Optimization of carbon ion and proton treatment plans using the raster-scanning technique for patients with unresectable pancreatic cancer

Pancreatic cancer patients are still having a dismal prognosis. About 277,000 people worldwide die each year due to this disease [1]. To date resection is considered to be the only curative treatment. In case of locally advanced unresectable pancreatic cancer (LAPC), neoadjuvant treatment approaches including combined chemoradiation with gemcitabine have proven efficacy towards tumor downsizing and lead to a secondary resectability in approximately 30 %, even in case of local relapse after primary surgery [2‐6]. Modern radiotherapy techniques including IMRT (Intensity-Modulated Radiotherapy) and IGRT (Image-Guided Radiotherapy) offer more conformal dose distributions and thus dose to organs at risk (OAR) can be reduced significantly. Recently it was shown that IMRT treatment in pancreatic cancer patients can provide comparable or less gastrointestinal toxicity than by conventional radiotherapy and therefore contributes to the clinical benefit of the treated patients, especially when used as dose-escalation treatment in combination with IGRT [7, 8]. Charged particle therapy represents an emerging technological advance in oncology. Particle therapy is characterized by an inverted depth-dose-curve, which leads to a low dose deposition within the entry channel and a well-defined high local dose deposition in the Bragg Peak region [9]. The so called Spread Out Bragg Peak (SOBP) allows to irradiate precisely, sparing surrounding normal tissue – both carbon ion and proton radiotherapy have highly conformal dose distributions with high dose deposit in the target volume and an increased sparing of the OARs [9]. High-LET (linear energy transfer) carbon ion beams are characterized by high dose deposition in their trajectory. This results in a high amount of clustered double-strand breaks in the cells‘DNA (Deoxyribonucleic acid). So, in contrast to photon and proton beams, carbon ion beams cause an enhanced reduction in clonogenic survival of pancreatic and also of hepatic cell lines [10‐13]. Carbon ions in particular offer a higher biological effectiveness due to enhanced and prolonged DNA damaging and induction of bulky lesions, which can be translated into higher RBE (Relative Biological Effectiveness) values [11, 13, 14]. Particle beams are notably appropriate in hepatobiliary and pancreatic malignancies, where radiosensitive normal tissues (e.g. liver, kidneys) are surrounding the target volume [15‐17].

There are also encouraging clinical results from Japanese particle therapy facilities, that have conducted small clinical trials and gained experience with carbon ion treatment, using different treatment protocols over the last few years [16, 18]. Nevertheless, particle therapy of abdominal organs is very complex. Dose application has to be analyzed carefully. That’s the reason why the purpose of this study is to evaluate different plan optimization strategies as a preparation for the clinical practice using active raster scanning technology [19]. Different proton and carbon ion field configurations are analyzed with regard to dose distribution and individual anatomy, using a customized rating scheme.

Our study focused on treatment planning strategies in patients with unresectable pancreatic cancer. 50 treatment plans for proton and carbon ion beams were evaluated. We developed a score taking into account target volume coverage and OAR parameters to provide a basis for decision making during the planning process for this patient group. The score demonstrates, that for target volume coverage, all beam angles and combinations provide nearly equivalent scores with a slight advantage for multiple field plans.

In summary, a three-field setup achieved consistently high values throughout the cases. A one-field arrangement with a single posterior field showed in some cases comparable results and overall very little score dispersion. Nevertheless, the maximum doses in the myelon were thoroughly high. But this field configuration seems to be of advantage in cases of topographic proximity of radiosensitive OARs to each other and to the target volume. The field-setup with two posterior oblique fields through the kidneys showed satisfying results, was able to spare the myelon, and score dispersion was quite small as well.

