Escalation of radiotherapy dose in large locally advanced drug-resistant gastrointestinal stromal tumors by multi-shell simultaneous integrated boost intensity-modulated technique: a feasibility study

Background Resistance to conventional dose schemes and radiotoxicity of healthy tissue is a clinical challenge in the radiation therapy of large locally advanced drug-resistant gastrointestinal stromal tumor (LADR-GIST). This study aimed to assess the feasibility of using multi-shell Simultaneous Integrated Boost Intensity-Modulated modality (SIB-IMRT) strategy to provide a safe and effective escalation dose regimen for LADR-GIST. Methods 7 patients with LADR-GIST were selected in this study. The modified SIB-IMRT plans for all patients were generated by delivering different escalation-dose gradients to four ring shaped regions (shells) within the gross tumor volume (GTV). The doses of the central volume of the tumor (GTVcenter) were escalated up to 70–92.5 Gy (25 fractions), while the doses of planning target volume (PTV) and shell-1 were kept at 50.0 Gy. Based on different escalation-dose gradients, the modified SIB-IMRT plans were divided into four groups (SIB-IMRT groups). For comparison purposes, plans obtained by conventional IMRT technique (Con-IMRT) with 50 Gy (25 fractions) were also generated for all patients (Con-IMRT group). All plans were normalized to cover 95% of the PTV with the prescribed dose of 50.0 Gy. The equivalent uniform dose (EUD), relative equivalent uniform dose (rEUD), dose volume histogram (DVH), dose profile, conformity index (CI) and monitor unit (MU) were evaluated in five groups. The Friedman Test was performed to determine whether there were significant differences (P < 0.05). Results Compared with the Con-IMRT group, the EUD of GTV (EUDGTV) and rEUD of SIB-IMRT groups were improved when escalation-dose gradient was increased, and the improvement became significant when the escalation-dose gradient reached 20% of the prescription dose. The rEUD tended to be stable as the escalation-dose gradient went up to 25% of the prescription dose. There were no significant differences in CIs and DVH metrics for OARs between the Con-IMRT group and any SIB-IMRT group, but the significant differences were observed between the SIB10-IMRT group and the SIB25-IMRT group. For the SIB-IMRT groups, as the dose gradient became steeper in the dose profiles, the higher dose was mainly accumulated in the inner part of GTV accompanied with a higher MU. Conclusions The proposed multi-shell SIB-IMRT strategy is feasible in dosimetry for LADR-GIST and can acquire higher therapeutic gain without sacrifice of healthy tissues. It appears that the scheme of delivering 20% of the prescribed escalation-dose gradient to the target volume can provide satisfactory dose irradiation for LADR-GIST, and it should be evaluated in future clinical study.


