The application of fine-needle aspiration (FNA) to the evaluation of the thyroid nodule has greatly enhanced the ability of the clinician to appropriately select patients for thyroidectomy. However, despite extensive experience with thyroid FNA, the cytological distinction of benign from malignant follicular neoplasia remains problematic. As a result, most patients with FNA findings of a follicular neoplasm are referred for thyroidectomy. The goal of the present study was to develop clinical criteria capable of predicting malignancy in patients with an FNA diagnosis of follicular neoplasm. Among 1121 patients undergoing thyroid FNA at two large teaching centers during the period 1990 to 1995, 149 patients had cytological findings consistent with a follicular neoplasm. Among 103 patients referred for thyroidectomy, 22 (21%) were found to have a malignancy in the biopsied nodule. Among patients subjected to thyroidectomy, the risk of malignancy was significantly higher when follicular neoplasia was present in a male (43% vs. 16% for females, p = 0.007), when the nodule was greater than 4 cm to palpation (40% vs. 13% for nodules less than 4 cm, p = 0.03), or when the nodule was judged to be solitary by palpation (25% vs. 6% for a dominant nodule in a multinodular goiter, p = 0.02). Bayesian analysis of the data reveals that after an FNA showing a follicular neoplasm, the risk of malignancy in males with large nodules was nearly 80%, compared with a rate of only 3% in females with small nodules. These results suggest that clinical features including gender, nodule size, and character of the gland by palpation can be systematically integrated into the decision analysis, thereby improving the selection of patients for surgical referral.