Warning: mkdir(): Permission denied in /home/virtual/lib/view_data.php on line 87 Warning: chmod() expects exactly 2 parameters, 3 given in /home/virtual/lib/view_data.php on line 88 Warning: fopen(/home/virtual/e-kjs/journal/upload/ip_log/ip_log_2025-02.txt): failed to open stream: No such file or directory in /home/virtual/lib/view_data.php on line 95 Warning: fwrite() expects parameter 1 to be resource, boolean given in /home/virtual/lib/view_data.php on line 96 A Nomogram for Predicting Overall Survival of Patients With Primary Spinal Cord Glioblastoma

A Nomogram for Predicting Overall Survival of Patients With Primary Spinal Cord Glioblastoma

Article information

Neurospine. 2024;21(2):676-689
Publication date (electronic) : 2024 June 30
doi : https://doi.org/10.14245/ns.2448082.041
1Department of Neurosurgery, The Second Affiliated Hospital of Soochow University, Soochow University, Suzhou, China
2Department of Neurosurgery, Xuanwu Hospital, Capital Medical University, Beijing, China
Corresponding Author Rujun Li Department of Neurosurgery, The Second Affiliated Hospital of Soochow University, Soochow University, Sanxiang Road, Gusu District, Suzhou 215004, China Email: newjun_li@163.com
Received 2024 January 15; Revised 2024 April 9; Accepted 2024 April 17.

Abstract

Objective

Primary spinal cord glioblastoma (PSCGBM) is a rare malignancy with a poor prognosis. To date, no prognostic nomogram for this rare disease was established. Hence, we aimed to develop a nomogram to predict overall survival (OS) of PSCGBM.

Methods

Clinical data of patients with PSCGBM was retrospectively collected from the neurosurgery department of Soochow University Affiliated Second Hospital and the Surveillance Epidemiology and End Results database. Information including age, sex, race, tumor extension, extent of resection, adjuvant treatment, marital status, income, year of diagnosis and months from diagnosis to treatment were recorded. Univariate and multivariate Cox regression analyses were used to identify independent prognostic factors for PSCGBM. A nomogram was constructed to predict 1-year, 1.5-year, and 2-year OS of PSCGBM.

Results

A total of 132 patients were included. The 1-year, 1.5-year, and 2-year OS were 45.5%, 29.5%, and 18.9%, respectively. Four variables: age groups, tumor extension, extent of resection, and adjuvant therapy, were identified as independent prognostic factors. The nomogram showed robust discrimination with a C-index value for the prediction of 1-year OS, 1.5-year OS, and 2-year of 0.71 (95% confidence interval [CI], 0.61–0.70), 0.72 (95% CI, 0.62–0.70), and 0.70 (95% CI, 0.61–0.70), respectively. The calibration curves exhibited high consistencies between the predicted and observed survival probability in this cohort.

Conclusion

We have developed and internally validated a nomogram for predicting the survival outcome of PSCGBM for the first time. The nomogram has the potential to assist clinicians in making individualized predictions of survival outcome of PSCGBM.

INTRODUCTION

Primary spinal cord glioblastoma (PSCGBM) is a rare disease, accounting for approximately 1.5% of intraspinal tumors [1]. To date, no standard treatment algorithm of PSCGBM was established. In contrast to its intracranial counterpart, PSCGBM was reported to have poorer survival outcome. The median survival time for PSCGBM is just about 9 months [2-4].

Several factors might contribute to its worse prognosis. Firstly, gross total resection (GTR) of PSCGBM is a great challenge due to dense nerve fiber in spinal cord and no definite margin between normal spinal cord and tumor [5]. As a result, patients with PSCGBM frequently had high postoperative residual tumor burden. Secondly, infertile blood supply of spinal cord might lead to insufficient chemotherapeutic drug permeability. Thirdly, MGMT promoter methylation, which is a prognostic marker for benefit from temozolomide (TMZ), infrequently occurred in spinal cord astrocytoma [6,7]. In conclusion, the current investigation is still ongoing to determine the precise efficacy of TMZ in the treatment of PSCGBM [1,8-10].

In previous studies, sex, ages, adjuvant treatment and surgical treatment were found to be prognostic factors of PSCGBM in Cox proportional hazards model [11-14]. However, Cox model could not be used to predict individual survival outcome and quantify survival probability. In recent years, nomograms are widely used for cancer prognosis [15-17]. As compared with traditional Cox regression model, a nomogram is a simple, visual and personalized scoring system for the prognostic prediction and can be used to predict individual survival probability. To date, no nomogram for predicting survival outcome of PGCGBM is established. Here, we have developed a nomogram for PSCGBM in our study to predict OS based on a large cohort.

