Performance of Afirma Genomic Sequencing Classifier in Binary Subcategories of Atypia of Undetermined Significance Thyroid Nodules: Single Versus Repeat Diagnosis

Abstract

Background:

Afirma Genomic Sequencing Classifier (GSC) testing has been utilized for further risk stratification of thyroid nodules categorized as atypia of undetermined significance (AUS). The 2023 Bethesda system subcategorizes AUS diagnosis into AUS with nuclear atypia (AUS-N) and other atypia (AUS-O). The current study aims to determine if performance of GSC testing differs between the two AUS subcategories and between single AUS cohort and repeat AUS cohort.

Methods:

This retrospective study analyzed consecutive thyroid nodule fine-needle aspiration with a single or a repeat AUS diagnosis and a diagnostic GSC testing result (benign vs. suspicious). All AUS nodules were divided into AUS-N or AUS-O subcategory and followed by either surgical intervention or at least 12 months of clinical and/or ultrasound monitoring. We then assessed performance of GSC testing in each subcategory and subsequently compared the individual performance in AUS-N or AUS-O subcategory between single AUS cohort and repeat AUS cohort.

Results:

The study identified a total of 365 thyroid nodules subcategorized as AUS-N (N = 106) and AUS-O (N = 259). Both cohorts showed a significantly lower GSC benign call rate (BCR) in AUS-N nodules compared with AUS-O nodules (43% vs. 71% in single AUS, p = 0.001; 58% vs. 74% in repeat AUS, p = 0.02). The proportion of histology-proven malignancies associated with a suspicious GSC result tended to be greater in AUS-N nodules than AUS-O nodules (28% vs. 10% in single AUS, p = 0.09; 38% vs. 27% in repeat AUS, p = 0.3). Compared with AUS-N nodules, AUS-O cohorts demonstrated significantly higher specificity in the single AUS group (73% vs. 51%, p = 0.01). In both subcategories, the repeat AUS cohort yielded greater specificity, positive predictive value, and diagnostic accuracy compared with the single AUS group. However, the differences did not reach statistical significance.

Conclusions:

GSC BCR and diagnostic performance of GSC testing may vary in AUS-N versus AUS-O subcategories. However, there were no statistically significant differences in GSC performance between single and repeat AUS cohorts.

Introduction

Since the first publication of The Bethesda System for Thyroid Cytopathology (TBSRTC),¹ it has been widely implemented in the cytological assessment of thyroid nodules. Following the second edition in 2018,² the 2023 TBSRTC (the third edition) was recently published.³ The three editions of TBSRTC^1
–3 consist of six diagnostic categories, and the newest TBSRTC³ enlists one term for each of the six diagnostic categories. These categories include (I) nondiagnostic, (II) benign, (III) atypia of undetermined significance (AUS), (IV) follicular neoplasm (FN), (V) suspicious for malignancy, and (VI) malignant. Among the three intermediate categories (III, IV, and V), AUS is the most heterogeneous category. The 2023 TBSRTC further divides AUS into two subcategories as follows: AUS with nuclear atypia (AUS-N) and AUS with other atypia (AUS-O).³ Subcategorizing AUS into AUS-N and AUS-O is supported by the findings that the risk of malignancy (ROM) differs among the different subcategories of observed atypia. Two studies have demonstrated higher ROM in AUS nodules subcategorized as AUS-N compared with AUS nodules subcategorized as AUS-O.^4,5

For AUS cytology, the American Thyroid Association (ATA) management guidelines suggest “repeat fine-needle aspiration (FNA) and/or molecular testing, diagnostic surgery, or surveillance.”⁶ Prior studies reported that repeat FNA of thyroid nodules initially categorized as AUS yielded a more definitive diagnostic categorization in 45–50% of cases, of which up to 45% were reclassified as benign nodules.^7,8 Molecular testing incorporated with FNA cytological assessment has been proven to be an effective modality in further risk stratification of AUS nodules and estimation of ROM, thereby optimizing the management of AUS nodules.^9,10 The Afirma Genomic Sequencing Classifier (GSC) was introduced in 2017, and compared with its predecessor (Gene Expression Classifier [GEC]), the GSC has demonstrated improved specificity and positive predictive value (PPV), high sensitivity, and negative predictive value (NPV).^11,12

At our institution, GSC testing is generally performed reflexively if the second FNA has a repeat AUS diagnosis; however, endocrinologists and endocrine surgeons may make individually tailored decisions. Previously published data from our institution demonstrated that when comparing between patient cohorts with a single AUS diagnosis and a repeat AUS diagnosis, the overall diagnostic performance of GSC was not statistically significant.¹³ The lack of difference in GSC testing performance was believed to be partially explained by selection bias; however, GSC testing performance in subcategories of AUS-N and AUS-O in single AUS compared with repeat AUS had not been assessed.

