Metastatic site discriminates survival benefit of primary surgery for differentiated thyroid cancer with distant metastases: A real-world observational study

Background: Role of surgery in the management of differentiated thyroid cancer (DTC) with distant metastases (DM) remains controversial. We aimed to determine the survival benefit of primary surgery on the basis of metastatic site. Methods: A retrospective cohort study based on the SEER database was conducted to identify DTC patients with DM diagnosed between 2010 and 2016. Patients were divided into surgery and non-surgery group, and propensity score weighting was used to balance clinicopathologic factors between groups. Results: Of 3537 DTC patients with DM, 956 (66.0%) cases were managed with surgery and 493 (34.0%) cases were managed without surgery. There were 798 all-cause deaths and 704 DTC specific deaths over a median follow-up of 22 months. The weighted 3-year overall survival (OS) for the surgery group was 55.2%, compared to 27.8% (P < 0.001) for the non-surgery group. The magnitude of the survival difference with surgery was significantly correlated with metastatic sites (Pinteraction<0.001). Significant survival improvements in surgery group compared with non-surgery group were observed in patients with lung-only metastasis (adjusted HR = 0.45, P < 0.001) , bone-only metastasis (adjusted HR = 0.40, P < 0.001), or liver-only metastasis (adjusted HR = 0.27, P < 0.001), whereas survival no improvement of surgery was found for patients with brain-only metastasis (adjusted HR = 0.57, P = 0.059) or multiply organs distant metastases (adjusted HR = 0.81, P = 0.099). Conclusion: The survival benefit offered by surgery for DTC patients with DM varies by metastatic sites. Decisions for primary surgery of DTC patients with DM should be tailored according to metastatic sites. multiply metastasis, specific survival in patients without brain metastasis, whereas surgery offered no overall survival DTC specific survival benefit for patients with brain metastasis. results suggest that individualized decisions for primary surgery of primary DM patients could be tailored on the basis of metastatic sites. Although there were no widely-accepted guidelines exist on the management of metastatic thyroid carcinoma, patients who underwent surgical resection had significantly longer survival than patients who did not in this study. We thus believe that

(34.0%) cases were managed without surgery. There were 798 all-cause deaths and 704 DTC specific deaths over a median follow-up of 22 months. The weighted 3-year overall survival (OS) for the surgery group was 55.2%, compared to 27.8% (P < 0.001) for the non-surgery group. The magnitude of the survival difference with surgery was significantly correlated with metastatic sites (Pinteraction<0.001). Significant survival improvements in surgery group compared with non-surgery group were observed in patients with lung-only metastasis (adjusted HR = 0.45, P < 0.001) , boneonly metastasis (adjusted HR = 0.40, P < 0.001), or liver-only metastasis (adjusted HR = 0.27, P < 0.001), whereas survival no improvement of surgery was found for patients with brain-only metastasis (adjusted HR = 0.57, P = 0.059) or multiply organs distant metastases (adjusted HR = 0.81, P = 0.099).
Conclusion: The survival benefit offered by surgery for DTC patients with DM varies by metastatic sites. Decisions for primary surgery of DTC patients with DM should be tailored according to metastatic sites. Background Differentiated thyroid carcinoma (DTC), which includes papillary thyroid carcinoma (PTC) and follicular thyroid carcinoma (FTC), is one of the most curable endocrine cancers. In the majority of patients with DTC, the cause of death is due to distant metastases (DM) rather than locoregional recurrence.
DTC with DM at initial diagnosis (primary DM) had markedly varying clinical outcomes from rapid progression and death to complete remission (4)(5)(6)(7)(8)(9)(10), the difference of metastatic disease site was considered as a possible reason for inconsistent outcome. The site of metastasis had been reported as significant association between extrapulmonary metastases and poor prognosis (4,(7)(8). Of note, the survival benefit of the removal of the primary tumor in patients with primary DM among those trials is controversial.
The aim of this real-world observational study was to determine the survival benefit of primary surgery among patient subpopulations stratified by metastatic sites who presented with DM at initial diagnosis. We hypothesized that the local surgery may confer a survival benefit to patients with low metastatic tumor burden.

