AUS in Thyroid Cytology: A 21‐Year Institutional Experience and the Impact of Atypia Subclassification

ABSTRACT

Background

The Atypia of Undetermined Significance (AUS) category remains one of the most controversial groups in The Bethesda System for Reporting Thyroid Cytopathology (TBSRTC). We aimed to assess AUS outcomes at our institution over a 21‐year period.

Methods

Thyroid fine needle aspiration (FNA) cases between 2000 and 2021 were retrospectively reviewed, and AUS diagnoses were identified. Cases were subclassified as nuclear (AUS‐N) or other atypia (AUS‐O). The AUS rates per year and cytopathologist, risk of malignancy (ROM), and radiological features were analyzed.

Results

There were 15,099 thyroid FNAs, of which 553 initial AUS cases were included. The AUS rate was 0.7% and 6.6% before and after the implementation of the TBSRTC. AUS rates among cytopathologists ranged from 2.8% to 8.8%. Of the 553 cases, 168 (30%) underwent surgery, 269 (48.6%) had a repeat FNA, and 116 (20.9%) were observed or lost to follow‐up. Pre‐TBSRTC the overall ROM was 20%, and post‐TBSRTC 30.1% and 26.7% when NIFTP was considered malignant, and nonmalignant, respectively. ROM was significantly higher in AUS‐N than AUS‐O (39% vs. 20.7% with NIFTP malignant; 36% vs. 17.6% without; p < 0.01). ROM did not differ significantly between patients with one or two consecutive AUS diagnoses (26.7% vs. 25%). On ultrasonographic findings, calcifications, and cystic changes were more frequent in AUS‐N, while nodule size and multiplicity were similar between groups.

Conclusion

In conclusion, AUS‐N and AUS‐O demonstrated significantly different ROMs, supporting the 2023 TBSRTC subclassification. Repeat FNA led to diagnostic reclassification in a substantial proportion of cases, supporting its role as an acceptable management strategy, particularly in settings with limited access to molecular testing.

1. Introduction

Fine‐needle aspiration (FNA) cytology is an essential diagnostic tool for evaluating thyroid nodules, identifying malignancies, and guiding appropriate management. To standardize the reporting of thyroid FNAs, a classification system was initially proposed in 1996 by the Papanicolaou Society of Cytopathology. This system gained global recognition with the publication of the first edition of “The Bethesda System for Reporting Thyroid Cytopathology” (TBSRTC) in 2010. The classification was further revised in 2017 and again in 2023, in the second and third editions [1, 2]. The most recent edition presents six diagnostic categories, which include three indeterminate categories: “atypia of undetermined significance” (AUS), “follicular neoplasm” (FN)/“oncocytic follicular neoplasm” (OFN), and “suspicious for malignancy” (SM), in addition to the nondiagnostic, benign, and malignant categories.

One of the most controversial categories of the TBSRTC has been the “atypia” category, namely, AUS. This category encompasses a wide range of abnormal architectural and cellular changes that do not meet the criteria for more definitive diagnoses [2]. Therefore, interpretations can vary significantly among different cytopathologists and laboratories, leading to considerable variation in the reported malignancy rates [3].

In the first and second editions, this category was referred to as “AtUS” or “Follicular Lesion of Undetermined Significance” (AUS/FLUS). Although “Follicular Lesion of Undetermined Significance” (FLUS) was regarded as an acceptable alternative to AUS, the inconsistent usage of these two terms has caused confusion. The third edition recommended using the preferred AUS terminology for this category to enhance clarity and consistency.

In the recent 2023 edition of TBSRTC, the term “AUS” is further classified to include subcategories: “AUS with nuclear atypia” and “AUS‐other,” which encompasses architectural atypia, oncocytic atypia, atypical lymphoid cells, and atypia not otherwise specified [2].

