A Simple-to-Use Nomogram for Predicting Postoperative Early Death Risk in Elderly Patients with Spinal Tumors: A Population-Based Study

Background For elderly patients with primary spinal tumors, surgery is the best option for many elderly patients, in addition to palliative care. However, due to the unique physical function of elderly patients, the short-term prognosis is often unpredictable. It is therefore essential to develop a novel nomogram as a clinical aid to predict the risk of early death for elderly patients with primary spinal tumors who undergo surgery. Materials and Methods In this study, clinical data were obtained from 651 patients through the SEER database, and they were retrospectively analyzed. Logistic regression analyses were used for risk-factor screening. Predictive modeling was performed through the R language. The prediction models were calibrated as well as evaluated for accuracy in the validation cohort. The receiver operating characteristic (ROC) curve and the decision curve analysis (DCA) were used to evaluate the functionality of the nomogram. Results We identified four separate risk factors for constructing nomograms. The area under the receiver operating characteristic curve (training set 0.815, validation set 0.815) shows that the nomogram has good discrimination ability. The decision curve analysis demonstrates the clinical use of this nomogram. The calibration curve indicates that this nomogram has high accuracy. At the same time, we have also developed a web version of the online nomogram for clinical practitioners to apply. Conclusions We have successfully developed a nomogram that can accurately predict the risk of early death of elderly patients with primary spinal tumors undergoing surgery, which can provide a reference for clinicians.


Introduction
Primary malignant tumors of the spine have an extremely low incidence as a rare bone malignancy. It accounts for approximately 5% of bone tumors and less than 0.2% of all malignant tumors [1,2]. Osteosarcoma, chondrosarcoma, Ewing's sarcoma, and chordoma are the most common primary malignant tumors of the spine, accounting for approximately 35%, 30%, 16%, and 8.4%, respectively [1,3].
Although the cause of some of these tumors is unclear (e.g., Ewing's sarcoma), neoadjuvant and adjuvant treatment with surgical resection has been adopted as the treatment of choice in the National Comprehensive Cancer Network (NCCN) guidelines and has been shown to have a 5-year survival rate of 66.6% [4]. However, due to the rarity of primary malignancies of the spine, studies have been conducted in almost all age groups. Te elderly, as a population with a high prevalence of malignancy, have a higher risk of complications and a poorer prognosis after surgery due to deterioration in physical function and other underlying diseases [5,6]. How to improve the prognosis of elderly patients with spinal malignancies after undergoing surgery is a challenge for clinical practitioners.
Nomograms are predictive models widely used by clinicians and have been applied to predict the prognosis of many types of cancer [7][8][9][10]. Te Surveillance, Epidemiology, and End Results (SEER) database contains basic data on 28% of the US cancer patients, providing adequate and reliable sample for this study. Early death was considered to be death within six months of the initial surgery, and this defnition has also been applied in other studies [11,12]. To predict the early death in elderly patients with primary spinal tumors undergoing surgery, this nomogram was successfully constructed.

Patients and Data Collection.
Patient data from the retrospective study came from the National Cancer Institute's Surveillance, Epidemiology, and End Results database (SEER) database, which covers cancer information for 28 percent of the total US population [13]. Because our study is retrospective and does not involve patient's personally identifable information, ethical approval and patient consent are not required. We extracted clinical pathological data and demographic data from patients including age, race, sex, tumor grade, tumor stage, tumor histology, marital status, radiotherapy, chemotherapy status, and year of diagnosis.
Te inclusion criteria were as follows: (1) the primary site was the spine, (2) a diagnosis of malignancy was made, and (3) the patient who was 60 years of age or older. Finally, 651 elderly patients diagnosed with primary malignant spinal tumors between 1975 and 2015 were included.

Statistical
Analysis. Te total information entries are randomly divided into a training cohort and a validation cohort at a ratio of 7 to 3 by using R software. Univariate logistic regression analysis was used to screen out variables in the training cohort that might afect the prognosis of elderly patients with primary spinal tumors undergoing surgery, and then, the above variables were screened again using multivariate logistic regression analysis to exclude confounding efects between diferent factors. Based on the risk factors screened by multivariate logistic regression analysis, we used R software to construct a nomogram. Using R software, the training cohort and the validation cohort are used to plot the receiver operating characteristic (ROC) curve, and the area under the curve (AUC) is closer to 1, indicating that the performance of this nomogram is better. Nomogram was evaluated for clinical efcacy using decision curve analysis (DCA). Calibration curves are used to assess the accuracy of predictions. All statistical analysis and charting were carried out by using SPSS (25.0) and R software version 4.0.3. Te p value <0.05 is set to be statistically signifcant.

Patients' Demographic Characteristics.
Based on the inclusion criteria, a total of 651 elderly patients were included in this study. By the R software, all patients were randomized into a training cohort and a validation cohort in a 7 : 3 ratio. Detailed demographic and clinicopathological information of the above patients is presented in Table 1.

Associated Risk Factors.
After univariate logistic regression analysis, age, tumor grade, tumor histology, and tumor stage became risk factors. Subsequently, the above variables were analyzed by multivariate logistic regression, and after removing the mixed factors between variables, it was determined that age, tumor grade, tumor histology, and tumor stage were all independent risk factors afecting the prognosis of elderly patients with primary spinal tumors who underwent surgery ( Table 2).