In this current retrospective cohort review, we used the 2023 TBSRTC³ as a guideline and further divided all AUS nodules from both single AUS and repeat AUS cohorts into either AUS-N or AUS-O subcategory. We then assessed performance of GSC testing in each subcategory and subsequently compared the individual performance in AUS-N or AUS-O subcategory between single AUS cohort and repeat AUS cohort.

Material and Methods

This retrospective cohort review was approved by the Institutional Review Board (study approval number: HUM00252213) at the University of Michigan in Ann Arbor, Michigan. A SNOMED search of the electronic pathology database in our institution for the period from July 2017 to June 2022 was conducted to retrieve cases in which thyroid nodules carried either a single AUS diagnosis or a repeat AUS diagnosis; both had concurrent diagnostic GSC results (benign vs. suspicious). There was no overlap between single and repeat AUS cohorts. All nodules were followed by either surgical intervention or at least 12 months of clinical and/or ultrasound monitoring. Nodules with a nondiagnostic GSC testing result (due to inadequate sampling) and nodules without surgical follow-up or at least 12 months of clinical and/or ultrasound monitoring were excluded from the study. In this regard, 12 cases were excluded from the study due to a nondiagnostic GSC testing result. Among which, seven had a single AUS diagnosis (four AUS-N and three AUS-O) and five had a repeat AUS diagnosis (one AUS-N and four AUS-O). Furthermore, 90 cases with a diagnostic GSC result were also excluded from the study due to the lack of either surgical follow-up or at least 12 months of clinical/ultrasound follow-up, including 46 with a single AUS diagnosis and 44 with a repeat AUS diagnosis.

Ultrasound-guided thyroid FNAs were performed by endocrinologists, radiologists, and surgeons with cytology-assisted rapid on-site adequacy evaluation yielding a combination of conventional smears, ThinPrep slides, and/or cell blocks. Two dedicated passes from each nodule were also collected into an Afirma-provided fixative vial during the procedure. FNA specimens were then assessed by subspecialty board-certified cytopathologists who have minimum of 3–10 years experiences in diagnostic thyroid cytopathology. The pathologists were aware of patients’ previous FNA results if any. Cytological diagnoses were reported using the TBSRTC system.³ When a single or a repeat diagnosis of AUS was rendered, the aforementioned precollected samples were sent to Veracyte CLIA laboratory (South San Francisco, CA) for Afirma GSC testing. For the purpose of this study, AUS nodules with cytological atypia with or without architectural atypia were grouped into the AUS-N subcategory, while the rest were grouped into the AUS-O subcategory defined by the 2023 TBSRTC.³

The following information from individual patients included in this study was collected from the electronic medical record: age, sex, size of thyroid nodule, and corresponding histological diagnosis if available. The electronic medical record was evaluated to determine the stability of the nonsurgically resected nodules via clinical and/or ultrasound monitoring during a period of at least 12 months after the FNA. Nodules that did not develop new high-risk features and showed no marked change in size were considered clinically stable.¹⁴

The following diagnostic parameters of GSC testing were calculated for each subcategory (AUS-N and AUS-O) in each cohort (single AUS vs. repeat AUS) as follows:

BCR = number of nodules with benign GSC result/total number of nodules with GSC testing

Sensitivity = number of nodules with GSC suspicious result and histology-proven malignancy (true positive)/number of all histology-proven malignant nodules (true positive + false negative)

Specificity = number of nodules with GSC benign result and a subsequent surgical and/or clinical benign diagnosis (true negative)/numbers of all benign nodules (true negative + false positive)

PPV = true positive/all nodules with GSC suspicious result (true positive + false positive)

NPV = true negative/all nodules with GSC benign result (true negative + false negative)

Diagnostic accuracy = (true positive + true negative)/total number of nodules

The above parameters were compared between two subcategories (AUS-N vs. AUS-O) and between two cohorts (single AUS vs. repeat AUS) using Social Science Statistics (https://www.socscistatistics.com/tests/). Pearson’s chi-square for categorical variables and Student’s t-test for continuous variables were performed. Statistical significance was defined as a two-tailed p-value of < 0.05 for all analyses.

Results

Study cohort

The study included a total of 365 thyroid nodules which fulfilled inclusion criteria, including 145 and 220 nodules in the single AUS and repeat AUS cohorts, respectively. In the single AUS cohort, 44 and 101 nodules were subcategorized as AUS-N and AUS-O, respectively. The repeat AUS cohort consisted of 62 and 158 nodules, which were subcategorized as AUS-N and AUS-O, respectively (Fig. 1). Female predominance (≥68%) was seen in all subcategories. Nodules measuring ≤4 cm represented the majority in each subcategory of the single AUS cohort (93% AUS-N and 88% AUS-O) and repeat AUS cohort (95% AUS-N and 94% AUS-O). There is no significant difference in nodule size among different groups (p > 0.05) (Table 1).

FIG. 1.

Participant flow diagram. AUS, atypia of undetermined significance; GSC, Genomic Sequencing Classifier; AUS-N, AUS with nuclear atypia; AUS-O, AUS with other atypia.

