Bethesda Category 3 Thyroid Treatment and Management

A Bethesda Category 3 thyroid nodule, formally called “atypia of undetermined significance” (AUS), sits in a gray zone where the cells collected during a needle biopsy look neither clearly normal nor clearly cancerous. The risk of malignancy in this category averages about 22%, though individual studies report rates anywhere from roughly 10% to 50% depending on the institution and population studied. Because that range is wide enough to make both “just watch it” and “operate immediately” feel like the wrong call, management hinges on narrowing the uncertainty with additional tools: repeat biopsy, molecular testing, ultrasound features, and clinical judgment. There is no single correct path for every patient, but the options are well defined and increasingly effective at sorting who needs surgery from who can safely avoid it.

What Bethesda Category 3 Actually Means

The Bethesda System for Reporting Thyroid Cytopathology classifies fine-needle aspiration (FNA) results into six categories, from nondiagnostic (Category 1) to malignant (Category 6). Category 3, AUS, is reserved for samples that show some worrisome features but not enough to push the diagnosis into a higher-risk bracket. The 2023 edition of the Bethesda System updated the implied risk of malignancy for Category 3 to about 22%, with an expected range of 12–30%.¹ That revision also simplified Category 3 into two subgroups: “AUS–nuclear atypia” and “AUS–other,” replacing a more complicated earlier scheme.² The distinction matters because nuclear atypia carries a meaningfully higher cancer risk than other patterns of atypia, a point that shapes the next steps.

When a condition called NIFTP (noninvasive follicular thyroid neoplasm with papillary-like nuclear features) is excluded from the malignancy count, the risk for Category 3 drops further, to around 16%.¹ NIFTP was reclassified as a non-malignant tumor in 2016 because it behaves in an extremely indolent way, so whether your pathologist still counts it as cancer changes the math. Multiple institutions have confirmed this drop: one large Indian cohort saw the AUS malignancy rate fall from about 43% to 32% once NIFTP was reclassified, and a U.S. single-center study reported a decline from 55% to 48%.³⁴ These numbers are higher than the Bethesda system’s average because both institutions are surgical referral centers where the patient population is enriched for malignancy.

Why the Type of Atypia Matters

Not all Category 3 results carry the same weight. Nodules flagged for nuclear atypia, meaning the cell nuclei look abnormal in ways that mimic papillary thyroid cancer, tend to be malignant at roughly twice the rate of nodules with only architectural atypia, where the arrangement of the cells rather than the cells themselves looks unusual. One study found a malignancy rate of about 37% for nuclear atypia versus 15% for architectural atypia.⁵ A separate study confirmed those numbers almost exactly, and a formal statistical analysis showed the odds of cancer with nuclear atypia were roughly six times higher than with architectural atypia.⁶

The practical implication is that a patient whose pathology report specifies nuclear atypia may be steered more quickly toward molecular testing or surgery, while someone with architectural atypia might reasonably start with a repeat biopsy or surveillance. In pediatric patients, this gap is even more pronounced: one study found that about 52% of AUS-nuclear atypia nodules in children were malignant, compared with 15% of AUS-other nodules.⁷ Children with nuclear atypia in particular deserve close follow-up or early intervention.

Repeat Biopsy as a First Step

For many patients, the initial management after a Category 3 result is simply to repeat the FNA. The logic is straightforward: the first sample may have been borderline because of limited material, blood contamination, or sampling error. A second pass sometimes yields a definitive benign or malignant reading, and guidelines from organizations like the American Thyroid Association recommend repeat FNA as a reasonable first move. One multicenter study found that repeat biopsies reduced unnecessary surgeries by catching more benign nodules on the second attempt, sparing those patients an operation they did not need.⁸

There is, however, a catch. If the second biopsy also comes back as Category 3, the risk of malignancy actually goes up. One Korean study of 67 nodules with repeat Category 3 results found a malignancy rate of about 73%, significantly higher than the roughly 60% seen on the initial Category 3 sample in that same surgical population.⁹ A separate study similarly supported surgical intervention after a second Category 3 result, finding that about 64% of those nodules proved malignant on final pathology.¹⁰ So a single repeat result of Category 3 should prompt a more decisive step, whether that is molecular testing or surgery, rather than yet another FNA.

