Follicular thyroid lesion risk calculator

Englum

Supplemental materials

Statistical methodology

Using ICD-O-3 (International Classification of Diseases for Oncology, 3rd edition) histology codes for thyroid cancer, adult patients (age ≥ 18 years at diagnosis) in the SEER database with C-PTC, FV-PTC, and FTC from 1988-2011 were identified (Appendix 1). Other histological variants of PTC and minimally invasive follicular carcinoma were excluded from analysis. Patients with tumors <1 cm in size were excluded. The study time frame was chosen due to significant improvement in the collection of key data points beginning in 1988. NCDB data from 1998-2011 with these inclusion and exclusion criteria were used for validation.

After exclusion of C-PTC, multivariable logistic regression analysis was used to determine the independent association of the following characteristics with final tumor histology (FTC vs. FV-PTC): patient age, gender, race, tumor size, nodal status, presence or absence of distant metastases, and extrathyroidal extension. Restricted cubic splines of continuous data in relation to the log odds ratio (OR) were evaluated for linearity. In cases where linearity assumptions were violated, variables were transformed using linear splines or categorization for the final model. Model discrimination and calibration were assessed with C-statistics and calibration plots. The coefficients from the final model were transformed into a simplified scoring system by normalizing against the lowest coefficients (normalization ratio = 0.2). Scores were then grouped, and predicted rates of FTC final diagnosis were calculated. The discrimination and calibration of the simplified scoring system were then assessed and internally validated using bootstrapping techniques. Finally, the predictive model and simplified scoring system were externally validated in the NCDB by applying the models developed in the SEER database and evaluating the C-statistics and calibration plots.

Appendix 1. ICD-O-3 codes used to develop study population and classifications
ICD-O-3 / Classification / Description
8050 / C-PTC / Papillary carcinoma, NOS
8260 / C-PTC / Papillary carcinoma of thyroid
8330 / FTC / Follicular adenocarcinoma, NOS
8331 / FTC / Follicular adenocarcinoma, well differentiated
8332 / FTC / Follicular adenocarcinoma, moderately differentiated
8340 / FV-PTC / Papillary adenocarcinoma, follicular variant
C-PTC: Classical Papillary thyroid carcinoma; FTC: Follicular thyroid carcinoma; FV-PTC: Follicular variant-papillary thyroid carcinoma; NOS: not otherwise specified.

Appendix 2. Full model discrimination and calibration in SEER data 1988-2011.

Appendix 3. Predictors of FTC in full predictive model.
Variable / Coefficient / AOR (95% CI) / p-value
Intercept / -9.16 / <0.01 (<0.01, <0.01) / <0.001
Age (per decade <45) / 0.17 / 1.18 (1.12,1.25) / < 0.001
Age (per decade >45) / 0.16 / 1.18 (1.13,1.23) / < 0.001
Male / 0.22 / 1.25 (1.15,1.35) / < 0.001
Race (ref = White)
Black / 0.16 / 1.18 (1.05,1.33) / 0.006
Other / 0.05 / 1.05 (0.94,1.18) / 0.413
Tumor size (ref = 1-2 cm)
2.1-4 cm / 1.01 / 2.74 (2.52,2.99) / < 0.001
>4 cm / 1.53 / 4.63 (4.19,5.11) / < 0.001
Nodal metastases / -1.72 / 0.18 (0.15,0.21) / < 0.001
Distant metastases / 1.34 / 3.81 (2.88,5.04) / < 0.001
Extrathyroidal extension / -0.16 / 0.85 (0.76,0.96) / 0.007
FTC: follicular thyroid carcinoma; AOR: adjusted odds ratio; CI: confidence interval

Appendix 4. Full model discrimination and calibration in NCDB data 1998-2011.


Appendix 5. Simplified scoring system discrimination and calibration in NCDB data 1998-2011.

1