Introduction
Craniospinal irradiation (CSI) planning is complicated due to the large treatment length and multiple isocentres needed in order to treat the patient accurately with radiotherapy using linear accelerator. Special care is needed at the field junctions, mainly to minimise setup variations and positional displacement, which lead to dose variations, most often noted with 3-dimensional conformal radiotherapy technique (3DCRT) planning. The main drawback of 3DCRT planning for CSI cases is the increases in the organ at risk (OAR) doses and difficulty in field matching at the junctions. Recent linear accelerator technical innovations have allowed conformal, homogenous and junctional shift-free Reference Seppala, Kulmala, Lindholm and Minn1 delivery, using either intensity-modulated radiotherapy technique (IMRT) or volumetric-modulated arc therapy (VMAT).
Georg et al. Reference Georg, Knoos and McClean2 reported that IMRT and VMAT technique fluences were modified by adjusting the multileaf collimator (MLC) speed, dose rate, gantry speed to get the desired fluence so that flattened beams (FBs) were not required when using these techniques. In medical linear accelerator, the flattening filter is located between primary collimator and monitor chamber, and the main function is to produce a uniform dose at reference depth. The characteristics of FBs and clinical characteristics have been described extensively in the literature. Monte Carlo simulation and experimental studies performed on flattening filter-free (FFF) beam in various linear accelerators have shown that the removal of flattening filter in the treatment head leads to a reduction in head scatter, increase in dose rate, reduced out-of-field dose and faster treatment delivery times. In the absence of the flattening filter, the beam intensity is maximum at the centre and reduces rapidly at the field edge. With VMAT, the non-tumour tissues get exposed to significant scatter dose, and the use of FFF beam has been as a solution to this issue. The adoption of these delivery techniques has also reduced the incidence of early and late effects of radiation, Reference Bloom3 such as hypothyroidism, pneumonitis, infertility and secondary cancers. Due to a large number of monitor units (MUs) required for IMRT planning, the probability of developing secondary cancers has been reported by several researchers. Reference Followill, Geis and Boyer4–Reference Kry, Salehpour and Followill6
The literature is limited on dosimetric comparisons of FB versus FFF in CSI cases studies using different calculation algorithms. In our study, the beams were placed posteriorly, particularly spine fields (partial arcs), and therefore encounter highly heterogeneous regions such as spine, lung and brain. Furthermore, fast and accurate modern algorithms incorporating heterogeneity corrections have also been developed recently. A variety of dose calculation algorithms such as pencil beam convolution (PBC) algorithm, analytical anisotropic algorithm (AAA) and Acuros XB (AXB) algorithm were published and implemented on Eclipse treatment planning platform (Varian Medical Systems, Palo Alto, USA). The original description of AXB algorithm was reported by Vassiliev et al., Reference Vassiliev, Wareing, McGhee, Failla, Salehpour and Mourtada7 and its implementation on Eclipse treatment planning system (TPS) was reported by Fogliata et al. Reference Fogliata, Clivio, Vanetti, Mancosu and Cozzi8 Failla et al. Reference Failla, Wareing, Archambault and Thompson9 reported that AXB algorithm takes in to account the heterogeneity of the patient directly, by solving the linear Boltzmann transport equation (LBTE) numerically. The LBTE describes the radiation particles behaviour at the macroscopic level for a given volume of interest. The AXB algorithm-calculated dose distributions were comparable to the Monte Carlo methods with relatively high speed. The present study was designed to distinguish any difference in FB versus FFF (6 and 10 MV) using two different dose calculation algorithms for CSI cases.
Materials and Methods
Patient selection, positioning, imaging and contouring
Patients were selected retrospectively from our institutional database. All patients were positioned supine on a vac-loc immobilisation system with concurrent thermoplastic immobilisation device for the head. An anteroposterior topograph was performed on the CT scanner (Siemens Somatom Sensation) to achieve a straightened spinal column. A CT scan (3 mm slice thickness) was then acquired for treatment planning without contrast administration. The following OAR were delineated: lens, eyes, optic nerves, brain stem, optic chiasm, pituitary, parotids, cochlea, oral cavity, dysphagia and aspiration-related structures (DARS), thyroid, oesophagus, common lung, heart, spleen, stomach, kidneys, bowel, bladder, rectum, femur and testis. Clinical target volume (CTV_CSI) for CSI included the entire brain (and meninges) and spinal cord (from the foramen magnum to S2–S3 caudally, as determined by MRI) and laterally, the volume was extended to cover the nerve roots. Planning target volume (PTV) was generated after applying an isotropic 7-mm margin around spinal CTV.
