Pancreatic cancer is a highly aggressive malignancy. Chemoradiation therapy (CRT) is used in many cases to improve local-regional control; however, toxicities associated with radiation can be significant given the location of the pancreas. The purpose of this study is to quantify the dosimetric changes seen when using photons or protons in patients receiving CRT for cancer of the pancreas.
Ten patients with pancreatic head adenocarcinoma treated between 2010 and 2013 were evaluated in this study. All patients underwent simulation with contrast-enhanced computed tomography imaging. Separate treatment plans using proton radiation therapy, intensity-modulated radiation therapy, and 3-dimensional photon radiation therapy modalities were created for each patient. Dose-volume histograms were calculated and analyzed to compare plans between the 3 modalities. The organs at risk evaluated in this study are the kidneys, liver, small bowel, and spinal cord. To determine statistical significance, analysis of variance and 2-tailed paired t tests were performed for all data parameters.
The proton radiation therapy plans resulted in significantly lower doses delivered to the kidneys, liver, small bowel, and spinal cord. This finding was seen consistently across the parameters analyzed in this study.
For patients receiving CRT, the proton plans are technically feasible and dosimetrically appealing with superior organs at risk sparing. Proton radiation therapy may improve the therapeutic ratio for patients receiving CRT for pancreatic cancer.
Pancreatic cancer accounts for more than 30 000 deaths annually in the United States . Surgical resection offers the best chance of long-term survival. Unfortunately, both local and systemic recurrences are common after a pancreaticoduodenectomy. Chemoradiation therapy is an integral part of pancreatic cancer management; however, toxicities associated with radiation may be significant given the location of the pancreas.
Conventional radiation therapy, such as 3-dimensional (3D) photon therapy, uses x-ray beams. These enter and exit the body, creating both entrance and exit dose. As a result, nontargeted organs and tissues surrounding the pancreatic target are also exposed to radiation. Improvements in radiation delivery techniques have resulted in methods to improve beam conformity around treatment targets. Intensity-modulated radiation therapy (IMRT) is one such method. It uses multiple beam angles at varying intensities to escalate dose at the target while sparing surrounding normal tissue from high-dose regions. One report  has suggested that IMRT can reduce high-grade gastrointestinal toxicity in the setting of pancreatic cancer. Proton radiation therapy (3D proton therapy), another form of radiation treatment, uses charged particle beams. A proton beam deposits most of its energy at a discrete depth within tissue, called the Bragg peak. This peak is predictable and can be created to match the exact depth and thickness of the tumor target. The entirety of the beam's energy is deposited at the target, thus avoiding any exit dose. Some previous dosimetric studies  have shown a potential role for proton radiation therapy in the postoperative setting. In patients with unresectable disease, radiation is sometimes given as part of definitive treatment. The purpose of this study was to quantify the dosimetric changes seen in using protons or photons with consistent planning parameters for patients receiving chemoradiation therapy for cancer of the head of the pancreas.
Patients and Methods
We retrospectively evaluated 10 patients with pancreatic head adenocarcinoma. The clinical staging (per the American Joint Committee on Cancer, 7th edition ) ranged from T1–4N0–1M0. (Table 1). All patients were treated between 2010 and 2013.
Simulation and Treatment Planning
Patients underwent simulation in the supine position by using intravenous and oral contrast-enhanced computed tomography imaging (GE Lightspeed VCT scanner; GE Healthcare, Little Chalfont, United Kingdom) with 2.5-mm slice thickness. All patients were scanned from above the diaphragm to 3 to 4 cm below the iliac crest. Separate treatment plans using proton radiation therapy, IMRT, and 3D photon radiation therapy modalities were created for each patient. Our IMRT plans each used 6-MV coplanar nonparallel opposed beams. Target volumes and critical structures were designated, and ring structures were drawn by the planners to optimize dose coverage and avoid hot spots. All treatment plans were created with the Odyssey 4.8 planning system (Optivus Proton Therapy, Inc, San Bernardino, California), which uses pencil beam algorithm. The gross tumor volume was defined as the pancreatic head lesion, or as the preoperative tumor bed for patients undergoing resection. The clinical target volume (CTV) was defined as gross tumor volume plus a 1-cm manual expansion in all directions. The planning target volume (PTV) was generated by expanding the CTV by 10 to 15 mm. Setup uncertainty from respiratory motion and diaphragm movement was accounted for with close attention to target expansions. The planning target volume (PTV) was generated by expanding the CTV by 1.5 cm and 1.2 cm for the 3DCRT and IMRT plans, respectively. All 3DCRT plans are given a 1 cm margin beyond the PTV to the block edge to account for beam penumbra. The lateral penumbra and distal margin of proton plans generated by the treatment planning system were between 1 to 1.5 cm and based on the beam energy selected.
