Clinical impact of testing for mutations and microRNAs in thyroid nodules

Abstract Background We report results of a multicenter clinical experience study examining the likelihood of patients with indeterminate thyroid nodules to undergo surgery or have malignant outcome based on multiplatform combination mutation and microRNA testing (MPT). Methods MPT assessed mutations in BRAF, HRAS, KRAS, NRAS, and PIK3CA genes, PAX8/PPARγ, RET/PTC1, and RET/PTC3 gene rearrangements, and the expression of 10 microRNAs. Baseline clinical information at the time of MPT and clinical follow‐up records were reviewed for 337 patients, of which 80% had negative MPT results. Kaplan Meier analysis for cumulative probability of survival without having a surgical procedure or malignant diagnosis over the course of patient follow‐up was determined for MPT results of 180 patients, among which only 14% had malignancy. Results A negative MPT result in nodules with Bethesda III or IV cytology (2009) conferred a high probability of non‐surgical treatment, with only 11% expected to undergo surgery and a high probability of survival without malignancy (92%) for up to 2 years follow up. A positive MPT result conferred a 57% probability of malignancy and was an independent risk factor for undergoing surgical treatment (Hazard Ratio [HR] 9.2, 95% confidence intervals 5.4‐15.9, P < .0001) and for malignancy (HR 13.4, 95% confidence intervals 4.8‐37.2, P < .0001). For nodules with weak driver mutations, positive microRNA test results supported high risk of cancer while negative results downgraded cancer risk. Conclusion MPT results are predictive of real‐world decisions to surgically treat indeterminate thyroid nodules, with those decisions being appropriately aligned with a patient's risk of malignancy over time.


| INTRODUCTION
When managing thyroid nodules physicians need to distinguish potentially malignant nodules from those with benign disease to limit unnecessary surgery. Ultrasound imaging followed by cytopathology review of fine needle aspirations (FNA) is the current standard of care for the diagnosis of thyroid nodule malignancy. Although cytopathology categorizes the majority of nodules as benign or malignant with a high degree of certainty, up to 30% of nodules are reported as indeterminate, where the presence of malignancy is less certain. [1][2][3][4] Only 12%-20% of surgically resected indeterminate nodules that have atypia of undetermined significance or a follicular lesion of undetermined significance (AUS/FLUS, Bethesda category III) diagnosis will have malignancy. 5 The malignancy rate only increases to 25%-33% in nodules that are a follicular neoplasm or are suspicious for follicular neoplasm (FN/SFN, Bethesda category IV). 5 Rates of malignancy in Bethesda category III and IV nodules are likely even lower when nodules that undergo observation rather than surgery are also considered. Regardless, indeterminate cytopathology often triggers unnecessary diagnostic lobectomy and even total thyroidectomy, given concern for cancer.
Molecular testing of nodules with indeterminate cytology diagnosis following a thyroid nodule FNA has been incorporated into routine care to help resolve diagnostic uncertainty. 6 While mutations strongly associated with malignancy, such as BRAF V600E, RET fusions, and TERT can assist in surgical decision making, other mutations considered weak drivers of cancer carry less certainty. 7-10 RAS mutations, the most common type in thyroid nodules, have been reported to have variable positive and negative predictive value for malignancy, often occurring in benign nodules. [11][12][13][14][15] Furthermore, residual risk of malignancy (5%-25%) is present in patients who lack any detectable mutational change. 16,17 Multiple studies have described the ability of RNA-based risk classifiers to help resolve diagnostic uncertainty in indeterminate thyroid nodules. [18][19][20] A messenger RNA risk classifier can help to rule out the need for surgery in indeterminate nodules due to its high negative predictive value for malignancy. 19,20 However, its less than optimal positive predictive value limits the ability to rule in the need for surgery, especially in AUS/FLUS nodules, where malignancy rates are low. [19][20][21] Comparatively, non-coding microRNA based risk classifiers have shown similarly high negative predictive value (NPV) for malignancy and superior positive predictive value when used in combination with analysis for oncogenic mutations and messenger RNA fusion transcripts. 18 Given the improved performance of the combination approach, this multiplatform combination of mutations and microRNA test (MPT) can effectively help to rule in and rule out the need for surgery. Clinical experience has shown that microRNA analysis can also help reclassify cancer risk under variable mutational profiles, including when patients have weak driver mutations such as RAS. 16 However, there has previously been limited data supporting the impact of MPT results on real-world decisions to surgically treat nodules or on the outcomes of patients who have undergone such testing in clinical practice.
Herein, we report the results of a multicenter clinical experience study of nodules that underwent MPT testing in clinical practice. We evaluated the impact of MPT results on real-world decisions to surgically treat indeterminate thyroid nodules with Bethesda category III or IV diagnosis. We also determined if those surgical decisions were appropriately aligned with a patient's risk of malignancy over a period of time based on MPT results.

| Study design and patient population
The study was an observational clinical experience study of consecutive patients who had multiplatform combination mutation analysis (ThyGenX) and microRNA classification (ThyraMIR) testing prescribed by their physician as standard of care. The study included medical record review, was non-interventional and was reviewed and approved by a central independent ethics review board (Quorum IRB #32262).
Informed consent was waived by the IRB due to minimal risk. The study included nine participating sites from across the United States. Subjects included patients ≥18 years of age who had initial, baseline molecular testing performed between July 2015 and January 2018 for indeterminate thyroid nodules due to Bethesda III or IV category cytology and who did not have a history of thyroid cancer. Study exclusion criteria included any subject that did not have past MPT results and those that had incidental thyroid cancer findings unrelated to their tested thyroid nodule.

| Baseline patient characteristics
Baseline patient demographic and clinical information was abstracted from de-identified patient history and physical examination records, ultrasound, cytology records, and clinical notes. This information included the patient age, nodule size, laboratory values at the time of molecular testing, family and personal histories of cancer (thyroid and non-thyroid), and non-malignant thyroid disease. Baseline information reflected that which was present when molecular testing was prescribed. Cytology results and corresponding Bethesda categorization (2009) were derived from cytology records of fine needle aspirates (FNA) for which molecular testing was prescribed.

| Baseline multiplatform mutation and microRNA testing (MPT)
Multiplatform mutation and microRNA testing (MPT) was performed clinically on thyroid nodule FNA aspirates as part of standard of care using ThyGenX and ThyraMIR commercial tests (Interpace Diagnostics).

| Statistical analysis
Statistical analysis was performed using R statistical software (r-project. org). P-values <.05 were considered statistically significant. Malignancyfree and surgery-free survival was calculated from the date of molecular testing using Kaplan-Meier survival analysis. A univariate and multivariate Cox proportional hazards regression model was used to assess the relationship between the independent clinical factors and MPT results compared to surgical treatment decisions and patient outcomes. The log rank test was used to compare the survival curves when appropriate.
The expected absolute risk of undergoing surgery over a range of pretest surgery probabilities was calculated for MPT results using rates observed in the study cohort and Bayes theorem. The probability of a negative MPT result (ie, the rate of negative test results) was calculated based on the performance characteristics of the test in the study cohort over a range of various baseline malignancy rates. 22

| RESULTS
Our clinical experience study included 337 consecutive patients with AUS/FLUS (Bethesda category III) or FN/SFN (Bethesda Category IV) nodules from nine participating institutions that were prescribed MPT testing as part of their current standard of care. In our initial analysis, 80% (270/337) of study subjects had a negative MPT result as previously defined, 18