Treatment planning and analysis
VMAT plans were generated on the Eclipse TPS version 11.0 and capable to deliver on True Beam platform, equipped with high-definition MLC (HDMLC). After removal of the flattening filter, the dose calibration was performed for the delivery platform using TRS 398 protocol 10 (1 cGy = 1MU at Dmax for 10 × 10, field size). The HDMLC contains 120 leaves (inner 32 MLC pairs are 2·5 mm width, and outer 28 leaf pair is 5 mm width at the isocentre). Optimization was performed after selecting a maximum dose rate of 600 MU/min for 6 MV FB plans and 1,400 MU/min for 6 MV FFF plans. For 10 MV FB plans, the dose rate was 600 and 2,400 MU/min for 10 MV FFF plans. In our study, five patients were selected because of longer optimization time and complicated planning. Each patients, eight plans were generated by changing the beam energy and calculation algorithm. The plans were 6 MV FB, 6 MV FFF, 10 MV FB and 10 MV FFF for AAA-calculated plans. Similarly, in the case of AXB algorithm, the plans were 6 MV FB, 6 MV FFF, 10 MV FB and 10 MV FFF. In total, 40 plans were generated.
Four isocentres were used to plan each patient due to the treatment length and the limitations of HDMLC. The source to isocentre distance was 100 cm. The isocentre locations were the brain, cervical, thoracic and lumbar spine. For the cranial field, dual arcs (360°) were used with collimator rotation of 30° for the clockwise (CW) arc and 330° for counter-clockwise (CCW) arc. In the case of spine fields, dual partial arcs (160°) were used posteriorly, and collimator rotations of 10° for CW and 350° for CCW were applied. The planning goals were that 95% of PTV should receive 100% prescribed dose and dose to OARs should be as low as possible.
All plans were analysed in terms of conformity index (CI), homogeneity index (HI) and low- and high-grade index. HI for TV was calculated to evaluate the dose homogeneity in the target.
HI = (D2%−D98%)/D50%, where D2%, D50% and D98% were dose received by 2%, 50% and 98% TV, respectively. HI value of zero represents homogeneous dose distribution. 11
CI: A ratio to evaluate the coverage criteria of the prescription dose (PD) for the treatment plans. CI = Volume within 98% isodose line divided by TV. Unity value of CI indicates the good dose conformity.
Low- and high-gradient indices were calculated using the following formula. Reference Paddick and Lippitz12
$${\rm{High }} - {\rm{gradient}}{\mkern 1mu} {\rm{index }}\left( {{\rm{G}}{{\rm{I}}_{{\rm{High}}}}} \right){\rm{ = }}{{\rm{V}}_{{\rm{50}}\% }}{\rm{PD/}}{{\rm{V}}_{{\rm{90}}\% }}{\rm{PD}}.$$
$${\rm{Low }} - {\rm{gradient}}{\mkern 1mu} {\rm{index }}\left( {{\rm{G}}{{\rm{I}}_{{\rm{Low}}}}} \right){\rm{ = }}{{\rm{V}}_{{\rm{25}}\% }}{\rm{PD/}}{{\rm{V}}_{{\rm{50}}\% }}{\rm{PD}}.$$
where V25%, V50% and V90% were volume receiving by 25%, 50% and 90% of the PD, respectively. Various dosimetric parameters such as D98%, D95%, D50%, D2%, D98% to PTV and dose to other OARs were noted from the dose–volume histograms. Statistical analysis was performed using paired sample t-test and a p-value of less than 0·05 significant.
Results
A comprehensive description of all comparisons performed is shown in Tables 1–4.
