Clinicians have become reliant on fine-needle aspiration (FNA) cytology in the diagnosis of thyroid nodules. We encountered several patients who had thyroid cancer at operation despite having had FNAs that never showed malignant or suspicious cells. Hence, we retrospectively studied 100 consecutive patients with histologically proven thyroid cancer who had undergone preoperative FNA. Fourteen of these patients had cancers that were not detected by FNA, three of whom developed widespread disease. The sensitivity of FNA was 79%, the false-negative rate 21%, and the inadequate rate 12%. FNA was less sensitive in detecting follicular and Hurthle cell carcinomas compared to papillary carcinomas. FNAs interpreted by cytopathologists at a major university center were more sensitive than those performed elsewhere. A single false-negative FNA delayed surgical treatment by 28 months, sometimes despite clinical evidence suggesting malignancy. Subjects whose tumors were not detected by FNA experienced delayed treatment, had higher rates of vascular and capsular invasion, and were more likely to have persistent disease at follow up (hazard ratio 2.28). False-negative results remain a concern in the cytologic diagnosis of thyroid cancer. Although FNA is a useful test, clinical findings should overrule cytologic data in order for timely treatment to occur.