Propensity score matching analysis of the prognosis for the rare insular subtype of thyroid cancer based on SEER database

Insular thyroid carcinoma (ITC) is an uncommon thyroid malignancy with an unclear prognosis. The aim of this study was to determine the prognoses of patients with ITC. We investigated a large cohort of patients with differentiated thyroid cancer from the Surveillance, Epidemiology, and End Results (SEER) database who were registered between 2004 and 2013, and compared the prognosis of patients with ITC to those with classic papillary thyroid cancer (CPTC) and follicular thyroid cancer (FTC). Patient mortality was determined using Kaplan-Meier analyses with log-rank tests, as well as Cox proportional hazards regression analyses. The study cohort comprised of 165 patients with ITC, 5419 patients with FTC, and 60739 patients with CPTC. The rate of cancer-specific mortality per 1000 person-years for ITC was higher than that for CPTC or FTC. According to multivariate Cox regression analysis, however, the cancer-specific and all-cause mortality rates of ITC were similar to those of CPTC and FTC. The cancer-specific survival rate in patients with ITC was higher than that in patients with CPTC, but similar to that in patients with FTC, after adjusting for potentially influencing factors using propensity score matching analysis. These findings, which contrast with previously published data, provide new implications for the treatment of patients with ITC.

ITC is a rare thyroid malignancy that was first defined by Carcangiu et al. in 1984 [11]. Its morphological features lie between those of well-differentiated carcinoma (papillary or follicular) and undifferentiated or anaplastic carcinoma of the thyroid [12,13]. This type of thyroid carcinoma is considered an aggressive cancer with a high propensity for local recurrence and distant metastasis [14,15]; it reportedly has a worse prognosis than classic thyroid carcinomas [9,12,13,15]. However, some studies did not find a significant difference in prognosis between patients with ITC and those with differentiated thyroid cancer: papillary and/or follicular thyroid carcinomas (FTCs) [16,17]. Additionally, most of the ITC-related literature comprises of case reports or small single institution-based case series, with no large population-level studies.
The Surveillance, Epidemiology, and End Results (SEER) program of the National Cancer Institute is Research Paper the largest publicly available database of its kind, and is a valued source of high-quality information for cancer incidence and survival in the United States [18,19]. Propensity score matching (PSM) is a statistical normalization method for analyzing observational data by estimating the effects of a treatment, policy, or other intervention and accounting for covariates that influence the administration of the treatment. The propensity score is a balancing score: The differences between groups on the covariates condensed down into a single score so if two groups balanced on the propensity score then balanced on all the covariates. PSM aims to reduce bias due to confounding variables [20]. In this study, we investigated the prognosis of ITC compared to PTC and FTC by analyzing the SEER database information from 2004-2013 using PSM methods due to the diversity of number of cases with different subtypes.

Risk factors for thyroid cancer-specific and allcause mortality rates
According to univariate Cox regression analyses, age, sex, race, TNM stage, tumor extension, radiation, and surgical approach were significant risk factors for cancer-specific mortality. In the multivariate Cox regression model, CPTC and FTC showed no significant risk for cancer-specific mortality compared to ITC after adjusting for influential risk factors (Table 3). According to univariate Cox regression analyses, age, sex, race, TNM stage, multifocality, tumor extension, radiation, and surgical approach were found to be significant factors influencing all-cause mortality. Multivariate Cox regression analysis also revealed that ITC showed no significant risk for all-cause mortality compared to CPTC and FTC (Table 3).

Adjusting for patient characteristics using PSM
The cancer-specific mortality rate of patients with ITC was significantly higher than that of patients with CPTC and FTC (p < 0.001). The all-cause mortality rate of patients with ITC was also higher than patients with CPTC and FTC (p < 0.001; Figure 1A-1D). To minimize selection bias, propensity scored matching analysis was performed for age, sex, race, T/N/M stage, multifocality, tumor extension, and radiotherapy approaches. On survival analysis, patients with ITC had a better prognosis for cancer-specific survival compared to patients with CPTC and FTC after PSM for age, sex, and race (p < 0.001 for both; Figure 2A-2B). Furthermore, there were no significant differences in cancer-specific mortality between ITC and CPTC patients after PSM for age, sex, race, T/N/M stage, multifocality, and tumor extension (p = 0.731; Figure 3A). Patients with ITC were observed to have cancer-specific survival rates similar to patients with FTC (p = 0.640; Figure 3B). After matching for all potential influencing factors (including radiotherapy), the prognosis for patients with ITC was similar to that for patients with FTC, but better than that for patients with CPTC (p = 0.177 and p = 0.021, respectively; Figure 4A-B). According to survival analysis for all-cause mortality, ITC had a better prognosis compared to CPTC and FTC after matching for age, sex, and race (all p < 0.001; Figure 5A-5B). Similar results were obtained after matching for age, sex, race, T/N/M stage, multifocality, and tumor extension between CPTC and ITC (p < 0.001; Figure 6A). However, the prognosis of patients with ITC showed no statistical difference compared to those with FTC (p = 0.640; Figure  6B). After matching for all influential factors including radiation treatment, patients with CPTC and FTC showed poorer all-cause mortality rates compared to patients with ITC (both p < 0.001; Figure 7A-7B).