Background
Gastrointestinal stromal tumors (GISTs) are uncommon neoplasms arising from the mesenchymal tissue of the gastrointestinal tracts [1]. Tumor size, mitosis rate, tumor location, kinase mutation status, and tumor rupture have been widely studied and considered as prognostic indicators [2]. The mutations in tyrosine kinase receptor (KIT) and/or platelet-derived growth factor receptor alpha (PDGFRA) gene have been considered as main factors of the pathogenesis of GISTs. Resection and/or tyrosine kinase inhibitors (TKIs) therapy is always the first choice for GISTs treatment [3]. Although TKIs have greatly improved the therapeutic effect of advanced GISTs, the secondary drug resistance is still common [1]. Some tumors, especially locally advanced drug-resistant gastrointestinal stromal tumor (LADR-GIST), are often technically unresectable and easily resistant to TKIs therapy. As for patients with technically unresectable tumors, there are only few treatment options left once they are resistant to systemic TKIs therapy. Therefore, the management of LADR-GIST has emerged as a challenging clinical problem. Under such circumstances, radiotherapy maybe a valuable option for LADR-GIST management [4].
Since GISTs is traditionally considered to be insensitive to radiation, radiotherapy is not recommended as a therapeutic option and is only for palliative intent in the current treatment guidelines [5,6]. There was not a large-scale, randomized, prospective trial to evaluate the efficacy of radiotherapy for GISTs. So far, a few case reports showed that GISTs were not uniformly radioresistant and that radiotherapy could be a valuable alternative in GISTs management [7,8]. Due to missing consensus on radiotherapy for patients with GISTs, various attempts of radiation dose schemes have been reported [4,9]. In a retrospective study of 15 patients, Cuaron et al. [9] suggested that GISTs were more sensitive to a higher radiation dose. In our previous study [10], the prescribed dose of PTV was set to 50.4 Gy in 28 fractions, while the dose for the center of GTV was boosted to 62-64 Gy. Based on Choi criteria [11], the partial response of the three LADR-GIST patients showed that a reasonable boosted dose for the center of tumor may obtain an effective tumor control with negligible treatment toxicities. These reports demonstrated that a higher and heterogeneous dose distribution in radiotherapy may offer a valuable alternative option for management of LADR-GISTs.
Actually, a uniform dose distribution is commonly applied within target volume and the maximum dose is limited within 110%-115% of the prescription in Con-IMRT. However, the prescription of a homogeneous dose cannot meet the hypoxia requirement of large tumors, as the central area of large tumors has obvious hypoxic changes and should be treated with higher radiation doses [12,13]. To improve the tumor response in Con-IMRT, we generally increase the radiation dosage for the whole target volume. Nevertheless, a higher radiation dose delivered to bulky tumors in Con-IMRT could cause serious side effects as a result of a significant increased irradiation of healthy tissue. Compared with Con-IMRT and sequential IMRT technique, SIB-IMRT allow a dose boost to different target volumes of the tumor and increase tumor response without significant increase of healthy tissue irradiation [14,15]. Due to the particularity of the LADR-GIST, it is a challenge to deliver a stereotactic-like dose to tumor using SIB-IMRT modality. Fortunately, the shell-structure optimization is able to provide a better dose conformity and control dose gradients around the target volume without compromise of target volume coverage [16,17]. Therefore, the shell-structure optimization has been widely used in Stereotactic body radiotherapy (SBRT) plans, and has helped the SBRT plan to get a better conformity, lower radiation dose for OARs, and smaller low-dose areas of normal tissue [18,19]. Given the above, the shell-structures applied in SIB-IMRT may have the potential to improve the tumor response without significant increase in the radiotoxicity of the adjacent normal in LADR-GISTs. In this paper, taking the advantages of multi-shell optimizing and SIB-IMRT technique, we planned a dose-escalation study for large LADR-GIST. This new planning scheme was named as multi-shell SIB-IMRT. The focus of this study was to explore a safe and effective radiation regimen for the LADR-GIST, which can generate a non-uniform and higher dose distribution inside the GTV without exposing OARs to higher radiation doses. Cui et al. Radiation Oncology (2022) 17:216

