Automated treatment planning as a dose escalation strategy for stereotactic radiation therapy in pancreatic cancer

Abstract Purpose To assess the feasibility of automated stereotactic volumetric modulated arc therapy (SBRT‐VMAT) planning using a simultaneous integrated boost (SIB) approach as a dose escalation strategy for SBRT in pancreatic cancer. Methods Twelve patients with pancreatic cancer were retrospectively replanned. Dose prescription was 30 Gy to the planning target volume (PTV) and was escalated up to 50 Gy to the boost target volume (BTV) using a SIB technique in 5 fractions. All plans were generated by Pinnacle3 Autoplanning using 6MV dual‐arc VMAT technique for flattened (FF) and flattening filter‐free beams (FFF). An overlap volume (OLV) between the PRV duodenum and the PTV was defined to correlate with the ability to boost the BTV. Dosimetric metrics for BTV and PTV coverage, maximal doses for serial OARs, integral dose, conformation numbers, and dose contrast indexes were used to analyze the dosimetric results. Dose accuracy was validated using the PTW Octavius‐4D phantom together with the 1500 2D‐array. Differences between FF and FFF plans were quantified using the Wilcoxon matched‐pair signed rank. Results Full prescription doses to the 95% of PTV and BTV can be delivered to patients with no OLV. BTV mean dose was >90% of the prescribed doses for all patients at all dose levels. Compared to FF plans, FFF plans showed significant reduced integral doses, larger number of MUs, and reduced beam‐on‐times up to 51% for the highest dose level. Despite plan complexity, pre‐treatment verification reported a gamma pass‐rate greater than the acceptance threshold of 95% for all FF and FFF plans for 3%‐2 mm criteria. Conclusions The SIB‐SBRT strategy with Autoplanning was dosimetrically feasible. Ablative doses up to 50 Gy in 5 fractions can be delivered to the BTV for almost all patients respecting all the normal tissue constraints. A prospective clinical trial based on SBRT strategy using SIB‐VMAT technique with FFF beams seems to be justified.


| INTRODUCTION
Pancreatic carcinoma is the fourth leading cause of cancer death in developed countries. 1 Despite the recent advancements in surgical, chemotherapy, and radiation therapy, the overall survival rates at 1 and 5 years are at 26% and 6%, respectively. 1 At the time of initial diagnosis, the tumor is usually locally advanced and infiltrates the main blood vessels, such as the superior mesenteric artery, the portal confluence, and the celiac trunk; thus, significantly increasing the likelihood of margin positive resection.
Radiation therapy as local treatment has been utilized as neoadjuvant, adjuvant, or definitive treatment with or without systemic therapy. When the treatment has a neoadjuvant purpose, the aim is to downstage the disease to radical resection, even if initially inoperable and improve local control. 2,3 However, the presence of radiosensitive surrounding organs at risk (OARs), in particular the duodenum, has limited the delivering of high doses to the target, giving a probability of success of about 25-30%. 4 The introduction of intensity-modulated radiotherapy (IMRT) and, later of, volumetric modulated arc therapy (VMAT) has greatly improved the ability to spare adjacent OARs while delivering a therapeutic dose to the target, with the potential to reduce treatment toxicity and improve local control. 5,6 In particular, since VMAT demonstrated to maintain similar or improved plan quality with respect to fixed-field IMRT but with a significant reduction in treatment delivery time, it was recently proposed as an optimal technique for pancreatic cancer treatment. [7][8][9] Furthermore, modulated techniques allowed the simultaneous delivery of different doses to different target volumes within a single fraction, an approach called simultaneous integrated boost (SIB). This last strategy was found to be more efficient in terms of treatment shortening and radiobiological improved effect as the biological equivalent dose to the tumor increases with higher dose per fraction. For pancreatic cancer, this strategy could deliver a boost dose to the portion of tumor infiltrating the peripancreatic vessels, with the aim of achieving tumor resectability, and a lower dose to the rest of the target volume, avoiding an overdosage to the portion of tumor overlapping the duodenal wall. 10 The dosimetric feasibility of IMRT and VMAT with SIB for pancreatic cancer has been recently successfully demonstrated for conventional dose fractionation showing excellent tumor coverage, conformity of dose distribution, and sparing of OARs. [11][12][13] At the same time, the technological advancements in immobilization and imaging, together with the ability to deliver high conformal doses and to account for organ motion have led to a widespread implementation of stereotactic body radiotherapy (SBRT) in a number of clinical settings. 14 SBRT has garnered a major interest for pancreatic cancer patients since the delivery of ablative doses in a few fractions may improve downstaging and local control and also the shorter treatment time results in an easy integration with chemotherapy. Recent reviews of the literature, [15][16][17] focused on the use of SBRT for pancreatic cancer in unresectable cases, reported that (i) the survival outcomes of patients treated with radiosurgical doses are similar to those recorded on series based on prolonged chemoradiation, (ii) radiation dose escalation can help prevent local tumor progression, (iii) SBRT can be easily integrated into a regimen of aggressive chemotherapy preventing unnecessary delays, and (iv) SBRT has a great potential for conversion to resectability in patients enrolled on a non-curative treatment regimen.
The simultaneous application of SBRT, VMAT and SIB techniques and strategies to pancreatic tumors suggests a new possible clinical paradigm, in which high ablative focused doses are delivered in few fractions to the portion of the tumor near the vascular infiltration and lower doses are simultaneously administrated to the rest of the target volume. This aim is very challenging and obtaining quality plans is a demanding task for medical physicists and dosimetrists.
Recently, various algorithms have been proposed for an automatic optimization of the planning procedure and the search for the optimal patient plan. In particular, fully automated VMAT quality plans for head-neck, 18,19 prostate, 20 and for SBRT treatments of liver 21 and lung 22 metastasis have been successfully generated for clinical application using the Autoplanning template-based optimization engine implemented in Pinnacle 3 treatment planning system (TPS, Philips Healthcare, Fitchburg, WI).
In this study we assessed the feasibility of automated SBRT-VMAT planning using a SIB approach as a dose escalation strategy for stereotactic radiation therapy in pancreatic cancer. Aiming to maximize the dose delivery to the target vascular infiltration while minimizing the probability for duodenum toxicity, we retrospectively re-planned 12 patients evaluating the performance of the Autoplanning module for flattened (FF) and unflattened (FFF) photon beams.

