The first reported case of a patient with pancreatic cancer treated with cone beam computed tomography-guided stereotactic adaptive radiotherapy (CT-STAR)

Background Online adaptive stereotactic radiotherapy allows for improved target and organ at risk (OAR) delineation and inter-fraction motion management via daily adaptive planning. The use of adaptive SBRT for the treatment of pancreatic cancer (performed until now using only MRI or CT on rails-guided adaptive radiotherapy), has yielded promising outcomes. Herein we describe the first reported case of cone beam CT-guided stereotactic adaptive radiotherapy (CT-STAR) for the treatment of pancreatic cancer. Case presentation A 61-year-old female with metastatic pancreatic cancer presented for durable palliation of a symptomatic primary pancreatic mass. She was prescribed 35 Gy/5 fractions utilizing CT-STAR. The patient was simulated utilizing an end-exhale CT with intravenous and oral bowel contrast. Both initial as well as daily adapted plans were created adhering to a strict isotoxicity approach in which coverage was sacrificed to meet critical luminal gastrointestinal OAR hard constraints. Kilovoltage cone beam CTs were acquired on each day of treatment and the radiation oncologist edited OAR contours to reflect the patient’s anatomy-of-the-day. The initial and adapted plan were compared using dose volume histogram objectives, and the superior plan was delivered. Use of the initial treatment plan would have resulted in nine critical OAR hard constraint violations. The adapted plans achieved hard constraints in all five fractions for all four critical luminal gastrointestinal structures. Conclusions We report the successful treatment of a patient with pancreatic cancer treated with CT-STAR. Prior to this treatment, the delivery of ablative adaptive radiotherapy for pancreatic cancer was limited to clinics with MR-guided and CT-on-rails adaptive SBRT technology and workflows. CT-STAR is a promising modality with which to deliver stereotactic adaptive radiotherapy for pancreatic cancer. Supplementary Information The online version contains supplementary material available at 10.1186/s13014-022-02125-z.


Background
Pancreatic cancer is a lethal malignancy with a five-year overall survival rate of 2-10% [1][2][3][4]. In recent years, there has been an increased focus on the utilization of stereotactic body radiotherapy (SBRT) for the definitive treatment of pancreatic malignancies [2,5,6]. SBRT for pancreatic cancer is also critical in the palliative setting, Open Access † Minsol Kim and Joshua P. Schiff should be noted as co-first authors as SBRT has been demonstrated to elicit durable local control and long-lasting relief of symptoms of local progression such as abdominal pain and gastric outlet obstruction [7][8][9]. However, the delivery of SBRT for pancreatic tumors is challenging given the close proximity of the mobile and radiosensitive luminal gastrointestinal tract [10]. Magnetic resonance imaging (MRI) guided radiotherapy has been shown to allow precise delineation of daily target and organ at risk (OAR) volumes, improving the efficacy of pancreatic SBRT while minimizing toxicity [11][12][13]. Recently, the implementation of daily online adaptive planning via stereotactic magnetic resonance guided adaptive radiotherapy (SMART) has yielded promising progression-free and overall survival rates as well as a favorable toxicity profile in the ablation of pancreatic cancer [4,[14][15][16].
Recently, a novel ring gantry computed tomography (CT) based radiotherapy machine has been developed with a high-quality cone-beam CT capable of yielding high resolution on-board volumetric images and an artificial intelligence (AI) enhanced treatment planning system (TPS), which is capable of daily adaptive planning (ETHOS, Varian Medical Systems, Palo Alto, CA) [17][18][19]. The use of cone beam CT-guided adaptive radiotherapy for the clinical ablation of pancreatic cancer has not yet been described. Herein we describe the first reported treatment of a patient with pancreatic cancer using cone beam CT-guided stereotactic adaptive radiotherapy (CT-STAR), including a discussion of the workflow and dosimetric analysis of the treatment.

