Intensity modulation in radiotherapy: photons versus protons in the paranasal sinus

Transcription

1 Radiotherapy and Oncology 66 (2003) Intensity modulation in radiotherapy: photons versus protons in the paranasal sinus Anthony John Lomax a, *, Michael Goitein b, Judy Adams b a Division of Radiation Oncology, Paul Scherrer Institute, 5232 Villigen-PSI, Switzerland b Department of Radiation Oncology, Massachusetts General Hospital, Boston, MA, USA Received 23 November 2001; received in revised form 14 July 2002; accepted 20 August 2002 Abstract Purpose: The purpose of this study is to investigate whether successive tightening of normal tissue constraints on an intensity modulated X-ray therapy plan might be able to improve it to the point of clinical comparability with the corresponding intensity modulated proton therapy plan. Materials and methods: Photon and proton intensity modulated plans were calculated for a paranasal sinus case using nominal dose constraints. Additional photon plans were then calculated in an effort to match the dose volume histograms of the critical structures to those of the proton plan. Results: On reducing the low dose contribution to both orbits in the photon plan by tightening the constraints on these structures, an increased dose heterogeneity across the target resulted. When all critical structures were more strictly constrained, target dose homogeneity and conformity was further compromised. An increased integral dose to the non-critical normal tissues was observed for the photon plans as dose was progressively removed from the critical structures. Conclusions: Both modalities were found to provide comparable target volume conformation and sparing of critical structures, when the nominal dose constraints were applied. However, the use of intensity modulated protons provided the only method by which critical structures could be spared at all dose levels, whilst simultaneously providing acceptable dose homogeneity within the target volume. q 2002 Elsevier Science Ireland Ltd. All rights reserved. Keywords: IMRT; Intensity modulated proton therapy; Proton therapy; Treatment plan comparison 1. Introduction In previous publications, we reported on our investigations of how proton treatment plans compare with both conventional and intensity modulated (IM) photon treatment techniques in a number of different sites and indications [2,7,10,14 16,28]. In all these works, although the high dose sparing of neighbouring critical structures and conformation of dose to the target volume was found to be similar between intensity modulated X-ray plans and both standard [2,10,15,16,28] and intensity modulated protons [7,14], we consistently found that the use of protons can significantly reduce low-to-mid dose levels for all organs at risk. Indeed, the physical characteristics of protons dictate that this must, almost, always be the case when considering all non-target tissues together that is, the total integral dose delivered to a patient. However, this does not necessarily have to be the case for the dose deposited in locally restricted regions such as individual critical organs. * Corresponding author. When comparing two radiation modalities it is usual to develop optimised treatment plans, using the same constraints and optimisation criteria for each, and then compare the resulting dose distributions and other derived quantities. However, the so-called optimised plans may not be the best that can be achieved. The dose constraints and search algorithm, for example, often drive a plan to meet but not necessarily to exceed the constraints imposed. When protons are compared in this manner with X-rays, the superior dose distribution of protons often leads to their delivering much less dose in some or all identified organs at risk than the corresponding X-ray plan. This study examines whether, in such a situation, a better X-ray plan than the optimised X- ray plan can be achieved if the constraints on the X-ray plan are successively lowered. If so, then the dosimetric advantage of the proton plan (and hence, presumably the biological advantage) may be less than that appears from a comparison with the original optimised plans. In this work, we investigate, in more detail, the differences between intensity modulated X-ray and proton therapies (henceforth referred to as IMXT and IMPT, /02/$ - see front matter q 2002 Elsevier Science Ireland Ltd. All rights reserved. doi: /s (02)

2 12 A.J. Lomax et al. / Radiotherapy and Oncology 66 (2003) respectively) for a paranasal sinus case. In particular, for this case, we wish to see how far doses in locally bounded regions of the patient, such as neighbouring, individual critical structures, can be reduced in IMXT, and how these then compare to the doses delivered by IMPT. An adaptive approach has been adopted for this work, in which progressively stringent dose and dose volume constraints were applied to the IMXT plans in order to match, as closely as possible, the dose volume histograms (DVH) of the IMPT plan for all identified critical structures. In order to eliminate possible artefactual differences that may arise from the use of different optimisation regimes, and/or different dose calculation and analysis techniques, we used the identical optimisation methods and assumptions in the calculation of both IMXT and IMPT plans. 