Evaluation and management of thyroid nodules with indeterminate cytology in adults

INTRODUCTION — When thyroid nodule fine-needle aspiration (FNA) cytologic results show atypia of undetermined significance (AUS, Bethesda III) or follicular neoplasm (Bethesda IV), the results are often called indeterminate. The risk of malignancy with these cytologic classifications ranges from approximately 10 to 40 percent [1]. Before the introduction of molecular testing, the majority of patients with a cytologic result showing AUS (confirmed on repeat aspiration) or follicular neoplasm had diagnostic thyroid surgery. However, most patients had surgery for what was ultimately confirmed to be benign disease. Improvement in the assessment of indeterminate FNA results with molecular testing allows for better risk stratification and reduces the need for diagnostic thyroid surgery.

The evaluation and management of thyroid nodules with indeterminate cytology in adults will be reviewed here. The diagnostic approach to thyroid nodules in general, including initial evaluation and selection of nodules for FNA, as well as the management of thyroid nodules with benign, suspicious, or malignant cytology, are reviewed separately. (See "Diagnostic approach to and treatment of thyroid nodules".)

FNA CYTOLOGY CLASSIFICATION SCHEME — The National Cancer Institute Thyroid Fine-Needle Aspiration State of the Science Conference ("Bethesda Conference") suggests the following cytologic classification scheme (table 1) [1]:

●I. Nondiagnostic

●II. Benign

●III. Atypia of undetermined significance (AUS)

●IV. Follicular neoplasm

●V. Suspicious for malignancy

●VI. Malignant

The terms used by different cytopathologists to describe follicular thyroid nodules vary (table 2). When cytologic results show AUS or follicular neoplasm, the results are often called indeterminate [1]. It is essential that clinicians interpreting these reports be familiar with the terminology used by their cytopathologist. A second review of indeterminate biopsies (AUS or follicular neoplasm) by an outside expert may be warranted, especially at low-volume centers. (See "Atlas of thyroid cytopathology" and "Thyroid biopsy", section on 'Diagnostic categories'.)

AUS — This category (Bethesda III) is subdivided into nuclear atypia and other atypia. Other atypia incudes architectural atypia (mixed macro- and microfollicular lesions where the proportion of macro- and microfollicles is similar), oncocytic atypia (lesions with extensive oncocytic [Hürthle cell] change but insufficient findings to be considered an oncocytic neoplasm), and lymphocytic atypia, among others. AUS with architectural atypia was previously referred to as follicular lesion with undetermined significance. (See "Atlas of thyroid cytopathology", section on 'Atypia of undetermined significance (AUS)'.)

The risk of malignancy with this classification (based on surgical resection data) ranges from 13 to 30 percent (table 1) [1]. AUS with nuclear atypia has a higher risk of malignancy compared with architectural atypia or oncocytic atypia [2,3]. In a meta-analysis, the overall malignancy rate was 29 percent for Bethesda III; however, it was 46 percent for nuclear atypia, 22 percent for architectural atypia, and 13 percent for oncocytic atypia [3].

Nuclear atypia is commonly seen in hyperplastic nodular goiter, papillary cancer, follicular variant papillary cancer, and noninvasive follicular thyroid neoplasm with papillary-like nuclear features (NIFTP) (table 2). Architectural atypia is commonly seen in nodular goiter (including autonomous nodules), follicular neoplasms, follicular cancer, follicular variant papillary cancer, and NIFTP. Oncocytic atypia is seen in degenerating nodular Hashimoto's thyroiditis, oncocytic neoplasms, and oncocytic cancer.

Follicular neoplasm — The cytologic features of this category (Bethesda IV) include cell crowding and/or microfollicle formation without nuclear features of papillary thyroid cancer. Postsurgical evaluation reveals a malignancy rate of approximately 23 to 34 percent [1]. Follicular neoplasms may be benign follicular adenomas (including autonomous nodules), follicular cancers, follicular variant papillary cancers, follicular neoplasms with oncocytic features (formerly Hürthle cell neoplasm), or NIFTP.