Patient presentation
A 61-year-old woman presented following an episode of abdominal pain due to acute pancreatitis. During the patient's work up, a CT chest/abdomen/pelvis demonstrated a mass in the pancreatic body. Biopsy of the mass confirmed pancreatic adenocarcinoma. The patient met with medical oncology and was recommended neo-adjuvant systemic therapy but declined and pursued alternative therapies. The patient returned to clinic several months later with abdominal pain and interval imaging demonstrating progression of local disease with encasement of the splenic and superior mesenteric veins ( Fig. 1) as well as the development of liver metastases. The primary mass measured 4.8 × 3.8 cm. The patient was referred to radiation oncology for consideration of palliative radiotherapy. On interview, the patient reported left upper quadrant and back pain as well as malaise and weight loss. Physical exam was otherwise unremarkable. The patient was recommended SBRT to her primary mass for durable palliation, 35 Gy in 5 fractions, 7 Gy per fraction. Given the high dose per fraction and adjacent critical organs at risk, the treating radiation oncologist elected to use daily online adaptation with cone beam CT-guidance.

Treatment planning and delivery
The patient was simulated utilizing an end-exhale breath-hold CT with intravenous and oral bowel contrast and a 4-dimensional CT. Intravenous contrast was administered at the 45-s delay phase per institutional protocol. The primary image used for planning was the end-exhale breath-hold CT. The 4D-CT is captured in the case that the patient is non-compliant with breathhold and requires treatment with a different modality and/or dose and fractionation. Of note, as contrast is not delivered with each subsequent daily cone beam CT, the density of the contrast is overrided on the simulation CT to the density of water so that the contrast has no dosimetric impact on the initial plan (P I ). The patient was positioned in a custom immobilization device with left arm down and right arm up, per institutional pancreatic SBRT practice. An MRI was obtained at time of simulation and fused to the simulation images for assistance in target delineation. All treatment planning was performed in the ETHOS (v.02.01.00) TPS. The gross tumor volume (GTV) comprised the gross tumor demonstrated on simulation imaging. As the patient was simulated and intended to be treated at end-exhale breath-hold, a internal GTV or internal target volume was not created. No clinical target volume (CTV) was utilized per standard institutional pancreatic SBRT practice. A 0.5 cm uniform volumetric expansion was applied to form a planning target volume (PTV). The relevant organs-at-risk (OARs) were contoured at the axial slices from 3 cm below to 3 cm above the PTV.
A PTV optimization (PTV opt ) structure was generated, made from the PTV minus any overlap with critical OARs plus a 5 mm margin on the OARs. The critical OARs were the luminal gastrointestinal structures, namely the stomach, duodenum, small bowel, and large bowel. This PTV opt was used to drive prescription dose to the tumor to drive target coverage, given that areas of direct PTV and OAR overlap are not prioritized for target coverage per our standard adaptive radiotherapy practices [4,[20][21][22]. Both the P I and adaptive plans (P A ) were generated using a strict isotoxicity approach, in which maximum OAR constraints are prioritized over target coverage [21,23]. However, a minimum dose of 25 Gy was maintained to the PTV to ensure some uncertainty margin coverage. Dose constraints and objectives are in Table 1. Conservative luminal gastrointestinal OAR constraints were used given the palliative nature of the case. We have provided our standard departmental pancreatic adaptive SBRT dose constraints in Additional file 1: Table S1. A beam arrangement of two ¾ co-planar arcs was used, with 30 and 330 degree collimator angles.
Daily P A were created based on the patient's anatomyof-the-day. The TPS automatically deformed the OAR and target contours from the P A onto the daily cone beam CT using a vendor supplied elastic deformation algorithm, and the TPS AI auto-adjusted the stomach, duodenum, and liver according to the anatomy-of-theday. The deformed GTV was then overwritten and the simulation GTV was ridigly copied onto the patient's anatomy-of-the-day. OARs within a 3-cm contour ring Table 1 OAR constraint and target volume metrics are presented for the initial non adaptive (P I ) and adapted (P A ) plans (per standard adaptive protocol [22]) were adjusted by the radiation oncologist in order to confirm accuracy. The initial simulation based treatment plan (P I ) was projected on the patient anatomy-of-the-day at the same time that the re-optimized daily adapted plan (P A ) was generated. The P I and P A were compared using dose volume histogram (DVH) objectives, and the superior plan that met all dosimetric goals was delivered. Of note, all acquired kV cone beam CTs were considered of sufficient quality for target and OAR delineation as well as daily adaptation per the treating radiation oncologist and medical physicist.

