Yield and costs of molecular diagnostics on thyroid cytology slides in the Netherlands, adapting the Bethesda classification

Abstract Objective To evaluate our institutional experience with molecular diagnostics (MD) on thyroid cytology smears, evaluate the costs and describe MD guided clinical management of indeterminate Bethesda III/V thyroid nodules. Methods We performed a retrospective review of 164 Bethesda III or V thyroid cytopathology reports subjected to MD from 2013 to 2020, that altered Bethesda classification or management. MD consisted of mutation and gene fusion analysis by next‐generation sequencing (NGS) of morphologically analysed and selected cytological slides. Findings were modelled to nationwide data on Bethesda incidences from ‘the Dutch Pathology Registry’ PALGA, and costs were estimated. Results 82 of 164 cases received an upgrade in Bethesda class. Twenty cases changed from Bethesda III to IV/V, 62 from Bethesda III or V to VI, and 72 remained unaltered. We estimate net savings with implementing MD, by preventing 454 repeat cytology and 326 (diagnostic) hemithyroidectomies, to be at least 2 million Euro annually in the Netherlands. Per Bethesda III and V patient, net savings would be about 100 Euro and 4100 Euro, respectively. Conclusion NGS‐based MD on nucleic acids extracted directly from cytology slides is a feasible and cost saving tool for personalized management in indeterminate Bethesda III/V thyroid cytology. Based on the interpretation of our retrospective data, we assume that this approach results in less disease burden for the patient, reduced surgical interventions and complication risks, reduced sick leave, among others. Further evaluation of structural implementation of the presented approach in routine thyroid Bethesda III/V cytology in a prospective setting is warranted.


| INTRODUC TI ON
Thyroid nodules have a high prevalence, up to 60% of the general population. 1 However, only a minority (1.6-12%) of thyroid nodules is malignant. 2 As a consequence, many (in essence) unnecessary thyroid surgeries and repeat FNACs are performed due to indeterminate diagnosis.
Accrued insights in molecular tumorigenesis 7 allow for the (auxiliary) implementation of molecular pathology in certain clinical cases and to help direct their subsequent management, as is also suggested in guidelines and literature. 1,4 Consequently, the extent of molecular diagnostics (MD) applied in thyroid cytology has expanded by the years. Single BRAF V600E detection with promising cost-effectiveness studies 8,9 was expanded to cancer hotspot panels, to larger gene panels recently. [10][11][12][13][14] In Bethesda III or V cytology cases in particular, MD may help direct a change in management. In contrast, MD would, with rare exceptions, not cause an alteration in management in case of Bethesda IV, in the setting of current (Dutch) guidelines.
Molecular analysis of Bethesda IV cases may serve mainly to affirm a correct cytological classification of follicular lesions.
While promising, drawbacks of these commercial tests are their high costs 10 and often the requirement of additional aspirates besides the regular FNAC; however, there are some recent rapports on the use of cytology slides also. 20,21 The application of molecular analysis to cytology slides is a suitable, affordable method and has shown promising results. [21][22][23][24][25][26][27] During routine microscopic examination, representative areas of the regular cytology slides are marked by the pathologist. Then, nucleic acids for molecular testing can be isolated from these selected areas.
Altogether, this technique may prove as a cheaper 28 and a more feasible method in its routine implementation into conventional cytopathologic examination when compared to aforementioned commercial kits. Besides, the present approach may yield even more representative analysis results while requiring only very small amounts of nucleic acids from cells suspected for neoplasm that are carefully selected. 25 In contrast, the use of commercial kits on additional cytological aspirates does not allow for microscopic examination or selection of the cells that are used for molecular analysis.
In this study, we evaluated our molecular diagnostics approach applied to microsections of routine thyroid FNAC slides selected during routine microscopic examination by the pathologist. To this end, we retrospectively reviewed our institutional cases with cytological indeterminate thyroid nodules graded as Bethesda III or Bethesda V, that were molecular diagnostically analysed, over seven years. The molecular diagnostics consist of targeted DNA and RNA next-generation se- In addition, nationwide data were retrieved on Bethesda III to VI classifications of thyroid lesions. Reduced costs were estimated and modelled to national level. Ultimately, the presented method may serve as a feasible approach in both applicability (no additional material to routine cytology slides), cost-effectiveness and patient safety by circumventing, in essence unnecessary, invasive procedures of repeat FNACs or diagnostic hemithyroidectomies.