Patient and volume definition
For dosimetry analysis, 7 patients with LADR-GIST treated in our center from January 2016 to June 2022 were selected in this study, and their tumors' maximum diameters in diagnostic CT imaging were all over 10 cm. The median age of patients was 58 years. These patients underwent R 0 resection of the primary tumor followed by systemic TKIs therapies, and they had never been treated with radiotherapy or embolization before. A few years later (1-10 years), tumor recurrence and/or metastasis, as well as drug resistance were found. Moreover, lesion progression and bulky tumor were detected in these patients (details in Table 1). The GTV diameter ranged from 12.0 to 20.0 cm (17.7 cm ± 2.6 cm). The total abdomen pelvic region was covered in the planning CT scan (GE Medical Systems, Milwaukee, WI). Target volume and organs at risk (OARs) were contoured with Eclipse ™ treatment planning system version 13.5 (Varian Medical Systems, Palo Alto, CA) by the same attending oncologist, and were reviewed by a senior oncologist. We registered the relevant MRI sequences and/or contrasted CT images with the treatment planning CT before contouring. The GTV of the LADR-GIST was delineated on the registration images. After that, the PTV was obtained by GTV plus 5 mm margin. The clinical target volume (CTV) was not defined considering the low rate of lymph node metastasis (1-2%) in GISTs [1,3]. The OARs mainly included rectum, bladder, intestine, and femoral heads [10]. The normal tissue (NT) structure was defined as the body minus PTV. Ring 1 was defined as 2 cm of PTV margin minus 1 cm of PTV margin, and ring 2 was defined as 3 cm of PTV margin minus 2 cm of PTV margin. In addition, GTV was divided into 5 parts by 4 shells based on the concentric contraction within GTV. The first inner shell (shell-1) was 1-1.5 cm concentric contraction from the outer contour of GTV. Each subsequent inner shell (shell-2, shell-3 shell-4) was another 1 cm concentric contraction from the previous one. The most inner part located 4-4.5 cm from the outer contour of GTV covered the center of GTV (GTV center ). Figure 1 illustrates these regions in three standard orthogonal planes and 3D display of OARs for patient 1.

Treatment planning
In this study, a Con-IMRT plan and four SIB-IMRT plans were generated in a dual-arc modality with 6 MV photon beams and calculated by high resolution collapsed cone convolution algorithm by Raystation ™ treatment planning system version 4.7.5 (Raysearchlabs, Stockholm, Sweden) for each patient. All plans were created by an experienced medical physicist and reviewed by a senior medical physicist.
All settings were kept the same for different plans of the same patient except for the dose-gradient of target volumes. Table 2 shows the dose goals used for planning in this study. The prescription dose for PTV and shell-1 were both set to 50 Gy in 25 fractions. The dose of shell-2 was set to 55 Gy (110% of prescription dose), while the dose of shell-3, shell-4 and GTV center were boosted to different dose in four SIB-IMRT groups (the subscript of the plan represents the escalation-dose gradient, see details in Table 2). For comparison, the Con-IMRT plan with 50 Gy for PTV and GTV were generated in the same beam arrangement.

Plan analysis and evaluation
For better comparison, all plans were normalized to 95% of the PTV covered by the prescription dose of 50 Gy and were analyzed using SPSS statistical analysis software package 23.0 (SPSS Inc., Armonk, NY). The Friedman test was performed among five groups with different escalation-dose gradients. Pairwise comparison between any two groups was performed when result of Friedman test was found to be significantly different. A p value < 0.05 was considered statistically significant.

The equivalent uniform dose indexes
To quantify the ability of multi-shell SIB-IMRT strategy to deliver high effective biological doses, the EUD based on DVH was used for GTV, OARs and NT structure [20][21][22]. The EUD was calculated by a MATLAB program from Hiram A. Gay's paper [20].
where the parameter a was a negative value (− 10) for target volume and a positive value (1) for NT structure in this study, Di and vi data pairs were obtained from    Dmax < 40 Gy differential dose volume histogram of a given radiotherapy plan, Vi was the part of the target volume irradiated by a dose Di. n f was the number of fractions. The relative EUD (rEUD) was defined as the ratio of the EUD of GTV (EUD GTV ) to that of NT structure (EUD NT ). A higher rEUD value represented a higher therapeutic gain ratio and indicated a better dose escalation and higher effective biological dose for target volume, but a low dose for normal tissue.

Dose-volume histograms and irradiation for healthy tissues
Besides, D mean (the mean dose), D 1cc (the received dose to 1cm 3 ), D 2cc , V 20 (the percentage volume of the OAR receiving ≥ 20 Gy), V 30 , V 40 , V 50 of OARs were compared between the Con-IMRT group and any SIB-IMRT group.