Study Design and Data Source
After receiving an exemption from the Partners HealthCare Institutional Review Board, we performed a retrospective longitudinal cohort study using data obtained from the SEER Program of the National Cancer Institute.
The Surveillance, Epidemiology, and End Results (SEER) database (http://seer.cancer.gov/) sponsored by the National Cancer Institute (NCI) covered 18 population-based registries, involving a large proportion (28%) of US people. We used the November 2018 SEER-18 submission for this retrospective longitudinal cohort study, which included patients from geographic regions covered as follows: Metropolitan Atlanta, Connecticut, Detroit, Hawaii, Iowa, New Mexico, San Francisco-Oakland, Seattle-Puget Sound, Los Angeles, San Jose-Monterey, Utah, Rural and Greater Georgia, Alaska, Greater California, Kentucky, Louisiana, and New Jersey. We identified 3537 patients who were first initially diagnosed as DTC with DM between January 1, 2010, and December 31, 2016 (Fig. 1). We excluded patients only diagnosed by autopsy or death certifications, moreover, those cases without histological confirm were also excluded. Patient receiving surgery for metastatic sits or unknown sites as well as cases with unknown metastatic sites were excluded. At last, eligible 1449 DTC patients with primary DM were included in this study, who were grouped according to whether they underwent primary tumor surgery (N = 956) or not (N = 493). Considering this study was reviewed and approved by the institutional review board at the Shaoxing Second hospital and determined to not be human participant research, patient consent was not involved.

Identification of key variables
The SEER*Stat software (version 8.3.6) was used to extract relevant information, including patient identification, age of diagnosis, year of diagnosis, sex, tumor size, regional lymph node status, race/ethnicity, marital status, distant metastatic site, histology type, nuclear grade, surgery, chemotherapy, radiation therapy, cause-specific death classification, other cause of death classification and survival month.
International Classification of Diseases for Oncology (ICD-O-3) (13)was used to identify the cancer site and histology type, and cases of papillary and follicular thyroid cancer were selected using the restrictions and . We placed Hurthle cell carcinoma (ICDO-3 = 8290) into the category of follicular carcinomas, as used in Lim et al. (14,15). To investigate the benefit of primary surgery on the basis of metastasis sites, the variable was categorized into single organ and multiple organs metastases. The single organ metastasis was further classified into bone-only, liver-only, lung-only and brain-only metastasis, and multiple organs metastases were classified into multiply organs metastases including brain or excluding brain.

Main Outcome Measure
The primary endpoint of this study was overall survival (OS) and disease-specific survival (DSS), which was defined as an internal from time of diagnosis to overall death (DTC specific death) or date of last contact and considered as censored statuses if patients were alive until date of last contact. SEER defines mortality data based on the International Classification of Diseases Revisions 8 to 10, which categorized the cause of death as DTC specific death and other cause death.

Statistical Analysis
For this study, we adopt the similar statistical analytic approaches with previous studies (16,17) that examined the benefit of interventions for breast cancer subsets. Clinicopathologic factors were compared between the surgery groups and non-surgery groups using Pearson χ2 tests. Multiple imputation of missing data was performed by a multivariate logistic regression model, and 10 cycles were repeated to produce a final data set. Imputation model included these variables as follows: race (white, black, or others), marital status (single, separated and married), nuclear grade (I,II,III,IV), tumor size classification (0-2cm, 2-4cm,or 4 cm), and regional lymph node status(positive or negative).
Propensity score weighting was then used to balance patient characteristics between the surgery and non-surgery groups (18.19). we calculated the propensity scores based on patient age, year of diagnosis, race, sex, tumor size, regional lymph node status, marital status, distant metastatic site, histology type, nuclear grade, chemotherapy, radiation therapy through a logistic regression model for receipt of surgery. From the model, the inverse predicted probability of breast surgery assignment was used to define weights for patients who received surgery(1/probability) and for those who did not receive surgery (1/[1 − probability]). Patient characteristics after propensity score adjustment are shown to be balanced in Table 2.
The hazard ratios for the DSS and OS of patients in the surgery group compared with patients in the non-surgery group were evaluated using propensity score weights for log-rank tests and Cox regression models. Hazard ratios (HRs) of OS and DSS were reported from multivariable models that adjusted for patient age, year of diagnosis, race, sex, tumor size, regional lymph node status, marital status, distant metastatic sites, histology type, nuclear grade, chemotherapy, radiation therapy. Similar procedures were also performed among subgroups defined by metastatic sites, and interaction tests were conducted using a likelihood ratio test to explore whether any survival benefit conferred by surgery varied across subgroups.
In addition, to assess the stability of our results, we conducted a series of sensitivity analyses. First, the entire analyses were repeated after imputation unknown data using random survival forest methodology. Then, proportional subdistribution hazards model was used to calculated HR of OS and DSS between surgery and non-surgery group after adjusting competing events (20) such as death from other causes. Second, we performed the analysis after restriction to patients in the SEER 9 registry, because the data in the SEER 9 registry are more accurate than the data in newer SEER registries (21). Last, since age under 55 years was a good prognosis factor whether there is distant metastasis, we excluded the patients with primary DM under the age of 55 years who often sought surgery for longer survival.
All P values were calculated from 2-sided tests with threshold of 0.05 to evaluate statistical significance of survival benefit by surgery, and all statistical analyses were performed using R software (version 3.6.1).