Core Needle Biopsy as an Alternative

When a repeat FNA is planned, some centers offer core needle biopsy (CNB) instead. Because CNB retrieves a small cylinder of tissue rather than loose cells, it provides the pathologist with more architecture to evaluate. In one head-to-head comparison, CNB produced inconclusive results far less often than repeat FNA for AUS nodules: about 27% of the time versus 49%.¹¹ A more recent study confirmed these findings, reporting a diagnostic rate of about 71% for CNB compared with roughly 36% for repeat FNA.¹² CNB is not universally available and carries a slightly different risk profile (more discomfort, rare risk of hematoma), but where it is offered, it can break the diagnostic logjam more effectively than simply repeating a fine-needle aspiration.

Molecular Testing and How It Changes the Decision

Molecular testing has become one of the most significant developments in managing indeterminate thyroid nodules. Two major commercial platforms dominate the landscape: Afirma (which includes the Gene Expression Classifier and the newer Genomic Sequencing Classifier, or GSC) and ThyroSeq (currently version 3). Both use leftover material from the original FNA sample, so no additional needle stick is needed if the lab saved the specimen.

These tests work differently, and the distinction is worth understanding. Afirma’s GSC is primarily designed as a “rule-out” test: when it returns a benign result, you can be quite confident the nodule is not cancerous. A meta-analysis of molecular testing platforms for indeterminate nodules found that Afirma GSC had the best rule-out performance among the tests studied.¹³ One institutional comparison found that the newer GSC significantly outperformed the older GEC, with a benign call rate of about 77% versus 52% and improved specificity.¹⁴ Across all AUS subcategories, the GSC maintained a negative predictive value of 100% in one study, meaning every nodule it called benign was in fact benign on surgical pathology.¹⁵

ThyroSeq v3 takes a broader approach, scanning for mutations and gene fusions across more than 100 genes.¹⁶ Its negative predictive value for AUS nodules was reported at about 99.5% in one institutional study, and the same meta-analysis mentioned above ranked ThyroSeq v3 as having the best overall discriminative ability among the platforms tested.¹⁷¹³ When ThyroSeq v3 returns a positive result, the probability of malignancy is high: one two-center study reported that about 68% of surgically resected ThyroSeq v3-positive Bethesda 3 nodules were malignant.¹⁸

Cost-effectiveness analyses generally favor molecular testing over proceeding straight to diagnostic surgery. One study found the cost per correct diagnosis was roughly $14,000 for ThyroSeq v3, about $18,000 for Afirma GSC, and over $38,000 for diagnostic lobectomy.¹⁹ A Canadian analysis found that adding routine molecular testing to the management of indeterminate nodules raised mean costs by about $2,000 per patient but improved the rate of appropriate management from 64% to 89%.²⁰ Not every health system covers the cost, which is a real-world barrier. In the United States, most private insurance plans cover at least one of the major platforms for indeterminate nodules, but out-of-pocket exposure varies.

How Ultrasound Features Refine Risk

The cytology result does not exist in a vacuum. Ultrasound findings add a second layer of risk stratification that can push the needle toward or away from intervention. Systems like ACR TI-RADS score nodules based on features including composition, echogenicity, shape, margins, and calcifications. A nodule that looks low risk on ultrasound and falls into Bethesda 3 on cytology is less worrying than one with suspicious imaging features and the same cytology. One large study found that ACR TI-RADS 3 nodules had a negative predictive value above 94%, meaning the vast majority not recommended for biopsy by the imaging score were truly benign.²¹

Specific ultrasound features that independently raise malignancy risk in Category 3 nodules include a “taller than wide” shape, microcalcifications, and irregular or microlobulated margins.²² Nodule size also plays a role: larger nodules (3 cm or more) with low- or intermediate-suspicion ultrasound features had independently increased malignancy risk in one study.²³ Male sex was also an independent risk factor in that same analysis. These findings suggest that even without molecular testing, clinicians can narrow the risk estimate substantially by integrating the imaging report and patient demographics with the cytology.

When Surgery Is the Right Choice

Surgery, typically a lobectomy (removing half the thyroid), is appropriate in several scenarios: when molecular testing comes back positive or suspicious, when a repeat biopsy again yields Category 3, when ultrasound features are highly concerning, or when the patient prefers a definitive answer. The extent of surgery depends on the clinical picture. Lobectomy is the default for a diagnostic operation because it removes the nodule while preserving half of the thyroid’s hormone-producing tissue. If the pathologist finds cancer on the lobectomy specimen and it meets certain criteria (larger size, aggressive features, or bilateral disease), completion thyroidectomy to remove the other lobe may follow.

A Canadian multi-institutional study found that larger nodule size was the strongest clinical predictor of a decision to proceed with surgery, while older patients were less likely to choose an operation.²⁴ Those findings make intuitive sense: a large nodule may already be causing compressive symptoms, and younger patients have a longer time horizon in which an undiagnosed cancer could progress.

