Thyroid nodules are a common clinical occurrence, with their prevalence varying from 20% to 65% in the general population [1]. The critical distinction between benign and malignant or premalignant lesions is paramount for effective management, particularly when determining the need for surgery. The initial evaluation of these nodules often involves ultrasound imaging coupled with fine-needle aspiration (FNA) cytological analysis [2,3], but these methods have limitations, including variable diagnostic accuracy, especially in cytologically indeterminate nodules where the risk of malignancy (ROM) hovers between 13% and 34% [4]. Hence, subjecting most patients with indeterminate nodules to diagnostic surgery results in a substantial proportion undergoing unnecessary procedures [5]. The emergence of molecular testing has shown promise in overcoming diagnostic challenges and improving clinical management [6]. Multigene panels, which analyze mutations across various relevant genes altering the molecular characterization of thyroid tumors, offer a broader perspective on thyroid cancer pathogenesis [7,8]. This comprehensive approach enhances the assessment of ROM, helping to make more informed decisions for individual patients [9].

Initially, molecular tests focused on a single or a small number of genetic variants in key driver genes, such as BRAF, RAS, peroxisome proliferator-activated receptor γ (PPARγ)-paired box 8 (PAX8), and/or rearranged during transfection/papillary thyroid carcinoma (RET/PTC) 1/3 (7-gene panel), but their diagnostic efficacy was insufficient [10]. Commercial multi-gene panels, including ThyroSeq v3 (ThyroSeq, Rye Brook, NY, USA), Afirma gene expression classifier (GEC), gene sequencing classifier (GSC) (Afirma, Washington, DC, USA), and ThyGeNext/ThyraMIR (Parsippany, NJ, USA), employing nextgeneration sequencing (NGS) methods, have demonstrated significantly improved diagnostic performance by including a large number of genes [6]. Thus, the American Thyroid Association 2015 guideline [2] and recent European guidelines [11] recommend that molecular testing can be considered for cytologically indeterminate nodules, although studies on the actual clinical and surgical implications remain limited [12]. However, considering the cost of these commercial NGS tests, European guidelines state these tests should be used “if possible” and advocate the development of less expensive tests. The price of commercial NGS tests can range from US$3,000 to over US$6,000, which includes the cost of RNA or miRNA sequencing in addition to DNA tests [13,14].

On the other hand, a substantial portion of indeterminate nodules identified as follicular pattern tumors harbor mutations associated with low-risk molecular subtypes, such as RAS-like or non-BRAF-non-RAS (NBNR), with a low ROM or risk of recurrence (i.e., low-risk mutations) [2,15,16]. More than 90% of RAS-like or NBNR tumors with genetic alterations possess a point mutation in RAS, eukaryotic translation initiation factor 1A X-linked (EIF1AX), dicer 1, ribonuclease III (DICER1), phosphatase and tensin homolog (PTEN), enhancer of zeste 1 polycomb repressive complex 2 subunit (EZH1), or tumor protein p53 (TP53). Therefore, a small DNA gene panel testing these genes with the addition of BRAF gene can yield superior diagnostic performance compared to the 7-gene panel, even though it cannot detect RET and PPARγ fusions that the 7-gene panel can [16]. Moreover, although the sensitivity or negative predictive value (NPV) of this small DNA gene testing alone may be lower than that of commercial NGS tests that include RNA or miRNA testing, it could still support clinical decision-making for the management of indeterminate nodules with a cost at least five to six times lower. Additionally, this small DNA gene testing with the addition of telomerase reverse transcriptase (TERT) promoter and phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha (PIK3CA) genes can also cover common double mutations often found in aggressive thyroid cancer [17]. Therefore, the small DNA testing of those 11 genes can aid not only in diagnosis but also in prognosis prediction. Therefore, this study investigated the clinical efficacy of molecular testing, specifically in the diagnostic performance and the ability of reducing the surgical resection rate (RR), through prospective DNA panel testing of the 11 genes. These genes included eight main driver genes of follicular patterned tumors (BRAF, NRAS, HRAS, KRAS, EIF1AX, DICER1, PTEN, and EZH1) and three genes associated with high-risk cancers (TERT promoter, PIK3CA, and TP53). We assessed its diagnostic accuracy by comparing it with three commercial gene panel tests and further analyzed the surgical RR through a comparison with propensity score-matched controls who did not undergo molecular testing.

