An ultrasound model to discriminate the risk of thyroid carcinoma

Thyroid nodules are common on ultrasonographic examination and are mostly benign. Ultrasound characteristics may help discriminate thyroid carcinoma (TC) from benign nodules. The aims of this study were to identify ultrasonographic characteristics associated with TC and to validate a previously proposed model based on the presence of three ultrasonographic characteristics.

From a protocolized prospective registry of 1108 fine needle aspiration biopsies performed during a 16-month period at an ambulatory center, the ultrasonographic characteristics of TC and non-TC biopsies were compared. Adjusted odds ratios (ORs) and likelihood ratios for TC were estimated for eight combinations of three previously identified characteristics (microcalcifications, hypoechogenicity, and irregular borders).

Microcalcifications (OR, 6.6; 95% confidence interval [CI], 4.4-9.9), hypoechogenicity (OR, 4.7; 95% CI, 2.8-8.0), and irregular borders (OR, 4.3; 95% CI, 2.8-6.5) were independently associated with TC. When added to a logistic regression model, the three ultrasonographic characteristics remained statistically significant. In the absence of these three features, the likelihood ratio for TC was 0.1 (95% CI, 0.0-0.2), while in their simultaneous presence, the likelihood ratio was 11 (95% CI, 6.6-19.0).

The absence or simultaneous presence of three simple ultrasonographic characteristics generates a large change of pretest probability of TC and could avoid unnecessary fine needle aspiration biopsy.