The management options for patients with the diagnosis of AUS include repeat FNA, molecular testing, diagnostic lobectomy, or surveillance. Molecular testing was suggested as a complement to cytopathologic examination in the AUS category, which can provide additional diagnostic information and significantly impact clinical management decisions. Some commercially available molecular thyroid tests, incorporating several genetic alterations, have been developed for the risk stratification of nodules with indeterminate diagnoses, which have different negative (NPV) and positive predictive values (PPV) [4, 5, 6, 7, 8]. Repeat FNA is often the preferred approach, particularly in centers without access to molecular testing [9, 10]. It provides an opportunity to re‐evaluate the nodule, potentially leading to a more definitive diagnostic classification. Surgical excision remains an option for many AUS nodules based on clinical factors such as size, cosmetic concerns, suspicious radiological findings, or patient preferences.

To promote uniform practice among laboratories and cytopathologists and encourage quality improvement, it is desirable to establish and follow performance metrics associated with TBSRTC, especially concerning the AUS category.

In this study, we aimed to assess the outcomes of the AUS category and share our experience over 21 years at our institution. Our study included cases from 2000 to 2021 to demonstrate how the atypia rate at our institution evolved over a 21‐year period, including the era before and after the adoption of TBSRTC. Our objective was to analyze the risk of malignancy (ROM) for the AUS category before and after the TBSRTC, examine the difference in ROM and ultrasonographic findings of the subcategories as “nuclear” versus “other” atypia.

2. Materials and Methods

This study was conducted under an institutional review board‐approved protocol.

All thyroid fine needle aspiration (FNA) biopsies, between January 2000 and December 2021, were retrieved from the institutional electronic medical record system (Sunquest CoPathPlus, Tucson, AZ). Throughout the entire 21‐year study period, all cytological specimens were prepared as conventional smear slides. “The Bethesda System for Reporting Thyroid Cytopathology (TBSRTC)” was first implemented at our institution in 2012, and after that date, thyroid nodules diagnosed with “atypia of undetermined significance” or “follicular lesion of undetermined significance” were selected and the reason for atypia (nuclear or architectural/other) was mentioned. Before TBSRTC, a 5‐tier classification system (Nondiagnostic, Benign, Atypical, Suspicious, Positive) with explanatory comments was used. Prior to the adoption of TBSRTC (before 2012), cases with a diagnosis of “atypia”/“atypical” were selected from the institutional database.

At our institution, the cytopathology reports of thyroid FNAs have a detailed microscopic description, particularly for indeterminate categories. Two pathologists (O.K., G.A.B.) independently reviewed all cytopathology reports of AUS cases and subclassified them into AUS‐N (nuclear atypia) or AUS‐O (other) based on the descriptive microscopic features. A predefined set of criteria was used to guide subclassification: AUS‐N included cases with features such as pale chromatin, nuclear enlargement, nuclear grooves, pseudoinclusions, and irregular nuclear contours; AUS‐O included cases with architectural atypia, microfollicular pattern, oncocytic changes, or other atypical features lacking nuclear changes characteristic of papillary thyroid carcinoma. In instances of disagreement, a consensus diagnosis was reached through joint review and discussion. The reviewers were blinded to final histologic outcomes and follow‐up data.

Patient records were reviewed for subsequent clinical follow‐up and radiological findings. The surgical outcome or follow‐up data of the patients were recorded. Patients with no subsequent histological diagnosis and clinical follow‐up were defined as “lost to follow‐up.”

For nodules that were resected, the histopathologic diagnosis of the biopsied nodule was recorded based on the review of surgical pathology reports. Nodule size and location were used to match the aspirated nodule with the resected nodule.

The ROM was calculated based on histologic outcomes in nodules that underwent surgical resection. The ROM calculations were made separately for cases diagnosed before and after implementation of TBSRTC criteria for a clearer, methodologically appropriate comparison of outcomes aligned with contemporary TBSRTC definitions.

The “noninvasive follicular thyroid neoplasm with papillary‐like nuclear features (NIFTP)” entity was introduced in 2016. It remains controversial to consider this entity as benign or malignant in ROM calculations. To address the variability in ROM estimates related to NIFTP classification, ROM was calculated using two approaches: (1) including NIFTP in the nonmalignant category and (2] considering NIFTP as malignant, consistent with molecular‐based risk stratification models that incorporate NIFTP in ROM and positive predictive value calculations.