MATERIALS AND METHODS

1. Study Population

Data were extracted from Surveillance Epidemiology and End Results (SEER) database (the Incidence-SEER 8Regs Custom Data, Nov 2021 Sub [1975–2019 varying] and IncidenceSEER 17Regs Custom Data, Nov 2021 Sub [2000–2019 varying]). Overlapped data between the 2 subdatabases were identified based on unique patient ID. Only patients who were diagnosed with glioblastoma (ICD-O-3 code: 9940, 9941) and lesions located at the spinal cord or cauda equina (ICD-O-3 code: C72.0 for “spinal cord”; C72.1 for “cauda equina”) were included. Additionally, patients who were diagnosed with PSCGBM and underwent surgery at the Department of Neurosurgery, the Second Affiliated Hospital of Soochow University, were also included. Patients meet the following criteria would be excluded: (1) metastasis, instead of primary lesion, which could be identified by sequence number and primary tumors were marked with “one primary only” nor “1st of 2 or more primaries”; (2) no surgery was performed or surgical strategy was unknown; (3) death from other causes or cause was unknown; (4) survival time was not available; (5) diagnostic confirmation was not based on pathological examination. The following data were collected: age groups, sex, race, tumor extension, extent of resection, adjuvant therapy, year of diagnosis, marital status at diagnosis, median household income (MHI) adjusted for inflation to 2019 and months from diagnosis to treatment. Detailed screening flow chart is shown in Fig. 1.

Fig. 1.

The final diagnosis flowchart for patient enrollment. SEER, Surveillance Epidemiology and End Results; GBM, glioblastoma.

The study was approved by the Institutional Review Board (IRB) of the Second Affiliated Hospital of Soochow University (IRB No. JD-HG-2024-047).

2. Definition of Variables

Age distribution was categorized into 3 groups: 5–17, 18–64, and ≥ 65 years. Race was divided into white, black, and other/unknown. Tumor invasion was stratified into localized, distant and unknown. Extent of resection was classified as biopsy, partial resection (PR), GTR, and unknown. Adjuvant treatment was divided into none, radiotherapy (RT) only, chemotherapy (CT) only, radiochemotherapy, and unknown. Years of diagnosis were categorized into 3 groups at 20-year intervals. Marital status was stratified into single (never married), married (including common law), and divorced/widowed. MHI inflation adjusted to 2019 was categorized as ≤ $50,000, $50,000–$59,999, $60,000–$69,999, ≥ $70,000 and unknown. Months from diagnosis to treatment were categorized into 2 groups based on whether patients received immediate treatment within 1 month. Survival outcome was dichotomized into alive and cancer-specific death.

3. Data Analysis and Diagnostic Prediction Model Building

Continuous variables were reported as mean± standard deviation or median ± interquartile range (IQR), as appropriate. Categorical data were presented as the frequency (percentage). Two-tailed t-test was used for normally distributed continuous variables and Mann-Whitney U-test for nonnormally distributed continuous variables; chi-square test or Fisher test was used for categorical variables. The Kaplan-Meier method was applied to calculate survival time and rates. The nomograms were built based on the results of multivariable Cox analyses of OS. The final model selection for the nomograms was performed by a backward step-down selection process using the Akaike information criterion. The performance of the nomogram was measured by the C-index. Calibration of the nomogram for 1-, 1.5-, 2-year survival was done by comparing the predicted with the observed survival. In the present study, the nomogram was subjected to 1,000 bootstrap resamples for internal validation. Furthermore, decision curve analysis (DCA) was performed to finalize the ranges of threshold probabilities within which the nomograms were clinically valuable by rmda (risk model decision analysis) package. A significance level of p< 0.05 was used to denote statistical significance. All statistical analysis was performed using R software (ver. 4.3.2, R Foundation for Statistical Computing, Vienna, Austria).

RESULTS

A total of 132 patients with PSCGBM were included (SEER database: 97 patients; our institute: 35 patients). 61 (46.2%) were female. The median age at diagnosis was 30 years (IQR, 15–46). The majority (78.8%) of lesions were localized, 7.6% of tumors exhibited distant dissemination. 71 (53.8%) and 47 (35.6%) of patients underwent PR and GTR, respectively. As to adjuvant therapy, 18 (13.6%) and 2 (1.5%) of patients received RT only and CT only, respectively, while 88 (66.7%) of patients received radiochemotherapy. The majority (77.3%) received treatment within one month after diagnosis (Table 1).

Demographic and treatment characteristics of patients with primary spinal cord glioblastoma

The median survival time was 11 months (Fig. 2). Kaplan-Meier survival curves by variable categories showed that age groups, extent of resection and adjuvant therapy were associated with survival outcome (Fig. 3A–J).

