproxen and placebo. The treadmill walking scores maintained the APS treatment differences but did not capture additional (dynamic) analgesic signals. IGADS scores also reflected treatment differences. Conclusions: The modified walking model showed pain intensity differences between treatments at t1⁄40 and during treadmill in a relatively small KOA population. The modifications expedited enrollment and time to completion but limited interpretation of pain on walking. 508 PREDICTORS FOR THE PROGRESSION OF KNEE PAIN IN SUBJECTS WITH EARLY SYMPTOMATIC KNEE OSTEOARTHRITIS FIVE YEAR RESULTS FROM THE CHECK STUDY A.N. Bastick y, S. Verkleij y, J. Damen y, J. Wesseling z, P.J. Emans x, S.M. Bierma-Zeinstra y. 1 Erasmus Univ. Med. Ctr., Rotterdam, The Netherlands; Univ. Med. Ctr. Utrecht, Utrecht, The Netherlands; Univ. Hosp. Maastricht, Maastricht, The Netherlands Purpose: To determine which patientor disease characteristics can help predict future pain trajectories in patients with early symptomatic osteoarthritis of the knee (knee OA). Methods: The data were acquired from the Cohort Hip and Cohort Knee (CHECK) study, a prospective follow-up cohort of 1002 participants with early symptomatic knee and/or hip OA. We restricted ourselves to participants who had knee pain; knee OA according to the ACR criteria; and a completed follow-up period. Baseline characteristics were assessed. The outcome measure was pain assessed by the Visual Analogue Score (VAS) over a five year period. The participants were categorized into three distinct pain trajectories retrieved by latent class growth analysis. This resulted in a favourable; a stable; and an unfavourable course of pain group. Multinomial logistic regression analyses were performed univariately and multivariately to calculate odds ratios (ORs) and corresponding 95% confidence intervals (95% CI) in order to estimate the association of baseline variables with the distinct trajectories. Additionally a subanalysis was performed for smaller subgroups with more extreme courses of pain. Results: 705 participants were included in our analyses (1⁄4 705 knees). Therewere significant differences between the pain trajectory groups in how participants cope with pain. Four variables remained statistically significant after the univariate and the multivariable analyses between the unfavourable and the favourable course of pain group: number of co morbidities (> 1 vs. 1) (OR 2.2 [1.2-4.1]); baseline VAS (OR 1.8 [1.52.1]); pain during active flexion of the knee (yes vs. no) (OR 2.2 [1.04.5]); and joint space tenderness of the index knee during physical examination (yes vs. no) (OR 2.1 [1.1-3.8]). The subanalysis showed similar significant variables. Conclusions: Individuals with early symptomatic knee OAwho present with more severe joint pain; lesser capability to cope with pain; multi morbidity; joint space tenderness of the index knee; pain in the ipsilateral hip; and painful range of motion of the ipsilateral hip have an increased risk of progression of knee pain during a five year follow-up period. 509 INVESTIGATION OF THE RELATIONSHIP BETWEEN HIP BONE MARROW LESIONS, HIGH CARTILAGE SIGNAL AND HIP AND KNEE PAIN IN TWO POPULATION-BASED COHORTS H. Ahedi y, D. Dore y, L. Blizzard y, F. Cicuttini z, G. Jones y. yMenzies Res. Inst., Univ. of Tasmania, Hobart, Tasmania, Australia; zDEPM, Sch. of Publ. Hlth. and Preventive Med., Monash Univ., Melbourne, Australia Aim: To examine the relationship between hip BMLs, high cartilage signal and hip and knee pain using data from two geographically different cohorts. Methods: The first cohort included 143 healthy subjects from the Melbourne Collaborative Cohort study (MCCS) with a right hip MRI conducted during 2009-2010. The second cohort involved 198 subjects from Tasmanian Older Adult Cohort (TASOAC) with right hip MRI conducted at two time points, approx. 