This case series examines the accuracy of a genomic classifier in diagnosis and treatment of patients with indeterminate thyroid nodules.
Key Points
Question
Does the diagnostic performance of a genomic classifier vary based on the size of indeterminate thyroid nodules (ITNs)?
Findings
In this case series of 218 ITNs, the positive predictive value and specificity of the genomic classifier tested significantly improved in ITNs larger than 4 cm. Negative predictive values remained high across all size categorizations.
Meaning
The findings of this study suggest that, in genomic classifier–negative ITNs larger than 4 cm, initial management with thyroid lobectomy may be sufficient.
Abstract
Importance
Genomic classifiers were developed to better guide clinicians in the treatment of indeterminate thyroid nodules (ITNs). To our knowledge, whether there is variation in the diagnostic accuracy of these tests depending on ITN size has not been previously studied.
Objective
To analyze the diagnostic performance of a genomic classifier in relation to ITN size.
Design, Setting, and Participants
A case series study with medical records review was conducted including all patients with a cytologic diagnosis of ITN managed with genomic classifier testing and surgery from January 2015 to December 2018 at NYU Langone Health. Demographics, ITN characteristics, genomic profiles, treatment, and final pathologic findings were recorded. Data analysis was conducted from March to April 2021.
Main Outcomes and Measures
The primary aim was to assess the positive predictive value (PPV), negative predictive value (NPV), sensitivity, and specificity of a genomic classifier test (ThyroSeq) in relation to ITN size (<2, 2-4, and >4 cm). The secondary aim was to investigate the risk of cancer associated with genetic signatures.
Results
Of the 212 patients with 218 ITNs, 158 (74.5%) were women; median (SD) age was 49 (15.6) years. Genomic classifier results were positive in 173 ITNs (79.4%) treated with surgery. In this group of 173 positive ITNs, 46 (26.6%) were malignant on final pathologic testing. Overall, the observed cancer prevalence in the population was 23.9% (52 ITNs). In 45 ITNs that underwent surgery despite a negative genomic classifier interpretation, 6 (13.3%) were malignant. The PPV of a positive test was 27% and the NPV was 87%. The PPV and NPV findings improved as the ITN size increased (<2 cm [n = 98]: PPV, 25%; NPV, 79% vs >4 cm [n = 33]: PPV, 50%; NPV, 89%). Test specificity was higher in larger ITNs (<2 cm: 15% vs >4 cm: 40%; P = .01). Isolated RAS sequence variations were the most common variant identified in malignant nodules (11 [21.1%] of all ITNs), followed by BRAF variants (7 [13.5%] of all ITNs).
Conclusions and Relevance
In this case series, the performance of the ThyroSeq test improved for larger ITNs. The risk of cancer in large ITNs with negative test results was low. These data suggest that, in genomic classifier–negative ITNs larger than 4 cm, initial management of thyroid lobectomy may be sufficient.
Introduction
Thyroid nodules are common, with palpable nodules presenting in up to 5% of the general population.1 Fine-needle aspiration (FNA) biopsy is the preferred and recommended method of assessing thyroid nodules that are clinically or radiologically indicated for further evaluation. Evaluation of large thyroid nodules remains an area of debate because FNA biopsies in large (>4 cm) thyroid nodules have shown variable false-negative rates ranging from 0% to 13%.2,3 Based on the Bethesda System for Reporting Thyroid Cytopathology,4 20% to 30% of all FNA specimens return as cytologically indeterminate. Indeterminate thyroid nodules (ITNs) include Bethesda III (atypia or follicular lesion of undetermined significance) and Bethesda IV (follicular neoplasm or suspicious for follicular neoplasm).5 Owing to the broad risk of cancer (6%-40%) in ITNs,5 diagnostic lobectomy is recommended for definitive diagnosis. Initial management of total thyroidectomy is currently recommended for ITNs larger than 4 cm to limit the need for remedial surgery.6 More recently, molecular assays, such as the ThyroSeq Genomic Classifier (Sonic Healthcare USA), have been implemented in clinical care to risk-stratify ITNs and potentially reduce the number of patients who may undergo surgery unnecessarily.
The goal of genomic classification is to identify molecular alterations that are associated with cancer. This identification can help to triage patients with ITNs at high risk for cancer to surgery and patients with ITNs at low risk for cancer to observation. ThyroSeq v2 is marketed to have a positive predictive value (PPV) of 66% and a negative predictive value (NPV) of 97%.7 A recent multi-institutional analysis of ThyroSeq demonstrated variable test performance, with PPVs ranging from 25% to 85%.8 When considering the removal of noninvasive follicular thyroid neoplasms with papillarylike nuclear features (NIFTPs) from the malignant category, PPVs decreased across many institutions, ranging from 23% to 43%. Negative predictive values across institutions have remained high (88%-100%) and consistent with the ThyroSeq marketed value.8
Given the reported variability on rates of false-negative FNA biopsies in larger ITNs and the decreased PPV observed in ThyroSeq performance,8 our team sought to explore whether ThyroSeq results were reliable as ITN size increased. To our knowledge, no previous reports have analyzed ThyroSeq performance based on the size of the ITN. We set out to analyze the clinical, genomic, and pathologic characteristics of ITNs at our institution. Our primary aim was to assess the association between thyroid ITN size and the diagnostic accuracy of ThyroSeq testing. Our secondary aim was to analyze genetic signatures associated with the risk of malignant thyroid nodules.
