A retrospective analysis of molecular testing in cytologically indeterminate thyroid nodules with histologic correlation: Experience at a heterogenous multihospital system

Thyroid malignancy is one of the most common types of cancer in developed nations. Currently, fine‐needle aspiration cytology (FNAC) is the most practical screening test for thyroid nodules. However, cytologically indeterminate samples comprise approximately 15%–30% of cases. These include cases classified as atypia of undetermined significance (AUS), follicular neoplasm (FN), and suspicious for malignancy (SFM). Indeterminate cases can be sent for molecular testing for more definitive classification to help guide management and prevent overtreatment of benign thyroid nodules. We conducted a retrospective review on molecular testing of indeterminate thyroid FNAC and reviewed subsequent histologic diagnoses in resection specimens to assess how molecular testing supported a diagnosis and its effect on clinical management of patients at our institution.

with lesions that are difficult to diagnosis based solely on FNAC morphology.The types of genetic alterations identified in the resected thyroid lesions were consistent with what has been previously described in the literature.Additionally, we found that in the patients with indeterminate thyroid FNAC with adjunct molecular testing, more than half did not undergo surgical resection.This finding emphasizes the value of adding molecular testing in patients, particularly when attempting to reduce unnecessary surgical intervention.

| INTRODUCTION
Thyroid nodules are common, with a prevalence of up to 6% detected by palpation, up to 35% detected with ultrasound examination, and up to 65% discovered at autopsy. 1 Additionally, thyroid malignancy is one of the most common types of cancer in the United States and other developed countries, currently being the 7th and 8th most common cancer in women and men, respectively.The overwhelming majority of thyroid nodules are benign, thus screening guidelines based on clinical, radiologic, and pathologic findings have been established in order to appropriately triage the management of patients with thyroid nodules.Currently, fine-needle aspiration cytology (FNAC) is the most practical, cost effective, and accurate screening test for thyroid nodules. 2 The Bethesda System for Reporting Thyroid Cytopathology (TBSRTC) has greatly influenced thyroid cytopathology by standardizing the way pathologists communicate diagnostic findings of FNAC to clinicians.Subsequent management of thyroid nodules after FNAC is largely dictated by the resulting cytologic Bethesda classification.Asymptomatic nodules with a low risk of malignancy (ROM) are most often followed with clinical and radiologic observation, while nodules with a high ROM warrant surgical intervention.
There are, however, limitations to thyroid FNAC.Diagnostically indeterminate samples comprise approximately 15%-30% of cases and include those classified as atypia of undetermined significance (AUS, TBSRTC category III), follicular neoplasm (FN, TBSRTC category IV), and suspicious for malignancy (SFM, TBSRTC category V). 3 Historically, algorithms for treating indeterminate nodules included diagnostic lobectomy or active surveillance; however, many indeterminate aspirates are benign on surgical resection.Presently, in the era of personalized medicine and preoperative risk assessment, molecular testing has become yet another tool to guide management of cytologically indeterminate thyroid nodules.The current American Thyroid Association (ATA) and the American Association of Endocrine Surgeons (AAES) recommend molecular testing on thyroid FNAC for select patients with indeterminate nodules classified under the TBSRTC as AUS, FN, and SFM. 4 The current 2023 National Comprehensive Cancer Network Guidelines recommend molecular testing on thyroid FNAC for select patients with indeterminate nodules classified under the TBSRTC as AUS or FN.6][7] The management of thyroid nodules requires a multidisciplinary approach, utilizing clinical features, laboratory values, radiologic findings, and cytomorphologic findings to develop an effective treatment plan.In this approach, molecular testing can serve as a complementary test to cytopathology for clinical triaging and surgical planning.
We conducted a retrospective review of thyroid nodule cases at our institution to evaluate how FNAC with adjunct molecular testing correlated with subsequent surgical resection specimens.As a large, multihospital institution servicing a diverse patient population, our robust cytopathology department evaluates a large volume of thyroid FNAC specimens per year.These cases originate from a number of different endocrinologists and other proceduralists practicing within different clinics throughout the hospital system.Within these diverse practice settings, there is the potential to utilize four different molecular testing platforms for thyroid FNAC.We were curious to investigate the heterogeneity of our practice conditions, which are not unlike many other institutions around the world.The purpose of this study is to characterize aspirated thyroid nodules with genetic aberrations found on preoperative molecular analysis and correlate final surgical histologic outcomes to explore the impact of molecular cytopathology on facilitating a more accurate, optimal assessment of risk of malignancy and its ultimate effect on patient management.