Conformity index and monitor unit
The conformity index (CI) of PTV and monitor unit (MU) were obtained and compared. A CI value closer to 1 indicated a more conformal dose distribution to PTV and a better normal tissue sparing [23].
where the parameter TV RI was the target volume covered by the prescribed dose, TV was the target volume, V RI was the volume covered by the prescribed dose. Figure 2A showed an example of the DVH comparison of GTV and NT between the Con-IMRT plan and SIB-IMRT plans. Significant dose escalations were observed

Evaluation of target volume in different plans
in GTV of the SIB-IMRT plans, while no significant dose difference in DVH information was found among NT of these plans. As shown in Fig. 3, The SIB-IMRT plans showed higher and more concentrated dose distribution of the target volume compared with the Con-IMRT plan. As shown in Fig. 3A, the dose profile was extracted along the dashed line of dose distribution in each plan. The comparison of profiles (Fig. 4) demonstrated that SIB-IMRT plans with higher EUD GTV had a steeper dose gradient within GTV, while the dose profiles excluding PTV were nearly consistent with that of Con-IMRT plan. Obviously, the SIB-IMRT groups had higher EUD GTV and rEUD than that of the Con-IMRT group (shown in Table 3). The Friedman test showed a significant difference in EUD GTV and rEUD among the five groups (P < 0.05). Subsequently, the result of Pairwise comparisons showed that the EUD GTV and rEUD of SIB-IMRT group were not significantly different from that of Con-IMRT group until the escalation-dose gradient was up to 20% of the prescription dose (For EUD GTV : Con-IMRT VS SIB 20 -IMRT, P = 0.004, Con-IMRT VS SIB 25 -IMRT, P = 0; For rEUD: Con-IMRT VS SIB 20 -IMRT, P = 0.004, Con-IMRT VS SIB 25 -IMRT, P = 0). However, there was no significant difference between any two SIB-IMRT groups optimized with different escalation-dose gradient, except that between the SIB 10 -IMRT group and SIB 25 -IMRT group (For EUD GTV , P = 0.041; For rEUD, P = 0.04). Moreover, with the increase of the dose-escalation gradient, the rEUD of different groups was found to become stable (SIB 20 -IMRT group and SIB 25 -IMRT group have the similar mean rEUD).

Evaluation of healthy tissues in different plans
Friedman test showed significant differences in EUD NT , the EUD of bladder (EUD bla ), the EUD of intestine (EUD ins ) and D mean of bladder among the five groups (P < 0.05) . Subsequently, significant differences between SIB 10 -IMRT group and SIB 25 -IMRT group (P < 0.05) were found in the Pairwise Comparisons. Tables 4, 5, 6 and 7 summarized the comparisons of dosimetry parameters and EUDs for OARs. With the increase of escalation-dose gradient, the DVH metrics (V 20 , V 30 , V 40 , V 50 , D mean , D 1cc , D 2cc ) and EUDs slightly decreased first and then increased. Most DVH indexes of SIB 10 -IMRT group were even lower than those of the Con-IMRT group. An   Fig. 2B. For the bladder, there were lower irradiated volumes for SIB-IMRT plans compared with that for the Con-IMRT plan, except the volumes irradiated with a dose level from 2000 to 4300 cGy in SIB 20 -IMRT plan and SIB 25 -IMRT plan. For the rectum, all SIB-IMRT plans provided lower irradiated volumes compared with the Con-IMRT plan. For intestine, five plans provided similar irradiated volumes as shown in DVH information. For ring 1 structure, the SIB 25 -IMRT plan provided the highest irradiated volume, followed by SIB 20 -IMRT, then followed by Con-IMRT and SIB 15 -IMRT, while the SIB 10 -IMRT plan provided the lowest irradiated volume. The DVHs for ring 2 structure was similar to that for ring 1 structure. Table 8 listed CIs and MUs of different plans. There was a slight change of the mean CIs among the rest groups except for that between the SIB 10 -IMRT group and SIB 25 -IMRT group (P < 0.05). Although the mean MUs (534-721) increased along with the dose escalation in SIB-IMRT groups, no statistically significant differences were found until the dose gradient was increased to 25%.

