Long Follow-up Surgical Results in 284 Cases of Clival Chordomas: The Risk Factors for Outcome and Tumor Recurrence

OBJECTIVE: Skull-base chordoma (SBC) is rare and one of the most challenging diseases to treat. We aimed to assess the optimal timing of adjuvant radiation therapy (RT) and evaluate the factors that inuence resection and long-term outcomes. METHODS: In total, 284 patients with 382 surgeries were enrolled in this retrospective study. Postsurgically, 64 patients underwent RT before recurrence (pre-recurrence RT), and 47 patients underwent RT after recurrence. During the rst attempt to achieve gross-total resection (GTR), when the entire tumor was resected, 268 patients were treated with an endoscopic midline approach, and 16 patients were treated with microscopic lateral approaches. Factors associated with the success of GTR were identied using c 2 and logistic regression analyses. Risk factors associated with chordoma-specic survival (CSS) and progression-free survival (PFS) were evaluated with the Cox proportional hazards model. RESULTS: In total, 74.6% of tumors were marginally resected [GTR (40.1%); near-total resection (34.5%)]. History of surgery, large tumor volumes and tumor locations in the lower clivus were associated with a lower GTR rate. The mean follow-up period was 43.9 months. At last follow-up, 181 (63.7%) patients were alive. RT history, histologic subtype (dedifferentiated and sarcomatoid), non-GTR, no postsurgical RT, and the presence of metastasis were associated with poorer CSS. Patients with pre-recurrence RT had the longest PFS and CSS, while patients without postsurgical RT had the worst outcome. CONCLUSION: GTR is the goal of initial surgical treatment. Pre-recurrence RT would improve outcome regardless of GTR.


Introduction
Chordoma is a rare bone malignancy with an incidence rate of 0.08 and 0.04 per 100,000 in the United States and Europe [20,11] and in Taiwan [13], respectively. Skull-base chordomas (SBCs) account for 32% of all chordomas [20]. Local recurrence is common with a late recurrence rate > 50% [24,9,30,14,10]. The average survival after surgery with or without radiation therapy (RT) is approximately 7.7 years [35]. Because of the deep location and close proximity to vital structures, surgical treatment is a challenge for SBCs with the total resection rate ranging from 0% to 73.7% [9]. The endoscopic midline approach (EMA) for SBCs yields better or similar resections than the microscopic lateral open approach (MLOA) [26,12].
Factors that in uence resection with EMA still need to be clari ed.
Although RT is recommended as an adjuvant treatment to surgery [25], the use of RT remains controversial [23,40,9,7]. In clinical practice, it is not clear whether and when RT should be administered [23,31], especially when the tumor was gross total resected [33]. Hence, a comprehensive investigation of the optimum management protocol conducted in a large study is warranted [9,14].
Our group, the rst neuroendoscopic group in China, was established in Beijing Tiantan Hospital in 1998 and has since then focused on skull-base diseases [18]. We have used both EMA and MLOA to resect SBCs during the past two decades. In this study, we retrospectively analyzed the clinical data of 284 patients with SBCs who were treated by our single group to identify factors for achieving complete resection and the optimal timing for RT.

Patients
Patients with histologically con rmed SBCs who were treated between December 31, 2003, and January 31, 2019, were included in the present study. c All patients signed informed consent forms.

Radiological evaluation
All patients underwent MRI and CT examinations preoperatively. The tumor volume was calculated as volume = (a × b × c)/2, where a, b and c represent the longest diameters in sagittal, coronal and axial views, respectively. According to the sellar oor and sphenoid sinus oor plane in sagittal view, the clivus was divided into superior, middle and inferior zones, as previously described by our group ( Figure 1A) [12].
According to the bilateral boundaries traditionally established for EMA, the skull base was divided into midline and paramedian regions ( Figure 1B) [12]. When > ½ circumference of the internal carotid artery (ICA) was encased, cavernous sinus invasion was de ned. Tumors that had invaded into the subdural space were de ned as exhibiting dural penetration [37].

