lnt. J. Radration Onrologv Biof. Phys. Vol. 21. pp. 751-165 Printed in the U.S.A All nghts reserved.
Copyright
0360-3016191 $3.00 + .@I 0 1991 Pergamon Press plc
??Technical Innovations and Notes
IMPORTANCE OF PRECISE POSITIONING FOR PROTON BEAM THERAPY IN THE BASE OF SKULL AND CERVICAL SPINE HIDEO TATSUZAKI,” M.D.,
PH.D. AND MARCIA M. URIE, PH.D.
Department of Radiation Medicine, MassachusettsGeneral Hospital and Harvard Medical School, Boston, MA 02114, USA Using proton beam therapy, high doses have been delivered to chordomas and chondrosarcomas of the base of skull and cervical spine. Dose inhomogeneity to the tumors has been accepted in order to maintain normal tissue tolerances, and detailed attention to patient immobiliiation and to precise positioning has minhnixed the margins necessary to ensure these dose constraints. This study examined the contribution of precise positioning to the better dose localization achieved in these treatments. Three patients whose tumors represeuted diiferent anatomic geometries were studied. Treatment plans were developed which treated as much of the tumor as possible to 74 Cobalt-Gray-Equivalent (CGE) while maintaining the central brain stem and central spinal cord at 5 48 CGE, the surface of the brain stem, surface of the spinal cord, and optic structures at i 60 CGE, and the temporal lobes at 5 5% likelihood of complication using a biophysical model of normal tissue complication probability. Two positioning accuracies were assumed: 3 mm and 10 mm. Both proton beam plans and 10 MV X ray beam plans were developed with these assumptions and dose constraints. In all cases with the same positioning uncertainties, the proton beam plans delivered more dose to a larger percentage of the tumor volume and the estimated tumor control probability was higher than with the X ray plans. However, without precise positioning both the proton plans and the X ray plans deteriorated, with a 12% to 25% decrease in estimated tumor control probability. In all but one case, the difference between protons with good positioning and poor positioning was greater than the difference between protons and X rays, both with good positioning. Hence in treating these tumors, which are in close proximity to critical normal tissues, attention to i~obili&ion and precise positioning is essential. With good positioning, proton beam therapy permits higher doses to significantly more of the tumor in these sites than do X rays. Proton therapy, Comparative treatment planning, Precise positioning, Patient immobilization, complication probability models, Tumor control probability models.
Normal tissue
distribution. Because protons have a finite range of penetration, distal tissues receive no dose. With X rays, all tissues along the beam path receive some dose because the photons are attenuated only near-exponentially. The proton beam treatments in our program at the Massachusetts General Hospital (MGH) and Harvard Cyclotron Laboratory (HCL) have emphasized minimizing the volume of normal tissue irradiated. Margins added for patient motion and positioning uncertainties have been minimized by paying careful attention to immobilization of the patient, to obtaining the treatment planning CT with the patient in the immobilization device and in the treatment position, and to confirming correct positioning with co-axial diagnostic radiographs before each treatment. This is particularly important to the proton beam therapy because of our treatment philosophy: as much of the tumor volume as possible is treated to the prescribed dose while certain critical struc-
INTRODUCTION
Proton beam therapy has the potential to deliver dose distributions that are superior to those of high energy X rays. Experience has demonstrated that higher doses to tumor volumes can be achieved while maintaining acceptable normal tissue reactions. For example, patients treated for chordomas and chondrosarcomas of the base of skull and cervical spine have received an average prescribed dose of 69.0 CGE* (2), while conventional treatment has limited the dose to 50-55 Gy because of the proximity of these tumors to critical structures such as the brain stem, spinal cord, and optic structures. These higher doses have resulted in actuarial local control rates at 5 years of 82% (2) as compared to 40% at 3 years for conventional therapy (1). The obvious advantage of proton beams over X ray beams is that they produce an intrinsically superior dose *Present address: Institute of Clinical Medicine, University of Tsukuba, Tsukuba, Ibaraki 305, Japan. Reprint requests to: Marcia M. Uric, Ph.D. Supported in part by Grant CA 21239 from the National Cancer Institute, DHHS. Accepted for publication 22 March 1991.
*Proton doses are quoted in cobalt-Gray-equivalent (CGE) which is the physical dose times the relative biological effectiveness (RBE) of 1.10 (14). Since all tissues are thought to have the very similar RBE’s, X ray doses in Gy can be directly compared to proton doses in CGE. 751
I. J. Radiation Oncology 0 Biology 0 Physics
758
August 1991, Volume 2 1, Number 3
Table 1. Constraintsto normal tissues used to develop treatment plans which delivered 48 CGE to the initial target volume (TV1) and 74 CGE to as much of the boost target volume (TV-2) while maintainingnormal tissues at or below these doses Normal tissue
Constraint
Optic chiasm Optic nerves
60 CGE 60 CGE
Brain stem-Spinal cord Central Surface
48 CGE 60 CGE
Temporal lobes A. >1 cm from TV-2 51 cm from TV-2 B. NTCP
60 CGE no limit 5%
tures are maintained at or below specified tolerance doses. When these constraints cause small portions of the tumor volume adjacent to these normal tissues to receive less than the prescribed dose, the dose inhomogeneity to the tumor is accepted. The larger the margins added for uncertainties in order to ensure these normal tissue doses, the larger the volume of tumor that does not receive the prescribed dose. The purpose of this study was to examine how important precise positioning is to the dose localization achieved in these treatments. The technique of the study was to develop the “best” 3-dimensional treatment plans possible with either protons or X rays with one of two positioning accuracy assumptions: 3 mm and 10 mm. All four plans were compared; dose-volume histograms and a model of tumor control probability (TCP) were used to assess the differences among them. By comparing “well-positioned” proton beam plans to “poorly positioned” proton beam plans and to “well positioned” X ray beam plans, the role of precise positioning in achieving dose localization was assessed. METHODS AND MATERIALS Three patients with chordomas, two of the base of skull and one of the cervical spine, who had been previously treated with protons were selected for this study because their tumors represent three different anatomic geometries. Patient N is a g-year-old boy with a large (128 ml) chordoma located midline in the upper clival area. He had had a subtotal resection through a temporal craniotomy, and the remaining gross disease extended from the top of the orbits to the first cervical vertebra. Patient V is a 60-yearold female with a small (18 ml) chondroid chordoma recurrent in the right parasellar-middle fossa region 4 years after surgery (no postoperative radiation therapy). Patient G is a 26-year-old female with a large (164 cc) cervical chordoma which wrapped 220“ around (lateral and anterior to) the spinal cord and extended from the petrous bone through the third cervical vertebra.