Table 1. Target, body-PTV and MU comparison between FB and FFF beam for AAA-calculated plans
Table 2. OARs comparison between flattened and FFF beam for AAA-calculated plans
Table 3. Target, body-PTV and MU comparison between flattened and FFF beam for AXB-calculated plans
Table 4. OARs comparison between flattened and FFF beam for AXB-calculated plans
Target coverage and OAR sparing for 6 MV FB versus 6 MV FFF using AAA algorithm
The 6 MV FB and FFF beam generated dosimetrically equivalent target coverage (D98%, D95%, D50% and D2%, p = NS for all comparisons) using AAA algorithm. The V110% of PTV was higher with 6 MV FFF plan compared to 6 MV FB plan (1·6 ± 0·63 versus 2·7 ± 0·73; p = 0·036). The 6 MV FB plan was more homogenous and conformal (HI = 0·11 ± 0·0 versus 0·12 ± 0·01, p = 0·004; CI = 1·07 ± 0·02 versus 1·08 ± 0·03, p = 0·034). The low-gradient indices were lower with 6 MV FFF plan (2·26 ± 0·16 versus 2·22 ± 0·15, p = 0·331), and the high-gradient index value is same 6 MV FB and FFF (1·93 ± 0·11) VMAT plans.
The 6 MV FFF plan spared the lens, eyes, parotids, cochlea, thyroid, common lung, heart, liver, kidney, bowel, bladder, rectum, femur, testis, body minus PTV region and low-dose spillage region (corresponding to 1 Gy, 2 Gy, 3 Gy, 4 Gy and 5 Gy). A significant difference was observed in HI, CI, V110%, common lung Dmean, bowel, treatment MUs and low-dose spillage region. The treatment MUs were higher in 6 MV FFF beam plan in comparison to 6 MV FB (1,296·4 ± 221 versus 1,616·8 ± 303, p = 0·004).
Target coverage and OAR sparing for 6 MV FB versus 6 MV FFF using AXB algorithm
The 6 MV FB and FFF also generated dosimetrically equivalent target volume coverage (D98%, D95%, D50% and D2%, p = NS for all comparisons) using AXB algorithm. The V110% of PTV was higher with 6 MV FFF plan (1·52 ± 0·73) when compared to 6 MV FB plan (1·22 ± 0·63, p = 0·023).
The 6 MV FB plan was more homogenous and conformal (HI = 0·12 ± 0·0 versus 0·13 ± 0·01, p = 0·016; CI = 1·08 ± 0·02 versus 1·08 ± 0·03, p = 0·041). The high- and low-gradient indices were lower with 6 MV FFF plan (1·86 ± 0·11 versus 1·85 ± 0·11, p = 0·137 for high-gradient indices (HGI) and 2·33 ± 0·12 versus 2·31 ± 0·1, p = 0·327 for low-gradient indices (LGI)).
The 6 MV FFF plan spared the lens, eyes, optic nerves, brain stem, parotids, thyroid, common lung, heart, liver, kidney, bowel, bladder, body minus PTV region and low-dose spillage region (corresponding to 1 Gy, 2 Gy, 3 Gy, 4 Gy and 5 Gy). A significant difference was observed in HI, CI, V110%, common lung, bowel, testis, treatment MUs and low-dose spillage region. The treatment MUs were higher in 6 MV FFF plan when compared to 6 MV FB (1,306 ± 221 versus 1,608·8 ± 303, p = 0·005).
Target coverage and OAR sparing for 10 MV FB versus 10 MV FFF using AAA algorithm
The dose calculated by the AAA algorithm generated dosimetrically equivalent plans using 10 MV FB and 10 MV FFF, except higher V110% of 6·3%, in FFF plan and 3·4% for FB plan (p = 0·045). The 10 MV FB plan was more homogenous and conformal when compared to FB (HI = 0·12 ± 0·02 versus 0·15 ± 0·02, p = 0·035; CI = 1·07 ± 0·03 versus 1·1 ± 0·03, p = 0·026). The high- and low-gradient indices were lower with 10 MV FFF plan (1·94 ± 0·08 versus 1·91 ± 0·09, p = 0·415 for HGI and 2·26 ± 0·14 versus 2·22 ± 0·10, p = 0·219 for LGI).
The 10 MV FFF plan spared the lens, eyes, oral cavity, common lung, heart, liver, spleen, stomach, kidney, bowel, bladder, rectum, body minus PTV region and low-dose spillage region (corresponding to 1–5 Gy). A significant difference was observed in HI, CI, V110%, eyes, common lung, stomach, kidney, bladder, testis, treatment MUs and low-dose spillage region. The treatment MUs were higher in 10 MV FFF in compared to 10 MV FB plan (1,217·8 ± 211 versus 1,529·4 ± 156, p = 0·001).
