Clinical practice of image-guided spine radiosurgery - results from an international research consortium

Rationales for the practice of spine radiosurgery compared to conventional palliative radiotherapy are similar between the five institutions: all agree on more durable pain control and long-term local tumor control. Four institutions state a more rapid pain relief as reason for spine radiosurgery (UPMC, UofT, UVAMC, RSMC), three institutions explicitly mention spine radiosurgery for radio-resistant histologies (UHW, UVAMC, UofT), two institutions use spine radiosurgery because of better patient convenience (UPMC, RSMC) and one institution describes the potential of improved overall survival in the oligometastatic setting as reason for spine radiosurgery (UHW). Radiosensitive histologies are excluded in all but one (RSMC) institution and a minimum performance status is required in all institutions (table 1). All four academic centers perform spine radiosurgery in the framework of a prospective protocol but not as a prospective trial.

In contrast, patient selection with respect to estimated live expectancy is substantially different: one institution strictly selects patients with good life expectancy using a predictive scoring system for overall survival (UHW), two centers exclude patients with very poor life expectancy (UofT, UVAMC) and life expectancy is no relevant factor for two institutions (UPMC, RSMC).

The inclusion and exclusion criteria of vertebral metastasis treated with spine radiosurgery are described in table 2. There is agreement that the relationship between the target volume and any OAR other than the spinal cord does not influence the indication for spine radiosurgery. Both lytic and sclerotic lesions are treated and all institutions but one (UHW) prefer a stabilization procedure prior to radiosurgery in cases of spinal instability. Compression fractures are always discussed with the neurosurgeon/spine surgeon and symptomatic spinal cord compression is a contraindication in all institutions. All vertebras in the cervical, thoracic and lumbar spine are treated and the number of vertebras within one target volume is limited to 3 except one institution, were only two vertebras are allowed within one target volume (RSMC).

Disagreement is evident whether epidural involvement or a small distance between the metastasis and the spinal cord are contraindications.

There is a good agreement in the imaging modalities and their technical application for staging and target definition (table 3). All institutions acquire dedicated CT and MRI images (a diagnostic MRI is allowed at the UPMC) for delineation and slice thickness is between 1-2 mm.

Differences, however, are observed in the MRI sequences and in acquisition of a dedicated planning FDG-PET.

Similar target volume concepts are used in the five institutions (table 4). All centers define the gross-tumor volume (GTV) based on CT and MR imaging, two centers perform co-registration of a FDG-PET (UPMC, RSMC). All centers treat the involved vertebras only without "prophylactic" irradiation of the superior and inferior vertebra. All institutions use an anatomical target volume concept where the target volume extends to uninvolved parts of the vertebras. Additionally, all institutions but one (RSMC) apply safety margins of 2-3 mm.

However, the details of the target volume concepts are different. Three institutions do always treated the entire vertebral body and/or the entire posterior elements in case of involvement (UPMC, UVAMC, RSMC). One institution has a similar concept, however, differentiates between the ipsilateral and contralateral posterior elements (UofT). One institution uses a two dose level approach where the high-dose target volume is defined as the GTV with a 3 mm safety margin and the low-dose target volume is the entire vertebra (UHW).

Regarding definition of the OAR spinal cord, all but one institution (RSMC) define the spinal cord in the MRI images; the spinal canal is delineated in CT images at the RSMC. Delineation is performed minimum 1 vertebra superior and inferior to the planning target volume (PTV) in all institutions and safety margins of 1-2 mm are applied for generation of the planning OAR spinal cord in all but one institution (UVAMC). On the level of the cauda equina, all institutions define the thecal sac as planning OAR.

Large variability is observed in terms of treatment dose and fractionation (tables 5). Two institutions treat the majority of their patients with single fraction radiosurgery of 16 - 24 Gy (UPMC, UVAMC). One center prefers 2 or 3 fraction radiosurgery (UofT), however, will treat with single fraction if no epidural disease is evident and single level disease. Two centers perform fractionated radiosurgery only (UHW, RSMC), and both of them choose between two fractionation schemas with estimated life expectancy as selection criterion. Fractionated radiosurgery is performed in 2-10 fractions with physical doses of 24 - 48.5 Gy. Despite the differences in dose and fractionation, all but one institution practice dose prescription to the D90, whereas one institution uses the ICRU reference point (UofT).

Based on an α/β = 10 Gy, the median 2 Gy-equivalent dose (EQD_{2 Gy}) is 50 Gy [minimum 36 Gy (3 × 8 Gy) and maximum 68 Gy (1 × 24 Gy)]. As described above, one institution (UHW) uses a two dose-level concept with conventional doses in the "elective" parts of the vertebra (10 × 3 Gy; 5 × 4 Gy) and dose escalated irradiation in the involved parts of the vertebra (10 × 4.85 Gy; 5 × 7 Gy).

No institution varies the spinal cord tolerance based on cervical, thoracic or lumbar target location. Otherwise, dosimetric parameters as well as tolerance doses for the spinal cord and thecal sac were substantially different between all five institutions (table 6).

Minor differences are observed in terms of treatment planning (table 7). All institutions treat their patients at an Elekta Synergy S linear accelerator equipped with the Beam Modulator (4 mm leaf width); one center does also perform spine radiosurgery on different linear accelerators (UVAMC). Treatment planning system is Pinnacle (Philips Radiation Oncology Systems, Milpitas, CA, USA) and intensity modulation is planed using step-and-shoot IMRT only (UPMC), both IMRT and VMAT (UHW, RSMC, UofT) and VMAT only (UVAMC). Technical details are summarized in table 7.