Table 1.

Clinical Characteristics of the Study Cohort

	Single AUS (145)		Repeat AUS (220)		Total
	AUS-N	AUS-O	AUS-N	AUS-O	Total
Number of nodules	44 (30%)	101 (70%)	62 (28%)	158 (72%)	365
Sex
F	33 (75%)	81 (80%)	44 (71%)	108 (68%)	266 (73%)
M	11 (25%)	20 (20%)	18 (29%)	50 (32%)	99 (27%)
Patient age(y/o)	54 (31–80)	55 (22–91)	57 (26–80)	60 (22–82)
Nodule size(cm)
≤2	19	40	25	78	162 (44%)
2–4	22	49	34	70	175 (48%)
>4	3	12	3	10	28 (8%)
Mean (range)	2.3 (0.8–6.3)	2.6 (1.1–5.5)	2.3 (1.0–6.3)	2.2 (1.0–8.8)

F, female; M, male; AUS, atypia of undetermined significance; AUS-N, AUS with nuclear atypia; AUS-O, AUS with other atypia.

BCR and follow-up of GSC-benign nodules

BCR of GSC testing for AUS-O subcategory was significantly higher compared with that of AUS-N subcategory in both the single AUS cohort (72/101 = 71% vs. 19/44 = 43%, p = 0.001) and repeat AUS cohort (117/158 = 74% vs. 36/62 = 58%, p = 0.02). However, BCR of GSC testing for AUS-O and AUS-N did not differ significantly between the single AUS and repeat AUS cohorts (Table 2).

Table 2.

Genomic Sequencing Classifier Results in Atypia of Undetermined Significance Thyroid Nodules

	Single AUS (145)		Repeat AUS (220)
	AUS-N	AUS-O	AUS-N	AUS-O
GSC Suspicious	25 (57%)	29 (29%)	26 (42%)	41 (26%)
GSC benign	19 (43%)	72 (71%)	36 (58%)	117 (74%)

GSC, Genomic Sequencing Classifier.

Patients with the GSC-benign nodules are significantly older than those with GSC-suspicious nodule, with mean ages of 59 versus 49 years in the single AUS cohort (p < 0.001) and 62 versus 52 years in the repeat AUS cohort (p = 0.03). However, there is no significant age difference between the single AUS and repeat AUS cohorts regardless of GSC testing results.

The majority of GSC-benign thyroid nodules in both cohorts were managed with clinical and/or ultrasound monitoring for an average of 21 months (12–63 months) and were deemed clinically stable. Of the surgically treated GSC-benign nodules in each cohort, histopathological examination revealed non-neoplastic changes and follicular/oncocytic adenoma in almost all but two nodules. One was categorized as AUS-O in single AUS cohort, and the subsequent lobectomy specimen revealed multifocal well-differentiated papillary thyroid carcinoma (PTC). Another was categorized as AUS-N in repeat AUS cohort, and the subsequent surgical specimen revealed follicular carcinoma with capsular invasion (Table 3).

Table 3.

Clinical Follow-Up and Histopathological Diagnosis of Genomic Sequencing Classifier Benign Nodules

	Single AUS		Repeat AUS
	AUS-N	AUS-O	AUS-N	AUS-O	Total
GSC benign	19	72	36	117	244
Clinically stable	14 (74%)	49 (68%)	30 (83%)	97 (83%)	190 (78%)
Surgically treated	5 (26%)	23 (32%)	6 (17%)	20 (17%)	54 (22%)
Histological result	5	23	6	20
Benign histopathology	5 (100%)	22 (96%)	5 (83%)	20 (100%)	52 (96%)
Follicular nodular disease	2	14	3	17
Hashimoto thyroiditis		5		2
Follicular adenoma	3	2	2
Oncocytic adenoma		1		1
Malignant histopathology	0 (0%)	1 (4%)	1 (17%)	0 (0%)	2 (4%)
Follicular carcinoma			1
PTC		1

PTC, papillary thyroid carcinoma.

Follow-up of GSC-suspicious nodules

A suspicious result in GSC testing was obtained in 54 of 145 (37%) thyroid nodules in the single AUS cohort (25 AUS-N and 29 AUS-O) and 67 out of 220 (30%) thyroid nodules in repeat AUS cohort (26 AUS-N and 41 AUS-O). Ninety-six percent (116 of 121) of the GSC-suspicious nodules underwent surgical interventions. Among these surgically treated GSC-suspicious nodules, 42 of 52 (81%) nodules in the single AUS cohort (17 AUS-N and 25 AUS-O nodules) and 43 out of 67 (64%) nodules in the repeat AUS cohort (14 AUS-N and 29 AUS-O nodules) proved to be benign on histological evaluation. The most common benign categorization was follicular nodular diseases followed by follicular/oncocytic adenoma and lymphocytic thyroiditis. Notably, three noninvasive follicular thyroid neoplasm with papillary-like nuclear features (NIFTP) cases were documented in the repeat AUS cohort (one AUS-N and two AUS-O nodules). It is worth mentioning that NIFTP was classified as a low-risk, borderline neoplasm by the 2022 World Health Organization Classification of Thyroid Neoplasms.¹⁵ Accordingly, NIFTP is morphologically and clinically intermediate between benign and malignant tumors, with a favorable prognosis.

