The differences in distant metastatic patterns and their corresponding survival between thyroid cancer subtypes

This study aimed to systematically elucidate the metastatic patterns and their corresponding survival of each thyroid cancer subtype at time of diagnosis.


| INTRODUCTION
Thyroid cancer is the most common type of endocrine cancer and its incidence has increased three-fold during the past 30 years. 1 Among thyroid cancers, papillary thyroid carcinoma (PTC) remains the most common histologic subtype, followed by follicular thyroid carcinoma (FTC), medullary thyroid carcinoma (MTC), poorly differentiated thyroid carcinoma (PDTC), and anaplastic thyroid carcinoma (ATC). Most thyroid cancers have an indolent course with an estimated 5-year survival of more than 95%. However, a subset of thyroid cancers may behave aggressively with distant metastasis (DM) at time of diagnosis, which is the main prognostic factor for disease-specific mortality. 2 Several clinicopathological and molecular parameters have been demonstrated to be significant risk factors associated with this rare event. [3][4][5][6] Clinical data on the dissemination patterns and survival of thyroid cancers with DM at presentation have been primarily gathered from retrospective institutional experience. Nationwide or large population-based analyses on this topic are needed to understand its full clinical implications. Several studies have described the metastatic patterns of thyroid cancer variants, mainly differentiated thyroid cancers (DTCs), [7][8][9][10] with only a few studies focusing on MTC, PDTC, or ATC. [11][12][13] Unfortunately, the number of included patients in these studies is relatively small and often limited to a specific subtype of thyroid cancer.
This study aimed to evaluate the DM patterns of different thyroid cancer subtypes and their impact on patient survival utilizing the Surveillance, Epidemiology, and End Results (SEER) program.

| MATERIALS AND METHODS
We accessed the SEER 18 registries custom database to search for thyroid carcinomas from 2010 to 2018. The year 2010 was selected as the cut-off year because the metastatic location data were tabulated in the SEER database from this year onward. DM at presentation was categorized as M0 (without DM) and M1 (with DM). During this period, a total of 119 206 thyroid cancers were identified and we subsequently removed 6256 cases with missing information for DM at presentation. Among 115 481 cases with available information, there were 2787 (2.4%) cases with DM at presentation included for analyses. We extracted the following covariates: patient ID, age at diagnosis, sex, race, primary site, histology diagnosis (Table S1), tumor diameter, nodal/distant metastasis, multifocality, sites of DM (bone, brain, liver, lung, distant nodes, and other sites), the extent of resection, radiotherapy, chemotherapy, overall survival (OS) time, and OS status.
The chi-squared and Wilcoxon rank-sum tests were used to compare clinical parameters between groups. For survival analyses, Kaplan-Meier analysis and Cox regression analysis were run to analyze all-cause mortality differences between M1 groups. Hazard ratio (HR) and corresponding 95% confidence interval (CI) were computed. All p-values are two-sided and a value of less than 0.05 was considered statistically significant. Statistical analyses were performed using R, version 3.6.1 (The R Foundation, Vienna, Austria).
Patients with thyroid cancer with DM at time of diagnosis presented at an older age and with a large tumor diameter. Females were slightly more common than males. PTC was the most common histologic subtype, followed by ATC and FTC. Nearly half of the cases were unresectable and diagnosed by biopsy only. The characteristics of the patient cohort are described in Table 1. 3.2 | Dissemination patterns of M1 thyroid cancer subtypes Among PTC, HCC, PDTC, and ATC, the lungs were the most common site for DM whereas bone metastases were most frequently observed in FTC and MTC (p < 0.001). The proportion of liver metastases was remarkably higher in MTC (42.7%) as compared to other histologic subgroups (p < 0.001) ( Figure 1). On the other side, distant node involvement mainly occurred with MTC (37.4%), PTC (30.5%), and ATC (27.9%). About 27.6% of patients were documented with metastases to multiple organs at the time of diagnosis. MTC (40.6%) and ATC (32.8%) most frequently metastasized to more than one site at presentation. Metastases to lung alone were seen in 53.2%, 49.1%, and 47.4%, of ATCs, PTCs, and PDTCs, respectively, whereas bone metastases as the only site of metastatic disease were found in 43.1% of FTCs and 26.5% of HCC. Single organ dissemination to the liver was generally uncommon; they were identified in 20.3% of MTCs but only occurred in 0-5% of other subtypes. Brain metastases alone were also extremely rare in thyroid cancers. Table 2 shows the correlation of single-and multi-organ metastases with various clinical parameters. Large tumor diameter, non-White race, ATC/FTC/MTC histology, and unifocal tumors were significant risk factors for the presence of multi-organ metastases at presentation. Other clinical covariates including age, sex, and regional nodal metastasis were not associated with increased risk for multi-organ spread. Patients presenting with DM to multiple organs were associated with a dismal outcome in comparison to those presenting with DM to only one site (HR = 1.810; 95% CI = 1.624-2.017; p < 0.001). Stratified by metastatic sites, patients with multi-organ metastases had a statistically comparable survival with brain metastases alone (median OS of 5 months versus 14.0 months, respectively; p = 0.158), but had a shortened survival in comparison to the liver (median OS of 15.0 months; p < 0.001), lung (median OS of 9.00 months; p < 0.001), and bone (median OS of 26.5 months; p < 0.001) (Figure 3). Metastases to the brain alone had a statistically comparable prognosis with metastases to the liver (p = 0.461) or lung alone (p = 0.791). On the other side, metastases to bone alone showed an improved outcome compared to liver (p = 0.016), lung (p < 0.001), and brain alone (p = 0.011). Patients with DM to another

