Quantifying the changes in the tumour vascular micro-environment in spinal metastases treated with stereotactic body radiotherapy - a single arm prospective study

Only adult patients who in whom SBRT for SM was clinically indicated were eligible for recruitment. Inclusion and exclusion criteria are described in Table 1. Patients who were undergoing systemic chemotherapy at recruitment, had to observe a one-week wash-out period between their last chemotherapy and first SBRT treatment.

SBRT for SM were carried out as per our department protocol. In brief, patients with lesions at T3 and above were immobilized using a rigid 5-point thermoplastic mask, and patients with lesions below T3 were immobilized using a rigid body bag (Elekta AB, Stockholm, Sweden). CT simulation imaging was performed and reconstructed at 2 mm for RT planning. A dedicated MRI in the region of interest was obtained within a few days of CT simulation. Axial T1 and T2 volumetric MRI sequences were co-registered with CT simulation images to define the target volumes and organs-at-risk. The gross tumour volume (GTV) included bony metastatic disease within the vertebral level, including any para-vertebral extension. Clinical target volume (CTV) included the adjacent compartment, as per contouring guidelines.¹⁴ A 2-5 mm margin was applied in patients with para-vertebral extension to form the final CTV. The CTV was expanded by 2 mm isotropically to create the planning target volume (PTV). The spinal cord was defined using the axial T2 imaging, and a 2 mm planning organ at risk volume (PRV) margin was applied. For lesions at L2 and below, the thecal sac was contoured and a PRV margin was not given. The portion of the PTV that overlapped with the PRV Cord was carved out. Recommended SBRT doses are 24 – 27 Gy in 3 fractions, or 24 Gy in 2 fractions (delivered on alternate days). SBRT treatment was planned using Monaco (Elekta AB, Stockholm, Sweden) using volumetric modulated arc therapy (1 or 2 arcs). We aimed for a CTV coverage of D90/90. Priority was given to the avoidance of critical organs at risk (such as spinal cord PRV, thecal sac, brachial and sacral plexuses), and published dose limits were adhered to.¹⁵ Image guided radiotherapy was performed on Elekta Infinity, with cone-beam CT (CBCT) guidance pre- and post-treatment. For long treatment sessions lasting more than 20 minutes, a mid-treatment CBCT was performed.

Patients were assessed clinically at time of CT simulation, 1-week and 3-months post RT as part of the study. Patients were followed up 3–6 months thereafter (as part of shared care between medical oncology and radiation oncology). Patients were recommended to have a MR imaging every 3–6 months as part of follow-up. Information collected at baseline and 3 months include: primary histology, baseline pain score (VAS), analgesia requirement and spinal instability score (SINS). Pain score was assessed at time of CT simulation (approximately 2 weeks prior to SBRT) and 12-week post-SBRT, using VAS 0-10.

Local tumour response was evaluated according to the MD Anderson criteria¹⁶ at 3 months (complete response (CR), partial response (PR), progressive disease (PD), stable disease (SD)) by a MSK radiologist (15 years’ experience) who was blinded to the treatment (SS). Local recurrence was assessed based on available clinical imaging at the last follow-up. Data was censored at the time of last follow-up. Acute toxicity was assessed at 1 week post RT and late toxicity was assessed at 3 months and during further clinical follow-up, using the Common Terminology Criteria for Adverse Events version 5.0.

All MRI examinations were performed using a 3-T MRI scanner (Siemens Biograph mMR) located at clinical imaging research centre (CIRC), National University of Singapore. MRI was performed at baseline (during time of CT simulation, approximately 2 weeks prior to the start of SBRT), 1-week after completing SBRT (median 8 days, range 6 to 13 days) and at 12 weeks post SBRT (median 92 days, range 82 to 100 days). MRI was obtained in the region of interest (target vertebral level, with 2 levels superior and inferior). The following conventional MRI sequences were acquired prior to contrast administration: sagittal T2, axial T2 (2 mm slice thickness), axial T2 TIRM (3 mm slice thickness), pre-contrast axial T1 VIBE with fat-saturation (2 mm slice thickness).