Prior to the introduction of the NIFTP entity in 2016, in the absence of retrospective histologic slide review, an inherent limitation of the study due to incomplete availability of original resection materials, cases diagnosed as encapsulated follicular variant of papillary thyroid carcinoma (FVPTC) were retained in the malignant category for outcome analysis, as not all encapsulated FVPTCs meet current diagnostic criteria for NIFTP.

The molecular studies were analyzed, when available. In our institution, the specimens are sent out for ThyroSeq molecular testing (CBLPath Laboratories, Rye Brook, NY).

We analyzed AUS rates per year and per cytopathologist. The ROM was calculated according to surgical outcome data separately for the AUS‐N and AUS‐O subcategories, before and after the implementation of the TBSRTC in 2012. The results were statistically compared between the groups using the chi‐square test. Radiological features and the type of surgery were compared between the groups using Fisher's exact test.

3. Results

3.1. AUS Rates

During a 21‐year span of time, 10 cytopathologists evaluated a total of 15,099 thyroid FNAs, of which 654 were diagnosed as AUS. After the exclusion of the 20 cases classified as “atypical lymphoid cells” and 81 cases re‐categorized as AUS on repeated FNA (76 cases on second FNA and 5 cases on third FNA), 553 cases remained. Of the 553 cases diagnosed as AUS on initial FNA, 526 unique patients were included in the study, as some patients underwent FNA of more than one nodule. A total of 168 (30%) cases of 553 directly underwent surgery, 269 (48.6%) had a repeat FNA, and 116 (20.9%) were observed or lost to follow‐up (shown in Figure 1).

FIGURE 1.

Study design with inclusion and exclusion criteria. [Color figure can be viewed at wileyonlinelibrary.com]

The mean age of the patients was 45.2 years; ranging from 6 to 89 years. There was a female predominance; 405 females (76.9%) and 121 males (23.1%), with a female‐to‐male ratio of 3.34:1. Of 553 nodules, 272 (49.1%) were located in the right lobe, 252 (45.5%) in the left lobe, 26 (4.7%) in the isthmus, and 3 (0.5%) in the thyroidectomy bed.

Prior to the implementation of TBSRTC in 2012, a total of 5877 thyroid FNAs were performed, of which 42 (0.7%) were classified as “atypical (AUS).” The cases were evaluated by six cytopathologists. Among these 42 cases, 19 (45.2%) underwent immediate surgical intervention, 8 (19%) had a repeat FNA, and 15 (35.7%) were observed or lost to follow‐up. After subclassification, 40 cases were categorized as AUS‐N and 2 as AUS‐O, resulting in an AUS‐N:AUS‐O ratio of 20.

Following the implementation of TBSRTC, a total of 9222 thyroid FNAs were performed over a 10‐year period, of which 612 (6.6%) were diagnosed as AUS. These cases were evaluated by nine cytopathologists. After excluding 20 cases reported as “atypical lymphoid cells” and 81 cases that were re‐categorized as AUS on repeat FNA, 511 cases remained for analysis. After subclassification, 162 cases were categorized as AUS‐N and 349 as AUS‐O, resulting in an AUS‐N:AUS‐O ratio of 0.46. Of these cases, 149 (29.1%) underwent immediate surgery, 261 (51%) had a repeat FNA, and 101 (19.7%) were observed or lost to follow‐up.

The laboratory's annual AUS rate ranged from 0.16% to 9.1%, with an overall rate of 4.2% over the 21‐year study period. After the implementation of TBSRTC, the AUS rate has changed between 3.9% and 7.1%, with an overall rate of 6.6% in 10 years (Figure 2).

FIGURE 2.

The annual rate of AUS in thyroid FNA's in a 21‐year span. [Color figure can be viewed at wileyonlinelibrary.com]

Each cytopathologist's AUS rates were analyzed and the mean AUS rates ranged between 0.9% and 8.8%. After the implementation of TBSRTC, each cytopathologist's AUS rates ranged between 2.8% and 8.84% (Figure 3). Before TBSRTC, all was under 1%.

FIGURE 3.

AUS rates among individual pathologists in thyroid FNA diagnosis, after the implementation of TBSRTC. [Color figure can be viewed at wileyonlinelibrary.com]

3.2. Follow‐Up Information

Follow‐up data were obtained, and the ROM was calculated separately for cases diagnosed before and after the implementation of TBSRTC criteria.

