Physicians' accuracy and interrator reliability for the diagnosis of unstable meniscal tears in patients having osteoarthritis of the knee

Objective: To determine clinicians' accuracy and reliability for the clinic al diagnosis of unstable meniscus tears in patients with symptomatic osteoa rthritis of the knee. Design: A prospective cohort study. Setting: A single tertiary care centre. Patients: One hundred and fifty-two patients with sy mptomatic osteoarthritis of the knee refractory to conservative medical tre atment were selected for prospective evaluation of arthroscopic debridement . Intervention: Arthroscopic debridement of the knee, including meniscal te ar and chondral flap resection, without abrasion arthroplasty. Outcome meas ures: A standardized assessment protocol was administered to each patient b y 2 independent observers. Arthroscopic determination of unstable meniscal tears was recorded by 1 observer who reviewed a video, recording and was bl inded to preoperative data. Those variables that had the highest interobser ver agreement and the strongest association with meniscal tear by univariat e methods were entered into logistic regression to model the best predictio n of resectable tears. Results: There were 92 meniscal tears (77 medial, 15 lateral). Interobserver agreement between clinical fellows and treating su rgeons was poor to fair (kappa < 0.4) for all clinical variables except rad iographic measures, which were good. Fellows and surgeons predicted unstabl e meniscal tear preoperatively with equivalent accuracy of 60%. Logistic re gression modelling revealed that a history of swelling and a ballottable ef fusion were negative predictors. A positive McMurray test was the only posi tive predictor of unstable meniscal tear. "Mechanical" sym toms were not re liable predictors in this prospective study. The model was 69% accurate for all patients and 76% for those with advanced medial compartment osteoarthr itis defined by a joint space height of 2 mm or less. Conclusions: This stu dy underscored the difficulty in using clinical variables to predict unstab le medial meniscal tears in patients with pre-existing osteoarthritis of th e knee. The lack of interobserver agreement must be overcome to ensure that the findings can be generalized to other physician observers.