Histopathology-proven malignancies were documented in 10 out of 54 (19%) nodules in the single AUS cohort (7 AUS-N and 3 AUS-O nodules) and 21 out of 67 (31%) nodules in repeat AUS cohort (10 AUS-N and 11 AUS-O nodules). Among the single AUS nodules subcategorized as AUS-N, five PTCs, one medullary thyroid carcinoma (MTC), and one extranodal marginal zone lymphoma of mucosa-associated lymphoid tissue (MALT) were identified, while three conventional PTCs were identified in single AUS-O nodules. It should be noted that in both nodules with a suspicious GSC testing result, in which MALT lymphoma and MTC were diagnosed by surgical pathology, the MTC, BRAF, RET/PTC1, or RET/PTC2 malignancy classifiers were not detected before surgery. Of the 10 cases in the repeat AUS cohort subcategorized as AUS-N nodules, 8 were PTCs (6 conventional and 2 follicular type), 1 was oncocytic carcinoma, and 1 was NIFTP. Similarly, in the 11 cases in the repeat AUS cohort subcategorized as AUS-O nodules, 7 were PTCs (including 3 conventional and 4 follicular type), 2 were follicular/oncocytic carcinomas, and 2 NIFTPs. Two patients in the single AUS cohort and three patients in the repeat AUS cohort declined surgical treatment. During an average of 29 months of clinical monitoring, their nodules remained stable and, thus, were presumed to be benign (Table 4).

Table 4.

Clinical Follow-Up and Histopathological Diagnosis of Genomic Sequencing Classifier Suspicious Nodules

	Single AUS		Repeat AUS
	AUS-N	AUS-O	AUS-N	AUS-O	Total
GSC suspicious	25	29	26	41	121
Clinically stable	1 (4%)	1 (3%)	2 (8%)	1 (2%)	5 (4%)
Surgically treated	24 (96%)	28 (97%)	24 (92%)	40 (98%)	116 (96%)
Histological result	24	28	24	40
Benign histopathology	17 (71%)	25 (90%)	14 (58%)	29 (73%)	85 (73%)
Follicular nodular disease	12	14	8	16
Hashimoto thyroiditis	2	1	1	1
Follicular adenoma	3	9	4	9
Oncocytic adenoma		1	1	3
Malignant histopathology	7 (29%)	3 (10%)	10 (42%)	11 (27%)	31 (27%)
PTC	5	3	6	3
PTC (follicular variant)			2	4
Follicular carcinoma				1
Oncocytic carcinoma			1	1
Medullary thyroid carcinoma	1
Lymphoma	1
Low-risk neoplasm NIFTP^a			1	2

NIFTP was classified as a low-risk neoplasm given its favorable prognosis.

NIFTP, noninvasive follicular thyroid neoplasm with papillary-like nuclear features.

When there was a suspicious GSC result, there was a nonstatistically significant trend toward higher proportion of histology-proven malignancies associated with AUS-N compared with AUS-O nodules in both single AUS (28% vs. 10%, p = 0.09) and repeat AUS cohorts (38% vs. 27%, p = 0.3). In addition, slightly higher malignant rates were observed in the repeat AUS compared with single AUS nodules in both AUS-N (38% vs. 28%, p = 0.42) and AUS-O categories (27% vs. 10%, p = 0.09). However, these observed differences did not reach statistical significance (p > 0.05).

Diagnostic performance of GSC in the single AUS cohort and the repeat AUS cohort

Table 5 compares the diagnostic performance of GSC testing between the single AUS cohort and repeat AUS cohort, subcategorized into AUS-N and AUS-O nodules with NIFTP included in the malignant category. Compared with AUS-O nodules, AUS-N cohorts demonstrated lower specificity in both AUS cohorts. However, this difference was significant in the single AUS group (51% vs. 73%, p = 0.01), but not in the repeat AUS nodule cohort (69% vs. 80%, p = 0.09). AUS-N cohorts exhibited a nonstatistically significant trend toward higher PPV in both single AUS (28% vs. 10%, p = 0.09) and repeat AUS nodules (38% vs. 27%, p = 0.3). In both AUS-N and AUS-O subcategories, GSC testing appeared to show greater specificity, PPV, and diagnostic accuracy in the repeat AUS cohort compared with the single AUS cohort. Nevertheless, these differences did not reach statistical significance.

Table 5.

