Introduction

Chordoma is a rare malignant neoplasm, arising from notochordal remnants, thus it is located almost exclusively in the axial skeleton [1]. It has an overall incidence of 0.08 per 100,000 individuals and accounts for 40 % of all primary sacral tumors [2]. Sacral chordoma is a typically slow growing and locally aggressive tumor, with a reduced ability to metastasize [3]. The diagnosis is often delayed because of the long standing, nonspecific initial symptoms, allowing the tumor to reach large sizes [4]. Because chordomas have shown to have a poor sensitivity towards radiotherapy and chemotherapy, they are mainly treated by surgical resection; a daunting undertaking given the complex, resource intensive, impairment inducing nature of the procedures and the high preponderance for local recurrence (LR) and eventual death [5].

Enneking oncologic management principles would recommend wide surgical en bloc resection of chordomas; however, this is difficult, even in the hands of the most experienced spine oncology surgeons [6]. Wide resection is not uniformly achieved in 35–75 % of cases, primarily due to the relatively inaccessible anatomical location, preference for neurological preservation and large size at the time of diagnosis [7–13]. The fact that chordomas grow in a lobulated fashion and have distant microscopic tumor outgrowths also makes wide surgical resection difficult [14]. Based on low-quality evidence, insufficient tumor resection is probably the main cause of LR and subsequently death [9, 10, 13]. Other factors that possibly influence survival and LR have been previously reported and include increased age, high sacral localization, lack of radiotherapy, prior resections, higher tumor grade, and increasing extent of tumor invasion [9, 10, 15–22]. Based on the dire consequences related to mortality and morbidity with the mismanagement of sacral chordomas, higher levels of evidence are needed to improve decision making and consequently patient outcome.

This study aims to identify prognostic factors that have an impact on the local recurrence-free survival (LRFS) and overall survival (OS) of surgically treated sacral chordomas (SC) patients from a multicenter ambispective database.

Methods

Study design

AOSpine International, through their Knowledge Forum Tumor conducted one of the first multicenter studies on primary spinal tumors [23]. They developed a database containing clinical and outcome data about surgically treated primary spinal tumor cases. A retrospective review of prospectively collected data, or ambispective design with cross-sectional follow-up, was performed by 12 of 13 leading spine oncology referral centers (Fig. 1); seven centers from North America (Johns Hopkins University School of Medicine, Baltimore, USA; University of British Columbia, Vancouver, Canada; MD Anderson Cancer Center, Houston, USA; University of Toronto, Toronto, Canada; Memorial Sloan-Kettering Center, New York, USA; Mayo Clinic, Rochester, USA; University of California San Francisco, San Francisco, USA), five from Europe (National Center for Spinal Disorders, Budapest, Hungary; Rizzoli Institute, Bologna, Italy; Queens Medical Centre, Nottingham, UK; Instituto Ortopedico Galeazzi, Milan, Italy; Oxford University Hospital NHS Trust, Oxford, UK), and one from Australia (Princess Alexandra Hospital, Brisbane, Australia). Patients met the inclusion criteria if they were diagnosed with a primary spinal tumor, received a surgical resection, and participated in at least one clinical follow-up. Patients with a secondary spinal tumor, spinal cord tumor, spinal lymphoma, or myeloma were excluded. The study was approved by the Scientific and Research Ethics Committee in each of the participating centers. The reporting of this work follows the guidelines of the strengthening the reporting of observational studies in epidemiology (STROBE) initiative [24].