Fig. 2.

In primary spinal cord glioblastoma, the median survival time is 11 months. The overall survival rates at 1, 1.5, and 2 years were 45.5%, 29.5%, and 18.9%, respectively.

Fig. 3.

In the Kaplan-Meier survival curve analysis, we considered variables such as age groups (A), sex (B), race (C), tumor extension (D), extent of resection (E), adjuvant therapy (F), year of diagnosis (G), marital status at diagnosis (H), median household income (MHI) inflation adjusted to 2019 (I), and months from diagnosis to treatment (J). The analysis revealed significant differences only in age groups, adjuvant therapy and extent of resection.

Multivariate Cox regression analysis demonstrated that age groups (18–64 years: hazard ratio [HR], 0.59; 95% CI, 0.38–0.92; p = 0.021), tumor extension (distant: HR, 2.71; 95% CI, 1.33–5.52; p= 0.006), extent of resection (GTR: HR, 0.36; 95% CI, 0.14–0.93; p= 0.034), adjuvant therapy (radiochemotherapy: HR, 0.37; 95% CI, 0.16–0.85; p= 0.019; unknown: HR, 0.27; 95% CI, 0.09–0.78; p= 0.016) were independent predictors of OS (Table 2).

Univariate and multivariate Cox analysis of cancer-specific survival

Although the median survival time of PSCGBM is relatively short, of note, we observed that 45 patients had a survival time of at least 18 months. Therefore, it will be helpful for clinician to discriminate patients with relative better survival outcome from those with relative poorer survival outcome by using a nomogram. Based on independent predictors of OS identified by multivariable Cox analysis, a nomogram for predicting 1-year OS, 1.5-year OS, and 2-year OS was constructed (Fig. 4). In terms of discrimination of the nomogram model, the C-index value was 0.71 (95% CI, 0.61–0.70), 0.70 (95% CI, 0.61– 0.70), and 0.72 (95% CI, 0.62–0.72) for the prediction of 1-year OS, 1.5-year OS, and 2-year OS, respectively and comparable C-index values were confirmed through bootstrapping validation (C-index for the prediction of 1-year OS, 1.5-year OS, and 2-year OS: 0.74, 0.74, 0.75) (Fig. 5A–C). As respect to calibration, the calibration curves of the score system showed high consistencies between the predicted and observed survival probability in this cohort (Fig. 5D–F). Finally, DCA was performed to evaluate the clinical usefulness of the nomogram. When the predicted threshold probability was 80%–100% for 1-year OS, 1.5-year OS, and 2-year OS, application of this model to predict survival outcome could add more benefit than the treat-all or treat-none strategy (Fig. 5G–I).

Fig. 4.

The nomogram functions as a visual tool, effectively depicting the correlation between each variable and the annual survival rate. The variables in the nomogram design hold different degrees of importance, where a higher cumulative score indicates a lower annual survival rate. OS, overall survival.

Fig. 5.

The area under the curve (AUC) represents the area under the blue curve, providing a method for predicting annual overall survival (OS) accuracy. Specifically, the AUC values for 1 year (A), 1.5 years (B), and 2 years (C) were determined as 0.71, 0.70, and 0.72, respectively. Importantly, an AUC exceeding 0.70 indicates a superior predictive effect. In plots for 1 year (D), 1.5 years (E), and 2 years (F), greater alignment between the red line segment and the blue dashed line signifies higher accuracy in survival prediction by the model. Finally, the decision curve analysis was employed to validate the clinical efficacy of 1- (G), 1.5- (H), and 2-year OS rates (I). CI, confidence interval.

DISCUSSION

PSCGBM accounts for only 1%–5% of central nervous system glioblastomas and 1.5% of all spinal cord tumors [18]. Consistent with the findings of our study, previous studies reported that the disease has a poor prognosis with a median survival time of 12–14 months [1,19]. Due to its rarity, prognostic factors associated with OS of PSCGBM are not well understood. In our study, univariate and multivariate Cox regression analysis indicated that age groups, tumor extension, extent of resection and adjuvant therapy were predictors of PSCGBM. Likewise, a multicenter study by Inoue et al. [11] suggested that adolescent and young adult (HR, 3.53; 95% CI, 1.17–10.64), intracranial dissemination (HR, 4.30; 95% CI, 1.29–14.36), and no radiationtherapy (HR, 57.34; 95% CI, 6.73–488.39) were risk factors for mortality of patients with PSCGBM.