Diagnostic Performance of Afirma Genomic Sequencing Classifier

	Single AUS		Repeat AUS
	AUS-N	AUS-O	AUS-N	AUS-O
Sensitivity (CI)	100% (1, 1)	75% (0.32, 1.17)	91% (0.74, 1.08)	100% (1, 1)
Specificity (CI)	51%* (0.35, 0.67)	73%* (0.64, 0.82)	69% (0.56, 0.81)	80% (0.73, 0.86)
PPV (CI)	28% (0.10, 0.45)	10% (−0.01, 0.21)	38% (0.20, 0.57)	27% (0.13, 0.40)
NPV (CI)	100% (1, 1)	99% (0.95, 1.01)	97% (0.92, 1.03)	100% (1, 1)
Diagnostic accuracy	59%	73%	73%	81%

True positive: Histology classified as malignant on surgical pathology; True negative: Benign on histology and in clinical follow-up.

p = 0.01.

PPV, positive predictive value; NPV, negative predictive value; CI, confidence interval.

Discussion

The previously published study of Afirma GSC performance from our institution consisted of 67 nodules with single AUS diagnosis and 135 nodules with repeat AUS diagnosis, respectively. None of the AUS nodules was further subcategorized, and a comparable performance of GSC testing was observed between single and repeat AUS cohorts.¹³ To further simplify subclassification while reflecting clinical risk and subsequent management of thyroid nodules diagnosed as AUS, the newly published 2023 TBSRTC has now divided AUS category (category III) into two broad subcategories as follows: AUS-N and AUS-O. Based on the newest recommendation, we subclassified the thyroid nodules which had a single or repeat AUS diagnosis accompanied by a diagnostic Afirma GSC testing result into AUS-N or AUS-O subcategory, aiming to compare GSC testing performance among four subcategories based on single versus repeat AUS diagnosis and binary AUS subcategorization. The current study had markedly increased number of cases in both the single AUS cohort (n = 145) and the repeat AUS cohort (n = 220) and revealed that AUS-N nodules were more likely to be GSC suspicious compared with AUS-O nodules. Compared with the single AUS cohort, the repeat AUS cohort appeared to show an improvement of some diagnostic parameters, namely, specificity, PPV, and diagnostic accuracy, in both AUS-N and AUS-O subcategories. However, these differences did not reach statistical significance.

The most recently published study from another institution demonstrated a significant difference in BCR between AUS-N (56%) and AUS-O (74%) subcategories among the thyroid nodules with a repeat AUS diagnosis.¹⁶ Similarly, our study showed that BCR was significantly lower in AUS-N subcategory compared with AUS-O subcategory in both the single AUS cohort (43% vs. 71%) and repeat AUS cohort (58% vs. 74%). These results are in keeping with our prior study focusing on Afirma GSC performance in thyroid nodules with a repeat AUS diagnosis (which has a dataset of 4-year overlap with this study from 2017 to 2021), which demonstrated a significantly greater PPV and lower BCR in thyroid nodule aspirates displaying combined nuclear and architectural atypia compared with thyroid nodule aspirates with merely architectural atypia.¹⁷

Benefits and advantages of repeat FNA in the management of thyroid nodules with a single AUS diagnosis have been well documented.^7,8,18,19 In addition, commercially available molecular testing has been implemented as an adjunct to FNA for further risk stratification of AUS nodules. The majority of previously published studies investigated GSC performance in thyroid nodules with a repeat AUS and/or FN/suspicious for follicular neoplasm (SFN) diagnosis.^20

–24 There are only a few published studies that have focused on the evaluation of GSC performance in thyroid nodules with a single AUS, and/or FN/SFN diagnosis only, or a mixture of nodules with both a single and a repeat AUS and/or FN/SFN diagnosis.^25,26 Nevertheless, in these studies, criteria for case selection (single AUS vs. repeat AUS) for GSC testing were not provided in detail. At our institution, endocrinologists/endocrine surgeons often apply GSC testing mainly in nodules with a repeat AUS diagnosis. However, the sample may be collected for GSC testing during a single FNA procedure at the endocrinologist’s discretion and patient-related variables. It is presumed that nodules with suspicious clinical presentations and/or ultrasound findings have a higher pretest probability of malignancy, which may impact the interpretation of GSC test performance between these two cohorts. In this regard, our previously published study demonstrated a comparable performance of GSC testing in thyroid nodules with single AUS diagnosis versus repeat AUS diagnosis.¹³ A separately published study on GEC testing conducted at another institution demonstrated an identical malignant risk associated with GEC-suspicious result for thyroid nodules with a single versus a repeat AUS diagnosis.²⁷ However, our current study contributes to the limited available data comparing GSC performance in AUS-N and AUS-O subcategories for single versus repeat AUS cohorts. Despite the trend in higher percentage GSC-suspicious single AUS nodules being benign (81%), compared with 72% of repeat AUS GSC-suspicious nodules, our findings demonstrate that repeat FNA on first-time AUS thyroid nodules does not significantly improve the diagnostic performance of GSC testing in either the AUS-N or AUS-O category.