A dose of 50.4 Gy given in 28 fractions was delivered to the PTV. All plans were optimized to allow 95% isodose coverage of at least 95% of the PTV. The 3D proton plan beam arrangements consisted of 2 to 4 beams at oblique angles. The median proton beam energy was 250 MeV, with some minor deviations depending on the distal depth of the target. Our institution uses a passive scattering beam system, which requires a patient portal-specific collimating aperture to shape the dose to the target field laterally. A range compensator is used to conform the dose to the distal aspect of the target volume. A spread-out Bragg peak to cover the target in the beam direction is achieved with a modulator wheel. All IMRT plans consisted of 6 to 9 coplanar, nonparallel opposed 6-MV photon beams delivered with a multileaf collimator using a step-and-shoot technique. Each 3D photon plan consisted of a 4-field box (anterior-posterior (AP), posterior-anterior (PA), right lateral (RL), left lateral (LL)) using 15 to 24 MV photon beams delivered with either a multileaf collimator or custom-cut block.
Plan Evaluation and Analysis
To compare the plans with the different modalities and beam arrangements, dose-volume histograms (DVHs) were calculated and analyzed. The organs at risk (OARs) being evaluated in this study are the kidneys, liver, small bowel, and spinal cord. These were contoured and designated as normal tissue. The small bowel was contoured per Radiation Therapy Oncology Group (RTOG) 0848 and extended from the jejunum to 2 cm below the lower extent of the CTV. Analysis was performed for the volume of kidney receiving 15 Gy (V15), 20 Gy (V20), and mean kidney doses were collected for the left, right, and bilateral kidneys. The small-bowel V15 and V50, the dose delivered to one-third of the liver (D1/3), mean liver dose, and the maximum spinal cord dose were also analyzed.
Conformity indices were also obtained and analyzed for plans between the 3 treatment modalities. The homogeneity index was defined as the difference between the maximum and minimum dose to the target volume (D1% and D99%, respectively) divided by the prescription dose [5, 6]. Uniformity index was also used; it was defined as the ratio of D5% to D95% [7, 8]. Both homogeneity index and uniformity index were used to assess overall plan uniformity per previously established methodology . The conformity index was defined, per RTOG guidelines, as the volume of the 95% isodose curve divided by the PTV volume. To determine statistical significance, analysis of variance and 2-tailed paired t tests were performed with P values < .05 considered to be statistically significant.
A total of 10 patient scans were used for this study. Three treatment plans were created on each scan: 3D proton, IMRT, and 3D photon. Tables 2 and 3 present dose-volume parameters obtained from these plans. Dose distributions for liver, kidney, and small bowel from 2 of our study patients are presented in Figure 1. The CTV was encompassed by the 95% isodose line in all cases. At least 95% of the PTV was encompassed by the 95% isodose line. Separate plans were generated and optimized for all 10 study patients. The gross tumor volume and CTV were held constant in each patient for each of the 3 plans.
Three-Dimensional Proton versus Intensity-Modulated Radiation Therapy
First, we compared dose-volume parameters of the IMRT plans with those of 3D proton plans. The 3D proton plans resulted in a lower mean total kidney dose (3.17 Gy versus 9.59 Gy, P = .039). The mean liver dose was reduced (1.83 Gy versus 5.24 Gy, P = .021). The maximum dose to the spinal cord was also reduced (3.64 Gy versus 26.16 Gy, P = .000). Intensity-modulated radiation therapy provided better homogeneity (0.40 versus 0.14), uniformity (1.33 versus 1.09), and conformity (0.85 versus 1.13) relative to the 3D proton plans. This was unsurprising given the inverse nature of the IMRT treatment planning.
Three-Dimensional Protons versus Three-Dimensional Photons
The next comparison looked at dose-volume parameters between 3D proton and 3D photon plans. The proton plans resulted in a lower mean liver dose (1.83 Gy versus 4.97 Gy, P = .021) and liver D1/3. The V15 and V50 bowel were also reduced: 45.87 versus 63.93, P = .035; and 5.81 versus 16.77, P = .018, respectively. The maximum dose to the spinal cord was reduced (3.64 Gy versus 32.23 Gy, P = .000). The proton plans also demonstrated better homogeneity (0.40 versus 0.14) and better conformity (0.85 versus 1.30). There was no difference in uniformity between the 3D proton and 3D photon plans (1.33 versus 1.20).
Intensity-Modulated Radiation Therapy versus Three-Dimensional Photons
Finally, we compared dose-volume parameters between IMRT plans with 3D photon plans. The 3D photon plans resulted in a lower maximum spinal cord dose (26.16 Gy versus 32.23 Gy, P = .049). No significant difference was seen in the other OARs. The IMRT plans resulted in better conformity (1.13 versus 1.30) but there was no difference in homogeneity (0.14 versus 0.14) or uniformity (1.09 versus 1.20).
The results of our study demonstrate a significant tissue-sparing benefit of 3D proton plans over the IMRT and 3D photon plans. Target coverage was adequate in each of these treatment planning modalities but the amount of normal tissue irradiated differed among them. Clinically acceptable plans were generated for all 10 patients with each of the 3 treatment modalities. Target coverage was achieved in each plan despite a significant patient-to-patient variation in target volume size and shape. The planning target volumes ranged from 196 to 1169 mL (median, 347 mL), suggesting that each of these 3 modalities could provide clinically acceptable treatment plans at least in terms of target coverage. Of the 10 patients in this study, 8 had locally advanced disease (Table 1), which explains some of the larger target volumes treated.
