Development of a malignancy predicting score in EBUS-TBNA

Endobronchial ultrasound‐guided transbronchial needle aspiration (EBUS‐TBNA) is an established and essential tool in lung cancer diagnosis and staging. While most studies show a specificity of 100%, sensitivity is often sub-optimal. The aim of this study was to develop a prediction model for malignancy, based on patient and lymph node characteristics. A retrospective cohort analysis of lymph nodes submitted to EBUS-TBNA between January 2016 and December 2017 was conducted. Lymph node characteristics (station, morphology, size, and vascularization) and histology were evaluated. A multivariate logistic regression analysis was performed and a prediction score for malignancy was developed from this model using a regression coefficient-based scoring method. A cut-off point was chosen for optimal sensitivity. 308 lymph node results were analysed, corresponding to 222 patients, with a mean age of 61 (range 27 - 85). 76 (33.8%) were female. Punctured lymph nodes had a mean short axis of 13.3 mm. 66 (21.4%) were malignant. On univariable analysis, age, lymph node size, borders, vascularization and station (pretracheal vs others) were associated with malignancy. Only lymph node size (β=0.169), vascularization (β=-1.370) and station (β=1.035) were retained in the final multivariate model. This model [ sizex0.169 + (1.035 if pre-tracheal) – (1.370 if vascularized)] had an area under curve of 0.777 (95% CI 0.715 - 0.839), p<0.001. For a cut-off of 1.5 in this score, sensitivity was 95%. A score based on lymph node size, vascularization and station can improve accuracy of EBUS-TBNA and help to optimize false-negative results.