European Thyroid Association Guidelines regarding Thyroid Nodule Molecular Fine-Needle Aspiration Cytology Diagnostics

Introduction/Background

Fine-needle aspiration (FNA) has high sensitivity and specificity in distinguishing benign from malignant thyroid lesions [1]. However, in 2–16% of cases cytology is not diagnostic [2], i.e., the material is insufficient for diagnosis and an FNA repetition is requested. In 5–20% of cases, moreover, it is not possible to discriminate between benign and malignant nodules because of an indeterminate cytology, including follicular proliferation and atypia of undetermined significance (AUS) [3]. As a minority (about 20%) [4] of indeterminate thyroid nodules has been proved to be malignant at final histology, a high proportion of these patients undergo diagnostic rather than curative thyroid surgery.

Reporting of FNA cytology has improved in the past 10 years with the introduction of classification schemes in order to standardize terminology, to facilitate communication among cytopathologists, endocrinologists and surgeons, and to provide the malignancy risk for specific diagnostic categories. The Bethesda System, a 6-diagnostic-category system, is at present the most used reporting system for thyroid FNA cytology. The cate­gories are as follows: (I) non-diagnostic/unsatisfactory; (II) benign; (III) atypia of undetermined significance/follicular lesion of undetermined significance (AUS/FLUS); (IV) follicular neoplasm/suspicious for follicular neoplasm (FN/SFN), a category that also encompasses the diagnosis of Hürthle cell neoplasm/suspicious for Hürthle cell neoplasm; (V) suspicious for malignancy, and (VI) malignant [5]. In a meta-analysis of 25,445 thyroid FNA samples reported from 8 studies using The Bethesda System for Reporting Thyroid Cytology, 9.6% of all samples were diagnosed as AUS/FLUS, 10.1% were diagnosed as FN/SFN, and 2.7% were diagnosed as suspicious for malignancy, with an average cancer risk of 15.9, 26.1, and 75.2%, respectively [6]. Excluding suspicious lesions, which have such a high malignancy risk level that a surgical strategy is called for, both the AUS/FLUS and FN/SFN, also named indeterminate cytology, have a non-negligible cancer risk. Therefore, diagnostic surgery appears to be frequently necessary even if the malignancy risk is not so high that it indicates definitive cancer with certainty.

Other classification systems for thyroid FNA used in Europe are the Italian consensus by the Italian Society for Anatomic Pathology and Cytology jointly with the Italian Division of the International Academy of Pathology (SIAPEC-IAP) [7] and the Royal College of Pathologist [8] classification. Table 1 shows the comparison of the diagnostic categories of these classification systems. The main difference resides in the Bethesda class IV/RCP Thy3f and the Italian TIR3B. The latter category includes both follicular and non-follicular patterned lesions instead of a pure follicular patterned group.

Table 1.

Comparison of the diagnostic categories of the classification systems

View Table

The need to improve the diagnostic performance of FNA cytology in the setting of indeterminate results has led to the search for diagnostic biomarkers, and several recently introduced molecular tests [9], such as gene expression classifier (GEC) [10], gene mutational panel [11], targeted next-generation sequencing (tNGS) [12], and the combination of mutation detection and miRNA expression [13], have been reported as promising in ruling in or out a substantial risk of cancer. Of note, the major limit to the use of molecular testing in thyroid FNA may stem from a clinicohistopathological finding: it has recently been suggested that the encapsulated, non-invasive follicular variant of papillary thyroid carcinoma (PTC) should be reclassified as a benign tumor. If this reclassification is unanimously accepted, it will lead to a dramatic reduction of the risk of malignancy for AUS/FLUS, FN/SFN, or indeterminate FNA cytologies [14] and consequently also a dramatic reduction of the risk of malignancy for associated mutations like RAS.

A specific setting among follicular lesions is represented by nodules with oncocytic (Hürthle cell) features. The clinical significance of oncocytic changes in thyroid tumors is controversial, and a more aggressive behavior of Hurthle cell carcinomas when compared with the respective usual variants is still being debated [15]. Cytomorphologic criteria alone, moreover, are not helpful in distinguishing between benign and malignant Hürthle cell neoplasms. Finally, a potential setting for molecular tests with diagnostic purpose could be the non-diagnostic material. The rate of non-diagnostic results that should not exceed 10% of cases demonstrated a percentage of malignancy estimated between 5 and 15% and represents an indirect measure of the expertise and quality of FNA practice [4]. A molecular test resolving non-diagnostic FNA cytologies has been regarded as an important goal for future research.

Cytological specimens represent a useful and valuable source of cellular material for the evaluation of various molecular diagnostic tests that contribute to optimal patient care. This source suffers, however, from a large variability of sampling and processing. Cytopathology laboratories utilize numerous different methods to obtain and process cytological specimens. Therefore, the development and careful validation of molecular assays for these different cytological samples, such as direct smears, cell blocks, and liquid-based preparations, are important.