| MATERIALS AND METHODS
Our study protocol was approved by our institutional investigational review board (IRB) and the study was granted a waiver of informed consent for patients whose data were collected for study analysis since all data was originally generated for patient care purposes.We retrospectively reviewed reports from thyroid FNAC specimens, corresponding molecular testing, and subsequent surgical resection specimens (including thyroid lobectomy and total thyroidectomy) from January 1, 2015 to October 14, 2021.A chart review of the patients included was performed to assess certain clinical characteristics and radiologic features related to their thyroid nodule(s).All specimens were collected in the Houston Methodist (HM) Hospital system (including the main hospital campus and six peripheral hospitals) located in the metropolitan area of Houston, Texas, USA.All cytologic and surgical cases have been diagnosed by board certified cytopathologists associated with the HM Department of Pathology and Genomic Medicine.FNAC were classified using the TBSRTC.In our study, we have updated the terminology used in retrieved surgical resection reports to the most recent 2022 World Health Organization (WHO) Classification of Thyroid Tumors. 8For example, the diagnoses of "adenomatous nodule" and "Hurthle cell adenoma" have been replaced with the updated terms "thyroid follicular nodular disease" (TFND) and "oncocytic adenoma of the thyroid" (OAT), respectively.0][11] The Yates Chisquare method was used to determine p-values, using the online statistics tool available at vassarstats.net.The majority of thyroid FNA specimens interpreted at HM are received from the HM radiology department with immediate assessment performed by a certified cytopathologist.It is our practice across all our hospital systems to prepare direct smear slides for thyroid FNAC.The needle rinses are placed in CytoLyt ® and may be subsequently used for ThinPrep ® , cell block, or molecular testing using the HMTCT.During the study period, dedicated passes for molecular testing by a commercial laboratory (i.e., Afirma ® , ThyGeNEXT ® , and Thyroseq ® ) were collected in the appropriate proprietary media at the time of FNA only if specifically requested by the ordering physician.Some endocrinology clinics affiliated with HM that do not have access to immediate assessment by a certified cytopathologist also collect specimen in designated media for molecular testing at the time of the initial FNA, with the type of testing performed varying according to the preferences of the requesting physicians.In some cases, material for molecular testing was collected during a second FNA procedure following an initial indeterminate FNAC diagnosis.
Our in-house HMTCT uses a combination of single base extension genotyping and targeted next-generation sequencing (NGS)-based assays to detect mutations and fusions in genes associated with thyroid cancer, as described above.This test is validated for formalin-fixed paraffin-embedded tissue specimens, as well as FNA needle rinses in CytoLyt ® , and therefore may be performed on initial FNA specimens that were not collected in proprietary media.If this testing option is performed, it is ideal to send the entire rinse material in CytoLyt ® to maximize DNA content.
ThyGeNEXT ® (expanded version of ThyGenX ® ) and ThyraMIR™ is a combined NGS testing platform.ThyGeNEXT ® uses a targeted NGS oncogene panel to detect alterations in genes associated with thyroid cancer.ThyraMIR™ is a miRNA classifier reflexively performed after a weak gene alteration (i.e., RAS-like) is detected by ThyGeNEXT ® .A ROM is estimated from the two tests based on genotype and miRNA profile. 12yroSeq ® V3 Genomic Classifier is a DNA and RNA targeted NGS platform used to detect gene alterations associated with thyroid neoplasms.Results are reported as positive or negative along with detailed molecular findings, risk of recurrence, and suggested personalized patient management. 13e Afirma ® Gene Expression Classifier (GEC) uses DNA microarray testing and the Afirma ® Genomic Sequencing Classifier (GSC) uses RNA NGS along with analysis by data-based machine learning to create a binary gene classifier model.Their proprietary algorithm classifies lesions as having either a benign (ROM 4%) or suspicious (ROM 50%) result.Suspicious lesions can then be analyzed with the Afirma ® Xpression Atlas (XA), which uses targeted RNA sequencing to detect expressed variants and fusions. 14lecular testing was performed in 16% (n = 178/1102) of indeterminate cytology cases (2% total, n = 178/10,253; AUS: 80%, n = 142/178; FN: 19%, n = 33/178; SFM: 2%, n = 3/178) (shown in Figure 1).HMTCT was performed in 22% (n = 40/178) of cases.ThyGenX ® and ThyGeNEXT ® /ThyraMIR™ was performed in 7% (n = 13/178) of cases.ThyroSeq ® V3 was performed in 7% (n = 13/178) of cases.).Afirma ® GEC and GSC was performed in 63% (n = 112/178) of cases including AUS.
