The predictor model shows a high accuracy (> 80%) for malignant thyroid nodule when it includes age (≥ 38.5 years.), border irregularity, microcalcifications, and nodule size (≥ 2 cm) using high-resolution ultrasound. In our study, scintigraphyc study was not useful to differentiate the two groups. The number of nodules (solitary vs. multiple) did not predict malignancy. No other clinical or laboratory parameters were significant in this study, including auto-immune disease and TSH level. Therefore, a risk stratification scheme would theoretically help both the patient and the surgeon to make a better decision upon the extent of recommended surgery.
Age in thyroid nodule patients was identified as an independent predictor for malignancy with an age cut off of ≥ 38.5. Baier et al (2009) reviewed reports of 944 thyroid nodules of four sonographic features and found statistical significance in malignant nodules in young patients (≤ 45 years) and solid nodule morphology . Several studies have tried to predict malignancy in thyroid nodules with indeterminate or suspicious FNA findings according to the M.D. Anderson Cancer Center series . The ROC-curve of our data shows a different age cutoff to predict malignancy during thyroid nodule evaluation, providing good accuracy and high sensitivity rate. In fact, the application of the International Union against Cancer (AJCC/UICC) classification system based on pTNM parameters and in age is recommended for tumors of all types, including thyroid cancer , because it provides a useful shorthand method to describe the extent of the tumor. Age is one of the criteria, with cutoff of over 45 years that is in disagreement with the findings showed in our data, which agree with the Banks and Baier et al study [16, 20].
The importance of TSH levels as a predictor of malignancy in thyroid nodule evaluation, have been discussed in recent studies showing that an elevated serum TSH concentration might be associated to increased risk of different thyroid cancers in patients with nodular goiter [11, 12, 15]. Higher TSH values, even within normal ranges, have been associated with a greater risk of thyroid malignancy in some studies [11–15]. Boelaert et al (2006) studied 1.500 consecutive patients without overt thyroid dysfunction and found a significant increase in adjusted odds ratios (AORs) for the diagnosis of malignancy in subjects with serum TSH 1.0-1.7 mU/liter compared to TSH less than 0.4 mU/liter (AOR 2.72), with further increases being evident in those with TSH 1.8-5.5 mU/liter (AOR 3.88). Males, younger patients, and those with clinically solitary nodules were also at increased risk . We did not observed correlation of gender or solitary nodule in our data. The TSH concentration was not significant after multivariate analysis in our study in accordance with some authors  and in disagreement with other ones [11, 12], remaining in this way unclear to date and needing further investigation.
In our data correlation with thyroid auto-immunity and malignancy was not found. In the majority of the previous retrospective studies, there is a support for the correlation between thyroid malignancy and Hashimoto's thyroiditis (HT) [24–30]. A current review of the American Thyroid Association guidelines for thyroid nodules and thyroid cancer stated that rate of malignancy in nodules in thyroid glands involved with HT could be possibly higher [6, 31]. In a recent study by Anil et al (2010), a malignancy rate of 1.0% in HT group (2 out of 191 nodules) vs. 2.7% in the control group (19/713) was demonstrated, although no statistical significance was found even at higher TSH levels  that is similar to our results.
Thyroid ultrasound is used to evaluate index of nodule size, location, characteristics, number and presence of additional thyroid nodules and to detect suspicious appearance of lymph nodes . Nodule size has been pointed out not to be a predictive of malignancy [6, 11, 12, 18]. Patients with multiple thyroid nodules have the same risk for malignancy as those with solitary nodules. Is recommended that all patients with nodular thyroid glands should be submitted to US evaluation [6, 18]. Our data showed correlations of thyroid malignancy with nodules which presented microcalcifications, border irregularity, size ≥ 2 cm, central flow by Doppler and hypoechogenicity after US study. After a multiple logistic regression border irregularity and microcalcifications were the strongest predictors of malignancy in thyroid nodule, followed by the nodule over 2 cm in diameter.
Gonzalez-Gonzales (2010) evaluated the efficiency of diagnostic of sonographic findings and compared to those of FNA biopsy of thyroid nodules to study US characteristics of 341 thyroid nodules. The multivariate logistic regression revealed that the only variable, which kept a significant association with malignancy, was the presence of microcalcifications . These data confirm the study by Li QS (2010) who retrospectively reviewed 115 nodules (104 patients) with PTC. They also analyzed thyroid nodules and cervical lymph nodes size, border, calcification, echotexture, hemodynamic on US. The microcalcifications showed an increased in suspicion for malignancy of thyroid nodule . A hypoechoic thyroid nodule with increased internal vascularity, ill-defined border and microcalcifications, PTC was strongly suggested, which is similar to our data.
The color Doppler analysis was not correlated to thyroid malignancy in our study, which agrees with currently data in the literature. Moon et al (2010) evaluated 1083 thyroid nodules, 814 benign and 269 malignant. The central flow was frequently seen in benign nodules and the absence of vascularity was more frequent in malignant nodules. Vascularity itself or a combination of vascularity and gray-scale US features was not as useful as the use of suspicious gray-scale US features alone for predicting thyroid malignancy , similar to the data of Cantisani et al (2010) of 1.090 assessed patients . In their study, they concluded that pattern III cannot be used to predict malignancy with confidence, and FNA still is mandatory to remove the nature of the nodule. Choi et al (2009) followed up 165 patients with indeterminate cytology diagnosed as follicular neoplasm and no difference in malignancy incidence on gender; age (≥ 45 years), nodule size and US features were found. Only central color Doppler flow was predictive for malignancy in follicular neoplasm . However, Anil et al (2010) showed that US features of nodule echogenicity, structure, margin, and Doppler flow were similar in patients with Hashimoto's thyroiditis and in control group .
Banks et al (2008) proposed a risk score analysis based on patient age (50 yrs), nodule size (2.5 cm) and cytopathological features (atypia) for patients with indeterminate or suspicious FNAC . They observed a nonlinear relationship between age and risk of malignancy, and patients at both age extremes were more likely to have malignant thyroid nodules.
A predictor model was created using variable of age (> 39 years), border irregularity, microcalcifications and nodule diameter (>2 cm) to identify thyroid malignancy with good accuracy (>80%). Is important to highlight that to understand the combination of age and US parameters in malignancy prediction is essential for clinicians to make decisions, and to guide surgical definition in many cases. The TSH level, the presence of auto-immune disease or scintigraphyc study was not useful to make differentiation in the two groups. Male gender, solitary nodule or Hashimoto's thyroiditis were also not considered predictors of malignancy in our study.
Risk prediction, based on clinical and US parameters, should be used as an adjunct the findings of FNA aiming to identify patients who require further investigation and/or surgical intervention. Prospective studies are required to define the role of this risk prediction to improve clinical management in a larger patient population.