What is the appropriate size criterion for proton radiotherapy for hepatocellular carcinoma? A dosimetric comparison of spot-scanning proton therapy versus intensity-modulated radiation therapy

We obtained approval from our institutional review board at Hokkaido University Hospital for this retrospective dosimetric study. Between January and October 2011, 10 consecutive patients with HCC treated at our institute, with GTV greater than 6 cm in diameter and portal vein tumor thrombosis (PVTT), were included in this study. The tumor size and that location data are summarized in Table 1. CT data sets were acquired using a slice thickness of 2 or 2.5 mm. All the patients received X-ray radiotherapy targeting the PVTT alone, not the whole tumor, as a palliative treatment.

In order to compare large volume irradiation, a GTV encompassing the whole tumor was re-contoured by a radiation oncologist using one of two treatment planning systems: (TPS) XiO (CMS Inc., St Louis, MO), Pinnacle³ (Philips, Inc., Madison, WI) or Eclipse TPS Ver.10.0.0 (Varian Medical Systems, Palo Alto, CA). The clinical target volume (CTV) was defined as a 5 mm expansion of the GTV minus areas of overlap with uninvolved extra hepatic structures (e.g., lung, abdominal wall, intestine) [8]. A 1 cm margin for set-up margin and organ motion was uniformly added to the CTV to generate the planning target volume (PTV), assuming the use of gated irradiation. Target and normal structures were formatted and transferred to the Eclipse TPS Ver.10.0.0. Here, for patients with extremely large targets (PTV >1500 cm³), we prepared two different plans: one with the GTV contoured PVTT alone (plan ID 1, 2 and 3 in Table 1) and the other with the GTV contoured whole tumor area (plan ID 11, 12 and 13 in Table 1). This is because we were concerned that patients with large tumors should receive RT targeting the PVTT alone, even though SSPT can spare the surrounding normal liver. A total of 13 GTVs in 10 patients were used for IMRT plans and SSPT plans (total of 26 plans were generated), in this comparative study. We made a comparative dosimetric study between simulated plans of photon and proton beams on various irradiation volumes of HCC utilizing IMRT and SSPT plan techniques. These simulated plans were generated in Eclipse TPS, assuming photon treatment with an MHCL-15SP v80m (Mitsubishi Electronics Co., Ltd., Tokyo) LINAC and proton treatment with a Probeat III (Hitachi Co Ltd, Japan) proton accelerator [9], respectively.

For the IMRT plans, five to nine evenly spaced intensity modulated fields were generated with a 6 MV photon beam. For Proton plans, the SSPT technique [10], which can be generated using the inverse planning approach like IMRT, was applied. For SSPT plans, simple arrangements of two or three proton fields were selected. The dose distributions of proton plans generated with only two of three fields were dependent on field incidence, achieving good results for the normal liver spared by beam entrance. First, one beam angle was selected so that beam paths can take the shortest way to cover the target. Then one or two more beams were added and each beam’s angle was adjusted so that dose distribution was homogenous. The SSPT plans were simulated as individually weighted proton Bragg peaks distributed throughout the PTV. Energies were selected from those actually deliverable. The energies required to cover the target homogeneously were 70 to 180 MeV. For SSPT plans, a relative biologic effectiveness factor for protons of 1.1 was employed.

The dose prescription for this study was 60 Gy for IMRT plans or 60 cobalt Gy equivalents (CGE) for SSPT plans covering 95% of the PTV, delivered in 15 fractions [11, 12]. The dose volume constraints for the liver used in the planning process were taken on the basis of normal tissue tolerances as estimated by Emami et al. [13]. The maximum doses delivered to one third and two thirds of the normal liver were planned in order not to exceed the tolerance dose (TD) 5/5 (the probability of 5% complications within 5 years from irradiation). The constraint for the normal liver was set at its volume receiving 33 Gy or CGE or more (V33) less than 67% and V42 less than 33%. We also considered the dose-volume limits for therapeutic partial liver RT which is recommended by Pan CC et al. [14]. Mean normal liver doses were planned in order not to exceed the tolerance dose 28 Gy (CGE) in 2 Gy or CGE fractions for primary liver cancer. Dose limits of 33, 42 and 28 Gy (CGE) in 15 fractions were specified. These are equivalent to 35, 50 and 28.6 Gy (CGE) using 2 Gy (CGE) per fraction assuming an α/β ratio of 2.5, respectively. Both IMRT and SSPT plans were optimized with the requirement that at least 95% of the PTV received the prescribed dose.

