Prediction Models to Improve the Diagnostic Value of Plain Radiographs in Children With Complete Discoid Lateral Meniscus

Abstract

Purpose: To develop prediction models to improve the diagnostic utility of plain radiographs for the diagnosis of complete discoid lateral meniscus by combining previously reported radiographic findings.

Methods: Patients ages 5 to 16 years with complete discoid lateral meniscus confirmed by arthroscopy or magnetic resonance imaging were included. Patients with insufficient radiographs were excluded. Normal control subjects were randomly sampled by age and sex matching. Subjects were divided into 2 groups considering skeletal maturation (5-9 and 10-16 years). Radiographic variables included were lateral joint space, height of the fibular head, height of the lateral tibial spine, obliquity and cupping of the lateral tibial plateau, condylar cutoff sign, and squaring and notching of the lateral femoral condyle. Prediction models were developed by regression analyses. The cutoff value (COV) for best accuracy was determined with its sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV).

Results: A total of 162 knees (126 patients) with complete discoid lateral meniscus and 151 age- and sex-matched knees (151 subjects) with normal meniscus were included. In subjects 5 to 9 years old, the prediction model was risk score = [-20.08 * height of the fibular head/femoral interepicondylar distance (FIED)] + [-42.26 * height of the lateral tibial spine/FIED]. The COV of -8.47 showed the best accuracy (74.4%), with sensitivity of 85.9%; specificity, 60.4%; PPV, 72.4%; and NPV, 78.0%. In subjects 10 to 16 years old, the prediction model was risk score = [77.04 * lateral joint space/FIED] + [-34.55 * height of the fibular head/FIED] + [-56.58 * height of the lateral tibial spine/FIED] + [-16.44 * condylar cutoff sign]. The COV of -18.03 showed the best accuracy (85.4%), with sensitivity of 79.6%; specificity, 90.4%; PPV, 87.6%; and NPV, 83.9%.

Conclusions: The prediction models combining the plain radiographic findings showed higher diagnostic values than the diagnostic values of the individual radiographic findings. The results of this study provide improved diagnostic utility of plain radiography for the detection of completed discoid lateral meniscus in children.

Level of evidence: Level III, diagnostic study.