EVALUATION AND MANAGEMENT — Our approach outlined below (algorithm 1) is based on molecular testing (where available) and is largely consistent with the American Thyroid Association (ATA) guidelines [4].

The response to thyroxine (T4)-suppressive therapy should not be used to select patients for surgery or observation. Previously, the response of a nodule to T4-suppressive therapy was used to select nodules that required excision; nodules that shrank were assumed to be benign, and those that failed to shrink were excised to exclude malignancy. However, this approach is not very specific, since only 17 to 25 percent of nodules regress when T4 is administered (see "Thyroid hormone suppressive therapy for thyroid nodules and benign goiter"). Furthermore, 13 percent of subsequently proven papillary cancers in one series decreased in size during T4 treatment [5].

Reflex molecular testing versus repeat FNA with molecular testing — For patients with indeterminate cytology (Bethesda III, AUS [nuclear or other atypia] or Bethesda IV, follicular neoplasm), the approach to management varies with institutional practices and availability of reflex molecular testing. Unless a sample for molecular testing was saved at the time of the initial biopsy (ie, reflex molecular testing for AUS or follicular neoplasm cytology), a repeat fine-needle aspiration (FNA) with a sample for molecular testing (if available) after a 6- to 12-week interval (or earlier) is indicated in most patients with indeterminate cytology. Alternatively, mutational analysis on the original sample may be available by scraping a sample from the cytology slides. (See 'Sample from initial FNA not available' below.)

There are some data to suggest that repeat FNA with molecular testing only if repeat FNA persistently shows indeterminate cytology is better than reflex molecular testing on initial FNA samples that show indeterminate cytology [6]. A study of 363 nodules with Bethesda III and IV cytology compared the initial reflex strategy with the strategy of repeat biopsy with molecular testing [6]. A repeat biopsy was done for all 363 nodules, and molecular testing (Afirma GEC) was sent regardless of the repeat cytology reading. The assumption was made that the repeat sample was obtained from the same nodular area as the initial sample. Among the 61 nodules with repeat FNA cytology showing Bethesda I and II classification and molecular testing showing suspicious results, there were four low-risk cancers and three noninvasive follicular thyroid neoplasm with papillary-like nuclear features (NIFTPs; 1.9 percent of the total) that would have been missed with the repeat FNA strategy; however, a diagnostic lobectomy was avoided in 42 patients.

Sample from initial FNA available for molecular analysis — Many centers collect an extra fine-needle aspiration (FNA) sample at the time of the initial biopsy to be used for molecular testing should the cytology be read as indeterminate (AUS or follicular neoplasm).

Three of the widely available molecular testing approaches include mutational analysis (ThyroSeq v3), mRNA genomic expression (Afirma genomic sequencing classifier [GSC]), and a combination of microRNA (miRNA) gene expression and mutational analysis (ThyraMIR/ThyGenX). (See 'Molecular markers' below.)

Molecular pattern — If molecular testing is obtained (either at the initial or subsequent FNA), further management depends on the results of molecular testing (algorithm 1).

●Benign molecular pattern – For patients with indeterminate FNA cytology who have a benign molecular pattern (no mutations on mutational analysis [ThyroSeq v3] or a benign GSC, or a benign miRNA gene expression classifier and mutational analysis result), we suggest observation rather than diagnostic lobectomy. The negative predictive value for malignancy with ThyroSeq v3 and Afirma GSC is 96 to 97 percent, which seems to be sufficiently high to warrant observation over diagnostic surgery [7-10]. Repeat ultrasound is performed in 12 to 24 months to assess stability. However, the decision to observe a patient with a benign genetic profile should be reassessed as more data become available.