Particle therapy of abdominal organs is very complex, as inter- and intraindividual changes in organ motion and bowel gas movement may have a serious impact on ion beam dosimetry [22, 23]. Kumagai and colleagues reported a treatment plan analysis of passive scattered carbon ion beams at their facility and found out, that both anterior-posterior and left-right field constellations caused the highest dose affections (mainly because of gastrointestinal gas bubbles) [22]. The three-field configurations include such fields, which is why they have to be evaluated critically. Our results also point to more robustness of a single posterior field compared to three-field arrangements. In this setting beam paths are practically not affected by gas fillings in stomach (left-right) and intestine (anterior-posterior). On top of that, inter- and intrafractional variations due to breathing may cause relevant changes in the beam path leading to adverse effects. Taniguchi and colleagues further analyzed doses in duodenum and stomach in patients with unresectable pancreatic carcinoma treated with a five-fraction protocol. Results show a decreasing dose to the OARs during expiration compared to inspiration [24]. The interaction of beam and organ motion can cause interplay effects and potentially lead to (unexpected) dose variations in target structures and thus unwanted normal tissue exposure [25]. However, intra- and inter-fractional changes are described but not totally understood, which is why we need re-planning scenarios, especially in scanned ion beam treatment, where slight changes may result in significant dose variations [22‐24, 26]. In general, we are referring to Japanese experiences in ion beam therapy of LAPC. But one has to be clear about the fact, that irradiation was performed by the use of scattered ion beam therapy. Our department made investigations in robustness in scanned ion beam therapy, which we can rely on. Batista et al. has presented data about pancreatic cancer, that supported our results. A single posterior field and two oblique posterior fields show good results in case of robustness [27]. On top of that Richter et al. has made investigations in the topic of robustness of ion beam irradiation of liver tumors. The group was able to show, that fractionation is a potential tool to reduce dose inhomogenity by interplay effects [28, 29]. That is one of the reasons why our PHOENIX trial has dose escalation steps starting with 12 fractions [21].

In out treatment plan computations doses to OARs were thoroughly uncritical. But distinct field setups lead to different risk profiles. Whereas posterior fields deposit higher doses to the myelon and partial volumes of the kidneys, right lateral fields are surely affecting partial liver volumes. Higher doses in intestinal structures were found when left lateral and anterior fields were used (three-field configurations). Even though only small sub-volumes received remarkable doses, this could theoretically result in clinically relevant complications. The experience on comparable dose protocols with intestinal structures in close proximity to the target volume is based on carbon ion therapy of sacral chordoma and locally recurrent rectal cancer, where no higher gastrointestinal toxicities were recorded [30, 31].

The single posterior field setup is of major concern, because of a steep RBE-increase at the distal end of the SOBP, which could result in unexpected high doses to the small intestine ventrally of the pancreatic tumor. Calculated doses to intestines were in the range of tolerance. With regard to the DVH, the single posterior field can spare most of the OARs. High doses in the myelon were calculated, but did not exceed general QUANTEC (Quantitative Analyses of Normal Tissue Effects in the Clinic) dose constraints [32].

Nevertheless, tolerance doses of the OARs are not well defined for carbon ion irradiation and are extrapolated from photon-based data as long as no new dose-volume thresholds are defined. An important limitation of the LEM-based TPS is also the underlying biological dataset. According to our current clinical practice we do not differentiate the radiobiological characteristics of normal tissue and tumor during treatment plan optimization. Experience with carbon ion beams based on newer biological optimized LEM-algorithms within the TPS are only theoretical and not debated in this work [33].

Topographical influence and advantage of the single-field configuration is mainly expressed by carbon ion beams. It is likely that the proton beams are at a disadvantage due to their wider lateral dose gradient [9]. Due to this beam broadening, especially with deeper located targets, smearing of the dose distribution possibly results in plans inferior to advanced photons. Basically, a direct comparison between carbon ion and proton beams could be drawn but this may not be useful, as we are currently still working with an earlier version of LEM [34]. Therefore the dose in the proximal part of the beam channel is rather overestimated and the dose in the target itself is underestimated as recently demonstrated by Grun and co-workers with an improved version of the local effect model (LEM IV) [35]. Carbon ion radiation would result in more beneficial dose distributions both in the target and OARs than shown in the presented data.

Nonetheless, our study was able to present possible field setups and evaluated them by the use of a customized score. With reference to our results, particle therapy of pancreatic cancer is possible by the use of all 5 field setups. Distinct topographical conditions should be taken into account.

Future work of our group will focus on even more differentiated radiobiological-based treatment planning, which includes different α/β ratios for normal and target tissues.

In summary, field configurations with three fields showed the best dose distributions – for both carbon ion beams and proton beams. Nevertheless, three-field configurations are highly influenced by gastrointestinal variations. A single posterior field deposits high doses in the myelon, but seems to be the most robust one and showed good results for difficult and varying topographical conditions. A setup with two oblique posterior fields generated reproducible results and can be set as a reasonable compromise between three fields and one field configurations.