Discussion
In recent years, radiotherapy has been beneficial to the management of GISTs [4,8,9]. The result has shown that even short-term radiation therapy can effectively alleviate local symptoms, and its toxic and side effects for the localized progression or metastasis of GISTs is acceptable [24,25]. Radiotherapy could play a role in the treatment of GISTs, either as an adjuvant therapy or a definitive    treatment with or without a TKI [7,8,26]. But in previous studies [7,9,27], the tumor was usually under control for only a few months because the total bioequivalent doses were 30-50 Gy by conventional fractionation. In prospective study of Joensu et al. [27], metastases were treated with external beam radiotherapy using either conformal 3D planning or intensity modulated radiotherapy and conventional fractionation to a cumulative PTV dose of 30-40 Gy. Only 2 of 25 GISTs patients achieved partial remission after conventional radiotherapy. There are two main reasons. Firstly, GIST is a most common sarcoma in the gastrointestinal tract. It is relatively radio-resistant to conventional dose regimens perhaps due to the histological relation to soft-tissue sarcoma which has a relatively slow clinical responsiveness to radiation therapy [28]. Secondly, tumor size is one of the most important factors to predict the malignant potential of GISTs. The larger the tumor, the larger the hypoxia area. Especially in the center of the target volume, an inadequate radiation dose will result in poor prognosis. Therefore, a higher equivalent uniform dose and a heterogeneous distribution seems to be needed in GTV for LADR-GIST. The SBRT or SIB-IMRT is the commonly used technique for delivering high and heterogeneous dose to

