Performance of multigene testing in cytologically indeterminate thyroid nodules and molecular risk stratification

Study cohorts and sample collection

Patients with at least one thyroid nodule that was clinically diagnosed and confirmed by ultrasound were retrospectively enrolled and screened in our study (3,175 cases). Patients under 18 years old, who did not have a primary diagnosis, or who were missing the required information (2,343 cases) were excluded, as detailed in Fig. 1. Fine-needle aspiration (FNA) samples were collected under ultrasound guidance by radiologists using a 22-gauge needle. The aspirate was smeared on microscope slides and stained after being fixed with 95% ethanol for cytological examination. Categorization was conducted according to The Bethesda System for Reporting Thyroid Cytopathology (TBSRTC). A total of 529 FNA samples from cytologically indeterminate thyroid nodules (i.e., diagnosed by clinical pathologists as Bethesda III, IV or V) were collected to analyze the DNA mutations and RNA fusions via NGS. The clinical diagnosis of the nodules and their subsequent follow-up reports following definitive surgery, if performed (Table 1) were analyzed retrospectively for further investigation. This retrospective study was approved by the Ethics Committee of Jiangsu Integrated Traditional Chinese and Western Medicine Hospital (No. 2022-LWKY-042) and the Ethics Committee of Guangzhou KingMed Medical Laboratory Center (No. 2023005). Patient consent was waived due to the observational nature of the retrospective study and patient identities were kept anonymous (No. 2023005).

Molecular analysis

Sequencing libraries were generated using one-step multiplex PCR targeted amplicons. Briefly, the genomic DNA were isolated from the FNA samples of thyroid nodules using the QIAamp DNA Mini kit (Qiagen, Hilden, Germany) in accordance with the manufacturer’s specifications. The concentrations were determined by Qubit 3.0 Fluorometer (Thermo Fisher Scientific, Waltham, MA, USA). Sample DNA were mixed with KAPA2G Fast PCR (Roche, La Posay, France) and were amplified to detect driver gene variants, utilizing our multigene testing panel (Table S1) for the detection of 14 thyroid cancer-related genes (AKT1, BRAF, CTNNB1, EIF1AX, HRAS, KRAS, NRAS, PAX8, PIK3CA, PTEN, RET, TERT, THADA, and TP53) and 21 types of gene rearrangements occurring in thyroid cancer (ACBD5, AFAP1L2, ALK, ATG10, BRAF, CALM2, CCDC6, ERC1, ETV6, FLNC, FMNL2, KIAA1217, KIAA1594, KIF20B, NCOA4, NTRK3, PAK1, PAX8, PIBF1, PPAR γ, PXK, RALGAPA2, RET, SND1, and STRN). Superfluous primers were purified using Agencourt AMPure XP 60 mL kit (Beckman, Brea, CA, USA). The concentrations of barcoded PCR produced library were then measured and diluted to 100 pM. A total of 20 µL pooled amplicons were sequenced at the Ion Proton system (Thermo Fisher Scientific, Waltham, MA, USA). Local alignments of reads to the hg19 genome were performed via Bowtie2 (version 2.2.4) in the paired-end mode. SAM alignment files were converted to BAM files, sorted, and indexed using Samtools (version 0.1.19), following by a procession with Bam-read count and a customized written Perl script.

Test performance evaluation and false-positive/-negative (FP/FN) results analysis

The actual status of each diagnosis was determined: (1) by surgical pathology or (2) a benign molecular test without surgical pathology. The results of the benign molecular test are only considered truly benign when they are based on ultrasound and/or clinical characteristics. Test performance characteristics including sensitivity, specificity, positive predictive value (PPV), negative positive predictive value (NPV), and Kappa-value for consistency check were calculated at 95% confidence interval (CI) for the overall cytologically indeterminate specimens, as well as separately for nodules in Bethesda III, IV and V following the established method (Altman & Bland, 1994). Based on observed sensitivity and specificity, hypothetical PPV and NPV curves were modeled over the entire range of possible disease prevalence (0–100%), allowing the observed and the anticipated PPV and NPV to be compared. In order to preclude misdiagnosis due to false-negative test results, histologic slides of nodules that had negative molecular results on FNA cytology but had been diagnosed as malignant on resection, were blindly reviewed by another pathologist.