Once an acceptable treatment plan was obtained, dose-volume histogram (DVH) analysis was performed. First, DVH comparisons of the IMRT and SSPT plans were made. Although many representations of the data are valid, we used the Lyman-Kutcher-Burman (LKB) NTCP model [15‐17] to calculate the risk of RILD. The LKB-NTCP model is used to estimate the volume dependence of normal tissue toxicity that permits comparisons between plans based on DVHs. The Lyman-NTCP model describes the probability of a complication after uniform radiation of a fractional volume of normal tissue (v) to a dose (D), as

where

TD ₅₀ (v) represents the tolerable dose associated with a 50% chance of complications for uniform partial liver irradiation, where TD ₅₀ (v) is related to the whole liver (v = 1) tolerance through the power law relationship:

TD ₅₀ (1) represents the tolerance of the whole organ to irradiation, m characterizes the steepness of the dose-response at TD ₅₀ (1), and n represents the volume effect, which relates the tolerance doses of uniform whole organ irradiation to uniform partial organ irradiation. Dawson and colleagues derived these three parameters as n = 0.97, m = 0.12, and TD ₅₀ (1) = 39.8 Gy (CGE) (hereafter referred to as the Michigan parameters) from the LKB-NTCP model fitted to the complication data for 203 patients with liver cancer treatment, 11 fr/wk, 1.5–1.65 Gy (CGE) /fraction [3]. The median dose of radiation delivered was 52.5 Gy (CGE) (range 24–90). In this study, the risk of RILD was estimated using the LKB-NTCP model with the Michigan parameters.

To correct for the difference in protocol between this study and the Michigan study [3], normal liver DVHs were normalized before computation of the dose distributions from which the DVHs were computed. The physical dose values of each plan were converted to normalized iso-biologic effective doses using the linear quadratic model (LQ-model). That is, the normal liver cumulative DVH dose bins were converted to a fraction-size equivalent dose (FED) [18] described as

where f _s denotes a reference fraction size. For example, the dose per fraction delivered using the schedule from which the modeled NTCP data was derived, d, is the physical dose per fraction, n is the total number of fractions, and α/β is the ratio of the linear LQ-model parameters for the organ and end point at risk. The mean FED is then described as

where N _b denotes the total number of dose bins in the differential DVH, FED ^fs _{α/β i} is the FED in the i-th dose bin, and v _i is the partial volume associated with the i-th dose bin (∑ ^Nb _n=1 v _i =1). In this study, we calculated the FED, continuing to work with the reference point dose per fraction f _s and α/β ratio for which Dawson’s NTCP model fit derived, following the treatment schedule employed by Dawson et al. [3], and using an α/β = 2.5.

To convert the non-uniform complex dose distributions into “equivalent” uniform dose distributions, the Kutcher-Burman effective volume (V_eff)-DVH reduction scheme [16] was used. The V_eff method transforms the histogram into a uniform histogram of height V_eff and dose, D_max, equal to the maximum dose in the histogram. This transformed histogram is assumed to indicate the same probability of complications as the original histogram. Each step in the histogram of height Δv_i and extension D_i is assumed to satisfy a power law relationship so that it is adjusted to one of smaller volume ΔV_eff and extension D_max using

where n is a size parameter. This procedure is applied to each bin of the histogram. The normalized normal liver DVH with a reduction of the DVH to the effective liver volume is then irradiated as

The LKB-NTCP was then calculated using the Michigan parameters and the normalized normal liver DVH with a reduction of the DVH to the V_eff.

Mean FED, V_eff and the risk of RILD for normal liver were calculated and compared between IMRT plan and SSPT plan. We investigated the size of GTV that could be delivered at a prescribed dose for targets with successful achievement of dose constraints to the normal liver. Assuming all GTVs were spherical, the nominal diameter was calculated and used in this study. The median volume of GTV was 539.0 cm³ (range, 20.1–2194.5) and the median nominal diameter was 8.6 cm (range, 3.4–16.1) as shown in Table 1.

