PURPOSE To prospectively estimate the accuracy of sonoelastography in the differentiation of benign and metastatic cervical lymph nodes (LNs) in patients suspected of having thyroid or hypopharyngeal cancer, with histologic nodal findings as the reference standard. MATERIALS AND METHODS The study protocol was approved by the hospital review board; each patient gave written informed consent. One hundred forty-one peripheral neck LNs (60 metastatic, 81 metastasis free) in 43 consecutive patients (22 men, 21 women; mean age, 58 years +/- 13 [standard deviation]) were examined. Patients referred for surgical treatment of suspected thyroid or hypopharyngeal cancer were examined with gray-scale ultrasonography (US), power Doppler US, and sonoelastography. At gray-scale and power Doppler US, the following LN characteristics were evaluated: short-axis diameter, short-to-long-axis diameter ratio, echogenicity, calcifications, and vascularity. A four-point rating scale was used to evaluate the US elastograms for LN visibility, relative brightness, margin regularity, and margin definition. In addition, strains of LN and surrounding neck muscles were measured on elastograms, and the muscle-to-LN strain ratio--that is, the strain index-was calculated. The diagnostic potential of the examined criteria for metastatic involvement was evaluated with univariate analysis and multivariate generalized estimating equation (GEE) regression. P < .05 indicated statistical significance. RESULTS A strain index greater than 1.5 had high utility in metastatic LN classification, with 98% specificity, 85% sensitivity, and 92% overall accuracy. These results were significantly better than those obtained by using the best gray-scale criterion--that is, a short-to-long-axis diameter ratio greater than 0.5-which had 81% specificity, 75% sensitivity, and 79% overall accuracy. CONCLUSION Sonoelastography had high accuracy (92%) in the differentiation of benign and metastatic cervical LNs in patients suspected of having thyroid or hypopharyngeal cancer.