The thyroid FNA procedure should be performed under ultrasound guidance and thyroid cells are aspirated using a 23-, 25-, or 27-gauge needle for cytological evaluation. Direct cytology smears are both air dried or fixed with alcoholic solution or spray preparation. It is important to note that this preparatory technique is performed without formalin fixation. Formalin fixation leads to the crosslinking of nucleic acids and proteins, fragmentation of isolated nucleic acid material, and the possibility of sequence alterations. This point needs to be considered when other sources of samples such as cell block or cytology preparation with preserving liquid are used and specific protocols have to be applied in order to obtain high-quality nucleic acid isolation [16] Different types of specimen obtained from thyroid FNA procedure may be used for molecular analysis: (a) cells collected from FNA during the procedure [11], (b) cells retrieved from FNA-stained smears [1], (c) cells retrieved from air-dried slides [17], and (d) paraffin sections from cell blocks. All these samples have been shown to be suitable for point mutation analysis as well as for detecting chromosome rearrangements [17]. The different assays should be optimized for small size specimens and validated for the different cytology specimens. Furthermore, standardization of the performance characteristics and appropriate calibration as well as of analytic interpretation before clinical interpretation is necessary [18].

Nucleic acid recovering from direct smears offers relevant advantages: first, it permits to check the cells before nucleic acid isolation, allowing the analysis of exactly those cells that are cytologically indeterminate and the enrichment of the pathological population [16, 17]. This is not possible if a dedicated pass of FNA is washed in a tube containing nucleic acid preservative solution. Tumor enrichment mitigates the error based on the possibility that tumors may be heterogeneous with respect to a specific molecular alteration. Second, molecular tests from direct smears are performed on the already available diagnostic material without additional reaspiration of the nodule, patient discomfort, and the risk of inadequate sampling due to recent intranodular hemorrhage [16, 17].

Molecular analysis for thyroid FNA should be restricted to centers that have a low percentage of inadequate samples, an appropriate, but not too high, percentage of AUS/FLUS diagnoses, and a good track record for cytological-surgical concordance and malignancy rates. Cytological and molecular interpretation should be integrated. All molecular testing intended for clinical use must be performed only in certified clinical laboratories after appropriate validation of at least analytical sensitivity and specificity of mutation detection. Analytical sensitivity should be correlated with clinical sensitivity, so that the sensitivity is not pushed to extremely high levels, as this would lead to false-positive mutation detection.

Methods

The European Thyroid Association initiated this task force with the overall aim to review the relevant medical evidence and provide clinical recommendations on the use of molecular diagnostics in the evaluation of indeterminate cytology thyroid nodules (including the respective Bethesda categories of AUS/FLUS or FN, or SFN, or equivalent). Task force members were selected by the Task Force Chair, Ralf Paschke, based on their expertise in relevant fields, including endocrinology, molecular oncology, epidemiology, and endocrine surgery. This task force received no funding from the European Thyroid Association, industry, or any other organizations, and it is editorially independent from the European Thyroid Association. The competing interests of all task force members were collected by the Task Force Chair, prior to writing the manuscript and updated prior to submission for publication of this document (reported at the end of this manuscript). The clinical questions for these guidelines were developed by the Task Force Chair and modified by the Task Force members, such that there was final agreement by the group on the relevant topics for review.

What Is the Value of Mutation Panel Assessment (Including as a Minimum BRAF, NRAS, HRAS, KRAS, PAX8/PPARG, RET/PTC) in the Diagnostic Assessment of an Indeterminate Cytology Thyroid Nodule?

The most frequent genetic alterations detected in PTC and follicular thyroid carcinoma (FTC) are BRAF and RAS mutations and RET/PTC and PAX8/PPARG rearrangements [20, 21]. TERT promoter mutations and TP53 mutations are less frequent in well-differentiated (PTC and FTC) carcinomas than in less-differentiated carcinomas, but their detection in thyroid FNA should also be considered in the appropriate clinicocytopathological setting [22, 23].

The search for mutations of BRAF, TERT promoter, and TP53 is recommended in cases classified in Bethesda class V as well as in the subgroups of Bethesda class III in which there are focal features suggestive of PTC. Besides helping to clarify whether or not one is dealing with a malignant tumor, the detection of BRAF (and, probably, also of TERT promoter and TP53 mutations) may guide the type of surgery, leading to total thyroidectomy if the clinicopathological setting is appropriate [21, 24, 25].

For non-diagnostic samples (Bethesda class I) in most cases, the problem stems either from a lack of expertise in the center that performed the defective procedure or from particular characteristics of the nodule [26]. In any of the aforementioned cases, the best approach is to perform a second FNA at another institution with FNA experience. Molecular testing may help in selected cases [27].

In Bethesda class II cytological specimens (benign lesion) it is not helpful (and may even be confusing) to look for any mutations and/or rearrangements in FNA material except for rare cases when there is a strong clinical suspicion for cancer but a discordant FNA cytology result [28].