Methods
All patients with a cytologic diagnosis of Bethesda III or IV status on thyroid FNA cytologic testing who underwent molecular testing using ThyroSeq v2 or ThyroSeq v3 (developed by the University of Pittsburgh Medical Center and Sonic Healthcare USA)7 from January 2015 to December 2018 at NYU Langone Health were reviewed. All patients at NYU Langone Health with Bethesda III or IV ITNs undergo routine molecular testing in their diagnostic workup. In this cohort, we included patients who underwent thyroid surgery with available results from pathologic testing of tissue samples obtained during surgery to have a complete data set. This study followed the reporting guideline for case series. Institutional review board approval for the study was obtained from the NYU Langone Health Institutional Review Board. The need for informed consent was waived by the board because the study was retrospective and examined data collected in the course of clinical care. The data were deidentified with no pertinent patient information.
Demographic characteristics (ie, age and sex; self-reported race and ethnicity data were not used in any of our analyses and were included only to further describe the demographics of the study population), clinical ITN size, Bethesda classification, ThyroSeq results, treatment detail, and final histopathologic diagnoses were tabulated. Sonographic characteristics were reviewed for all patients in the study cohort. Because our study period of 2015-2018 predated the adoption of the American College of Radiology Thyroid Imaging Reporting and Data System (TI-RADS) at NYU Langone Health, not all ITNs were scored using the TI-RADS classification system. Only ITNs diagnosed after 2017 had consistent reporting of TI-RADS classifications. Thus, to ensure consistency in our data reporting, we chose not to incorporate TI-RADS classification into our analysis. Genomic classification results were considered positive if alterations with more than a 30% risk of cancer were reported. Results with cancer risks less than 30% or no alterations identified were considered negative. Fine-needle aspiration–biopsied ITNs were matched with findings on surgical pathologic testing by correlating the laterality and location within the lobe and ITN size on preoperative ultrasonography and final pathologic test reports. Cases in which the ITN location and size did not match were excluded. A diagnosis of NIFTP was not considered to be malignant owing to its reclassification as a nonmalignant entity in 2016 but was considered clinically important.9,10,11 Incidental carcinomas separate from the biopsied ITN were excluded from the calculations of the sensitivity and specificity and the incidence of cancer.
Our primary objective was to analyze the diagnostic performance (sensitivity, specificity, PPV, and NPV) of ThyroSeq by ITN size (<2, 2-4, and >4 cm). ThyroSeq diagnostic performance was calculated with NIFTP considered alternatively as both a nonmalignant and clinically important entity. The secondary aim of the study was to analyze genetic signatures associated with the risk of cancer.
Statistical Analysis
Data analysis was conducted from March to April 2021. Categorical variables were compared by the Pearson χ2 test and Fisher exact test when appropriate. A 2-sided α level <.05 was used as the threshold of statistical significance. Analyses were conducted using R, version 3.5.1, statistical software.12
Results
During the study period, 908 nodules were diagnosed as Bethesda III or IV on FNA. From this group, 82 ITNs (9.0%) underwent genomic testing using Afirma Genomic Sequencing Classifier (Veracyte) were excluded. The remainder (826 [91.0%]) with ThyroSeq (v2 or v3) testing were further reviewed. Of these, 608 ITNs (73.6%) were observed and 218 ITNs (26.4%) underwent surgery. More than half (396 [65.1%]) of the observed ITNs were ThyroSeq-negative. The 218 ITNs that underwent surgery had a complete clinical and pathologic data set composed the study population. Patient demographics are described in Table 1. Patient management is presented in the Figure.
Table 1. Clinical, Genomic, and Pathologic Characteristics of 212 Patients With 218 Bethesda III and IV Thyroid Nodules.