Before the implementation of TBSRTC, after a single AUS diagnosis, 19 of 42 patients underwent surgery, of which 3 were diagnosed as papillary thyroid carcinoma, 4 as follicular/oncocytic adenoma, and 12 as follicular nodular disease. Eight patients had a repeat FNA and were classified as benign. One of them underwent surgery and was diagnosed with papillary thyroid carcinoma. Of these 20 cases that underwent surgery, 19 were AUS‐N and 1 was AUS‐O. The overall ROM was 20% (4:20); the ROM for AUS‐N and AUS‐O was 21% and 0%, respectively.

Following the implementation of TBSRTC, after a single AUS diagnosis, 149 (29.1%) patients underwent surgery. Of these 149 cases, 56 were AUS‐N, and 93 were AUS‐O. The histopathological outcomes were as follows: Benign: 75 (50.3%), Follicular/Oncocytic adenoma: 30 (20.1%), NIFTP: 9 (6%), Follicular carcinoma: 7 (4.6%), Papillary thyroid carcinoma: 25 (16.7%), the others (Langerhans cell histiocytosis, large cell lymphoma, hyalinizing trabecular tumor): 3 (1.9%). The ROM after a single AUS diagnosis was 28.1% (42/149) when NIFTP was counted as malignant, and 22.1% (33/149) as nonmalignant (shown in Table 1).

TABLE 1.

Surgical outcomes of cases that underwent surgery after a single FNA with AUS‐N and AUS‐O diagnoses, before and after the implementation of TBSRTC.

Surgical outcome	Before TBSRTC			After TBSRTC
	AUS‐N (n = 18)	AUS‐O (n = 1)	Overall (n = 19)	AUS‐N (n = 57)	AUS‐O (n = 92)	Overall (n = 149)
FND	11 (61.1%)	1 (100%)	12 (63.1%)	32 (56.1%)	43 (46.7%)	75 (50.3%)
FA/OA	4 (22.2%)	—	4 (21%)	2 (3.5%)	28 (30.4%)	30 (20.1%)
FC	—	—	—	3 (5.2%)	4 (4.3%)	7 (4.6%)
PTC	3 (16.6%)	—	3 (15.7%)	13 (22.8%)	12 (13%)	25 (16.7%)
NIFTP	—	—	—	5 (8.7%)	4 (4.3%)	9 (6%)
Other B	—	—	—	1 (1.7%)	1 (1%)	2 (1.3%)
Other M	—	—	—	1 (1.7%)	—	1 (0.6%)

Abbreviations: AUS‐N: atypia of undetermined significance‐nuclear, AUS‐O: atypia of undetermined significance‐other, B: benign neoplasms, FA/OA: follicular adenoma/oncocytic adenoma, FC: follicular carcinoma, FND: follicular nodular disease, M: malignant neoplasms, NIFTP: noninvasive follicular thyroid neoplasm with papillary‐like nuclear features, PTC: papillary thyroid carcinoma.

After the initial AUS diagnoses, 261 (51%) cases underwent a repeat FNA. Repeat FNA reclassified 66.6% of the AUS cases to a different category: 135 (51.7%) cases were reclassified as benign, 23 (8.8%) as FN, and 16 (6.1%) as SM or malignant. Eleven (4.2%) cases were nondiagnostic (shown in Figure 1).

The diagnoses of 76 (29.1%) patients remained the same as AUS; of the 20 cases initially diagnosed as AUS‐N, 11 remained the same, while 9 were reclassified as AUS‐O. Of the 56 cases initially diagnosed as AUS‐O, 51 remained the same while 5 were reclassified as AUS‐N. Of 76 patients who had two consecutive AUS diagnoses, 40 (52.6%) underwent surgery. The ROM after two consecutive AUS diagnoses was 25% (10:40) when NIFTP was counted as malignant, and 20% (8:40) as nonmalignant (six papillary carcinomas, one follicular carcinoma, two NIFTP, and one medullary carcinoma). The second FNA diagnoses and surgical outcomes are shown in Table 2.

TABLE 2.

During 21 years, second repeated FNA diagnoses of AUS cases with a following thyroidectomy and their surgical outcomes.