The mean FED for normal liver was checked against the Michigan study [3] in which Dawson et al. [3] reported that no cases of RILD were observed when the mean liver dose was under 31 Gy, and the mean liver doses associated with a 5% risk of RILD for patients with metastatic and primary liver cancer are 37 Gy and 32 Gy, respectively (in 1.5 Gy per fraction, assuming an α/β = 2.5 for the liver). The values of mean FED, V_eff and the risk of RILD for normal liver were aligned as a function of the nominal diameter of GTV. For the values of mean FED and V_eff, the Wilcoxon signed-rank test was performed to compare differences between IMRT and SSPT plans, with a p value of < 0.05 being considered significant.

For all 10 patients, a total of 26 plans for each technique of IMRT and SSPT were evaluated through careful review of the PTV isodose distributions on all slices by a radiation oncologist. Each plan provided good coverage of the target, covering 95% of the PTV. For the patients with relatively small targets having nominal diameters less than or equal to 6.3 cm (volume of GTV ≤ 130.9 cm³), the target dose optimization according to dose constraints to one third and two thirds of the volume of the normal liver was successfully achieved for IMRT plans (Table 2). However, when the nominal diameter was greater than or equal to 7.8 cm, none of the IMRT plans were able to deliver a prescribed dose of 60 Gy covering 95% of the PTV for larger targets without sacrificing normal liver. SSPT plans achieved dose constraint for the normal liver successfully in all plans except for the plan with a 16.1 cm nominal diameter GTV (Table 2). SSPT plans obtained clearly superior results to IMRT plans for every case. Examples of the DVH and dose distributions obtained with IMRT and SSPT plans for the cases of nominal diameter of GTV = 5.1 cm (a), 7.8 cm (b) and 16.1 cm (c) are shown in Figure 1 and Figure 2.

The mean FED of normal liver, summarized in Table 2, increased with the size of the GTV. The averages of the mean FED of normal liver for IMRT plans and SSPT plans were 52.1 Gy (range, 30.3–78.4) and 25.9 CGE (range, 15.2–33.3), respectively. SSPT plans maintained a lower mean FED of normal liver that of IMRT plans (p =0.01) The mean FED s for SSPT plans reached a plateau at around 30 CGE, while that for IMRT plans increased with the size of the tumor. The values of V_eff for each plan are also summarized in Table 2. The mean values of V_eff for normal liver in IMRT plans and SSPT plans were 0.64 (range, 0.40–0.86) and 0.42 (range, 0.29–0.60), respectively. SSPT achieved a lower V_eff for normal liver than did IMRT ( p < 0.001 ). The V_eff for normal liver of the SSPT reached a plateau around 0.5 even if the size of the tumor increased while that for IMRT plans increased with tumor size. For IMRT plans, both the mean FED and V_eff for normal liver drastically increased from between 6.3–7.8 cm nominal diameter of GTV, and the difference between IMRT and SSPT plans was significant when the nominal diameter of GTV was more than 6.3 cm (Table 2).

Figure 3 shows the relationship between risk of RILD probability and tumor size. As expected, the risk of RILD probability for HCC for all plans increased in relation with the value of V_eff and mean FED for normal liver; that for IMRT varied from 0.03 to 1.00, while that for SSPT varied from 0.0003 to 0.087. IMRT plans were able to keep the risk of RILD low in cases of GTV nominal diameter below 6.3 cm, but then the risk increased drastically with tumor size. For tumors of GTV nominal diameter less than or equal to 6.3 cm, the averages of RILD probability for IMRT and SSPT plans were 0.016 and 0.009, respectively. For GTV nominal diameter greater than 6.3 cm, the average RILD probabilities for IMRT and SSPT plans were 0.942 (range, 0.822–1.00) and 0.045% (range, 0.001–0.087), respectively. The RILD probabilities were lower than IMRT when using SSPT for all cases, and they were significantly lower for nominal diameters of GTV greater than 6.3 cm.