Prediction of mediastinal lymph node metastasis in papillary thyroid cancer.

Because of the rarity of mediastinal lymph node metastasis, no evidence exists regarding risk factors indicating its presence. This study aimed to identify risk factors representing potential triggers of imaging for mediastinal lymph node metastasis. Carried out was an analysis of 392 consecutive patients with node-positive papillary thyroid cancer referred to a tertiary surgical center. Significant univariate associations (P </= 0.001) existed, after correction for multiple testing, between mediastinal lymph node metastasis and poor tumor differentiation (24% versus 2%), number of positive lymph nodes (mean 14.4 versus 7.3 nodes), and distant metastasis (43% versus 11%). Only nominally significant were age at first tumor diagnosis (47.7 versus 39.1 years; P = 0.042), extrathyroidal tumor extension (76% versus 53%; P = 0.044), and bilateral lymph node metastasis (81% versus 49%; P = 0.006). In the initial and optimized multivariate logistic regression models, which included all significant variables from the above analysis, only poor tumor differentiation (odds ratio 11.6-14.6) and distant metastasis (odds ratio 5.4-6.1) represented significant (P </= 0.005) predictors of mediastinal lymph node metastasis. These two variables featured excellent negative predictive values of 96%. Regarding prediction of mediastinal lymph node metastasis, poor tumor differentiation was more discriminatory than distant metastasis, as reflected in a better accuracy (94% versus 86%) and positive predictive value (42% versus 18%). We conclude that, subject to validation in independent series, patients with poorly differentiated papillary thyroid cancers, which are rare, should be ideal candidates for mediastinal imaging because 5 of 12 patients (42%) having this condition in the present investigation harbored mediastinal lymph node metastasis.