Objective: To determine if any preoperative or intraoperative factors can reliably predict malignancy in patients with Hurthle cell neoplasms. Summary Background Data: Most experienced surgeons recommend total thyroidectomy for Hurthle cell carcinomas and reserve thyroid lobectomy for Hurthle cell adenomas. However, delineation between Hurthle cell adenoma versus carcinoma often cannot reliably be made either before or during surgery. Methods: Medical records from 57 consecutive patients who underwent thyroid resections for Hurthle cell neoplasms between October 1984 and April 1995 at The Johns Hopkins Hospital were analyzed to determine if any factors were predictive of malignancy. Results: Of the 57 patients with Hurthle cell neoplasms, 37 had adenomas and 20 had carcinomas, resulting in a 35% prevalence of malignancy. Patients with adenomas did not differ from those with carcinoma with respect to age, sex, or history of head and neck irradiation. However, patients with Hurthle cell carcinomas had significantly larger tumors (4.0 ± 0.4 cm vs. 2.4 ± 0.2 cm, p < 0.005). Furthermore, although the incidence of malignancy was only 17% for tumors 1 cm or less and 23% for tumors 1 to 4 cm, tumors 4 cm or greater were malignant 65% of the time (p < 0.05). Both fine-needle aspiration and intraoperative frozen section analysis had low sensitivities in the detection of cancer (16% and 23%, respectively). With up to 9 years of follow-up, there has been no tumor- related mortality. Conclusions: These data demonstrate that the size of a Hurthle cell neoplasm is predictive of malignancy. Therefore, at the time of initial exploration for large Hurthle cell neoplasms (>4 cm), definitive resection involving both thyroid lobes should be considered because of the higher probability of malignancy.