2. Material and methods We have previously reported on the development of methods for the optimisation and calculation of intensity modulated proton plans [9]. For this work, these methodologies have been modified to allow, also, for optimisation of photon dose distributions, through the substitution of precalculated photon pencil beams for the fixed extent spreadout-bragg-peak (SOBP) proton kernels employed in the IMPT approach described in Ref. [9]. The photon kernels used here were calculated in water, using the code developed at the University of Wisconsin [1,12]. With this code, three-dimensional (3D) dose kernels for a collimated 6 MV beam with a SSD of 60 cm, passing through a 4 mm wide aperture were produced. In order to verify the accuracy of the resulting kernel, a depth dose curve and lateral profile for cm 2 open field were calculated and compared to measured data. Qualitatively good agreement was found between the two data sets. For the proton pencil beams, we used the model described by Scheib and Pedroni [23] and further developed by Schaffner et al. [22]. For our calculations, we assumed that the initial proton pencil beam cross section in air is Gaussian, has a full-width-at-half-maximum (FWHM) of 6 mm and is invariant as a function of incident proton energy. Multiple Coloumb scattering in the patient was modelled following the methods described by Scheib and Pedroni [23]. For both radiation types, tissue heterogeneities in the patient (taken from the computed tomographic (CT) study) were taken into account by assigning a depth to each dose grid point equal to its water equivalent depth (see e.g. Petti [19] or Schaffner et al. [22]). The optimisation scheme is identical to that described in Ref. [9], but with the additional capability of dose volume constraints following the approach described by Bortfeld [3]. Using the nomenclature introduced in Ref. [9], all proton plans were calculated using full 3D modulation (3D-IMPT). For all plans, pencil beam weights for the optimisation were calculated on a 4 mm grid normal to the incident beam direction, and additionally at 4.5 mm depth increments for the 3D-IMPT plan. For the paranasal case investigated here, three target Table 1 Dose and dose volume constraints for all structures for plans P1, X1, X2 and X3 Plan Structure Dose constraint (Gy) Dose volume constraint Importance factor P1,X1 T T T Left globe 50 5 Right globe 50 5 Brainstem 53 5 Left optic nerve Right optic nerve X2 T T T Left globe 22.8 Gy, 50% 5 Right globe 22.8 Gy, 20% 5 Brainstem 53 5 Left optic nerve Right optic nerve X3 T T T Left globe 22.8 Gy, 30% 5 Right globe 15.2 Gy, 20% 5 Brainstem 15.2 Gy, 20% 5 Left optic nerve 38 Gy, 50% 10 Right optic nerve 38 Gy, 40% 10

3 A.J. Lomax et al. / Radiotherapy and Oncology 66 (2003) volumes, T1, T2 and T3 were defined, for which doses of 76, 66, and 54 Gy, respectively, were prescribed. Dose constraints to the critical structures for the nominal photon and proton plans are shown in Table 1. In applying these optimisation methods to the case described here, we used the proton plan, without further iterations, as the reference plan. That is, first the 3D-IMPT plan (P1) was calculated using a set of nominal, maximum dose constraints (Table 1). An IMXT plan was then calculated (X1) using the same constraints as the IMPT plan. Two additional IM photon plans were subsequently developed. In the first of these (X2), additional dose volume constraints to both globes were set in such a way as to constrain the low dose (#30% of the target prescription dose) part of the DVHs for these organs to essentially the same levels as were obtained in the initial proton plan (see Table 1). In the remaining photon plan (X3), dose volume constraints were added to all defined critical structures, in an attempt to match the IM photon DVHs for the critical structures to those of the IM proton plan (see Table 1). In all plans, the constraints to the PTVs remained constant. For all photon and proton plans, the beam geometry was identical, consisting of nine equally spaced, coplanar fields, each separated by 408. The resulting three IM photon plans were compared with the nominal IM proton plan using DVH analysis and visual comparison of the 3D dose distributions. Although only one section is illustrated in the figures, all dose calculations and DVHs etc. were fully three-dimensional. 3. Results Dose distributions for the nominal plans (P1 and X1) at the level of the globes, optic nerves and chiasm and at the level of the petrous ridge are shown in Fig. 1. DVHs for all targets, both globes and the brain stem are shown in Fig. 2. Both plans produce good conformation of dose to each of the three target dose levels, and provide similar target coverage (for example, the volume of T1 with dose.68 Gy is and 99.6% for plans P1 and X1, respectively), and both plans successfully constrained all critical structures to the required maximum dose levels. However, at dose levels below those at which the constraints were imposed, the proton plan provides significantly greater sparing of all critical structures in comparison to the initial photon plan, particularly in the right and left globes. This result is consistent Fig. 1. Dose distributions for plans (a) P1 and (b) X1 at the level of the globes and petrous ridge (upper and lower images, respectively).