One of the FTUMP cases had insufficient RNA for molecular testing.

| DISCUSSION
Our study focused on 178 cases of indeterminate FNAC cases that underwent molecular testing.A variety of different genetic alterations were observed in the molecular tests, but only a few were in the "BRAF-like" family of alterations associated with classical PTC and more aggressive behavior.This shows the efficacy of cytology in identifying these lesions as positive for malignancy by cytomorphology alone in the majority of cases.Most of the positive findings were in the "RAS-like" family associated with less aggressive follicularpatterned lesions that are more subtle cytologically.We found that lesions that were classified as cytologically indeterminate were predominantly low grade follicular derived lesions on histologic correlation.This is not unexpected as follicular derived lesions pose an especially difficult challenge when interpreting FNAC, and many of these lesions require histologic examination of the entire capsule to evaluate for invasion for definitive diagnosis.Molecular testing enabled refinement of the ROM for this subset.For example, our case with a PAX8-PPARG translocation had a ROM of 25%-40%, versus other cases with NRAS mutations that had a ROM as high as 95%.
Lesions previously classified as adenomatous nodules, colloid nodules, or adenomatous hyperplasia (now reclassified as TFND) have traditionally not been considered neoplastic.6][17][18][19] In our study, there were 5 cases of TFND that harbored genetic alterations.Foci of malignant transformation are known to occur in nodules previously T A B L E 1 Results of molecular testing in fine needle aspiration cytology (FNAC) and subsequent surgical resection histologic diagnosis.diagnosed as TFND, which can be explained by the detected clonality in some of these lesions. 8n-invasive follicular thyroid neoplasm with papillary-like nuclear features (NIFTP) was first introduced in 2016 as a nomenclature revision to tumors that were previously classified as noninvasive encapsulated follicular variant of papillary thyroid carcinoma and was recognized as a distinct entity by the World Health Organization (WHO) in 2017. 20,21This change resulted from data supporting the low risk biologic potential of this neoplasm and introduced a paradigm shift to reduce overtreatment.0][11] Currently, the NCCN and AAES consider lobectomy alone as adequate treatment for NIFTP. 22,23Therefore, for the purposes of this study, NIFTP was considered in the benign group for final histologic diagnosis as these tumors have been shown to have extremely indolent behavior and clinically are treated similarly to follicular adenoma.The introduction of NIFTP has posed diagnostic dilemmas in molecular testing.The available commercial molecular tests tend to classify NIFTP as being suspicious for malignancy or low risk of malignancy. 24,25Suspicious results on molecular testing may lead to overtreatment of this entity if total thyroidectomy or lymph node dissection is performed over a hemithyroidectomy. 26yroid tumors of uncertain malignant potential are low risk follicular cell-derived neoplasms which are generally encapsulated or welldemarcated with equivocal capsular and/or vascular invasion. 27These tumors have low risk of distant metastases, tumor recurrence, and tumor related death and are commonly associated with RAS-like mutations. 28,29e borderline tumors in our cohort were classified as being of uncertain malignant potential based on the inability to definitively classify the invasive nature of the lesion.In the case where a molecular alteration was identified, a more definitive risk stratification could be determined for the patient based on the type of mutation or translocation found.For example, a BRAF-like mutation or TERT promoter found in a tumor of uncertain malignant potential would indicate more aggressive tumor biology and prompt surgical removal. 29,30This is in contrast to the situation in which no mutation was found and clonality of the lesion cannot be proven and a more conservative approach to follow-up may be chosen.
Follicular patterned tumors, such as follicular carcinoma and IEFVPTC are known to harbor somatic mutations in the RAS gene family. 8The most common of these are NRAS mutations involving codon 61, which we did detect in our study group.RAS mutated tumors tend to confer a better prognosis over some other mutations associated with thyroid cancer, such as BRAF. 31PAX8-PPARG translocations are seen in less than 10% of follicular adenomas, up to 30% of NIFTP, and up to 40% of IEFVPTC.8.This gene rearrangement is frequently reported in follicular patterned lesions and in our cohort was detected in an OAT. 8 BRAF p.K601E mutation is the second most common BRAF mutation found in thyroid nodules. 32The p.K601E mutation is associated with follicular patterned thyroid cancers, as was seen in our study, and tends to confer a better prognosis than BRAF p.V600E positive tumors. 324][35] In our study, a case of an OAT was found to harbor both NRAS and EIF1AX mutations, allowing for a higher risk of malignancy classification in this tumor than if a RAS mutation alone was found.
