Impact of Treatment Modalities Upon Survival Outcomes in Skull Base and Clival Chordoma: An NCDB Analysis

Objectives: Skull base chordomas are locally aggressive malignant tumors derived from the notochord remnant. There are limited large-scale studies examining the role and extent of surgery and radiation therapy. Design: Analysis of the National Cancer Database (NCDB) was performed to evaluate the survival outcomes of various treatments, and to assess for predictors of overall survival (OS). Participants: Retrospective, population-based cohort study of patients diagnosed with a clival/skull base chordoma between 2004-2015 in the NCDB. Main Outcome Measures: The primary outcome was overall survival (OS). Results: 468 cases were identified. 49% of patients received surgery and 20.7% had positive margins. Mean age at diagnosis was 48.4 years in the surgical cohort, and 55% were male. Of the surgical cohort, 33.8% had negative margins, 20.7% had positive margins, and 45.5% had unknown margin status. Age ≥ 65 (HR 3.07, 95% CI 1.63-5.76, p<0.001), diagnosis between 2010-2015 (HR 0.49, 95% CI 0.26-0.90, p=0.022), tumor size >5 cm (HR 2.29, 95% CI 1.26-4.15, p=0.007) and government insurance (HR 2.28, 95% CI 1.24-4.2, p=0.008) were independent predictors of OS. When comparing surgery with or without adjuvant radiation, no survival differences were found, regardless of margin status (P=0.66). Conclusions: Surgery remains the mainstay of therapy. Advanced age over 65, large tumor size, and government insurance were predictors of worse OS. While negative margins and the use of adjuvant radiation did not appear to impact OS, these may very well reduce local recurrences. A multidisciplinary approach is critical in achieving optimal outcomes in this challenging disease.


Introduction
Chordomas are locally destructive neoplasms that originate from the remnant of the notochord.
[1] While their appearance is benign on histopathology, they exhibit "malignant" behavior and are locally invasive, likely to recur, and are noted to disseminate in advanced stages of disease. [2] While the most frequent location of origin is the sacral spine, 30-40% of chordomas arise in the cranial base, most notably the clival and paraclival regions. [1] Chordomas are rare, with an estimated incidence of 0.08 per 100,000 based on prior population-based data. [3] Within the clivus, the lower third is the least likely primary site (upper third 72%, middle third 82%, lower third 42%), but the most frequent site of residual tumor. [4] Three histological subtypes of chordomas have been described, which are classical, chondroid, or dedifferentiated chordomas, all of which characteristically stain positive for epithelial markers on immunohistochemistry. [5] Left untreated, patient mortality associated with progressive local disease generally occurs in 12 months.
[6] Given the scarcity of this disease, the optimal therapeutic strategy continues to be elucidated, as do the clinical determinants of overall survival (OS). Treatment modalities that have been described for clival chordoma include surgery, radiation, and chemotherapy.
Given its infiltrative nature and predilection to invade critical structures, surgical resection with negative margins remains highly challenging to obtain for chordomas.

Accepted Manuscript
This article is protected by copyright. All rights reserved.
Additionally, in certain cases, the morbidity of large intradural resection (e.g., brainstem injury, basilar injury, cerebrospinal fluid leak) may preclude the ability to achieve negative margin status. [7] While recent database analysis has assessed the roles of radiation, chemotherapy, and surgery, there has not been to our knowledge an analysis of the clinical outcomes of positive versus negative margin status in the context of modern adjuvant care. [5] In addition, the role of socioeconomic factors with respect to OS has not been fully interrogated in this patient population. With these outcomes in mind, we undertook an analysis to assess the determinants of survival for clival chordoma in a large-sample, population-based database.