4 14 A.J. Lomax et al. / Radiotherapy and Oncology 66 (2003) Fig. 2. Comparative DVHs for plans P1 and X1 for all targets, both globes and brain stem. with the work reported in Refs. [2,7,10,14 16,28] and is clearly seen by visual inspection of the two dose distributions in Fig. 1, and in the critical structure DVHs of Fig. 2. In order to investigate the potential of IMXT to spare critical structures to the same extent as that achieved in the nominal proton plan, we applied additional constraints to the critical structures in an effort to force the optimisation process to reduce the volume of critical organs receiving doses below the maximum dose constraints. This approach was first applied to the orbits alone (plan X2), by applying an additional dose volume constraint as shown graphically in Fig. 3 and tabulated in Table 1. Fig. 4 compares DVHs for: the high dose target (T1), both globes, and the brain stem for plans P1, X1 and X2. From this figure, it can be seen that, by applying an additional dose volume constraint to the globes, a similar degree of dose sparing of the globes could be achieved as was observed with the nominal proton plan. However, this was at the cost of an increased heterogeneity of dose in the target volumes (volume of T1 with dose.68 Gy, 97.7%) and a small increase in the volume of brain stem irradiated to doses greater than about 22 Gy (see Fig. 4). These features are confirmed through visual inspection of the dose distribution (Fig. 5c), in which some loss of conformation to the high dose target is seen, as well as an increase in dose to the brain stem resulting from dose contributions from some highly weighted pencil beams which pass through this organ. In plan X3, dose volume constraints were applied to all the defined critical structures (brainstem and optic nerves, as well as left and right globes, as tabulated in Table 1), again with the purpose of matching the IMXT DVHs to those of the nominal IMPT plan (plan P1). The resulting DVHs are shown in Fig. 6, as is the DVH for the non-critical normal tissues (all normal tissues excluding the defined critical structures). Although the DVHs for the brain stem and both globes now compare very favourably with the corre- Fig. 3. DVHs of the right globe for plans P1 and X1, together with a schematic representation of the form of the maximum dose constraint used for plan X1 and the subsequent dose volume constraint applied in plan X2.