In contrast to intracranial GBM, PSCGBM frequently occurred in younger aged population and showed no sex predilection [3,19,20]. The median age in our cohort was 30 years (IQR, 16–45). Consistently, the study by Konar et al. [21] revealed that mean age of PSCGBM was 27 years and 51% of patients were below 18 years of age. However, in their study, age was not identified as an independent predictor of mortality. A retrospective study by Moinuddin et al. [14] included 190 patients with PSCGBM, the mean age was 40.8± 22.3 and age was found to be significantly associated with OS (p= 0.046).

The majority of PSCGBM primarily occurs in the cervical, thoracic and conus medullaris regions [22,23]. In line with our findings, Moinuddin et al. [14] found that extended lesion was associated with unfavorable survival outcome of PSCGBM. The study conducted by Ardeshiri et al. [24] also revealed that patients with more than 3 segments involved exhibited a higher likelihood of experiencing neurological deterioration in comparison to those with only one segment involved. It was reported that 40%–50% patients with PSCGBM could develop cerebrospinal fluid (CSF) dissemination and CSF dissemination was significantly related to poor OS [10,11,21]. Closer anatomic proximity of spinal tumors to the subarachnoid space compared to their intracranial counterparts might contribute to the high rate of CSF dissemination.

As to extent of resection, the role of surgical resection in PSCGBM is not well understood. The study of Lam et al. [13] indicated no statistically significant effect of the extent of resection on the length of survival among patients with PSCGBM. Even worse, McGirt et al. [3] and Wolff et al. [25] found that radical resection could worsen survival outcome of PSCGBM. Conversely, a large cohort study involving 208 PSCGBMs by Chalif et al. [26] demonstrated that GTR independently conferred a survival benefit to patients with PSCGBM (HR, 0.194; p< 0.001). Consistently, Kahn et al. [27] and Corradini et al. [28] also found that GTR could improve survival outcome of PSCGBM. Additionally, cordectomy, which is a more radical type of GTR and viewed as a salvage treatment, is expected to improve long-term survival by restricting or delaying intracranial dissemination of PSCGBM [3,29-36]. In our study, GTR was identified as a protective factor of favorable survival outcome (HR, 0.36; 95% CI, 0.14–0.93; p = 0.034). Due to no definite margin between normal spinal cord and PSCGBM, GTR without neurological compromise is a great challenge. Although GTR can improve patient outcomes, we need to make a balance between the patient’s neurological function and survival prognosis. In our view, for patients with intact neurological function (McCormick grade ≤ 3), we do not recommend pursuing complete tumor resection at the risk of functional compromise, while for paraplegic patients (McCormick grade= 5), we recommend GTR or even cordectomy in the case of preoperative informed consent is available and intraoperative frozen pathology indicates high-grade glioma.

The role of radiochemotherapy in PSCGBM is controversial. An aggressive approach involving whole-brain irradiation along with focal spinal irradiation has been suggested, even in the absence of evidence indicating intracranial dissemination [19]. However, Chalif et al. [26] found that radiation was not independently associated with improved survival of PSCGBM, while chemotherapy was significantly related to improved survival in patients with PSCGBM. In the retrospective study of Kaley et al. [37], it was found that both TMZ and bevacizumab could improve the survival rate. However, the impact of CT on PSCGBM is less pronounced as compared to cerebral GBM [1]. The study by Hernandez-Duran et al.38 did not show a significant relationship between TMZ and prolonged survival. Although the effect of RT on PSCGBM is not well established, RT was frequently prescribed to PSCGBM following surgical treatment, which repurpose the treatment strategy of intracranial GBM [17,22]. It is worth noting that the study by Inoue et al. [11] demonstrated that RT was associated with prolonged survival time of PSCGBM, but chemotherapy did not. In our study, only radiochemotherapy showed protective effect in patients with PSCGBM, while radiation only or chemotherapy only did not confer survival benefit. Likewise, the study by Cheng et al. [1] revealed that radiation plus TMZ could prolong survival time as compared to TMZ only or none (p = 0.002). Additionally, immunotherapy showed favorable efficacy in hematological malignant tumor [39]. In recent years, immunotherapy was tested in brain glioblastoma and shown promise in intracranial gliomas with some research suggesting benefit for spinal cord gliomas [40]. The application of immunotherapy, such as immune checkpoint inhibitors or chimeric antigen receptor T-cell therapy, to PSCGBM stands for future direction of research [41].

In our study, a nomogram was constructed to predict individual survival outcome of PSCGBM. In recent years, nomograms were widely used to predict survival outcome of other tumors, such as gastric cancer [42], non–small-cell lung cancer [43], hepatocellular Carcinoma [44] and proved to be an excellent tool of predicting individual survival outcome in recent years. As to generalizability, the development of our nomogram was based on a large public dataset and patients treated in our institute, which contained different populations. In addition, the nomogram showed good discrimination and calibration. Therefore, our nomogram has good generalizability. Although several Cox proportional hazards models of PSCGBM were developed and some survival predictors were found [11-14], the advantage of nomogram over traditional Cox hazard-proportional model is that nomogram could be used to predict individual survival outcome of PSCGBM and quantify the survival probability of an individual with PSCGBM. Therefore, this tool could be used to guide clinical decision-making based on quantized survival outcome and to discriminate patients with relatively better survival outcome and patients with relatively poorer survival outcome. Moreover, this tool could be applied to facilitate preoperative clinician-patient communication.

