Incremental utility of expanded mutation panel when used in combination with microRNA classification in indeterminate thyroid nodules

Abstract INTRODUCTION Focused and expanded mutation panels were assessed for the incremental utility of using an expanded panel in combination with microRNA risk classification. METHODS Molecular results were reviewed for patients who underwent either a focused mutation panel (ThyGenX®) or an expanded mutation panel (ThyGeNEXT®) for strong and weak oncogenic driver mutations and fusions. microRNA results (ThyraMIR®) predictive of malignancy, including strong positive results highly specific for malignancy, were examined. RESULTS Results of 12 993 consecutive patients were reviewed (focused panel = 8619, expanded panel = 4374). The expanded panel increased detection of strong drivers by 8% (P < .001), with BRAFV600E and TERT promoters being the most common. Strong drivers were highly correlated with positive microRNA results of which 90% were strongly positive. The expanded panel increased detection of coexisting drivers by 4% (P < .001), with TERT being the most common partner often paired with RAS. It increased the detection of weak drivers, with RAS and GNAS being the most common. 49% of nodules with weak drivers had positive microRNA results of which 33% were strongly positive. The expanded panel also decreased the number of nodules lacking mutations and fusions by 15% (P < .001), with 8% of nodules having positive microRNA results of which 22% were strongly positive. CONCLUSIONS Using expanded mutation panels that include less common mutations and fusions can offer increased utility when used in combination with microRNA classification, which helps to identify high risk of malignancy in the cases where risk is otherwise uncertain due to the presence of only weak drivers or the absence of all drivers.

