The results of the SA and DA comparison do not show a clear advantage of one technique over the other.
Nevertheless the differences are highly significant: a slight benefit for SA at the minimum dose in the PTV which is yet below the specified value in the DVO is in opposition to lower dose values in the rectum for DA. In most cases the minimum dose to the PTV is localized in the posterior region. Keeping the minimum dose here is correlated to a higher dose to the rectum. The DA technique favours the dose reduction by focusing on one half of the PTV for each arc. This is promoted by a higher MDT. A decrease of the objective weight for the posterior part of the rectum and an increase of the weight for the minimum dose in the PTV might improve the DA results. This could be a future step to find an optimal set of DVO.
SA optimization comes out with fewer MU than DA, because DA focuses on one half of the PTV for each arc. Fewer MU for SA treatments are an indicator for shorter treatment times. Additional time is also needed by the record and verify system and the control system of the linear accelerator to prepare the second arc. These circumstances could be confirmed in the plan verification measurements: The treatment times for the selection of SA were all smaller than 2 min 15 s, whereas all DA treatments took more than 3 min 10 s up to nearly 5 min and are therefore above the level of three minutes where considerable organ movements might occur. The studies of Ghilezan  and Nederveen  have shown an influence of the delivery time on intrafraction organ motion: the longer the fractional treatment lasts the higher is the risk of anatomic deviation. Kupelian  argued with this potentially mismatch for daily image guidance and adaptive radiotherapy. In contrast actually there exists no defined standard recommendation for online correction in daily practice when using a VMAT technique for treating prostate cancer. An option to increase the minimum dose to the PTV in two arcs might be using orthogonal collimators for each arc. However, DA is not possible with different collimator values.
Regarding the collimator angle there is a clear outcome that an angle of 45° should be preferred to an angle of 0°. It is first advantageous for the dose distribution; this might be explained by the hypothesis that the leaves of the MLC in parallel opposed beams move in orthogonal directions and therefore these beams are not redundant . Furthermore Otto explains  that without collimator rotation only a single leaf pair can be used to modulate the intensity within one CT slice. And second the number of MU is 8% lower with collimator 45° than using a collimator angle of 0° which can be explained by the fact that with collimator 45° it is possible to irradiate the right and left side of the PTV at the same time sparing the rectum and urinary bladder in the centre, which is not possible with collimator 0°.
For SA treatments a MDT of 80 s seems to be too short to achieve an acceptable dose distribution. All evaluated dose parameters were in most cases significantly worse than in the two other groups. Only for the number of MU the best value is achieved. This is not surprising, as there are only two options for the optimizer to reduce the treatment time: higher dose rates and fewer MU. They are closely interconnected during the optimization process . Although a higher value than 80 s for the MDT is necessary, the decision between MDT 110 s and MDT 150 s is not clear without ambiguity: the dose distribution in the CTV and PTV is nearly equivalent, the doses to the OAR are a little bit higher for MDT 110 s, but the MU are lower than for MDT 150 s. A slight advantage for MDT 150 s might be derived from the incidence of plans meeting the endpoints. However, regarding the importance of short treatment times as discussed above and the slight difference in the dose values a MDT 110 s seems preferable, which is affirmed by the graphical evaluation.
The decision in the DA groups is clear: the significantly highest minimum dose to the PTV is achieved with MDT 80 s in two arcs, yielding very similar results for the homogeneity to the CTV and the dose values to the OAR without statistically significant differences and again the lowest number of MU which is statistically significant. Consequently the shortest MDT provides the best results.
It might be supposed that for a given MDT all delivery times shorter are considered by the optimizer. Our measurements indicate that this maximum is exploited. A long MDT results in more arc sections with a dose rate at the lower bound allowing less modulation between the control points. Therefore plans with a longer MDT can be worse than others with a shorter one.
Analyzing the results of the GS comparison we find the best value for the minimum dose to the PTV for GS 4°, here the group with GS 6° has the significantly lowest value. However, in this group we achieve the significantly lowest median dose to the urinary bladder and the lowest number of MU. Nevertheless the advantage of GS 4° versus GS 2° is quite small and might only be traced back to the lower modelling accuracy and might disappear in delivery. Regarding the results of the measurements (Table 3) we find the better accordance for the γ value the smaller the GS. However this is only significant for the DA technique (p = 1.0%), not significant (p = 6.8%) for the SA technique using MDT 110 s, either for GS 6° compared to the other SA measurements. The passing rates are similar as at Feygelman et al.  between 95.6% and 100.0% for GS 2°, 94.5% and 99.6% for GS 4° and 93.7% and 99.6% for GS 6°. The increase of the number of pixels failing the gamma criterion can be explained by the worse modelling of the continuous movement using a coarser gantry resolution. This model, the small-arc approximation, has theoretically been described by Webb and McQuaid . We conclude that the approximation is still valid for GS 4°, when we compare the passing rates with IMRT, which are nearly the same. According to Feygelman et al.  and also suggested by Bzdusek et al.  we would use the largest GS consistent with good dosimetric results (GS 4°) to minimize the calculation time. Furthermore in this group nearly as many plans complied with the DVO (26) as in the GS 2° group (27). Only 21 were found with GS 6°. Consequently GS 6° is not recommended for VMAT planning of prostate cancer.
The IMRT calculations lack of a similar systematic variation of parameters as done for VMAT and therefore detailed statistic intercomparisons would not be appropriate. As the main parameters and DVO were taken (and adapted) from an IMRT study  the results should be quite characteristic. At most one might therefore expect an advantage for the IMRT plans. On the contrary to  no modifications to the DVO were made to improve VMAT plans. Figure 1 and Table 2 show that all VMAT groups with collimator 45° give comparable or better results for target volumes and OAR. IMRT attains the lowest number of MU. That is a discrepancy to other prostate planning studies referring to RapidArc® [12–14], but for the most part it is due to the fact that the number of MU in the mentioned studies with sliding window IMRT technique is higher than ours with step-and-shoot. This has also been observed by Alvarez-Moret et al. . They also report that for equal or slightly more MU even a DA technique takes only 30% of an IMRT delivery time, for SA it is only 15%. Dobler et al.  list one prostate case with a time reduction from a seven field IMRT to a SA VMAT plan to 43% at comparable MU. Table 3 shows that our results for a larger number of patients confirm this benefit with statistical significance with a time reduction to around 50% for DA and to about 30% respectively 20% for SA.
It might be a surprise that higher MU are not generally related to improved plan quality. Obviously additional MU are not always exploited in smaller MLC apertures for better dose modulation. A similar effect has not only been observed in an IMRT TPS intercomparison , IMRT planning study comparisons of different algorithms within the same TPS [27, 28, 44], but also in the VMAT comparison of Palma et al. . Surely there remains some potential to improve the algorithm.
Up to now the TPS does not offer the option of dynamic collimator rotation, which is technically available on the Elekta Synergy®S linear accelerator. As Webb has shown  this would help to avoid “parked gaps” for closed leaf pairs, which are needed during the treatment, but cannot be parked below the fixed diaphragm due to limited leaf speed. Avoiding such ‘unwanted fluence’ might be a next step to improve dose distributions in VMAT plans.
The results of this planning study may in detail be valid only for the chosen set of DVO. However the references [16, 32] have shown that authors using different equipment and protocols but similar algorithms have achieved comparable results regarding GS and collimator angle. Although the majority of the plans failed the DVO, it could be shown that all planning aims are met using an appropriate set of parameters. The DVO for the rectum and for the PTV are somehow counterworking goals which are not met for all settings, as also was found by Crijns et al. , where all five RapidArc planning approaches failed achieving the rectum maximum dose.