Materials and Methods
This research was a retrospective, population-based cohort study. Our data was obtained from the National Cancer Database (NCDB), a partnership between the American Cancer Society and the Commission on Cancer (CoC) of the American College of Surgeons, which collects data on a large proportion of newly diagnosed oncologic diseases in the nation every year and is a comprehensive clinical surveillance resource that includes >34 million records from patients diagnosed at over 1500 CoC-accredited programs. [8] This investigation was given Institutional Review Board exemption due to the public use and anonymity of patient data within

Accepted Manuscript
This article is protected by copyright. All rights reserved. criteria consisted of the following: 1) patients receiving other additional treatments, 2) treatment provided at a different facility from the diagnosing facility, 3) patients with additional malignancies, 4) patients receiving palliative care, and 5) patients with follow-up unspecified or less than 30-days from the start of treatment. Of note, within the NCDB additional therapy is used to identify treatment of hematopoetic diseases (such as phlebotomy and transfusions), and as such this modification does not exclude treatment with chemotherapy, radiation therapy, or immunotherapy. There were two patients who received adjuvant chemotherapy, one patient who received adjuvant chemoradiotherapy, and one patient who received adjuvant immunotherapy who were not included in multivariate analysis given the small size of these cohorts.
Cases were classified as either receiving surgery only (SO) or surgery with adjuvant radiotherapy (SXRT) depending on the therapy received and the surgery and radiation sequence as defined by the NCDB variable "RX_SUMM_SURGRAD_SEQ". SO with negative margins was interpreted as GTR without adjuvant radiation, SXRT and negative margins as GTR with adjuvant radiation, SXRT and positive margins as STR with adjuvant radiation, and SO with positive margins as STR without adjuvant radiation. Though margin status and macroscopic extent of resection are not interchangeable definitions, negative margin status generally implies that GTR has occurred. The use of negative margin status as a surrogate marker for total resection has previously been utilized in analyses of esthesioneuroblastoma, sinonasal mucosal melanoma, osseous-based skull and mandibular tumors, and skull base chondrosarcoma. [9][10][11][12] Clinical covariates included age (<65 or ≥65 years), sex, race (Caucasian, African American,

Accepted Manuscript
This article is protected by copyright. All rights reserved.
included facility type (academic or non-academic) and location (east, central, or west), insurance status (private, government, or uninsured), income quartile (<$48,000 or ≥$48,000), education level of residence (<13% or ≥13% without a high school diploma), population size (<250,000 or ≥250,000 individuals), and distance from provider to patient. Treatment-related covariates Statistically significant variables on univariate were included in the multivariate analysis. This study utilized a p-value of <0.05 for statistical significance.

Results
A total of 468 cases (213 SO, 255 SXRT) diagnosed from 2004-2015 were identified from the NCDB. 97 patients had treatment data available for at least seven years of follow up.
Baseline patient characteristics of clinical and sociodemographic covariates are presented in Table 1. Chi-squared analysis between SO and SXRT demonstrated significant differences for

Accepted Manuscript
This article is protected by copyright. All rights reserved.
year of diagnosis (p=0.031), facility type (p=0.014), facility location (p=0.002), and surgical margin status (p<0.001) ( Table 1) Kaplan Meier log-rank test comparing OS outcomes between patients receiving SO with positive margins, SO with negative margins, SXRT with positive margins, and SXRT with negative margins demonstrated no significant differences in OS (p=0.66) (Figure 1). The 1-, 2-, 5-, and 10-year OS for these cohorts are provided in Table 3. However, there was a significant difference in OS comparing patient insurance status (p<0.001), with patients on private insurance demonstrating improved OS compared to those on government insurance (Figure 2). Specific radiation modalities of intensity-modulated radiotherapy (IMRT; N=49), stereotactic radiosurgery (SRS; N=54), other photon (N=37), and proton beam (N=92) radiation were not associated with differences in OS (p=0.16). Similarly, cumulative adjuvant radiation therapy dosage using 60 and 70 Gy as thresholds were not associated with OS benefit (p=0.63 and p=0.50, respectively).