Target coverage and OARs sparing for 10 MV FB versus 10 MV FFF in AXB algorithm
The dose calculated by AXB algorithm also generated dosimetrically equivalent plans using 10 MV FB and 10 MV FFF, except higher V110% of 4·26%, in FFF plan and 3·54% for FB, p = 0·041. The 10 MV FB plan was more homogenous and conformal (HI = 0·13 ± 0·02 versus 0·17 ± 0·02, p = 0·040, CI = 1·08 ± 0·03 versus 1·11 ± 0·03, p = 0·41). The high- and low-gradient indices were lower with 10 MV FFF plan (1·85 ± 0·08 versus 1·85 ± 0·09, p = 0·871 for HGI and 2·35 ± 0·07 versus 2·3 ± 0·06, p = 0·327 for LGI).
The 10 MV FFF plan spared the lens, eyes, parotids, cochlea, oral cavity, common lung, spleen, stomach, kidney, bowel, bladder, body minus PTV region and low-dose spillage region (corresponding to 1–5 Gy). A significant difference was observed in CI, HI, V110%, stomach, bowel, treatment MUs and low-dose spillage region. The treatment MUs were higher in 10 MV FFF plan in compared to 10 MV FB (1,226·2 ± 212 versus 1,529·4 ± 156, p = 0·002).
Discussion
Historically, CSI patients were treated with 2D planning, and target localization was based on bony landmarks. The main drawback of 2D planning Reference Mah, Danjoux, Manship, Makhani, Cardoso and Sixel13 is that MUs were calculated manually at a specified depth, and computer-based dose calculations were not performed. The lack of accurate dose calculation was also correlated with the probability of neuraxial recurrences Reference Vijay, Arun and Chakraborty14 due to inadequate dose in the PTV. However, the low-dose spillage has been reported to be significantly more with IMRT/VMAT technique compared to the conventional technique. Saini et al. Reference Saini, Aggarwal and Sharma15 study showed that the low-dose spillage volume is more in VMAT plan, and this directly increases the integral dose leading to an increased risk of secondary cancers. Also, most of the dosimetric studies on CSI reported in the literature have compared 3DCRT and IMRT/VMAT using FBs.
AAPM report TG 158 Reference Kry, Bednarz and Howell16 defined the out-of-field dose or non-target dose may be defined as dose outside the PTV, which is unintentionally irradiated. There are two types of non-target doses. The first is located near the field border that is along the beam entry and exit direction and is called in-field non-target dose. The dose which is located at a distance from the field border is called out-field non-target dose. The non-target dose is further classified into three levels. High doses (>50% of PD or >30 Gy), intermediate doses (5–50% of the PD or 3–30 Gy) and low doses (3 Gy or 5% of the PD). Fogliata et al. Reference Fogliata, Bergström and Cafaro17 reported that the non-target mean dose in CSI cases was in the range of 3–8 Gy. In our analysis, the non-target mean dose (body-PTV) was in the range of 6·28–6·38 Gy for both 6 MV FB and FFF beam plans. In the case of 10 MV FB and FFF beam plans, the non-target mean dose was in the range of 6·28–6·6 Gy.
Recently, the utilisation of FFF beam in stereotactic body radiotherapy (SBRT) lung Reference Vassiliev, Kry and Chang18 using gated delivery has increased due to the increased dose rate, lesser beam-on time (BOT), increased beam intensity at the central axis, reduction in peripheral dose and decrease in out-field scattered dose. The BOT for tomotherapy is 32 minutes, which is higher compared to smart arc and 3DCRT plans. Reference Pamela, Sotirios, Alonso, Carlos, Panayiotis and Niko19,Reference Wang, Golden and Ting20 In our analysis, all plans had a BOT of approximately 5 minutes which would minimise the probability of intra-fraction motion of the patient. Fu et al. reported that the treatment time depends on dose rate, MLC characteristics and delivered dose per fraction. The time reduction is insignificant with conventional dose per fraction (2 Gy/fraction). Reference Fu, Dai and Hu21
Operating the linear accelerator at higher energies (greater than 8 MV), neutron production increases, due to the interaction with high-density materials such as primary collimator, flattening filter and jaws. Kry et al. Reference Kry, Titt and Pönisch22 reported that with 18 MV FFF beam, IMRT plan reduces the secondary cancer risk from 2·9% to 0·9% due to lesser neutron production as compared to 18 MV FB.
