Introduction

Skull base chordomas are rare malignant tumors that account for < 0.2% of all intracranial tumors [1]. They are locally destructive tumors and tend to compress or encase nearby brainstem and cranial nerves, causing headache, diplopia, and other cranial nerve deficits [2]. Although maximal surgical resection increases patient survival rates, complete resection is challenging due to adjacent critical structures and may increase morbidity [3]. Conventional radiation techniques have difficulties in delivering adequately high radiation dose for chordomas and thus have low local control rates [4]. Adjuvant high-dose radiation techniques using particles, such as protons or carbon ions, have been reported to have improved 5-year local tumor control rates of > 80% after incomplete resection [2, 5]. Gamma Knife radiosurgery (GKRS) delivers high radiation doses to the target while minimizing injury to the surrounding tissue with its steep dosimetry decline. A recent multicenter study suggests that GKRS is a safe and effective alternative treatment for skull base chordomas with a 10-year overall survival rate of 70% [6].

Magnetic resonance imaging (MRI) plays an important role in pre- and posttreatment evaluations for chordomas [7]. Generally, chordomas appear hyperintense on T2-weighted imaging (T2WI) and heterogeneously enhancing on contrast-enhanced T1-weighted imaging (T1WI) [1]. On diffusion-weighted imaging (DWI), the signal of skull base bony structures and cerebrospinal fluid are suppressed, and chordomas can be clearly demarcated. Furthermore, previous studies have shown that chordomas could be distinguished from chondrosarcomas by apparent diffusion coefficient (ADC) measurements [8, 9]. However, the associations between these MRI signal characteristics and prognosis remain relatively unknown. Identification of MRI signal characteristics with prognostic value may provide guidance in therapeutic decision-making. Therefore, the purpose of the present study was to investigate the influence of MRI signal characteristics on GKRS outcomes of skull base chordomas.

Methods

Patients and study design

The TRIPOD reporting guideline was implemented. The local institutional review board approved the study protocol and waived the need for consent. For the period from 2001 to 2021, there were total 30 consecutive patients who received GKRS for skull base chordomas at our institute. After the exclusion of 6 patients without DWI data before GKRS for evaluations, 24 patients were enrolled in this study.

Stereotactic radiosurgery and follow-up strategy

Stereotactic radiosurgery was performed using Gamma Knife (Elekta AB, Stockholm, Sweden) Model B (from 2001 to 2006), Model 4C (from 2007 to 2012), or Perfexion (after 2012). Patients underwent MRI performed using a 1.5-T MR imager (Signa HDxt, General Electric Healthcare, Waukesha, WI, USA) for radiosurgical targeting. The GKRS planning MRI protocols included axial 3 mm-thick fast spin-echo T2WI, spin-echo T1WIs before and after administration of intravenous gadolinium-based contrast agent at a dose of 0.1 mmol/kg body weight, and DWI (repetition time/echo time, 4000–5000/60–75 ms, matrix size 128 × 128, field of view 300 mm; 3 directions; b value = 0 and 1000 s/mm2) with corresponding ADC maps. To reduce field inhomogeneity, DWI acquisition was performed before stereotactic frame placement. Based on the integrated MRI findings, an effective radiation dose was delivered to the delineated tumor while avoiding radiation-induced damage to the surrounding vital structures, particularly the brainstem. After GKRS, the patients were scheduled to undergo regular neurological evaluations every 1 to 3 months and MRI every 3 to 6 months.