The reported sensitivities, specificities, negative (NPVs), and positive predictive values (PPVs) for studies investigating panels detecting BRAF, NRAS, HRAS, KRAS, PAX8/PPARG, and RET/PTC in AUS/FLUS, FN/SFN, or indeterminate FNA cytologies range from 18 to 100%, from 82 to 100%, from 56 to 100%, and from 19 to 100%, respectively (Table 2). Varying risks of malignancies and different rates of FTCs or PTCs for the investigated samples or knowledge of the pathologist about the molecular results are important variables that profoundly influence these statistics. Moreover, especially the risk of malignancy for RAS-positive FNA results varies dramatically between these studies, i.e. from 19 to 85% [11, 16, 17, 27]. Therefore, the impact of the detection of NRAS, HRAS, KRAS, PAX8/PPARG, and RET/PTC in AUS/FLUS, FN/SFN, or indeterminate FNA cytologies has to be determined in each specific setting.

Table 2.

7-mutation panel reports (modified from Nishino, 2016 [56])

View Table

The detection of this panel of mutations has been performed in fresh FNA material, sometimes obtained by extra FNA passes and in air-dried, stained smears or liquid cytology material. Detecting these mutations in material obtained from air-dried, stained smears or liquid cytology material has obvious advantages such as no need for obtaining extra FNA material, storing extra FNA material, and most importantly performing molecular diagnostics on exactly the same sample as the cytologist used for his morphologic diagnosis [16, 17].

Should a Microarray GEC Test Be Routinely Used on an Indeterminate (AUS/FLUS or FN/SFN) Cytology Thyroid Nodule to Exclude a Thyroid Malignancy in Lieu of Diagnostic Thyroid Surgery or Continued Close Follow-Up?

The evaluation of mRNA expression in indeterminate cytology thyroid nodules using microarray technology is referred to as GEC testing. The core data of the studies are summarized in Table 3. The largest study describing analytic validity (diagnostic performance) of the GEC test was reported by Alexander et al. [10], who conducted a prospective, multicenter study in the USA; this study was funded by industry (Veracyte). The authors performed central, blinded, gold-standard, histologic confirmation of all specimens included in the final analysis [10]. In this study, 577 cytologically indeterminate aspirates were obtained, 413 of which had histopathological confirmation (which was required for inclusion in the study), and of these 389 were allocated to GEC testing (excluding those used in training or analytic verification) [10]. The cytological diagnoses included in the indeterminate category for this study included AUS/FLUS, FN/SFN, Hürthle cell neoplasm, or suspicious for malignancy, with the AUS/FLUS category comprising almost half of the specimens in the final analysis (49%, 129/265). Of the 389 samples allocated to GEC testing, 379 were processed for a GEC result, 328 had a valid GEC result, and 312 had a valid GEC result with gold-standard histopathological confirmation [10]. However, ultimately 265 were included in the final analysis [10]. Of these 265 non-consecutively obtained indeterminate cytology thyroid nodules, 85 (32%) were confirmed to be histopathologically malignant. In the 265 indeterminate cytology nodules, the sensitivity of the GEC test was 92% (95% confidence interval [CI], 84–97%, 78/85) and the specificity was 52% (95% CI, 44–59%, 93/180) [10]. It is important to note that all the patients included in the final analysis underwent thyroid surgery (as a pre-established inclusion criterion); therefore, long-term outcomes of patients avoiding thyroid surgery on the basis of a negative GEC test cannot be extrapolated from this study.

Table 3.

Evidence of molecular FNA diagnostics can reduce diagnostic lobectomies in individuals with indeterminate thyroid cytology who do not have thyroid cancer and clinical utility of GEC

View Table

The diagnostic performance of the GEC test was subsequently reported in 4 respective studies from the USA [29-32]. In each of these studies, treating physicians were aware of the GEC result at the time of surgical decision-making and not all patients who underwent a GEC test had gold-standard histopathological confirmation of the diagnosis; furthermore, there was no reported blinding of the histopathologists who ascertained the gold-standard diagnosis (when performed) [29-32]. McIver et al. [29] reported in a prospective register study that patients at the Mayo Clinic (Rochester and Jacksonville sites) were offered the GEC test if the FNA was performed on site and the cytology was reported as SFN, suspicious for Hürthle cell neoplasm, or AUS/FLUS; approximately 83% (30/36) of the specimens included in the final analysis were SFN or suspicious for Hürthle cell neoplasm. In this study, criteria for offering GEC testing to patients with indeterminate cytology thyroid nodules included willingness of the patient to defer surgery, lack of compressive symptoms, and absence of higher-risk features (including head and neck radiation history, family history of thyroid cancer, prior history of thyroid cancer, or worrisome imaging characteristics) [29]. A GEC “suspicious” result prompted a recommendation for surgery in this study, whereas a GEC “benign” result prompted a recommendation for repeat ultrasound in 3–6 months and thereafter yearly for 5 years [29]. McIver et al. [29] reported that of 105 indeterminate cytology FNAs obtained during the study, 72 were subjected to GEC testing, 60 had GEC results, 12 had insufficient samples for a GEC result, and of those 36 underwent surgery with histopathological confirmation of the diagnosis. Of the 36 samples with GEC results and histopathological confirmation of the diagnosis, the prevalence of malignancy was 17% (6/36). The authors of this study estimated that the sensitivity of the GEC test was 83% (5/6) and the specificity was 10% (3/30) in histopathologically confirmed specimens. The median follow-up for GEC “benign” nodules in this study was only 9.5 months at the time of publication [29].