Table 2. Final portal arrangementsfor the plans Proton beams
1OMV X ray beams
Pt N Initial target
RL, LL,
RL + LL
TV-2 Reduced TV-2
RS(35”), LS(35”) RL, LL RL. LL
RI_ + LL RL + LL
Pt v
Initial target
RL. LL
TV-2 Reduced TV-2
RP(20”) RS(IS”)
Pt G Initial target TV-2
RL, LL RL + PA
Reduced TV-2
RL + PA
RP(20”) + LA(20”), LA(30”) + RP(30”) RP(30”) + LA(30”) RP(4O”)S(30”) + LA(40”)1(30”) RL + LL RA(U”,wdg) + LA(45”,wdg) + PA(cord blk) RA(45”,wdg)+ LA(4S’,wdg) + PA(cord blk)
Note: + indicates treated in combination; indicates treated independently. Initial target: portal for treatment to 48 CGE (reduced at 45 CGE). TV-2: portals for treatment from 48 to 58 CGE. Reduced TV-2: portals for treatment from 58 to 74 CGE (often with field reductions for temporal lobe dose constraints). RL = right lateral; LL = left lateral; P = posterior; A = anterior; S = superior.
Each patient had a CT scan in his or her immobilization device in the treatment position; 3 mm thick sections were obtained at 3 mm intervals throughout the target volumes and several centimeters superiorly and inferiorly. Using the treatment planning system developed by Goitein et al. at the MGH (8,9), target volumes and normal tissues of interest were delineated on the CT scans. For each patient, two target volumes were defined: TV-l, the initial target volume, was to receive 48 CGE and included regions of possible microscopic extension. The second, the boost target volume (TV-2), was the region of gross disease to which the aim was to deliver 74 CGE. The extent of the TV-2 and its relationship to normal tissues for each patient are indicated in Figure 1. For all three patients, “best” plans were developed with proton beams and with 10 MV X ray beams with different assumptions. The goal was defined as the delivery of 74 CGE to as much of TV-2 as possible while maintaining the critical normal tissues listed in Table 1 at or below their dose constraints. Plans were developed by experienced planners with many iterations of beam design and weights. The one judged to meet the goals best was selected by the planners; no computer-assisted or automated optimization
Importance of precise positioning in proton therapy 0
H. TATSUZAKIAND M.
M.
URIE
Fig. 1. Sample transverse CT section (left panels) and reconstructed sagital section (central panels) of each patient with the TV-2 and normal tissues for which there were dose constraints delineated. The right panel indicates the relationship of the TV-2 to the brain stem/spinal cord, optic structures, and temporal lobes from a viewpoint at the top of the head looking toward the feet. The top row is Patient N, the middle row is patient V, and the bottom row Patient G.
was used. Ten MV X rays** were selected for the X ray plans because their penetration is suitable for these tumors located deep in the head and because their penumbra is similar to that of protons from the HCL cyclotron (80% to 20% dose decrement distance -6mm). Thus portals for both modalities were equal in size. For each modality, two positioning uncertainties were assumed: 10 mm, representing minimal patient immobilization without pre-treatment position confirmation and referred to here as poor position**Varian USA.
Clinac
18, Varian
Associates,
Palo Alto,
CA,
ing, and 3 mm, representing rigid immobilization and position confirmation prior to each treatment and referred to here as good positioning (13,17). These positioning uncertainties were chosen to represent the extremes of conventional treatment techniques to help distinguish between the effects of positioning and the effects of the type of radiation. Ten mm is larger than is representative of most treatments in many departments, but not unreasonable. For example, in Rabinowitz et al.‘s study (13), discrepancies
760
I. J. Radiation Oncology 0 Biology 0 Physics
August 1991, Volume 21, Number 3
Fig. 2. Dose distributions of the “well positioned” treatment plans for each patient superimposed on a representative CT section. The temporal lobe dose constraint is that the NTCP be 5 5%. The TV-2s and the normal tissues of interest are in black, The top row is Patient N, the middle row is patient V, and the bottom row Patient G. The proton beam plans are on the left and the X ray beam plans on the right; doses are in CGE for protons, Gy for X rays.
between the simulator film and treatment portal were 210 mm in -20% of the patients. These positioning uncertainties, 3 mm and 10 mm, were the margins added to the portal around each target volume, except when sparing a dose constraining normal tissue. In these cases, the 3 or 10 mm margins were the margins added to ensure the desired sparing. For example, if a portal was designed to deliver
no more than 50% of the tumor dose to the brain stem surface with the 3 mm positioning assumption, the field edge (50% dose) was placed 3 mm away from the surface, and with the 10 mm assumption, it was 10 mm away. For the proton portals, the exact calculations of the compensating boluses were modified (expanded) to assure that the target volumes would be treated as long as the patient was aligned
Importance of precise positioning in proton therapy 0 H. Table 3. Estimated tumor control probability (in percentage) of the proton plans and X ray plans for positioning uncertainties of 3 mm and 10 mm with two alternate temporal lobe dose constraints
Tumor control probability (%) PtN PtV
TATSUZAKI
Pt N. \-j _____............... x3
100 -
PtG
TL* limits: 5 60 CGE more than 1 cm from tumor Positioning X rays Protons
1Omm 43 53
3mm 62 67
3mm 57 69
1Omm 45 52
3mm 66 71
Lobe limit
125 -
Large wrap-around cervical spine
Small temporal
Large midline upper clivus
-
?5-
s 3 0,
50-
Positioning
3mm
1Omm
67 79
53 67
3mm 1Omm 72 51 77 65
‘,
-e--m
x10
_.m,_._._.
PlO
P3
0 0
1Omm 45 46
10
20
30
40
50
60
70
80
90
(CGE)
Pt v. 20 ... ... ... ... ... ... ..
x3
.
_-___
P3
.
_._._._._.
PlO
16-
*Temporal lobes. G 0,
,2-
w
within either the 3 mm or the 10 mm assumption (15). No differences in the ability to predict the range of the protons and the absorption of the X rays from the CT values were assumed. All four plans were developed independently; they did not necessarily use the same beam orientations and weights. For both modalities and both positioning accuracy unconventional portals and non-coplanar assumptions, beams were explored. The goal was to deliver 48 CGE to TV-l and 74 CGE to as much of the TV-2 as possible while maintaining the tolerances of the specified normal tissues. The philosophy of the portal design was to deliver to the normal tissues nearly their tolerance doses (to within 1 to 2 CGE) and then to spare them for the remaining portals. In practice, this meant several field reductions: a) a planned field reduction after 45 CGE to TV-l ; b) after 47 CGE to the TV-2, in order to spare the centers of the brain stem and spinal cord; c) after 58 CGE, in order to spare the optic structures and the surfaces of the brain stem and spinal cord; and d) to spare temporal lobes. Two alternative dose constraints were used for the temporal lobes. One limited any portion of the temporal lobes further than 1 cm from the TV-2 to 5 60 CGE; there was no limit on the dose the temporal lobe within 1 cm of TV-2 could receive. (For brevity, this is referred to as the ~60 CGE limit.) An absolute dose constraint makes treatment planning straightforward, although admittedly this is a relatively arbitrary constraint and is prejudiced toward our confidence in predicting and controlling the end of range of the proton beams. To eliminate this bias, the plans were developed using an alternative constraint based on an estimate of normal tissue complication probability (NTCP) to
b “\
25 -
1Omm 45 46
DOSE
3mm 69 72
1 ’4 i
\
TL* limits: NTCP I 5% X rays Protons
761
URIE
M. M.
NTCP Temporal
8
Tumor size and location
AND
Z
II 1 I i i
8-
$
i \A \ ’ I. \, y.