Study parameters and MRI evaluations

The patients’ data on demographic characteristics at GKRS [age, sex, Karnofsky Performance Status (KPS), neurological symptoms, cranial nerve deficits, prior resection grade, histopathologic type of chordoma, previous radiotherapy, tumor volume, and location], MRI features, GKRS parameters, and outcomes (neurological status, patient survival, radiological tumor response, and salvage treatment after GKRS) were collected for analysis. Prior resection grades were divided into > 90% and ≤ 90% of the tumor volume removed by reviewing their respective pre- and postoperative MRI. The tumor volume was measured from the sum of tumor areas in each image slice multiplied by the slice thickness. Tumor locations were classified as sphenoclival, occipitocervical, sphenopetrous, petrous-occipital, ethmoid-sphenoid, and extensive (involving at least 2 of the 5 locations mentioned above) types according to Wang et al. [3]. The pre-GKRS MRI was measured for the MRI signal characteristics of tumor, including RT2, RCE, and mean ADC, by using the measurement instruments of the picture archiving and communication system (version 2.1.0.11; SmartIris, Taiwan Electronic Data Processing Co., Taipei, Taiwan). To reduce the effect of image noise, the brainstem was selected as control for normalization of chordoma MR signal intensity according to Tian et al. [10]. RT2 and RCE were defined as signal intensity ratio of tumor to brainstem on T2WI and contrast-enhanced T1WI, respectively. Figure 1a and b demonstrate how regions of interest (ROIs) of tumor and adjacent brainstem were outlined on T2WI and contrast-enhanced T1WI for mean signal intensity measurements. At least 3 images in each MRI sequence were selected and their mean signal intensities of tumor and brainstem were averaged, respectively, and used for RT2 and RCE calculation. Figure 1c and d show tumor delineation according to the tumor geometry on ADC maps in conjunction with DWI. The pixel-by-pixel mean ADC values for the entire volume were calculated. Two neuroradiologists with 27 and 8 years of experience in neuroimaging, respectively, independently delineate the ROIs manually and their measurement results were averaged for subsequent analysis. We recorded patients’ data until patient death or loss to follow-up, because salvage treatments were performed for tumor progression after GKRS and patient death was the primary outcome. Tumor responses were evaluated by comparing follow-up MRI to pre-GKRS MRI and were categorized as follows: progression, > 10% increase in tumor volume; stable, ≤ 10% increase or decrease in tumor volume; and regression, > 10% decrease in tumor volume. We routinely choose 10% to classify tumor progression during the post-GKRS follow-up, because an error margin of < 10% is achievable with volume measurements using the trapezoidal rule formula and at least 5 slices of area contour [11, 12]. Initial tumor response was assessed in the immediate follow-up MRI 3 months after GKRS. Local control was defined as the absence of tumor progression within the 20% isodose treatment volume [13]. Remote tumor progression was defined as tumor progression beyond the 20% isodose treatment volume. Adverse radiation-induced events, including hypopituitarism, eye disorders, and nervous system disorders, were evaluated based on the Common Terminology Criteria for Adverse Events version 5.0 [14].

Fig. 1
figure 1

Magnetic resonance images at Gamma Knife radiosurgery for a progressed sphenoclival chordoma after prior resection with the tumor outlined in yellow and brainstem outlined in white. RT2 and RCE are obtained from the mean signal intensity ratio of tumor to brainstem on axial T2-weighted imaging (a) and contrast-enhanced T1-weighted imaging (b), respectively. The tumor is well delineated on the diffusion-weighted imaging (c) and mean apparent diffusion coefficient (ADC) value is measured on the corresponding ADC map (d)

Full size image

Statistical analyses

Statistical analyses were conducted using SPSS (version 22; SPSS, IBM, Dublin, Ireland). The results are presented as medians (interquartile ranges) and numbers (percentages) for continuous and categorical variables, respectively. The Mann–Whitney U test or Wilcoxon signed-rank test, as appropriate, was used to compare the differences in continuous variables. The interobserver reliability between the 2 evaluators was measured using the intraclass correlation coefficient. Univariable and multivariable Cox regression analyses with hazard ratios (HRs) were used to evaluate factors associated with patient survival and local tumor progression after GKRS. Variables with P < 0.10 in univariable analyses were included in the multivariable analysis. Receiver operating characteristic curve analysis with the Youden Index was used to determine cut-off RT2, RCE, and mean ADC values for discriminating between the dead and alive groups. The Kaplan–Meier method with log-rank test was used to determine and compare the survival data of the different groups. Statistical significance was set at P < 0.05.

Results

Patient characteristics and outcomes after GKRS

Table 1 presents the characteristics of the 24 patients with chordomas. The median age of the patients at the time of GKRS was 41 years (interquartile range, 31–55 years), and 12 (50%) of them were women. Before GKRS, all patients had surgical resection for the chordoma with resection grades of > 90% and ≤ 90% in 10 (41.7%) and 14 (58.3%) patients, respectively. The median number of surgical resection before GKRS was 2 (1–2). Four (16.7%) patients had undergone adjuvant radiotherapy, including proton in one and photon in three patients. Twenty-two (91.7%) patients were diagnosed with conventional chordoma and 2 (8.3%) patients with chondroid chordoma. The median interval between the first diagnostic neuroimaging and GKRS was 12 months (5–35 months). GKRS was performed as adjuvant therapy for residual tumor in 19 (79%) patients and as salvage treatment for tumor progression in 5 (20.8%) patients at a median period of 5 months (2–6 months) after last surgical resection. At the time of GKRS, the median KPS was 80 and 23 (95.8%) patients had neurological symptoms. The median tumor volume was 12.6 cm3 (4.5–15.8 cm3) and 8 (33.3%) cases were in extensive locations. The median margin dose was 16 Gy (14.5–17.8 Gy).