Surgical outcomes	2ND FNA diagnoses							Overall (n = 91)
	AUS‐N (n = 8)	AUS‐O (n = 32)	ND (n = 3)	B (n = 15)	FN (n = 18)	SM (n = 6)	M (n = 9)
FND	4	11	3	12	4	2		36 (39.5%)
FA/OA	2	13			12	1		28 (30.4%)
FC		1						1 (1%)
PTC	2	4		3	2	3	9	23 (25.2%)
NIFTP		2						2 (2.1%)
Other M		1						1 (1%)

Abbreviations: AUS‐N: atypia of undetermined significance‐nuclear, AUS‐O: atypia of undetermined significance‐other, B: benign neoplasms, FA/OA: follicular adenoma/oncocytic adenoma, FC: follicular carcinoma, FND: follicular nodular disease, M: malignant neoplasms, NIFTP: noninvasive follicular thyroid neoplasm with papillary‐like nuclear features, PTC: papillary thyroid carcinoma.

Ultimately, after TBSRTC, 239 cases (46.7%) underwent surgical resection, with or without repeated FNA. After the first FNA, 149 cases, and after second or more repeated FNAs, 90 cases underwent surgery. The overall ROM was 30.1% (n = 72) when NIFTP was counted as malignant, and 26.7% (n = 64) as nonmalignant.

During 21 years, the ROM after a single AUS diagnosis was 26.7% (45:168) when NIFTP was counted as malignant, and 21.4% (36:168) as nonmalignant. The ROM for AUS‐N and AUS‐O subcategories was 39% and 20.7%, respectively when NIFTP was considered malignant and 36% and 17.6% when nonmalignant. The surgical outcomes are shown in Table 3.

TABLE 3.

During 21 years, overall surgical outcomes of AUS‐N and AUS‐O cases.

Surgical outcome	AUS‐N (n = 100)	AUS‐O (n = 159)	Overall (n = 259)
FND	53 (53%)	70 (44%)	123 (47.5%)
FA/OA	7 (7%)	55 (34.6%)	62 (23.8%)
FC/OC	3 (3%)	5 (3.1%)	8 (3%)
PTC	31 (31%)	23 (14.4%)	54 (20.8%)
NIFTP	3 (3%)	5 (3.1%)	8 (3%)
Other B	1 (1%)	1 (0.6%)	2 (0.8%)
Other M	2 (2%)	—	2 (0.8%)

Abbreviations: AUS‐N: atypia of undetermined significance‐nuclear, AUS‐O: atypia of undetermined significance‐other, B: benign neoplasms, FA/OA: follicular adenoma/oncocytic adenoma, FC: follicular carcinoma, FND: follicular nodular disease, M: malignant neoplasms, NIFTP: noninvasive follicular thyroid neoplasm with papillary‐like nuclear features, PTC: papillary thyroid carcinoma.

Of 260 patients who underwent surgical resection, total thyroidectomy was performed in 121 (46.5%) patients and hemithyroidectomy in 139 (53.4%) patients. The rate of total thyroidectomy procedures for the AUS‐N versus AUS‐O group was 54.4% and 41.5%, respectively. Patients in the AUS‐N group were significantly more likely to undergo total thyroidectomy compared with those in the AUS‐O group (Fisher's exact test, p = 0.045).

3.3. Imaging Findings

The ultrasound (US) findings were available for 533 cases, 341 AUS‐O and 192 AUS‐N cases. The mean nodule diameter was similar between the two groups: in the AUS‐O group 2.31 cm (range: 0.5–7.8 cm) and in the AUS‐N group 2.28 cm (range: 0.6–11.9 cm).

Multinodularity was more common than solitary nodules in both groups and was slightly more frequent in the AUS‐N group than in the AUS‐O group 79.6% versus 75.6% (153 vs. 258 cases), although the difference was not statistically significant (Fisher's exact test, p = 0.32).

Calcifications were observed in 57 (16.7%) AUS‐O cases and in 51 (26.6%) AUS‐N cases. Cystic features were present in 69 nodules (20.2%) in the AUS‐O group and 66 nodules (34.4%) in the AUS‐N group. Both calcifications and cystic changes were significantly more common in the AUS‐N group (Fisher's exact test, p = 0.017 and p < 0.001, respectively).