Construction of Nomograms.
Nomograms that can predict the risk of early death in elderly patients with primary spinal tumors undergoing surgery are constructed based on four independent risk factors derived from the logistic regression analysis. Te risk factors of the patients were corresponding to each point of the horizontal axis, and the relevant risk factors were scored separately. Te prognostic prediction of the patient is obtained by adding the individual scores together to obtain a total score corresponding to the total score axis. We upload the nomograms that we have made and validated to the Internet so that they can be more easily used by more clinicians. Te URL is https://hzhspine. shinyapps.io/hzhspine/ ( Figure 1).

Evaluation of the Utility of Nomograms.
We evaluated this nomogram using the ROC curve and the DCA. Te AUC of the ROC curve of the training and validation cohorts was 0.815 and 0.815, respectively, showing that the model had good diferentiation ( Figure 2). On the other hand, the correction curves of both the training and validation cohorts obtained an ideal curve close to 45°, demonstrating that the prediction tool has some accuracy ( Figure 3). Decision-curve analysis shows that the nomogram can achieve an average good yield for patients under a certain risk threshold, which supports the clinical utility of this nomogram ( Figure 4). Finally, we risk-stratifed the cohort. Te results of Kaplan-Meier test demonstrated signifcant variability between the diferent risk strata ( Figure 5).

Discussion
Due to the rarity of primary spinal tumors, treatment strategies have only been progressively refned over the past two decades [14]. Even less attention has been paid to elderly patients. In the absence of relevant experience, reducing the risk of early death of elderly patients undergoing surgery for primary tumors of the spine is a major 2 Journal of Oncology challenge for clinicians. Terefore, the development of a nomogram to predict the risk of early death of these elderly patients is of great clinical relevance. In this study, tumor grade, tumor stage, histological type of tumor, and patient's age were found to be statistically signifcant indicators that had a signifcant impact on patients' shortterm prognosis. We can enter the above information for diferent individual patients into the nomogram to obtain an evaluation. It helps to take appropriate clinical measures according to patients' diferent risks of early death to improve prognosis. It has been established that the occurrence of primary malignancies of the spine is signifcantly correlated with age, and that cancer mortality increases dramatically with age from the age of 65 years [15,16]. Tis generally accepted view was also confrmed in the present study. In addition to the accumulation of exposure to predisposing factors, we believe that it is also related to the prevalence of underlying diseases in the elderly. Te accumulation of these carcinogenic factors and the presence of underlying disease make the elderly less tolerant of adjuvant or neoadjuvant therapy, surgery, and postsurgical rehabilitation and also have a greater likelihood of tumor metastasis [17].
In addition to age, logistic regression analysis showed that tumor grade, tumor stage, and histological type of tumor are also key factors. Te prognosis of survival in patients with primary tumors of the spine has been welldocumented to correlate with these factors, and many authors have constructed related nomograms [10,[18][19][20][21][22][23][24][25][26]. In our previous studies, we have shown that tumor grade, tumor stage, and histological type of tumor are the key factors infuencing postoperative OS in patients with primary spinal tumors, which is consistent with the results of this study [25]. However, almost all existing studies focus on all age groups, and no investigator has yet explored elderly patients as a study group. Although in the present study, radiation therapy and chemotherapy were considered to be irrelevant factors for prognosis, and the view is consistent with many fndings in the literature [10]. However, this still requires a different judgment on a case-by-case basis. In terms of tumor histology type, chondrosarcoma is resistant to radiation and chemotherapy, but multimodality therapy is still recommended in the treatment of Ewing's sarcoma and osteosarcoma [18,23,[26][27][28]. In contrast, a subgroup analysis of the study by Zhou et al. showed that the malignant outcome associated with chemotherapy was instead mostly osteosarcoma and chondrosarcoma [29]. However, certainly, the application of radiotherapy and chemotherapy is closely related to the histological type of the tumor. On the other hand, it has also been pointed out that for spinal confned tumors, chemotherapy is one of the signifcant worsening infuences, while for spinal metastatic tumors, chemotherapy has a good therapeutic efect [18,29]. Unlike chemotherapy, radiotherapy is thought to have a signifcant tumor metastasis suppression efect, signifcantly reducing the probability of metastasis in the group of patients who receive radiotherapy immediately after surgery [30,31].
Adequate presurgical evaluation is extremely important for elderly patients undergoing surgery for primary tumors of the spine [5]. Te nomogram developed in this study was able to predict the risk of early death of surgery in these elderly patients. It has great clinical implications for assessing whether patients can tolerate the physical challenges associated with surgery and postoperative rehabilitation. We must acknowledge the limitations of this study, particularly the unavoidable problem of bias in retrospective studies. But this nomogram can also be used as a reliable preoperative assessment tool for risk reduction.

Conclusions
We have developed a nomogram to predict the risk of early death of elderly patients with primary spinal tumors undergoing surgery.

Abbreviations
SEER: Surveillance, Epidemiology, and End Results OS: Overall survival AUC: Area under the curve DCA: Decision-curve analysis

Data Availability
Te publicly available datasets were analyzed in this study. Tese data can be found at the SEER dataset repository (https://seer.cancer.gov/).

Consent
Te study is based on data from public databases, and therefore does not require informed consent from patients.

Conflicts of Interest
Te authors declare that they have no conficts of interest.

Authors' Contributions
ZH H and YZ H conceived and designed the study. ZH H and Z Z performed the literature search. ZH H and YH L generated the fgures and tables. ZH H, ZG Z, and WF Z analyzed the data. ZH H wrote the manuscript, and YZ H critically reviewed the manuscript. ZH H and QQ K supervised the research. All the authors have read and approved the manuscript. Zhangheng Huang and Zhen Zhao contributed equally to this work.