This study demonstrated that a small 11-gene DNA panel, composed of genes prevalent in indeterminate nodules or associated with prognosis, could result in a clinically significant decrease in the RR and an increase in active surveillance for low-risk cancers. Although the diagnostic accuracy of this panel test was lower than that of other existing commercial tests with a larger number of genetic mutations, it still yielded reasonably good results.

Our findings demonstrate that the utilization of the small DNA panel can significantly reduce the RR of thyroid nodules. In particular, the RR of both all nodules and indeterminate nodules was notably lower in the panel group compared to the control group (27.0% vs. 52.5% and 23.5% vs. 68.2%, respectively). The RR in this study was lower than the PCR rates (41.7% and 40.0%, respectively). Most commercial panel tests recommend surgery in cases with positive results, but they suggest surveillance as an alternative approach if the molecular results suggest a low risk of cancer progression [37]. In this study, only 25.7% of patients with RAS mutations, 12.5% with NBNR mutations, or 18.5% without mutations in 11 genes of our panel underwent surgery, contrasting with 68.2% of control patients with indeterminate nodules. After the panel test, in Bethesda III nodules, 47.1% of RAS tumors were resected, while only 16.7% of NBNR tumors, similar to the 17.7% rate in the no mutation group, underwent surgery. This suggests that a larger proportion of patients with no or low-risk mutations preferred surveillance or delayed surgery instead of immediate surgery in our subjects, leading to a 50% (from 52.5% to 27%) reduction in surgery in all nodules and a 65% (from 68.2% to 23.5%) reduction in indeterminate nodules by 11-gene DNA panel testing. A gene mutation detected by our panel does not automatically categorize the nodule as malignant; each mutation’s significance is assessed individually. BRAF mutations typically indicate nearly a 100% ROM, and double mutations are also associated with a high ROM with aggressive behavior [38]. RAS mutations generally signify potential malignancy or premalignant lesions, with a ROM ranging 44% to 66% [38,39], often favoring surgical intervention in consideration with other clinical factors. Other single mutations (NBNR), such as EZH1 mutation, are generally considered low-risk mutations more likely associated with adenomas, with a ROM ranging from 6% to 47.6% [39-41]. It is crucial to note that despite the improvement in molecular testing, current clinical decision-making for surgical procedures involves a myriad of considerations beyond cytopathological findings, including nodule size, growth pattern, ultrasound features, patient history, and patient preferences.

There are risks associated with the reduced surgical RR and increased reliance on active surveillance, particularly in the context of cancer patients. These risks include the potential progression of premalignant lesions into cancer, the risk of cancer advancement and metastasis, as well as the potentially increased surgical complexity and associated complications. These are crucial aspects that deserve careful consideration. However, it is noteworthy that the risks associated with active surveillance of thyroid nodules or even thyroid cancer without mutations or with single mutations, particularly RAS-like or NBNR mutations, may not be extensively significant, as highlighted in previous studies [42,43]. Nevertheless, our small 11-gene panel does not detect copy number alterations common in OAs or low-risk fusions found in FAs or FTCs, which are detectable by larger commercial panels. Thus, while our panel includes aggressive mutations prevalent in thyroid carcinomas, a broader genetic assessment is necessary to fully evaluate the potential risks associated with active surveillance of thyroid nodules or cancers. Furthermore, gene panel testing results can offer predictions regarding the ROM, potentially making it easier to accept the decision to pursue active surveillance for lower-risk patients. On the other hand, for patients inclined towards surgical resection, the typical recommendation is diagnostic lobectomy. However, the use of the panel introduces the option of considering an initial total thyroidectomy instead of lobectomy, dependent on the number and type of mutations identified. Additionally, the genetic information provided by the gene panel can guide long-term patient management, allowing for more active and careful follow-up for those at higher risk and less intensive follow-up for those at lower risk. As a result, the panel emerges as a vital clinical tool that may assist in determining the optimal surgical extent and guide personalized patient care, and thus, integrating the panel into routine clinical practice may be beneficial.