| Current evidence in the literature and interpretation of molecular diagnostic results
With recent advancements in the understanding of pathophysiology and molecular mechanisms, panels for molecular testing, but also their clinical implications, keep evolving. For instance, detection of BRAF V600E gene variants 6,29 or gene fusions involving ALK, [29][30][31][32] NTRK [33][34][35] and BRAF 31 upgrades the Bethesda class to a Bethesda VI, as these are likely associated with papillary thyroid cancer (PTC) or may occur in poorly differentiated thyroid cancer (PDTC), anaplastic thyroid cancer (ATC) or, occasionally, medullary thyroid cancer (MTC), and malignancy should be expected in these cases, see Figure 1. The detection of (MTC-associated) RET gene variants may upgrade the grade to Bethesda VI. RET fusions are likely associated with PTC or may occur in PDTC, ATC and MTC. 29 As gene fusions occur more frequently in paediatric thyroid carcinoma than in the adult population, multiple studies have reported RET/PTC fusions in PTC, 36 including the study of molecular diagnostics on thyroid nodules by Monaco et al reporting PTC for all detected RET/PTC fusions. 37 Also in our experience, no RET fusions have been detected using NGS that were related to benign lesions so far (expert opinion, present study).
Remarkably, as a diagnostic test RET fluorescence in situ hybridization is inferior due to substantial false-positive results. 38 In the cases harbouring the aforementioned gene alterations that suit Bethesda VI, if the lesion is evaluated as >1 cm, according to recommendations of current Dutch clinical guidelines, a total thyroidectomy may be considered. Thereby, two-staged surgery in malignant Bethesda III and Bethesda V cases, including the exposure to risks for complications (including bleeding, wound infection) twice, but also additional costs, second hospitalization and recovery burden, may be circumvented with single-staged surgery.
A (diagnostic) hemithyroidectomy may be considered (and modification of class to Bethesda IV, and/or occasionally (indicated with a *) Bethesda V), if variants BRAF non−V600E , 29 Indeed, these molecular alterations may yet occur in benign lesions, and RAS variants are reported to occur in advanced thyroid carcinomas promoting tumorigenesis 51 ; however, also NIFTP cases have been described harbouring concurrent EIF1AX and RAS variants. 44,52 TERT promoter variants, concurrent TERT and RAS variants or TERT and BRAF V600E gene variants are also reported to indicate a more aggressive nature with worse prognosis. [53][54][55][56][57] In case of GLIS translocations (modification to Bethesda V), a lobectomy is the appropriate treatment, for this translocation is highly specific for hyalinizing trabecular tumour (a benign lesion). 58,59 Ultimately, according to histopathologic assessment of the diagnostic hemithyroidectomy, a completing thyroidectomy may be per-

| Case selection, retrospective review and data analysis
Nationwide data of anonymized reports encompassing Bethesda III to VI thyroid cytopathology over 7 years (01-01-2013 through 31-12-2019) including all thyroid cytopathology/surgery records per patient F I G U R E 1 Proposed decision chart according to MD findings on thyroid cytopathology categorized as Bethesda III or V. The molecular alterations shown in the first box may modify the Bethesda grade to a Bethesda VI, followed by total thyroidectomy depending on diameter of the lesion. Please see main text for further details and clinical considerations. Please see main text for further details and clinical considerations. The molecular alterations shown the second box may, in order to acquire definite histopathological diagnosis, be followed by a (diagnostic) hemithyroidectomy. The molecular alterations shown the second box are mostly associated with follicular lesions and may occur in follicular adenoma/NIFTP/FVPTC/FTC; so would suit Bethesda IV. However, gene variants of TP53, TERT and TERT gene fusions (indicated with *) are not specific for a follicular lesion per se, and may be found in various types of thyroid lesions, and rather associated with a higher aggressive nature; so, on its own, may justify the modification a Bethesda III to a Bethesda V. GLIS gene fusions are associated with hyalinizing trabecular tumours and may modify the Bethesda class as Bethesda V. Molecular diagnostics include gene variant and gene fusion analysis of total nucleic acid using NGS with a custom Ampliseq Cancer Hotspot Panel and a custom Archer FusionPlex CTL Panel. N.B. GLIS and TERT gene fusions to be included in our updated custom panels, according to recent insights. **depending on diameter. BIII, Bethesda III; BV, Bethesda V; BVI, Bethesda VI; FNAC, fine-needle aspiration cytology; MD, molecular diagnostics; NGS, next-generation sequencing in PALGA: Dutch Pathology Registry were requested from and kindly provided by PALGA (The Nationwide Network and Registry of Histoand Cytopathology in the Netherlands). 60