Patient characteristics
We identified 3537 eligible DTC patients with DM at the time of initial treatment on the basis of our inclusion and exclusion criteria (Fig. 1). Of this cohort, 956 (66.0%) received the primary surgery, and 493 (34.0%) patients were stratified into non-surgery group. Clinicopathologic factors and SEER cancer registries according to receipt of primary surgery were listed in Table 1. The final data after multiple imputations was exhibited in Table 2. Balance in patient characteristics was achieved after propensity score adjustments for estimating average treatment effect, as shown in Table 2. The proportion of patient with age under 55 years, earlier year of diagnosis, white/ethnicity, male, follicular, small tumor size, regional node negative, and nuclear grade was larger for the surgery group compared with the non-surgery group.  (17) 127 (13) Other a 82 (17) 130 (14) NA  Abbreviations: RAI: radioactive iodine; EBRT: external beam radiation therapy.
a American Indian/AK Native, Asian/Pacific Islander.
Survival benefit of primary surgery  Fig. 3).
In sensitivity analyses performed after the exclusion of patients with age under 55 years, after restriction of patients within SEER 9, after repeating analyses using the proportional subdistribution hazards model, we observed similar findings.

Discussion
Although DTC is a disease with generally a good outcome, patients presenting with distant metastatic disease have less favorable outcomes. Although distant metastasis from DTC are usually slowgrowing compared with other malignancies, some of the patients with these conditions die from disease-specific causes. For this reason, many risk stratification algorithms consider such cases to be high risk. Despite the higher chance of poor outcome, current treatment guidelines advocate an aggressive approach to management with surgery and postoperative radioactive iodine (RAI) therapy (22,23). Treatment consists of total thyroidectomy, neck dissection as indicated by the detection of disease in the central and/or lateral neck, followed by RAI therapy in most patients.
The prognostic value of distant metastatic site has been widely studied. In this study, the commonest site of metastases was lung (42%), bone (17%), liver (17%), brain (5%) and then the group with multiple sites affected (19%). These results were similar to those of previous reports (4,11,24). A number of studies finding that the prognosis of patients is related to the location of metastases (4,(7)(8). Patients with brain metastases have a worse outcome compared to other groups. In this study, the 3-year DSS rate of brain metastasis from DTC (28.7%) was the poorest compared with other organ metastasis (lung metastasis 45.9%, liver metastasis 40.9%, and bone metastasis 50.4%).
In this study, we observed higher overall and DTC specific survival in patients with primary DM managed with surgery than individuals without surgery, and this finding based on the real-world study was consistent with prior studies (25). This may be due to a possible heterogeneous treatment effect of primary surgery when metastatic sites varied. Our findings indicated that definitive local surgery of patients with only-brain metastasis offered no significant survival benefit over non-operative management, but a significant survival improvement for surgery was observed in other single organ metastasis DTC patients. For patients with multiply organs distant metastasis, surgery could also improve DTC specific survival in patients without brain metastasis, whereas surgery offered no overall survival or DTC specific survival benefit for patients with brain metastasis. These results suggest that individualized decisions for primary surgery of primary DM patients could be tailored on the basis of metastatic sites. Although there were no widely-accepted guidelines exist on the management of metastatic thyroid carcinoma, patients who underwent surgical resection had significantly longer survival than patients who did not in this study. We thus believe that the presence of DM alone does not automatically exclude the indication of aggressive local radical resection to clear margins. DM is undoubtedly the most common primary cause of cancer death in DTC, but other characteristics of DTC in the high-risk group, such as invasion to surrounding organs or anaplastic transformation in the neck lesion, can also become fatal. Haq et al. reported that lesser surgery (biopsy or nodulectomy) of the primary neoplasm in patients with DM was associated with worse survival compared to radical surgery (11). Although aggressive radical resection sometimes results in substantial perioperative or long-term morbidity, such as dysphonia, death from local causes including gradual suffocation from the tracheal involvement, dysphasia, and bleeding, most of them can be prevented.