The information of echogenicity was available for 284 nodules. Hypoechogenicity was the most common finding in both groups, observed in 99 (29%) AUS‐O cases and 52 (27.1%) AUS‐N cases, but the difference wasn't statistically significant (Fisher's exact test, p = 0.09).

Radiological features are summarized in Table 4. Thyroid Image Reporting and Data System (TI‐RADS) classification was not available for the majority of the specimens and was not evaluated.

TABLE 4.

Ultrasound characteristics of thyroid nodules in AUS‐N and AUS‐O cases.

Ultrasound feature	AUS‐N (n = 192)	AUS‐O (n = 341)	Overall (n = 533)
Nodule size (cm)
Median (range)	2.28 (0.6–11.9)	2.31 (0.5–7.8)	2.3 (0.5–11.9)
Nodule number
Multinodular	153 (79.6%)	258 (75.6%)	411 (77.1%)
Solitary	39 (20.3%)	83 (24.3%)	122 (22.8%)
Composition
Cystic component	66 (34.4%)	69 (20.2%)	135 (25.3%)
Calcification
Calcification positive	51 (26.5%)	57 (16.7%)	108 (20.2%)

Echogenicity	AUS‐N (n = 85)	AUS‐O (n = 199)	Overall (n = 284)
Hypoechoic	52 (61.1%)	99 (49.7%)	151 (53.1%)
Isoechoic	25 (29.4%)	64 (32.1%)	89 (31.3%)
Hyperechoic	8 (9.4%)	36 (18%)	44 (15.5%)

3.4. Molecular Findings

Seventeen cases, 5 AUS‐N and 12 AUS‐O cases, were sent for molecular analysis. Seven cases without a mutation underwent follow‐up, while five were lost to follow‐up. Five cases that were diagnosed as AUS‐O underwent surgery. Three cases harboring KRAS, HRAS, and PAX8/PPARγ mutations, respectively, were ultimately diagnosed as follicular adenoma. Two of them had a low mutational status in DICER1 and TSHR and were diagnosed as follicular nodular disease on histologic examination.

4. Discussion

TBSRTC has been in use since 2010 and was revised in 2017 and 2023. This system has established standardized terminology and management recommendations to reduce ambiguous reports, particularly for the “indeterminate” categories. Among these categories, the “AUS” has been the most controversial and heterogeneous, as the associated ROM and clinical significance have been subjects of ongoing debate and investigation.

The findings of this study highlight several important aspects of the AUS category. The recommendation that accompanied TBSRTC for thyroid cytopathology was that the AUS rate not exceed 10%. Our institution's AUS rate was 0.7% and 6.6% before and after the implementation of the TBSRTC, respectively. Overall rate of AUS was 4.2%, which is well below the recommended threshold. A noteworthy increase in the AUS rate was observed following the implementation of TBSRTC in daily practice in 2012. This increase could be explained by the designation of atypia before the adoption of TBSRTC in 2012, which may not fully meet contemporary AUS criteria, and pre‐TBSRTC “atypical” diagnoses may represent a narrower subset. And also, cytology descriptions can vary significantly among pathologists, especially over long study periods. This contributes to observed differences in AUS rates over time. In the literature, particularly during the early adoption of TBSRTC, AUS rates were reported to be significantly higher [11, 12, 13, 14]. It was explained by the challenges of getting used to a new classification system. Because the AUS category became an officially recognized and standardized part of reporting, cytopathologists and clinicians may have felt more comfortable using it for uncertain cases, leading to a higher number of AUS diagnoses.

However, several recent publications from large academic centers also report rates substantially exceeding this threshold in a range of 17.5%–22% [15, 16, 17, 18]. These findings suggest that a universal 10% target may not accurately reflect the practice patterns or the unique patient populations seen at high‐volume academic institutions. The variation in AUS rates is increasingly attributed to pre‐FNA factors, such as the implementation of advanced imaging reporting systems like TI‐RADS, which divert radiologically benign cases away from biopsy and increase the proportion of neoplastic nodules in the remaining population. Consequently, a higher AUS rate may represent a more selective and higher‐risk clinical cohort rather than diagnostic overutilization. On the other hand, prior to the advent of molecular testing, cytopathologists recognized that an AUS diagnosis often necessitated potentially redundant diagnostic surgeries. However, the integration of molecular profiling has likely shifted the threshold for AUS classification, as genomic data now facilitate a more precise differentiation between low‐risk and high‐risk nodules. Given this context, recent literature advocates for institutional benchmarking rather than a fixed “one‐size‐fits‐all” target. Laboratories are encouraged to determine appropriate thresholds based on their specific laboratory metrics and pathologist experience.