However, several limitations should be noted. Firstly, preoperative neurological findings and spinal lesion levels were not recorded in SEER database, which might influence the interpretation of our findings. Secondly, the dosage of RT and chemotherapeutic drug was not indicated in SEER database. Thirdly, H3 K27M mutation, a diagnostic and prognostic marker of diffuse midline glioma, H3 K27-altered, World Health Organization (WHO) grade 4, frequently occurred in primary spinal cord astrocytoma, but was not recorded in SEER database [45]. PSCGBM with H3 K27M mutation will be assigned an integrated diagnosis of diffuse midline glioma, H3 K27-altered, WHO grade 4. Therefore, integrated analysis without selecting out these cases with H3 K27M might confound our final result. Finally, due to the rarity of PSCGBM, no extra samples are available for an external validation. A multicenter study with large sample size is warranted to validate our findings.

CONCLUSION

A robust population-based survival-predicting model for PSCGBM is established and internationally validated. This nomogram offers clinicians a simple-to-use method for assessing mortality risk in patients with PSCGBM.

Notes

Conflict of Interest

The authors have nothing to disclose.

Funding/Support

This study received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Author Contribution

Conceptualization: YW, MS, FJ, RL; Formal analysis: YW, QM; Investigation: YW, MS, YC; Methodology: YW, QM, FJ, RL; Project administration: YW, QM, RL; Writing – original draft: YW, QM, MS; Writing – review & editing: YW, QM, YC, FJ, RL.

Acknowledgements

The authors express our sincere gratitude to Dr. Lei Cheng from Xuanwu Hospital, Capital Medical University for his invaluable contributions in project conception, language editing, and image enhancement.