Fig. 1
figure 1

Thirteen leading spine oncology referral centers. 1 University of British Columbia, Vancouver, Canada; 2 University of Toronto, Toronto, Canada; 3 University of California San Francisco, San Francisco, USA; 4 Mayo Clinic, Rochester, USA; 5 MD Anderson Cancer Center, Houston, USA; 6 Johns Hopkins University School of Medicine, Baltimore, USA; 7 Memorial Sloan-Kettering Center, New York, USA (this site did not contribute cases to the sacral chordoma cohort); 8 Queens Medical Centre, Nottingham, UK; 9 Oxford University Hospital NHS Trust, Oxford, UK; 10 Instituto Ortopedico Galeazzi, Milan, Italy; 11 Rizzoli Institute, Bologna, Italy; 12 National Center for Spinal Disorders, Budapest, Hungary; 13 Princess Alexandra Hospital, Brisbane, Australia. Centers in yellow contributed to the collection of sacral chordomas, while centers in gray did not

Full size image

Data collection

Data about demographics, baseline patient characteristics, surgical treatment, local disease recurrence, morbidity, and cross-sectional survival were gathered and entered into a database. The database was built and managed through a secure, web-based application (REDCap—Research Electronic Data Capture), hosted at AOSpine International [25].

Preoperative

Preoperative inpatient and outpatient clinical records were used to identify demographic and clinical data including age, gender, detailed medical history, preoperative symptoms, presence of pathologic vertebral fractures, and different neurological signs. Previous tumor surgery was defined as a surgical intervention beyond biopsy before the surgical resection. Motor deficit was assessed according to the Frankel or ASIA scales. Preoperative motor deficit was combined into two categories distinguishing the intact motor function (Frankel E) from paresis (Frankel D-A). Signs of cauda equina syndrome were also recorded. Results from imaging (CT, MRI, X-Ray, bone scan, PET-CT) and histological diagnosis were used to determine the localization, the local extension, and the oncologic stage of the tumor. The staging was performed according to the main categories of the Enneking surgical staging system (Ia—low grade malignant, confined to compartment; Ib—low grade malignant, invasive; IIa—high grade malignant, confined to compartment; IIb—high grade malignant, invasive; III—metastasis) [26].

Intraoperative

Intraoperative surgical data including surgical approach, nerve root and cauda equina sacrifice, type of resection, type of reconstruction, and the amount of blood loss were recorded. The parameters for type of resection (wide, marginal, intralesional or palliative) were determined by the surgeon. The surgeon’s impression about the surgical margins was validated by the pathologist during the histological analysis. The resections were also categorized according to the Enneking principles. Based on this classification, the resection of a malignant primary tumor such as chordoma (Enneking Ia, Ib, IIa, IIb, III) is defined as Enneking appropriate (EA) when a wide or marginal pathology is reported and Enneking inappropriate (EI) in the case of intralesional or palliative resection [27]. Tumor volume was measured on the histopathologic specimens. The height, width, and depth of the tumor were recorded, and the volume was calculated using the formula of an ellipsoid mass (volume = π/6 × height × width × depth) [28]. Tumor volume was transformed into a categorical variable where tumors were grouped as <100 and ≥100 cm3.

Postoperative

Follow-up data were obtained by direct examination of the patient and by performing the required imaging modalities. Follow-up data included any early and late postoperative complications, adjuvant radiation and chemotherapy, LR, any further surgeries for complications or recurrence, and current vital status. Postoperative complications were considered “early” if they occurred within 6 weeks after surgery and “late” if they occurred more than 6 weeks postoperative.

At the end of the study period, a cross-sectional follow-up of the vital status was performed in the form of an outpatient visit, telephone interview or accessing governmental vital statistic databases, if necessary.

Statistical analysis

The Kaplan–Meier method (K–M) was used to estimate OS and the LR. LR-free survival was defined as the length of time from the spine tumor surgery to the diagnosis of the first LR. The analysis was restricted to events that occurred within the first 10 years to adjust for patients who recently were diagnosed and had shorter follow-up times. Similarly, OS was defined as the length of time from the spine tumor surgery to death. Observations were censored when the patient was tumor free (LRFS analysis) or was alive (OS analysis) at the time of last clinical follow-up. The effect of individual variables on LR and OS was evaluated by assessing K–M curves with log-rank tests. To test for significance, selected continuous and categorical variables were re-categorized. Age, previous surgery, motor deficit, presence of cauda syndrome, tumor volume, adjuvant therapy, pathology, reconstruction, nerve root sacrifice, and tumor recurrence were evaluated. Variables with at least a marginally significant effect on survival (P < 0.1) were selected for the multivariate proportional hazards regression modeling. In the multivariate analysis, prognostic variables were identified when P ≤ 0.05. Statistical analysis was performed using STATA 12.0 software.