Table 1 Demographics, clinical presentations, MRI features, GKRS parameters, and outcomes
Full size table

The intraclass correlation coefficients for interobserver reliability between the 2 evaluators were 0.984 (95% CI, 0.75–0.999) for the RT2, 0.98 (95% CI, 0.685–0.999) for the RCE, and 0.966 (95% CI, 0.76–0.995) for the mean ADC. The median neuroimaging and clinical follow-up durations after GKRS were 45 months (32–92 months) and 46 months (35–93 months), respectively. After GKRS, worsened or new neurological deficits developed in 3 (12.5%) patients due to tumor progression. The initial tumor responses were progression in 1 (4.2%), stable in 12 (50%), and regression in 11 (45.9%) patients. On follow-up MRI, local and remote tumor progression occurred in 16 (66.7%) and 4 (16.7%) patients, respectively. Additional treatments were performed in 16 patients for tumor progression after GKRS, including 13 sessions of resection, 12 sessions of repeat GKRS, and 9 sessions of resection with adjuvant GKRS. Ten (41.7%) patients received more than 1 adjuvant GKRS session during the follow-up period. At the last follow-up, 9 (37.5%) patients died of tumor progression, and no radiation-induced adverse events were observed. In the 11 patients with at least 5 years of follow-up, 8 patients (73%) were alive at a median follow-up period of 10.7 years (6.3–12.9 years) after GKRS.

Factors associated with patient survival and local tumor progression after GKRS

Patients who died (n = 9) had chordomas with significantly lower pre-GKRS mean ADC (median: 1036.7 × 10–6 mm2/s vs 1335.3 × 10–6 mm2/s, P = 0.046) and more local tumor progression events (median: 2 vs 0, P = 0.012) than did those who survived (n = 15). The leading discriminators between the succumbed and alive groups were a RT2 of 1.5 (sensitivity: 44.4%, specificity: 93.3%, area under the curve: 0.689 [P = 0.128; 95% CI, 0.452–0.926]), a RCE of 1.57 (sensitivity: 88.9%, specificity: 66.7%, area under the curve: 0.778 [P = 0.025; 95% CI, 0.584–0.971]), and a mean ADC of 1270 × 10–6 mm2/s [sensitivity: 88.9%, specificity: 60%, area under the curve: 0.744 (P = 0.049; 95% CI, 0.542–0.947)]. Table 2 presents results of the univariable and multivariable Cox regression for overall survival and local tumor progression after GKRS. After adjustment for the interval between first diagnostic neuroimaging and GKRS, higher KPS (HR: 0.86, 95% CI: 0.768–0.964, P = 0.009) and mean ADC (HR: 0.992, 95% CI: 0.986–0.999, P = 0.016) were independently associated lower rates of death. Kaplan–Meier analyses (Fig. 2) revealed that mean ADC < 1270 × 10–6 mm2/s (log-rank test, P = 0.03) was associated with a lower probability of overall survival, and RT2 < 1.5 (P = 0.038) and RCE > 1.57 (P = 0.022) were associated with a lower probability of local tumor control. Mean ADC < 1270 × 10–6 mm2/s (P = 0.198) was not associated with a lower probability of local control. RT2 < 1.5 (P = 0.107) and RCE > 1.57 (P = 0.053) were not associated with a lower probability of overall survival.

Table 2 Univariable and multivariable Cox regression analyses for factors associated with patient survival and tumor progression after GKRS
Full size table
Fig. 2
figure 2figure 2