DCE-MRI was performed using a T1-weighted three-dimensional fast field-echo sequence in the axial plane. Before contrast injection, the pre-contrast T1-weighted fast field echo sequences were acquired at 4 flip angles (5, 10, 15 and 20 degrees) according to the same geometry to calculate the baseline T1 maps. DCE-MRI was performed after an intravenous bolus injection Dotarem (gadoterate meglumine), at 3 ml/s and dose of 0.2 ml/kg. This was done using an automatic injector and was followed by a 15 ml saline flush. Seventy phases were acquired over 5 minutes with a temporal resolution of 4.3 s (flip angle 15 degrees), using parallel imaging. Delayed post-contrast axial T1 VIBE sequences with fat-saturation (2 mm slice thickness) were then acquired after DCE imaging.

The same sequences were performed at 1-week and 12-weeks post SBRT. The length of the aorta or major artery was included in the scan where possible, to minimize inflow artefacts.

Blood was collected, in EDTA tube, at baseline (within 2 weeks prior to SBRT) and immediately after the last fraction of SBRT. Samples were centrifuged for 15 minutes at 1000×g at 2 - 8^oC within 30 minutes of collection, and plasma was removed to be stored at -80^oC. Once plasma was collected from all patients, ASM (Human Acid Sphingomyelinase ELISA Kit, Elabscience) and S1P (Sphingosine 1-phosphase ELISA kit, Echelon Biosciences, UT, USA) was batch evaluated using a semi-quantitative method, using the Sandwich-ELISA method. The optical density (OD) is measured spectrophotometrically at a wavelength of 450 ± 2 nm. The OD value is proportional to the concentration of S1P or ASM. The concentration of S1P and ASM were calculated by comparing the OD of the samples to the standard curve.

Tumour regions of interest (ROI) were drawn in consensus by spine radiation oncologist (BV 12 years’ experience) and two musculoskeletal radiologists (SS 15 years’ experience, LCH 12 years’ experience) on the non-contrast T1 and T2-weighted sequences. The soft tissue component of the metastatic deposits was delineated. For tumours without a predominant soft tissue component, the area showing T1 signal change was delineated.

Control ROI were drawn in the red marrow outside of the radiation volume. The same process was repeated for post-treatment imaging taking into account tumour regression.

All ROIs were drawn taking care to avoid venous structures, hemangiomas, disc spaces, cortical bone and spondylotic changes. ROIs drawn on anatomic reference images were simultaneously and automatically transferred to the corresponding location on the DCE parameter maps.

Raw blood perfusion data, obtained from DCE-MRI was processed and analysed using an in-house programme written on Matlab (MathWorks, Natick, MA, USA). Pre-processing steps included background spatial & temporal smoothing, noise removal. AIF was individually calculated for each acquisition of every patient, using a 3 x 3 (voxel averaged) window placed over an adjacent large vessel (aorta, or large calibre artery). Linear assumption between change in signal intensity and gadolinium concentration was made to convert the signal intensity curve to the concentration-time curve. The “Distributed Parameter” model was used for the calculation of quantitative perfusion parameters.

Both tumour and control ROIs were subjected to the same pre-processing, processing steps and analysis. Parameters were normalized using the readout from the control ROI on each scan. The median parameter value in the tumour, and control, ROIs were taken as the representative for each ROI. For quantitative parameters, including perfusion parameters and correlative plasma markers, mean, median, minimum, and maximum values were calculated. Quantitative perfusion parameters were also compared with the control cases.

We hypothesized a 30% reduction in DCE-parameters post SBRT. Working with a power of 80%, and alpha of 5%, we estimated a sample size of 8 - 10 patients. Differences between pre- and post-treatment values (of DCE-MRI metrics and plasma biomarkers) were compared using a paired T-test. Two-sided p-values of ≤ 0.05 were considered statistically significant. Statistical analysis performed using STATA v 14.