. 40 0
I
I
I
I
I
1
I
I
I
10
20
30
40
50
60
70
80
90
DOSE
(CGE)
Pt G. 175 -I
.................... . -
-_--_
X-J
P3
25
0
10
20
30
40 DOSE
50
60
70
80
90
(CGE)
Fig. 3. Dose-volume histograms for the TV-2 for each treatment plan developed for each patient with the temporal lobes constrained to no more than 5% NTCP.
limit the dose to the temporal lobes. For the NTCP estimate we used a model developed by Kutcher and Burman (lo), following the ideas of Lyman (11,12), which calculates the complication probability of inhomogeneously irradiated normal tissues. Parameters for the model were selected and adjusted by our clinicians? until they were ?John E. Munzenrider, M.D., Hideo Tatsuzaki, and Andrew P. Brown, MRCP, FRCR.
M.D.,
Ph.D.,
762
I. J. Radiation Oncology 0 Biology @ Physics
August 1991, Volume 2 1, Number 3
‘-
x3
-
P3
111-11..
x,0
.I.,.#.1.1.1.,. p,o
0
10
20
30
40
50
60
7b
DOSE (CGE) Fig. 4. Dose-volume
histograms
the temporal lobes for pat&t V.
for the right temooral
lobe for the treatment plans with a 5 5% NTCP constraint for
-
satisfied with the estimates of the relative risk to the temporal lobes. All plans, X rays and protons with both positioning assumptions, were then redeveloped with the constraint that the NTCP of the temporal lobes be I 5%. These are the results emphasized here. (Both constraints resulted in similar conclusions.) After assuring that the dose constraints of the normal tissues were met, doses to the TV-2s for all plans were quantitated in dose-volume histograms. (TV-l was adequately treated in all plans.) Various dose statistics were extracted to help analyze the differences. In addition, the histograms were used as input to a model of tumor control probability (TCP) for inhomogeneous irradiation developed by Goitein (7). The TCP’s predicted by this model (assuming a tumor control probability of 75% for a tumor uniformly irradiated to 70 CGE and a yso (slope of the dose response curve at 50% TCP as a ratio of percentage increase in TCP per 1% dose increment (3)) of 1.3 for the population) were then compared.
sitioning in both the proton plans and the X ray plans. These samples are included to give a flavor of the dose distributions achieved for each patient with the two modalities. The CT section is through the region of the temporal lobe for Pt N and Pt V and the dose distributions for all three patients are those limited by the constraint of I 5% NTCP for the temporal lobes. In Figure 3 are the dose-volume histograms for the TV-2 of each patient for each of the treatment plans developed with the temporal lobes limited to a complication likelihood of 5 5%, as predicted by the model. Results of the plans with the temporal lobe constrained to an absolute dose limit of 5 60 G CGE gave very similar results (as indicated in Table 3). These histograms were used as input to a model of TCP; the probability of local tumor control predicted by this model are presented in Table 3.
DISCUSSION RESULTS Eight treatment plans were developed for each of the three patients: proton beams for positioning uncertainties of 3 mm and 10 mm, and 10 MV X ray beams with 3 mm and 10 mm positioning uncertainties, each with two different limits on the doses to the temporal lobes. In all instances, the normal tissue constraints listed in Table 1 were met. The portal arrangements for the plans of each patient are outlined in Table 2. In general, the beam directions for a given modality were very similar with the different positioning accuracy assumptions. The different constraints to the temporal lobes did not alter the beam direction selection, only the size and relative weights of the portals. Dose distributions superimposed on representative transverse CT sections are shown in Figure 2 for good po-
In all instances, the proton beam plans delivered the prescribed dose to a larger percentage of the TV-2s than did the X ray beam plans with the same positioning accuracy assumptions and normal tissue constraints. Not only did a greater percentage of the TV-2 receive the prescribed dose, but the dose received by any given volume of the TV-2 was higher with the proton beam plans. In addition, the proton beam plans delivered the dose more uniformly, that is, the difference between the minimum and maximum TV-2 dose was less with protons. The improvements in TV-2 coverage with the proton plans depended upon the relative geometries of the tumor and critical normal tissues for which there are dose constraints. For Patient N, the TV-2, which is anterior to the brain stem, is treated with lateral portals with both protons and X rays. With good positioning, it is possible to deliver
Importance
of precise positioning
in proton therapy
270 CGE (95% of the prescribed dose) to 86% of the TV-2 with the proton beams. With the X ray beams only 46% of the TV-2 is treated to this dose (same temporal lobe limits). With poor positioning, the volume of TV-2 receiving 270 CGE is reduced significantly for both modalities, particularly in the proton plan with the NTCP temporal lobe constraint. In this geometry, the potential advantage of not treating the contra-lateral temporal lobe with lateral proton portals is lost when precise positioning is not assured. Patient V has a lateral tumor. For both temporal lobe dose constraints there was a stepwise improvement in TV-2 coverage progressing from X rays-poor positioning, to protons-poor positioning, to X rays-good positioning, to protons-good positioning. With good positioning, proton beams deliver ~70 CGE to 84% of the TV-2 with the NTCP temporal lobe limits. Poor positioning reduced this to 22%. These are approximately 20% higher than the comparable X ray beams plans. With this right-sided tumor, proton beams reduce the dose to the right temporal lobe significantly as compared to X rays, as shown in the dose-volume histograms in Figure 4. In this geometry, proton beams have a significant advantage over X ray beams but only if good positioning is assured. Patient G has a tumor that wraps 220’ around the spinal cord and does not extend superiorly enough for the temporal lobes to be dose constraining. The dose delivered to the TV-2 is driven by the positioning assumptions, as both modalities require multiple portals, each one sparing the spinal cord. For the proton beam plans, the TV-2 anterior to the spinal cord is treated with lateral fields; the TV-2 lateral to the spinal cord is treated with a posterior field in which the protons range out in the penumbra of the lateral field, a technique similar to that described by Castro et al. (4). Uncertainty in stopping the protons precisely in the penumbra of the lateral field results in small regions of double treatment within the TV-2, which accounts for the high dose tail in the dose-volume histogram (Fig. 3). The X ray plan for the TV-2 treatment consists of a pair of 45” anterior oblique portals, each with a 45” wedge and designed to spare the spinal cord (the right anterior portal treats to the left of the spinal cord, the left anterior oblique to the right). A posterior field with a cord block boosts the dose to the TV-2 lateral to the spinal cord. This field arrangement results in “hot spots” in the anterior portion of the TV-2 (and in the pharynx), accounting for the high dose tail in the dose volume histogram in the well positioned X ray plan (Fig. 3). With good positioning, proton beams deliver ~70 CGE to 75% of the TV-2; poor positioning reduces this percentage to 33%. In both positioning accuracy assumptions, approximately 10% more of the tumor received 70 CGE in the proton plan than in the X ray plan. Since the patient’s tumor is in the cervical spine, the temporal lobes were not dose limiting. Positioning precision was necessary to take advantage of the proton dose deposition characteristics and deliver higher doses to larger percentages of the TV-~‘S than the
??H.