Pancreatic tumors arise most frequently in the head of the pancreas, which is typically located to the right of midline. It is not uncommon for one kidney to receive more radiation dose than the other. This is also reflected in our data, which show the right kidney receiving a higher mean dose than the left kidney. The intrinsic properties of proton beams should reduce both total and unilateral kidney dose. Protons have an inherent dosimetric advantage over photons owing to their absence of exit dose. This results in a reduction in the volume of kidney receiving low doses of radiation when compared with IMRT and 3D photon plans (Figure 1C). There is a wide separation of the curves in the low-dose region of the DVH, indicating a large difference in the volume of tissue being irradiated by low doses of radiation. The difference begins to dissipate at volumes receiving closer to the prescribed dose. Some studies  have reported that 3D photon plans may cause small increases in kidney volumes exposed to higher doses of radiation. On the other hand, it is unsurprising that there was a lack of difference in mean total kidney dose between the IMRT and 3D photon plans. It would be difficult to avoid the nearest kidney with photon beams given its proximity to the pancreatic target.
Proton beams have a Bragg peak that allow for a lower entrance dose. The Bragg peak, in addition to the characteristic absence of exit dose, results in a lower integral dose delivered to the liver. Our results demonstrate a significantly lower mean liver dose in the 3D proton plans, which is consistent with a lower integral liver dose. Even the liver partial-volume doses were significantly lower. Figure 1B shows distinct separation of the curves in the low-dose region of the DVH. In comparison to the kidney DVH, there is a less striking separation of the curves in the liver DVH.
Radiation toxicity to the bowel is another significant matter that has been evaluated in numerous pancreatic irradiation studies. Previous study findings include a significant dose-response relationship for every increment of 5 Gy above ≥15 Gy, as well as a significant volume effect with V15 volumes greater than 150 cm3 [11–13]. The dose-volume and fraction-size dependence of bowel toxicity is challenging to interpret given the wide range of toxicities reported in the literature. Modern series reviewed by QUANTEC (Quantitative Analyses of Normal Tissue Effects in the Clinic) generally confirm the established dose tolerances commonly used in the clinical setting , and institutions commonly impose a V50 < 5% to limit late toxicities such as obstruction and perforation.
In our study, the volume of irradiated bowel was lower in the proton plans at both low (V15) and high (V50) dose regions. Proton beam characteristics, along with the flexibility to select a number of beam arrangements, resulted in a lower integral bowel dose. As seen in Figure 1A, it is evident that there is a large volume of bowel receiving a low dose in all 3 plans. The greatest difference in volume irradiated occurred at the lower-dose region of the DVH. However, given that the 3D proton plans also have a significantly lower V50 volume than the 3D photon (5.81% versus 16.77%, P = .018) plans, one clinical implication may be that 3D proton therapy offers a better modality for dose escalation. The bowel is a large organ and is difficult to avoid when treating the postoperative pancreatic target (Figure 2).
The IMRT and 3D photon plans generate relatively large low-dose regions in comparison to 3D proton plans. In our study, there was no significant difference between the volumes of low-dose regions treated by the IMRT and 3D photon plans. This was indicated by the lack of a significant difference between the MLD, V15 bowel, and V15 kidney parameters. In contrast, the 3D proton plans resulted in significantly smaller low-dose regions in comparison to IMRT. Again, the lack of exit dose seen in 3D proton plans contributes to a much smaller integral dose. The optimal radiation therapy technique for pancreatic cancer, however, is still unique for each patient.
The aim of our study was to quantify the dosimetric differences seen when using proton or photon radiation therapy. There was consistent overlap between the PTV and OARs so that no one technique could simultaneously achieve full target coverage while fully respecting OAR constraints. Furthermore, the size and extent of the target volume may preclude the use of certain modalities. Nonetheless, significant conclusions may still be drawn from these generated plans, in which full target coverage was obtained with reasonable uniformity and conformity. Another limitation of our study was the lack of 4-dimensional computed tomography simulation. The abdomen is a very mobile part of the body, which creates many challenges with normal breathing motion. Target volume motion during respiration may significantly affect beam selection during the planning process. Current studies at our institution use SpyroDyn'RX (Dyn'R, Aix-en-Provence, France) voluntary breath-hold technique as a means of active breathing control for more accurate target delineation and treatment delivery. There is speculation that the dosimetry benefits of proton beams may be lost when used to treat small target regions. Our findings, however, suggest that there may still be a significant benefit in treating small fields with proton radiation therapy.
In conclusion, 3D proton plans consistently delivered lower mean total kidney doses, mean liver doses, and liver D1/3 than the IMRT plans. The 3D proton plans also gave less mean liver dose, liver D1/3, bowel V15, and bowel V50 than the 3D photon plans. There was no difference between the IMRT and 3D photon plans in dose delivered to the kidneys, liver, or bowel.
ADDITIONAL INFORMATION AND DECLARATIONS
Conflicts of Interest: The authors have no conflicts of interest to disclose.
Acknowledgments: The authors would like to thank William Preston, EdD, for his assistance with manuscript preparation.