| Selection of thyroid cytopathologic cells and molecular diagnostics
MD was performed in the ISO15189 accredited Molecular Diagnostics Unit of the Pathology department of the LUMC (Leiden University Medical Center). The molecular analysis procedure of on nucleic acids isolated from cytology smears has been described previously. 61 Briefly, cytologic smears are fixed in methanol and Giemsa stained. During routine microscopic examination by pathologists (HM), the area of interest, if discernible, on the inspected thyroid cytopathology smear classified as Bethesda III or V is precisely demarcated for tumour cell enrichment.
In separate tubes, slides are emerged in xylene until the coverslips detach. Slides are washed using ethanol washing steps (100%-70%-50%). After tissue rehydration, the demarcated cells are scraped off. Total nucleic acid was isolated using a fully automated DNA/RNA isolation system. 62 In most cases, a single smear, yielding between 2 and 10 ng of nucleic acids, was sufficient for molecular analysis using NGS.
Subsequently, MD results were interpreted by a registered molecular scientist in pathology (HM and TvW), and Bethesda class was

| Estimations of reduced costs by implementing MD and modelling to nationwide data
Patient data from our institution regarding Bethesda III and V with molecular analysis were used for making estimations of reduced costs by implementing MD and subsequently modelled to the nationwide data. The proposed decision chart in Figure 1, based on the current practice at our institution, was used for modelling of saved costs. Additionally, a success rate for MD in Bethesda III and in Bethesda V cases, respectively, based on the institutional data, was adjusted for in the estimations.  tively, www.opend isdata.nl) and productivity costs (€200 per FNAC and €1600 per surgery). These productivity costs were estimated assuming 67% labour participation with on average 27 h per week (www.CBS.nl), with one day sick leave for FNAC and two weeks for surgery. Also, the share of female patients, age distribution and inflation was taken into account hereby.

| Bethesda classification of thyroid nodules in the Netherlands
Estimations of Bethesda class incidences in the Netherlands, based on reported codes in PALGA from 2013 through 2019, were 7230 for Bethesda III and 2557 for Bethesda V (Table 1). For Bethesda III, the annual incidence increased through the years and ranged be-

| Patient characteristics
In total, 164 FNAC cases with Bethesda III or V were analysed using MD at our institution from 2013 through 2019 (Table S1)

| Molecular diagnostics results
MD results from the MD analysis are listed in Table 2. In 54 cases (33%) a BRAF V600E gene variant and in 25 cases (15%) a RAS variant was detected; 9 (5%) cases had other gene variants. Gene fusions were detected in 10 (6%) cases. In 63 (38%) of all MD tested cases, no gene alterations were found. In seven cases (4% of total; six cases (9%) of Bethesda III and one case (1%) of Bethesda V), the quality of the material was insufficient for genetic variant testing.
Gene fusion analysis was performed in 75 cases in total (46% of all cases with MD). Ten gene fusions were detected in 10 cases (13% of cases with (intended) fusion analysis), and no fusions were detected in 30 cases (40% of cases with (intended) fusion analysis); the quality of the material was insufficient for fusion analysis in 35 cases (47% of cases with (intended) fusion analysis).
As described earlier, the Bethesda classification could be altered based on the MD results. The frequencies of (un)altered Bethesda III or V cases upon MD are presented in Table 3. In retrospect, the Bethesda III that were upgraded upon MD to a Bethesda IV or V prevented repeat FNACs (20 of 65 cases, 31% within initial Bethesda III). Regarding Bethesda V cases that were reclassified as Bethesda IV upon MD, the appropriate management remained a diagnostic hemithyroidectomy (10 of 99 cases, 10% within initial Bethesda V). The Bethesda III cases that were reclassified as Bethesda VI prevented at least one repeat FNAC, an unnecessary diagnostic hemithyroidectomy and an unnecessary delay in treatment due to inappropriate surveillance in four cases (6% within initial Bethesda III). The shifts from a Bethesda V to a VI prevented an unnecessary diagnostic hemithyroidectomy in 58 cases (59% within initial Bethesda V).
The detected gene alterations with MD and the histopathologic diagnostic outcome are described in Table S2 cases, concerned microcarcinomas of which no FNAC was obtained from in three of these five cases; instead the obtained FNAC correlated to co-existent benign nodules in the thyroid. The other two of these five malignant 'unaltered Bethesda III' cases were found to harbour gene fusions on the resected material; however, fusion analysis had not been performed on the cytological material at the time.
The Bethesda III/V classes that were altered and unaltered after MD and corresponding histopathologic diagnoses are shown in Table S3.