| MATERIAL AN D METHODS
A total of 12 patients with unresectable pancreatic head carcinoma due to vascular infiltration were included in this retrospective planning study.

2.A | Simulation and volumes definition
Patients were simulated supine with arms up using a Vac-Lok and the Elekta (Elekta (TM), Crawley, UK) stereotactic body frame (SBF) for immobilization. An abdominal compressor attached to the SBF by a rigid arc was used with the aim to minimize the mobility of targets close to the diaphragm by mechanically pressing the patient's epigastrium. A study on organ motion due to residual respiratory movements was performed during which the extent of tumor displacement caused by respiration was assessed.
For small bowel visualization, 2 cc of oral Gastrografin diluted in ½ liter of water were given to each patient, 30 minutes before CT scans acquisition, for small bowel visualization purposes.
Target volumes and OARs contouring were performed by a radiation oncologist and a radiologist, using the CT simulation performed in the arterial phase. The site of vascular infiltration was identified CILLA ET AL. and the involved vessel was contoured, with a circumferential margin of 5 mm, for the whole craniocaudal extension of infiltration or contact between the gross tumor volume (GTV) and vessel (or vessels, in case of the involvement of more than one vascular structure). This volume was defined as CTVvasc. The CTVvasc plus an anisotropic margin of 5 mm in the craniocaudal direction and 3 mm in the other directions was defined as BTV (boost target volume).
The tumor PTV (PTV) was defined as the GTV plus an anisotropic margin (5 mm in the craniocaudal direction and 3 mm in the other directions) and including the BTV.
The OARs were delineated as indicated in the RTOG atlas. 23 The duodenum was delineated from the pylorus to the duodeno-jejunal junction. Then, a PRV_duodenum was defined by adding 5 mm in craniocaudal direction and 3 mm in the other directions. The kidneys, liver, stomach, and spinal cord were also outlined from 20 cm above the GTV cranial margin to 20 cm below the GTV caudal margin.
To quantify the relationship of the BTV and duodenum for each patient and its impact on dosimetric outcomes, an overlap volume (OLV) was created as the overlap between BTV and the PRV_duodenum. Figure 1 shows the target, the OARs definition, and the OLV for a representative patient.

2.B | Dose prescription
For each patient three treatment plans were retrospectively calculated for each of conventional 6 MV FF and 6 MV FFF photon energies. At the first dose level, dose prescription was 30 Gy in 5 fractions (6 Gy/fraction) to PTV (including BTV). Then, two simultaneous different boost dose levels were prescribed to BTV: level2, 40 Gy (8 Gy/fraction) and level3, 50 Gy (10 Gy/fraction). In order to facilitate a comparison with treatments performed with a conventional fractionation (2 Gy/fraction), these doses were recalculated in terms of "Equivalent Dose in 2-Gy fractions" (EQD2) using a α/β ratio equal to 10 for the tumor. The EQD2 to PTV and BTV were equal to 40.0 Gy at dose level 1; the equivalent doses to BTV at dose levels 2 and 3 were equal to 60.0 Gy and 83.3 Gy, respectively.
The primary goal for targets coverage was that 95% of BTV and PTV received 95% of their prescription doses. When this request cannot be fulfilled for BTV, a secondary goal demands that the BTV mean dose should be at last the 95% of prescription. Normal tissue constraints were based on the AAPM TG101 recommendations 24 and are summarized as follows: PRV_duodenum and stomach: V 32Gy < 1 cc; Spinal cord: D 0.35cc < 23 Gy; liver: V 21Gy < 700 cc; kidneys: V 17.5Gy < 200 cc.

