Quality of life and radiographic severity of knee osteoarthritis predict total knee arthroplasty: data from the osteoarthritis initiative

Abstract

Purpose: Total knee arthroplasty (TKA) is a clinically important outcome in persons with symptomatic osteoarthritis (OA) of the knee. Few studies, however, have examined predictors of TKA in community-based cohorts. We analyzed data from the Osteoarthritis Initiative (OAI) to determine the most parsimonious group of baseline variables that robustly predicted undergoing a TKA during 5 years of follow-up. Methods: Data for this analysis were obtained from the OAI public access database (http://www.oai.ucsf.edu). At each annual visit through 48 months, subjects completed an extensive battery of questionnaires, weight was measured with a balance beam scale, and fixed flexion, weight-bearing PA radiographs of both knees were obtained using a standard protocol and the Synaflexer™ platform; the 60-month follow-up visit only involved completion of questionnaires. Subjects reported TKAs at the time of their annual follow-up visits; these were confirmed by the Coordinating Center after review of medical records obtained by the Clinical Centers. We examined selected variables measured at the baseline visit for their role as predictors of TKA using knee-specific multiple variable logistic regression models. The best models were selected for each knee based on Chi-square values and area under the receiver operating characteristic curve (AUC). Results: Data from 4674 subjects enrolled in the OAI who either had symptomatic radiographic knee OA or were at risk for this condition were included in this analysis. At entry, subjects had a mean (SD) age of 61.32 (9.17) years and body mass index (BMI) of 28.74 (4.82) kg/m2; 2729 (58.4%) were women and 3677 (78.75%) were white. Baseline knee-specific WOMAC pain (0-20 scale) and function (0-68 scale) scores were 2.56 (3.26) and 8.26 (10.63) for the right knee and 2.39 (3.43) and 8.39 (11.44) for the left knee, respectively. Baseline KOOS pain and symptoms scores (0-100 scale) were 83.56 (17.37) and 85.95 (14.60) for the right knee and 84.80 (18.26) and 86.49 (15.87) for the left knee, respectively; KOOS quality of life (QoL, 0-100 scale) was 66.10 (22.43). Distribution of Kellgren-Lawrence (KL) grades of severity of radiographic knee OA at baseline was similar for both knees: 37, 18, 27, 14 and 4 percent had Grade 0, 1, 2, 3, 4 changes, respectively. During 60 months of follow-up, there were 91 right and 102 left TKAs. In models containing clinical variables and adjusted for age, gender, BMI and race, the best models included WOMAC pain and KOOS QoL (AUC = 0.80 and 0.82 for right and left knees, respectively). When KL grade was added to these models, the AUC improved to 0.89 and 0.90 for the right and left knees, respectively; race, BMI and WOMAC pain were no longer significantly associated with TKA outcome. The best parsimonious model included age, gender, KOOS QoL score and KL grade with an AUC = 0.89 and 0.90 for the right and left knee, respectively. Subjects with highest odds for a TKA were those in the lowest quartile of QoL scores and with KL grade 3 or 4 radiographic severity. Conclusions: These data suggest that knowledge of only a patient's age, gender, KOOS QoL score and KL grade of severity of radiographic knee OA can predict with high accuracy the odds of undergoing a TKA over the ensuing 5 years. These findings extend our prior report (Shurupoff K, et al: Osteoarthritis Cart 2011;19S1:S134 [abstract #284]) that KOOS QoL and radiographic severity can be used to model the outcome of "Virtual TKA".