Searching standard parameters for volumetric modulated arc therapy (VMAT) of prostate cancer

Ten consecutive patients with a mean age of 71 years, which had been treated with primary external-beam radiotherapy for localized prostate cancer, were included in the retrospective planning study. Written informed consent was obtained from the patients for publication of this report and any accompanying images. All ten patients were immobilized in a vacuum mattress (Blue-BAG™ BodyFIX®, Medical Intelligence, Schwabmünchen, Germany) in supine position according to Boehmer et al. [20] and had three-dimensional treatment planning with a CT slice thickness of 5 mm. In the treatment-planning system Oncentra® MasterPlan, v3.3 (Nucletron B.V., Veenendaal, The Netherlands) the gross target volume (GTV: prostate gland and seminal vesicles), the clinical target volume (CTV, 5-mm three-dimensional margin added to the GTV excluding the rectal volume), the planning target volume (PTV, 10-mm three -dimensional margin added to the GTV without respect to the rectum), and normal tissues were delineated in each slice. The delineation of the volumes of interest followed the description of Bos et al. [21], who also showed that a little overlap of PTV and rectum results from the lateral expansion of the PTV. The rectal volume (according to Guckenberger et al. [22]) and urinary bladder as well as the femoral heads were delineated as organs at risk (OAR). Additional regions were defined to improve the plan quality: the PTV plus a margin of 5 mm (PTVm) was first subtracted from the patient outline (OL) for a non target volume (OL-PTVm) to avoid hot spots in the normal tissue; It was second subtracted from the rectum volume resulting in the posterior part of the rectum (R-PTVm) to achieve reduced dose in this region as similarly described in [12]. The last structure constructed was PTV minus CTV (PTV-CTV) to model the dose gradient from PTV to CTV.

For planning and measurements the data of a Synergy®S linear accelerator (Elekta Ltd, Crawley, United Kingdom) with 6MV photons were used, equipped with a BeamModulator™ head, an iViewGT™ electronic portal imaging device, and an on-board cone-beam CT XVI. The multileaf collimator has 40 leaf pairs, each of nominal width of 4 mm projected to isocentre, diaphragms limit the maximum field size of 21 cm × 16 cm. There are no moveable jaws. Leaf interdigitation is allowed without limitations, which means that each leaf can travel across the whole field size, independent on the position of neighbouring leaves. Closed leaf pairs are automatically shifted below the fixed diaphragm to minimize transmission between the leaf ends of opposed leaves.

The following linac specific parameters for VMAT delivery have to be fed into the planning system as described similar for Eclipse™ and RapidArc® [23]:

Minimum and maximum number of MU per degree of gantry rotation 0.10 MU/° and 20.0 MU/° respectively, minimum MU per cm leaf travel 0.30 MU/cm, maximum gantry speed 6.00°/s. Maximum leaf speed is 2.4 cm/s, the dynamic minimum leaf gap 0.2 cm, and the static minimum leaf gap 0.0 cm [24]. The maximum nominal dose rate is 500 MU/min. Seven fixed dose rate levels are available, each half the dose rate of the next higher level, continuous variation is not possible. Actual dose rates may differ from nominal dose rates by ±25%. For VMAT delivery the console software Precise Desktop® 7 determines automatically the fastest combination of dose rate, gantry speed and leaf speed [19].

The treatment planning was performed with Oncentra MasterPlan® v3.3 SP1, released clinically in December 2009, on a 64 bit Windows system. From the above mentioned data the two lowest dose rate levels had to be omitted, as the TPS allows only five different levels [19]. In the Beam Modeling module the user appoints only the treatment unit, the number of beams, the energy and the collimator angle. Although the linac could vary the collimator angle during VMAT delivery this feature is not implemented in the current TPS software version. Switching to the Optimizer module the user has further parameters to determine: gantry start angle, rotation direction, arc length, gantry angle spacing between subsequent control points (2°, 3°, 4°, or 6°), maximum delivery time (MDT), number of arcs, and leaf motion constraint in cm/°, which was set to 0.5 cm/°. The optimizer called RayArc has been developed by RaySearch and is described in [25] and in a White Paper [26]. RayArc is designed to create deliverable plans for Elekta and Varian machines. It exists in two versions, as SmartArc module integrated in Pinnacle³ Version 9 (Philips Healthcare, The Netherlands) and the mentioned Oncentra version.

The plans were set up with simultaneous integrated boost (SIB) in 33 fractions aiming for 59.4 Gy minimum dose to the PTV and 71.0 Gy minimum dose and 74.2 Gy maximum dose to the CTV, which was used as the boost treatment volume. As in [13, 18] and also in [25] (with little modifications) the same set of dose volume objectives (DVO) was used (Table 1) for all VMAT and IMRT plans, based on experiences with the same optimization module for IMRT [27]. The DVO are set as planning goals to be met if possible. As Dobler et al. [28] stated it might be possible to achieve better plan quality using DVO specifically designed for the algorithm and delivery method. However, a direct comparison of results would be impossible.