Furthermore, with the advancement of molecular tests for triaging indeterminate thyroid nodules, it has been proposed to combine molecular testing results with AUS diagnoses as an additional tool to monitor the performance of thyroid cytopathology practices [15, 16, 17, 18, 19, 20]. Ohori et al. demonstrated that positive call rate and molecular‐derived risk of malignancy (MDROM) could be used as a quality metric to provide potential strategies for practice improvement [21]. Thus, to promote and encourage quality improvement, it is desirable to establish and follow performance metrics associated with TBSRTC, especially concerning the AUS category.

In our study, the AUS rate showed considerable variation among the cytopathologists. After the adoption of TBSRTC, the mean AUS rate of each cytopathologist has ranged from 2.8% to 8.8%, yet it did not exceed the recommended target of 10%. This variance of cytopathologists' AUS reporting rate is not unexpected; in fact, interobserver variability in interpreting indeterminate categories is well documented in many studies. Studies by Padmanabhan et al. [22], Słowińska‐Klencka et al. [3], and Kocjan et al. [23] reported that the interobserver variability for indeterminate categories can range from fair to poor. The interpretation of indeterminate cytomorphological features in thyroid FNAs is affected by several factors, including the cytopathologist's experience, the methods of specimen preparation, the quality of the preparation, the number of FNA passes, and the availability of rapid on‐site evaluation [24].

In our study, we analyzed the ROM of AUS cases. When calculating the ROM, there are two points to consider. First, a small fraction of AUS nodules are surgically removed based on clinical or imaging features, leading to a selection bias that overestimates the ROM. Second, the presence of NIFTP lowers the ROM. According to 2023 TBSRTC, the estimated ROM for the AUS category is 13%–30%. Notably, when nodules diagnosed on surgical pathology as NIFTP are excluded from the malignant cohort, the risk declines to a range of 6%–20% [2].

In the literature, there is a broad variation in ROM among laboratories, ranging from 11.3% to 75.8%, and these rates are mostly higher than the recommended ranges considered in the TBSRTC [25, 26, 27, 28, 29, 30]. With the implementation of NIFTP diagnosis in the nonmalignant category, these rates decreased but still remained higher than the expected 6%–18% range [31, 32, 33]. A meta‐analysis by Haaga et al. indicated an overall ROM of 36.6% and 29.2% for the AUS category before and after the implementation of NIFTP diagnosis, respectively [25]. We calculated ROM values both ways, as counting NIFTP as malignant and nonmalignant. Before TBSRTC, the overall ROM was 20%, and after TBSRTC the overall ROM was 30.1% when NIFTP was considered malignant and 26.7% when nonmalignant. These data suggest that AUS nodules in certain practice settings may carry a higher ROM than previously thought, indicating that guidelines recommending repeat FNA or observation should be reconsidered.

When cases were further subdivided according to nuclear atypia (AUS‐N) and other atypia (AUS‐O), the ROM was 39% and 20.7%, respectively, when NIFTP was counted as malignant and 36% and 17.6% when nonmalignant. The ROM for AUS‐N was significantly higher than the ROM observed in AUS‐O (chi‐square p < 0.01). These findings are consistent with previous studies suggesting that nuclear or cytologic atypia in thyroid FNAs carries a higher malignancy risk than architectural atypia [34, 35, 36, 37, 38, 39, 40, 41]. As a result, the most recent edition of the TBSRTC, the AUS category has been subdivided into AUS‐N and AUS‐other subcategories. This distinction is crucial for clinical management, as the risk associated with cytologic atypia may necessitate a more aggressive approach, including earlier surgery or more frequent follow‐ups.

