MicroRNA-Based Risk Score for Predicting Tumor Progression Following Radioactive Iodine Ablation in Well-Differentiated Thyroid Cancer Patients: A Propensity-Score Matched Analysis

Abstract

To identify molecular markers that can accurately predict aggressive tumor behavior at the time of surgery, a propensity-matching score analysis of archived specimens yielded two similar datasets of DTC patients (with and without RAI). Bioinformatically selected microRNAs were quantified by qRT-PCR. The risk score was generated using Cox regression and assessed using ROC, C-statistic, and Brier-score. A predictive Bayesian nomogram was established. External validation was performed, and causal network analysis was generated. Within the eight-year follow-up period, progression was reported in 51.5% of cases; of these, 48.6% had the T1a/b stage. Analysis showed upregulation of miR-221-3p and miR-222-3p and downregulation of miR-204-5p in 68 paired cancer tissues (p < 0.001). These three miRNAs were not differentially expressed in RAI and non-RAI groups. The ATA risk score showed poor discriminative ability (AUC = 0.518, p = 0.80). In contrast, the microRNA-based risk score showed high accuracy in predicting tumor progression in the whole cohorts (median = 1.87 vs. 0.39, AUC = 0.944) and RAI group (2.23 vs. 0.37, AUC = 0.979) at the cutoff >0.86 (92.6% accuracy, 88.6% sensitivity, 97% specificity) in the whole cohorts (C-statistics = 0.943/Brier = 0.083) and RAI subgroup (C-statistic = 0.978/Brier = 0.049). The high-score group had a three-fold increased progression risk (hazard ratio = 2.71, 95%CI = 1.86–3.96, p < 0.001) and shorter survival times (17.3 vs. 70.79 months, p < 0.001). Our prognostic microRNA signature and nomogram showed excellent predictive accuracy for progression-free survival in DTC.