Regarding the surgically resected GSC-suspicious thyroid nodules in the current study, a similar proportion of conventional PTC was identified in the single AUS cohort (11%) and repeat AUS cohort (10%). However, PTC occurred more frequently in the AUS-N subcategory of both single AUS (17% AUS-N vs. 8% AUS-O) and repeat AUS (20% AUS-N vs. 5% AUS-O) cohorts. These findings are similar to prior studies, which demonstrated that nodules classified as AUS with cytological atypia exhibited a higher malignant risk than those with architectural atypia alone.^28,29 Conversely, thyroid nodules with architectural atypia only were more likely to be benign.^4,28,29

Our study also showed that neoplasms with a follicular growth pattern represented more than one-third of the surgically removed nodules subcategorized as AUS-O in the single AUS (36%) and repeat AUS (35%) cohorts. The most common of these neoplasms were follicular adenoma followed by follicular type PTC, NIFTP, and follicular carcinoma. It has been known that the characteristic nuclear features of PTC are often focal and patchy in follicular type of PTC,³⁰ which may result in diagnostic challenges in separating follicular type PTC from NIFTP and other follicular neoplasms. As such, all these entities may be interpreted preoperatively as AUS-O or follicular neoplasm. It should be noted that, in comparison with reported studies, the current study showed relatively lower PPV and ROM in both single AUS and repeat AUS cohorts. This could be partially explained by the finding that our study had no (in single AUS cohort) or less proportion (3/22 = 13% in repeat AUS cohort) of surgically removed nodules diagnosed as NIFTP. On the contrary, most of the previous studies from other institutions categorized NIFTP into malignant (positive) category, and NIFTP represented notable proportion of malignant nodules, ranging from 22% to 60% in their studies.^{20,25,31
–33}

There are several limitations of this study. The first limitation is that this study was a retrospective cohort review, and as such, we did not have the ability to randomize subjects nor blind cytopathologists to prior FNA evaluations. A second limitation, similar to our previous study,¹³ is the potential selection bias in the single AUS cohort in the current study as a subset of these patients may have more clinically or radiologically worrisome features. A third potential limitation of this study is the relatively limited number of cases in each subcategory (AUS-N, AUS-O) composed of each cohort (single AUS, repeat AUS) in the current study. Therefore, cases with AUS-N or AUS-O with GSC suspicious results may have a reported differing ROM in comparison with the currently accepted 50% (and further, the < 4% ROM if GSC is benign). Given our single retrospective cohort review and relatively limited number of subjects in each subcategory (AUS-N, AUS-O), it is difficult to offer a differing or modified treatment recommendation than the current ATA guidelines for management and/or surveillance.⁴ A fourth limitation is that the reported test performance (sensitivity, specificity, PPV, NPV) is only based on the definition for true negative by histopathology for cases who underwent surgical resection, which is the recognized gold standard, as not every case underwent surgical resection. However, the cases that did indeed undergo surgical resection would meet these gold-standard criteria, and in a clinical, real-world retrospective study, it is appropriate to not have patients undergo surgery for benign results on cytopathology or GSC.

In conclusion, our study demonstrated that AUS-N nodules were more likely to be GSC suspicious, and histology-proven malignancies linked to a suspicious GSC result also tended to be higher in AUS-N compared with AUS-O nodules. However, these variations do not appear to be related to whether the diagnosis is a single or repeat AUS, particularly when nodules with single AUS diagnosis present with worrisome clinical and/or imaging characteristics.

Footnotes

Acknowledgment

The authors thank Dr. Wei Hao for taking time to assist them with data analysis during the revision.

Authors’ Contributions

X.J.: Acquired/analyzed/interpreted data and resource materials, drafted the article and contributed significant revisions on subsequent drafts, contributed to final approval of version to be published, and agrees to be accountable for all aspects of the work in ensuring questions related to accuracy or integrity of any part of the work. D.T.B.: Contributed significant revisions for intellectual content, contributed to final approval of version to be published, and agrees to be accountable for all aspects of the work in ensuring questions related to accuracy or integrity of any part of the work. M.L.: Contributed revisions for intellectual content, contributed to final approval of version to be published, and agrees to be accountable for all aspects of the work in ensuring questions related to accuracy or integrity of any part of the work. A.H.: Contributed revisions for intellectual content, contributed to final approval of version to be published, and agrees to be accountable for all aspects of the work in ensuring questions related to accuracy or integrity of any part of the work. M.R.H.: Contributed revisions for intellectual content, contributed to final approval of version to be published, and agrees to be accountable for all aspects of the work in ensuring questions related to accuracy or integrity of any part of the work. M.P.: Contributed revisions for intellectual content, contributed to final approval of version to be published, and agrees to be accountable for all aspects of the work in ensuring questions related to accuracy or integrity of any part of the work. D.C.: Contributed revisions for intellectual content, contributed to final approval of version to be published, and agrees to be accountable for all aspects of the work in ensuring questions related to accuracy or integrity of any part of the work. J.J.I.: Contributed revisions for intellectual content, contributed to final approval of version to be published, and agrees to be accountable for all aspects of the work in ensuring questions related to accuracy or integrity of any part of the work. N.E.: Contributed revisions for intellectual content, contributed to final approval of version to be published, and agrees to be accountable for all aspects of the work in ensuring questions related to accuracy or integrity of any part of the work. Z.S.: Contributed revisions for intellectual content, contributed to final approval of version to be published, and agrees to be accountable for all aspects of the work in ensuring questions related to accuracy or integrity of any part of the work. L.D.: Contributed revisions for intellectual content, contributed to final approval of version to be published, and agrees to be accountable for all aspects of the work in ensuring questions related to accuracy or integrity of any part of the work. D.T.H.: Contributed revisions for intellectual content, contributed to final approval of version to be published, and agrees to be accountable for all aspects of the work in ensuring questions related to accuracy or integrity of any part of the work. B.S.: Contributed revisions for intellectual content, contributed to final approval of version to be published, and agrees to be accountable for all aspects of the work in ensuring questions related to accuracy or integrity of any part of the work. X.J.: Designed concept of article, acquired/analyzed/interpreted data and resource materials, drafted the article and contributed significant revisions on subsequent drafts, contributed to final approval of version to be published, and agrees to be accountable for all aspects of the work in ensuring questions related to accuracy or integrity of any part of the work.