Following a diagnosis of AUS from an initial thyroid FNA, performing a repeat FNA is generally considered appropriate in most cases to achieve a more definitive risk stratification. In our study, the majority of cases were reclassified into a more definitive category: 53.1% cases reclassified as benign, and 5.9% as SM or malignant. The diagnoses of 28.2% cases remained the same as AUS.

The ROM after a single AUS diagnosis versus two consecutive AUS diagnoses was 26.7%, and was 25%, respectively when NIFTP was considered malignant, 21.4% and 20%, when nonmalignant. There was no statistically significant difference in the ROM between cases with a single AUS diagnosis and those with repeated AUS diagnoses (p = 0.82). In the series by Evranos Ogmen et al. [42], the ROM for patients who underwent surgery after an initial AUS diagnosis and two repeated AUS diagnoses was reported as 30.6% and 27.3%, respectively. This finding aligns with several studies by Vanderlaan et al. [43], Broome et al. [44], and Ho et al. [45], which indicated that the malignancy rates were not significantly different after one or two AUS diagnoses.

In our cohort only a small subset of cases underwent molecular testing [ThyroSeq molecular testing (CBLPath Laboratories, Rye Brook, NY)], and the results provided some additional insights into the potential role of molecular studies in risk stratification. Molecular testing has been increasingly utilized to refine the management of indeterminate thyroid nodules, particularly for cases classified as AUS [5, 6]. In our cohort, molecular testing revealed that some cases with negative or low mutation status had benign outcomes, whereas cases with mutations (e.g., KRAS, HRAS, PAX8/PPARγ) were more likely to undergo surgery and were often found to be follicular adenomas. These findings are consistent with existing literature, which suggests that molecular markers can provide additional predictive value, especially for FNs.

Our study has several limitations inherent to its retrospective design. First, original cytology slides from a substantial proportion of cases, particularly those obtained more than 10 years ago, were no longer available at our institution for retrospective re‐review. Consequently, cytologic interpretations were based on detailed original microscopic descriptions documented in the cytopathology reports. All reports were systematically reviewed according to predefined criteria and subsequently subclassified as AUS‐N or AUS‐O. However, the designation of atypia in cases reported prior to 2012 did not follow the formalized AUS criteria of TBSRTC, and therefore some earlier cases may not fully meet contemporary AUS definitions and pre‐TBSRTC “atypical” diagnoses may represent a narrower subset. In addition, cytologic descriptions may vary among pathologists, particularly across extended study periods, potentially introducing interpretive variability. Prior to the adoption of the 3rd Edition of TBSRTC, our institution used a descriptive reporting structure that divided atypia into AUS‐N and AUS‐O categories, and this variability may introduce interpretive bias across the extended study period.

Second, surgical resection specimens were not re‐reviewed by the study authors. Original resection slides from earlier years were not universally available for review, representing an inherent limitation of the retrospective study design. Nonetheless, all thyroid resection specimens at our institution are evaluated within a subspecialty sign‐out system by dedicated head and neck pathologists, which likely contributes to diagnostic consistency. Cases diagnosed as encapsulated FVPTC prior to the introduction of NIFTP in 2016 were therefore categorized as malignant in outcome analyses. Given the ongoing debate regarding the most appropriate classification of NIFTP as benign or malignant for ROM calculations, both approaches were considered; however, this may still introduce variability in ROM estimates across the study cohort.

Third, this study is based solely on morphologic assessment without the incorporation of molecular testing. Although our findings are largely concordant with those of molecular‐based studies, the lack of molecular correlation represents an additional limitation. Unlike the study by Słowińska‐Klencka [10], molecular testing in our laboratory cannot be used as a surrogate marker for quality control due to the limited number of tests requested and the fact that it is a send‐out test rather than an in‐house procedure. However, implementing an ongoing quality dashboard, as proposed by Cibas et al., could enhance QA performance metrics by allowing real‐time access to data, facilitating the analysis of recent results, and tracking trends over time [34].

In conclusion, this study provides additional information about the clinical significance of the AUS category and supports the applicability of the 2023 Bethesda System in categorizing AUS subtypes, leading to improved diagnostic accuracy for thyroid cytopathology.

Funding

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.