| Modelled data and costs in Bethesda III/V thyroid cytology
When taking together the 164 Bethesda III/V cases that were tested with MD (also see Table 3), we could model rough numbers to nationwide incidences of Bethesda III cases and Bethesda V cases in

2019.
About 1179 unnecessary surgical interventions are done annually in the Netherlands (743 in Bethesda III and 436 in Bethesda V).

Bethesda IV 3907
Bethesda V 2557 Bethesda VI 1946 Compared to usual care without MD, a strategy with the implementation of MD (also see Figure 1) reduces the number of diagnostic and completing hemithyroidectomies (see Table S4).
In Bethesda III patients, the average number of surgical procedures is reduced from 0.77 procedures without MD (ie a diagnostic hemithyroidectomy in 55%, followed by a completing hemithyroidectomy in 22%) to 0.72 procedures with MD (ie a total thyroidectomy in 6% and a diagnostic hemithyroidectomy in 31%, of which half is followed by a completing hemithyroidectomy; a 91% success rate for the MD test has been adjusted for

| DISCUSS ION
Many unnecessary surgical interventions for diagnostic means, over one thousand annually in the Netherlands, are performed due to in-  sen. Yet, in the face of de-escalation, also in the latter case if <4 cm, a hemithyroidectomy may be chosen.
The costs of the commercial assays are high (ranging from about 3000 to 5000 USD). 19 Nevertheless, despite these high costs, molecular testing was still more cost-effective than diagnostic hemithyroidectomy with hypothetical modelling. 19 Other studies also show the applicability and validity of molecular analyses to cytology smears, including thyroid cytology. [21][22][23][24][25][26][27] The costs of MD as used in the present study on cytology smears are much lower (about 650 Euro). As in our study, material for NGS (as minimal as 2 ng of nucleic acids) is obtained from routine cytology smears upon dedicated morphological selection. Commercial kits mostly require additional aspirates for analysis, with no further morphological review of that material.
Limitations of this study include the retrospective nature of realworld data analyses and modelled estimates of reduced costs. The detection range of genetic alterations by the panels used through time has expanded. Also, MD use is increasingly applied and panels were updated over the years. Consequently, the detected alterations may be an underrepresentation of actual alterations present.
Furthermore, the proportion of unusable material for cDNA fusion analysis was much higher as compared to DNA gene variant analysis.
Consequently, gene fusions, if present in those cases, may not have been picked up. Also, the number of cases in which MD was performed during the reviewed timeframe was limited due to exclusion of cases that were included in another ongoing study at the time (EfFECTS study, www.effec ts-studie.nl). Nevertheless, the data give a good impression of overall advances that can be achieved by the routine implementation of MD in Bethesda III/V cases.

| CON CLUS ION
In this study, we used molecular profiling applied to cells selected from routine thyroid cytology smears graded as BIII or BV.
Preoperative diagnostic indices in nodules with indeterminate cytology may ultimately serve as an auxiliary tool in directing towards a single-, two-staged thyroidectomy or diagnostic hemithyroidectomy.
A more personalized treatment strategy can be followed using the MD guided diagnostic strategy. Thereby, patient safety is insured by circumventing invasive procedures, that is repeat FNAC or diagnostic hemithyroidectomy, if proven unnecessary by MD outcome. Also in its implementation on routinely obtained cytology smears without the need for separately obtained FNAC material, the presented method appeared to be feasible. Moreover, prevented (surgical) procedures showed cost-effectiveness of the presented approach.
The results of this study will pave the way for more research on its structural implementation in routine cytopathology in a prospective setting at larger scale.

ACK N OWLED G EM ENTS
This publication and the underlying study have been made possible partly due to the molecular analyses by Clinical Molecular Biologists

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available in the supplementary material of this article.