Abstract
Introduction: EBUS TBNA(ET) has become first-line Lymph Nodes(LN) staging procedure for lung cancer.
Aims/Objectives: To develop & validate risk stratification model to categorize ET negative LN in low vs high probability(LNS,HNS) false negative.
Methods: Retrospective analysis of 2014 database.Model built from derivation set(independent predictors of malignancy)and validation set(evaluate constructed model). Analyzed as potential predictors: CT LN size axis,LN SUV,SUV LN/tumor ratio,sonographic characteristics. To develop model only ET negative LN were used in analysis. Negative LN randomly divided(60:40 ratio)into derivation and validation set.Logistic regression analysis identified(derivation set)significant independent predictors of malignancy. We constructed prognostic index using regression analysis coefficients.Appropriate score cutoff was assessed;validation sample was used to evaluate diagnostic accuracy of constructed score.
Results: On 76 ET (all with rapid on-site pathological evaluation),47 LN were included in analysis(derivation set n=28,validation set n=19). SUV LN/tumor ratio & heterogeneous echogenicity were independent malignancy predictors.Using simplified scoring system based on natural logs of odds ratios from multivariable analysis on derivation sample,LN can be stratified into LNS & LNS. 20/21 and 13/14 LNS LN in derivation and validation set were benign proven; 5/7 and 3/5 HNS LN were proven malignant.Negative predictive value of model for derivation and validation set was 97.3% and 94.5%.
Conclusions: This model could assist multidisciplinary team,after negative ET,to establish which patients need further staging procedures and which may proceed directly to treatments.
- Copyright ©ERS 2015