Acceptance criteria for treatment plans vary substantially with all centers stating that no strict criteria exist, mostly because of the large variability of the target volumes in terms of size, shape and distance to the spinal cord and other relevant OARs. However, all institutions agree that PTV coverage is sacrificed until the dose limits of the critical OARs, especially of the spinal cord, are fulfilled.

Differences in these steps of spine radiosurgery are small (table 8). All patients are treated in supine position and immobilization is performed using thermoplastic head masks for cervical/upper thoracic lesions and the BodyFIX for thoracic and lumbar lesions. Daily pre-treatment image guidance is performed using cone-beam technology and set-up errors are corrected in six degrees of freedom using the robotic HexaPOD couch. Action level for translational errors is 1 mm in all but one institution where a larger action level of 2 mm is used (UVAMC); the action level for rotational errors is 1° in four institutions (UHW, UPMC, UofT, UVAMC), 0.3° in one institution (RSMC). A second cone-beam CT scan for verification of the IGRT shift is performed in all institutions and all institutions but one (UHW) perform intra-treatment cone-beam CT scanning for patient monitoring. A final scan after treatment delivery is performed in 3/5 institutions (UHW, UofT, UVAMC).

Follow up is performed in-house whenever possible in all institutions; the interval is most frequently every 3 months (table 9). Local tumor control is defined as tumor shrinkage or no tumor progression in serial imaging, with MRI as the preferable imaging modality. One institution (RSMC) uses routine FDG-PET imaging for evaluation of local tumor control. Pain is assessed in 4/5 institutions using either the Visual Analog Scale (UHW, UPMC, UVAMC) or NRS-11 (RSMC).

As previously mentioned, no dose response has been established for conventional palliative radiotherapy with doses between 8 Gy in 1 fraction and 40 Gy in 20 fractions. One explanation could be that all tested doses are well below established thresholds from radical radiotherapy: e.g. even the "high dose" approach of 40 Gy in 20 fractions would result in tumor control of less than 5% for NSCLC [33]. This hypothesis is supported by three studies, which reported a significant dose response relationship for primary [17, 31] and re-irradiation [34] spine radiosurgery. The Memorial Sloan-Kettering Cancer Center (MSMCC) group reported excellent local control of > 95% for a prescription dose of 24 Gy and a minimum dose of > 15.1 Gy; however, the results are based on a retrospective review of serially dose-escalated patients and require validation before conclusions can be drawn. For pain control as primary endpoint, Ryu et al. reported a lower prescription dose threshold of 14 Gy; this dose response was not statistically significant [25]. It is important to interpret all these treatment doses in the context of the target and OAR volumes, which were used in the specific trials. Clinical application of any dose specification without detailed knowledge and application of the respective target and OAR concept cannot be recommended.

Centers in this study use a large rage of fractionations between single fraction to 10 fractions. Moreover, even within the single fraction treatment, the applied doses range between 16 Gy to 24 Gy. In the US, radiosurgery and stereotactic body radiotherapy are defined as treatment with maximum five fractions. In this study, one institution uses a 10 fraction approach, which does not fall under this definition. Nevertheless, we use the term radiosurgery even for this 10 fraction regimen because of two reasons: 1) the clinical and technical practice of the 10 fraction regimen is identical to the regimens using 1- 5 fractions; 2) using the LQ model, the biological effective dose of the 10 fraction regimen is expected to be at least equivalent to the 1- 5 fractions regimens. However, the limitations of the LQ model need to be considered for very high single fraction doses [35].

The third area of uncertainty and disagreement is the tolerance dose of the spinal cord. There was agreement between 4/5 institutions that a planning organ-at-risk should be generated with a safety margin around the true spinal cord. However, this safety margin varied between 1 mm around the spinal cord to 2 mm around the spinal canal. The dosimteric parameter used as dose threshold varied between D_max, D10, D_{0.1 cc} and D_{2 cc}. The closest agreement was observed for single fraction radiosurgery, where dose thresholds of 10 - 11 Gy are used.

A very low incidence of myelopathy after spine radiosurgery has been described in the literature, despite dose escalation and hypo-fractionation. Ryu et al. observed a myelopathy in 1 out of 86 patients with a minimum follow-up of one year; these authors recommended a tolerance dose of 10 Gy as D10 [36]. Combined data from Stanford University Medical Center and University of Pittsburgh Medical Center reported an incidence of 5 out of 1075 patients, and it was recommended limiting the volume of spinal cord treated above an 8-Gy equivalent dose [37]. Sahgal et al. collected five cases of myelopathy and compared various dose volume parameters with 29 patients, who did not develop myelopathy [38]; all patients were treated with various radiosurgical fractionations. The thecal sack was delineated as the planning organ-at-risk volume for all cases and controls with centralized review of the DVH data. Doses were converted to 2 Gy-equivalent doses based on an α/β = 2 Gy to cope with variation of fractionation. The maximum point volume EQD_{2 Gy} within the thecal sac significantly correlated with the risk of myelopathy as opposed to the larger volumes investigated (0.1 cc, 2 cc, and 5 cc). An EQD_{2 Gy} threshold of 30 - 35 Gy was recommended for 1 - 5 fractions. In conventionally fractionated radiotherapy, the tolerance of the spinal cord is usually accepted between 45 - 50 Gy, despite that even a dose of 60 Gy results in a risk of myelopathy of only approximately 5% [39]. This disagreement between EQD_{2 Gy} thresholds based on hypo-fractionation and thresholds based on conventionally fractionated radiotherapy has recently been confirmed by Daly et al. [40]. Further research is consequently required in this field as we do not yet understand the biologic ramifications of high dose per fraction on the normal tissues or tumor itself.