| Characteristic | No. (%) |
|---|---|
| Age, median (SD), y | 49 (15.6) |
| Sex | |
| Female | 158 (74.5) |
| Male | 54 (25.5) |
| Race | |
| Black | 20 (9.4) |
| White | 153 (72.2) |
| Othera | 39 (18.4) |
| Nodule size, mean (SD), cm | 2.6 (1.5) |
| Nodule size categorization, cm | |
| <2 | 98 (45.0) |
| 2-4 | 87 (39.9) |
| >4 | 33 (15.1) |
| Bethesda classification | |
| III | 167 (76.6) |
| IV | 51 (23.4) |
| ThyroSeq version | |
| v2 | 196 (89.9) |
| v3 | 22 (10.1) |
| ThyroSeq resultb | |
| Positive | 173 (79.4) |
| Negative | 45 (20.6) |
| Final histologic diagnosis | |
| Adenomatoid nodule | 70 (32.1) |
| NIFTP | 61 (28.0) |
| Follicular variant PTC | 32 (14.7) |
| Hürthle cell adenoma | 15 (6.9) |
| Benign nodular hyperplasia | 15 (6.9) |
| Classical PTC | 11 (5.0) |
| Tall cell PTC | 6 (2.8) |
| Benign other | 5 (2.3) |
| Hürthle cell carcinoma | 3 (1.4) |
| Treatment | |
| Thyroid lobectomy | 142 (65.1) |
| Total thyroidectomy | 76 (34.9) |
Abbreviations: NIFTP, noninvasive follicular thyroid neoplasm with papillarylike nuclear features; PTC, papillary thyroid carcinoma.
Includes Asian, Hispanic, and other racial and ethnic groups.
Results were considered positive if alterations with more than a 30% risk of cancer were reported. Results with cancer risks less than 30% or no alterations identified were considered negative.
Figure. Schematic Representation Showing ThyroSeq Results, Treatment, Histologic Diagnosis, and Adjuvant Therapy in Patients With Cytologic Diagnosis of Bethesda III and IV Nodules, Stratified by Clinical Nodule Size.
CT indicates completion thyroidectomy; FNA, fine-needle aspiration; ITNs, indeterminate thyroid nodules; NIFTP, noninvasive follicular thyroid neoplasm with papillarylike nuclear features; RAI, radioactive iodine; TL, thyroid lobectomy; and TT, total thyroidectomy.
Of the 212 patients, the median (SD) age at diagnosis was 49 (15.6) years (range, 14-84 years); the population included 158 females (74.5%) and 54 males (25.5%). Twenty patients (9.4%) were Black, 153 (72.2%) were White, and 39 (18.4%) were reported as other (Asian, Hispanic, and other racial and ethnic groups). Most patients had a solitary ITN (206 [97.2%]); 6 patients (2.8%) had 2 ITNs each. Of the 218 ITNs, approximately three-quarters were categorized as Bethesda class III (167 [76.6%]). The mean (SD) ITN size was 2.6 (1.5) cm (range, 0.5-8.9 cm). Most ITNs underwent genomic testing using ThyroSeq v2 (196 [89.9%]) (Table 1). Among 173 positive nodules, most (155 [89.6%]) had single sequence variations identified on molecular testing. Sequence variations of the RAS gene family (HRAS, NRAS, and KRAS) were the most common singular variant identified (95 [61.3%]), followed by variants in the BRAF v600E gene (7 [4.5%]). A minority of the 173 positive nodules had dual alterations (17 [9.8%]); TERT + RAS variants (7 [41.2%]) were the most common in this group, followed by RAS + EIF1AX variants (3 [17.6%]). Only 1 nodule (0.6%) had 3 variants (BRAF v600E + TERT + PTEN) identified on ThyroSeq testing. Among the 45 negative nodules, 36 (80.0%) had no alterations identified on molecular testing. The remainder of negative nodules (9 [20.0%]) had sequence variations with reported cancer risks less than 30% (Table 2).
Table 2. Sequence Variation Profiles and Final Histologic Diagnoses of 218 Bethesda III and IV Nodules.