RET gene fusions are common in PTC and are the most common genetic aberrancy found in PTC associated with radiation exposure. 36,37TERT promoter mutations are associated with aggressive tumor biology either as an isolated mutation or when present as a second hit on top of another driver mutation. 38TERT promoter mutations may act synergistically with concurrent mutations, such as BRAF to promote aggressive tumor behavior. 39TERT mutations are found more frequently in poorly differentiated and anaplastic thyroid carcinomas than in other thyroid carcinoma types. 40 our study cohort, there were 4 malignant cases in which no genetic aberration was detected.In 1 of the 4 cases (tested by Afirma), it was noted that nucleic acid yield was insufficient for complete testing.However, for the other 3 cases, which were also tested using the HMTCT and Afirma panels, discrepancies may be explained by a number of reasons.For example, genetic alterations outside of the scope of the targeted panels may have been present or allele fraction may have been below the limit of detection.Sampling error may have also played a role in these findings, as overall tumor percentage may differ between each pass.
Seven of these cases (54%, n = 7/13) had subsequent surgical treatment.All cases analyzed using ThyroSeq ® V3 were negative for copy number alterations and gene expression profile.All the cases that had negative results by ThyGenX ® /ThyGeNEXT ® /ThyraMIR™ and ThyroSeq ® V3 (46%, n = 6/13) and a low predicted ROM did not receive surgical treatment.[Color figure can be viewed at wileyonlinelibrary.com] surgical decision making.We found that surgery was performed in a higher percentage of cytologically indeterminate nodules with adjunct molecular testing that detected a genetic alteration compared to those without a detectable genetic alteration (75%, n = 52/69 vs. 18%, n = 20/109).4][35] In our study, the majority of indeterminate nodules that underwent molecular testing and subsequent surgical resection had a lobectomy versus a total thyroidectomy (75%, n = 54/72 vs. 25%, n = 18/72, respectively) (shown in Figure 5).However, although we observed that in our institution diagnostic thyroid lobectomy was performed more often than total thyroidectomy in indeterminate nodules with molecular testing, we found no statistically significant evidence that molecular testing alone influenced this decision.Our patient cohort demonstrated that, on average, patients who did not receive surgical treatment were 5 years older than patients who did receive surgery.It is possible that older patients are less likely to be good surgical candidates.
Some multicenter studies have shown that negative results on molecular testing decreased rates of surgical intervention, as was the observed trend in our study.However, other studies have shown conflicting results.F I G U R E 4 Afirma ® Gene Sequencing Classifier (GSC) and Xpression Atlas molecular testing results from fine needle aspiration cytology specimen and subsequent surgical resection histologic diagnosis.A total of 112 FNAC cases (63%, n = 112/178) were analyzed using this test.Forty-one of these cases (37%, n = 41/112) resulted in surgical treatment.Of note, few cases were analyzed with the Afirma ® GEC, however none of these cases resulted in surgical resection.Of the Afirma ® cases that did not undergo surgical treatment (37%, n = 66/178), 50 cases (76%, n = 50/66) were classified as benign (ROM 4%) by either Afirma ® GEC or GSC testing, 12 cases (20%, n = 13/66) were deemed as suspicious (ROM > 50%), and three cases (5%, n = 3/66) had insufficient material for molecular testing.One case that was suspicious for malignancy on FNAC was sent only for the Afirma ® XA testing and was found to harbor SPOP c.281C>G (p.P94R), a mutation of unknown clinical significance.[Color figure can be viewed at wileyonlinelibrary.com] platforms determine ROM by different methods and algorithms, they all provided useful information for patients in our cohort.

| CONCLUSION
Molecular testing on cytologically indeterminate thyroid nodules can help provide a more accurate risk of malignancy assessment in patients with lesions that are difficult to diagnose based solely on FNAC morphology.Most of the positive findings were in the "RASlike" family associated with low grade, less aggressive follicularpatterned lesions that are more subtle cytologically.Molecular testing enabled refinement of the ROM for this subset.The types of genetic alterations identified in the resected thyroid lesions were consistent with what has been previously described in the literature.Additionally, we found that patients with genetic aberrations identified on molecular testing of indeterminate FNAC tended to be more likely to receive surgical treatment than those whose molecular testing results were negative.Molecular testing in cytologically indeterminate thyroid nodules is an evolving field and our findings highlight the utility of this ancillary testing modality and its ability to aid in the management of cytologically indeterminate thyroid nodules.