2.6 years apart. Subjects inMCCS had a T2-weighted MRI sequence and those in TASOAC had a Short T1 Inversion Recovery (STIR) MRI sequence. These MRI images were used to quantitatively assess hip BMLs size and determine high cartilage signal presence. Hip WOMAC (Western Ontario and McMaster Universities Osteoarthritis) was assessed in both cohorts, but knee WOMAC was available only in TASOAC. Log binomial regression was used to estimate the relative risk (RR) of presence of pain and of high cartilage signal, and linear regressionwas used in the analysis of severity of pain. Results: In the MCCS, 17% (24/143) of subjects had femoral and/or acetabular BML. Presence of any acetabular BML (n1⁄420, 13%) was associated with higher relative risk of hip pain [RR: 1.17, 95% CI: 1.03, 1.17]. Having a large femoral BML was associated with more severe hip pain [mean difference in hip pain score: 40.6, 95% CI: 6.00, 75.3]. Hip BML size was not associated with either hip or knee pain. In TASOAC, 28% (55/198) of subjects had a hip BML at either or both sites. In crosssectional analysis, presence of any large BML (n1⁄414, 8%) was associated with higher risk of hip pain [RR 2.14, 95% CI: 1.34, 3.41] and having a large acetabular BMLwas associatedwith greater severity of knee pain [mean diff 2.11, 95% CI: 0.36, 3.84] and greater hip pain [mean diff 3.92, 95% CI: 1.63, 9.57]. Longitudinally, resolution of femoral BMLs was associated with a decrease in hip [mean diff 2.75, 95% CI: 4.52, 0.99] and knee [mean diff 3.46, 95% CI: 5.91, 0.88] pain while incident acetabular [mean diff +5.84, 95% CI: +3.63, +8.05] and femoral [mean diff +0.98, 95% CI: +0.28, +1.78] BMLs were associated with worsening hip pain. High cartilage signal was associated with hip BMLs in each cohort [MCCS, RR: 1.70 95% CI: 1.41, 2.11; TASOAC, RR 1.79 95% CI: 1.52, 1.92] but not with pain. Conclusion: The consistent cross-sectional evidence from the MCCS and TASOAC cohorts, which was confirmed by longitudinal results from the TASOAC cohort, suggests that large hip BMLs are associated with both hip and knee pain. High cartilage signal was asymptomatic but strongly associated with hip BMLs. 510 ASSOCIATIONS OF FREQUENT PREDICTABLE AND UNPREDICTABLE PAIN WITH FUNCTIONAL AND PSYCHOLOGICAL OUTCOMES K.D. Allen y,z, H.B. Bosworth y,z, C.J. Coffman y,z, A.S. Jeffreys y, E.Z. Oddone y,z, W.S. Yancy, Jr. y,z. yDurham VA Med. Ctr., Durham, NC, USA; zDuke Univ. Med. Ctr., Durham, NC, USA Purpose: Studies have shown that greater pain severity is associated with poorer functional and psychological outcomes among patients with osteoarthritis (OA), but little is known about the impact of other aspects of the pain experience. This study examined associations of the frequency of predictable (e.g., occurs after a specific trigger) and unpredictable (e.g., occurs without warning) pain with self-reported functional outcomes and depressive symptoms among patients with physician diagnoses of hip and/or knee OA. Methods: Participants were n1⁄4293 patients enrolled in a clinical trial of a combined patient and provider intervention for managing OA at the Durham Veterans Affairs Medical Center (mean age 1⁄4 61, SD1⁄49, 90% male, 53% non-white primarily African American). All measures were from baseline assessments. The frequency of predictable and unpredictable pain were assessed with two items from the Measure of Intermittent and Constant Osteoarthritis Pain Index (ICOAP), both categorized as frequent (often, very often) vs. infrequent (never, rarely, sometimes). Physical function outcomes included the Western Ontario and McMasters University Osteoarthritis Index (WOMAC) function subscale and the Satisfaction with Physical Function Scale. Depressive