Intraoperative frozen section, where the pathologist evaluates tissue during the surgery itself, has mixed utility for Category 3 nodules. A meta-analysis found that frozen section is excellent at confirming cancer when it calls the result malignant (specificity near 99%), but its sensitivity is only about 43%, meaning it misses more than half of cancers.²⁵ A single-center study reported a somewhat better sensitivity of 71%, though specificity remained high at 95%.²⁶ In practice, a frozen section that says “malignant” during surgery can justify completing the thyroidectomy then and there, avoiding a second operation. But a benign or deferred frozen section result should not be taken as definitive reassurance.

Active Surveillance After a Benign Molecular Test

For patients whose molecular test returns benign or negative, active surveillance with periodic ultrasound is now the standard recommendation. One study following 162 nodules with benign or negative molecular results over a median of about three years found only one malignancy during the surveillance period, a minimally invasive Hürthle cell carcinoma that was caught because the nodule grew on ultrasound. That translated to an overall false-negative rate of 0.6%.²⁷ Fourteen of the 176 nodules in that cohort went to immediate surgery anyway (patient or physician preference), and none of those were malignant.

Surveillance typically involves a follow-up ultrasound at 12 to 24 months, then at increasing intervals if the nodule remains stable. Growth alone does not necessarily mean cancer, but significant growth, development of new suspicious features, or changes in the patient’s symptoms (voice changes, difficulty swallowing, palpable firmness) should prompt re-evaluation with another biopsy. The reassurance from a negative molecular test is strong but not absolute, and a small number of patients will still end up needing surgery during follow-up.

The Reproducibility Problem

One of the underappreciated challenges with Category 3 is that pathologists do not always agree on what belongs in this category. The cytological assessment involves subjective judgment, and interobserver variation is well documented, especially for AUS.²⁸ What one pathologist calls AUS, another might label benign or bump up to Category 4. This variability partly explains why published malignancy rates for Category 3 range so widely across institutions: the category captures different populations of nodules depending on the threshold each pathology department applies.

This is not a flaw that patients can fix, but it is worth knowing about. If your Category 3 result comes from a high-volume thyroid center where pathologists evaluate hundreds of these samples per year, the classification is likely more reliable than one from a lab that sees thyroid specimens infrequently. Some patients seek a second pathology opinion, and while guidelines do not mandate this, it is a reasonable step when the management decision is on the fence. Molecular testing partly sidesteps this problem by looking at the nodule’s genetic profile rather than relying on a visual assessment of the cells.

Shared Decision-Making and Patient Experience

Because Category 3 lives in genuine diagnostic uncertainty, how the treatment discussion is handled matters more than in straightforward cases. Guidelines consistently recommend shared decision-making, where the physician lays out the options and the patient’s values help determine the path. One study of surgical consultations for indeterminate thyroid nodules found that patients generally perceived high levels of shared decision-making involvement, but about 13% still reported clinically significant decisional conflict, meaning they felt uncertain about whether they made the right choice.²⁹ Greater patient involvement in the decision correlated with less conflict afterward.

Anxiety is a real part of the Category 3 experience. Hearing that your biopsy result is “indeterminate” can be more stressful than hearing a clear benign or malignant diagnosis, precisely because the uncertainty is the diagnosis. Some patients find that molecular testing, even when it is not strictly necessary for clinical management, provides psychological relief by converting an ambiguous result into a clearer one. Others prefer the certainty of surgery. Neither preference is wrong, and a physician who pushes one path without exploring what matters to you is shortchanging the conversation.

How Malignancy Rates Vary Across Institutions

The Bethesda system’s average malignancy rate for Category 3 sits at about 22%, but individual institutional experiences scatter widely around that number. One single-center study found a rate of about 26% in 97 operated Category 3 nodules, with roughly three-quarters ultimately proving benign.³⁰ Another institution reported about 9% papillary carcinoma among all its Category 3 cases that went to surgery.³¹ A Saudi study of 193 operated Bethesda 3 nodules found nearly half were malignant, though there was no significant difference in malignancy rate among the subcategories of atypia in that particular population.³²

This scatter is partly due to the reproducibility issues discussed earlier, partly due to referral patterns (surgical centers inherently see more suspicious cases), and partly due to real population-level differences in thyroid cancer epidemiology. The takeaway for patients is that a Category 3 result does not carry a fixed cancer probability. The number your clinician quotes should ideally reflect your institution’s own data, your ultrasound findings, and your atypia subtype, not just the Bethesda system’s published average.