DISCUSSION
ITC is a distinctive and rare thyroid cancer that is characterized by nests or "insulae" of small uniform neoplastic cells, and has a propensity to metastasize to regional lymph nodes and distant sites [11]. ITC is considered morphologically and biologically intermediate compared to well-differentiated thyroid carcinomas (PTC and FTC) on one hand and anaplastic carcinoma of the thyroid on the other; it was incorporated into the World Health Organization classification of poorly differentiated thyroid carcinoma in 2004 [21,22]. According to the literature, disease-related mortality due to ITC is higher than that due to DTC [15,[23][24][25], and its insular histotype is independently associated with patient survival [26,27]. www.impactjournals.com/oncotarget Because of sample size limitations owing to the overall rarity of ITC, the paucity of available studies have precluded the accurate determination of this disease's prognosis.
In our analysis of the SEER data, histological subtype was not an independent risk factor for cancerspecific or all-cause mortality. Before performing PSM, ITC was observed to have a poorer prognosis than CPTC and FTC; however, after adjusting for all potentially influencing factors including demographic data and radiation treatment, the cancer-specific mortality rate of patients with ITC was similar to that of patients with FTC and better than that of patients with CPTC. Likewise, the all-cause mortality rate in patients with ITC was better than those in patients with CPTC and FTC. This result was unexpected and maybe have potential prognostic and therapeutic implications for patients with ITC , as it was inconsistent with previously published data [15,24,25,28,29] ,and it maybe more trustworthy because we performed PSM analyses to adjust for biases from different baseline characteristics of two histotype groups , Surgeons have not yet established the best treatment option for ITC, and our results may provide guidance for clinicians and may also influence the new American Thyroid Association guidelines.   Patients with ITC have reportedly been recommended for total thyroidectomy, lymph node dissection, and radioactive iodine owing to a high incidence of nodal involvement as well as their poor prognoses [15,30]. However, in our current study, the prognosis of patients with ITC was better than that of CPTC and FTC after matching for all potentially influencing factors, and only 24% of patients with ITC experienced lymph node metastasis. Therefore, patients with ITC may not require aggressive treatment unless certain risk factors such as multifocality, tumor extension, and lymph node metastasis (as detected by preoperative ultrasonography) exist.
Previous molecular studies have exposed the roles of various oncogenes in different types of thyroid tumors [31,32]. The discovery of novel molecular-based prognostic markers for thyroid cancer, such as RET-PTC, RAS, BRAF (V600E), PTEN and TERT mutations, are among the most notable developments in thyroid cancerrelated medicine [33]. However, the role of molecular mechanisms and genetic events in determining the growth and biologic aggressiveness of ITC remains to be explained. Pilotti et al. [16] reported RAS gene family point mutations involving a high proportion of CAA→AAA transversions at codon 61 of the N-RAS gene in ITC; however, the same abnormality was found with similar frequency in invasive follicular thyroid carcinoma [34].
Likewise, mutations in the p53 gene are also nonspecific, as they have been described in both patients with ITC and in those with anaplastic carcinoma [23,35]. The further exploration of the molecular changes in ITC may be helpful for tailoring therapeutic interventions, and patient enrollment in clinical trials should be encouraged when systemic therapy for disseminated disease is required.  Our study had several limitations. First, our dataset lacked information on recurrence, which may have led to the overestimation of cancer-specific and all-cause mortality rates. Another limitation is that family history, vascular invasion, and other histologic findings were not evaluated or considered in our study. Furthermore, the SEER is a United States database, and the findings may not be applicable worldwide.
In summary, we found that patients diagnosed with ITC had similar prognoses as those with FTC and better prognoses than those with CPTC; this was contrary to the findings of previous (albeit smaller-sized) studies. Our findings ought to serve as a reference for the planning of future therapies by adjusting the intensity of the interventions based on the prognosis of this disease being more favorable than previously thought.

Ethics statement and study population
This investigation was conducted in accordance with the ethical standards set by the Declaration of Helsinki, and pursuant to national and international guidelines. The study was approved by Union Hospital's institutional review board.
We used the National Cancer Institute's SEER database to investigate patients with thyroid cancer, including CPTC, ITC, and FTC. The SEER project began in 1973; it is the only comprehensive source of population-based cancer information in the United States, and is supported by both the Centers for Disease Control and Prevention and the National Cancer Institute. This database reports cancer-related data that include the incidence, prevalence, mortality, population-  based variables, and primary tumor characteristics (i.e., histological subtype) from multiple geographic regions within the United States.

Data collection and analysis
Patients diagnosed with CPTC, ITC, and FTC between 2004 and 2013 were identified using the SEER database and combinations of the International Classification of Diseases for Oncology site code C73.9 (i.e., thyroid, papillary, and/or follicular histology). The diagnosis codes included in the study were: "papillary carcinoma", "papillary adenocarcinoma", "follicular carcinoma", "papillary carcinoma, insular", "follicular adenocarcinoma", and "papillary and follicular adenocarcinoma". Thyroid cancer-specific and all-cause survival rates were compared among 66323 patients with CPTC, ITC, and FTC. Age, sex, race, T/N/M stage, multifocality, tumor extension, and radiotherapy (i.e., none or refused, external beam radiation therapy, and radioactive I-131 ablation) were evaluated in patients with different histological subtypes.

Statistical analyses
Patient survival curves depicting thyroid cancerspecific and all-cause mortality were calculated by using Kaplan-Meier analyses; the log-rank test was used to determine statistical differences in survival. PSM analysis was conducted to control for potential baseline confounding factors. The propensity score analysis generally calculated using logistic regression with group as dependent, covariates as independent variables [36,37]. Cox proportional hazards regression analyses were performed to estimate hazard ratios with 95% CIs and to show the magnitude of the effect of different histological subtypes on cancer-specific and all-cause mortality rates. All p-values were 2-sided, and p-values < 0.05 were considered statistically significant throughout the study. Analyses were performed using the SPSS software version 23.0, Stata/SE version 12 (Stata Corp.), and GraphPad Prism version 6 (GraphPad Software Inc.).