Groups V 20 (%) V 30 (%) V 40 (%) V 50 (%) D 195cc (cGy) D 1cc (cGy) D 2cc (cGy) D mean (cGy) EUD ins (cGy)
Con-IMRT 28    hypoxic tumor. However, SBRT technique is not suitable for a huge tumor, especially for the tumor located in abdomen because of the radiosensitivity of OARs. In modern radiation therapy techniques, unconventional approaches have also been allowed to enable a large tumor to receive highly heterogeneous doses. Spatial modulation of megavoltage therapy beams, commonly referred to as spatially fractionated radiation therapy (SFRT) (e.g., "Lattice" radiation therapy), is a novel technique which purposefully enables the treated tumor to receive high degree of heterogeneous doses. Developed from the previous high-dose GRID radiotherapy, "Lattice" allows for localized 3D high-dose array within the tumor. This method provides lattice vertices with highly concentrated hot-spot doses, while provides rapid decreased doses between lattices, resulting in a periodic three-dimensional peak-to-valley dose distribution. It has shown a promising future in clinical studies as a method to improve treatment response of advanced and bulky tumors. However, this technique has been applied by only a few centers in clinical practice, more wide and effective application are still needed [29][30][31]. Fortunately, it has been proven that an appropriate SIB-IMRT strategy is able to deliver high biologically effective dose to large tumors without significant increase of healthy tissue irradiation [32][33][34][35]. In the study of Nomiya et al. [35], the dose for central of target area was boosted to 120% of the prescription dose by SIB-IMRT plan without upper dose constraint. Significant tumor regression was observed after radiotherapy, and no related toxic reactions were found. Savino et al. [36] proposed a modified SIB-IMRT for a patient having a huge chordoma with large swelling infiltrating in the right gluteal region and the ipsilateral thigh root. The proposed SIB-IMRT delivered 200% of PTV prescription dose to boost target volume (BTV) with an acceptable increased irradiation dose for healthy tissues surrounding PTV. In our previous study [10], three patients with LADR-GIST were treated by SIB-IMRT plans, and the radiation dose for GTVcenter was escalated up to 125% of the prescribed dose in PTV. During follow-up, three patients were assessed as partial response based on Choi criteria [11]. These results have indicated that a heterogeneous or stereotactic-like distribution based on an appropriate SIB-IMRT modality has advantages in management of LADR-GIST. Undoubtedly, it is a challenge to design an appropriate SIB-IMRT modality for LADR-GIST. In general, LADR-GIST has two key characteristics. One is that the closer to the center of the target area, the more severe the hypoxia is; the other is that the tumor is surrounded by many OARs. Based on these characteristics, the escalation dose multi-shell SIB-IMRT planning strategy was proposed in this study. It was different from the conventional SIB-IMRT technique in which the dose-escalation was simultaneously delivered to different target volumes [37,38]. In this study, only one target volume GTV was intentionally divided into several parts from outside to inside by multi-shell structures. In previous studies, the shell-structures around target volume have been commonly used in SBRT plans to adjust the dose gradient. The number and width of shells should be adjusted according to the target volume, clinical justifications, and planning goals. The use of a shell structure of 1-3 mm was recommended for limiting the high dose conformity. Besides, the two-shell structure was recommended with a distance between each other of 5-10 mm for the optimization of the prescription dose conformity, and a distance between each other of 15-20 mm for the optimization of the proximate dose fall-off. Additionally, a shell structure of 30-50 mm was suggested for limiting the skin doses and hot spots outside the proximate target region [18,39]. Those studies reported that a steeper dose fall-off could be achieved with the increase of the number of shells. However, the quality of the plan exhibits relatively small improvement when the number of the shells was more than 6 [18,19]. Different from the previous studies, four shells inside target volume were utilized for dose escalation in this study. The inner shells (shell-3, shell-4, GTV center ) were utilized for delivering as much radiation dose as possible to the central of GTV to improve the response of tumor. The maximum dose in GTV center could be escalated up to 197% of the prescription dose (Fig. 4). It was worth noting that the maximum EUD GTV could be accumulated to 62.9 Gy (ranging from 59.1 to 66.6 Gy) (Table 3) in the SIB-IMRT groups, which  was approximately 122% of that in the Con-IMRT group. This stereotactic-like dose based on SIB-IMRT strategy formed a "hot core" in the central of target volume (Fig. 3), and induced a higher rate of tumor cell apoptosis in bulky and hypoxic tumors. The exterior shells (shell-1, shell-2) played a major role in avoiding higher radiation dose for OARs, the radiation dose of shell-1 and shell-2 were set to 100%, 110% of the prescribed dose in GTV, respectively. As shown in Tables 3, 4, 5, 6 and 7, the delivered target volume was increased to a significant higher dose without any compromise of healthy tissue sparing. The size of shell-1 was determined by actual position of the target volume and OARs. In this study, shell-1 was defined as 1-1.5 cm concentric contraction from the outer contour of GTV in view of the overlap region between OARs and GTV. Compared with the shell-structure around PTV, the number and width of shells inside PTV was more dependent on the target