Although the ROM was significantly higher in the panel group than in the control group for all nodules (81.5% vs. 63.9%), suggesting that the panel’s molecular information could potentially contribute to sparing patients from unnecessary surgery, the difference was not very large, and there were no significant differences in the ROM among specific BC cytologies. This lack of significance may be attributed to the relatively small sample size, but it is also important to consider that RAS mutations and some other single gene mutations are not exclusive to carcinomas and can also occur in adenomas [44]. For certain genes, such as EZH1, the likelihood of a nodule being an adenoma with these mutations is relatively higher [45,46]. Thus, if the surgical resection of all nodules with a positive mutation of any gene is considered for malignant or premalignant lesions, it can heighten the sensitivity in cancer diagnosis. However, this lowers the specificity, impacting the ROM associated with nodules containing such mutations. Moreover, in cases where a gene mutation is present and the nodule is actually malignant, the decision to opt for active surveillance due to it being categorized as low-risk cancer can also impact ROM. Given that the ROM is influenced by these paradoxical decisions stemming from mutation results, it is imperative to interpret the outcomes of molecular tests with caution.

Unlike some other commercial panels, our multi-gene panel exhibited lower diagnostic performance, especially in terms of sensitivity or NPV, compared with Afirma GSC and ThyroSeq v3, which was anticipated due to the smaller number of genes covered in our panel. In theory, an increase in the number of genes associated with thyroid cancers covered in the panel would lead to higher NPV. However, it is important to note that this would also result in a greater inclusion of mutations associated with adenomas, making it challenging for PPV to improve. Consequently, even established commercial panels have faced difficulties in achieving optimal diagnostic accuracy [13]. Given these inherent challenges, our panel was not expected to attain perfect accuracy. While our panel exhibited a lower NPV than other commercial gene panels, the PPV was similar.

Hence, if we solely evaluate the NPV of our small DNA panel, it may be challenging to regard it as a viable alternative to larger commercial panels. Nevertheless, it is crucial to emphasize that guidelines for thyroid nodule management recommend active surveillance, as an alternative to surgery, for low-risk PTCs and indeterminate nodules, including follicular neoplasms. This consideration depends on various factors such as nodular size, growth rate, imaging characteristics, and patient preference, in addition to molecular findings. Even when dealing with nodules that could potentially be FTC or follicular variant PTC, active surveillance becomes a viable option if they are relatively small, exhibit a low TIRADS score, show no significant changes on ultrasonography, and notably lack high-risk mutations. Although the small multi-gene panel may not cover all known mutations associated with thyroid cancers, as demonstrated by the reduction in RR in this study, it can effectively reduce diagnostic surgery and increase active surveillance of thyroid cancer with low risk at an economical price, provided that it covers prevalent mutations associated with the risk of cancer development or progression.

The cost-effectiveness of multi-gene panels remains a subject of debate. It has been postulated that molecular testing can help avoid unnecessary surgery and reduce costs [47], partially based on the fact that, despite the high costs of large-scale molecular tests themselves (estimated costs for tests like ThyroSeq v3, Afirma GEC, Afirma GSC, and ThyGeNext/ThyraMIR ranging from US$3,000 to over US$6,000) [14], these molecular tests are cheaper than cost estimate for a diagnostic lobectomy [48]. However, this assertion holds particularly true in countries or regions where surgical expenses are notably high, such as in the United States. In contrast, in locations like Korea, an open thyroid lobectomy, inclusive of preoperative assessments and immediate postoperative visits, costs approximately US$712 [49], and this cost is even lower when a substantial part is covered by National Health Insurance in patients diagnosed with cancer. Therefore, it is crucial to consider the application of appropriate molecular tests, taking into account varying cost-effectiveness across different countries. To assess the cost-effectiveness, we indirectly compared our 11-gene panel to the established ThyroSeq3 commercial panel for thyroid nodules. ThyroSeq3, priced at $3,200 [6], demonstrated a reduction in RRs from 54% to 24% [50]. Given the United States diagnostic lobectomy cost of approximately $9,602 [48], the cost per avoided lobectomy was approximately –$21,340. In contrast, our 11-gene panel, priced at $550, showcased a decline in RRs from 53% to 25%. Considering the Korean diagnostic lobectomy cost of $712 [49], the cost per avoided lobectomy was approximately –$580. A negative cost per avoided lobectomy suggests cost savings, indicating our smaller panel may offer a more cost-effective option for thyroid nodule evaluation. However, further comprehensive studies with larger samples are needed to validate these findings.