●Suspicious pattern – Most patients with a suspicious pattern require thyroid surgery. Several mutations (eg, BRAF, TERT, RET/PTC, RAS) are strongly associated with thyroid cancer. For example, the risk of malignancy with a BRAF mutation approaches 100 percent, while the risk of malignancy with various RAS mutations varies from 40 to 72 percent [8,11]. The choice of lobectomy versus total thyroidectomy may depend on tumor size, the specific mutation, patient preference, and the presence of nodes or contralateral nodules on ultrasound, as well as the availability of a high-volume thyroid surgeon. (See "Differentiated thyroid cancer: Surgical treatment", section on 'Choice of procedure'.)

For patients with a suspicious pattern using the GSC, the positive predictive value is 47 percent [7]. For these patients, we suggest a diagnostic lobectomy; however, a total thyroidectomy may be preferred depending upon the clinical characteristics and patient preference as noted immediately above.

In the setting of oncocytic atypia (AUS, Bethesda III) or follicular neoplasm with oncocytic features (previously called Hürthle cell neoplasm, Bethesda IV), the current Afirma GSC reports a 41 percent false-positive rate for oncocytic lesions [7], while ThyroSeq v3 has a 38 percent false-positive rate [8], both based on small numbers of nodules. Thus, patients with oncocytic atypia (versus nuclear or architectural atypia) continue to have a high rate of diagnostic lobectomy for benign disease, despite the use of molecular testing. (See 'Performance with oncocytic (Hürthle cell) lesions' below.)

Sample from initial FNA not available — If a sample for molecular testing was not obtained at the time of the initial FNA, the FNA is repeated in most patients after a 6- to 12-week interval (or earlier), and an extra sample is collected for molecular testing (if available) during the repeat aspiration (algorithm 1) [12,13]. If the thyroid-stimulating hormone (TSH) is subnormal (indicating overt or subclinical hyperthyroidism) and the nodule shows architectural atypia (AUS or follicular neoplasm), the possibility that the nodule is hyperfunctioning is increased, and thyroid scintigraphy should be obtained (if not previously performed) prior to repeating the FNA. Some experts also perform thyroid scintigraphy if the TSH is in the lower end of the normal range (eg, <1 mU/L), as thyroid hormone production from some autonomous nodules may suppress TSH only within the lower portion of the normal range (see 'Thyroid scintigraphy' below). Since hyperfunctioning nodules rarely are cancer, a nodule that is hyperfunctioning on radioiodine imaging does not require repeat (or even initial) FNA. Repeat FNA should be performed in patients with nonfunctioning nodules.

Repeat FNA with or without molecular testing should not replace clinical judgement as repeat FNA or molecular testing may not be needed to inform surgical decisions. For example, a diagnostic lobectomy is prudent in a young patient with a growing, large (eg, ≥4 cm) nodule with an indeterminate cytology. In a study of 266 consecutive AUS nodules, the overall malignancy rate was 24.7 percent, but for nodules over 4 cm, the malignancy rate was 67 percent [14].

Repeat FNA benign — If the repeat aspirate is benign, molecular testing is not indicated. Repeat ultrasound is performed in 12 to 24 months to assess stability.

Repeat FNA indeterminate — For patients with repeat aspirates showing AUS or follicular neoplasm, additional evaluation may include molecular testing, thyroid scintigraphy (for AUS with architectural atypia or follicular neoplasm, if not already performed), or diagnostic lobectomy.

For oncocytic (Hürthle cell) lesions, there is a high false-positive rate with molecular testing (see 'Performance with oncocytic (Hürthle cell) lesions' below). In this setting, the cost of molecular testing as well as the clinical circumstances should be considered when deciding whether to proceed with molecular testing or referral for a diagnostic lobectomy.

●Molecular testing available – If repeat aspirates continue to show AUS or follicular neoplasm, the extra FNA sample collected at the time of the repeat aspirate can be used for molecular testing. Further management then depends on the results of molecular testing. (See 'Molecular pattern' above.)