Foo et al. Reference Foo, McCullough, Foote, Pisansky and Shaw23 also reported the probability of developing hypothyroidism after radiotherapy, though this analysis was performed with Head and Neck anthropomorphic tissue-equivalent phantom. The doses recorded on the thermoluminescent dosimeter and diode at thyroid gland were in the range of 10–80 Gy, and the authors concluded that this could potentially lead to hyperparathyroidism. Emami et al. Reference Emami, Lyman and Brown24 reported that the clinical probability of developing hypothyroidism was about 8% after completion of RT dose of 45 Gy. Another study by Yoden et al. Reference Yoden, Maruta and Soejima25 suggested that the volume of the thyroid gland receiving 30 Gy was predictive of the risk of developing hypothyroidism and this was also reported by Johansen et al. Reference Johansen, Reinertsen, Knutstad, Olsen and Fossa26 and Tell et al. Reference Tell, Lundell and Nilsson27 In our study, the V30Gy was within 3% in all the plans and differences were not significant.
Fogliata et al. Reference Fogliata, Nicolini and Clivio28 studied 6 and 10 MV flattened and FFF beam beams, and their characteristics were measured in a water phantom. They reported a measured difference of 2% for mechanical wedge and 1% for open beam in AXB algorithm compared with the AAA algorithm. Kan et al. Reference Kan, Leung and Yu29 studied the impact of AXB algorithm on patients undergoing radiotherapy for nasopharyngeal carcinoma. IMRT/VMAT plans were generated using AAA and AXB algorithms, and the authors concluded that calculations performed using the AXB algorithm resulted in lower target coverage and least dose to surrounding OARs. In our analysis as well, the AXB algorithm reduced the doses to nearby OARs with lower target coverage in both FB and FFF plans. Fogliata et al. Reference Fogliata, Nicolini and Clivio30 compared the dose distribution of AXB and AAA in non-small cell lung cancers and reported the dose deviation in the PTV was 3% and in the OARs was 3%.
Cao et al. Reference Cao, Ramaseshan and Corns31 reported that dose to the lens dose in CSI cases was 9·1 Gy, and Lee et al. Reference Lee, Brooks, Bedford, Warrington and Saran32 reported a dose in the range of 18–20 Gy. In our analysis, 6 MV FFF (< 8.1 Gy) and 10 MV FFF (< 9 Gy) plan led to better sparing of the lens compared to the FB plan. Shaitelman et al. Reference Shaitelman, Grills and Liang33 reported the incidence of Grade 3 pneumonitis after RT for non-small cell lung cancer, which was 2%, 4% and 24% for bilateral lung volumes receiving V5 < 35%, V5 35–50% and V5 > 50%, respectively. In our analysis, V5Gy was 50%, the V5Gy volume was less in 6 MV FFF and 10 MV FFF plans when compared to 6 and 10 MV FB plans.
Khan et al. Reference Khan, Villarreal-Barajas, Lau and Liu34 studied the effect of AXB and AAA in SBRT lung, and the authors concluded that AXB required 2% more MUs compared to AAA to deliver similar PDs. Our analysis demonstrated that the AAA-calculated 6 MV FFF plan needed 25% more MUs and 26% more MUs for 10 MV FFF plan in comparison to FB VMAT plans. Similarly, AXB-calculated 6 MV FFF beam plan needed 23% more MU and 25% more MUs for 10 MV FFF plan in comparison to FB VMAT plans. Many studies Reference Seppala, Kulmala, Lindholm and Minn1,Reference Fogliata, Bergström and Cafaro17,Reference Pamela, Sotirios, Alonso, Carlos, Panayiotis and Niko19,Reference Lee, Brooks, Bedford, Warrington and Saran32 on CSI cases by using five patients data to plan in FB by comparing the different treatment techniques (3DCRT versus IMRT or VMAT) and multi-institutional treatment experience. There are a limited number of studies comparing flattened and FFF beam in CSI cases. In our study, 40 plans were generated using five patient CT images and compared the FB and FFF beam plans (6 and 10 MV) using AAA and AXB algorithm by VMAT technique.
Conclusions
The 6 and 10 MV FFF beam with VMAT technique generated a highly conformal, homogenous plan and resulted in better sparing of OARs in CSI cases. The shorter BOT and lower normal tissue dose will lead to a lowered probability of intra-fraction motion and consequent dosimetric errors besides reducing the probability of developing a secondary cancer.