5 A.J. Lomax et al. / Radiotherapy and Oncology 66 (2003) Fig. 4. Comparative DVHs for plans P1, X1 and X2 for the high dose target (T1), both globes and the brainstem. sponding DVHs of plan P1 at all dose levels, the target dose homogeneity is further compromised (volume of T1 with dose.68 Gy, 94.0%), and some increase in the volume of both optic nerves irradiated to high doses (.55 Gy) can also be observed. Visual inspection of the X3 dose distribution (Fig. 5d) shows poor conformation of dose to the T1 target, and significant volumes of high doses (.72.2 Gy) in the lower dose targets and neighbouring normal tissues. The DVHs for other tissues i.e. normal tissues outside the PTVs and not included in the identified critical structures is also shown for all plans in Fig. 6. In terms of the dose to other tissues, all X-ray plans are significantly worse than the proton plan, and X3 is worse than X2, which is worse than X1 (an increase of 10.3% between plans X1 and X3). The differences appear small in Fig. 6, but, because the DVH corresponds to a large overall volume, the differences are relatively substantial in terms of absolute volume. 4. Discussion In this work, we have used a trial and error adaptive approach to optimising IMXT plans, through the manual definition of successively more stringent constraints applied to critical organs, in order to attempt, to match the critical organ DVHs of the IMXT plans to the nominal IMPT plan. Through this approach, we have found that, for this example at least, it is indeed possible to match some of the photon DVHs to those of the proton plan. However, It has not been possible to match all the DVHs (for example, the optic nerves) and the reduction of dose in the globes and brainstem has been achieved only at the cost of reduced dose heterogeneity in, and conformity to, the target volumes. The reasons for this loss of conformity can be inferred from the dose distributions and anatomy shown in Fig. 5d. Both the globes and the brainstem, at the level shown, are close to the target volume, and are also relatively large in comparison to it. When dose constraints to these organs are made more stringent, the optimisation algorithm causes the weights of pencil beams passing through these organ to be significantly reduced, if not switched-off completely. Such pencil beams can therefore contribute little dose to the target volume. As the globes and brainstem are relatively evenly spread around the targets, this reduces the degrees of freedom for the optimisation and more dose must be brought in through the remaining regions which are not shadowed by these organs. This results in an inevitable loss of conformity and, in this case at least, a loss of target dose homogeneity. By contrast, for the IMPT plan, the 3D modulation of Bragg peaks can recover some of these lost degrees of freedom, albeit, in a different manner, and can thus better reduce dose to the normal tissues whilst preserving target dose homogeneity. For both photon and proton plans in this study, we have

6 16 A.J. Lomax et al. / Radiotherapy and Oncology 66 (2003) used a set of nine, evenly spaced incident beams, without considering the possibility of optimising the beam orientations. For photons, this arrangement has been chosen as a reasonably standard arrangement, which can readily be delivered by a centre with a multi-leaf capability. The same number of fields has been selected for the proton plan in order to provide a one-to-one comparison of the techniques. The use of nine beams is certainly less than that typically considered for a tomotherapy treatment, and this could perhaps have some influence on the quality of the photon plans. On the other hand, the advantage of many versus few photon beams is the subject of some debate in the literature [4,5,13,17,18,26], with some studies indicating that for particular cases, there may be little advantage in using more than nine beams [21,27]. It is also unclear whether beam orientation optimisation could have an effect on the quality of the plans presented here. For example, the use of non-coplanar beams in this case could be problematic, due to the fact that the bulk of the mid- and lowdose targets (T2 and T3) are caudal to the optic structures, making the blocking of these structures difficult for vertex beams. On the other hand, Pugachev and colleagues have shown that the use of optimised, non-coplanar beam orientations and IMXT could improve the quality of plan in a nasopharynx case [20]. Nevertheless, there is little reason to believe that beam orientation optimisation, when applied to both modalities, would be of a greater advantage for photons than it would be for protons. Indeed, at least from the point of view of total integral dose, the use of nine proton fields could well be sub-optimal in this case, and the use of a smaller number of well chosen fields could achieve a similar level of target coverage and sparing of critical structures, while significantly reducing overall integral dose (see e.g. the comparison in 11). Having said this, the issue of optimisation of the number and orientation of beams, and its effect on such comparisons, is one that certainly needs to be addressed in more detail. A central premise of this work is that there is a clinical rationale for wanting to reduce doses in critical structures well below their presently accepted constraint values. However, one must address the problem of how valid this premise is. After all, many radiation oncologists work, to certain dose tolerances, in the belief that keeping organs at risk below these levels will keep the chance of complication sufficiently small (usually less than 5%). We consider that there can and will be situations, where further reductions may be clearly necessary. First, there are three groups of patients where lower tolerances than those normally used should be applied. These are paediatric patients, patients with concomitant therapies and patients Fig. 5. Dose distributions for (a) P1, (b) X1, (c) X2 and (d) X3 at the level of the globes.