●Molecular testing unavailable – If molecular testing is unavailable and repeat aspirates continue to show AUS (nuclear, architectural, oncocytic atypia) or follicular neoplasm, we suggest diagnostic surgery (typically thyroid lobectomy). If not already performed, assess the need for thyroid scintigraphy before proceeding to diagnostic surgery (algorithm 1). Autonomous (hyperfunctioning) nodules are rarely cancerous, and diagnostic surgery would not be indicated. (See 'Thyroid scintigraphy' below.)

For patients with AUS and only mild architectural atypia, ie, greater than 50 percent macrofollicular fragments (flat sheets with uniform, noncrowded cells) without suspicious sonographic features and in whom there are no risk factors for thyroid cancer (childhood head and neck radiation, family history), an alternative option is monitoring. We typically obtain a follow-up thyroid ultrasound in 12 months.

For patients who have had diagnostic lobectomy, the absence of capsular or vascular invasion (on surgical histology) is classified as a benign follicular adenoma, or NIFTP (in the absence of papillary architecture), and no further treatment is required. For patients whose surgical histology shows follicular thyroid cancer or follicular variant papillary thyroid cancer, completion thyroidectomy may be necessary in selected cases (eg, when radioiodine treatment is indicated due to the size or invasiveness of the tumor or because of the presence of metastatic disease). (See "Differentiated thyroid cancer: Surgical treatment", section on 'Choice of procedure'.)

DIAGNOSTIC TOOLS

Molecular markers — Previously, the majority of patients with an indeterminate cytologic result (AUS or follicular neoplasm), confirmed on repeat aspiration, had diagnostic thyroid surgery (usually lobectomy). However, most patients (75 to 95 percent) had surgery for what was ultimately confirmed to be benign disease. Improvement in the assessment of indeterminate fine-needle aspiration (FNA) results with molecular testing allows better risk stratification and reduces the need for diagnostic thyroid surgery.

The routine use of molecular testing for indeterminate cytology may also reduce the number of patients who will require a completion thyroidectomy (to facilitate radioiodine administration), as a lobectomy may still be chosen if the nodule has a mutation associated with a low risk of malignancy [15]. A total thyroidectomy may be more appropriate for a large nodule, especially with concerning ultrasound features and especially if the nodule has one or more mutations associated with more aggressive histology.

Available tests — There are three approaches to the molecular characterization of FNA aspirates that are widely commercially available in the United States [16,17]:

●Identification of particular molecular markers of malignancy, such as BRAF and RAS mutational status

●Use of high-density genomic data for molecular classification (a genomic sequencing classifier [GSC])

●Use of an FNA-trained microRNA (miRNA) classifier combined with molecular markers of malignancy

These tests assess the risk of malignancy or noninvasive follicular thyroid neoplasm with papillary-like nuclear features (NIFTP, formerly called noninvasive follicular variant of papillary thyroid cancer but subsequently reclassified as a nonmalignant or possibly premalignant variant) (see "Papillary thyroid cancer: Clinical features and prognosis", section on 'Variant forms'). Since surgery is required to diagnose NIFTP, there is no change in the use of these tests for selecting patients for surgery. However, the positive predictive values noted below represent the risk of malignancy or NIFTP.

●Mutational analysis – Mutational analysis identifies molecular markers of malignancy. Assays that test for a larger number of genetic alterations have superior negative predictive values and are therefore useful to enable patients with benign nodules to avoid surgery. Earlier versions of this assay [18-20] have been replaced by ThyroSeq v3, which tests for mutations, insertions/deletions, fusions, gene expression, and copy number variations in 112 thyroid-related genes. In a multicenter validation study of 257 indeterminate thyroid nodules, sensitivity was 94 percent, specificity 82 percent, negative predictive value 97 percent, and positive predictive value 66 percent [8]. In another validation study of this assay, 238 surgically removed tissue samples were used as a training set and 175 indeterminate FNAs were used as a validation set [21]. The sensitivity and specificity were 94 and 89 percent, respectively, in the training set and 98 and 82 percent in the validation set, with an accuracy of 91 percent. The sensitivity and specificity for oncocytic (Hürthle cell) lesions was 92.9 and 69.3 percent, respectively. Samples for ThyroSeq v3 are usually collected at the time of the initial or follow-up biopsy, but mutational analysis has an 80 percent success rate for samples obtained by scraping cytology slides when the cellular material has 200 to 300 cells [22].

●mRNA genomic sequencing classifier

•The first version of the Afirma test was called a gene expression classifier (GEC) [9,10,23]. In one analysis, the use of the test in patients with indeterminate thyroid cytology would save over 60 percent of patients from diagnostic thyroid surgery, which would result in an overall lower cost of care [24]. The second improved version of the Afirma test is called a genomic sequencing classifier (GSC), which adds a cascade of classifiers utilizing several thousand genes to better discriminate oncocytic (Hürthle cell) neoplasms from non-neoplastic oncocytic lesions, as well as classifiers that identify medullary thyroid cancer, parathyroid lesions, BRAF V600E mutations, and RET/PTC fusions.

The validation study for the GSC was based on the same samples used for validation of the gene expression classifier (adequate sample was available for 191 of 210 lesions) [7]. The overall sensitivity and specificity was 91 and 68 percent, respectively, with negative and positive predictive values of 96 and 47 percent, respectively, assuming a 24 percent prevalence of malignancy. The improved sensitivity and specificity for oncocytic lesions was 88.9 and 58.8 percent, respectively. (See 'Performance with oncocytic (Hürthle cell) lesions' below.)

•Afirma's Xpression Atlas utilizes RNA sequencing and is an add-on test, available for GSC-suspicious Bethesda III/IV nodules and Bethesda V/VI nodules, which assesses 593 genes, 905 variants, and 235 fusions [25,26].

●miRNA gene expression combined with mutational analysis – The miRNA classifier is part of a multiplatform test based on the expression level of 11 miRNAs involved in growth promotion or growth suppression, RNA fusions, and mutational analysis of 10 genes (ThyraMIR and ThyGenX). Version two (MPTXv2) utilized the same validation set as the original classifier [27,28]: 178 indeterminate nodules with histologic confirmation, 54 of which were cancer or NIFTP. Sensitivity was 98 percent, specificity 86 percent, positive predictive value (PPV) 75 percent, and negative predictive value (NPV) 99 percent [29]. The improved accuracy was primarily due to a benign reading for 18 RAS-mutated samples with negative miRNA (comprising 49 percent of the RAS-mutated samples), all of which had benign histology. MPTXv2 is able to use scrapings from previously obtained cytology slides [30].

The validation studies for all these methods involve very small numbers of malignant nodules [31]. It is also notable that ThyroSeq v3 recommends surgical excision for nodules with RAS mutations, while MPTXv2 classifies almost half of the RAS-mutated nodules negative on the basis of miRNA expression [29]. Before utilizing these expensive tests, one should also consider the sonographic characteristics and the size of the nodule, the degree of patient concern, and the availability of follow-up imaging, as well as the patient's candidacy for surgery.

One study [32] assessed the added benefit of using ThyroSeq or Afirma to assess the risk of malignancy compared with the American Thyroid Association (ATA) malignancy risk [4] based on the ultrasound appearance of the nodule. For ATA low- and intermediate-risk nodules (based on ultrasound appearance) the preoperative malignancy risk was 21 to 24 percent, and the histologically proven malignancy rate was 56 to 67 percent for nodules with a positive molecular test, while for ATA high-risk nodules, the malignancy risk was 78 to 80 percent, with the histologically proven malignancy rate of 80 to 88 percent for nodules with a positive molecular test. These data suggest that cytologically indeterminate nodules with ATA high suspicion for malignancy may not require molecular testing to make a recommendation for surgery.

The negative and positive predictive values noted above are highly dependent on the prevalence of thyroid cancer in any clinician's cytologically indeterminate nodule population, with the negative predictive value decreasing as the cancer prevalence increases. Thus, the negative predictive value of a benign test is more reliable the lower the risk of malignancy in patients with indeterminate nodules, and it is optimal when the prevalence is <30 percent. It would be helpful if the prevalence of malignancy in indeterminate nodules were known for the institution in which the patient is being evaluated, but this information is rarely available. The decision to observe a patient with a benign profile using ultrasound surveillance should be reassessed as more data regarding the reliability of these tests become available. (See 'Evaluation and management' above.)

Performance with oncocytic (Hürthle cell) lesions — In the setting of oncocytic atypia (AUS, Bethesda III) or follicular neoplasm with oncocytic features (oncocytic neoplasm, previously called Hürthle cell neoplasm, Bethesda IV), the earlier versions of the molecular tests were suboptimal. For example, the original Afirma GEC (no longer marketed) was frequently suspicious in oncocytic neoplasms (63 percent in one study), despite a low rate of malignancy (14 percent) [33]. The current Afirma GSC is improved (sensitivity and specificity for oncocytic lesions 88.9 and 58.8 percent, respectively) but still reports a 41 percent false-positive rate for oncocytic lesions [7], while ThyroSeq v3 has a 38 percent false-positive rate [8], both based on small numbers of nodules. ThyroSeq v3 has a sensitivity and specificity for oncocytic lesions of 92.9 and 69.3 percent, respectively.

Comparison of ThyroSeq v3 with Afirma GSC — ThyroSeq v3 and Afirma GSC appear to perform similarly. In a randomized trial evaluating the performance of these two molecular testing methods, the method was switched every other month at a single institution [34]. Adequate data were available for 313 indeterminate nodules. There was no difference in the benign cell rate (53 versus 61 percent), avoiding diagnostic lobectomy (51 versus 49 percent), sensitivity (100 versus 97 percent), specificity (80 versus 85 percent), PPV (53 versus 63 percent), and NPV (100 and 91 percent) in the Afirma GSC and ThyroSeq v3, respectively. There was also no difference in these parameters for the subset of patients with oncocytic lesions.

Unintended consequences of molecular testing — Studies have suggested that cytopathologists are using the Bethesda III (AUS) classification more frequently than intended by the Bethesda conference. For example, at one center, prior to the introduction of molecular testing, Bethesda III was assigned to 9 percent of 393 FNA samples, while after the introduction of molecular testing, Bethesda III was assigned to 21 percent of 380 FNA samples. As a result, the cost of care increased. Notably, the overall rate of thyroid cancer was unchanged at 48 percent after the introduction of molecular testing [35].

A possible additional explanation is that fewer benign nodules are being biopsied with the introduction of Thyroid Imaging, Reporting and Data System (TIRADS). In one center after the introduction of TIRADS, the rate of benign cytology readings fell from 50 to 19 percent between 2014 and 2021, while the rate of Bethesda III readings increased from 21 to 52 percent [36]. (See "Diagnostic approach to and treatment of thyroid nodules", section on 'Sonographic criteria for FNA'.)

Thyroid scintigraphy — Thyroid scintigraphy is used to determine the functional status of a nodule. A subnormal serum TSH, indicating overt or subclinical hyperthyroidism, increases the possibility that a thyroid nodule is hyperfunctioning. However, thyroid hormone production from some autonomous nodules may suppress TSH only within the lower portion of the normal range (eg, <1 mU/L). Scintigraphy may be informative in such patients as well, especially if prior TSH levels were subnormal, or when the results of an FNA suggest AUS with architectural atypia or a follicular neoplasm. In a study of 1029 thyroid nodules over 10 mm with normal TSH, 5 percent appeared hyperfunctioning on [99mTc] pertechnetate scintigraphy, and 65 percent of these met ATA ultrasound criteria for FNA [37]. However, this study may slightly overestimate the percentage of true benign autonomous nodules since 5 percent of cancers will trap pertechnetate, but not radioiodine. (See "Diagnostic approach to and treatment of thyroid nodules", section on 'Thyroid scintigraphy'.)