| DISCUSSION
Thyroid cancers, especially DTCs, generally have a good prognosis and show a good response to initial thyroidectomy. DM at presentation is not a common event in thyroid cancers and its incidence ranges from 3% to 15% in published literature. 7,14,15 It has been well established as one of the leading adverse events in thyroid cancers. 2 Published studies on DM in thyroid cancers are limited given its relatively rare occurrence and most of these studies solely focused on a specific subtype of thyroid cancers such as DTC, MTC, PDTC, or ATC. Additionally, studies reporting DM data less frequently reported the locations of metastatic spread limiting the understanding of the dissemination patterns and differences in patient survival between thyroid cancer subtypes with DM. This study accessed a large population-level database and demonstrated that each thyroid cancer subtype has a unique remote spread pattern and survival.
The lung and bone have been shown as the most common site of DM in DTCs. 8 16 Another single institution reported an incidence of 48%, 24%, and 19% of metastatic DTC metastasizing to lung only, bone only, and multiple sites, respectively. 8 For MTCs, the most common disseminated sites were liver and bone. 18 ATC is the most lethal form of thyroid cancer and DM is a frequent event among these neoplasms. Besic et al. reported that 78% of metastatic ATCs were seen in the lung, followed by intrathoracic lymph nodes while bone metastases were uncommon. 19 In a recent nationwide study on autopsy of patients with thyroid cancer, there were major differences in hematogenous spread patterns between thyroid cancer subtypes with the lung being the most frequent site for DTC and ATC whereas MTC more likely metastasized to the liver. 20 Our results affirmed that each thyroid cancer histological subtype has a distinctive metastatic pattern. Some of our findings differed from known patterns reported in prior clinical studies. This information is extremely helpful for clinicians to further guide the search for metastatic disease when the histological diagnosis has been confirmed.
Besides metastases to remote organs, our results also highlighted that spread to distant lymph nodes was not uncommon in M1 thyroid cancers, particularly MTC, PTC, and ATC. The risk of DM is known to be associated with nodal involvement. 21 Unfortunately, we were unable to investigate the location distribution of distant node metastases since these data are not included in the SEER database.
The prognostic impact of different metastatic sites in thyroid cancers with DM is still controversial. It is of clinical interest to accurately evaluate the risk stratification of these patients. Our study demonstrated that patients F I G U R E 3 Kaplan-Meier curve illustrating the prognostic impact of metastatic sites on overall survival of thyroid cancers with DM at presentation with thyroid cancer with metastases to multiple sites have the worst prognosis. This prognostic information is useful during patient counseling and may help clinicians consider additional local or systemic therapies. These high-risk patients might be suitable candidates for future clinical trials on novel targeted therapies. Wang et al. reported an estimated 5-year OS of 77.6% in DTCs with single-organ metastases and only 15.3% in those with multi-organ spread. 22 In this study, we also demonstrated that metastatic thyroid cancers did not have a universally similar outcome. Metastatic ATC had the worst outcome compared to MTC, PDTC, or DTC with an estimated 1-year OS of less than 5%. Therefore, it is important to determine the accurate histological diagnosis and the extent of DM at time of diagnosis. We also found several important risk factors associated with multi-organ metastases at presentation including non-White race and ATC/MTC/FTC histology. Additionally, large solitary thyroid cancers might have a higher risk for multi-organ involvement than smaller ones with multiple foci. These data are crucial to alert clinicians to search for multiorgan metastases in high-risk patients to better assess patient outcomes.