Surgical methods
We used both EMAs (transnasal or transoral or combined) and MLOAs, which were chosen according to the tumor location [12]. When a tumor was located in the midline region or was slightly extended to the paramedian region, EMA was the rst choice. Based on our belief that the dura is a natural barrier against intracranial invasion by chordomas, we did not cut off the inner layer of dura but thoroughly resected the tumor tissue that attached to the dura. When dural penetration existed [37], we followed the tumor passage and resected it. When the tumor was mainly located in the paramedian region, we chose the MLOA. The resection rate was divided into four grades according to the postsurgical MRI and CT combined with the intraoperative impressions of surgeons [30]. Gross-total resection (GTR) indicated that the entire tumor was resected, the surrounding heathy tissue was obviously exposed, and no suspected tumor could be found on postsurgical images. No residual tumor was found during surgery, and > 90% tumoral resection on images was de ned as near-total resection (NTR); > 70% tumoral resection on images was de ned as subtotal resection (STR), and a lesser extent of resection was de ned as partial resection (PR). For the sake of comparison, GTR and NTR were pooled and classi ed as marginal resection, and STR and PR were classi ed as intralesional resection [30,24,25].

Follow-up
The rst follow-up was performed in outpatient center after surgery. Thereafter enhanced MRI was done every six months. RT was recommended to all patients. The date of last follow-up, which was conducted by telephone, was November 9, 2019. Inquiries were made regarding RT modalities and dates. RT modalities include radiosurgery (Gamma Knife and Cyberknife), intensity-modulated photon radiotherapy (IMRT) and charged particle radiotherapy (CPRT, includes both proton and carbon ion RT). Tumor recurrence was con rmed by radiological imaging. Progression-free survival (PFS) was calculated from the date of surgery to the date of radiographic recurrence. Chordoma-speci c survival (CSS) was de ned as the time between the date of surgery and the date of death caused by chordoma. If a patient was lost to follow-up or died of surgical complications or non-disease-speci c reasons, censor data were applied.

Statistical analyses
Statistical analyses were performed using SPSS 19 (IBM Corp, Armonk, New York) or R version 3.6.1 (R Foundation for Statistical Computing, Vienna, Austria). The difference in age distributions between groups was tested using the Mann-Whitney U test. PFS and CSS were estimated with the Kaplan-Meier method. A Cox proportional hazards model was used to identify prognostic factors. To evaluate the in uence of RT timing on CSS, patients were classi ed into three categories: RT before recurrence (prerecurrence RT), RT after recurrence (late RT) and no RT. The potential factors associated with the resection rate were rst analyzed with Pearson's c 2 test or Kruskal-Wallis test, and factors with P ≤ .05 were jointly analyzed using logistic regression. P < .05 was considered statistically signi cant.

Patients' demographic and clinical characteristics
A total of 284 patients underwent 382 surgeries (Table 1), including 380 skull-base surgeries (349 EMAs and 31 MLOAs) and two metastatic-lesion resections. Common presenting symptoms included diplopia (50.7%), headache or neck pain (34.5%) and blurry vision (24.6%). The median duration from self-reported initial symptoms to diagnosis was 6.0 months (range 1.0-108.0 months). A total of 184 patients were   represented between 20 and 60 years of age, and group C represented ≥ 60 years of age. b One patient whose tumor was located in the paramedian region was included in the group with median extension to the paramedian region. represented between 20 and 60 years of age, and group C represented ≥ 60 years of age. b One patient whose tumor was located in the paramedian region was included in the group with median extension to the paramedian region.  (Fig. 2D), no prerecurrence RT (HR = 3.33, 95% CI = 2.11-4.21, P < .001) (Fig. 2E), history of RT (HR = 1.80, 95% CI = 1.05-2.37, P = .033) and history of surgery (HR = 1.59, 95% CI = 1.04-1.97, P = .034) were associated with shorter PFS (Table 3). represented between 20 and 60 years of age, and group 3 represented ≥ 60 years of age. b One patient whose tumor was located in the paramedian region was included in the group with median extension to the paramedian region.