●Nonfunctioning nodule – Repeat FNA in approximately 6 to 12 weeks should be performed in patients with nonfunctioning nodules (algorithm 1). An extra sample should be collected for molecular testing (if available) at the time of repeat FNA. (See 'Sample from initial FNA not available' above.)

●Autonomous nodule – Since hyperfunctioning (autonomous) nodules rarely are cancerous, a nodule that is hyperfunctioning on radioiodine imaging does not require repeat FNA or assessment with molecular markers. Patients with hyperfunctioning nodules are followed (repeat ultrasound in 12 to 24 months to assess stability), or if the patient is hyperthyroid, methimazole, radioiodine therapy, or surgery is advised. (See "Treatment of toxic adenoma and toxic multinodular goiter".)

●Indeterminate nodule – Because scintigraphy is two dimensional, its limitations result from the superimposition of abnormal nodular tissue and normally functioning thyroid tissue (image 1). Autonomous nodules that do not make sufficient thyroid hormone to suppress serum TSH concentrations may appear indeterminate on thyroid scintigraphy. Smaller nodules may also appear indeterminate on thyroid scintigraphy. They could represent either small, nonfunctioning nodules anterior or posterior to normally functioning thyroid tissue or autonomous nodules that do not produce sufficient thyroid hormone to suppress TSH (image 2).

These indeterminate nodules should not be referred to as warm or functioning, since the majority are, in fact, nonfunctioning nodules. Some UpToDate authors and editors obtain suppression scans in appropriate patients for further evaluation, whereas others proceed directly to molecular testing (see 'Molecular pattern' above) or diagnostic lobectomy if molecular testing is not available.

Suppression scanning is most commonly indicated in patients who have an indeterminate nodule on radionuclide scan and an FNA that was interpreted as an AUS with architectural atypia (microfollicles) or as a follicular neoplasm (microfollicular) since these cytologic findings may be associated with autonomous nodules. (See "Atlas of thyroid cytopathology", section on 'Follicular neoplasm'.)

During a suppression scan, patients are given thyroid hormone in a dose sufficient to suppress TSH secretion, and a second scan is done once TSH suppression is documented, which may occur in one to two weeks or may require several dose adjustments. Suppression can usually be accomplished by administering T4 (levothyroxine; 2 mcg/kg for 10 days) [38]. Uptake of radioiodine will be low or undetectable in nonautonomous tissue but persist in autonomous tissue (image 3). If the region of nonsuppressible uptake corresponds to the sonographic or palpable nodule, then the nodule can be assumed to be autonomous and benign [39]. Patients with autonomously functioning nodules may become hyperthyroid with even small doses of T4 and should be warned of hyperthyroid symptoms [40]. In such patients, the dose needed to suppress TSH may be considerably less than 2 mcg/kg. This test should not be done in older patients or in those with angina or cardiac arrhythmias.

FDG-PET scan — Fluorodeoxyglucose positron emission tomography (FDG-PET) is not recommended for the diagnostic work-up of indeterminate nodules. However, there are a growing number of trials evaluating the utility of FDG-PET, usually in combination with computed tomography (CT), to distinguish benign from malignant nodules with indeterminate cytology [41-45]. In various reports, sensitivity and specificity range from 77 to 100 percent and 33 to 64 percent, respectively [45]. In one meta-analysis, PET negativity was able to exclude the diagnosis of cancer if the lesion was >1.5 cm in diameter [46]. Until additional data are available to determine if PET-negative nodules with cytology consistent with a follicular neoplasm can be safely observed, we do not recommend ordering a PET scan to assess follicular neoplasms.

SOCIETY GUIDELINE LINKS — Links to society and government-sponsored guidelines from selected countries and regions around the world are provided separately. (See "Society guideline links: Thyroid nodules and cancer".)