Molecular risk stratification

Thyroid nodules with molecular aberrations were categorized into three molecular risk groups (MRGs): a high-risk group (TERT or TP53 alterations), a BRAF-like group (mainly BRAF^{V 600E}) and a RAS-like group (HRAS, KRAS, NRAS, and others). The classification of all variants into this three-category system is shown in Table S2. A sample would be considered positive for molecular pathology if some genetic alteration was detected; the sample would be negative if no variants were detected in our panel. The MRG results would be assessed for the relevance with clinicopathologic diagnosis and further analyzed for the aggressive features of tumor-node-metastasis (TNM) staging. Nodules with concurrent molecular alterations were isolated and their distribution in MRGs and their TNM classification were assessed individually. Slides of surgical histopathology for special cases (nodules suspected as false-negative or false-positive, presenting concurrent molecular alterations, and other infrequent genetic alterations) were retrospectively analyzed by another pathologist.

Statistical analyses

Descriptive summaries of the histopathology and molecular pathology reports were shown with counts and percentages. Pearson chi-square tests were used to compare the categorical variables of age, gender, Bethesda category, and thyroid surgery between benign/suspicious molecular result groups. Statistical analysis was conducted utilizing SPSS software (version 26.0) with P-value < 0.05 being considered as statistically significant.

Study cohort and demographics

A total of 529 cases with cytologically indeterminate thyroid FNA results were identified between February 2021 to December 2021 at Jiangsu Province Hospital on Integration of Chinese and Western medicine in China. These cases were analyzed for molecular testing via NGS. All data in the cohort are accessible in Table S2. The patient characteristics and thyroid nodules are shown in Table 1. A total of 391 (79.88%) patients with suspicious nodules as determined by molecular tests were female and 403 (81.41%) of the cases for the cohort were below 55 years old at time of diagnosis. The majority of thyroid nodules (410 cases, 77.50%) were interpreted as Bethesda V, and 14.56% (77 cases), and 7.94% (42 cases) were Bethesda III and IV, respectively. Overall, the results of our thyroid NGS panel determined benign nodules (34 cases, 6.43%) and suspicious nodules (495 cases, 93.57%). The accuracy of the test was checked to evaluate the performance of the multigene panel. Furthermore, 574 molecular aberrations with 89 samples containing coexisting nodules were further investigated for their associated histopathologic cancer types and TNM staging based on molecular risk stratification (Fig. 1).

Test performance

The test performance for sensitivity, specificity, NPV, PPV, and consistency of the cytologic groups of thyroid nodules is presented in Table 2. Since sensitivity and specificity are intrinsic characteristics for each test, the NPV and PPV depended on the prevalence of the disease in the screened population (Smith, 2012). Predicated PPV and NPV with 95% CI were calculated based on the observed sensitivity and specificity; hypothetical PPV and NPV curves were modeled over the entire range of possible disease prevalence (0–100%), as depicted in Fig. 2.

Among the positive test samples, 490 cases (98.99%) were malignant and five cases (1.01%) were benign on surgery resection, while 23 cases (67.65%) were benign and 11 cases (32.35%) were cancerous within the negative test samples. Overall, in the cohort of cytologically indeterminate nodules had a cancer prevalence of 94.71%; multigene testing presented a sensitivity of 97.80% (95% CI [96.11–98.90]%), a specificity of 82.14% (95% CI [63.11–93.94]%), a PPV of 98.99% (95% CI [97.79–99.54]%) and an NPV of 67.65% (95% CI [53.20–79.36]%) (Table 2). The tumor prevalence in our cohort was obviously higher than clinical actuality due to the difficulty of identifying “true” benign nodules without surgery, risks of malignancy (with NIFTP) published in the TBSRTC diagnostic categories were referenced as the expected prevalence range (marked with green rectangles in Fig. 2). Moreover, the consistency of the multigene testing performed better in Bethesda III and IV (Kappa = 0.748 and 0.778, respectively, ranked as substantial agreement) than V (Kappa = 0.599, ranked as moderate agreement) specimens. When three categories of cytologically intermediate nodules were pooled, the consistency check for molecular test with histopathologic diagnosis manifested substantial agreement (Kappa = 0.726) (Table 2).