follicular neoplasm or is suspicious for follicular neoplasm (FN/SFN).
Malignancy risk can range from 12% to 33% in these indeterminate nodules. 5 Without additional testing, diagnostic lobectomy is often required for definitive diagnosis, concluding most frequently in benign disease in the form of nodular hyperplasia, follicular adenoma, or related noncancerous processes for which tissue resection may have been unnecessary. These unnecessary surgeries result in significant healthcare costs and can negatively impact patient's quality of life. 6 When faced with an indeterminate cytology diagnosis, molecular testing is often used in routine clinical practice to further assess the risk of malignancy. There are a variety of commercially available tests on the market, each of which takes a different approach to assessing malignancy risk. One approach includes the use of mutation and messenger RNA fusion biomarker panels (ie, mutation panels) to identify oncogenic driver changes that may be present. More recently, these mutation panels have been expanded to include additional markers for mutations and fusions, with the goal of better predicting increased risk for malignancy and limiting diagnostic dilemmas encountered when nodules lack detectable oncogenic changes, where 5% to 25% risk of cancer exists. [7][8][9] However, inclusion of additional mutations and fusions that is not highly specific for malignancy may result in panels with lower positive predictive value (PPV).
Certain oncogenic drivers are strongly predictive of malignancy and aggressive thyroid cancer and as such can be considered "strong drivers." These include BRAFV600E, TERT promoter mutations (C228T and C250T), and RET mutations as well as BRAF and RET-related messenger RNA fusion transcripts. BRAFV600E and RET mutations and RET fusions have high PPV for malignancy, can be found in aggressive thyroid cancers, and are rarely if ever found in benign adenomas or hyperplastic nodules. [10][11][12] BRAF-related fusion transcripts have also been associated with BRAFV600E-like properties, although they are more commonly found in pediatric thyroid cancer populations 11,[13][14][15] TERT promoter mutations have been strongly correlated to persistent disease, aggressive forms of cancer, distant metastasis, and mortality. [16][17][18][19][20][21] Other oncogenic drivers can present a challenge to guiding patient management when they alone are used to assess malignancy risk.
These drivers have been more weakly associated with malignancy in thyroid nodules and as such can be considered "weak drivers." RAS mutations are the most common to indeterminate nodules and have been found in both benign and malignant nodules, presenting an uncertain PPV ranging from 15% to 70%. 8,[22][23][24][25] Other mutations and fusions can occur at a much lower frequency, making their PPV difficult to study and consequently not well understood. Rare BRAF mutations, excluding BRAFV600E, have RAS-like properties rather than BRAFV600E-like properties, and have been found in both benign and malignant thyroid nodules. 15,26,27 Although PIK3CA and PTEN mutations have been reported in follicular thyroid cancer, poorly differentiated thyroid cancers, and anaplastic thyroid cancer, [28][29][30] they can also be found in benign thyroid adenomas, as can GNAS and ALK mutations and THADA-and PPARG-related fusions. 28,[31][32][33][34][35][36][37][38][39][40][41][42] Furthermore, PPARG-and THADA-related fusion transcripts can have RAS-like properties, 15 while other fusions such as those related to NTRK and ALK have properties that are neither RAS-like nor BRAFV600E-like. 15 Although the predictive value for malignancy of these oncogenic changes is not well understood when found individually, it is well established that coexistence of many of these oncogenic changes along with other oncogenic drivers is generally associated with aggressive forms of thyroid cancer and poor prognosis. 28,29,33,[43][44][45] microRNA risk classifier testing has been used to help resolve diagnostic dilemmas encountered with the use of mutation panels alone. The microRNA classifier further assesses malignancy risk in nodules when mutation panels result in no oncogenic changes detected or in identification of oncogenic changes that have lower or less certain positive predictive value for malignancy. microRNAs reflect the output results of signaling pathways in a dynamic fashion providing valuable information of the behavior of the cells on the neoplastic spectrum. In contrast to mutations and fusions that occur at the intracellular level, microRNAs are uniquely designed to travel from one cell to another, regulating intercellular communication across multiple pathways, which places them central to understanding the transition from pre-cancer states through malignant transformation and spread of cancers. [46][47][48][49][50] A multiplatform approach using a combination of a mutation panel and a microRNA risk classifier has been shown to effectively "rule-in" and "rule-out" high risk of malignancy with results being predictive of surgical treatment decisions that are appropriately aligned with cancer risk. 6,51,52 Furthermore, microRNA risk classification has been shown to help to reclassify cancer risk in both the absence of mutational change and the presence of mutations that have lower positive predictive values for malignancy, with strong positive microRNA classifier results offering extremely high specificity for malignancy. 8 We aimed to better understand the incremental utility in using expanded mutation panels and how microRNA classifier testing can provide additional diagnostic information to expanded panel test results. We examined the frequency of strong and weak driver changes or the lack thereof in patients who underwent either focused mutation panel testing for more commonly tested strong and weak drivers or more robust, expanded mutation panel testing for additional strong and weak drivers. The latter cohort was also evaluated for  were reported as having Bethesda Diagnostic categories III or IV cytology results, or did not have a Bethesda Diagnostic category available in the data set examined. FNA specimens and corresponding cytology results were those from independent pathology practices, community hospital pathology departments, large metropolitan medical centers, and tertiary care academic centers in the United States and Canada. All molecular and cytology data were held in a secure central database as part of standard clinical practice. Use of de-identified molecular and cytology data from the secure database was IRB approved (Quorum Review#: 31963) for use in this study. Informed consent was waived by the IRB due to minimal risk.  Table 1 using polymerase chain reaction (PCR) to amplify the regions of interest prior to sequencing. A sequencing read depth of 1 000 was required for variant calls. Specimens were required to contain at least 3% of BRAFV600E for a positive variant call, 10% of GNAS, or 5% of the other individual DNA variants in the panel. For a positive mRNA fusion transcript call, specimens were required to contain at least 5% of an individual mRNA fusion transcript.

| Molecular analysis
Mutations and fusions common to both the focused and expanded mutation panels and those unique to the expanded panel are listed in Table 1. Mutations and fusions were categorized as strongly associated with thyroid malignancy and aggressive thyroid cancer (ie, strong drivers) or weakly associated (ie, weak drivers) based on published evidence described in the Introduction.
Mutations and fusions in each of these categories for the focused and expanded panels are listed in Table 1 Notes: Mutations and fusions were categorized as being strongly associated with malignancy and aggressive cancer (ie, strong drivers, bold font) or more weakly associated (ie, weak drivers, regular font) as described in the Introduction. BRAF X* indicates BRAF mutation other than BRAFV600E.

| Statistical analysis
Demographic differences in cohorts that underwent focused panel testing compared with expanded panel testing were compared by the Z test for proportions using the R statistical software (r-project.org).
The percent differences in the frequency of patients with strong and weak driver mutations or the lack thereof between the cohorts that underwent focused compared with expanded mutation panel testing were performed using the Z test for proportions using the R statistical software. These differences were also examined in the subset of patients who had detectable oncogenic change. The P values of <.05 were considered statistically significant. and/or association with aggressive disease, as described in Table 1 (bold font). Other mutations and fusions were considered weak drivers based on the literature supporting their presence in both benign and malignant thyroid nodules, their RAS-like signatures, and/or the lack of literature supporting their high positive predictive value for malignancy or aggressive behavior, as described in Table 1 (regular font).

