EUS-guided FNA: A multivariate analysis of procedural characteristics associated with a positive pathological diagnosis of malignancy

AIM To determine technical characteristics of EUS-guided fine needle aspiration (EUSGFNA) associated with a positive pathological diagnosis of cancer. METHODS: Procedural characteristics of EUSGFNA passes on lesions suspect for malignancy were prospectively recorded. The Wilson-Cooke and GIP needles were used according to endoscopist preference. Each pass produced 3-4 slides. Cytopathologists in the room studied 1 slide per pass for cellular adequacy. All slides were later stained and studied for the final diagnosis. Non-parametric univariate analysis determined variables independently correlated with malignant histology. Multivariate logistic regression (with control for multicolinearity) was used to control for cross-variable effects (α=0.05). RESULTS: There were 177 passes in 46 lesions (avg. 3.8 passes/lesion), in 45 patients. 80(45%) passes showed malignancy; 97(55%) did not. Variables studied include: type (101 nodes, 76 masses), location (70 mediastinum, 63 pancreas, 20 celiac axis, 12 esophagus, 5 peri-rectum, 5 rectum, 2 peri-gastric), hardness (79 soft, 58 hard, 36 medium, 4 unspecified), stroke length (96 short, 74 long, 7 unspecified), sampling site (111 center, 59 edge, 7 unspecified), aspirate size (131 ample, 30 scant, 16 unspecified), and aspirate bloodiness (56 bloody, 55 non-bloody, 53 visible blood, 13 unspecified). Based on available data in a subgroup of 19 malignant lesions, the 1st pass was positive in 11/19(58%), 2nd pass in 7/19(37%), 3rd pass in 1/19(5%). PREDICTORS OF POSITIVE DIAGNOSIS OF CANCER BY EUSGFNA Lesion characteristic + or-predictor p (Univariate) p (Multivariate) mediastinal location positive 0.0002 0.0002 sample site "edge" positive 0.03 0.02 lymph node positive NS 0.05 pancreatic location negative 0.007 0.004 mass negative NS 0.05 sample site "center" negative NS 0.04 SUMMARY: Independent predictors of an EUS-guided FNA diagnosis of malignancy were mediastinal location and sampling at the edge of the lesion. Pancreatic location was the only independent predictor of failure. Malignancy could be diagnosed after 2 passes in 95% of patients, and in 100% after 3 passes. CONCLUSIONS: To optimize the yield of EUS-guided FNA, lesions should be sampled at the edge. 4 passes should be adequate to diagnose malignancy in a suspicious lesion.