PREDICTIVE MODEL OF THYROID SURGERY SUFFICIENCY IN PATIENTS WITH BETHESDA III AND IV: "THE RULE OF FOUR"

AUTHORS:
X. Guirao (Sabadell, Spain) , S. Barcons (Sabadell, Spain) , V. Perez-Riverola (Sabadell, Spain) , B. Bella (Sabadell, Spain) , C. Molero-Cano (Sabadell, Spain) , T. Landa (Sabadell, Spain) , L. Albert Fabregas (Sabadell, Spain) , M. Rigla (Sabadell, Spain)
Background:
Thyroid lobectomy (TL) in patients with thyroid nodule (TN) Bethesda III (BIII) and IV (BIV) purports a totalization when pathology informs of high risk malignancy (HRM: Cell high risk phenotype and/or widely invasive-angioinvasive thyroid carcinoma). Studies on predictive variables correlating with HRM are lacking.
Methods:
Data on demographics, TIRADS, Bethesda, size of TN, extension of thyroidectomy of patients operated on for BIII and BIV TN have been retrieved from a prospective database and compared between low risk (LR: Benign, low risk neoplasm and malignancy, minimally invasive/non angioinvasive) and HRM groups. A predictive model has been constructed and sensitivity (SE), specificity (SP), positive (PPV) and negative predictive value (NPV) and area under curve (AUC), have been calculated. Quantitative data are summarized as a mean ± SD and analysis of qualitative variables have done with X2 test.
Results:
One-hundred eight patients with a mean age of 53±25 years, 70,4% females and 70,4 and 29,6% of BIII and BIV were assessed. TL and TT were performed in 62 and 38%. Thyroid malignancy has been reported in 28,9 and 56,3 in BIII and BIV. TN >4 cm and TIRADS 4-5 have been correlated with final HRM pathology (p=0,004 and p=0,003). Elicited significant variables predicted model construct showed a SE, SP, PPV, NPV and AUC of de 0,45, 0,94, 0,72, 0,82 and 0,77, respectively.
Conclusions:
The predictive model construct to detect patients at risk for HRM thyroid nodules, may be useful in the shared decision of the extension of thyroid surgery.