TATSUZAKI AND M. M. URIE
763
X ray plans. Poorly positioned proton beam plans did not do any better, and in 5 of the 6 cases were actually worse, than the well-positioned X ray beam plans. The differences between the good and poor positioning plans were similar for both modalities. Regardless of radiation type, good positioning can allow significantly more dose to be delivered to these target volumes which are in close proximity to critical normal tissues. Precise localization and position confirmation can contribute to better radiation therapy for patients treated with conventional photons as well as with protons. In developing these plans, only the constraints enumerated in Table 1 were considered; no other normal tissues were considered to be dose limiting. Some plans developed with the X ray beams delivered high, in fact, unacceptably high, doses to other normal tissues but were accepted for this comparison. For instance, for Pt G the TV-2 treatment created a 30% “hot spot” which would have delivered more than 90 Gy to portions of the pharynx, thereby making the plan unacceptable clinically. With Pt N, the X ray plans delivered the target dose (74 Gy) to the parotid glands while the proton plans maintained 50% of their volume below 40 CGE. These X ray plans were judged significantly inferior to the proton beams plans, and by a considerably greater margin than the differences in the TCPs or the dose-volume histograms for the TV-2s would indicate. Considerations such as these are very important in assessing the overall merits of different plans but are difficult to include as explicit constraints, since clinicians will accept some complications, such as xerostomia, if they are necessary to deliver the needed dose to the tumor but would certainly chose to avoid them if possible. Three-dimensional treatment planning capability was essential to this study, not only because it allowed the 3 dimensional dose analysis, but also because portals were designed by geometric optimization in the beam’s eye view. For example, with Patient V, the oblique angles for both protons and X rays were selected to optimize TV-2 treatment while sparing the brain stem. For Patient G the conventional X ray plan would have been opposed lateral fields, stopping the dose at the tolerance of the spinal cord, 53 Gy. With the 3-dimensional system, a significant portion of the TV-2 was treated to significantly higher doses using the combination of an anterior 45” wedged pair and a posterior field with a spinal cord block. These types of field arrangements would not have been designed without the three dimensional capability. Very many X ray beams, each with its own field defining aperture, were considered while developing the X ray plans; in essence, conformal therapy was simulated. This too could not have been done without 3 dimensional planning. The proton beams for these plans assumed an uncertainty in the ability to predict and control the end of range of the protons and hence of how tightly the dose can be tailored to the target volume distally (6). The primary contributors to this uncertainty are the accuracy of the CT values (Hounsfield units), the spatial resolution of the CT
764
I. J. Radiation Oncology 0 Biology ??Physics
scans (particularly in the caudad-cephalad direction) (5), and the scattering effects of multiple fine heterogeneities (16). Because of these uncertainties, the prescribed range is typically 5% greater than the calculated maximum required range. In addition, the compensating bolus is designed not to contour exactly the distal end of range to the target volume, but rather to assure that the target will be treated as long as the patient and bolus are aligned within the positioning uncertainty (15). These factors contribute to intentionally treating more non-target tissue in the nominal situation represented by the computer plan. The uncertainties associated with the CT values were also considered for the X ray beams, but their effects on the dose distribution are much less than those for the proton beams. The predicted exponential absorption of the X rays is only very modestly affected by differences in CT values of the magnitudes appropriate here. The dose-volume histograms summarize the dose that the TV-2s receive. Since it is not known which parameter(s) of these histograms determines the likelihood of controlling the tumor, it is difficult to judge the importance of the differences among the histograms. Various statistics were extracted and could have been presented, such as minimum tumor dose, mean dose, volume of tumor receiving, for example, 95% of the prescribed dose. These single parameters are not indicative of the overall differences among the histograms. Therefore, we elected to present the histograms themselves and the predictions of a model of tumor control probability for inhomogeneously irradiated tumors (7). The model divides the tumor into subgroups of tumorlets that are small enough to be considered as receiving a uniform dose; it then uses dose response data derived from clinical and experimental data to predict the relative likelihood of tumor control of the inhomogeneously irradi-
August
1991, Volume 21, Number 3
ated target to what it would have been had the target been homogeneously irradiated to the prescribed dose. This model allows the differences among the plans to be reduced to a comparison of a single number, TCP, and may give an indication of the relative merits of the plans. In these treatment plans, those with good positioning assumptions had estimated TCPs 12-25% higher than those with poor positioning. The differences were similar for both modalities, protons and X rays. With good positioning assumptions, the estimated TCP’s of the proton beam plans were 5% to 12% higher than X ray beam plans. For these TCP’s, a shallow (1.3) yso was used, which was estimated from sparse clinical data. Were the true gamma factor to be larger than this, the differences among the estimated TCPs would be larger. The adequacy of this model of TCP is yet to be proven: even if its concepts are correct, the differences in the TCPs predicted are very dependent upon the assumed slope of the dose-response curve (yso parameter). A comparison of the isodose contours and of the dose-volume histograms is certainly essential to assess the differences among plans, including those presented here. CONCLUSIONS Precise positioning of patients is essential if the dose deposition characteristics of protons are to be used to full advantage to treat tumors adjacent to critical normal tissues. With rigid immobilization and confirmation of correct position prior to each treatment, significantly more dose can be delivered to larger percentages of the tumors than with X rays. However, without good positioning, the proton beam treatments deteriorate and may not be as good as conventional radiation therapy with good positioning.
REFERENCES 1. Austin-Seymour;M.; Munzenrider, J.; Goitein, M.; Gentry, R.; Gragoudas, E.; Koehler, A.; McNulty, P.; Osborne, E.;
2.
3. 4.
5.
6.
Ryugo, D.; Seddon J.; Urie, M.; Verhey, L.; Suit, H. Progress in low-LET heavy particle therapy: intracranial and paracranial tumors and uveal melanomas. Rad. Res. 104(2): S219-S226; 1985. Austin-Seymour, M.; Munzenrider, J.; Goitein, M.; Verhey, L.; Urie, M.; Gentry; R.; Bimbaum, S.; Ruotolo, D.; McMannus, P.; Skates, S.; Ojemann, R.; Rosenberg, A.; Schiller, A.; Koehler, A.; Suit, H. Fractionated proton radiation therapy for chordomas and low grade chondrosarcomas of the base of skull. J. Neurosurgery 70:13-17; 1989. Brahme, A. Dosimetric precision requirements in radiation therapy. Acta Radiol. Oncol. 23 (Fasc.5):379-391; 1984. Castro, J.; Collier, J.; Petti, P.; Nowakowski, V.; Chen G., Lyman, J.; Lindstadt, D.; Gauger, G.; Gutin, P.; Decker, M.; Phillips, T.; Baken, K. Charged particle radiotherapy for lesions encircling the brain stem or spinal cord. Int. J. Radiat. Oncol. Biol. Phys. 17:477-484; 1989. Goitein, M. Compensation for inhomogeneities in charged particle radiotherapy using computed tomography. Int. J. Radiat. Oncol. Biol. Phys. 4:499-508; 1978. Goitein, M. Calculation of the uncertainty in dose delivered
during radiation therapy. Med. Phys. 12:608-612; 1985. 7. Goitein, M Tumor control probability for an inhomogeneously irradiated target volume. NC1 report, section 5.8, S. Zink, project officer; 1987. 8. Goitein, M.; Abrams, M. Multi-dimensional treatment planning: I. Delineation of anatomy. Int. J. Radiat. Oncol. Biol. Phys. 9:777-788; 1983. 9. Goitein, M.; Abrams, M.; Rowell, D.; Pollari, H.; Wiles, J. Multi-dimensional treatment planning: II. Beam’s eye view, back projection and projection through CT sections. Int. J. Radiat. Oncol. Biol. Phys. 9:789-797; 1983. 10. Kutcher, G.; Burman, C. Calculations of complication probability factors for non-uniform normal tissue irradiation: the effective volume method. Int. J. Radiat. Oncol. Biol. Phys. 16:1623-1630; 1989. 11. Lyman, J. Complication probability as assessed from dosevolume histograms. Rad Res. 104:S13-S19; 1985. 12. Lyman, J.; Wolbarst, A. Optimization of radiation therapy III: a method of assessing complication probabilities from dose-volume histograms. Int. J. Radiat. Oncol. Biol. Phys. 13(1):103-109; 1987. 13. Rabinowitz, I.; Broomberg, J.; Goitein, M.; McCarthy, K.; Leong, J. Accuracy of Radiation Field Alignment in Clinical
Importance of precise positioning in proton therapy 0 H. Practice. Int. J. Radiat. One. Biol. Phys. 11(10):1857-1867; 1985. 14. Urano, M.; Verhey, L.; Goitein, M.; Tepper, J.; Suit, H.; Phil, D.; Mendiondo, 0.; Gragoudas, E.; Koehler, A. Relative biological effectiveness of modulated proton beams in various murine tissues. Int. J. Radiat. Oncol. Biol. Phys. 10(4):509-514; 1984. 15. Urie, M.; Goitein, M.; Wagner, M. Compensating for hetero-
TATSUZAKIAND M.
