Study design and patients
This was an observational retrospective cohort study performed in a single institution. Patients with unresectable PHC who underwent PBT between January 2002 and August 2020 were eligible for the study. Patients were excluded if they had nodal metastasis at presentation or if they underwent surgery or other radiation therapies before PBT. Before the initiation of PBT, all patients were diagnosed based on biopsy or “radiologic studies and detection of elevated tumor markers (carbohydrate antigen 19-9 [CA19-9] and/or carcinoembryonic antigen [CEA])”. The location of tumor origin was determined based on radiologic and endoscopic studies. Staging was according to the TNM Classification of Malignant Tumors, 8th edition (Union for International Cancer Control) and Bismuth–Corlette classification. PHC was deemed unresectable by individual surgeons due to extensive hilar invasion, and/or bilateral liver involvement, and/or vascular encasement, and/or greater age, and/or severe pre-existing disease, such as chronic obstructive pulmonary disease. Acute adverse events (AAEs) and late adverse events (LAEs) were evaluated using the Common Terminology Criteria for Adverse Events (version 5.0). Written informed consent was obtained from all patients. The study protocol conformed to the ethical guidelines of the 1975 Declaration of Helsinki as reflected in a priori approval by the local institutional review board, identification number 2-16. We collected data pertaining to age, sex, performance status (PS) on the Eastern Cooperative Oncology Group scale, Child–Pugh classification, stent treatment for biliary stenosis before PBT, surgical spacer placement before PBT, chemotherapy before PBT, underlying liver disease, involvement of the portal vein, concurrent chemotherapy, chemotherapy before PBT, CA19-9, and CEA as risk factors.
Surgical spacer placement for PBT
When the tumor was widely adjacent to the gastrointestinal tract, it was difficult to administer a curative-intent PBT. In such cases, a spacer was placed between the inferior surface of the liver and gastrointestinal tract. Greater omentum and/or Gore-Tex sheets (W.L. Gore and Associates, Newark, DE, USA) were superimposed and applied as spacers .
Each patient was immobilized using a custom-made thermoplastic cast in the supine or prone position, and images were acquired by computed tomography (CT) at slice thicknesses of 2 mm and magnetic resonance imaging (MRI) at a slice thicknesses of 2-5 mm. All three-dimensional treatment planning was performed using the XiO-M treatment planning system (Mitsubishi Electric, Tokyo, Japan). The gross tumor volume (GTV) included the primary tumor; the clinical target volume (CTV) was defined as the GTV plus a 10 mm basic margin, and adjacent structures were not included. The CTV margin, which was defined by the radiation oncologist according to the degree of microscopic extension of the tumor, was poorly defined in PHC. It was reported that microscopic extension of PHC was usually <10 mm [16, 17]. The CTV margin was defined as 10 mm in our study. The planning target volume (PTV) was defined as the CTV plus a setup margin (5 mm) and a respiratory gating margin (1-3 mm), which was measured on CT images between inspiratory and expiratory phases. The stomach; small bowel, including the duodenum; liver; kidneys; and spinal cord were defined as organs at risk (OAR). The target volumes and OAR were delineated on the CT-MRI fusion images.
Dose was expressed as Gy (relative biological effectiveness [RBE]). The RBE of the proton beam had been determined to be 1.1. The dose prescription for curative intent was selected as follows: 72.6 Gy (RBE) in 22 fractions, 70.2 Gy (RBE) in 26 fractions, and 67.5 Gy (RBE) in 25 fractions were selected to the tumor adjacent to the main trunk of the portal vein; 76 Gy (RBE) in 20 fractions and 66.0 Gy (RBE) in 10 fractions were selected to the tumors adjacent only to the first branch of the portal vein; and 60.0 Gy (RBE) in 30 fractions was selected to reduce the risk of gastrointestinal adverse events because total gastrectomy was performed before the initiation of PBT. The prescribed dose was decided based on the earlier experience of soft tissue sarcoma treatment in our institution, because there were few reliable reports on PHC.
To compare different dose fractionations, we used the linear quadratic model; we assumed α/β of 10 Gy (RBE) for the tumor response, which was expressed as biologically effective dose 10 (BED10), and a/b of 3 Gy (RBE) for the late effect, which was expressed as biologically effective dose 3 (BED3) . The maximum dose restrictions for the stomach, small bowel, and spinal cord were 50 Gy (RBE), 50 Gy (RBE), and 45 Gy (RBE), respectively, at a standard fraction of 2 Gy (RBE) per fraction with a 3 Gy (RBE) α/β value [19, 20]. Additionally, irradiated volumes of the stomach, small bowel, liver, and kidney were made to be as low as possible. Typical dose distribution is shown in Figure 1.
Detailed dose volume histogram (DVH) data were retrospectively obtained using MIM Maestro version 6.9.6. (MIM Software Inc., Cleveland, OH, USA). The following dosimetric variables for BED10 of the GTV, CTV, and PTV were generated from the DVH, respectively: the maximal absolute dose (Dmax) and that covering 2 ml (D2 ml) of the GTV, CTV, or PTV; maximal absolute dose covering 50% (D50) of the GTV, CTV, or PTV; mean dose (Dmean); and minimum absolute dose (Dmin) and that covering 95% (D95) or 98% (D98) of the GTV, CTV, or PTV. To evaluate the dose of GTV, CTV, and PTV, we adopted the value of 79.2 Gy (RBE) [BED10] and 84.0 Gy (RBE) [BED10]) as indexes: the former was equal to the value of BED10 in 66 Gy (RBE) in 33 fractions, and the latter was equal to the value of BED10 in 70 Gy (RBE) in 35 fractions. These dose fractionations are commonly used for definitive radiotherapy in solid cancers.
The follow-up of each patient started on the date on which PBT was initiated. The Kaplan–Meier method was used to estimate overall survival (OS), MST, and progression-free survival (PFS) of the eligible subjects. We defined PFS as evidence of neither tumor regrowth nor other recurrence.
Univariable differences in Kaplan–Meier curves for OS were evaluated using the log-rank test. Cox regression was also used to assess the effects of the potential risk factors for OS in multivariate analysis, including those covariates that appeared to be significantly associated with OS (p<0.05) in univariate analysis. Statistical analyses were performed using R software version 3.6.2 (R Foundation for Statistical Computing, Vienna, Austria).