ABSTRACT
The role of ultrasonography (US) in the diagnosis of cancer in thyroid nodules is not well-established. The aim of the present study was to evaluate US performance in predicting cancer in thyroid nodules using a novel approach. Two hundred and eighty-nine patients with thyroid nodular disease were evaluated with clinical, biochemical and cytopathological examinations. Eighty patients with palpable solitary thyroid nodules or multinodular goiters who were to undergo surgery were included, and had a US exam performed by one of us. Some US characteristics of thyroid nodules were associated to cancer: absent halo, hypoechogenicity and microcalcifications, with sensitivity, respectively, of 56, 44 and 56%, and specificity of, respectively, 80, 83 and 94%. These findings were considered positive and were studied in two different combinations: simultaneous, when two or more were positive, and parallel, when any positive finding was present. When positive findings were studied simultaneously, sensitivity ranged 25 to 38% and specificity ranged 89 to 97%. Microcalcifications, associated or not to other findings, were highly specific for thyroid cancer, but they were only present in half of the malignancies. When positive findings were studied in parallel, sensitivity ranged 69 to 81% and specificity ranged 70 to 81%. The parallel combination of hypoechogenicity or microcalcifications or absent halo improved US sensitivity to 81% with an acceptable specificity (70%). This method is potentially useful to help us select patients for surgery when fine-needle aspiration biopsy is repetitively non-diagnostic or select for biopsy incidentally discovered non-palpable nodules.