M. URIE
765
geneities in proton radiation therapy. Phys. Med. Biol. 29: 553-566; 1984. 16. Urie, M.; Goitein, M.; Holley , W.; Chen, G. Degradation of the Bragg peak due to inhomogeneities. Phys. Med. Bio. 31(1):1-15; 1986. 17. Verhey, L.; Guotein, M.; McNulty, P.; Munzenrider, J.; Suit, H. Precise positioning of patients for radiation therapy. Int. J. Radiat. Oncol. Biol. Phys. 8:289-294, 1982.
Copyright
0360-3016191 $3.00 + .@I 0 1991 Pergamon Press plc
??Technical Innovations and Notes
IMPORTANCE OF PRECISE POSITIONING FOR PROTON BEAM THERAPY IN THE BASE OF SKULL AND CERVICAL SPINE HIDEO TATSUZAKI,” M.D.,
PH.D. AND MARCIA M. URIE, PH.D.
Department of Radiation Medicine, MassachusettsGeneral Hospital and Harvard Medical School, Boston, MA 02114, USA Using proton beam therapy, high doses have been delivered to chordomas and chondrosarcomas of the base of skull and cervical spine. Dose inhomogeneity to the tumors has been accepted in order to maintain normal tissue tolerances, and detailed attention to patient immobiliiation and to precise positioning has minhnixed the margins necessary to ensure these dose constraints. This study examined the contribution of precise positioning to the better dose localization achieved in these treatments. Three patients whose tumors represeuted diiferent anatomic geometries were studied. Treatment plans were developed which treated as much of the tumor as possible to 74 Cobalt-Gray-Equivalent (CGE) while maintaining the central brain stem and central spinal cord at 5 48 CGE, the surface of the brain stem, surface of the spinal cord, and optic structures at i 60 CGE, and the temporal lobes at 5 5% likelihood of complication using a biophysical model of normal tissue complication probability. Two positioning accuracies were assumed: 3 mm and 10 mm. Both proton beam plans and 10 MV X ray beam plans were developed with these assumptions and dose constraints. In all cases with the same positioning uncertainties, the proton beam plans delivered more dose to a larger percentage of the tumor volume and the estimated tumor control probability was higher than with the X ray plans. However, without precise positioning both the proton plans and the X ray plans deteriorated, with a 12% to 25% decrease in estimated tumor control probability. In all but one case, the difference between protons with good positioning and poor positioning was greater than the difference between protons and X rays, both with good positioning. Hence in treating these tumors, which are in close proximity to critical normal tissues, attention to i~obili&ion and precise positioning is essential. With good positioning, proton beam therapy permits higher doses to significantly more of the tumor in these sites than do X rays. Proton therapy, Comparative treatment planning, Precise positioning, Patient immobilization, complication probability models, Tumor control probability models.
Normal tissue
distribution. Because protons have a finite range of penetration, distal tissues receive no dose. With X rays, all tissues along the beam path receive some dose because the photons are attenuated only near-exponentially. The proton beam treatments in our program at the Massachusetts General Hospital (MGH) and Harvard Cyclotron Laboratory (HCL) have emphasized minimizing the volume of normal tissue irradiated. Margins added for patient motion and positioning uncertainties have been minimized by paying careful attention to immobilization of the patient, to obtaining the treatment planning CT with the patient in the immobilization device and in the treatment position, and to confirming correct positioning with co-axial diagnostic radiographs before each treatment. This is particularly important to the proton beam therapy because of our treatment philosophy: as much of the tumor volume as possible is treated to the prescribed dose while certain critical struc-
INTRODUCTION
Proton beam therapy has the potential to deliver dose distributions that are superior to those of high energy X rays. Experience has demonstrated that higher doses to tumor volumes can be achieved while maintaining acceptable normal tissue reactions. For example, patients treated for chordomas and chondrosarcomas of the base of skull and cervical spine have received an average prescribed dose of 69.0 CGE* (2), while conventional treatment has limited the dose to 50-55 Gy because of the proximity of these tumors to critical structures such as the brain stem, spinal cord, and optic structures. These higher doses have resulted in actuarial local control rates at 5 years of 82% (2) as compared to 40% at 3 years for conventional therapy (1). The obvious advantage of proton beams over X ray beams is that they produce an intrinsically superior dose *Present address: Institute of Clinical Medicine, University of Tsukuba, Tsukuba, Ibaraki 305, Japan. Reprint requests to: Marcia M. Uric, Ph.D. Supported in part by Grant CA 21239 from the National Cancer Institute, DHHS. Accepted for publication 22 March 1991.