Author Disclosure Statement

X.J. speaking at Diaceutics Webinar on June 12, 2024. None for other authors.

Funding Information

No funding was received for this article.

References

Ali

, Cibas

. The Bethesda System for Reporting Thyroid Cytopathology: Definitions, Criteria, and Explanatory Notes. New York, NY: Springer; 2009.

Ali

, Cibas

. The Bethesda System for Reporting Thyroid Cytopathology: Definitions, Criteria, and Explanatory Notes. Second edition. New York, NY: Springer; 2018.

Ali

, Baloch

, Cochand-Priollet

, et al. The 2023 Bethesda system for reporting thyroid cytopathology. Thyroid, 2023; 33(9):1039–1044; doi: 10.1089/thy.2023.0141

Olson

, Clark

, Erozan

, et al. Spectrum of risk of malignancy in subcategories of ‘atypia of undetermined significance’. Acta Cytol, 2011; 55(6):518–525; doi: 10.1159/000333232

VanderLaan

, Marqusee

, Krane

. Clinical outcome for atypia of undetermined significance in thyroid fine-needle aspirations: Should repeated FNA be the preferred initial approach? Am J Clin Pathol, 2011; 135(5):770–775; doi: 10.1309/AJCP4P2GCCDNHFMY

Haugen

. 2015 American Thyroid Association Management Guidelines for Adult patients with thyroid nodules and differentiated thyroid cancer: What is new and what has changed? Cancer2017, 2017; 123(3):372–381; doi: 10.1002/cncr.30360

Elomami

, Elhag

, Alseddeeqi

. Cytological sub-classification of atypia of undetermined significance may predict malignancy risk in thyroid nodules. Acta Cytol, 2021; 65(3):205–212; doi: 10.1159/000513066

Hathi

, Rahmeh

, Munro

, et al. Rate of malignancy for thyroid nodules with AUS/FLUS cytopathology in a tertiary care center - A retrospective cohort study. J Otolaryngol Head Neck Surg, 2021; 50(1):58; doi: 10.1186/s40463-021-00530-0

Nishino

, Krane

. Role of ancillary techniques in thyroid cytology specimens. Acta Cytol, 2020; 64(1–2):40–51; doi: 10.1159/000496502

10.

Ohori

. Molecular testing and thyroid nodule management in North America. Gland Surg, 2020; 9(5):1628–1638; doi: 10.21037/gs-2019-catp-26

11.

Zhang

, Smola

, Lew

, et al. Performance of Afirma genomic sequencing classifier vs gene expression classifier in Bethesda category III thyroid nodules: An institutional experience. Diagn Cytopathol, 2021; 49(8):921–927; doi: 10.1002/dc.24765

12.

Vuong

, Nguyen

TPX

, Hassell

, et al. Diagnostic performances of the Afirma Gene Sequencing Classifier in comparison with the Gene Expression Classifier: A meta-analysis. Cancer Cytopathol, 2021; 129(3):182–189; doi: 10.1002/cncy.22332

13.

Jin

, Lew

, Pantanowitz

, et al. Performance of Afirma genomic sequencing classifier and histopathological outcome in Bethesda category III thyroid nodules: Initial versus repeat fine-needle aspiration. Diagn Cytopathol, 2023; 51(11):698–704; doi: 10.1002/dc.25203

14.

Haugen

, Alexander

, Bible

, 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

15.

Baloch

, Asa

, Barletta

, et al. Overview of the 2022 WHO classification of thyroid neoplasms. Endocr Pathol, 2022; 33(1):27–63; doi: 10.1007/s12022-022-09707-3

16.