| RESULTS
In    have RAS-like signatures, and/or lack support for their PPV for malignancy, 8,14, and as such can be considered more weakly associated with malignancy (ie, weak drivers). We examined strong and weak driver mutations and fusions in thyroid nodules that underwent clinically prescribed molecular testing with either focused or expanded mutation panels to better understand the incremental utility in using an expanded mutation panel and how microRNA classifier testing can provide additional diagnostic information to expanded panel test results.
Expansion of the panel increased detection of strong drivers by 8% in patients who had oncogenic change. The higher frequency of strong drivers in patients who underwent expanded panel testing was largely due to inclusion of TERT promoter mutations in the expanded panel. TERT promoter mutations have been associated with poorly differentiated and anaplastic thyroid cancer. [16][17][18][19]21 They are also considered an independent risk factor for persistent disease, distant metastases, and mortality for well differentiated thyroid cancer. 16,20 Given this prognostic information, inclusion of TERT promoter mutation testing in the expanded panel may provide enhanced clinical utility. In our study, nodules with strong drivers, such as TERT, typically had positive microRNA results consistent with high risk of malignancy, the majority of which were strong positive microRNA results that are highly specific for malignancy. 8  Coexistence of weak and strong drivers not only elevates concern for malignancy but also elevates concern for aggressive thyroid cancer.
Coexisting TERT promoter mutation with BRAFV600E or RAS mutation was the most common, with such coexisting mutations having the potential to promote aggressive tumor behavior and predict poor patient survival. 17,19,43,45 It is well established that poorly differentiated and anaplastic thyroid cancer harbors multiple oncogenic drivers, including coexisting RAS and PIK3CA mutations detected in our study. 21 Well-differentiated papillary cancers can also have multiple oncogenic drivers, which typically indicate aggressive tumor behavior. 28,29,33,44 In such cases, these molecular findings may enable an optimal surgical approach to include lymph node sampling.
All oncogenic changes can contribute to neoplastic growth and progression, and therefore both strong and weak drivers should be F I G U R E 4 The frequency of (A) positive and negative microRNA classification in all patients who underwent expanded panel testing who had strong drivers (n = 270), weak drivers (n = 865), or the lack thereof (n = 3239) and the frequency of (B) positive and strong positive microRNA classification among all patients who had positive microRNA results based on the presence of strong drivers (n = 222), weak drivers (n = 428), or the lack thereof (ie, no mutations or fusions, n = 246) as determined by expanded mutation panel testing considered clinically important. However, the individual detection of drivers that are more weakly associated with malignancy presents a challenge to guiding patient management. 25  to either mechanism, the lack of detectable oncogenic change cannot provide complete assurance that a nodule is benign and as such those nodules have a residual cancer risk of 5% to 25%. [7][8][9] Our results support that microRNA testing can help to overcome limitations in assessing malignancy risk in the absence of detectable mutations and fusions. The extracellular distribution of microRNAs may help to identify higher risk lesions even when mutational heterogeneity and sampling variability may be present and no mutations or fusions are detectable. 51 Our results demonstrate that strong positive microRNA results that are highly specific for malignancy 8 and prevalent in nodules with strong drivers can also be found in a subset of patients who lack any detectable mutations and fusions even when a more expanded mutation panel is used. Additional molecular testing that is highly specific for malignancy can help to identify patients at higher risk of malignancy in this subgroup. Given the reported high specificity of positive microRNA results and their strong association with strong driver oncogenic changes predictive of malignancy demonstrated here, positive and even more so strong positive microRNA results can support elevated concern for malignancy and can justify more aggressive treatment options even in the absence of detectable mutations and fusions. 8,51 This is further supported by results of a multicenter study for the outcomes of patients who underwent clinically prescribed combination mutation and microRNA testing concluding that more aggressive management options are warranted in patients that lack mutations and fusions but have positive microRNA results given confirmed high risk of malignancy. 52 Although combination mutation and microRNA testing has been validated in patients with surgically confirmed outcomes, 51