There are gaps of knowledge on why thyroid cancers have distinctive spread patterns. Our results support the premise that different thyroid cancer subtypes may have different underlying biology and tumorigenesis. Several molecular biomarkers are associated with an increased risk for hematogenous spread in thyroid cancers including TERT promoter, RET, PLEKHS1, and TP53 mutations. 4,6,23 The distinct genetic landscape of each subtype might explain why different tumor cell types favor distant migration to a specific organ. Oncogenic mutations in the BRAF kinase gene have been shown to cooperate in the pathogenesis and development of lung metastases in melanomas. 24 BRAF mutation is the most predominant activating mutation in PTC and ATC, which could explain why these two entities likely develop lung metastases. Coexisting occurrence of BRAF and TERT promoter mutations, which is commonly seen in ATC and PDTC, could enhance the risk of DM at presentation. 25,26 ATC carries the highest tumor mutation burden among thyroid cancers resulting in the high likelihood of developing DM in these neoplasms. 27 Our study is the largest study to date providing data on disseminated locations of thyroid cancer subtypes with DM. Besides the outlined clinical importance, it is also essential to discuss the limitations of this study. Firstly, we could not avoid the certain selection biases inherent to retrospective patient collection and nature and design of population-level databases.. 8 It should be noted that HCC was only considered as a separate entity from FTC in 2010. 28 Hence, data on FTC and HCC before the time that SEER started adding an ICD-O-3 code for HCC may be biased. Next, data on DM during follow-up and recurrence-free survival were not available for analysis. A few studies have outlined the differences between thyroid cancers presenting with DM at diagnosis versus developing DM during follow-up. 16,29 Also, data for radioactive iodine treatment and chemotherapy regimen details are not fully complete in the SEER database, and it is not recommended to make statements about treatment effectiveness if confounding variables could not be adjusted. In the SEER database, metastases to rare distant sites such as adrenal glands, pancreas, kidney, or peritoneum are also not documented which could limit or confound the extent of our analyses. Finally, a small subset of M1 thyroid cancers (3.5%) had a history of prior malignant tumors and we could not determine whether the DM foci at the time of thyroid cancer diagnosis were thyroid origin or from the prior tumors.
In conclusion, this study demonstrated that different thyroid cancer subtypes remotely spread to specific organs and have unique metastatic patterns, which are important for patient counseling and mortality risk stratification. Given the relatively frequent occurrence of multi-organ metastases in thyroid cancers with DM at presentation and its adverse impact on patient outcome, clinicians should consider searching for additional metastases according to each specific histology and confounding risk factors.

CONFLICT OF INTEREST
The authors declare no conflicts of interest.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.