Actuarial probabilities of overall survival and local tumor control within 120 months after Gamma Knife radiosurgery for chordomas based on whether RT2 is < 1.5 (a, b), RCE is > 1.57 (c, d), or mean apparent diffusion coefficient (ADC) is < 1270 × 10–6 mm2/s (e, f). The values below the X-axis represent the number of patients remaining in the analysis. a The 5-year probability of overall survival was 40% and 81.6% for RT2 < 1.5 and ≥ 1.5, respectively. b The 5-year probability of local control was 0% and 34% for RT2 < 1.5 and ≥ 1.5, respectively. c The 5-year probability of overall survival was 58.3% and 88.9% for RCE > 1.57 and ≤ 1.57, respectively. d The 5-year probability of local control was 15.4% and 38.1% for RCE > 1.57 and ≤ 1.57, respectively. e The 5-year probability of overall survival was 54.7% and 88.9% for mean ADC < 1270 × 10–6 mm2/s and ≥ 1270 × 10–6 mm2/s, respectively. f The 5-year probability of local control was 13% and 41.1% for mean ADC < 1270 × 10–6 mm2/s and ≥ 1270 × 10–6 mm2/s, respectively

Full size image

Follow-up mean ADC values in cases with local tumor progression after GKRS

During follow-up, 28 events of post-GKRS local tumor progression, including 16 after first GKRS and 12 after repeat GKRS, occurred in 16 patients, and their mean ADC value was significantly higher than that before GKRS (median: 1243.9 × 10–6 mm2/s vs 1109.3 × 10–6 mm2/s, P = 0.005), but still lower than 1270 × 10–6 mm2/s. The changes of mean ADC values in the 28 progressed tumors after GKRS had little associations with margin dose (r = -0.199) and max dose (r = -0.007) by the Spearman’s rank correlation coefficient. Figure 3 illustrates a case example of chordoma with a pre-GKRS mean ADC of < 1270 × 10–6 mm2/s and three events of post-GKRS local tumor progression.

Fig. 3
figure 3

Chordoma with a pre-Gamma Knife radiosurgery (GKRS) mean apparent diffusion coefficient (ADC) value of < 1270 × 10–6 mm2/s and three events of post-GKRS local tumor progression in a 31-year-old man. Rows from top to bottom: axial T2WI, contrast-enhanced T1WI, DWI, and ADC maps. The contrast-enhanced T1WI selected from the Gamma Plan dose-planning program demonstrate the targets treated with margin doses of 18 Gy, 18 Gy, and 17 Gy at the first, second, and third GKRS session, respectively. The recurrent tumors are conspicuous on DWI at 37, 63, and 70 months after the first GKRS session. The corresponding mean ADC values (853 × 10–6 mm2/s, 1015 × 10–6 mm2/s, and 1090 × 10–6 mm2/s) were higher than the pre-GKRS mean ADC value (796 × 10–6 mm2/s), but still below 1270 × 10–6 mm2/s. The patient died of tumor progression

Full size image

Discussion

Skull base chordomas are rare intracranial tumors with poor prognoses due to their locally destructive nature, vicinity to vital neurovascular structures, and high recurrent rates [2]. The standard of care for skull base chordomas is safe maximal resection and adjuvant radiotherapy to preserve neurological function and improve tumor control [2]. By using a cohort of 238 patients treated primarily through resection, Wang et al. [3] reported a 5-year overall survival rate of 76%. Patients who underwent resection grade ≤ 90% and presented with recurrent tumor had lower survival rates. Despite recent progress in endoscopic techniques and intraoperative neuromonitoring, repeat resection remains challenging and did not improve survival for patients with recurrent tumor after radiotherapy [15]. Therefore, stereotactic radiosurgery alone or in combination with resection may be used for these patients to reduce morbidity [15]. A meta-analysis of 25 studies showed that the 5-year overall survival rates were 78% for 351 patients treated with protons and 87% for 361 patients treated with carbon ions [16]. In a multicenter cohort of 93 patients underwent GKRS with a mean margin dose of 17 Gy for a mean tumor volume of 8 cm3, the 5-year overall survival rate was 83% [6]. This study reported a slight lower 5-year overall survival rate of 70% in 24 patients underwent GKRS, probably because of their larger median tumor volume (12.6 cm3) and lower median margin dose (16 Gy). In the present study, 10 (42%) of 24 patients had received more than 1 adjuvant GKRS session during the follow-up period. In comparison with repeated particle therapy, repeat GKRS seems to be an affordable and accessible treatment option with fair tumor control and low adverse radiation-induced event rates for small- to medium-sized chordomas.