| Risk of cancer | Molecular alteration | Malignant, No. (%) | Nonmalignant, No. (%) | ||
|---|---|---|---|---|---|
| Alteration | No. | NIFTP | Benign | ||
| High-risk group | |||||
| 98%-100% | ALK | 1 | 1 (100) | 0 | 0 |
| NTRK3 fusion | 2 | 2 (100) | 0 | 0 | |
| TP53 + RAS | 1 | 1 (100) | 0 | 0 | |
| TERT + RAS | 7 | 5 (71.4) | 1 (14.3) | 1 (14.3) | |
| BRAF v600E + TERT | 1 | 1 (100) | 0 | 0 | |
| BRAF v600E + TERT + PTEN | 1 | 1 (100) | 0 | 0 | |
| BRAF-like group | |||||
| 95%-100% | BRAF v600E | 7 | 7 (100) | 0 | 0 |
| BRAF C1406 | 1 | 1 (100) | 0 | 0 | |
| TERT | 4 | 4 (100) | 0 | 0 | |
| PAX8/PPARG | 8 | 2 (25.0) | 4 (50.0) | 2 (25.0) | |
| BRAF K601E + EIF1AX | 1 | 1 (100) | |||
| CNA group | |||||
| 40%-80% | CNA | 7 | 1 (14.3) | 0 | 6 (85.7) |
| RAS-like group | |||||
| 30%-80% | RAS | 95 | 11 (11.6) | 41 (43.1) | 43 (45.3) |
| BRAF K601E | 4 | 2 (50.0) | 2 (50.0) | 0 | |
| THADA fusion | 7 | 1 (14.3) | 5 (71.4) | 1 (14.3) | |
| TP53 | 3 | 1 (33.3) | 0 | 2 (66.7) | |
| DICER1 | 3 | 0 | 1 (33.3) | 2 (66.7) | |
| PTEN | 1 | 0 | 0 | 1 (100) | |
| PIK3CA | 2 | 1 (50.0) | 1 (50.0) | 0 | |
| EIF1AX | 1 | 0 | 1 (100) | 0 | |
| RAS + EIF1AX | 3 | 0 | 1 (33.3) | 2 (66.7) | |
| RAS + PTEN | 2 | 1 (50.0) | 1 (50.0) | 0 | |
| Other non–high-risk alterations | |||||
| 30%-80% | TSHR | 2 | 0 | 0 | 2 (100) |
| MET overexpression | 7 | 2 (28.6) | 2 (28.6) | 3 (42.8) | |
| PTEN + TSHR | 2 | 0 | 0 | 2 (100) | |
| ThyroSeq-negativea | |||||
| <30% | EIF1AX | 4 | 0 | 0 | 4 (100) |
| TP53 | 2 | 0 | 0 | 2 (100) | |
| RET/PTC | 1 | 1 (100) | 0 | 0 | |
| 22q | 1 | 1 (100) | 0 | 0 | |
| PTEN | 1 | 0 | 0 | 1 (100) | |
| 3%-4% | No alterations | 36 | 4 (11.1) | 1 (2.8) | 31 (86.1) |
Abbreviations: CNA, copy number alterations; NIFTP, noninvasive follicular thyroid neoplasm with papillarylike nuclear features.
Results with cancer risks less than 30% or no alterations identified were considered negative.
Primary Treatment and Final Histologic Testing
Thyroid lobectomy was the most common surgical treatment (142 [65.1%]), followed by total thyroidectomy (76 [34.9%]). Twenty-six (18.3%) of the ITNs that underwent thyroid lobectomy had cancer identified on final surgical pathologic testing, and cancer was diagnosed on final pathologic testing in 26 ITNs (34.2%) that underwent total thyroidectomy (P = .008). Overall, the final diagnosis was benign in 105 nodules (48.2%), NIFTP in 61 nodules (27.9%), and malignant in 52 nodules (23.9%). Of the 105 benign nodules, 70 (66.7%) were interpreted as adenomatoid, followed by Hürthle cell adenoma (15 [14.2%]) and benign nodular hyperplasia (15 [14.2%]). Follicular variant of papillary thyroid carcinoma was the most common malignant diagnosis (32 [14.7%]), followed by classical papillary thyroid carcinoma (11 [5%]) (Table 1).
Prevalence of Cancer by Nodule Size
Rates of cancer differed significantly across the 3 size categorizations of less than 2 cm (24 [24.5%]), 2 to 4 cm (15 [17.2%]), and greater than 4 cm (13 [39.4%]) (P = .04). However, this association was primarily mediated by the slightly lower prevalence of cancer in nodules that measured 2 to 4 cm. Rates of cancer in nodules less than 2 cm compared with those larger than 4 cm did not differ significantly (24.5% vs 39.4%; P = .09). Follicular variant of papillary thyroid carcinoma was the most common malignant histologic diagnosis in each size category, with an increasing prevalence in larger nodules (<2 cm: 12 [50%], 2-4 cm: 10 [66.7%], and >4 cm: 10 [76.9%]). The prevalence of classical and tall cell variants of PTC decreased with increased nodule size (<2 cm: 11 [11.2%], 2-4 cm: 5 [5.7%], and >4 cm: 1 [3%]) (Table 3).
Table 3. Rates of Cancer in 218 Bethesda III and IV Nodules Stratified by Clinical Nodule Size.
| Size, cm | No. | Malignant, No. (%) | Nonmalignant, No. (%) | |||||
|---|---|---|---|---|---|---|---|---|
| Total malignant | Follicular variant PTC | Classical PTC | Tall cell PTC | Hürthle cell carcinoma | NIFTP | Benign | ||
| <2 | 98 | 24 (24.5) | 12 (12.2) | 7 (7.1) | 4 (4.1) | 1 (1.0) | 28 (28.6) | 46 (46.9) |
| 2-4 | 87 | 15 (17.2) | 10 (11.5) | 3 (3.4) | 2 (2.3) | 0 | 26 (29.9) | 46 (52.9) |
| >4 | 33 | 13 (39.4) | 10 (30.3) | 1 (3.0) | 0 | 2 (6.1) | 7 (21.2) | 13 (39.4) |
| Total | 218 | 52 (23.9) | 32 (14.7) | 11 (5.0) | 6 (2.8) | 3 (1.4) | 61 (28.0) | 105 (48.2) |
Abbreviations: NIFTP, noninvasive follicular thyroid neoplasm with papillarylike nuclear features; PTC, papillary thyroid carcinoma.