Results

Patient demographic and clinical characteristics

Between December 1985 and May 2012, a total of 1,495 primary spinal tumors were treated and the data were entered in the AOSpine Tumor Knowledge Forum Primary Spinal Tumor database (Fig. 2). Three hundred and forty-four patients had a chordoma and 173 patients received surgical treatment for a primary chordoma localized in the sacrum. Six patients who had Enneking Grade III (metastases) tumors were excluded from the study. Table 1 shows the demographic characteristics of the final cohort (167 patients). The male/female ratio was 98/69 with a mean age of 57 ± 15 years at the time of surgery (range 18–89). The majority of patients (n = 152, 96 %) presented with tumor-related spinal pain at the time of the diagnosis. Presence of motor deficit (Frankel/ASIA C and D) was also relatively common (n = 37, 24 %), and serious neurological deterioration was also a frequent symptom, where 41 (27 %) patients had cauda equina syndrome. Fifteen (9 %) patients had at least one previous spinal tumor surgery.

Fig. 2
figure 2

Flow-chart for patient selection

Full size image
Table 1 Demographic and clinical characteristics of 167 patients diagnosed with a primary sacral chordoma
Full size table

Tumor characteristics

Sixty-three (38 %) patients had chordomas with only sacral involvement, 89 (54 %) patients had sacrococcygeal chordomas, nine (6 %) patients had a sacral tumor involving the lumbar spine, and three (2 %) patients had only coccygeal chordomas (Table 2). The majority of the tumors (n = 128, 79 %) were Enneking Ib tumors (conventional and chondroid chordoma), and only 30 (19 %) tumors were Enneking IIb tumors (dedifferentiated chordoma). Only four (2 %) patients had a relatively small size tumor, that was confined only to the sacrum (two patients Enneking Ia and two IIa tumors). The mean tumor volume was 588.1 ± 1,423.1 cm3 (range 0.8–14,137 cm3).

Table 2 Sacral chordoma characteristics
Full size table

Treatment

The majority of tumors (n = 125, 76 %) were removed by a posterior only approach, 38 (23 %) by a combined anterior/posterior approach, and only two (1 %) were managed by an anterior only approach (Table 3). In 125 (82 %) patients, the sacrifice of one or more nerve roots was necessary during the tumor resection; in 10 (7 %) patients, the whole cauda equina was resected. The mean blood loss was 2,646 ± 3,613.5 ml (range 100–22,000 ml).

Table 3 Treatment details and outcome for the primary sacral chordoma patients
Full size table

The surgeon rated the intervention as marginal or wide in 131 (86 %) patients, and as intralesional in 21 (14 %) patients. The final pathologist rated specimen was widely or marginally resected in 129 (81 %) patients and intralesionally resected in 30 (19 %) patients. The difference between the two ratings was not significant (P = 0.34, Chi2 = 0.907, df = 1). Based on Enneking principles, 129 (81 %) patients had EA resection and 30 (19 %) patients had EI resection.

Thirty-nine (23 %) patients received adjuvant chemotherapy, conventional radiotherapy, carbon beam irradiation, or a combination.

Follow-up

The average follow-up of the patients was 3.2 years (range 5 days–16.2 years). The majority of the patients (n = 106, 63 %) were alive with no evidence of local or systemic disease at last clinical follow-up (Table 4). Twenty-six (15 %) patients were alive with evidence of local disease only, 11 (6 %) patients with systemic disease only, and 9 (5 %) patients with both local and systemic disease. Nine patients died due to propagation of the disease or due to disease-related complications. The cause of death for six patients was possibly unrelated to the sacral chordoma. The cross-sectional follow-up revealed that after the last clinical follow-up, 35 additional patients died from different causes.

