Patients selection and treatment planning
Between November 2010 and September 2011 at Istituto Clinico Humanitas (ICH) 55 patients with 73 primary or metastatic liver tumours were treated by TrueBeam in an on-going phase II prospective protocol approved by the internal ethical committee. Aim of the protocol, approved in late 2009, was the evaluation of local control in patients with liver malignancies; secondary endpoints were the acute and late toxicities, and evaluation of overall survival. According with Fleming approach, to demonstrate local control of at least 80% with a power of 90%, at least forty-four patients in three years were necessary to complete the study. At present, all the patients are recruited and the follow-up are ongoing therefore no analysis regarding local control or toxicities will be reported in the present study.
Of the patients evaluated, 11 had primary liver tumours, and 44 had hepatic metastases from colon (23), biliar duct-pancreas (6), breast (5), and other sites (10). All the patients had been considered unfit for surgery or other non-surgical treatments at the time of radiation. Chemotherapy was stopped at least 3 weeks before SBRT and withheld until disease progression. Median patient age was 63 years (range: 43-83). Only patients with at least 1000 cm3 of liver free from the disease and with ≤ than 3 lesions were considered eligible to the protocol.
Disease extension was evaluated in all cases by Computed Tomography (CT) with contrast. Although they were not inclusion criteria of the study, magnetic resonance imaging (MRI) and positron emission tomography/CT using FDG tracer (PET/CT) were acquired in respectively 2 and 14 patients. Both MRI and PET/CT were acquired for those patients whose treatment indication for radiotherapy was justified by these diagnostic exams; thus PET and MRI were only an integration in the target definition since the use of PET and MRI in defining the tumour is under evaluation and will be topic of future work. Computed tomography scans for planning were acquired for all patients positioned supine with their arms above the head; patients were immobilized by means of a thermoplastic body mask including a Styrofoam block for abdominal compression to minimize internal organ motion. Contrast free and 3 phases contrast-enhanced planning CT scans were acquired in free quiet breathing mode at 3 mm slice thickness with a stereotactic body frame to localise the isocentre. Breath hold in simulation CT was not mandatory in the study as many patients were unfit to maintain breath hold for many seconds with the compressor. In case of collaborative patients the simulations CT were performed in voluntary exhale breath hold.
Furthermore in case of lesions located in the VII or VIII hepatic segment or in case of liver cupola shift greater than 5 mm on the four simulation CTs, a four dimensional CT (4D-CT) was performed to best define the target margin. In addition, in two cases presenting internal clips due to previous surgery, the 4D-CT scan was acquired and the SBRT was performed in gated modality with internal marker tracking by 2D imaging. This option was released in July 2011 with TrueBeam version 1.5, and thus the first patients were not treated with internal marker tracking. This topic will be deepened in a specific paper.
The gross tumor volume (GTV) included macroscopic disease defined on CT as well as on PET if available. The clinical target volume (CTV) was defined equal to the GTV. The planning target volume (PTV) was generated by taking into account both the internal margin (IM) and the set-up margin (SM). Since SM was maintained at a minimum by the cone-beam CT (CBCT) daily verification of set up variations, the overall CTV-PTV margin was prescribed as 8-12 mm in the cranial-caudal axis and 4-6 mm in the anterior-posterior and lateral axes, allowing mainly for residual intra-fraction target motion as well as for inaccuracies in CBCT image interpretation [
26,
27]. The organs at risks (OAR) considered were: healthy liver, spinal cord, kidneys, stomach, duodenum, heart, small bowel, oesophagus and ribs, in relation with the lesion location.
