BACKGROUND: In the USA, about 30 200 well-differentiated thyroid carcinomas were diagnosed in 2007, but the prevalence of thyroid nodules is much higher (about 5% of the adult population). Unfortunately, the preoperative characterisation of follicular thyroid nodules is still a challenge, and many benign lesions, which remain indeterminate after fine-needle aspiration (FNA) cytology are referred to surgery. About 85% of these thyroid nodules are classified as benign at final histology. We aimed to assess the diagnostic effect of galectin-3 expression analysis in distinguishing preoperatively benign from malignant follicular thyroid nodules when FNA findings were indeterminate. METHODS: 544 patients were enrolled between June 1, 2003, and Aug 30, 2006. We used a purified monoclonal antibody to galectin-3, a biotin-free immunocytohistochemical assay, and a morphological and phenotypic analysis of FNA-derived cell-block preparations. Galectin-3-expression analysis was applied preoperatively on 465 follicular thyroid proliferations that were candidates for surgery, and its diagnostic accuracy was compared with the final histology. FINDINGS: 31 patients were excluded because they had small galectin-3-negative thyroid nodules; we did not have data for 47 patients; and one patient with an oncocytic nodule was excluded. 331 (71%) of the assessable 465 preoperative thyroid FNA samples did not express galectin-3. 280 (85%) of these galectin-3-negative lesions were classified as benign at final histology. Galectin-3 expression was detected, instead, in 134 of 465 (29%) thyroid proliferations, 101 (75%) of which were confirmed as malignant. The overall sensitivity of the galectin-3 test was 78% (95% CI 74-82) and specificity was 93% (90-95). Estimated positive predictive value was 82% (79-86) and negative predictive value was 91% (88-93). 381 (88%) of 432 patients with follicular thyroid nodules who were referred for thyroidectomy were correctly classified preoperatively by use of the galectin-3 test. However, 29 (22%) of 130 cancers were missed by the galectin-3 method. INTERPRETATION: Our findings show that if the option of surgery was based theoretically on galectin-3 expression alone, only 134 thyroid operations would have been done in 465 patients; therefore a large proportion (71%) of unnecessary thyroid surgical procedures could be avoided, although a number of galectin-3-negative cancers could be potentially missed. The galectin-3 test proposed here does not replace conventional FNA cytology, but represents a complementary diagnostic method for those follicular nodules that remain indeterminate.