Accepted Manuscript
This article is protected by copyright. All rights reserved.
Surgery is the mainstay of therapy for clival and skull base chordomas, with a recent analysis of the NCDB and Surveillance, Epidemiology, and End Results (SEER) database by Holou et al. demonstrating that 86% of patients received surgery. [5] Numerous surgical approaches have been described for these tumors, including transcranial, transnasal, high anterior cervical retropharyngeal, and transoral techniques. [13][14][15] Endoscopic endonasal techniques in particular are associated with high rates of gross total resection (GTR) for midline disease up to 50-90%. [16][17][18] For disease necessitating resection of dura, reconstruction with vascularized tissue and employing a lumbar drain have been noted to decrease the incidence of CSF leak, though transclival defects generally lead to a high flow CSF-leak and are amongst the most challenging defects to primarily repair. [18,19] In light of the diversity of strategies and the frequency of operative intervention, developing an evidence-driven approach to surgical resection is paramount. Our findings contribute uniquely to this discussion in that no apparent survival benefit was found when comparing GTR versus STR, with or without adjuvant radiation, or regardless of margin status. For our analysis negative margin status was considered as indicating that GTR had occurred, as a limitation of analysis within the NCDB is that GTR is not a separately coded variable.  Figure S1). As GTR is not expressly coded for within the NCDB, and many prior studies have used differing definitions of GTR, caution must be taken in applying our findings too broadly given the robust body of research demonstrating improved survival and decreased recurrence in association with GTR. One possible explanation of our data is that the increased efficacy of adjuvant therapy has provided a similar survival benefit to that

Accepted Manuscript
This article is protected by copyright. All rights reserved.
previously accomplished by GTR. In light of these data, STR may be considered in cases where GTR entails significant morbidity, as long as adjuvant radiation therapy is provided. One example is in patients with limited intradural disease, where extradural resection may be considered followed by adjuvant radiation therapy, thereby avoiding the possible risks and morbidity associated with a transclival skull base defect.
With respect to radiation therapy, there is emerging literature that postoperative chordoma patients treated with adjuvant radiation therapy have improved overall survival with proton therapy at five years relative to conventional radiation therapy, and that improved outcomes are associated with high-dose treatment, if tolerated by the patient. [31] Chordomas are known to be relatively radioresistant, often requiring treatment with at least 70 Gy. [32] Modern radiation treatment strategies include IMRT, SRS, and particle therapies including proton and carbon ions. [33] There is emerging evidence that particle radiation therapy (either proton or carbon ion therapy) offers high local control while minimizing damage to organs at risk, though their effectiveness is decreased by higher gross tumor volume. [34] Current best practices from the Chordoma Global Consensus Group advise that in selecting between radiation therapy and surgical resection for locoregional recurrence, there are insufficient data to support generalized recommendations and treatment must by tailored to the individual. [32] In our analysis, no survival benefit was associated with higher doses of radiation therapy (Supplementary Figure   S2) or specific radiation modality (Figure 3).
Chemotherapy is not commonly utilized for clival and skull base chordoma, though targeted therapies are under investigation based on whole-transcriptome analysis, with potential targets including the epidermal growth factor receptor, c-Met, and HER2/neu pathways. [35] Accepted Manuscript This article is protected by copyright. All rights reserved.
As in prior studies, patients with larger tumors (size ≥ 5 cm) and older age (age ≥ 65) had worse OS. Larger volume disease represents a greater challenge to both surgical resection and radiation therapy given proximity to critical structures. [7] Proximity to the optic apparatus and brainstem is particularly associated with high local recurrence rates. [36,37] Prior NCDB analysis has shown that age greater than 60 is an independent risk factor for decreased survival in cranial chordoma, which is concordant with our analysis. [5] Treatment of clival and skull base chordoma appears to be becoming more effective with time, as OS was improved in patients diagnosed with the disease after 2009 as compared to prior to this date. This finding likely reflects treatment trends towards improved surgical techniques and knowledge of surgical access and anatomy, multimodal treatment, and the advent of techniques to deliver high dose radiation. [5] Anatomic limitations of external surgical approaches have been significantly expanded by advent and wide-spread adoption of endoscopic approaches. [38] Of particular note are the trans-cavernous corridor approach to the superior clivus, the supravidian transpterygoid approach to the middle clivus, and the transpterygoid infravidian approach to the inferior clivus. [38] While an endoscopic approach is often sufficient for gross total resection of midline lesions, a combination of endoscopic and external approaches may be appropriate in tumors with significant lateral or inferior extension. [38] With an experienced skull base team, gross total resection can be achieved in up to 88.9% of primary cases. [39] Though not yet universally available, treatment modalities such as proton therapy are increasing in availability throughout the US, which is promising in light of studies demonstrating excellent outcomes. [40] While proton therapy was first described for chordomas over 20 years ago, availability of this technique and subsequent utilization has increased to the point that it is now more commonly employed than photon radiation within NCDB studies, including our