Guzman-Arocho

, VanderLaan

, Nishino

. Binary subclassification scheme (AUS-Nuclear versus AUS-Other) adequately risk-stratifies thyroid fine needle aspiration specimens classified as Atypia of Undetermined Significance. J Am Soc Cytopathol, 2024; 13(1):23–32; doi: 10.1016/j.jasc.2023.10.001

17.

Jin

, Lew

, Pantanowitz

, et al. Performance of Afirma genomic sequencing classifier and histopathological outcome are associated with patterns of atypia in Bethesda category III thyroid nodules. Cancer Cytopathol, 2022; 130(11):891–898; doi: 10.1002/cncy.22625

18.

Papazian

, Dublin

, Patel

, et al. Repeat fine-needle aspiration with molecular analysis in management of indeterminate thyroid nodules. Otolaryngol Head Neck Surg, 2023; 168(4):738–744; doi: 10.1177/01945998221093527

19.

Köseoğlu

, Özdemir Başer

, Çetin

. Malignancy outcomes and the impact of repeat fine needle aspiration of thyroid nodules with Bethesda category III cytology: A multicenter experience. Diagn Cytopathol, 2021; 49(10):1110–1115; doi: 10.1002/dc.24823

20.

Angell

, Heller

, Cibas

, et al. Independent comparison of the Afirma genomic sequencing classifier and gene expression classifier for cytologically indeterminate thyroid nodules. Thyroid, 2019; 29(5):650–656; doi: 10.1089/thy.2018.0726

21.

San Martin

, Lawrence

, Bena

, et al. Real-world comparison of Afirma GEC and GSC for the assessment of cytologically indeterminate thyroid nodules. J Clin Endocrinol Metab, 2020; 105(3); doi: 10.1210/clinem/dgz099

22.

Nishino

, Mateo

, Kilim

, et al. Repeat fine needle aspiration cytology refines the selection of thyroid nodules for Afirma Gene Expression Classifier Testing. Thyroid, 2021; 31(8):1253–1263; doi: 10.1089/thy.2020.0969

23.

VanderLaan

, Nishino

. Molecular testing results as a quality metric for evaluating cytopathologists’ utilization of the atypia of undetermined significance category for thyroid nodule fine-needle aspirations. J Am Soc Cytopathol, 2022; 11(2):67–73; doi: 10.1016/j.jasc.2021.10.001

24.

Rajab

, Payne

, Forest

, et al. Molecular testing for thyroid nodules: The experience at McGill University Teaching Hospitals in Canada. Cancers (Basel), 2022; 14(17); doi: 10.3390/cancers14174140

25.

Wei

, Veloski

, Sharda

, et al. Performance of the Afirma genomic sequencing classifier versus gene expression classifier: An institutional experience. Cancer Cytopathol, 2019; 127(11):720–724; doi: 10.1002/cncy.22188

26.

Gortakowski

, Feghali

, Osakwe

. Single institution experience with Afirma and Thyroseq testing in indeterminate thyroid nodules. Thyroid, 2021; 31(9):1376–1382; doi: 10.1089/thy.2020.0801

27.

Baca

, Wong

, Strickland

, et al. Qualifiers of atypia in the cytologic diagnosis of thyroid nodules are associated with different Afirma gene expression classifier results and clinical outcomes. Cancer Cytopathol, 2017; 125(5):313–322; doi: 10.1002/cncy.21827

28.

, Inman

, Cramer

. Subclassification of “atypia of undetermined significance” in thyroid fine-needle aspirates. Diagn Cytopathol, 2014; 42(1):23–29; doi: 10.1002/dc.23052

29.

VanderLaan

, Marqusee

, Krane

. Usefulness of diagnostic qualifiers for thyroid fine-needle aspirations with atypia of undetermined significance. Am J Clin Pathol, 2011; 136(4):572–577; doi: 10.1309/AJCPO0BQ2YSKPXXP

30.

Sheahan

, Mohamed

, Ryan

, et al. Follicular variant of papillary thyroid carcinoma: Differences from conventional disease in cytologic findings and high-risk features. JAMA Otolaryngol Head Neck Surg Surg2014; 140(12):1117–1123; doi: 10.1001/jamaoto.2014.2548

31.

Endo

, Nabhan

, Porter

, et al. Afirma gene sequencing classifier compared with gene expression classifier in indeterminate thyroid nodules. Thyroid, 2019; 29(8):1115–1124; doi: 10.1089/thy.2018.0733

32.

Geng

, Aguilar-Jakthong

, Moatamed

. Comparison of Afirma gene expression classifier with gene sequencing classifier in indeterminate thyroid nodules: A single-institutional experience. Cytopathology, 2021; 32(2):187–191; doi: 10.1111/cyt.12920

33.

Polavarapu

, Fingeret

, Yuil-Valdes

, et al. Comparison of Afirma GEC and GSC to nodules without molecular testing in cytologically indeterminate thyroid nodules. J Endocr Soc, 2021; 5(11):bvab148; doi: 10.1210/jendso/bvab148