Validation of a thyroid malignancy risk calculator in a Portuguese cohort

The Carral et al. calculator integrates clinical, laboratory and ultrasound variables to estimate thyroid malignancy risk. We independently evaluated its performance in a Portuguese surgical cohort.

We conducted a retrospective study including patients who underwent total thyroidectomy between 2021 and 2023. For patients with multiple nodules, the calculator was applied to the index nodule, defined as the one with the highest EU-TIRADS score or, when equal, the largest diameter. Original predicted probabilities were compared with histopathology. Discrimination was assessed by area under the receiver operating characteristic curve (AUC); calibration by plot, Hosmer-Lemeshow test, intercept and slope.

Among 153 patients (mean age 52.1 ± 12.6 years; 86.3% female), 32 (20.9%) had differentiated thyroid carcinoma. Median predicted risk was 37.7% (IQR 21.8-58.6%) in malignant and 1.56% (IQR 0.37-6.2%) in benign nodules (p < 0.001). AUC was 0.877 (95% CI 0.82-0.93). Calibration was imperfect (Hosmer-Lemeshow p = 0.28; intercept 0.85, 95% CI 0.51-1.23; slope 0.78, 95% CI 0.63-0.91). At the 9.2% Youden threshold, sensitivity, specificity, positive predictive value and negative predictive value were 75.0%, 84.3%, 55.8% and 92.7%, respectively. Subgroup analyses were exploratory.

The calculator showed excellent discrimination but imperfect calibration in this selected surgical cohort. Prospective validation and clinical-impact assessment in unselected populations are required before it is used to alter fine-needle aspiration decisions.