The isodose distribution applied during SBRT typically includes planned heterogeneity within the tumor intended to intensify the dose within the tumor. In this protocol, the isodose line prescribed to cover the PTV was at least 67% of the prescribed dose (range 67-95%), trying to maximizing it up to 95% [
28]. Dose prescription was set to 75 Gy in 3 consecutive daily fractions. For OARs, plans were required to meet the following objectives: V
15 Gy (volume receiving 15 Gy) < (total liver volume - 700 cm
3) for healthy liver, D
0.1 cm3 for spinal cord < 18 Gy (dose at a volume of 0.1 cm
3 should be lower than 18 Gy), V
15 Gy < 35% for both kidneys, V
21 Gy < 1% for duodenum, small bowel, oesophagus, and stomach, V
30 Gy < 1% for heart; D
30 cm3 < 30 Gy for ribs were considered as a secondary objective [
3,
29]. In case of overlap between PTV and duodenum or stomach, the priority was given to the OAR cropping the PTV to comply with the OAR limits.
Treatment delivery
All plans were designed and optimised with RA technique using the optimizer PROIII for a Varian TrueBeam equipped with a Millennium multi-leaf collimator (MLC) with a leaf width of 5 mm at the isocentre. RA plans were designed using full (i.e. 360°) or partial (i.e. around 200°) multiple arcs according in order to achieve the best dose distributions. Specifically, partial arcs were used in cases of (1) lesion located far from the median axis (i.e. more than 10 cm) to do not collide the gantry with the couch induced by laterality of the couch, and (2) lesion very close to serial OARs (i.e. heart, gastro-intestinal organs) to best protect them. Where possible, coplanar arcs were employed to fasten the delivery time, otherwise, non coplanar arcs arrangements were used with two perpendicular couch positions. In particular, the non-coplanar approach was adopted in multi-lesions cases only. All dose distributions were computed with the Analytical Anisotropic Algorithm (AAA) (version 10.0.28) implemented in the Eclipse planning system with a calculation grid resolution of at maximum 2.0 mm.
Treatment was delivered in 3 consecutive working days, with the patient keeping a 3-hour fast to avoid gross displacement of stomach. Treatment delivery included stereotactic frame localization in the first session aiming to a preliminary isocentre positioning followed by image guidance with on-line couch adjustment at each fraction by means of cone beam CT (CBCT). Couch repositioning was operated after automatic matching of CBCT images to reference planning CT, followed by manual refining. The shift values were analyzed for all patients and random (σ) and systematic (Σ) population errors were calculated according to Van Herk approach [
30]. In two cases the delivery was performed in respiratory gated modality. This approach allows the radiation beam to be turned off when respiratory movements place the target outside of the predetermined positioning parameters, and to resume the radiation when the target falls back within the accepted alignment. In particular, the respiration path was revealed with RPM system (Varian) and internal markers, previously detected on the simulation CT. The markers were detected by instantaneous kV-portal images acquired before each beam-on phase. On each of these kV-portal images, a circular region of interest (ROI) of 5 mm radius defined the theoretical marker's position for each projection; the radiation oncologist could then verify the instantaneous marker's positions to be inside the ROI, highlighting possible internal organ motions.
Before treatment, each plan was verified to assess dosimetric agreement between computed and delivered dose distributions. This quality assurance process was performed with two independent quality assurance systems (MatriXX and Gafchromic). The results of these measurements were scored in terms of the Gamma Agreement Index (GAI) based on the γ of Low analysis [
31] with thresholds: Distance to Agreement = 3 mm, ΔDose = 3%.
Data analysis and statistics
Technical parameters of delivery were scored in terms of number of arcs, total number of monitor units (MU), monitor units per Gy (MU/Gy), total beam on time, total treatment time (the time in which the patient is in the bunker), and isocenter shift. Dosimetric quality of treatments was measured on the basis of dose volume histogram (DVH) analysis. For CTV and PTV the following data were reported: target coverage (mean, D1%, D95%, V95%, V107%) and conformity for PTV. Conformity index (CI95%) was defined as the ratio between the volume of patient irradiated at 95% of the prescribed dose and the target volume. For OARs, the mean dose, the maximum dose (Dxcm3) and appropriate values of VxGy (volume receiving at least x Gy) were scored. The data were reported separately for patients with 1 lesion and with 2-3 lesions. The Kolmogorov-Smirnoff test was applied to the data to evaluate if the data were normally distributed.