Survival and Analysis for Mixed Medullary-Follicular Thyroid Carcinoma: based on SEER database

Purpose Due to lack of proper diagnostic tools, we aimed to establish a nomogram for Mixed Medullary-Follicular Thyroid (MMFTC) and comparison with AJCC staging in prognosis. Methods Data regarding 203 patients with MMFTC (ICD-O-3 codes 8346, 8347) between 2000 and 2016 from The Surveillance, Epidemiology, and End Results (SEER) database. X-tile program was used to evaluate the optimal cut-off values for continuous variables. Univariate and multivariate regression analyses were performed with the Cox proportional hazards regression model to analyze the independent factors related to prognosis. Construct cancer-specic survival (CSS) and overall survival (OS) were analyzed. The resulting values were compared with the nomogram and the American Joint Committee on Cancer (AJCC) staging using C-index, verication curve, internal validation and decision curve analysis (DCA). nomograms


Introduction
Thyroid carcinoma is the most common endocrine malignancy and more than 50,000 new cases have been diagnosed in the United States in 2019, with an increasing mortality rate (1). Thyroid carcinoma is generally classi ed into two histopathological groups depending on their origin: one group of Follicular cell-derived carcinoma (papillary carcinoma and follicular carcinoma) and the other group of parafollicular C cell-derived carcinoma (medullary carcinoma). Despite different origins, carcinomas of various origins can be found to coexist in the same thyroid gland; this principally involves two cases: the same thyroid gland comprises multiple carcinomas of various origins, which are independent of each other, and the single carcinoma consists of various non-homologous components, which also includes two cases: the non-homologous components of the carcinoma are adjacent to each other at their surfaces, and the non-homologous components are mixed with each other, namely, MMFTC (2). MMFTC is an extremely rare malignant epithelial carcinoma. According to the 1988 World Health Organization Classi cation of tumors of endocrine organs, it can be de ned as "a tumor showing the morphological feature of medullary carcinoma together with immunoreactivity for calcitonin, and the morphological features of follicular thyroid carcinoma together with immunoreactivity for thyroglobulin" At present, most studies on MMFTC were presented in the format of literature review or case report, which focused on the pathologic characteristics of the carcinoma (4,5,6). Therefore, it is particularly important to acquire medical data for MMFTC from large databases. The present study, with the aim of investigating the survival and analysis of prognostic factors, offers a novel perspective on MMFTC.

SEER database
The data was obtained from SEER, which has been established by the National Cancer Institute. The database represented approximately 34.6% of the US population (7). The SEER is a premier source of population-based cancer information in the US that includes therapy information, morphological features, survival data, and cancer incidence (8).

Patient selection
We selected the patients diagnosed with MMFTC (ICD-O-3 codes 8346, 8347) between 2000 and 2016 from the SEER database. The considered variables included year of diagnosis, age at diagnosis, race, sex, tumor size, extrathyroidal extension, multifocality, AJCC staging information (version 8), surgical approach, radiation, Chemotherapy, survival months and vital status. We excluded the patients with missing information, as well as those without survival time and vital status.

Statistical analysis
Descriptive statistics were used to analyze the basic characteristics of the selected patients with MMFTC.
The p-value of the Categorical variables was determined by Pearson χ2 or the exact Fisher's tests. The optimal cut-off values for continuous variables (e.g., year of diagnosis, age at diagnosis, and tumor size) were determined through the use of X-tile program (Yale University, New Haven, Connecticut, USA).
Univariate and multivariate regression analyses were performed utilizing the Cox proportional hazards regression model to analyze the independent factors related to prognosis. Meanwhile, 95% con dence intervals and hazard ratios were calculated and analyzed with P < 0.05 considered statistically signi cant. The above statistics analyses were conducted under SPSS version 26.0 (IBM Corp., Armonk, NY). Multivariable Cox regression was performed to draw CSS and OS nomograms, and to evaluate the accuracy of the predicted nomograms by calculating the C-index, drawing a veri cation curve and conducting internal validation. Finally, the clinical utility of the predictive model was evaluated by the DCA, and the DCA of the AJCC staging predictive model was drawn for comparison and the conclusion was drawn. Nomogram, veri cation curve, and DCA were developed and adjusted using R version 1.2.5033 (The R Development Core Team, Vienna, Austria) in the R Studio environment.

Results
The optimal cut-off values for continuous variables The optimal cut-off value of MMFTC patients' year of diagnosis, age at diagnosis, and tumor size was determined by the x-tile program ( gure 1). Based on the overall survival rate, the optimal cut-off value for the year of diagnosis was determined to be 2011, the optimal cut-off value for age and the optimal cutoff value for tumor size were determined to be 63 years and 49 mm, respectively.