References

1. Cheng X, Lou S, Huang S, et al. Primary spinal cord glioblastoma multiforme: a retrospective study of patients at a single institution. World Neurosurg 2017;106:113–9.
2. Santi M, Mena H, Wong K, et al. Spinal cord malignant astrocytomas. Clinicopathologic features in 36 cases. Cancer 2003;98:554–61.
3. McGirt MJ, Goldstein IM, Chaichana KL, et al. Extent of surgical resection of malignant astrocytomas of the spinal cord: outcome analysis of 35 patients. Neurosurgery 2008;63:55–60. discussion 60-1.
4. Stupp R, Brada M, van den Bent MJ, et al. High-grade glioma: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up. Ann Oncol 2014;25 Suppl 3:iii93–101.
5. Adams H, Avendano J, Raza SM, et al. Prognostic factors and survival in primary malignant astrocytomas of the spinal cord: a population-based analysis from 1973 to 2007. Spine (Phila Pa 1976) 2012;37:E727–35.
6. Mansouri A, Hachem LD, Mansouri S, et al. MGMT promoter methylation status testing to guide therapy for glioblastoma: refining the approach based on emerging evidence and current challenges. Neuro Oncol 2019;21:167–78.
7. Sun P, Fan DJ, Fan T, et al. A prospective clinical study on MGMT protein expression and the effect of gene promoter methylation on sensitivity to chemotherapeutics in spinal glioma. J Inflamm Res 2021;14:4777–84.
8. Stupp R, Mason WP, van den Bent MJ, et al. Radiotherapy plus concomitant and adjuvant temozolomide for glioblastoma. N Engl J Med 2005;352:987–96.
9. Stupp R, Hegi ME, Mason WP, et al. Effects of radiotherapy with concomitant and adjuvant temozolomide versus radiotherapy alone on survival in glioblastoma in a randomised phase III study: 5-year analysis of the EORTC-NCIC trial. Lancet Oncol 2009;10:459–66.
10. Tendulkar RD, Pai PA, Wu S, et al. Irradiation of pediatric high-grade spinal cord tumors. Int J Radiat Oncol Biol Phys 2010;78:1451–6.
11. Inoue T, Endo T, Muto J, et al. Shorter survival time of adolescents and young adult patients than older adults with spinal cord glioblastoma: a multicenter study. J Neurosurg Spine 2023;40:196–205.
12. Cheng L, Yao Q, Ma L, et al. Predictors of mortality in patients with primary spinal cord glioblastoma. Eur Spine J 2020;29:3203–13.
13. Lam S, Lin Y, Melkonian S. Analysis of risk factors and survival in pediatric high-grade spinal cord astrocytoma: a population-based study. Pediatr Neurosurg 2012;48:299–305.
14. Moinuddin FM, Alvi MA, Kerezoudis P, et al. Variation in management of spinal gliobastoma multiforme: results from a national cancer registry. J Neurooncol 2019;141:441–7.
15. Zhang W, Ji L, Wang X, et al. Nomogram predicts risk and prognostic factors for bone metastasis of pancreatic cancer: a population-based analysis. Front Endocrinol (Lausanne) 2022;12:752176.
16. Zaak D, Burger M, Otto W, et al. Predicting individual outcomes after radical cystectomy: an external validation of current nomograms. BJU Int 2010;106:342–8.
17. Wang Y, Li J, Xia Y, et al. Prognostic nomogram for intrahepatic cholangiocarcinoma after partial hepatectomy. J Clin Oncol 2013;31:1188–95.
18. Strik HM, Effenberger O, Schafer O, et al. A case of spinal glioblastoma multiforme: immunohistochemical study and review of the literature. J Neurooncol 2000;50:239–43.
19. Morais N, Mascarenhas L, Soares-Fernandes JP, et al. Primary spinal glioblastoma: a case report and review of the literature. Oncol Lett 2013;5:992–6.
20. Mallick S, Madan R, Julka PK. Primary spinal glioblastoma treated with adjuvant radiation and temozolomide: report of two cases. J Cancer Res Ther 2015;11:654.
21. Konar SK, Maiti TK, Bir SC, et al. Predictive factors determining the overall outcome of primary spinal glioblastoma multiforme: an integrative survival analysis. World Neurosurg 2016;86:341–8.e1-3.
22. Shrivastava RK, Epstein FJ, Perin NI, et al. Intramedullary spinal cord tumors in patients older than 50 years of age: management and outcome analysis. J Neurosurg Spine 2005;2:249–55.
23. Baleriaux DL. Spinal cord tumors. Eur Radiol 1999;9:1252–8.
24. Ardeshiri A, Chen B, Hutter BO, et al. Intramedullary spinal cord astrocytomas: the influence of localization and tumor extension on resectability and functional outcome. Acta Neurochir (Wien) 2013;155:1203–7.
25. Wolff B, Ng A, Roth D, et al. Pediatric high grade glioma of the spinal cord: results of the HIT-GBM database. J Neurooncol 2012;107:139–46.
26. Chalif EJ, Foster C, Sack K, et al. Impact of extent of resection and adjuvant therapy in diffuse gliomas of the spine. Spine J 2023;23:1015–27.
27. Kahn J, Loeffler JS, Niemierko A, et al. Long-term outcomes of patients with spinal cord gliomas treated by modern conformal radiation techniques. Int J Radiat Oncol Biol Phys 2011;81:232–8.
28. Corradini S, Hadi I, Hankel V, et al. Radiotherapy of spinal cord gliomas: a retrospective mono-institutional analysis. Strahlenther Onkol 2016;192:139–45.
29. Marchan EM, Sekula RJ, Jannetta PJ, et al. Long-term survival enhanced by cordectomy in a patient with a spinal glioblastoma multiforme and paraplegia. Case report. J Neurosurg Spine 2007;7:656–9.
30. Konar SK, Maiti TK, Bir SC, et al. Spinal cordectomy: a new hope for morbid spinal conditions. Clin Neurol Neurosurg 2017;152:5–11.
31. MacCarty CS, Kiefer EJ. Thoracic lumbar and sacral spinal cordectomy; preliminary report. Proc Staff Meet Mayo Clin 1949;24:108–15.
32. Kyoshima K, Ito K, Tanabe A, et al. Malignant astrocytoma of the conus medullaris treated by spinal cordectomy. J Clin Neurosci 2002;9:211–6.
33. Ewelt C, Stummer W, Klink B, et al. Cordectomy as final treatment option for diffuse intramedullary malignant glioma using 5-ALA fluorescence-guided resection. Clin Neurol Neurosurg 2010;112:357–61.
34. Viljoen S, Hitchon PW, Ahmed R, et al. Cordectomy for intramedullary spinal cord glioblastoma with a 12-year survival. Surg Neurol Int 2014;5:101.
35. Nakamura M, Tsuji O, Fujiyoshi K, et al. Cordotomy for patients with thoracic malignant astrocytoma. J Neurosurg Spine 2010;13:418–23.
36. Raco A, Polli FM, Palmieri M, et al. Spinal cordectomy for the management of thoracic malignant intraspinal tumors in paraplegic or irreversibly, severely paraparetic patients: a technical remark. J Clin Neurosci 2019;68:308–11.
37. Kaley TJ, Mondesire-Crump I, Gavrilovic IT. Temozolomide or bevacizumab for spinal cord high-grade gliomas. J Neurooncol 2012;109:385–9.
38. Hernandez-Duran S, Bregy A, Shah AH, et al. Primary spinal cord glioblastoma multiforme treated with temozolomide. J Clin Neurosci 2015;22:1877–82.
39. Tang L, Huang Z, Mei H, et al. Immunotherapy in hematologic malignancies: achievements, challenges and future prospects. Signal Transduct Target Ther 2023;8:306.
40. Grady C, Melnick K, Porche K, et al. Glioma Immunotherapy: advances and Challenges for Spinal Cord Gliomas. Neurospine 2022;19:13–29.
41. Hu J, Liu T, Han B, et al. Immunotherapy: a potential approach for high-grade spinal cord astrocytomas. Front Immunol 2021;11:582828.
42. Dong D, Tang L, Li ZY, et al. Development and validation of an individualized nomogram to identify occult peritoneal metastasis in patients with advanced gastric cancer. Ann Oncol 2019;30:431–8.
43. Liang W, Zhang L, Jiang G, et al. Development and validation of a nomogram for predicting survival in patients with resected non-small-cell lung cancer. J Clin Oncol 2015;33:861–9.
44. He Y, Luo L, Shan R, et al. Development and validation of a nomogram for predicting postoperative early relapse and survival in hepatocellular carcinoma. J Natl Compr Canc Netw 2023;22(1D):e237069.
45. Cheng L, Zhang F, Zhao X, et al. Mutational landscape of primary spinal cord astrocytoma. J Pathol 2023;260:317–28.