Molecular Analysis and Cancer Probabilities per Sequence Variation
One-quarter of positive ITNs (46 [26.6%]) had malignant findings on final surgical pathologic testing. In the 45 ITNs that underwent surgery despite a negative genomic classifier interpretation, 6 (13.3%) were malignant. RAS variants were the most common sequence variation identified in malignant nodules (11 [21.2%] of all malignant nodules). The remainder of RAS nodules were NIFTP (41 [43.1%]) or benign entities (43 [45.3%]). Other common variants identified in malignant nodules were BRAF v600E variants (7 [13.5%] of all malignant nodules), TERT + RAS variants (5 [9.6%] of all malignant nodules), and variants in the TERT promoter gene (4 [7.7%] of all malignant nodules) (Table 2).
ThyroSeq Performance Analysis
The overall sensitivity of ThyroSeq in identifying malignant nodules was 88%, and the specificity was 23%. The PPV was 27% and NPV was 87%. With NIFTP considered as a clinically significant finding, the PPV increased and the NPV decreased slightly (PPV: 61%; NPV: 84%). ThyroSeq performance was further assessed based on ITN size (<2, 2-4, and >4 cm). In ITNs less than 2 cm (98 [45.0%]), test sensitivity was 88% and specificity was 15%. In this group, PPV was 25% and NPV was 79%. In ITNs 2 to 4 cm (87 [39.9%]), sensitivity was 87% and specificity was 28%. In this group, PPV was 20% and NPV was 91%. In ITNs larger than 4 cm (33 [15.1%]), ThyroSeq sensitivity was 92% and specificity was 40%; PPV was 50% and NPV was 89%. With NIFTP considered a significant finding, PPV increased across ITN sizes (<2 cm: 58%; 2-4 cm: 60%; and >4 cm: 75%) and NPV remained consistent (<2 cm: 79%; 2-4 cm: 91%; and >4 cm: 78%). Specificity significantly improved in ITNs larger than 4 cm compared with those smaller than 2 cm (40% vs 15%; P = .01). There were no significant differences in rates of false-negative results across ITN size categorizations (<2 cm: 11%; 2-4 cm: 13%; and >4 cm: 12%; P = .95). These findings were true when the performance of the test was calculated in identifying both cancer and NIFTP. The test specificity remained significantly higher in ITNs larger than 4 cm compared with those smaller than 2 cm (54% vs 24%; P = .03) (Table 4).
Table 4. ThyroSeq Test Performance in 218 Bethesda III and IV Nodules Stratified by Clinical Nodule Size.
| Variablea | Nodules, No. (%) | Diagnostic performance, % | ||||
|---|---|---|---|---|---|---|
| Total | Malignant | NIFTP | Benign | Malignant | Malignant + NIFTP | |
| Clinical nodule size, cm | ||||||
| <2 | ||||||
| ThyroSeq-positive | 84 | 21 (25.0) | 28 (33.3) | 35 (41.7) |
|
|
| ThyroSeq-negative | 14 | 3 (21.4) | 0 | 11 (78.6) | ||
| 2-4 | ||||||
| ThyroSeq-positive | 65 | 13 (20.0) | 26 (40.0) | 26 (40.0) |
|
|
| ThyroSeq-negative | 22 | 2 (9.1) | 0 | 20 (90.9) | ||
| >4 | ||||||
| ThyroSeq-positive | 24 | 12 (50.0) | 6 (25.0) | 6 (25.0) |
|
|
| ThyroSeq-negative | 9 | 1 (11.1) | 1 (11.1) | 7 (77.8) | ||
| Overall institutional performance | ||||||
| ThyroSeq-positive | 173 | 46 (26.6) | 60 (34.7) | 67 (38.7) |
|
|
| ThyroSeq-negative | 45 | 6 (13.3) | 1 (2.2) | 38 (84.5) | ||
Abbreviations: NIFTP, noninvasive follicular thyroid neoplasm with papillarylike nuclear features; NPV, negative predictive value; PPV, positive predictive value; SENS, sensitivity; SPEC, specificity.
Results were considered positive if alterations with more than a 30% risk of cancer were reported. Results with cancer risks less than 30% or no alterations identified were considered negative.
Rates of Completion Thyroidectomy and Radioactive Iodine Therapy
A total of 142 ITNs (65.1%) from 140 patients underwent thyroid lobectomy, and 8 patients (5.7%) underwent completion thyroidectomy. Rates of completion thyroidectomy were analyzed in positive and negative ITNs across nodule size categorizations. Few patients with positive ITNs that were treated with thyroid lobectomy underwent completion thyroidectomy (<2 cm: 3 [5.4%]; 2-4 cm: 1 [2.2%]). Three patients (27.3%) with large (>4 cm) positive ITNs underwent completion thyroidectomy. The need for completion thyroidectomy in negative ITNs was rare; only 1 patient with a negative ITN larger than 4 cm underwent completion thyroidectomy (Figure).