*Proton doses are quoted in cobalt-Gray-equivalent (CGE) which is the physical dose times the relative biological effectiveness (RBE) of 1.10 (14). Since all tissues are thought to have the very similar RBE’s, X ray doses in Gy can be directly compared to proton doses in CGE. 751
I. J. Radiation Oncology 0 Biology 0 Physics
758
August 1991, Volume 2 1, Number 3
Table 1. Constraintsto normal tissues used to develop treatment plans which delivered 48 CGE to the initial target volume (TV1) and 74 CGE to as much of the boost target volume (TV-2) while maintainingnormal tissues at or below these doses Normal tissue
Constraint
Optic chiasm Optic nerves
60 CGE 60 CGE
Brain stem-Spinal cord Central Surface
48 CGE 60 CGE
Temporal lobes A. >1 cm from TV-2 51 cm from TV-2 B. NTCP
60 CGE no limit 5%
tures are maintained at or below specified tolerance doses. When these constraints cause small portions of the tumor volume adjacent to these normal tissues to receive less than the prescribed dose, the dose inhomogeneity to the tumor is accepted. The larger the margins added for uncertainties in order to ensure these normal tissue doses, the larger the volume of tumor that does not receive the prescribed dose. The purpose of this study was to examine how important precise positioning is to the dose localization achieved in these treatments. The technique of the study was to develop the “best” 3-dimensional treatment plans possible with either protons or X rays with one of two positioning accuracy assumptions: 3 mm and 10 mm. All four plans were compared; dose-volume histograms and a model of tumor control probability (TCP) were used to assess the differences among them. By comparing “well-positioned” proton beam plans to “poorly positioned” proton beam plans and to “well positioned” X ray beam plans, the role of precise positioning in achieving dose localization was assessed. METHODS AND MATERIALS Three patients with chordomas, two of the base of skull and one of the cervical spine, who had been previously treated with protons were selected for this study because their tumors represent three different anatomic geometries. Patient N is a g-year-old boy with a large (128 ml) chordoma located midline in the upper clival area. He had had a subtotal resection through a temporal craniotomy, and the remaining gross disease extended from the top of the orbits to the first cervical vertebra. Patient V is a 60-yearold female with a small (18 ml) chondroid chordoma recurrent in the right parasellar-middle fossa region 4 years after surgery (no postoperative radiation therapy). Patient G is a 26-year-old female with a large (164 cc) cervical chordoma which wrapped 220“ around (lateral and anterior to) the spinal cord and extended from the petrous bone through the third cervical vertebra.
Table 2. Final portal arrangementsfor the plans Proton beams
1OMV X ray beams
Pt N Initial target
RL, LL,
RL + LL
TV-2 Reduced TV-2
RS(35”), LS(35”) RL, LL RL. LL
RI_ + LL RL + LL
Pt v
Initial target
RL. LL
TV-2 Reduced TV-2
RP(20”) RS(IS”)
Pt G Initial target TV-2
RL, LL RL + PA
Reduced TV-2
RL + PA
RP(20”) + LA(20”), LA(30”) + RP(30”) RP(30”) + LA(30”) RP(4O”)S(30”) + LA(40”)1(30”) RL + LL RA(U”,wdg) + LA(45”,wdg) + PA(cord blk) RA(45”,wdg)+ LA(4S’,wdg) + PA(cord blk)
Note: + indicates treated in combination; indicates treated independently. Initial target: portal for treatment to 48 CGE (reduced at 45 CGE). TV-2: portals for treatment from 48 to 58 CGE. Reduced TV-2: portals for treatment from 58 to 74 CGE (often with field reductions for temporal lobe dose constraints). RL = right lateral; LL = left lateral; P = posterior; A = anterior; S = superior.
Each patient had a CT scan in his or her immobilization device in the treatment position; 3 mm thick sections were obtained at 3 mm intervals throughout the target volumes and several centimeters superiorly and inferiorly. Using the treatment planning system developed by Goitein et al. at the MGH (8,9), target volumes and normal tissues of interest were delineated on the CT scans. For each patient, two target volumes were defined: TV-l, the initial target volume, was to receive 48 CGE and included regions of possible microscopic extension. The second, the boost target volume (TV-2), was the region of gross disease to which the aim was to deliver 74 CGE. The extent of the TV-2 and its relationship to normal tissues for each patient are indicated in Figure 1. For all three patients, “best” plans were developed with proton beams and with 10 MV X ray beams with different assumptions. The goal was defined as the delivery of 74 CGE to as much of TV-2 as possible while maintaining the critical normal tissues listed in Table 1 at or below their dose constraints. Plans were developed by experienced planners with many iterations of beam design and weights. The one judged to meet the goals best was selected by the planners; no computer-assisted or automated optimization
Importance of precise positioning in proton therapy 0
H. TATSUZAKIAND M.
M.
URIE
Fig. 1. Sample transverse CT section (left panels) and reconstructed sagital section (central panels) of each patient with the TV-2 and normal tissues for which there were dose constraints delineated. The right panel indicates the relationship of the TV-2 to the brain stem/spinal cord, optic structures, and temporal lobes from a viewpoint at the top of the head looking toward the feet. The top row is Patient N, the middle row is patient V, and the bottom row Patient G.
was used. Ten MV X rays** were selected for the X ray plans because their penetration is suitable for these tumors located deep in the head and because their penumbra is similar to that of protons from the HCL cyclotron (80% to 20% dose decrement distance -6mm). Thus portals for both modalities were equal in size. For each modality, two positioning uncertainties were assumed: 10 mm, representing minimal patient immobilization without pre-treatment position confirmation and referred to here as poor position**Varian USA.
Clinac
18, Varian
Associates,
Palo Alto,
CA,
ing, and 3 mm, representing rigid immobilization and position confirmation prior to each treatment and referred to here as good positioning (13,17). These positioning uncertainties were chosen to represent the extremes of conventional treatment techniques to help distinguish between the effects of positioning and the effects of the type of radiation. Ten mm is larger than is representative of most treatments in many departments, but not unreasonable. For example, in Rabinowitz et al.‘s study (13), discrepancies
760
I. J. Radiation Oncology 0 Biology 0 Physics
August 1991, Volume 21, Number 3
Fig. 2. Dose distributions of the “well positioned” treatment plans for each patient superimposed on a representative CT section. The temporal lobe dose constraint is that the NTCP be 5 5%. The TV-2s and the normal tissues of interest are in black, The top row is Patient N, the middle row is patient V, and the bottom row Patient G. The proton beam plans are on the left and the X ray beam plans on the right; doses are in CGE for protons, Gy for X rays.
between the simulator film and treatment portal were 210 mm in -20% of the patients. These positioning uncertainties, 3 mm and 10 mm, were the margins added to the portal around each target volume, except when sparing a dose constraining normal tissue. In these cases, the 3 or 10 mm margins were the margins added to ensure the desired sparing. For example, if a portal was designed to deliver
no more than 50% of the tumor dose to the brain stem surface with the 3 mm positioning assumption, the field edge (50% dose) was placed 3 mm away from the surface, and with the 10 mm assumption, it was 10 mm away. For the proton portals, the exact calculations of the compensating boluses were modified (expanded) to assure that the target volumes would be treated as long as the patient was aligned
Importance of precise positioning in proton therapy 0 H. Table 3. Estimated tumor control probability (in percentage) of the proton plans and X ray plans for positioning uncertainties of 3 mm and 10 mm with two alternate temporal lobe dose constraints
Tumor control probability (%) PtN PtV
TATSUZAKI
Pt N. \-j _____............... x3
100 -
PtG
TL* limits: 5 60 CGE more than 1 cm from tumor Positioning X rays Protons
1Omm 43 53
3mm 62 67
3mm 57 69
1Omm 45 52
3mm 66 71
Lobe limit
125 -
Large wrap-around cervical spine
Small temporal
Large midline upper clivus
-
?5-
s 3 0,
50-
Positioning
3mm
1Omm
67 79
53 67
3mm 1Omm 72 51 77 65
‘,
-e--m
x10
_.m,_._._.
PlO
P3
0 0
1Omm 45 46
10
20
30
40
50
60
70
80
90
(CGE)
Pt v. 20 ... ... ... ... ... ... ..
x3
.
_-___
P3
.
_._._._._.