Article information Continued

Fig. 1.

The final diagnosis flowchart for patient enrollment. SEER, Surveillance Epidemiology and End Results; GBM, glioblastoma.

Fig. 2.

In primary spinal cord glioblastoma, the median survival time is 11 months. The overall survival rates at 1, 1.5, and 2 years were 45.5%, 29.5%, and 18.9%, respectively.

Fig. 3.

In the Kaplan-Meier survival curve analysis, we considered variables such as age groups (A), sex (B), race (C), tumor extension (D), extent of resection (E), adjuvant therapy (F), year of diagnosis (G), marital status at diagnosis (H), median household income (MHI) inflation adjusted to 2019 (I), and months from diagnosis to treatment (J). The analysis revealed significant differences only in age groups, adjuvant therapy and extent of resection.

Fig. 4.

The nomogram functions as a visual tool, effectively depicting the correlation between each variable and the annual survival rate. The variables in the nomogram design hold different degrees of importance, where a higher cumulative score indicates a lower annual survival rate. OS, overall survival.

Fig. 5.

The area under the curve (AUC) represents the area under the blue curve, providing a method for predicting annual overall survival (OS) accuracy. Specifically, the AUC values for 1 year (A), 1.5 years (B), and 2 years (C) were determined as 0.71, 0.70, and 0.72, respectively. Importantly, an AUC exceeding 0.70 indicates a superior predictive effect. In plots for 1 year (D), 1.5 years (E), and 2 years (F), greater alignment between the red line segment and the blue dashed line signifies higher accuracy in survival prediction by the model. Finally, the decision curve analysis was employed to validate the clinical efficacy of 1- (G), 1.5- (H), and 2-year OS rates (I). CI, confidence interval.

Table 1.

Demographic and treatment characteristics of patients with primary spinal cord glioblastoma