The typical histopathologic finding of chordoma is vacuolated cells with intracytoplasmic mucus [1]. The high fluid content gives chordoma high signal intensity on T2WI [1, 7]. Following administration of gadolinium-based contrast agents, chordoma demonstrates variable enhancement on T1WI characterized as a “honeycomb” appearance, possibly due to mucinous content or necrosis [1, 7]. However, these MRI features cannot reliably differentiate skull base chordoma from chondrosarcoma [9]. DWI with ADC map measures the extracellular water diffusion quantitatively [17], which may reflect cell density and enable preoperative differentiation between chordoma and chondrosarcoma [9, 18]. Furthermore, DWI demarcates the high signal of chordoma in the signal-suppressed background of bone and cerebrospinal fluid [8]. Our institute included DWI in MRI protocols to facilitate GKRS planning for remnant chordomas after resection and post-GKRS evaluation for tumor response and detection of recurrent tumor.

Because of the rarity of skull base chordomas, there are few studies on the MRI signal characteristics associated with treatment outcomes (Table 3) [10, 19, 20]. Tian et al. [10] measured the relative signal intensity of chordoma to brainstem on T2WI and contrast-enhanced T1WI in 156 patients before surgery and proposed the MRI grading system for skull base chordomas. High grade chordomas with RT2 of ≤ 2.49 and RCE of > 0.77 were found to have more abundant tumor blood supply and higher risks of tumor progression than low grade chordomas do [10]. Their subsequent serial studies demonstrated that the high MR grade was associated with a higher chordoma growth rate and more prevalent in older patients with a shorter progression-free survival [21, 22]. The low RT2 value reflects a low fluid content and may corroborate the low T2 signal intensity appearance of poorly differentiated chordomas [9]. Following their approach, we showed that RT2 of < 1.5 and RCE of > 1.57 were associated with a lower probability of local tumor control in univariable anaylsis.

Table 3 MRI signal characteristics associated with treatment outcomes of skull base chordomas in the literature
Full size table

In the present study, we found that lower mean ADC values were independently associated with shorter patient survival after adjustment for KPS and the interval between GKRS and prior MRI for initial diagnosis. Previous studies also revealed that lower mean ADC values were associated with lower overall survival rates in patients with skull base chordomas [19, 20]. The underlying mechanism of these finding remains to be elucidated. Although the mean ADC values were not statistically different among the histopathologic types of chordomas in prior studies [8, 9], a lower mean ADC value was supposed to reflect higher cellularity and aggressiveness in chordomas [18,19,20]. The cut-off mean ADC values were similar in respective studies with a slight difference in the timings of measurement, namely 1198 × 10–6 mm2/s before surgery [19], 1270 × 10–6 mm2/s before GKRS, and 1494 × 10–6 mm2/s after radiotherapy [20]. In addition, we found that the mean ADC value of progressed tumors after GKRS was significantly higher than that before GKRS (1244 × 10–6 mm2/s vs 1109 × 10–6 mm2/s), but was still below 1270 × 10–6 mm2/s. These findings suggest that despite the elevation of mean ADC value after radiation treatment [20], the low mean ADC value indicating aggressive chordomas may determine the patient prognosis after treatment [19]. Accordingly, chordomas with lower mean ADC values warrant a more vigilant follow-up after treatment. Dose-escalated protons or extended-field GKRS with higher margin dose ≥ 20 Gy if possible may improve patient survival [23, 24].

This study had several limitations. First, the study included a small number of patients with a relatively short follow-up period because skull base chordoma is a rare intracranial tumor. Second, this was a single-center retrospective study with patients treated through GKRS, and the results are subject to selection bias and not generalizable to institutes using different radiation techniques. Third, the MRI signal of postoperative chordomas may be heterogeneous due to internal necrosis or hemorrhage [8], and thus the minimum and maximum ADC values were not used for analysis in our study to avoid measurement errors. Fourth, ADC < 1270 × 10–6 mm2/s showed a trend toward significance in the multivariable model of patient death (HR: 57.253, 95% CI: 0.891–3677.472, P = 0.057), which may be due to small patient number. Finally, although mean ADC value may reflect cellular density, correlations of the MRI signal characteristics with histopathologic data were not performed in this study. Our results need to be further validated in multicenter studies involving larger number of patients, such as the one from International Radiosurgery Research Foundation [6].

Conclusions

Management of skull base chordomas requires a multidisciplinary approach, and GKRS is a safe and effective option in treating residual or progressed tumor after surgical resection. Integration of DWI and conventional MRI sequences may help in treatment planning and posttreatment evaluation for skull base chordomas. In addition, lower mean ADC values were associated with shorter patient survival after GKRS. Further studies are necessary to determine the role of ADC in guiding therapeutic strategy of skull base chordomas.