Twelve participants were consented for the prospective study. Two patients withdrew consent before SBRT. Ten patients (with 13 lesions) were treated with SBRT. All 10 patients underwent blood sample collection. Nine of the ten patients underwent MRI assessment at 1 week and 3 months post SBRT. One patient died 1-month post SBRT secondary to pneumonia (unrelated to treatment). Overall, the median age was 62 years (range 51–75 years) and 70% were male. Majority of the patients (46%) had lesions in the thoracic spine, followed by 30% involving the lumbar spine. Kidney and lung were the most common primary tumour sites (30% each respectively). Median follow-up for alive patients was 42 months (range 22.3–54.3 months). Patient demographics and treatment related details are shown in Table 2. The most common indication for SBRT was oligometastasis (8/10). Median prescribed dose was 27 Gy (24–27 Gy) delivered in 3 fractions, with detailed breakdown of the dose and fractionation in Table 2. A majority of participants had prior systemic therapy (6/10). Three patients had prior immunotherapy, and none had prior anti-VEGF therapy.

The normalised mean and median DCE-parameter values at baseline, 1 week and 12 weeks post SBRT are reported in Figure 2A-E. Representative images of Patient 7a are shown in Figure 3.

Figure 2

Figure 3

K trans (mL/100g/min): The normalised mean, median value at baseline was 12.08, 3.65 (range 0.03–62.49), at 1-week post SBRT, 3.10, 2.66 (range 0.01–10.55), and at 12-weeks post SBRT, 1.30, 0.92 (range 0.001–4.60) respectively. P-value comparing baseline & 1 week (p = 0.14), comparing baseline & 12 weeks (p = 0.13), and comparing 1 week & 12 weeks (p = 0.08) were obtained.

Permeability surface product PS (mL/100g/min): The normalised mean, median value at baseline was 13.89, 3.59 (range 0.02–70.83), at 1-week post SBRT, 1.66, 1.21 (range 0.01–6.56), and at 12-weeks post SBRT, 1.29, 0.66 (range 0.001–5.60) respectively. P-value comparing baseline and 1 week (p = 0.12), comparing baseline and 12 weeks (p = 0.12), comparing 1 week and 12 weeks (p = 0.16).

Figure 3

Vp (%): The normalised mean, median value at baseline was 4.14, 2.21 (range 0.82–14.31), at 1-week post SBRT, 4.73, 3.27 (range 0.67–16.90) and at 12-weeks post SBRT, 2.40, 1.45 (range 0.20–9.30) respectively. P-value comparing baseline and 1 week (p = 0.41), comparing baseline and 12 weeks (p = 0.14), comparing 1 week and 12 weeks (p = 0.13).

Ve (%): The normalised mean, median value at baseline was 11.04, 4.03 (range 0.29–43.46), at 1-week post SBRT, 18.88, 5.92 (range 0.06–22.44) and at 12-weeks post SBRT, 6.01, 5.86 (range 0.65– 13.51) respectively. P-value comparing baseline and 1 week (p = 0.11), comparing baseline and 12 weeks (p = 0.17), comparing 1 week and 12 weeks (p = 0.15).

Perfusion, F (mL/100g/min): The normalised mean, median value at baseline was 2.65, 2.37 (range 0.03–9.03), at 1-week post SBRT, 3.62, 3.54 (range 0.08–8.10) and at 12-weeks post SBRT, 1.77, 0.93 (range 0.07–5.23) respectively. P-value comparing baseline and 1 week (p = 0.25), comparing baseline and 12 weeks (p = 0.20), comparing 1 week and 12 weeks (p = 0.01).

ASM levels did not show any significant change post SBRT (baseline mean 6.15 ng/ml vs. post SBRT mean 5.46 ng/ml, p = 0.71). Similarly, S1P levels did not show any significant change (baseline mean 0.67 μM vs. post-SBRT 0.63 μM, p = 0.52) (Figure 4)

Figure 4

All patients tolerated SBRT well, with 3 patients experiencing mild acute toxicity (Grade 1 esophagitis). Three patients developed vertebral compression fractures (VCF) approximately 6 months post SBRT (23% on a per-lesion analysis). Of these, 2 were mild (Grade 1) and pre-existing, and did not require any intervention. One patient developed severe (Grade 3) symptomatic VCF and required stabilisation surgery. All patients had pain improvement 3 months post SBRT, with a median reduction of VAS of 5 (range 3 to 8). Response (MD Anderson criteria) at 3 months – with all patients, who were alive, demonstrating Partial Response (70%) or Stable disease (30%). There was one local recurrence (at 54 months) in the cohort, with data censored at last follow-up. These results are summarised in Table 3.