The predictive value of quantitative joint space width loss for total hip replacement

Abstract

Purpose: Change in quantitative joint space width obtained from conventional AP radiographs of the pelvis is the most commonly recommended and studied measure of hip OA structural outcome. Yet there is surprising little data on the predictive validity of hip JSW loss for clinical outcomes. The minimum JSW (mJSW) is the most widely used measure, but several alternative hip JSW metrics are available. No studies have compared the predictive validity of different metrics of hip JSW loss for clinical outcomes. Therefore, the purpose of this study was to determine how change in JSW over 4 years assessed by different metrics was predictive or associated with subsequent total hip replacement (THR) using data from the Osteoarthritis Initiative (OAI). Methods: Baseline and 48 month follow-up visit pelvis radiographs of all subjects in the OAI were assessed for hip OA (HOA) by an experienced radiologist and an experienced rheumatologist. The radiographs of 843 subjects with possible or definite radiographic HOA in either hip at the 48 month timepoint were measured quantitatively for hip JSW with computer assisted edge-detection tools. Two readers segmented the femoral head contours and acetabular sourcils in a 40 degree arc over the femoral head centers and perpendicular to the inter-head axis. 1326 hips with possible or definite radiographic OA by the 48 month time point were included in the analysis. Hips with very low baseline JSW (<1 mm) and THR for reasons other than OA were excluded. Changes in four JSW metrics were analyzed as predictors of THR: minimum JSWs (mJSW), change in medial and lateral compartment minimum JSW (medJSW and latJSW), mean change in JSW (meanJSW - equivalent to a measure of JSW area), and the mean change in JSW for a 5 degree arc region of interest at the location of the minimum JSW at 48 month (roiJSW). We calculated the mean changes (in mm), standardized response means (SRM) for all hips combined, and area under receiver operating characteristic curves (AUC) for the JSW changes calculated by these 4 methods. Odds ratios (OR) for each 0.5 mm JSW loss (or difference) were estimated using binary logistic regression with GEE (to account for more than one hip per subject) and adjusted for age, sex, BMI, and race. Results: 46 hips had confirmed THR for OA between the 48 and 96 month visits. In all subjects, the mean age was 63.2 years (±9.2) and mean BMI was 28.3 (±4.4). Age, sex, and BMI were not significantly different in subjects with confirmed THR. JSW by all metrics decreased more in the first 4 years in the hips that underwent THR within the next 4 years than those that did not (Table 1). AUCs for JSW metrics predicting THR ranged from 0.83 to 0.87. All JSW loss metrics were significantly associated with subsequent THR within 4 years, with ORs that ranged from 2.3–3.3 per 0.5 mm loss. The ORs were greatest for medJSW and meanJSW. All metrics were significantly associated with THR independent of the minimum JSW at 48 months, but with lower odds; ORs ranged from 1.7–2.4.Tabled 1Table 1. JSW average and SD (mm), sensitivities and predictive performancesHips OA with THRHips OA without THRAll Hips CombinedMeanSDMeanSDMeanSDSRMAUCOR95% CImJSW1.221.080.200.490.240.560.430.842.62.1–3.4medJSW1.311.030.200.480.240.550.430.873.32.5–4.4latJSW1.241.110.200.520.240.590.410.832.41.9–3.0meanJSW1.261.040.210.430.250.510.490.863.32.4–4.4roiJSW1.471.200.340.530.380.610.630.842.31.8–2.8mJSW Baseline3.451.423.730.843.720.87n/a0.581.30.9–1.7mJSW 48 months2.221.283.520.883.480.92n/a0.832.31.8–2.9 Open table in a new tab Conclusions: These JSW loss metrics all performed similarly in terms of sensitivity to change and prediction of subsequent THR. The sensitivity to change was higher in metrics that average the JSW loss in a hip (meanJSW and roiJSW) than vs. those based on minimum JSW (mJSW, medJSW, and latJSW).The higher SRM in meanJSW is reflects lower standard deviations, but the higher SRM in roiJSW is due to greater measured decreases. The predictive value was greatest (in both AUC and OR) for medJSW and meanJSW metrics. JSW loss was also significantly associated to THR risk independent of the minimum JSW at 48 months, which may suggest that the amount or rate of JSW loss has clinical impact irrespective of the thickness of the cartilage at any given time. Our study demonstrates the strong predictive validity of hip JSW change for subsequent THR, an important clinical outcome. All JSW metrics performed similarly.