In our sample, the overall need for radioactive iodine (RAI) therapy was low (19 [8.7%]). The need for RAI in patients with positive ITNs was lower in small ITNs (<2 cm: 5 [6.0%]; 2-4 cm: 4 [6.2%]). One-third (8 [33.3%]) of patients with positive ITNs larger than 4 cm received RAI. Patients with negative ITNs rarely received adjuvant RAI therapy (<2 cm: 1 [7.1%]; 2-4 cm: 1 [4.3%]). No patients with negative ITNs larger than 4 cm underwent adjuvant RAI therapy (Figure).
Characteristics of ITNs Larger Than 4 cm
Of the 33 ITNs larger than 4 cm, 24 (72.7%) were interpreted as positive on ThyroSeq testing. In this group, 16 patients (48.5%) underwent lobectomy (11 positive, 5 negative); 17 patients (51.5%) underwent total thyroidectomy (13 positive, 4 negative). Of all patients with negative results, only 1 patient, who was treated with lobectomy, harbored a malignant nodule. This patient underwent completion thyroidectomy and ultimately had benign findings in the contralateral lobe on surgical pathologic testing. Twelve of the 24 patients (50%) with positive results were diagnosed with malignant nodules (5 patients underwent lobectomy, 7 underwent total thyroidectomy). Of the 5 patients treated with lobectomy, 3 subsequently underwent completion thyroidectomy and 1 was diagnosed with a malignant nodule in the contralateral lobe.
Discussion
In this analysis of ThyroSeq performance, we observed high sensitivity and NPV across all ITN sizes. We found that PPV and specificity significantly improved in ITNs of larger size, specifically, those larger than 4 cm, without any corresponding decrease in NPV.
Although ThyroSeq is widely used in the diagnostic workup of ITNs, evidence-based literature to aid in the interpretation of these test results is scarce. Recent single-institutional and multi-institutional ThyroSeq performance analyses report variable and decreased real-world PPVs, which has led to uncertainty concerning the clinical utility of the molecular information provided by the genomic classifier. We noted marked improvements in the diagnostic performance of ThyroSeq in larger ITN sizes in this study. These data may aid in the diagnosis and management of large ITNs.
The PPV of ThyroSeq at our institution for identifying malignant lesions (27%) was lower than originally reported (81%) when NIFTPs were considered nonmalignant.7,13 Conversely, the NPVs remained relatively consistent with values from original reports (87% vs 96%).7,13 However, it is well established that variability exists in ThyroSeq PPV and NPV in different clinical settings; this variability is attributable to differing disease prevalence, which directly affects PPV and NPV.8 Sensitivity and specificity characterize a test independent of disease prevalence. Sensitivity was consistent with the values stated in marketing (88% vs 90%), but specificity was again found to be lower (23% vs 93%).7,13 Alternatively, with NIFTP considered a clinically significant finding, we observed a considerably higher PPV of 61%. This finding is not surprising because ThyroSeq was introduced into clinical care before the reclassification of NIFTP as a nonmalignant entity. Considering these changes, the PPV of ThyroSeq was more consistent with values from previously reported series.9,10 Because thyroid lobectomy is still necessary to render a diagnosis of NIFTP, including NIFTP in the calculations of PPV is helpful in this respect. Because the ultimate management of NIFTP remains a matter of further study, we present the data as they were identified. Whether all NIFTPs should undergo surgical excision or whether this entity can be safely observed is not clear. Prior reports noted a very low rate of progression to follicular neoplasm and a very low potential for distant metastases.14,15
ThyroSeq performance was further analyzed by stratifying ITN size. The PPV was low in ITNs smaller than 2 cm (25%) and in those 2 to 4 cm (20%). The PPVs improved (<2 cm: 58%; 2-4 cm: 60%) for these ITN sizes once NIFTPs were considered significant findings. In ITNs larger than 4 cm, there was substantial improvement in PPV (50%). This pattern was maintained when calculating the performance of ThyroSeq in identifying both malignancies and NIFTPs. The NPV was less variable (79%-91%) across ITN sizes and was consistent with values from prior institutional analyses and original reports.7,8,13 Sensitivity was high (87%-92%) across ITN sizes. ThyroSeq specificity was significantly higher in ITNs larger than 4 cm (40%) compared with those smaller than 2 cm (15%) owing to the slightly higher prevalence of cancer in ITNs larger than 4 cm. The observed improvement in PPV and specificity occurred without any loss in NPV or sensitivity. In addition, the rates of false-negative findings on ThyroSeq tests remained consistent across ITN sizes (11%-13%). Although data on rates of false-negative FNA biopsy vary,2,3 our findings are consistent with prior studies that reported a low rate of false-negative aspirates in larger ITNs.3,16 Although the utility of ThyroSeq as a rule-in test for cancer remains unclear, these data suggest that ThyroSeq may be used to rule out cancer in ITNs larger than 4 cm.