Harrell and Bimston [30] also reported the results of a single-center retrospective study from Florida, in which all histopathological specimens were read centrally by an experienced pathologist, who was not blinded to the GEC result. These authors reported on 58 indeterminate cytology specimens, including the categories of either AUS/FLUS or FN (the latter being follicular or Hürthle cell predominant). The prevalence of malignancy in those with histopathological confirmation in this study was 51% (18/35). In this study, the GEC test sensitivity was 94% and specificity was 24%. Harrell et al. [30] indicated that the patients with a GEC “benign” result were advised to return for a yearly follow-up, including neck imaging, but no follow-up data were reported.

In another single-center retrospective, unblinded study, Lastra et al. [31] reported on the University of Pennsylvania experience, in which 132 cases of indeterminate cytology nodules were subject to GEC testing, including the categories of AUS/FLUS, FN, or FN with oncocytic features, with more than half of the cases being in the AUS/FLUS category (52%, 68/132). The prevalence of malignancy in the 50 cases in this study that had histopathological confirmation of diagnosis was 44% (22/50). In calculations using data from this study [31], the GEC test sensitivity was 100% (22/22) and specificity was 7% (2/28). No follow-up data were reported on patients who did not undergo surgery in this study.

Most recently, Marti et al. [32] reported a retrospective analysis of all indeterminate thyroid nodules evaluated with the GEC at 2 New York-based institutions, including Memorial Sloan Kettering Cancer Center (94 nodules) and Mount Sinai Beth Israel Hospital (71 nodules). In this study, the GEC test was reported as “benign” in 38.3% of cases at Memorial Sloan Kettering Cancer Center and 52.1% of cases at Mount Sinai Beth Israel Hospital. The authors estimated that the PPVs of the GEC “suspicious” results were 57.1 and 14.3% at respective institutions, whereas the NPVs of a GEC “benign” result were estimated to be 86–92% at Memorial Sloan Kettering Cancer Center and 95–98% at Mount Sinai Beth Israel Hospital. The authors of this study stressed the importance of knowing the prevalence of malignancy in indeterminate thyroid nodules at respective institutions, as such information was considered critical in interpreting the GEC results.

In summary, based on the above studies that reported on sensitivity and specificity, GEC test sensitivity was reported to range from 83 to 100% and specificity from 7 to 52%, whereas the prevalence of malignancy in histopathologically confirmed study populations ranged from 17 to 51% [10, 29-31].

Long-term outcomes of patients undergoing GEC testing, including those who do not undergo thyroid surgery based on a negative GEC result, are important to be considered for clinical recommendations. Unfortunately, currently, such long-term outcome data are lacking, with available studies reporting median follow-up periods of less than 1 year in duration [10, 29].

What Is the Value of tNGS in the Diagnostic Assessment of an Indeterminate Cytology Thyroid Nodule?

tNGS sequencing is a promising method to examine multiple genes simultaneously with high sensitivity potentially achieving not only high PPVs but also high NPVs [33] and with low input of starting material. Preliminary results for thyroid FNAs have thus far only been published by 2 groups [12, 33, 34], and clinical validation results are not yet available. Data for lung and other tumors show that next generation sequencing is robust like Sanger sequencing in routine diagnostics and, in addition, is able to reveal mutations at low percentage and to screen the mutational status of different critical samples offering innovative diagnostic opportunities [35-37]. Furthermore, methodological problems like result interpretation (e.g., for unknown mutations), definition of cutoffs for mutation calling, and bioinformatics analysis need to be solved, and common standard operation procedures (SOPs) need to be defined.

tNGS analyzing larger mutation panels was used for the molecular diagnostics of limited numbers of FN/SFN, AUS/FLUS, and indeterminate FNA samples reporting sensitivities from 71 to 91%, specificities from 89 to 93%, NPVs from 92 to 97%, and PPVs from 63 to 83%. If reproduced in larger sample sets this tNGS approach has the potential to become a simultaneous rule-in and rule-out test if NPVs >95% can be confirmed.

Possible Surgical Consequences of Molecular FNA Diagnostic Results of Thyroid Nodules

The preoperative knowledge of a verified thyroid carcinoma will have an impact on the surgical strategy and therefore will considerably increase the success of the oncologic treatment.