Summary of molecular alterations

All the somatic single nucleotide variants (SSNVs), insertions, deletions, duplications, and fusions detected in cytological indeterminate thyroid nodules were summarized in Table 3. The most observed gene aberration in this cohort was BRAF^{V 600E} (446 cases, 82.59%), followed by RET aberrations (20 fusions and two mutants, 4.07%), NRAS mutants (22 cases, 4.07%), KRAS mutants (17 cases, 3.15%), and other alterations. Notably, CCDC6-RET fusion was detected in 17 cases, which may be related to 16 patients with papillary thyroid carcinoma or microcarcinoma (PTC/PTMC) and one benign case with concurrent molecular alteration. Moreover, some molecular aberrations may co-exist with others rather than appear alone; these include BRAF rearrangements and deletions, PTEN deletions, TERT^Q302R, and mutants of AKT1, EIF1AX, TP53, and PIK3CA. There are some molecular alterations that had never been reported in solid tumors before, including BRAF-PXK fusion, FLNC-BRAF fusion, BRAF^{N486_T491delinsS}, PAX8^C147R, PTEN^D252Rfs∗42, and TERT^Q302R, as well as the EIF1AX c.338-2A>T splice site mutation and the KIAA1217-RET fusion in thyroid cancers.

False-negative and false-positive molecular test results with histopathologic diagnosis

The clinicopathologic characteristics of 11 false-negative (Case 1 to Case 11) and five false-positive (Case 12 to Case 16) cases are listed in Table 4. In total, 11 malignant nodules tested negative in our cohort, including one in Bethesda III, one in Bethesda IV, and nine in Bethesda V. These patients underwent surgery due to other suspicious features, subject to their preference. All of the thyroid nodules proved to be PTC/PTMC after diagnostic surgery, in which the presence of extrathyroidal invasion (T4a_staging) was observed in three cases. Four cases revealed locoregional nodal metastases (N1a/b_staging). Two nodules that tested negative before surgery (Case 1 and Case 4) were verified positive with the BRAF^{V 600E} mutant on resection, indicating that molecular misdiagnosis may be the result of a limited tumor volume, sampling technique, or preservation method.

The false-negative result of Case 5 may be specious, since the value of FNA-Tg was significantly high (above 500.000 ng/mL) in the puncture fluid of the left lymph nodule and FNAC manifested chronic lymphocytic thyroiditis. However, none of the molecular aberrations were detected in either the thyroid and lymph nodules at diagnosis. Similarly, the false-positive result of Case 12 may require further study for a potentially malignant transformation, given that the final histological diagnosis was benign follicular adenoma (FA), three cm in diameter. This nodule was suspected as being follicular neoplasia by FNA cytology and the BRAF^{V 600E} mutant was detectable before surgery.

Molecular risk stratification (concurrent molecular alterations excluded)

The cancer subtypes in specific molecular alterations and MRGs are summarized in Table 5. After the exclusion of concurrent molecular alterations, only two nodules with TP53 alterations were divided into the high-risk group, one (TP53^P82Rfs∗66) of which was follicular thyroid carcinoma (FTC) with partial differentiation. The other (TP53^{G262−S269del}) was determined to be PTMC on resection, however it was initially cytologically categorized as Bethesda IV (suspicious for a follicular neoplasia). The prevalence of PTC/PTMC in the BRAF-like group reached 100% in 426 tumors, while 22 malignant nodules in the RAS-like group were composed of 14 PTC/PTMC (63.64%) and eight FTC or follicular patterned tumors (follicular variant PTC/PTMC, FV-PTC/FV-PTMC) (36.36%).