Variable Total (n=132) Alive (n=24) Cancer-specific death (n=108) p-value
Age (yr), median (IQR) 30 (15–46) 37.5 (24.0–53.5) 28 (14.8–46.0) 0.102
Age groups (yr) 0.106
 0–17 39 (29.5) 3 (12.5) 36 (33.3)
 18–64 83 (62.9) 18 (75.0) 65 (60.2)
 ≥ 65 10 (7.6) 3 (12.5) 7 (6.5)
Sex 0.967
 Female 61 (46.2) 11 (45.8) 50 (46.3)
 Male 71 (53.8) 13 (54.2) 58 (53.7)
Race 0.194
 White 75 (56.8) 10 (41.7) 65 (60.2)
 Black 7 (5.3) 1 (4.2) 6 (5.6)
 Other/unknown 50 (37.9) 13 (54.2) 37 (34.3)
Tumor extension 0.074
 Localized 104 (78.8) 23 (95.8) 81 (75.0)
 Distant 10 (7.6) 0 (0) 10 (9.3)
 Unknown 18 (13.6) 1 (4.2) 17 (15.7)
Extent of resection 0.087
 Biopsy 14 (10.6) 0 (0) 14 (13.0)
 Partial resection 71 (53.8) 12 (50.0) 59 (54.6)
 Gross total resection 47 (35.6) 12 (50.0) 35 (32.4)
Adjuvant therapy 0.936
 None 10 (7.6) 2 (8.3) 8 (7.4)
 Radiotherapy only 18 (13.6) 4 (16.7) 14 (13.0)
 Chemotherapy only 2 (1.5) 0 (0) 2 (1.9)
 Radiochemotherapy 88 (66.7) 15 (62.5) 73 (67.6)
 Unknown 14 (10.6) 3 (12.5) 11 (10.2)
Year of diagnosis < 0.001*
 1975–1994 10 (7.5) 0 (0) 10 (9.3)
 1995–2014 63 (47.7) 4 (16.7) 59 (54.6)
 2014–2023 59 (44.7) 20 (83.3) 39 (36.1)
Marital status at diagnosis 0.191
 Single (never married) 73 (55.3) 10 (41.7) 63 (58.3)
 Married (including common law) 50 (37.9) 13 (54.2) 37 (34.3)
 Divorced/widowed 9 (6.8) 1 (4.2) 8 (7.4)
MHI inflation adjusted to 2019 (USD) 0.397
 < 50,000 40 (30.3) 10 (41.7) 30 (27.8)
 50,000–59,999 21 (15.9) 5 (20.8) 16 (14.8)
 60,000–69,999 27 (20.5) 4 (16.7) 23 (21.3)
 > 70,000 36 (27.3) 5 (20.8) 31 (28.7)
 Unknown 8 (6.1) 0 (0) 8 (9.3)
Months from diagnosis to treatment 0.003*
 Immediate treatment within 1 mo 102 (77.3) 13 (54.2) 89 (82.4)
 No immediate treatment within 1 mo 30 (22.7) 11 (45.8) 19 (17.6)

Values are presented as number (%) unless otherwise indicated.

IQR, interquartile range; MHI, median household income; USD, United States dollar.

*

p<0.05, the groups exhibited statistically significant differences.

Table 2.

Univariate and multivariate Cox analysis of cancer-specific survival

Variable Univariate analysis
Multivariate analysis
HR 95% CI p-value HR 95% CI p-value
Age groups (yr)
 0–17 Reference
 18–64 0.64 0.42–0.97 0.034* 0.59 0.38–0.92 0.021*
 ≥ 65 1.26 0.55–2.85 0.586 1.21 0.48–3.05 0.690
Sex 0.92 0.63–1.34 0.651
Race
 White Reference
 Black 0.70 0.30–1.63 0.410
 Other/unknown 0.96 0.64–1.45 0.855
Tumor extension
 Localized Reference
 Distant 2.34 1.20–4.56 0.012* 2.71 1.33–5.52 0.006*
 Unknown 1.48 0.88–2.51 0.143 0.76 0.32–1.79 0.532
Extent of resection
 Biopsy Reference
 Partial resection 0.62 0.35–1.12 0.117 0.70 0.29–1.66 0.417
 Gross total resection 0.38 0.20–0.72 0.003* 0.36 0.14–0.93 0.034*
Adjuvant therapy
 None Reference
 Radiotheraphy only 0.48 0.20–1.17 0.106 0.45 0.17–1.24 0.123
 Chemotherapy only 1.85 0.39–8.80 0.438 2.23 0.46–10.81 0.320
 Radiochemotherapy 0.43 0.20–0.91 0.027* 0.37 0.16–0.85 0.019*
 Unknown 0.39 0.15–1.02 0.054 0.27 0.09–0.78 0.016*
Year of diagnosis
 1975–1994 Reference
 1995–2014 0.57 0.29–1.12 0.104 0.77 0.23–2.63 0.677
 2014–2023 0.46 0.23–0.93 0.031* 0.44 0.11–1.69 0.230
Marital status at diagnosis
 Single (never married) Reference
 Married (including common law) 0.76 0.51–1.15 0.196
 Divorced/widowed 0.89 0.42–1.86 0.748
MHI inflation adjusted to 2019 (USD)
 < 50,000 Reference
 50,000–59,999 0.92 0.05–1.68 0.777
 60,000–69,999 1.53 0.89–2.63 0.128
 > 70,000 0.83 0.49–1.39 0.474
 Unknown 1.74 0.79–3.81 0.166
Months from diagnosis to treatment
 Immediate treatment within 1 mo Reference
 No immediate treatment within 1 mo 0.63 0.38–1.03 0.067

HR, hazard ratio; CI, confidence interval; MHI, median household income; USD, United States dollar.

*

p<0.05, a statistical difference with the first subgroup within the group.