Effects of RT on long-term outcomes when GTR was achieved
A total of 89 patients with primary SBCs achieved GTR. PFS was signi cantly longer in the group with pre-recurrence RT than in the others (HR = 3.41, 95% CI = 1.39-8.35, P = .007) after adjusting for age, sex and tumor volume. CSS was not evaluated due to the limited number of death events.

Radiotherapy modalities and outcomes
To study the relationship between RT modalities and outcomes, we analyzed 45 patients with primary SBCs who were treated with either CPRT or other RT modalities. However, because of the limited statistical power due to the short follow-up and limited number of cases in the group with CPRT, we did not observe signi cant differences in CSS or PFS between the two groups (Supplementary Table 3).

Choice of surgical approach
Because EMA is the shortest route to the tumor and majority of SBCs located in the midline region [10,23,26]. we agree that EMA is appropriate for most SBCs [7]. To achieve maximal resection, MLOA was still required in selected cases. We did not nd a signi cant difference in the GTR rate between EMA and MLOA, which was consistent with the previous ndings of our group and others [30,34,38]. This supports the philosophy that a team focusing on skull-base disease should master different approaches [1].

Resection rate and outcome
The resection rate strongly in uences CSS and PFS [9,38,36]. Maximum resection was traditionally recommended with the goal of marginal resection [24,25,30]. The recurrence rate ranges from 16-45% at 10 years after marginal resection, and most recurrences occur within 2 or 3 years [23,34,14]. After a more detailed analysis, we found that the advantage of marginal resection was mainly contributed by GTR and that no obvious differences were found among NTR, STR and PR. Therefore, GTR should be the goal if possible [23]. This result also indicates that the standard for assessing the resection rate should be postsurgical images rather than the intraoperative impression of the surgeon [6].
Other risk factors for outcome Dedifferentiated subtype, history of RT, lack of postsurgical RT, and metastasis increased the risk of both shorter survival and recurrence, which has been shown by previous studies [9, 17, 10, 5, 34]. However, it is controversial whether different survival rates exist between conventional and chondroid types [29,7]. We found no signi cant difference, which is consistent with several studies [32,2]. Several previous studies showed that dural penetration [37,7] and age [29,5] were associated with long-term outcomes, and we demonstrated that they were not independent risk factors after multivariate analysis [23,36].

Risk factors reducing the chances of GTR
We found that younger age (≤ 20 years) increased the risk of non-GTR. This may be explained by our previous nding that most children with SBCs had these tumors in the lower clivus and cranio-cervical junction area [4]. Tumors in the lower clivus have the lowest GTR rate [14,17,6,4]. Not surprisingly, history of surgery, large tumor volume and tumor location are independent risk factors for GTR, as supported by previous studies [30,14,17,26,7].
Surgical complications and treatments CSF leakage and cranial nerve injury were the main complications in prior studies [17,14,6]. Bene ting mainly from the application of vascularized nasoseptal aps [17,7], CSF leakage was low in our group (3.9%). We emphasize the importance of cleaning tumors that are attached to the dura and maintaining the integrity of the dura, which effectively prevent CSF leakage. Radical resection of all involved dura might account for high rates of CSF leakage (14%) [6,8]. We deem the dura to be a natural barrier against tumor growth intracranially, which has been supported indirectly by two autopsy studies [28,21]. Both studies found that the dura was intact even in advanced stages.
Chordomas with cavernous sinus invasion have a lower GTR rate (23.0% vs 47.7%), although it is not an independent risk factor. The major di culty for achieving GTR in this setting was fear of ICA injury, which is another serious complication [17]. Endovascular treatment is an effective remedial treatment when the ICA is injured [39]. Therefore, we support the notion that EMA is safe and has few complications once adequate experience is gained [26].