Clinicians should carefully evaluate clinical characteristics during follow-up, even in mutation-negative nodules, which may turn out to be a cancer showing progression [51]. Additionally, proper education on molecular testing is crucial for clinicians, considering the lack of specificity in existing guidelines regarding the appropriate usage and interpretation of these tests. Given the significant added costs and potential challenges in interpretation, clinicians must exercise caution when ordering and interpreting molecular tests [12]. Recognizing the inherent limitations of molecular testing and staying updated with evolving research are essential for optimizing the application of molecular markers in clinical settings [52].

Although probing only the surface of underlying tumor biology, molecular testing offers accessible prognostic information. Multi-gene panels can identify mutations highly specific for malignancy, such as BRAF^V600E or double mutations with TERT or TP53 [53]. Preoperatively identifying these higher-risk mutations empowers clinicians to identify potentially high-risk thyroid nodules and tailor management accordingly, including the appropriate surgical extent for those undergoing surgery [54]. Our study detected four nodules with double mutations, indicating that the discovery of high-risk mutations is possible even with a panel focused on a limited number of essential genes.

This study has some limitations. First, in cases where surgery was not performed, we could not definitively determine whether the nodules were truly malignant or benign, leading to uncertainty regarding the actual ROM. Second, being conducted by a single clinician at a single center increases the risk of bias and limits the depth of research compared to multicenter studies. However, for our investigation using a small 11-gene panel, we deliberately chose a single-center approach to minimize result interpretation variations among clinicians, given the panel’s novelty and specific focus. Although we acknowledge the evolving utility of this 11-gene panel, future research is encouraged to leverage more extensive multicenter studies with larger sample sizes and a broader spectrum of clinicians for comprehensive insights. Third, our findings may not be generalizable to other populations or settings. While expensive commercial panels may be suitable in countries like the USA, where the higher cost is justified by slightly higher diagnostic rates, they are not as suitable for Korea. Thus, we developed this 11-gene panel originally for the Korean setting, considering the local need for a cost-effective and genetically appropriate panel. However, with satisfactory results, the 11-gene panel might also prove useful in other populations, although further research is needed to confirm its effectiveness and cost-efficiency in different contexts. Fourth, the decision to conduct molecular testing was not performed in all consecutive nodules because of patient preferences, introducing a potential selection bias. We anticipated that some patients might opt out of the study if they were required to undergo an additional FNA for gene testing after receiving cytology results. To mitigate this risk, we streamlined the process by performing both the cytological analysis and multi-gene panel testing during the same FNA procedure, although this approach deviates from the standard clinical practice. Furthermore, the study periods differed between the prospective panel and retrospective control groups, potentially introducing bias. Given the non-randomized nature, we saw that identical time frames would likely introduce more patient selection bias. To address this, we intentionally set distinct study periods to minimize unwarranted influence on patient selection and outcomes. Specifically, the control group had a minimal 1-year difference from the prospective panel, chosen strategically to reduce variations in practices impacting outcomes. Importantly, no significant changes in guidelines, diagnostic procedures, or surgical practices occurred within the 1 to 2 years of the study period. Lastly, the inherent limitations of the panel itself have implications for its diagnostic performance. The panel exclusively tests for point mutations in 11 genes from DNA samples, lacking the capacity to detect other types of genetic alterations such as fusions and copy number variations. Therefore, further research is required to develop a multi-gene panel that encompasses the most essential mutations for the molecular diagnosis of thyroid cancer while remaining cost-effective.

In conclusion, our small 11-gene DNA panel is not only an economical test, but it also is capable of aiding in the clinical management of thyroid nodules by providing information on the ROM or aggressiveness of thyroid cancer based on accompanying mutations, complementing standard clinical, radiological, and cytological evaluations. This approach holds the potential to reduce unnecessary surgery, particularly for patients seeking active surveillance for non-aggressive malignant nodules. By integrating molecular information into risk assessment and management strategies, multi-gene panels contribute to refining thyroid nodule management practices and advancing patient care. Further research is necessary to identify the optimal diagnostic molecular tool for managing indeterminate thyroid nodules, focusing on cost-effective gene panels with an appropriate number and combination of genes.