In our sample, singular RAS variants were the most common sequence variation identified. Nodules with singular RAS variants represented 21.2% of all malignant nodules. However, of all ITNs with isolated RAS variants, only 11.6% were ultimately malignant (the remainder were NIFTP [43.1%] or benign entities [45.3%]), which is lower than values from prior studies and estimated probabilities of cancer in ThyroSeq reports.7,8,13,17 Sequence variations at higher risk for cancer included BRAF v600E (100%), TERT (100%), and TERT + RAS (71.4%). Although the number of cases with high-risk variants in these series was small, our observed cancer risks were similar to those in earlier reports.8,18 Overall, we were unable to identify a specific genomic signature associated with cancer.
The observed improvement in ThyroSeq diagnostic performance in larger ITNs could potentially be attributed to a few different factors. First, the prevalence of cancer was highest in large (>4 cm) ITNs in this series. It is well established that higher disease prevalence leads to increased PPVs; therefore, this is the most likely explanation for the observed increase in test performance. Second, the improved diagnostic performance could be associated with increased operator performance in ultrasonography-guided FNA biopsy in larger ITNs, characterized by improved accuracy in targeting larger ITNs and retrieving adequate cellular samples in biopsy aspirates. Third, larger ITNs may have been present for a longer time and therefore had the opportunity to develop more molecular alterations in comparison with smaller ITNs. However, these hypotheses warrant further research. Use of ThyroSeq missed only 6 malignant nodules, and none were high risk; therefore, the risk of ThyroSeq failing to identify cancer in large ITNs appears to be low. Furthermore, the risk of completion thyroidectomy in ITNs larger than 4 cm that were initially managed with thyroid lobectomy was low. Our data suggest that negative interpretations on ThyroSeq testing in larger ITNs may be considered accurate.
Based on our data, further de-escalation of surgical treatment for patients diagnosed as having large negative ITNs could potentially be considered. The 2015 American Thyroid Association management guidelines recommend total thyroidectomy for ITNs larger than 4 cm to limit the need for completion surgery if malignant lesions are identified on final pathologic testing.6 However, in our sample, the need for completion thyroidectomy and RAI therapy was low in negative ITNs, and no patients with negative ITNs larger than 4 cm required RAI therapy. Our results suggest that patients diagnosed with large (>4 cm) negative ITNs who have a sonographically clean contralateral lobe may be safely treated with thyroid lobectomy. It might be prudent to suggest that patients diagnosed with large ITNs should undergo a careful review of cytologic test and imaging findings. The outcome data on the extent of surgery in these cases are limited, and whether thyroid lobectomy is the appropriate first step in treatment of ITNs larger than 4 cm and positive, but without any high-risk genomic alterations, is an issue that requires further research.
Limitations
This study has limitations, including the retrospective design and single-institutional setting. Different disease prevalence in other clinical settings may lead to variations in test performance. Furthermore, the involuntary inclusion of incidental microcarcinomas would inflate the PPV and specificity of the test. There is also the possibility for variation in the cytologic interpretation of aspirate and histologic interpretation of surgical specimens. In addition, ThyroSeq v2 was used in the testing of most ITNs in our cohort. Newer versions of ThyroSeq with expanded sequencing panels may lead to improvements in diagnostic accuracy, particularly in Hürthle cell lesions. Despite such limitations, our study allowed us to identify clinical parameters in which ThyroSeq could be used to guide the treatment of patients with ITNs and potentially limit the number of total thyroidectomies.
Conclusions
The overall institutional performance of ThyroSeq in cytologically ITNs is commonly lower than marketing reports. In this study, we noted improved ThyroSeq PPVs and specificity in larger-sized ITNs, specifically those larger than 4 cm. The NPVs and sensitivity remained high throughout each size categorization. These data suggest that ThyroSeq may be used to rule out cancer in ITNs larger than 4 cm. Therefore, in ITNs larger than 4 cm with a negative ThyroSeq interpretation or no high-risk sequence variations, initial management of lobectomy may be warranted. These results should be validated in a prospective controlled study.