Dosimetric and clinical results
Constraint and coverage metrics for the P I and P A are demonstrated in Table 1 (Fig. 2). The P A achieved hard constraints in all five fractions for all four critical luminal gastrointestinal structures. Figure 4 illustrates how the use of daily adaptive planning allowed for a specific radiotherapy fraction to achieve the small bowel hard constraint, where as delivery of the P I would have violated that constraint. Treatment component times were recorded and are demonstrated in Table 2. Mean (standard deviation) total treatment time was 70 min (68.3-81.7) and treatment time decreased each consecutive fraction. The patient completed all five fractions of CT-STAR without issue. The patient ultimately progressed locally and distantly, and passed away several months after treatment.

Discussion
Herein we describe the first reported use of CT-STAR for the treatment of a patient with pancreatic cancer using a novel ring gantry device. These data demonstrate that the delivery of the P I would have led to nine critical OAR hard constraint violations across all five fractions, and that the daily P A met all critical OAR hard constraints in all five fractions. Furthermore, the use of daily adaptation improved PTV opt , GTV V100, and GTV D95 coverage (Table 1) while alleviating the hard constraint violations. With regards to workflow, the overall treatment times were within the range of previously described treatment times for daily adaptation, and the decreased time per each consecutive fraction suggests that treatment times decrease with increased patient/staff familiarity [22,24].
The utility of adaptive stereotactic radiotherapy for the treatment of pancreatic cancer can not be understated. The effective ablation of pancreatic cancers requires the delivery of biologic effective dose of at least 100 Gy [10]. However, this is difficult to achieve as the pancreas is adjacent to several mobile and radiosensitive OARs. Initial studies evaluating the use of ablative doses of standard CT-guided stereotactic radiotherapy for the treatment of pancreatic cancer proved efficacious with regards to local control, but also displayed high rates of luminal gastrointestinal organ toxicity [25][26][27]. Adaptive radiotherapy can improve the therapeutic index of SBRT for pancreatic cancer. Recently, our institution published outcomes for patients with inoperable pancreatic cancer treated with SMART and demonstrated durable progression-free and overall survival rates as well as a favorable toxicity profile [4]. While these data are promising, it's notable that their application is limited to MR-guided workflows. Prior to the advent of the ETHOS platform, adaptive SBRT for pancreatic cancer was limited to clinics with MR-guided or CT-on-rails workflows [4,16,28,29].
While CT-STAR has the capacity to expand access to adaptive pancreatic SBRT, there are potential limitations of using a CBCT-guided platform instead of a MR-guided platform. The improved soft tissue contrast of MRI can be useful in pancreatic and abdominal contouring, which can be of particular importance when delineating gross organ invasion. In contrast with MR-guidance, gross organ invasion is challenging to delineate on cone beam CT. In our experience [4], approximately 10% of patients with locally advanced pancreatic cancer present with evidence of gross organ, and the use of CBCT-guided adaptive SBRT may be limited in that subset of patients. This may be of consideration when planning to install either a MR-or cone beam CT-guided adaptive platform.
Herein we demonstrate that stereotactic adaptive radiotherapy is able to be delivered on a cone beam CTguided modality, which promises to increase access to adaptive pancreatic SBRT world wide. This case presentation demonstrates the potential for CT-STAR to