Prognostic Factors and Treatment of Spinal Astrocytomas

A Multi-institutional Cohort Analysis

Yingjie Zou, MD; James Sun, BA; Yangying Zhou, BS; Harrison Xiao Bai, MD; Xiangyan Huang, MD; Ranjith Babu, MD; Alessandro Landi, MD; Kap Sum Foong, MD; Zishu Zhang, MD; John H. Woo, MD; Yongguang Tao, PhD; Xuejun Li, MD, PhD; Xiangqi Tang, MD, PhD; Bo Xiao, MD, PhD; Paul J. Zhang, MD; Li Yang, MD, PhD


Spine. 2018;43(10):E565-E573. 

In This Article


Multi-institutional Cohort

Our institutional ethical committee granted a waiver to retrospectively search our institutional pathology database. HIPAA compliance was maintained throughout this portion of the study. Our multi-institutional cohort consisted of 98 patients treated at four institutions between 1976 and 2016. The following variables were extracted from chart review: age, gender, extremity weakness/paresthesia/pain, bowel/bladder incontinence, WHO grade (I–IV), most proximal location of the tumor (cervical vs. thoracic vs. lumbar), pre- and postoperative McCormick grade, extent of surgical resection (GTR vs. STR), use of RT and/or chemotherapy, progression-free survival (PFS), and overall survival (OS).

Literature Search

Electronic database search was performed using PubMed. Search keywords included a combination of "astrocytoma," "spinal," "spine," "spinal cord," and "tumor" or their variations. Only human studies that were published in English from 1965 to 2016 were considered. References in relevant articles were examined to identify additional eligible studies. Our patient selection criteria are shown in Supplementary Figure 1,

Studies were included if they reported original research data on the long-term outcomes of individual patients with spinal astrocytoma. Studies were excluded if they (1) did not report individual patient data, (2) presented data that were already included in another study, (3) primarily focused on assessing the performance of surgeries to improve patients' daily function, (4) did not provide certain key patient information (e.g., extent of resection, etc.), and (5) were not original research articles. Two authors (Y.Z. and H.X.B.) examined the titles, abstracts, and full texts of all eligible studies independently to identify those that met the inclusion criteria. Discrepancies were reconciled between two authors in consensus. Literature studies included are listed in Supplementary Table 1,

Statistical Analysis

Analysis was performed on the multi-institutional cohort followed by an integrative analysis of the literature data. Summary statistics were performed for proportions and parametric variables (median and range). Baseline characteristics of the multi-institutional and literature cohorts were compared using Fisher exact test. Univariate followed by multivariate analysis was performed to identify factors associated with PFS and OS. Variables with a P value less than 0.10 on univariate analysis were included in the multivariable Cox proportional hazards regression. Differences in pre- and postoperative functional status, as measured by the modified McCormick score (Supplementary Table 2,, were compared using Wilcoxon signed-rank test. Subgroup analysis on change in McCormick score by extent of resection was performed using Fisher exact test. All statistical analyses were performed and figures generated using StataIC 14.2 (StataCorp, College Station, TX).