References
- 1.Straccia P, Rossi ED, Bizzarro T, et al. A meta-analytic review of the Bethesda System for Reporting Thyroid Cytopathology: has the rate of malignancy in indeterminate lesions been underestimated? Cancer Cytopathol. 2015;123(12):713-722. doi: 10.1002/cncy.21605 [DOI] [PubMed] [Google Scholar]
- 2.McCoy KL, Jabbour N, Ogilvie JB, Ohori NP, Carty SE, Yim JH. The incidence of cancer and rate of false-negative cytology in thyroid nodules greater than or equal to 4 cm in size. Surgery. 2007;142(6):837-844. doi: 10.1016/j.surg.2007.08.012 [DOI] [PubMed] [Google Scholar]
- 3.Megwalu UC. Risk of malignancy in thyroid nodules 4 cm or larger. Endocrinol Metab (Seoul). 2017;32(1):77-82. doi: 10.3803/EnM.2017.32.1.77 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Durante C, Grani G, Lamartina L, Filetti S, Mandel SJ, Cooper DS. The diagnosis and management of thyroid nodules: a review. JAMA. 2018;319(9):914-924. doi: 10.1001/jama.2018.0898 [DOI] [PubMed] [Google Scholar]
- 5.Cibas ES, Ali SZ. The 2017 Bethesda System for Reporting Thyroid Cytopathology. Thyroid. 2017;27(11):1341-1346. doi: 10.1089/thy.2017.0500 [DOI] [PubMed] [Google Scholar]
- 6.Haugen BR, Alexander EK, Bible KC, et al. 2015 American Thyroid Association management guidelines for adult patients with thyroid nodules and differentiated thyroid cancer: the American Thyroid Association Guidelines Task Force on Thyroid Nodules and Differentiated Thyroid Cancer. Thyroid. 2016;26(1):1-133. doi: 10.1089/thy.2015.0020 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Steward DL, Carty SE, Sippel RS, et al. Performance of a multigene genomic classifier in thyroid nodules with indeterminate cytology: a prospective blinded multicenter study. JAMA Oncol. 2019;5(2):204-212. doi: 10.1001/jamaoncol.2018.4616 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Marcadis AR, Valderrabano P, Ho AS, et al. Interinstitutional variation in predictive value of the ThyroSeq v2 genomic classifier for cytologically indeterminate thyroid nodules. Surgery. 2019;165(1):17-24. doi: 10.1016/j.surg.2018.04.062 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Baloch ZW, Seethala RR, Faquin WC, et al. Noninvasive follicular thyroid neoplasm with papillary-like nuclear features (NIFTP): a changing paradigm in thyroid surgical pathology and implications for thyroid cytopathology. Cancer Cytopathol. 2016;124(9):616-620. doi: 10.1002/cncy.21744 [DOI] [PubMed] [Google Scholar]
- 10.Nikiforov YE, Seethala RR, Tallini G, et al. Nomenclature revision for encapsulated follicular variant of papillary thyroid carcinoma: a paradigm shift to reduce overtreatment of indolent tumors. JAMA Oncol. 2016;2(8):1023-1029. doi: 10.1001/jamaoncol.2016.0386 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Strickland KC, Howitt BE, Marqusee E, et al. The impact of noninvasive follicular variant of papillary thyroid carcinoma on rates of malignancy for fine-needle aspiration diagnostic categories. Thyroid. 2015;25(9):987-992. doi: 10.1089/thy.2014.0612 [DOI] [PubMed] [Google Scholar]
- 12.R Development Core Team 3.0.1. The R Project for Statistical Computing. 2013. Accessed April 1, 2020. https://www.R-project.org/
- 13.Nikiforov YE, Carty SE, Chiosea SI, et al. Impact of the multi-gene ThyroSeq next-generation sequencing assay on cancer diagnosis in thyroid nodules with atypia of undetermined significance/follicular lesion of undetermined significance cytology. Thyroid. 2015;25(11):1217-1223. doi: 10.1089/thy.2015.0305 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Chung R, Guan H, Ponchiardi C, et al. Noninvasive follicular thyroid neoplasm with papillary-like nuclear features: epidemiology and long-term outcomes in a strictly defined cohort. Thyroid. 2021;31(1):68-75. doi: 10.1089/thy.2019.0616 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Canini V, Leni D, Pincelli AI, et al. Clinical-pathological issues in thyroid pathology: study on the routine application of NIFTP diagnostic criteria. Sci Rep. 2019;9(1):13179. doi: 10.1038/s41598-019-49851-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Yoon JH, Kwak JY, Moon HJ, Kim MJ, Kim EK. The diagnostic accuracy of ultrasound-guided fine-needle aspiration biopsy and the sonographic differences between benign and malignant thyroid nodules 3 cm or larger. Thyroid. 2011;21(9):993-1000. doi: 10.1089/thy.2010.0458 [DOI] [PubMed] [Google Scholar]
- 17.Marti JL, Avadhani V, Donatelli LA, et al. Wide inter-institutional variation in performance of a molecular classifier for indeterminate thyroid nodules. Ann Surg Oncol. 2015;22(12):3996-4001. doi: 10.1245/s10434-015-4486-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Xing M. Genetic-guided risk assessment and management of thyroid cancer. Endocrinol Metab Clin North Am. 2019;48(1):109-124. doi: 10.1016/j.ecl.2018.11.007 [DOI] [PMC free article] [PubMed] [Google Scholar]