The clinical setting may lead to two different basic diagnostic approaches in the decision-making for thyroid surgery due to suspicious nodes:

1. Rule out malignancy in order to avoid surgery

2. Verify malignancy preoperatively and determine the type of thyroid cancer to optimize surgical strategy and treatment success, while avoiding secondary interventions

In order to comply with the rule-in and rule-out strategies the necessary NPV for rule-out tests should be >95% according to the National Comprehensive Cancer Network guidelines [38], while the ideal PPV should be >95% for rule-in strategies leading to more radical resections (total thyroidectomy). However, apart from the indication for primary radical surgery a lower risk of malignancy prompts the recommendation for diagnostic resection. Current guidelines recommend total or near total thyroidectomy or lobectomy in case of FNA results suspicious for malignancy (Bethesda class V), which is associated with a 60–75% risk of malignancy. Similarly lobectomy is recommended in case of Bethesda class IV with a risk of malignancy ranging from 15 to 30% [5].

What Is the Evidence that Molecular FNA Diagnostics Can Reduce Diagnostic Lobectomies in Individuals with Indeterminate Thyroid Cytology Who Do Not Have Thyroid Cancer?

In a recent meta-analysis [6], 22% of 25,448 FNAB specimens were classified in one of the indeterminate categories (“follicular lesion of undetermined significance (FLUS),” “follicular neoplasm (FN),” or “suspicious for malignancy”) according to the Bethesda Cytologic Classification System. 6,362 (25%) underwent surgery confirming malignancy among indeterminate FNAB specimens in 31%. The cytology categories AUS/FLUS, FN, and suspicious for malignancy corresponded to ascending malignancy rates of 16, 26, and 75%, respectively. If additional diagnostics could rule out malignancy this would potentially prevent 84, 74, and 25% of the interventions assuming that there is no other reason (e.g., compression symptoms, growing nodules, thyrotoxicosis) for surgery.

GEC (based on Affymetrix Human Exon 1.0 GeneChip) [10] aims to rule out malignancy in FNABs that were classified as either “follicular lesion of undetermined significance (FLUS)” or “follicular neoplasm (FN)” according to the Bethesda Cytologic Classification System. Five nonrandomized studies (4 prospective, 1 retrospective) reported on patients with intermediate thyroid cytology but benign classification based on the GEC. Based on the results of the GEC, thyroid surgery was prevented in 32–98% of patients. However, only 2 studies provided limited follow-up information on a subgroup of patients that did not undergo surgery (11 and 8 months). Moreover, the calculation of the NPV of the test was based on the assumption that all thyroid nodules classified as benign and that were not operated on did not include false negative results. Despite promising initial results generated by industry-funded studies in North America, the value of the GEC still has to be demonstrated in the routine clinical setting [10, 29, 39]. Marti et al. [32] evaluated the test performance in 2 different clinical centers. Patient populations of both centers demonstrated significantly different risks for thyroid cancer, which had a profound impact on the test performance as well as NPV and PPV. The authors demonstrated that PPV and NPV depend heavily on the prevalence of malignant disease in the population under study. In the studied centers GEC results had little or no impact on the decision for or against surgery. The GEC would be expected to provide the most useful information in a practice setting with a prevalence of malignancy in intermediate thyroid nodules with a risk of malignancy of 15–21%. However, in most practice settings the associated risk of malignancy differs from this range.

In summary, the percentage of individuals undergoing thyroid surgery after GEC testing has been reported to range from 38 to 60% with the majority of those preceded by a “suspicious” GEC result [10, 29-31]. In the only study reporting the rate of surgery in a concurrently followed cohort of individuals with indeterminate cytology who did not undergo GEC testing (likely in part due to not meeting pre-established institutional criteria for offering the GEC test), McIver et al. [29] reported that the rate of thyroid surgery was relatively comparable at 55%. However, Duick et al. [28] reported a historical thyroid surgery rate of 74% in patients with indeterminate cytology nodules. Some of the most frequently reported reasons for individuals with indeterminate cytology thyroid nodules and a GEC benign result to undergo surgery included the presence of compressive symptoms, higher risk of malignancy based on history or imaging characteristics, and personal preference. This leads to the question why the GEC was performed in these patients with other indications for surgery. Molecular diagnostics that rule out malignant disease may reduce the number of thyroid resections in the short-term follow-up if the risk of malignancy associated with indeterminate thyroid cytology ranges between 15 and 21% and the test results in an NPV are >95% in a defined patient population.

Mutation panels that screen for the most frequent genetic alterations in differentiated thyroid carcinoma including BRAF, NRAS, KRAS, RET/PTC, NTRK, and PAX8/PPARG are predominantly intended to identify thyroid cancer and are not able to rule out malignancy with an NPV >95%. Therefore, surgery has to be recommended despite the exclusion of these mutations [11, 16, 17, 28, 40]. A tNGS analysis of the above-mentioned genes plus additional genes (AKT, TP53, GNAS, PTEN, TERT, CTNNB1, PIK3CA, and TSHR) [12] included 143 FNA samples with indeterminate cytology. Among all FN/SFN nodules, the extended preoperative mutation detection was performed with 90% sensitivity (95% CI, 80–99%), 93% specificity (95% CI, 88–98%), a PPV of 83% (95% CI, 72–95%), an NPV of 96% (95% CI, 92–100%), and 92% accuracy (95% CI, 88–97%). While these results almost meet the criteria of an ideal test, 91 FNA samples in the study were analyzed retrospectively and the pathologist was aware of the histologic diagnosis. Therefore, the impact on the decision to recommend surgery and the surgical strategy cannot be estimated.