Table 4 Vital and oncologic status of patients at last clinical follow-up and the cross-sectional follow-up
Full size table

Local recurrence analysis

Fifty-seven (35 %) patients had LR after surgery. The median LRFS was 4 years (Fig. 3a). In the univariate analyses, previous tumor surgery at the same site (P = 0.002), type of resection (P < 0.001), and tumor volume (P = 0.030), were significantly associated with LR (Table 5). When these three variables were combined in a multivariate model, previous surgery and type of resection were significantly related to LR (P = 0.048, HR = 2.05, CI 95 % = 1.00–4.18 and P = 0.009, HR = 2.43, CI 95 % = 1.25–4.73, respectively; Table 6). Undergoing a previous spine tumor operation and having an intralesional resection are associated with an increased risk of LR.

Fig. 3
figure 3

A Kaplan–Meier analysis of local recurrence (a) and survival (b), following surgery for all patients. Gray lines and shading represent 95 % confidence intervals

Full size image
Table 5 Results from the univariate (Kaplan–Meier) analysis evaluating survival and incidence of local recurrence over a 10-year period following surgical resection
Full size table
Table 6 Result of the multivariate Cox regression analysis of patient local recurrence and survival over a 10-year period following surgical resection
Full size table

Survival analysis

By the end of the study period, 50 (30 %) patients died and 117 (70 %) patients were alive. The median OS was 6 years (Fig. 3b). In the univariate analyses, age at surgery (P < 0.001) and motor deficit (P = 0.003) were significantly associated with OS (Table 5). The nerve root sacrifice was only trending towards significance (P = 0.088). When these three variables were combined in a multivariate model, age and motor deficit remained significantly associated with OS (P = 0.039, HR = 1.02, CI 95 % = 1.00–1.04 and P = 0.002, HR = 0.83, CI 95 % = 1.46–5.48, respectively; Table 6). Increasing age and a motor deficit of C or D are associated with a poor OS.

Discussion

Sacral chordomas (SC) are rare and thus difficult to manage and study. We report, to our knowledge, the largest multicentric ambispective cohort study of surgically treated SC. Our survival analysis of 167 patients with sacral chordoma assessed the effect of several variables both on LRFS and OS. The results from Kaplan–Meier and log-rank analyses were first evaluated to identify variables for multivariate Cox modeling. The multivariate model showed that EA surgery (en bloc resection with wide or marginal margins based on the pathology data) does improve the LRFS. Another interesting finding was the negative effect of previous surgery on LR. Furthermore, age and motor deficit (Frankel or ASIA score of C or D) were independently associated with poor survival.

The postoperative LR and the mortality can be influenced by several factors (Table 7). Several publications tried to identify prognostic factors, but the majority of these studies are statistically underpowered. In contrast, this study uses a large population-based multicentric database and statistical modeling to identify prognostic factors.

Table 7 Literature review on prognostic factors for OS and LRFS of sacral chordoma
Full size table

The first study which used survival analysis to assess the effect of different factors on LRFS in 21 surgically treated SC was published by Samson et al. in 1993 [17]. The authors used univariate Cox regression analysis and found old age to have an impact on LR, but only showing a trend towards significance. Cheng et al. reviewing their 31-year experience with sacral chordoma resection had similar findings, old age and higher sacral localization with or without lumbar involvement were independently associated with high LR [20]. In our analysis, old age had a negative impact only on OS.

In 1999, York et al. [9] reported a survival analysis of 27 surgically treated sacral chordoma cases. They assessed only the LRFS, which was negatively influenced in the univariate survival analysis by subtotal tumor resection and by the lack of radiotherapy after surgery. One year later in 2000, Bergh et al. [18] analyzed 39 consecutive patients, and found that inadequate surgical margins have a negative impact on LRFS and on disease-specific survival. In the case series of Fuchs et al. [10] the authors reported that surgical margins were the most important predictor of OS and LRFS. In 2010, Ruggieri et al. analyzing their institutional experience with sacral chordoma resection (56 patients during 30 years practice) found that surgical margins and previous intralesional surgery had a negative impact on LRFS. The inadequate surgical margin and the previous surgery was a prognostic factor for LRFS in our multivariate model. This indicates that EA resection reduces LR. In the group of patients who underwent EI resection, the occurrence of LR was higher 64 % versus the 29 % of cases where the EA resection was feasible (P < 0.001, Chi2 = 12.383, df = 1). This difference was not significant in the term of survival (Table 8).