PlO
16-
*Temporal lobes. G 0,
,2-
w
within either the 3 mm or the 10 mm assumption (15). No differences in the ability to predict the range of the protons and the absorption of the X rays from the CT values were assumed. All four plans were developed independently; they did not necessarily use the same beam orientations and weights. For both modalities and both positioning accuracy unconventional portals and non-coplanar assumptions, beams were explored. The goal was to deliver 48 CGE to TV-l and 74 CGE to as much of the TV-2 as possible while maintaining the tolerances of the specified normal tissues. The philosophy of the portal design was to deliver to the normal tissues nearly their tolerance doses (to within 1 to 2 CGE) and then to spare them for the remaining portals. In practice, this meant several field reductions: a) a planned field reduction after 45 CGE to TV-l ; b) after 47 CGE to the TV-2, in order to spare the centers of the brain stem and spinal cord; c) after 58 CGE, in order to spare the optic structures and the surfaces of the brain stem and spinal cord; and d) to spare temporal lobes. Two alternative dose constraints were used for the temporal lobes. One limited any portion of the temporal lobes further than 1 cm from the TV-2 to 5 60 CGE; there was no limit on the dose the temporal lobe within 1 cm of TV-2 could receive. (For brevity, this is referred to as the ~60 CGE limit.) An absolute dose constraint makes treatment planning straightforward, although admittedly this is a relatively arbitrary constraint and is prejudiced toward our confidence in predicting and controlling the end of range of the proton beams. To eliminate this bias, the plans were developed using an alternative constraint based on an estimate of normal tissue complication probability (NTCP) to
b “\
25 -
1Omm 45 46
DOSE
3mm 69 72
1 ’4 i
\
TL* limits: NTCP I 5% X rays Protons
761
URIE
M. M.
NTCP Temporal
8
Tumor size and location
AND
Z
II 1 I i i
8-
$
i \A \ ’ I. \, y.
. 40 0
I
I
I
I
I
1
I
I
I
10
20
30
40
50
60
70
80
90
DOSE
(CGE)
Pt G. 175 -I
.................... . -
-_--_
X-J
P3
25
0
10
20
30
40 DOSE
50
60
70
80
90
(CGE)
Fig. 3. Dose-volume histograms for the TV-2 for each treatment plan developed for each patient with the temporal lobes constrained to no more than 5% NTCP.
limit the dose to the temporal lobes. For the NTCP estimate we used a model developed by Kutcher and Burman (lo), following the ideas of Lyman (11,12), which calculates the complication probability of inhomogeneously irradiated normal tissues. Parameters for the model were selected and adjusted by our clinicians? until they were ?John E. Munzenrider, M.D., Hideo Tatsuzaki, and Andrew P. Brown, MRCP, FRCR.
M.D.,
Ph.D.,
762
I. J. Radiation Oncology 0 Biology @ Physics
August 1991, Volume 2 1, Number 3
‘-
x3
-
P3
111-11..
x,0
.I.,.#.1.1.1.,. p,o
0
10
20
30
40
50
60
7b
DOSE (CGE) Fig. 4. Dose-volume
histograms
the temporal lobes for pat&t V.
for the right temooral
lobe for the treatment plans with a 5 5% NTCP constraint for
-
satisfied with the estimates of the relative risk to the temporal lobes. All plans, X rays and protons with both positioning assumptions, were then redeveloped with the constraint that the NTCP of the temporal lobes be I 5%. These are the results emphasized here. (Both constraints resulted in similar conclusions.) After assuring that the dose constraints of the normal tissues were met, doses to the TV-2s for all plans were quantitated in dose-volume histograms. (TV-l was adequately treated in all plans.) Various dose statistics were extracted to help analyze the differences. In addition, the histograms were used as input to a model of tumor control probability (TCP) for inhomogeneous irradiation developed by Goitein (7). The TCP’s predicted by this model (assuming a tumor control probability of 75% for a tumor uniformly irradiated to 70 CGE and a yso (slope of the dose response curve at 50% TCP as a ratio of percentage increase in TCP per 1% dose increment (3)) of 1.3 for the population) were then compared.
sitioning in both the proton plans and the X ray plans. These samples are included to give a flavor of the dose distributions achieved for each patient with the two modalities. The CT section is through the region of the temporal lobe for Pt N and Pt V and the dose distributions for all three patients are those limited by the constraint of I 5% NTCP for the temporal lobes. In Figure 3 are the dose-volume histograms for the TV-2 of each patient for each of the treatment plans developed with the temporal lobes limited to a complication likelihood of 5 5%, as predicted by the model. Results of the plans with the temporal lobe constrained to an absolute dose limit of 5 60 G CGE gave very similar results (as indicated in Table 3). These histograms were used as input to a model of TCP; the probability of local tumor control predicted by this model are presented in Table 3.
DISCUSSION RESULTS Eight treatment plans were developed for each of the three patients: proton beams for positioning uncertainties of 3 mm and 10 mm, and 10 MV X ray beams with 3 mm and 10 mm positioning uncertainties, each with two different limits on the doses to the temporal lobes. In all instances, the normal tissue constraints listed in Table 1 were met. The portal arrangements for the plans of each patient are outlined in Table 2. In general, the beam directions for a given modality were very similar with the different positioning accuracy assumptions. The different constraints to the temporal lobes did not alter the beam direction selection, only the size and relative weights of the portals. Dose distributions superimposed on representative transverse CT sections are shown in Figure 2 for good po-
In all instances, the proton beam plans delivered the prescribed dose to a larger percentage of the TV-2s than did the X ray beam plans with the same positioning accuracy assumptions and normal tissue constraints. Not only did a greater percentage of the TV-2 receive the prescribed dose, but the dose received by any given volume of the TV-2 was higher with the proton beam plans. In addition, the proton beam plans delivered the dose more uniformly, that is, the difference between the minimum and maximum TV-2 dose was less with protons. The improvements in TV-2 coverage with the proton plans depended upon the relative geometries of the tumor and critical normal tissues for which there are dose constraints. For Patient N, the TV-2, which is anterior to the brain stem, is treated with lateral portals with both protons and X rays. With good positioning, it is possible to deliver
Importance
of precise positioning
in proton therapy
270 CGE (95% of the prescribed dose) to 86% of the TV-2 with the proton beams. With the X ray beams only 46% of the TV-2 is treated to this dose (same temporal lobe limits). With poor positioning, the volume of TV-2 receiving 270 CGE is reduced significantly for both modalities, particularly in the proton plan with the NTCP temporal lobe constraint. In this geometry, the potential advantage of not treating the contra-lateral temporal lobe with lateral proton portals is lost when precise positioning is not assured. Patient V has a lateral tumor. For both temporal lobe dose constraints there was a stepwise improvement in TV-2 coverage progressing from X rays-poor positioning, to protons-poor positioning, to X rays-good positioning, to protons-good positioning. With good positioning, proton beams deliver ~70 CGE to 84% of the TV-2 with the NTCP temporal lobe limits. Poor positioning reduced this to 22%. These are approximately 20% higher than the comparable X ray beams plans. With this right-sided tumor, proton beams reduce the dose to the right temporal lobe significantly as compared to X rays, as shown in the dose-volume histograms in Figure 4. In this geometry, proton beams have a significant advantage over X ray beams but only if good positioning is assured. Patient G has a tumor that wraps 220’ around the spinal cord and does not extend superiorly enough for the temporal lobes to be dose constraining. The dose delivered to the TV-2 is driven by the positioning assumptions, as both modalities require multiple portals, each one sparing the spinal cord. For the proton beam plans, the TV-2 anterior to the spinal cord is treated with lateral fields; the TV-2 lateral to the spinal cord is treated with a posterior field in which the protons range out in the penumbra of the lateral field, a technique similar to that described by Castro et al. (4). Uncertainty in stopping the protons precisely in the penumbra of the lateral field results in small regions of double treatment within the TV-2, which accounts for the high dose tail in the dose-volume histogram (Fig. 3). The X ray plan for the TV-2 treatment consists of a pair of 45” anterior oblique portals, each with a 45” wedge and designed to spare the spinal cord (the right anterior portal treats to the left of the spinal cord, the left anterior oblique to the right). A posterior field with a cord block boosts the dose to the TV-2 lateral to the spinal cord. This field arrangement results in “hot spots” in the anterior portion of the TV-2 (and in the pharynx), accounting for the high dose tail in the dose volume histogram in the well positioned X ray plan (Fig. 3). With good positioning, proton beams deliver ~70 CGE to 75% of the TV-2; poor positioning reduces this percentage to 33%. In both positioning accuracy assumptions, approximately 10% more of the tumor received 70 CGE in the proton plan than in the X ray plan. Since the patient’s tumor is in the cervical spine, the temporal lobes were not dose limiting. Positioning precision was necessary to take advantage of the proton dose deposition characteristics and deliver higher doses to larger percentages of the TV-~‘S than the
??H.