What Is the Evidence that Molecular FNA Diagnostics Can Increase the Number of Primary Total Thyroidectomies and Reduce the Number of 2-Stage Surgeries?

The reduction of the number of 2-stage surgeries is dependent on the preoperative diagnosis of thyroid cancer especially in cases with non-malignant or with indeterminate FNA cytology. Molecular markers that are almost exclusively associated with thyroid cancer qualify for the “diagnosis” of malignant disease in the absence of histologic proof. However, studies that report on the clinical impact do not exist.

Several studies investigated the use of mutation screening for predominant genetic alteration in PTC and FTC. Various genes were included, e.g. BRAF, RET/PTC, TRK, PAX8/PPARG, and RAS. Cantara et al. [40] reported the results of the analysis of 235 FNAs in 174 patients. All non-malignant FNAs (53 non-diagnostic, 41 indeterminate, 87 benign, and 54 suspicious for malignancy) were included. Indeterminate FNA samples revealed mutations in 7/41 cases [40]. Excluding the detection of RAS mutations in follicular adenomas the molecular genetic information would have a potential impact in 61 patients for all FNAs but only in 4/41 patients with indeterminate FNA. Moreover, the indication for surgery was determined prior to the inclusion into the study and not dependent on the molecular genetic analysis. However, the authors recommended primary thyroidectomy in case of mutation-positive FNA including RAS mutations. It is not reported how many patients would have undergone primary thyroidectomy for bilateral disease irrespective of the molecular genetic result or in how many cases the diagnosis could be established by frozen section.

The study of Beaudenon-Huibregtse et al. [41] subjected patients to molecular genetic analysis of BRAF, KRAS, HRAS, NRAS mutations, and PAX8/PPARG and RET/PTC rearrangements irrespective of the cytology result. They reported the identification of oncogenic mutations in 25/53 with indeterminate cytology, 4/26 with a benign cytology result, and 2/2 with non-diagnostic cytology. Potentially in 31 of 618 patients (546 with benign cytologies; 5%) 2-stage surgery or suboptimal surgery could have been avoided. However, the authors do not provide information on intraoperative findings or on how the surgical strategy was changed in the patients with mutation-positive nodules.

Similarly Moses et al. [42] and Eszlinger et al. [17] demonstrated that in about 11% of all FNAs that were classified as non-malignant (Bethesda classes I–V) the identification of oncogenic mutations could potentially alter the surgical strategy. Again in both studies, no information is provided on how the surgical strategy was changed in patients with mutation-positive nodules.

Nikiforov et al. [11] reported their results including 967 FNAs of 762 patients with AUS/FLUS, FN/SFN, and also 52 suspicious FNAs. Oncogenic mutations were present in 89% of FNAs. The additional mutational analysis of BRAF V600E, NRAS codon 61, HRAS codon 61, and KRAS codons 12/13 point mutations and RET/PTC1, RET/PTC3, and PAX8/PPARG rearrangements in the different cytology categories enabled the identification of thyroid malignancies and potentially a preoperative stratification of the surgical therapy between lobectomy and thyroidectomy in 38 of 513 FNAs including Bethesda classes IV and V. Again, information on the exact impact on the surgical strategy was not provided but primary thyroidectomy in case of oncogenic mutations is recommended.

The same group reported on the extended tNGS mutation panel including the additional genes AKT, TP53, GNAS, PTEN, TERT, CTNNB1, PIK3CA, and TSHR [12]. 143 FNA samples with indeterminate cytology were analyzed. The extended mutation panel achieved a PPV of 83% (95% CI, 72–95%) and an NPV of 96% (95% CI, 92–100%). 91 FNA samples in the study were analyzed retrospectively, and the pathologist was aware of the histologic diagnosis. Therefore, the impact on the decision to recommend surgery and the surgical strategy is unclear and has to be evaluated in a larger prospective trial.

All reviewed studies are based on mutational panels including at least 5 different genes that are associated with thyroid cancer. However, the risk of malignant disease varies for each mutation. While BRAF, RET/PTC, and PAX8/PPARG mutations predict thyroid malignancy with more than 95% probability, the risk of malignancy associated with RAS mutations varies between 19 and 87% [11, 16, 17, 43].

In summary, predominantly retrospective, non-randomized data indicate that potentially in 11–47% of patients (Table 2) with indeterminate FNAs, which also include suspicious cytologies in some studies, the surgical strategy may be changed from lobectomy or partial resection to primary thyroidectomy due to additional information on molecular genetic alteration. However, the potential impact is significantly influenced by the following:

Unpublished results of the PETS II study including 22,008 patients revealed that 78% of the patients underwent surgery without preoperative FNA leading to a carcinoma rate of 12% [45]. Similar results were reported by Wienhold et al. [46].