TNM staging and aggressive features (concurrent molecular alterations excluded)

The extrathyroidal extension and TNM stage of thyroid carcinomas were proposed to be associated with molecular risk subtyping (Hong et al., 2022). Thus, we investigated the aggressive features of the molecular alterations via the categorization of MRGs in 485 nodules with known TNM staging information (data displayed in Table S2). Among 174 BRAF-like tumors with confirmed diagnosis (PTC or PTMC) and Tx staging excluded, 57 cases (32.76%) were observed infiltration of thyroid capsule (T4a or T4b staging) and 115 cases (46.56%) of 247 assessable data (diagnosis of PTC or PTMC and Nx staging excluded) happened lymph node metastases (N1a or N1b staging). In RAS-like groups, however, two (22.22%) of the nine malignant nodules and four (40.00%) of the 10 cases were observed to have infiltration of the thyroid capsule and lymph node metastases, respectively. The dominating aberrations may be the BRAF^{V 600E} mutant and deletion-insertions, along with RET mutants and fusions.

Concurrent molecular alterations and histopathological characteristics

A total of 44 nodules with concurrent molecular alterations were isolated and separately analyzed to determine whether some aberrations tended to co-occur with others and present different histopathological characteristics. Figure 3 shows that all cases were classified as PTC or PTMC, with the exception of two benign nodules and one FTC. The most common BRAF^{V 600E} could be detected in 39 nodules (88.64%) concurrently with other 25 mutants from eight genes, including TERT, TP53, PIK3CA, AKT1, BRAF^K601E, HRAS, KRAS, and NRAS, followed by the NRAS^Q61R mutant, which occurred in nine nodules (20.45%). Among the remaining five cases without the BRAF^{V 600E} mutant, two PTC nodules were related to the BRAF rearrangements with the PXK, FLNC, or SND1 gene, while one PTC nodule underwent double deletions in the PTEN gene and one benign nodule revealed concurrent mutants in KRAS^A146V and NRAS^Q61R. An unknown mutant in the protein detected in EIF1AX^c.338−2A>T was concurrent with HRAS^Q61R in a FTC nodule. These findings agreed with previous reports (Fagin & Wells Jr, 2016).

Previous studies have focused on the relationship of molecular markers with distant metastases, such as the BRAF^{V 600E} mutation and TERTp mutations. These were frequently found to co-exist and their presence was considered to be valuable for PTC relapse risk assessment (Ulisse et al., 2021; Soares et al., 2021). Likewise, in our cohort, potentially aggressive nodules may also be related to concurrent molecular alterations involving BRAF^{V 600E}being accompanied by TERT^Q302R, AKT1^L52R, NRAS^G12C, NRAS^Q61R and CCDC6-RET fusion. Five cases with lymph node metastases (N1 staging, marked with triangle symbols in Fig. 3) presented with infiltration of the thyroid capsule (T4 staging, marked with pentagram symbols in Fig. 3) after surgery resection, although one nodule with concurrent mutations of BRAF^{V 600E} and NRAS^G12C was an exception. Histopathologic slides of a resection in two cases of FV-PT(M)C were reviewed and the photomicrographs are presented in Fig. 4. The nodule with the combination mutations of BRAF^{V 600E}, TP53^G244D and KRAS^Q61R were poorly differentiated, while the lesion with concurrent alterations of BRAF^{V 600E} and TERT^Q302R displayed an extrathyroidal extension with regional nodal involvement and local invasion of the skeletal muscle (data not shown) and adipose tissue (Fig. 4). These observations indicated that thyroid nodules with concurrent molecular alterations may be related with poor differentiation, aggressiveness, and a follicular variant of the PTC subtype.