Timing of RT and outcome
Adjuvant RT is a widely accepted treatment for residual chordomas [40,19], although no strong evidence supports it [31]. We found that PFS and CSS were signi cantly longer in patients with pre-recurrence RT than in other patients, which is similar to prior studies [14,34]. In addition, the PFS in the late-RT group was shortest in the present series (data not shown); however, the CSS was still longer than that of the patients who had no RT. Similarly, Olabisi Sanusi et al. found that a second dose of RT at any point either as sole treatment or as adjuvant treatment for recurrence showed a statistically signi cant effect on PFS (P = 0.009) [22]. Taken together, these ndings indicate that adjuvant RT is an independent factor for better outcomes [7,34], and the importance of pre-recurrence RT is worth emphasizing. Once recurrence occurs, postsurgical RT is also recommended to prolong CSS when the patient has no RT history [24].
It remains debatable whether to undertake pre-recurrence RT after GTR [9,23,19]. Lately, Yagiz Yolcu et al.'s found that compared with GTR alone, GTR plus RT did not offer any signi cant survival bene t for patients with sacral or spinal chordomas, at the price of higher complications rate [33]. However, our result is consistent with a recent nding that pre-recurrence RT in GTR group will prolong PFS [22].
Therefore, pre-recurrence RT is recommended to prevent recurrence even after GTR for patients with SBCs [19].

Radiation modalities and outcome
In recent years, increasing studies found that CPRT has advantages over other modalities in the chordoma treatment. Satoshi Takahashi et al. [27] found that the PFS of carbon ion treated group was longer than that of the other groups treated with other radiation modalities or untreated, however, a recent meta-analysis did not nd signi cant differences between radiosurgery, proton RT and carbon ion RT at 3 and 5 years survival [40]. Li H et al. also found that IMRT serves as an effective alternative to CPRT based on their retrospective analysis of 46 cases. Selection bias was likely present in Li's study and might be conducive to a result that contradicts previous ndings; for example, economic status (CPRT is much more expensive than other RT modalities) and education status (which will affect whether the patients would follow the clinical recommendation) in uence the accessibility of CPRT. For now, we cannot compare the effectiveness of CPRT and other radiation modalities due to short follow-up and limited number of cases with CPRT. Thus, we cannot rule out the possibility that there is no bene t of CPRT over other types of radiation, which might be associated with the improved RT techniques of radiosurgery or IMRT [15,16]. Although we recommend that patients with chordomas seek help from radiologists who administer particle-radiation therapy, we rmly believe that a randomized trial is warranted even in its rarity [14].

Limitations
As a retrospective study, main limitations were no well-designed timing of RT and no random choice of radiation modalities. Another limitation is that tumor biological characteristics were not analyzed, and such characteristics might be an underlying factor for GTR [9]. Recently, we found that genomic alterations were associated with recurrence and CSS [3]. Therefore, combining tumor biology with clinical characteristics will produce more robust conclusions in the future.

Conclusions
To the best of our knowledge, this is the largest series of SBCs surgically treated by single surgical team, and we found that surgery should be the initial treatment for primary SBCs. GTR is the surgical goal and should be applied in both primary and recurrent SBCs [36]. History of surgery, larger tumor volume, and tumor location (lower clivus, extension from the midline to the paramedian region) are independent risk factors for GTR. Pre-recurrence RT will postpone recurrence, even after GTR is achieved.   Risk factors for chordoma-speci c survival (CSS) and progression-free survival (PFS) in all 284 patients. A, The CSS was signi cantly longer in the gross-total resection (GTR) group than in the near-total resection (NTR) group, subtotal resection (STR) group or partial resection (PR) group (P<.05). B, When the NTR group, STR group and PR group were classi ed into the non-gross-total resection group (non-GTR), the CSS was longer in the GTR group than in the non-GTR group (P<.05). C, The group with pre-recurrence