Development of an integrated, multi-modal approach to improve diagnostic accuracy and reliability for pre-operative planning in pediatric thyroid carcinoma
Mentor Name: Ted Laetsch
Thyroid carcinoma is second most common carcinoma in adolescent girls in the United States. Treatment typically consists of surgery, either lobectomy or total thyroidectomy, as well as central and/or lateral lymph node dissection. Therapeutic radioactive iodine (RAI) is also used for patients with higher risk disease. These therapies carry the potential for lifelong side effects including hypothyroidism, pulmonary fibrosis, and secondary malignancy. Thus, risk stratifying patients to allow them to receive the least toxic therapy required to cure their disease is a critical need. There are multiple independent risk classification systems for thyroid cancer (ultrasound - American College of Radiology Thyroid Imaging and Reporting Data System [ACR-TIRADS] and fine needle aspiration (FNA) - The Bethesda System for Reporting Thyroid Cytopathology [TBSRTC]). Results from ultrasound and FNA can be inconclusive because both TIRADS and TBSRTC are subjective with broad inter- and intra-observer variability in interpretation. Further, multiple recent studies have shown that specific genetic alterations correlate with the risk of malignancy and invasive/metastatic behavior. However, there is currently no consistent risk stratification scheme for patients with newly diagnosed thyroid nodules that incorporates ultrasound, FNA, and molecular testing data to generate a single risk classification/score. Using data from more than 350 patients treated at CHOP, this study will generate an ordinal logistic regression model to create a combined risk stratification score for pediatric thyroid cancer. The predictors to be evaluated in the model are the ACR-TIRADS score from ultrasounds, TBSRTC score from FNA, and risk stratification for genetic mutations, along with disease status data. The outcome of the model will be an ordinal variable to predict the disease status of 1) intrathyroidal (i.e. thyroid only) 2) local metastasis (i.e. spread to lateral or central lymph nodes), or 3) distantly metastatic (past the neck). With the fitted model, we will be able to calculate the probability of a patient falling into one of these three categories based on the values of the predictors. An integrated diagnostic approach has great potential to more accurately select patients for surgery and inform preoperative planning.