size. For the same target volume, the number of shells was inversely proportional to the width of shells, that is, the smaller the width, the more the number of shells. A steeper dose escalation (higher EUD GTV ) could be achieved with more shells or higher escalation-dose gradient. But at the same time, the risk of dose spillover will increase (in Fig. 4). Therefore, the setting of shells and escalation-dose gradient would lead to a significant trade-off between the EUD GTV and the healthy tissue dose. In this study, the setting of shells and the escalation-dose gradient were the optimal solution when the dose gradient reached 20% of the prescription dose. If wider shells were provided for plans with higher dose gradient, the contribution of higher dose gradient to EUD GTV will be weakened by the reducing of the number of shells. Therefore, an optimal therapeutic gain ratio can be achieved by a reasonable shell-structure and escalation dose scheme. In order to quantitatively evaluate the therapeutic gain ratio of radiotherapy, the rEUD was proposed. As shown in Table 3, all SIB-IMRT groups have better EUD GTV and rEUD compared with the Con-IMRT group, the EUD GTV and rEUD were improved when escalation-dose gradient was increased, and the improvement became significant when the escalation-dose gradient reached 20% of the prescription dose (P < 0.05). Along with the increase of escalation-dose gradient, the value of rEUD tended to be stable (the rEUD of SIB 20 -IMRT group was similar to that of SIB 25 -IMRT group). The reasons may be as follows, firstly, shell-1 and shell-2 which accounted for nearly 60% of the GTV volume (shown in Fig. 1) received almost the same radiation dose in four SIB-IMRT groups (shown in Table 2); secondly, the difference of escalation-dose gradient was just 5% of prescription dose between any two neighboring SIB-IMRT groups; thirdly, the EUD NT increased along with the increase of the escalation-dose gradient. As shown in Tables 4, 5, 6 and 7, the DVH comparison results in OARs demonstrated that a reasonable utilization of shell structure enabled SIB-IMRT plan to generate a required escalation-dose gradient. In our study, the DVH metrics and EUDs for OARs showed a trend of slightly decreasing first and then increasing with the increase of escalation-dose gradient (see Fig. 2B and Tables 4, 5, 6 and 7). There were no significant differences of DVH parameters for OARs between the Con-IMRT group and any SIB-IMRT group. However, significant differences were found in EUD NT , EUD bla , EUD ins , CI and D mean of bladder between the SIB 10 -IMRT group and the SIB 25 -IMRT group (P < 0.05). In addition, most DVH indexes of the SIB 10 -IMRT group were found to be even lower than those of the Con-IMRT group (see Tables 4,  5, 6 and 7). The results implied that the SIB 10 -IMRT group had the advantage of OAR's sparing caused by the relaxation of GTV upper dose constraint in SIB-IMRT plans in which the relative weight of all other constraints increased [10]. However, the advantage was weakened even offset with the increase of escalation-dose gradient. Thus, when the escalation-dose gradient reached 25% of the prescribed dose, the difference became significant compared with the SIB 10 -IMRT group. It was implied that 20% of the prescribed dose may be the optimal escalation-dose gradient, which can provide a high therapeutic gain ratio without overdosing the OARs compared with the Con-IMRT group. It was also safe and effective to set the escalation-dose gradient to 25% of the prescribed dose in this target volume, while the DVH metrics of OARs tended to increase. These results also demonstrated that the escalation-dose gradient delivered to target volume could not be increased limitlessly. Given the above, the proposed multi-shell SIB-IMRT strategy is able to provide high escalation dose to a LADR-GIST while maintaining the similar dose level for OARs. The detailed escalation dose scheme must be synthetically considered together with tumor size, overlap region between the tumor and OARs, prescribed dose, and shell-structure and so on.
Nonetheless, there are still several limitations in this study. Firstly, the sample size was small, and only seven patients were studied because of the rigid inclusion criteria. Maybe 20% of the prescribed dose is not the optimal dose gradient as more samples are included in the future. However, it would not affect the direction provided in this study, that is, an optimal therapeutic gain ratio can be achieved by reasonable shell-structures and escalation dose scheme. In our study, the results indicated that initial optimization dose gradient could be set to 20% of the prescribed dose. The final optimization dose gradient should be adjusted according to the actual situation. A large sample size is required and further investigation need to be implemented for the optimal dose gradient. Secondly, the variation of shell-structure was not intensively investigated, because our focus was on the exploration of dose gradient regimen. The selection of shell parameters would be an attractive topic, because it is dependent on the target volume and OARs. Thirdly, it is only a dosimetry comparison, which means further clinical trials is still required.

Conclusion
The proposed multi-shell SIB-IMRT strategy is safe and reliable in dosimetry. It is able to provide a feasible scheme for patients having LADR-GIST by providing high dose distribution for the center of a large tumor. In our study, the dosimetry evaluation demonstrated that it is safe and optimal to apply 20% of the prescribed dose as the escalation-dose gradient for LADR-GISTs. However, the scheme of delivering 20% of the prescribed dose gradient should be further evaluated in future clinical study.