Surgery, radiotherapy, and RAI therapy have been widely used to treat the DTC with distant metastases. In our study, the benefit of surgical intervention of locally disease was limited in patients with brain metastasis. RAI is an important systemic therapy for patients with brain metastases with RAI-avid disease who respond to this therapy (26). Unfortunately, based on current case reports and retrospective series, RAI uptake by cranial metastatic lesions is quite low, with a reported range from 0-25% of cases (27)(28)(29)(30)(31), possible explanations may be decreased expression of the sodium iodide symporter (NIS) or diminished membrane targeting of NIS in metastatic lesions (32). Furthermore, some studies have suggested that the prognosis in patients who present initially with metastases versus those who subsequently develop metastases may be different (7,11,12,33). The patients presenting initially with metastases appear to have relatively favorable outcomes compared with patients developing metastases after initial treatment. This result will due to the patients with distant metastasis at presentation are 'treatment-naive,' in particular with respect to RAI, and therefore strongly RAI avid (24). One study has found that complete local control and metastatic site RI avidity were independent predictors of an increased DSS or OS relative risk, whereas only complete local control was an independent OS predictor in patients presenting with primary DM (7).
The age at diagnosis of the initial cancer is known to be a valuable prognostic factor for the recurrence and mortality of DTC (34). In the current series multivariate analysis identified age was an independent risk factor for bad prognostic feature. We found that age over 55 years was significant predictors of a poor outcome, the results showed that an improved survival with patients younger than 55 years at presentation compared to those older than 55 years (3-year DSS 52% vs. 45%, respectively). Age, as a factor of poor prognosis, is mainly directly related to the degree of differentiation of thyroid cancer. The association with age is directly related to tumor differentiation and hence RAI avidity. Nixon et al. found that age and RAI avidity were associated, and younger patients often had higher rates of RAI-avid which was associated with a good outcome (33), while loss of RAI avidity was associated with a poor outcome (35).
The number and location of metastases also affect the prognosis of DTC patients with primary DM. In the present study, the most frequent locus and the number of the brain metastasis was not analyzed.
Al-Dhahri et al. noted that brain metastasis occurs more frequently in the cerebral hemispheres, other sites for intracranial metastasis are the cerebellum, brainstem and pituitary (36). Previous study reported that brain metastases in the brainstem as well as with cranial neuropathy or vision changes could lead to a poor prognosis (37,38). In addition, patients with multiple cranial metastases seemed to had a worse outcome than patients with a single metastasis.   The distribution of metastatic sites in differentiated thyroid cancer patients with distant metastases.

Figure 3
Hazard ratio comparing OS/DSS between surgery group and non-surgery group according to metastatic site for patients with distant metastases. (*) Weighted by inverse propensity score. (Ϯ) Multivariate analysis adjusted by patient age, year of diagnosis, race, sex, tumor size, regional lymph node status, marital status, distant metastatic sites, histology type, nuclear grade, chemotherapy, radiation therapy. Abbreviations: OS, overall survival; DSS, disease specific survival; HR, hazard ratio.