F I G U R E 5 Surgical thyroid resection specimen by type of surgery (thyroid lobectomy vs. total thyroidectomy), whether a genetic alteration was present, and the associated fine needle aspiration cytology diagnosis.AUS, atypia of undetermined significance; FN, follicular neoplasm; SFM, suspicious for malignancy.[Color figure can be viewed at wileyonlinelibrary.com] Several molecular testing platforms were utilized in this study including an in-house panel, the Houston Methodist Thyroid Cancer Test (HMTCT), which assesses gene mutations in BRAF, HRAS, NRAS, and KRAS by multiplex PCR, single base probe extension, and MALDI-TOF analysis (OncoCarta Panel and Sequenom MassArray instrument, Agena Biosciences) and gene rearrangements of RET, PPARG, NTRK1, and NTRK3 by next generation sequencing (CTL FUSIONPlex lung and thyroid kit, ArcherDX, Boulder, CO and NextSeq550 instrument, Illumina).Commercial reference laboratory molecular tests included ThyGeNEXT ® Thyroid Oncogene Panel (including the former ThyGenX ® ) and ThyraMIR™ microRNA expression test (Interspace Diagnostics, Parsippaney-Troy Hills, NJ), ThyroSeq ® (Sonic Healthcare, Rye Brook, NY), and Afirma ® (Veracyte, South San Francisco, CA).
Oncocytic adenoma of the thyroid (n=3) Follicular variant of papillary thyroid carcinoma (n=1) Classic papillary thyroid carcinoma (n=1) Follicular variant of papillary thyroid carcinoma (n=1) Non-invasive follicular thyroid neoplasm with papillary-like nuclear features (n=2) Follicular carcinoma (n=1) Follicular tumor of uncertain malignant potential (oncocytic) (n=1) Follicular variant of papillary thyroid carcinoma (n=1) Follicular adenoma (n=1) Follicular tumor of uncertain malignant potential (n=1) Follicular adenoma (n=1) Indeterminate Cytology Thyroid follicular nodular disease (n=2) Follicular tumor of uncertain malignant potential (oncocytic) (n=1) Classic papillary thyroid carcinoma (n=1) Follicular adenoma (n=1) Thyroid follicular nodular disease (n=1) F I G U R E 2 Houston Methodist Thyroid Cancer Test (HMTCT) molecular testing results from fine needle aspiration cytology specimen and subsequent surgical resection histologic diagnosis.A total of 40 FNAC cases (22%, n = 40/178) were analyzed using this test.Twenty of these cases underwent surgical resection (50%, n = 20/40).Of the 20 cases that did not result in surgical treatment, 17 had no mutation detected and 3 cases showed genetic alterations ((HRAS (p.Q61R) (n = 1) and NRAS (p.Q61R) (n = 2)).One case diagnosed as follicular variant of papillary thyroid carcinoma had insufficient RNA for translocation testing.[Color figure can be viewed at wileyonlinelibrary.com] follicular thyroid neoplasm with papillary-like nuclear features (n=7) Follicular variant of papillary thyroid carcinoma (n=4) Follicular tumor of uncertain malignant potential (n=1) Medullary thyroid carcinoma (n=1) Thyroid follicular nodular disease (n=3) Follicular adenoma (n=1) Follicular variant of papillary thyroid carcinoma (n=1) Oncocytic adenoma of the thyroid (n=1) Classic papillary thyroid carcinoma (n=3) Oncocytic adenoma of the thyroid (n=2) Xpression Atlas: positive Follicular tumor of uncertain malignant potential (n=1) Parathyroid gene expressionsignature Parathyroid hyperplasia (n=1) No results due to inadequate RNA yield Oncocytic adenoma of the thyroid (n=1) Follicular variant of papillary thyroid carcinoma (n=1) Follicular tumor of uncertain malignant potential (n=1) NRAS Classic papillary thyroid carcinoma (n=1) Thyroid follicular nodular disease (n=1) Follicular carcinoma (n=1) /A; Xpression Atlas: negative 6,32It is important to note that current indications for Our data allows for an analysis of different approaches to molecular testing, as we are able to analyze a variety of different testing platforms ordered by clinicians.However, this study was not designed to compare or evaluate the performance characteristics of the different available proprietary tests.Although the various molecular testing