A proposal for a semiquantitative scoring system for lymphedema using Non-contrast Magnetic Resonance Lymphography (NMRL): Reproducibility among readers and correlation with clinical grading

Purpose: To assess the ability and reproducibility of Non-contrast Magnetic Resonance Lymphography (NMRL) in detecting and quantify lymphedema, using a semiquantitative scoring system. Methods and material: This is a monocentric retrospective study of 134 consecutive patients with a clinical diagnosis of limb lymphedema who performed a Non-contrast Magnetic Resonance Lymphography (NMRL) at our Institution between November 2014 and February 2017. Lymphedema was classified based both on clinical and radiologic evaluation. An NMRL total score was obtained for each limb's segment and compared to the clinical grade, used as reference standard. NMRL intra-observer, inter-observer variability and intraclass correlation were calculated. NMRL sensitivity, specificity, and accuracy in identifying lymphedema were provided. Based on score distribution an NMRL four-stage system was developed. Results: NMRL showed 92% sensitivity, 77% specificity and 82% accuracy in identifying lymphedema. An almost perfect agreement was obtained by expert operators, while substantial agreement was obtained by non-expert operators. Substantial agreement resulted also for the inter-observer variability (Cohen's Kappa K = 0.73, CI 95% [0.69–0.78]). The intra-class correlation showed an almost perfect relationship both by expert and non-expert operators. Excellent correlation between clinical grade and NMRL score and between clinical grade and NMRL stage were found for each segment. Conclusions: NMRL is a confident and reproducible exam with high sensitivity, good specificity and high accuracy in lymphedema detection; the semiquantitative NMRL score resulted a reliable and reproducible tool able to quantify lymphedema severity.