Detailed epidemiologic studies about the frequency of completion thyroidectomy do not exist but some studies report a frequency of completion thyroidectomy in about 3–16% of thyroid carcinomas [47, 48].

What Is the Evidence that a Strategy of Decision-Making on the Extent of Thyroidectomy according to Molecular FNA Diagnostics Reduces Complication Rates (Permanent Hypoparathyroidism, Recurrent Nerve Paresis, and Complications Caused by Secondary Anesthesia) in Individuals with Thyroid Nodules?

Secondary surgical interventions are associated with the risk of permanent hypoparathyroidism, recurrent nerve paresis, and risks associated with surgical procedures and anesthesia in general. The frequency of complications following completion thyroidectomy is dependent on the timing of the secondary intervention [47] as well as the type of the primary resection (bilateral resection, lobectomy).

Assuming that the FNA molecular diagnostic profile only leads to primary total thyroidectomy in patients with malignant disease and overtreatment can be avoided, the long-term health outcome would be improved in a subgroup of patients that otherwise might experience permanent complications (1–4% permanent recurrent nerve paresis, about 3% permanent hypoparathyroidism).

What Is the Evidence that a Strategy of Decision-Making on the Extent of Thyroidectomy according to Molecular FNA Diagnostic Profile Reduces the Local Recurrence Rate of Thyroid Carcinomas or Improves Long-Term Survival?

Unpublished results of the PETS II study including 22,008 patients with 2,577 thyroid carcinomas (12%) revealed that completion thyroidectomy within 6 months was necessary in 11% of carcinoma patients [45]. 16% of patients with thyroid carcinoma were operated on for recurrent disease. Predominantly, these recurrences are located within the thyroid bed or lymph nodes of the central compartment. Prophylactic central lymph node dissection is usually not performed following total thyroidectomy and postoperative diagnosis of thyroid cancer despite the fact that it might be indicated if the diagnosis was made prior to surgery.

Therefore, the following clinical situation may contribute to a possible improved health outcome:

Total thyroidectomy instead of hemithyroidectomy or other subtotal resection

Total thyroidectomy and central lymph node dissection instead of total thyroidectomy without central lymph node dissection

Total thyroidectomy for differentiated thyroid carcinoma has been recommended in many countries in recent guidelines. The identification of specific mutations such as BRAF V600E and TERT in thyroid nodules >1 cm, which are associated with malignant tumors in almost 100% of cases, justify a total thyroidectomy and possibly prophylactic central lymph node dissection. However, as outlined in recent revisions of the guidelines of the British Thyroid Association [49] and the American Thyroid Association [50] a more conservative resection is advocated for low-risk thyroid carcinomas.

The benefit of prophylactic central node dissection for PTC is continuously debated, and prospective randomized controlled trials comparing total thyroidectomy with and without prophylactic lymph node dissection are unfeasible. Metareviews of retrospective data imply a reduction of local recurrences by prophylactic central lymph node dissection of about 5% [51, 52].

In summary, there is minor evidence that long-term health outcome may be improved by FNA molecular diagnostics.

What Is the Evidence that Molecular FNA Diagnostics with a GEC Can Alter the Follow-Up for Thyroid Nodules?

GEC assigns a thyroid FNA to 1 of 2 categories, “benign” or “suspicious,” on the basis of an expression pattern of mRNA extracted from 1 or 2 dedicated FNA needle passes [10, 39]. Thus, it is designed as a rule-out test that recognizes benign tumors. Several subsequent papers [29, 30, 32] evaluated the clinical validity of GEC but their results were referred to the histopathological outcome of the surgery and not to the follow-up of non-operated patients who did not undergo surgery because their GEC result was benign. A short overview of their data is given in Table 3.

The National Comprehensive Cancer Network (NCCN) thyroid carcinoma guidelines, version 2.2013, recommends molecular testing of indeterminate thyroid lesions with the following limitation: “If molecular testing predicts a risk of malignancy comparable to the risk of malignancy seen with a benign FNA cytology (approximately 5% or less), consider observation” [38].

The careful review of all studies did not reveal analyses devoted to the follow-up of nodules classified “benign” by GEC. The question is even how long should the follow-up be. According to Nou et al. [53], who followed up 2,010 benign nodules for an average of 8.5 years (range 0.25–18), repeated thyroid nodule evaluation should be done 2–4 years after the initial benign FNA. 18 false-negative thyroid malignancies were diagnosed at a mean 4.5 years (range 0.3–10) after the initial benign biopsy. Noteworthy, no distant metastases were observed and all patients were alive at an average of 11 years following the FNA. According to Lee et al. [54] a long-term follow-up of thyroid nodules, initially benign by FNA, does not improve the cancer detection rate. A group of 738 patients with FNA-confirmed benign thyroid nodules were observed at the MD Anderson Center. 92 of them were operated on after initial ultrasonography, whereas the 646 ones remaining were subjected to further follow-up. Among them, 366 patients had at least 1 ultrasonography. 226 of them were monitored for a period shorter than 3 years (median 13 months, range 1–35), while 140 subjects had a long-term follow-up (>3 years, median 57 months, range 36–154). The number of ultrasonographies and repeated FNAs was significantly higher in the long-term group; however, surgical resection of the nodule as well as malignancy rate at final histopathological examination was similar between the 2 groups. Thus, stopping further follow-up after 3 years may be considered [54]. However, a longer follow-up of 8.1 years was proposed by Liel et al. [55]. Based on the data of Nou et al. [53] a 2-year follow-up should be sufficient.