Accepted Manuscript
This article is protected by copyright. All rights reserved. current study. [5,36] Carbon ion therapy is much less accessible, with the first US site to be created at the Mayo Clinic in Minnesota.
[41] Two benefits to particle therapy over photon therapy are secondary to the sharp penumbra, allowing 1) better tumor coverage and 2) dose escalation despite tumors that are close to dose-limiting structures such as the optic apparatus, brainstem, and spinal cord. While the radiation doses are not universally agreed upon, recommendations have been made for at least 74 GyE to gross disease.
[42] Improved imageguidance and IMRT techniques often make such high doses possible using conventional photon therapy. [43] Alternatively, there have been very good results with the use of stereotactic radiosurgery, with excellent local control rates in one small series from MSKCC. [44] Chordomas are radioresistant and less vulnerable to sublethal DNA damage, which means there is a theoretical advantage to higher doses per fraction that can deliver more irreparable DNA damage. [45] Another notable finding in our analysis was decreased OS in patients with government insurance relative to private insurance. This is consistent with a recent NCDB study by Carey et al., who demonstrated decreased OS for patients with no insurance or government insurance who were diagnosed with head and neck malignancy. [46] Part of this effect has been attributed to advanced stage at presentation, as highlighted in one study, where patients with Medicaid were both more likely to present with metastatic head and neck cancer and more likely to not receive definitive treatment relative to patients with private insurance. [47] Limitations of this study include the retrospective nature of our data. While our overall study size was comparable to prior observational studies, our work was nevertheless underpowered to evaluate the role of adjuvant chemotherapy, chemoradiotherapy, or immunotherapy in the treatment of this disease, which represents an opportunity for future

Accepted Manuscript
This article is protected by copyright. All rights reserved.
research. The NCDB does not offer data on disease-specific survival, nor does it provide detailed descriptions of the surgical approaches utilized during treatment. Given the deidentified nature of the NCDB, review of individual patient data to identify the specific means of determining margin status was not possible. In addition, postoperative complications are not readily available for analysis, which may represent a potential confounding variable with respect to mortality outcomes. Prospective data are needed to further elucidate the optimal treatment of skull base and clival chordoma.

Conclusions
Surgery remains the mainstay of therapy for skull base and clival chordoma. Advanced age, large tumor size, and government insurance were all predictors of worse OS. General consensus for skull base chordomas is to pursue maximal safe resection, and the decision for adjuvant radiation is generally based on margin status. In this study neither negative margins nor adjuvant radiation were associated with improved survival, although progression and recurrence data were unknown. A multidisciplinary approach to these challenging tumors is critical to optimize treatment outcomes.

Financial Disclosure
Not Applicable.

Author Conflict of Interest
Not Applicable.

Accepted Manuscript
This article is protected by copyright. All rights reserved.

Accepted Manuscript
This article is protected by copyright. All rights reserved.

Accepted Manuscript
This article is protected by copyright. All rights reserved.

Accepted Manuscript
This article is protected by copyright. All rights reserved.
Pearson's chi-squared test or unpaired two-sample t-test *P-value below threshold for statistical significance (p<0.05)