Demographic and Clinicopathological Characteristics of MMFTC
A total of 203 individuals were included in this study (table 1), of which, 57 (28%) were mixed medullary follicular carcinoma and 146 (72%) were mixed medullary papillary carcinoma. According to the inclusion and exclusion criteria source, patient data were obtained from the SEER database. The basic information of the patient was shown in Table 1. The person demonstrated a large age span, ranging from 18 to 92 years old, and the optimal cut-off value for age was determined to be 63 years. In this study, 81 were male (39.9%) and 116 were female (60.1%). The ratio of male to female was about 1: 1.5. The selected race was mainly Caucasian, accounting for 85.2%. Interestingly, the difference between the two types of carcinoma was apparent only in the condition that the tumor foci number (p <0.001) and pT stage (p =0.033) are signi cantly different, of which mixed medullary papillary carcinoma is mainly multifocal cancer and mixed medullary follicular carcinoma is mainly single cancer; however, there was no signi cant effect on survival time (p = 0.989). The survival time of patients was signi cantly different, with a large standard deviation. The median survival time was 75 months, while the shortest survival time was only 3 months.

Risk factors for survival
Based on univariate and multivariate Cox regression analysis listed in table 2, independent prognostic factors of cancer-speci c mortality and all-cause mortality of MMFTC were determined, respectively. Allcause mortality is related to the year of diagnosis (p = 0.011), age at diagnosis (p = 0.010), tumor size (p = 0.013), extrathyroidal extension (p = 0.008), pT2 stage (p = 0.021) and Radioactive implants or Radioisotopes (p = 0.031), while tumor-speci c mortality is related to the year of diagnosis (p = 0.045), tumor size (p = 0.003), extrathyroidal extension (p = 0.009) and pN stage (p = 0.008).

Construction and Validation of the Nomograms for overall survival and cancer-speci c survival
The variables including year of diagnosis, age at diagnosis, tumor size, extrathyroidal extension, T stage, and radiotherapy information were adopted to construct the nomogram for OS, while the nomogram for CSS was developed based on the variables including year of diagnosis, tumor size, extrathyroidal extension and N stage. The C-indexes of the nomograms of the OS and CSS were 0.794 and 0.872, respectively. In the internal validation cohort ( gure 3), the C-indexes of prediction accuracy for OS and CSS were 0.796 and 0.873, respectively. Moreover, 3, 5, and 10 years of OS and CSS calibration curves were also plotted, and it was found that the predicted results of the nomogram are consistent with the actual observed results. The results of both the quantitative and graphical evaluations show the reliability of the nomogram. Besides, on DCA ( gure 4), the predictive model showed great net bene t compared with AJCC 8th edition over a wider range of threshold probabilities, indicating the favorable potential clinical effect.

Discussion
At the moment, the histogenetic characteristics and pathogenetic origin of MMFTC are still obscure and have remained controversial. Several hypotheses were proposed to explain the MMFTC; one of them is the stem cell hypothesis, which assumes that follicular cells and C cells are derived from the same original stem cell and have a common proto-oncogene, and the original stem cell is capable of differentiating towards both follicular and C-cell lineage (9). The second hypothesis is a collision, which was developed on the basis of the assumption that the carcinomas occurring simultaneously are just a coincidence, which is based on the high incidence of PTC in thyroid cancer; this hypothesis was proposed speci cally for the situation where two different carcinoma types collide within the same thyroid gland (10). Another is the eld-effect hypothesis, assuming that a common oncogenetic stimulus triggers the neoplastic transformation of two cell-derived carcinoma (11). In some studies, the hypothesis of divergent differentiation might be involved, where the C cells adjacent to the follicular phenotype by the acquisition of additional molecular defects (12). The last hypothesis is hostage; Volante et al. (13) found that MTC is covered with hyperplastic follicles. After the normal follicular cells are encapsulated in MTC, they are stimulated and embedded by trophic factors, and the follicular cells are trapped into tumoral phenotype during the proliferation process.
In our study, we have established the nomograms to estimate the OS and CSS of extremely rare cancer. MMFTC has its unique characteristics compared to other histotypes of thyroid cancer. First, the extremely low incidence rate has led to a vague understanding of MMFTC. From 2000 to 2016, there were about 100000 cases of thyroid cancer (14), of which, MMFTC were only diagnosed in 203 cases (0.2%), which was consistent with those previously reported by Papotti et al. (15) and Kashima et al. (16). In addition, the unique pathological characteristics impede the diagnosis, leading to more signi cant prognostic difference. It is worth noting that the average survival time is 74.62 ± 51.41. However, so far, the previous reports on MMFTC are mostly case-series analyses and case reports, there is no established standard treatment scheme and no multicenter randomized controlled trial to guide treatment decisions. Therefore, up to now, there have been no previous studies that speci cally develop MMFTC predictive model. X-tile program was used to provide the optimal cut-off value for continuous variables that affect the prognosis of tumors (17). The optimal cut-off value for the year of diagnosis was determined to be 2011 by X-tile program. Despite continuous improvements in medical standards, the detection rate of MMFTC remains increasing. At the same time, the increase of carcinogenic factors such as chemical radiation and environmental pollution has also had a signi cant impact on the survival rate of MMFTC. The age of diagnosis is an independent prognostic factor for OS by X-tile program, and we found that males have higher morbidity and mortality than females. It is considered that most male patients with thyroid cancer are often discovered in the middle or late stage (18). Meanwhile, we found some tumor-related factors, such as tumor size, extrathyroidal extension, multifocality, TNM clinical stage, and postoperative radiotherapy; these factors could serve as independent prognostic factors for OS or CSS.
Through taking into due account all the aforementioned factors, AJCC staging may not well predict MMFTC survival. Therefore, we developed a nomogram for predicting MMFTC by combining all independent prognostic factors. The reliability of the developed model can be veri ed by calculating the C-index, conducting internal validation, and drawing a veri cation curve. In the nomogram of OS, extrathyroidal extension and postoperative radiotherapy have higher risks of poor prognosis, while in the nomogram of CSS, tumor size and lymph node metastasis have higher risks of poor prognosis, indicating that postoperative risk increases when the tumor enlarges and breaks through the thyroid capsule and extrathyroidal extension occurs (19), and lymph node metastasis is the most common form of thyroid cancer metastasis. When the above occurs, combined thyroidectomy with radioactive implants or radioisotopes may improve the e cacy of treatment for MMFTC. Meanwhile, we found that the independent prognostic factors of CSS are signi cantly less than OS, and age as an independent prognostic factor only appears in the nomogram of OS. As such, elderly patients who suffer from thyroid cancer are likely to be more susceptible to other infectious diseases.
In this study, some potential limitations should still be highlighted and discussed herein. The main limitation results from the SEER database. Speci cally, the lack of recurrence information will result in a positive offset in the evaluation of OS and CSS; moreover, the prediction accuracy is also limited owing to the lack of molecular biomarkers (such as BRAF mutation, RAS mutation, etc.), calcitonin, and other related variable information. In addition, it is worth noting that the data for rare diseases are considerably sparse and a validation cohort is absent in our study (20). To minimize the effect of this bias, we used all available data and comprehensively analyzed the relevant variables. However, compared with the AJCC staging, the nomogram based on independent prognostic factors shows satisfactory prediction accuracy.
Besides, DCA demonstrates the excellent clinical applicability of the predictive nomogram in our study. Analogously, several studies on different types of thyroid cancer have utilized DCA to verify the bene ts and clinical utility of the predictive capacity of the established models (21,22,23). These studies suggest that the nomogram model is a superior risk prediction method in comparison with AJCC staging in terms of both OS and CSS, allowing clinicians to develop more appropriate and e cient therapeutic strategies for MMFTC patients.

Conclusions
Although MSTC is an extremely rare disease in multiple countries, the incidence of thyroid cancer have shown increasing trends across in the recent years, more and more patients cases of thyroid cancer being described in MSTC. Meanwhile, with further exploration of the molecular mechanism of MMFTC and development of clinical technology, the means of tumor treatment are continuously improved and perfected. We are con dent of that, in the next few years, more accurate therapeutic methods will be developed for MMFTC and can achieve superior clinical outcomes.

Declarations Acknowledgements
The SEER database provides valuable data for manuscripts, Thanks to Dr Tian, Dr Xi, Dr Wang, Dr Miao, Dr Chen, Dr Huang and Dr Zhang for the contribution to this anuscripts, Authors'contributions WZ conceived, designed and wrote the initial draft of the manuscript. WZ, WB, MX and CZD performed the statistical analyses. TW, XHQ and WZ reviewed, revised and approved the nal version of the manuscript. All authors read and approved the nal manuscript.

Funding
This study is supported by Beijing Nova Program (No.Z181100006218011).

Availability of data and materials
The datasets used and/or analysed during the current study are available from the corresponding author upon reasonable request.

Ethics approval and consent to participate
The article uses the seer database, which is publicly available free of charge, so ethical exemptions are available at the First Medical Center of Chinese People's Liberation Army General Hospital. Figure 1 Identi cation of optimal cut-off values of year of diagnosis (A,B), age at diagnosis (C,D), and tumor size (E,F) by X-tile program. The optimal cut-off value for the year of diagnosis was determined to be 2011, the optimal cut-off value for age was determined to be 63 years and the optimal cut-off value for tumor size was determined to be 49 mm .