In the multicenter clinical study, reported by Alexander et al. [10], of the 174 GEC benign tumors ultimately 11 patients were operated on, mostly because of personal preference. Histopathological examination revealed a benign thyroid tumor in 10 of them and PTC in 1 patient [10]. This case raises some concerns on the validity of a GEC benign outcome for follow-up of non-operated patients.

The above-mentioned studies by McIver et al. [29], Harrel and Bimston [30], and Marti et al. [32], devoted to GEC clinical utility, did not present follow-up data.

To date there have been no studies validating the use of GEC in European patients. It is noteworthy that the frequency of nodular goiter in Europe, especially in the central part, is much higher than that in the USA, an iodine-sufficient country. Thus, if the frequency of nodular goiter is higher, the relative prevalence of malignant tumors will be lower than that in the USA. It should be emphasized that the NPV depends on the prevalence and therefore an independent validation for moderately iodine-deficient areas is necessary.

What Is the Evidence that Molecular FNA Diagnostics by Analyzing a Panel of Somatic Mutations Can Alter the Follow-Up?

It is commonly accepted that positive rule-in tests constitute an indication for surgery due to a substantial risk of malignancy [9]. However, the important question arises of what to do when the FNA result is indeterminate and the panel of somatic mutations is positive only with respect to RAS mutation? Total thyroidectomy has been proposed for such cases due to a high risk of cancer [11, 43]. This is not supported by results of others who demonstrated that the presence of RAS mutation was associated with thyroid cancer in 19, 38, and 57% of cases while BRAF mutations were associated with cancer in 100% cancer samples [16, 27]. Even those who propose total thyroidectomy found that RAS mutations in benign thyroid adenomas do not require extensive surgery as for thyroid cancer [11]. No data regarding the follow-up of patients with these nodules have been published.

The analysis of a panel of somatic mutations together with FNA in follicular lesions increased the diagnostic sensitivity from 60 to 80% in comparison with FNA alone. The risk of malignancy increased to 71% for mutation-positive FN and decreased to 18% for mutation-negative follicular lesions when compared to a 28% risk of malignancy evaluated by FNA only [16]. Unfortunately, also in this study no data with reference to follow-up were given.

Conclusion

Recently, several molecular tests have been introduced into the clinical routine, such as GEC, seven-gene mutational panel, and tNGS. GEC recognizes benign lesions on the basis of an expression pattern of mRNA extracted from 1 or 2 dedicated FNA needle passes. GEC NPVs and PPVs are controversial in different studies and even the diagnostic performance needs to be clarified by the evaluation of the follow-up of patients who avoided surgery as a consequence of a GEC benign result. We recommend that in an individual with an indeterminate cytology a GEC test should not be routinely used to exclude malignancy.

Mutation panels intended to identify malignancies must include at least BRAF, RAS point mutations and RET/PTC, NTRK, PAX8/PPARG rearrangements. Several “homemade” methods comprising PCR with final Sanger sequencing and some commercial methods are available to screen for these alterations with the limitation that they cannot rule out malignancy with an NPV >95%. Therefore, surgery has to be recommended despite the exclusion of these mutations. Moreover, the role of RAS mutations is still unclear. While BRAF, RET/PTC, and PAX8/PPARG mutations predict thyroid malignancy with more than 95% probability, the risk of malignancy associated with RAS mutations is 57–87%. Thus, total thyroidectomy may not be recommended for RAS mutations due the lower PPV associated but lobectomy is preferred.

The use of tNGS with larger panels of genes offers an alternative method which is possibly able to achieve an NPV of 95% or more and a high PPV and sensitivity. However, issues such as result interpretation, cutoffs for mutation calling, bioinformatics analysis, and common SOPs need to be addressed. Therefore, the use of tNGS should currently be limited to research settings.

New molecules are emerging in thyroid cancer. In particular, investigations of the role of miRNAs in discriminating benign from malignant lesions in FNAC reported very preliminary results which need to be corroborated in larger studies with final histology as gold standard and adequate follow-up. Moreover, issues such as result interpretation, cutoffs, bioinformatics analysis, and common SOPs need to be addressed.

In conclusion, molecular tests may help clinicians to drive patient care and the surgical decision if the analysis is performed in specialized laboratories. These molecular tests require standardization of performance characteristics and appropriate calibration as well as analytic interpretation before clinical interpretation.

Disclosure Statement

The authors have no conflicts of interest to declare.