SUMMARY AND RECOMMENDATIONS

●AUS or follicular neoplasm – When thyroid nodule fine-needle aspiration (FNA) cytologic results show atypia of undetermined significance (AUS) or follicular neoplasm, the results are often called indeterminate. The risk of malignancy with these cytologic classifications ranges from 13 to 30 percent and 23 to 34 percent, respectively (table 1). (See 'AUS' above and 'Follicular neoplasm' above.)

●Approach to evaluation and management – For patients with follicular neoplasm or AUS (nuclear, architectural, oncocytic, or other atypia), the approach varies with institutional practices and availability of molecular testing (algorithm 1). (See 'Evaluation and management' above.)

Unless a sample for molecular testing was saved at the time of the initial biopsy, a repeat FNA with a sample for molecular testing (if available) after a 6- to 12-week interval (or earlier) is indicated in most patients with indeterminate cytology. It also may be possible to do mutational analysis on a sample obtained by scraping the cytology slide from the initial biopsy, which eliminates the need for a repeat FNA.

If the TSH is subnormal (indicating overt or subclinical hyperthyroidism) and the nodule shows AUS with architectural atypia or follicular neoplasm, the possibility that the nodule is hyperfunctioning is increased, and thyroid scintigraphy should be obtained (if not previously performed) prior to repeating the FNA. Some experts also perform thyroid scintigraphy if the TSH is in the lower end of the normal range (eg, <1 mU/L), as thyroid hormone production from some autonomous nodules may suppress TSH only within the lower portion of the normal range. Since hyperfunctioning nodules rarely are cancer, a nodule that is hyperfunctioning on radioiodine imaging does not require repeat (or even initial) FNA. Repeat FNA should be performed in patients with nonfunctioning nodules. (See 'Sample from initial FNA not available' above and 'Thyroid scintigraphy' above.)

•If repeat aspirates are read as benign, molecular testing is not needed. (See 'Repeat FNA benign' above.)

•If repeat aspirates continue to show AUS or follicular neoplasm, the extra FNA sample collected at the time of the repeat aspirate can be used for molecular testing (if available). (See 'Repeat FNA indeterminate' above.)

●Molecular testing available – If molecular testing is available (either at initial or subsequent FNA), further management then depends on the results of molecular testing (algorithm 1).

•For patients with indeterminate FNA cytology who have a benign molecular pattern on molecular testing, we suggest observation (Grade 2C). However, the decision to observe a patient with a benign profile should be reassessed as more data become available. Diagnostic lobectomy remains an alternative option for patients with a benign molecular pattern, depending upon the nodule's clinical characteristics and patient preference. (See 'Molecular pattern' above.)

•For patients with a suspicious molecular pattern using these techniques, diagnostic lobectomy or total thyroidectomy is indicated in most patients. (See 'Molecular markers' above and 'Molecular pattern' above.)

●Molecular testing unavailable – If molecular testing is unavailable and repeat aspirates continue to show follicular neoplasm or AUS, we suggest diagnostic surgery (typically thyroid lobectomy) (Grade 2C). For patients with AUS and only mild architectural atypia, ie, greater than 50 percent macrofollicular fragments (flat sheets with uniform, noncrowded cells) without suspicious sonographic features and in whom there are no risk factors for thyroid cancer (childhood head and neck radiation, family history), monitoring is an alternative option. We typically obtain a follow-up thyroid ultrasound in 12 months. (See 'Repeat FNA indeterminate' above.)

●Diagnostic lobectomy findings – For patients who have had a diagnostic lobectomy, the absence of capsular or vascular invasion (on surgical histology) is classified as a benign follicular adenoma, or noninvasive follicular thyroid neoplasm with papillary-like nuclear features (NIFTP) in the absence of papillary architecture, and no further treatment is required. For patients whose surgical histology shows follicular thyroid cancer (or follicular variant papillary thyroid cancer), completion thyroidectomy may be necessary in selected patients who require radioiodine. (See "Differentiated thyroid cancer: Surgical treatment", section on 'Choice of procedure'.)