7 A.J. Lomax et al. / Radiotherapy and Oncology 66 (2003) Fig. 6. DVHs for all plans, for the high dose target (T1), both globes, brain stem, optic nerve and non-critical normal tissues. requiring re-irradiation due to recurrent disease. In such cases, the radiation tolerance of all normal tissues, whether defined as critical or not, may well be compromised in relation to the normally accepted tolerances. In such cases therefore, there may be a value in reducing doses to normal tissues to the lowest possible level. Second, the data needed to establish the response of organs to ionising radiation is rather sparse, and well known tolerance doses may not be as accurate as thought. As an example, although not relevant to this case, there is now some evidence to suggest that the rectum, hitherto regarded as essentially a serial organ (n ¼ 0:12 in Burman s fit to the Lyman model [6,8,11]), may well be sensitive to large volumes of lower dose [25]. Additionally, van der Kogel and his collaborators have reported that even the tolerance of the spinal cord of rats to small hot spots can be modified by a presence of a bath of lower dose [24]. Indeed, although the effects of large volumes of mid-to-low doses are largely unknown for most organs, this does not imply that such doses are necessarily benign, especially when long-term side effects are considered. Lastly, the relative reduction of doses to critical structures with respect to those delivered to the target by which these results can be alternatively interpreted is a necessary pre-requisite for any form of dose escalation. The ability to, generally and specifically, reduce doses to normal tissues could, therefore, have important implications for further increasing the doses delivered to particularly radiation resistant tumours. A final point should be made about the IMPT plan presented in this work. As no attempt has been made to further reduce doses to the critical structures for this plan, it should not necessarily be considered as being the best that can be achieved with IMPT. It would, of course, be possible to introduce tighter constraints to this plan also, and thus reduce critical organ doses even more, with, at least at some point, a corresponding loss of target dose homogeneity. Indeed, it is an inherent property of 3D-IMPT that, due to the localisation of dose in the Bragg peak, and the subsequent possibility of modulating fluences in three dimensions, it provides many more degrees of freedom for the optimisation than are available to IMXT. 5. Conclusions This work has attempted to explore the limits of what is achievable with IMXT methods in selectively avoiding defined critical structures, and to compare these to the possibilities provided by IMPT. Although the optimisation method, treatment geometries and indication presented are all very specific, the fact that the same parameters have been applied to both treatment modalities leads us to believe that the results presented here could well be generally applicable. That is, although IMXT can selectively be defined to spare some critical structures to similar levels as a nominally calculated proton plan, this cannot be achieved for an

8 18 A.J. Lomax et al. / Radiotherapy and Oncology 66 (2003) indefinite number of critical structures, and not without consequence on target dose homogeneity. The fact that the IMPT plan presented here was a first guess plan implies that the additional degrees of freedom inherent in the delivery of 3D-IMPT can be exploited to further reduce doses to critical neighbouring critical structures, or alternatively, for achieving similar results with a smaller number of fields. Either way, it is likely that IMPT could provide a technique for further advancing the great steps achieved by the introduction of intensity-modulated methods into X-ray based radiotherapy. References [1] Download of code can be found at www-madrad.radiology.wisc.edu. [2] Baumert BG, Lomax AJ, Miltchev V, Davis JB. A comparison of dose distributions of proton beams and conformal stereotactic photon beams for irradiation of brain lesions. Int J Radiat Oncol Biol Phys 2000;49: [3] Bortfeld T, Stein J, Preiser K. Clinically relevant intensity imodulation optimisation using physical criteria. In: Leavitt DD, Starrkschall G, editors. XII International Conference on The use of computers in radiation therapy, Salt Lake City, pp [4] Brahme A. Optimisation of radiation therapy and the development of multileaf collimation. Int J Radiat Oncol Biol Phys 1993;25: [5] Brahme A. Optimisation of radiation therapy. Int J Radiat Oncol Biol Phys 1993;28: [6] Burman C, Kutcher GJ, Emami B, Goitein M. Fitting of normal tissue data to an analytical function. Int J Radiat Oncol Biol Phys 1991;21: [7] Cella L, Lomax AJ, Miralbell R. Potential role of intensity modulated proton beams in prostate cancer radiotherapy. Int J Radiat Oncol Biol Phys 2001;49: [8] Emami B, Lyman JT, Brown A, Coia L, et al. Tolerance of normal tissue to therapeutic irradiation. Int J Radiat Oncol Biol Phys 1991;21: [9] Lomax AJ. Intensity modulation methods for proton radiotherapy. Phys Med Biol 1999;44: [10] Lomax AJ, Bortfeld T, Goitein G, et al. A treatment planning intercomparison of proton and intensity modulated photon radiotherapy. Radiother Oncol 1999;51: [11] Lyman JT, Wolbarst B. Optimisation of radiation therapy. Int J Radiat Oncol Biol Phys 1987;13: [12] Mackie TR, Ahnesjø A, Dickof P. Development of a convolution/ superposition method for photon beams. In: Bruinvis AD, van der Giessen PH, van Kleffens HJ, Wittkamper FW. IX International Conference on The use of computers in radiation therapy, Amsterdam, pp [13] Mackie TR, Deasy J, Holmes T, Fowler J. Letter in response to Optimisation of radiation therapy and the development of multileaf collimation by Anders Brahme. Int J Radiat Oncol Biol Phys 1994;28: [14] Miralbell R, Cella L, Weber D, Lomax AJ. Optimizing radiotherapy of orbital and paraorbital tumors: intensity modulated X-ray beams versus intensity modulated proton beams. Int J Radiat Oncol Biol Phys 2000;47: [15] Miralbell R, Lomax AJ, Borfeld T, Rouzaud M, Carrie M. Potential role of proton therapy in the treatment of pediatric medulloblastoma/ primitive neuro-ectodermal tumors: reduction of the supratentorial target volume. Int J Radiat Oncol Biol Phys 1997;38: [16] Miralbell R, Lomax AJ, Russo M. Potential role of proton therapy in the treatment of pediatric medulloblastoma/primitive neuro-ectodermal tumors: spinal theca irradiation. Int J Radiat Oncol Biol Phys 1997;38: [17] Mohan R, Ling CC. When becometh less more? (Editorial). Int J Radiat Oncol Biol Phys 1995;33: [18] Mohan R, Ling CC, Stein J, Wang X-H. The number of beams in intensity-modulated treatments: In response to Drs Söderström and Brahme. Int J Radiat Oncol Biol Phys 1996;34: [19] Petti P. Differential-pencil-beam dose calculations for charged particles. Med Phys 1992;19: [20] Pugachev A, Li JG, Boyer AL, Xing L. Beam orientations in IMRT: to optimize or not to optimize? In: Schlegel W, Bortfeld T, editors. XIII International Conference on The use of computers in radiation therapy. Heidelberg, pp [21] Sauer OA. Vergleich verschiedener Optimierungskonzepte für die intensitätsmodulierte Strahlentherapie. Z Med Phys 1999;9: [22] Schaffner B, Pedroni E, Lomax AJ. Dose calculation models for proton treatment planning using a dynamic beam delivery system: an attempt to include density heterogeneity effects in the analytical dose calculation. Phys Med Biol 1999;44: [23] Scheib S, Pedroni E. Dose calculations and optimizations for 3D conformal voxel scanning. Radiat Environ Biophys 1992;31: [24] Schippers JM, van Luijk P, Bijl HP, Coppes R, Konings AWT, van der Kogel AJ. Modeling of volume effects after proton irradiation of the rat spinal cord (abstract). Int J Radiat Oncol Biol Phys 2000;46:782. [25] Skwarchuk MW, Jackson A, Zelefsky MJ, Venkatraman ES, et al. Late rectal toxicity after conformal radiotherapy of prostate cancer. Int J Radiat Oncol Biol Phys 2000;47: [26] Söderström S, Brahme A. Small is beautiful and often enough: in response to Drs Mohan and Ling. Int J Radiat Oncol Biol Phys 1996;34: [27] Stein J, Mohan R, Wang X-H, et al. Number and orientations of beams in intensity-modulated radiation treatments. Med Phys 1997;24: [28] Zurlo A, Lomax AJ, Hoess A, et al. The role of proton therapy in the treatment of large irradiation volumes: a comparative planning study of pancreatic and biliary tumours. Int J Radiat Oncol Biol Phys 2000;48: