Prognosis of the Keratinizing Squamous Cell Carcinoma of the Tongue Based on Surveillance, Epidemiology, and End Results Database

Background The objective of this study is to determine the prognostic factors of keratinizing squamous cell carcinoma of the tongue (KTSCC) and to establish a prognostic nomogram of KTSCC to assist clinical diagnosis and treatment. Methods This study identified 3874 patients with KTSCC from the Surveillance, Epidemiology, and End Results (SEER) database, and these patients were randomly divided into the training (70%, (n = 2711) and validation (30%, n = 1163) cohorts. Cox regression was then used to filter variables. Nomograms were then constructed based on meaningful variables. Finally, the concordance index (C-index), net reclassification index (NRI), integrated discrimination improvement (IDI), calibration charts, and decision-curve analysis (DCA), were used to evaluate the discrimination, accuracy and effectiveness of the model. Results A nomogram model was established for predicting the 3-, 5-, and 8-year overall survival (OS) probabilities of patients with KTSCC. The model indicated that age, radiotherapy sequence, SEER stage, marital status, tumor size, American Joint Committee on Cancer (AJCC) stage, radiotherapy status, race, lymph node dissection status, and sex were factors influencing the OS of patients with KTSCC. Verified by C-index, NRI, IDI, calibration curve, and DCA curve, our model has better discrimination, calibration, accuracy and net benefit compared to the AJCC system. Conclusions This study identified the factors that affect the survival of KTSCC patients and established a prognostic nomogram that can help clinicians predict the 3-, 5-, and 8-year survival rates of KTSCC patients.


Introduction
Tongue cancer refers to tumors occurring in the back of the tongue, two-thirds of the front of the tongue, the side edge of the tongue, and the bottom of the tongue. It is believed that the incidence of tongue cancer rises with age. However, there has been a noticeable rise among patients under 45, particularly among women, in recent years [1]. More than 90% of tongue cancers are tongue squamous cell carcinomas (TSCC), and their mortality rate has increased dramatically in recent years. Human papillomavirus, alcohol, and tobacco are indicated as risk factors for TSCC [2][3][4]. Te WHO divides oral tongue squamous cell carcinoma into four classes based on the level of keratinization and diferentiation of cancer cells. Grade I is well diferentiated and has >50% keratinization (K4), grade II is moderately diferentiated and has 20-50% keratinization (K3), grade III is poorly diferentiated and has 5-20% keratinization (K2), while grade IV is undiferentiated and has 0-5% keratinization (K0-K1) [5]. Keratinizing squamous cell carcinoma of the tongue (KTSCC) is classifed as having the keratinizing subtype if any amount of keratinization is present, which accounts for approximately 50% of TSCC [6]. Te prognosis of diferent histological types of TSCC is signifcantly diferent. Te 5-year survival rate of KTSCC is >80%, while the 5-year survival rate of hypokeratinizing and nonkeratinizing squamous cell carcinoma is >50% [1,7]. Tough KTSCC has a relatively good prognosis and low malignancy, the tongue has abundant blood vessels and lymph nodes, and the primary tumor is prone to metastasizing. Because of this, Cooper et al. found that KTSCC had both lower disease-free and disease-specifc survival rates compared with other nonkeratinizing oral squamous cell carcinoma, indicating that keratinization is an important factor that afects the prognosis [8]. However, previous research into clinical prognosis assessments of patients with KTSCC have been inadequate, and hence a more accurate clinical prognosis assessment scheme is still needed to improve the reliability of predictions by doctors and improve the survival rate of patients.
Te current standard method for evaluating the prognosis of various tumors is the unifed tumor staging system developed by the American Joint Committee on Cancer (AJCC) [9,10]. However, this method is only based on the histological characteristics of tumors and ignores human sociology and individual characteristics, such as sex, race, living environment, and other potentially important factors, which often results in highly inaccurate predictions [11]. Prognostic nomograms not only incorporate oncological characteristics but also combine the demographic and sociological characteristics of patients to more reasonably and accurately estimate their prognoses, making them more valuable for clinical applications [12]. Terefore, our study established a KTSCC prognostic nomogram using data from KTSCC patients in the Surveillance, Epidemiology, and End Results (SEER) database. Te nomogram can be used to guide clinicians in evaluating the prognosis of KTSCC patients.  [13,14]. Te examined variables included the third edition of the International Classifcation of Cancer Diseases (ICD-O-3), age, tumor size, AJCC stage, race, SEER summary stage, sex, surgery status, marital status, radiotherapy status, lymph node dissection status, radiotherapy sequence, chemotherapy status, income, and vital status. Diagnoses made during 2004-2015 were selected. All relevant information was obtained from the SEER database (https:// seer.cancer.gov/), and consent was not required from the patients [15,16]. All-cause mortality in KTSCC was the outcome of this study.

Statistical Analysis.
Categorical variables were expressed as percentages, and diferences among the variables were evaluated. Continuous variables were expressed using median and interquartile values. A prognostic nomogram for KTSCC was then established according to the variables screened using Cox regression. Te concordance index (Cindex) and time-dependent area under the receiver operating characteristic curve (AUC) were used to test the discrimination performance between the predicted value from this model and the actual value. Quantitative evaluation indicators such as net reclassifcation improvement (NRI) and integrated discrimination improvement (IDI) index were also used to evaluate whether the predictive ability of the model was improved over that of previous models in a more comprehensive and multilevel manner [17,18]. Te accuracy of the nomogram was tested based on the ft between the calibration and standard curves [19]. Te validity of the nomogram was assessed using decision-curve analysis (DCA) [20].
SPSS Statistics software (version 27.0, Chicago, IL, USA) and R software (version 4.0.1; https://www.Rproject.org) were used for the statistical analysis, and statistical results were considered signifcant at p < 0.05.

Nomogram Creation and Evaluation.
We used the fnal results from the multivariate Cox regression to construct a nomogram ( Figure 2) for predicting the probabilities of overall survival (OS) at 3, 5, and 8 years in patients with KTSCC. Te results of the nomogram indicated that age has the greatest impact on KTSCC prognosis, followed by radiotherapy sequence, SEER stage, tumor size, AJCC stage, marital status, radiotherapy status, race, lymph node dissection status, and fnally, sex. Te next step was to evaluate the nomogram. Te Cindex in both the training (0.688) and validation (0.691) cohorts was better than that of the AJCC staging system (0.617 and 0.611). Te AUC values for the 3-, 5-, and 8year OS probabilities were 0.709, 0.723, and 0.742 in the training cohort and 0.716, 0.725, and 0.741 in the validation cohorts; the corresponding values for the AJCC staging system were 0.653, 0.644, 0.632 and 0.646, 0.621, 0.620, respectively. Tese fndings indicated that the new model had better predictive power than the AJCC staging system ( Figure 3).
Te Establish and evaluate the nomogram "C01.9", "C02.0", "C02.1", "C02.2", "C02.3", "C02.4", "C02.8" and "C02.9"  International Journal of Clinical Practice the training and validation cohorts were 10.5% and 10.6%, respectively, at 3 years (p < 0.001), 11.2% and 11.6% at 5 years (p < 0.001), and 11.5% and 11.9% at 8 years (p < 0.001). Te NRI and IDI values were both higher than zero, indicating that the predictive power of the new model is better. Te calibration curve indicated that the overall slopes of the curves of the training and validation cohorts were very similar to the reference line, and the points on the line were evenly distributed, indicating that our model had the strong predictive ability (Figure 4). Figure 5 shows the results from the DCA of the nomogram for patients with KTSCC. It indicates that the net beneft of this model in predicting survival probability was greater than that of the AJCC system, indicating that the nomogram that we have established is efective.

Discussion
KTSCC is a subhistological type of squamous cell carcinoma of the tongue. To the best of our knowledge, TSCC is the most researched at present, and few people study KTSCC independently and systematically [21,22]. Te incidence and mortality of KTSCC are increasing, and the age of onset is getting younger. It is important for the prognosis study of KTSCC.
Te AJCC system is currently available to assess the prognoses of most tumors but this method is based only on tumor histology, and does not take into account sociological and individual factors such as the economic status, race, and treatment conditions of patients. Our new nomogram model incorporated various comprehensive factors, including economic status, race, marital status, treatments, and other factors that independently afect the prognosis of KTSCC, and it can better predict the OS rate of patients with KTSCC. Notes. * p < 0.05; * * p < 0.01; * * * p ≤ 0.001, HR: hazard ratio, CI: confdence interval, RX summ-scope reg LN sur: regional lymph node dissection after surgery; RX summ-surg/rad seq: surgery and radiation sequence.
Comparing it with the AJCC staging system revealed that our new model can provide more accurate information, and for clinicians to use when performing diagnoses and formulating treatment plans for patients. Age has been an independent factor afecting TSCC prognosis. Best et al. [23] found that 56.1% of young patients with TSCC had lymph node metastasis. However, the prognosis deteriorated as age increased, which may be related to excessive drinking and smoking and other factors [24]. Te incidence and recurrence of TSCC in young patients have also recently been increasing [4]. Age was also the strongest risk factor for KTSCC prognosis in our new nomogram. For sex and race, like most TSCC [25], females with KTSCC have better OS than males, and black patients have signifcantly worse prognoses than white patients. We also found that single and divorced individuals had worse outcomes than married individuals. Studies have found that unmarried patients are more likely to have metastasis in various cancers, and the risk of death due to cancer also increases signifcantly, which may be related to poor compliance as well as the relatively low economic and educational levels of unmarried patients [26,27]. Studies have found that the early diagnosis rate of most cancers is higher for married patients than for unmarried patients [28]. We, therefore, plan to conduct future prospective experiments to confrm the exact relationship between radiotherapy status and KTSCC prognosis.
Tis study also found that among the factors infuencing KTSCC prognosis, tumor size, AJCC stage, SEER tumor stage, radiotherapy status, and postoperative lymph node dissection also afected the overall survival probability. Our fndings indicated that larger tumors caused worse KTSCC prognoses. Similarly, in SEER staging, more-widespread tumor metastasis is associated with a worse prognosis, with distant metastasis having the worst prognosis. Tese results were closely related to the special location of the tongue that is rich in blood vessels with its adjacent lymphatic vessels. With the development of tumors, localized invasion, and lymph node metastases often occur at an early stage, and distant metastasis eventually occurs. Tumor metastasis will adversely afect the prognosis.
It is worth noting that our study found that radiotherapy status and postoperative lymph node dissection were protective factors for KTSCC prognosis. Te current main treatment method for KTSCC is surgery. Postoperative selective neck lymph node dissection can increase the OS of patients and control recurrence. Consistent with the study by Kurita et al. [29] on oral squamous cell carcinoma, this study found that postoperative lymph node dissection was benefcial to the clinical prognosis of patients. However, surgical treatment can afect the speaking and eating functions of patients, while radiotherapy and chemotherapy can better protect their pronunciation, eating, and other functions and improve their quality of life [30,31]. Our nomogram also demonstrates that radiotherapy alone, or preoperative or     postoperative radiotherapies, could improve the OS rate of patients with KTSCC, but radiotherapy before and after surgery was a risk factor for imaging prognosis, indicating that the sequence and status of radiotherapy also have greater impacts on KTSCC prognosis. Tis may be related to excessive radiotherapy, which can destroy local normal tissues, cause swallowing disorders, hearing disorders, and other discomforts, or even accelerate tumor metastasis. Clinical treatments should therefore integrate various factors, control the frequency and sequence of radiotherapy within an appropriate range, and seek a trade-of between prolonging survival time and maintaining the quality of life. Tis will also be addressed in the next step of our research, that is, designing prospective experiments to clarify the relationship between radiotherapy and KTSCC prognosis.
We have established a new predictive model ( Figure 2) that was mostly based on individual patient conditions to predict the probabilities of OS at 3-, 5-, and 8-year after the diagnosis of KTSCC in individual patients. Clinical medical staf can calculate the total score for the nomogram based on the combined situation of each patient and then make more reasonable and reliable decisions based on the survival probability corresponding to that total score. C-index and AUC were used to test the discrimination of the new model, which would be found to have better discrimination ability. Te results for the quantitative indicators NRI and IDI demonstrated that the accuracy of our model in predicting survival probabilities was signifcantly better than that of the AJCC staging system. Te calibration curve plot of this model was found to be highly consistent with the standard curve, and the DCA plot indicated that the net beneft of our model in predicting the prognosis of patients with KTSCC was superior to that of the AJCC staging system [32]. Tis suggests that our model is of greater value in clinical applications.
Tis study had certain shortcomings. Selection bias: this study is a retrospective design, which can easily lead to selection and information biases. More sociological and biological factors need to be included in the research: clinicopathological factors, certain biological indicators, and patient behavior habits that afect prognosis are not included in the SEER database. Further confrmation is needed in prospective studies: prospective studies of the developed nomogram are still needed to further test and confrm its predictive power, which will be the focus of our future research.

Conclusion
Tis study has successfully established and verifed a prognostic nomogram for OS in KTSCC. Notably, age and the sequence and status of radiotherapy signifcantly afected KTSCC prognosis. Te nomogram can predict the 3-, 5-, and 8-year os of KTSCC patients. Terefore, it can help clinicians to formulate more reasonable treatment strategies. Area under the time-dependent receiver operating characteristic curve NRI:

Abbreviations
Net reclassifcation index IDI: Integrated discrimination improvement DCA: Decision-curve analysis AJCC: American joint committee on cancer OS: Overall survival HR: Hazard ratio CI: Confdence interval.

Data Availability
Te datasets generated and analyzed during the current study are available in the SEER database (https://seer.cancer. gov/). SEER is supported by the Surveillance Research Program (SRP) of the NCI Division of Cancer Control and Population Sciences (DCCPS). Te aim is to inform the science of cancer surveillance and the collection, analysis, interpretation, and dissemination of reliable populationbased statistics. SEER releases a standard set of research data every spring based on the previous November's submission of data from the registries. Te data we used is based on the November 2021 submission. We accessed these through the SEER * Stat software with additional approvals. Te data that support the fndings of this study are available from the SEER * Stat software, but restrictions apply to the availability of these data, which were used under license for the current study, and so are not publicly available. Data are, however, available upon reasonable request and with the SEER Research Data Agreement.