2.C | Treatment planning
Auto-planning (AP) is a module in Pinnacle 3 Version 16.0 designed to automate the inverse planning optimization process by utilizing a so-called "Technique", ie, a template of parameters that can be customized for each treatment protocol and tumor site. 25 The Technique includes the definition of all beam parameters, dose prescriptions, and planning objectives for PTVs and OARs. The For each patient, six plans were generated, one for each dose level, using the VMAT optimization process for both coplanar 6 F I G . 1. Axial slices of two patients showing the relationship among BTV, PTV, and PRV_duodenum. MV_FF and 6 MV_FFF photon beams from an Elekta VersaHD linac (Elekta Ltd., Crawley, UK). Each plan consists of a dual arc; a gantry rotation is described by a sequence of 90 control points, ie, one every 4°around the patient with no overlap. All plans were calculated with the Pinnacle 3 collapsed cone algorithm and a dose grid resolution of 2 mm.

2.D | Plans evaluation
Dosimetric quality of plans was evaluated by means of dose-volume histograms (DVHs). According to the ICRU report 83, 26 evaluated metrics for BTV and PTV were the minimal dose delivered to the 98% and 95% of the target volume (D98%, D95%), the median dose (D50%), and the maximum dose delivered to the 2% of the target volume (D2%).
Following the suggestion of Van't Riet et al, 27 we evaluated the dose conformity to both target volumes by means of the conformation number (CN) defined as: where TV RI is the target volume covered by the prescription isodose,    Figure 3a.

2.E | Dosimetric verification
In particular, when the OLV exceeded 8%, no advantage for BTV coverage in terms of D95% metric is observed despite the dose escalation. On the contrary, BTV mean dose was >90% of the prescribed dose for all patients, including those with major OLV (Figure 3b).
The effect of the dose difference on the OARs was negligible between FF and FFF plans. As reported in As reported in Table 3, the use of FFF plans resulted in a significantly reduced integral doses compared with FF plans, with P < 0.05 at all dose levels. In particular, FFF plans showed a ID reduction by 3.1% (P = 0.003), 4.9% (P = 0.002), and 4.9% (P = 0.002) for the three increasing dose levels. The Figure 4c shows the integral dose to normal tissues as a function of the BTV dose; the reduction in integral doses of FFF plans with respect to FF plans gradually becomes larger with increased BTV dose. In particular, when the BTV dose was escalated up to 50 Gy, the percentage increase in integral dose was 19.1% for FF plans and 16.8% for FFF plans.
An example of dose distributions for two representative patients with and without OLV and planned at the highest dose level with FFF beams is shown in Figure 5.  Recently, the Pinnacle 3 Autoplanning engine has demonstrated several benefits in other complex anatomical sites as head-neck cancers 18,19 and liver and lung SBRT, 21,22 improving the overall treatment planning quality and efficiency. In particular, in our clinic, we demonstrated that the implementation of Autoplanning for complex cancer cases planned with SIB strategy as head-neck and high-risk prostate cancers translated into a significant increase in dose conformity and reduction in integral dose. 31 A recent paper addressed the feasibility of automated planning for pancreas SBRT using step-andshoot IMRT technique. 32 The authors reported that, also in this com-

| 55
A limitation of this feasibility study is that the effect of residual respiratory motion on dose distribution was not investigated. We are aware that the delivery of SBRT for pancreatic cases is complicated by tumor and normal tissue motion induced by respiration and that techniques for organ motion control are critical for successful implementation of this strategy. A recent study 37 39 This new technology not only offers a superior soft tissue imaging compared to cone beam CT but also allows the opportunity for adaptive replanning when significant interfraction variation is highlighted, 40 potentially increasing the safety and effectiveness of treatment.

| CONCLUSIONS
In the present study, we evaluated the potential of automated planning to deliver ablative integrated boost doses to critical vasculature that limits resectability of pancreatic tumors. We reported that ablative doses up to 50 Gy in 5 fractions can be delivered to the BTV for almost all patients respecting all the normal tissue constraints.
Autoplanning then can represent an effective way to generate complex treatment plans also in a SBRT strategy. Based on the promising aforementioned results, a prospective clinical trial for pancreas SBRT using automated planning with SIB-VMAT technique and FFF beams seems to be justified.

CONFLI CT OF INTEREST
No conflict of interest.