Table 8 Prevalence of LR and mortality in the EA and EI groups
Full size table

McGirt et al. published the only population-based study until now, which assessed surgically treated chordoma patients (67 sacral chordoma and 47 mobile spine chordoma patients). They revealed from the SEER registry that increasing age, increasing extent invasion, more recent year of surgery and sacral localization is associated with poor survival in chordoma [16]. In the publications of Bergh and McGirt, large tumor size was a prognostic factor of poor survival. In our analysis, it was significant only in the univariate model (P = 0.03).

In the primary spinal tumor literature, there is no reference on the preoperative neurological deficit as a prognostic factor of mortality. However, Tokuhashi et al. [29] reported the severity of spinal cord injury as an important factor of poor prognosis in patients with secondary spinal tumors. In SC, neurological deficit is rare and is limited mainly to the L5, S1 nerve roots. In the most severe cases, the whole cauda equina can be affected. In our analysis, the presence of cauda equina syndrome was not a prognostic factor. In contrast, we identified Frankel or ASIA score below E as a negative prognostic factor for survival in the multivariate analysis. Another interesting observation is that patients with postoperative neurologic deficit due to planed nerve root sacrifice have poorer survival. However, this prognostic factor was significant only in the univariate analysis. Patients with neurologic deficit usually have an impaired quality of life (QOL) which was suggested to shorten the survival [30].

In the majority of these publications, old age and inadequate surgical margins were common prognostic factors for OS and LRFS. However, a common drawback in interpreting their results is that they used the simplest form of survival analysis (Kaplan–Meier and log-rank analysis) based on statistically underpowered studies and retrospective limitations. Only McGirt et al. used multivariate Cox regression modeling—which is the gold standard in survival analysis—to identify the possible prognostic factors associated with OS. The problem with this population-based study is that it lacked granularity due to its registry design, specifically around surgical details and pathology. In our multivariate Cox models, the number of events per variable was 19.5 and 16.7 in LFRS and OS analyses, respectively, which is superior than the literature recommendation [31].

Despite an ambispective design and dedicated, detailed data collection, our study has numerous limitations. The major limitation is with respect to follow-up. Based on best available literature, the current 5- and 10-year survival for chordomas is 72 and 48 % [32]. The follow-up in our study, therefore, is too short to specifically deal with the issue of OS. It is not unreasonable from a theoretical perspective however that if LR occurs the OS is likely reduced; only longer term follow-up data would answer this question. Similarly, the follow-up is probably a little early for LR, but the results of statistical significance of EA and decreased LR are probably very robust. Furthermore, the fact that the analysis was based on a retrospective review of prospective data constitutes a limitation. To overcome this, we performed a cross-sectional follow-up of the vital status at the end of the study period. The final limitations are around the error and variability in surgical and pathology impressions, which are difficult to control at rare conditions and a multicenter design. Multicenter collection has been initiated.

Due to the intensive research in oncology, the therapeutic strategies in the management of primary spinal tumors are changing. On one hand, the proton and carbon beam therapies are showing promising effect in the cases of otherwise radioresistant solid tumors (chordoma and chondrosarcoma). On the other hand, in the past years, some molecular pathways and possible target molecules were identified (e.g., brachyury in the case of chordoma), which can lead in the near future to the development of novel therapeutic agents. Until then, regardless of its morbidity, the surgical intervention is the treatment of choice in PSTs. To improve the surgical decision making, and to better understand the positive and negative effects of surgery, prospective multicenter studies are needed—incorporating health-related QOL assessment; but the results of this study would suggest that surgeons treating SC strongly adhere to EA surgical margins to minimize the risk of LR and its miserable, relentless sequelae.