The isocentre was positioned in the centre of the CTV. All calculations were performed by a pencil beam algorithm with a calculation grid spacing of 4 mm. Both the slice thickness of 5 mm and the calculation grid spacing of 4 mm were rather coarse regarding a leaf width of 4 mm. The reason is that the main memory of the hardware was too small to handle a better resolution in the optimization process. To verify that there was no misinterpretation by interpolation, for a subset of ten different plans, one for each patient, a forward calculation was done with a finer resolution. An artificial CT with a slice thickness of 2.5 mm was generated by means of the TPS. The calculation grid spacing was set to 2 mm. All plans were forward calculated without new optimization and without any change. There was either no change, or the values shifted slightly for all plans systematically to the same direction, keeping the general conclusion unchanged.

Altogether 360 plans were calculated, 36 per patient. The following parameters were varied systematically for each patient:

Besides the median values in the OAR the following parameters were evaluated: the total number of MU, the homogeneity H and D₅ and D₉₅ in the CTV with H defined as H = (D₅–D₉₅)/D_Average[33, 34]. D₅ and D₉₅ are defined as the dose to 5% and 95% of the CTV, respectively. Furthermore the maximum dose to the posterior part of the rectum and the minimum dose in the PTV were analyzed as endpoints of the optimization. The results of all plans were grouped according to the modified parameters and analyzed statistically using SPSS PASW Statistics programme version 18.0.0. Groups were compared applying student’s test (t-test) for paired samples. As level of significance α = 5% (two-sided) was chosen. To obtain a good statistical significance as few groups of plans as possible were established for each parameter.

Moreover all plans meeting the endpoints mentioned above were selected and the incidence in the different groups was determined.

Five groups of VMAT plans and the IMRT group containing ten plans for the different patients each were measured with a MatriXX Evolution® 2D-array with gantry angle sensor (IBA Dosimetry, Schwarzenbruck, Germany) [35, 36] positioned in the horizontal isocentre plane. The gantry angle sensor was attached to the gantry and connected to the array. Every 200 ms a dose matrix and the corresponding gantry angle were acquired automatically. The matrices were corrected for angular dependencies, including couch attenuation as described in [35], using a correction factor matrix implemented in the software. Lastly they were summed up to the dose of the complete plan. This was compared to the calculated dose matrices by gamma evaluation [37] with a dose tolerance of 3% of the maximum dose and a distance to agreement of 3 mm. We performed the evaluation for the area with dose values above 10% as recommended in [38]. The percentage of pixels out of range (γ > 1) was analyzed. In addition radiation times were measured from pressing the start button to end of delivery.

All results are shown in Table 2; significant differences between groups are described in the corresponding sections. On behalf of brevity, differences and corresponding p-values are not mentioned if not statistically significant except of an intermediate range (α = 10%), otherwise they are described as significant (α = 5%) or highly significant (α = 1%).

The median dose to the femoral heads in a range from 24.5 Gy to 33.4 Gy as average for the subgroups of ten patient plans stayed far below the DVO of 50 Gy. According to Kjær-Kristoffersen et al. [13] this parameter was disregarded furthermore.

In a last step for each parameter combination the plans of the ten different patients were evaluated and the average values for the specified parameters graphically depicted. Our examples (Figure 1) show the bar charts of the endpoints, which did not meet the specified DVO in every case as average over ten plans. For the sake of clarity, the bars of the groups with collimator 0° have been omitted, as it had been shown above that they gave poorer results. The minimum dose in the PTV was only met in two groups, [SA, MDT 110 s, GS 2°] and [SA, MDT 110 s, GS 4°]. Two groups with DA technique remained just under the prescribed value of 59.4 Gy: [DA, MDT 80 s, GS 2°] and [DA, MDT 80 s, GS 4°]. Closest to this value with GS6° was the group [SA, MDT 80 s, GS 6°] reaching 59.3 Gy. Furthermore these five groups fulfilled the DVO for the posterior rectum. Figure 2 shows the dose distributions for one specific IMRT plan, and two VMAT plans ([DA, MDT 80 s, GS 4°] and [SA, MDT 110 s, GS 4°]).

The four groups resulting as the optimal groups of VMAT from the graphical evaluation, the just mentioned group with GS 6° [SA, MDT 80 s, GS 6°] and the IMRT group were selected for plan verification. Table 3 shows the average values and standard deviations for the delivery times and the percentage of pixels out of range.