TATSUZAKI AND M. M. URIE
763
X ray plans. Poorly positioned proton beam plans did not do any better, and in 5 of the 6 cases were actually worse, than the well-positioned X ray beam plans. The differences between the good and poor positioning plans were similar for both modalities. Regardless of radiation type, good positioning can allow significantly more dose to be delivered to these target volumes which are in close proximity to critical normal tissues. Precise localization and position confirmation can contribute to better radiation therapy for patients treated with conventional photons as well as with protons. In developing these plans, only the constraints enumerated in Table 1 were considered; no other normal tissues were considered to be dose limiting. Some plans developed with the X ray beams delivered high, in fact, unacceptably high, doses to other normal tissues but were accepted for this comparison. For instance, for Pt G the TV-2 treatment created a 30% “hot spot” which would have delivered more than 90 Gy to portions of the pharynx, thereby making the plan unacceptable clinically. With Pt N, the X ray plans delivered the target dose (74 Gy) to the parotid glands while the proton plans maintained 50% of their volume below 40 CGE. These X ray plans were judged significantly inferior to the proton beams plans, and by a considerably greater margin than the differences in the TCPs or the dose-volume histograms for the TV-2s would indicate. Considerations such as these are very important in assessing the overall merits of different plans but are difficult to include as explicit constraints, since clinicians will accept some complications, such as xerostomia, if they are necessary to deliver the needed dose to the tumor but would certainly chose to avoid them if possible. Three-dimensional treatment planning capability was essential to this study, not only because it allowed the 3 dimensional dose analysis, but also because portals were designed by geometric optimization in the beam’s eye view. For example, with Patient V, the oblique angles for both protons and X rays were selected to optimize TV-2 treatment while sparing the brain stem. For Patient G the conventional X ray plan would have been opposed lateral fields, stopping the dose at the tolerance of the spinal cord, 53 Gy. With the 3-dimensional system, a significant portion of the TV-2 was treated to significantly higher doses using the combination of an anterior 45” wedged pair and a posterior field with a spinal cord block. These types of field arrangements would not have been designed without the three dimensional capability. Very many X ray beams, each with its own field defining aperture, were considered while developing the X ray plans; in essence, conformal therapy was simulated. This too could not have been done without 3 dimensional planning. The proton beams for these plans assumed an uncertainty in the ability to predict and control the end of range of the protons and hence of how tightly the dose can be tailored to the target volume distally (6). The primary contributors to this uncertainty are the accuracy of the CT values (Hounsfield units), the spatial resolution of the CT
764
I. J. Radiation Oncology 0 Biology ??Physics
scans (particularly in the caudad-cephalad direction) (5), and the scattering effects of multiple fine heterogeneities (16). Because of these uncertainties, the prescribed range is typically 5% greater than the calculated maximum required range. In addition, the compensating bolus is designed not to contour exactly the distal end of range to the target volume, but rather to assure that the target will be treated as long as the patient and bolus are aligned within the positioning uncertainty (15). These factors contribute to intentionally treating more non-target tissue in the nominal situation represented by the computer plan. The uncertainties associated with the CT values were also considered for the X ray beams, but their effects on the dose distribution are much less than those for the proton beams. The predicted exponential absorption of the X rays is only very modestly affected by differences in CT values of the magnitudes appropriate here. The dose-volume histograms summarize the dose that the TV-2s receive. Since it is not known which parameter(s) of these histograms determines the likelihood of controlling the tumor, it is difficult to judge the importance of the differences among the histograms. Various statistics were extracted and could have been presented, such as minimum tumor dose, mean dose, volume of tumor receiving, for example, 95% of the prescribed dose. These single parameters are not indicative of the overall differences among the histograms. Therefore, we elected to present the histograms themselves and the predictions of a model of tumor control probability for inhomogeneously irradiated tumors (7). The model divides the tumor into subgroups of tumorlets that are small enough to be considered as receiving a uniform dose; it then uses dose response data derived from clinical and experimental data to predict the relative likelihood of tumor control of the inhomogeneously irradi-
August
1991, Volume 21, Number 3
ated target to what it would have been had the target been homogeneously irradiated to the prescribed dose. This model allows the differences among the plans to be reduced to a comparison of a single number, TCP, and may give an indication of the relative merits of the plans. In these treatment plans, those with good positioning assumptions had estimated TCPs 12-25% higher than those with poor positioning. The differences were similar for both modalities, protons and X rays. With good positioning assumptions, the estimated TCP’s of the proton beam plans were 5% to 12% higher than X ray beam plans. For these TCP’s, a shallow (1.3) yso was used, which was estimated from sparse clinical data. Were the true gamma factor to be larger than this, the differences among the estimated TCPs would be larger. The adequacy of this model of TCP is yet to be proven: even if its concepts are correct, the differences in the TCPs predicted are very dependent upon the assumed slope of the dose-response curve (yso parameter). A comparison of the isodose contours and of the dose-volume histograms is certainly essential to assess the differences among plans, including those presented here. CONCLUSIONS Precise positioning of patients is essential if the dose deposition characteristics of protons are to be used to full advantage to treat tumors adjacent to critical normal tissues. With rigid immobilization and confirmation of correct position prior to each treatment, significantly more dose can be delivered to larger percentages of the tumors than with X rays. However, without good positioning, the proton beam treatments deteriorate and may not be as good as conventional radiation therapy with good positioning.
REFERENCES 1. Austin-Seymour;M.; Munzenrider, J.; Goitein, M.; Gentry, R.; Gragoudas, E.; Koehler, A.; McNulty, P.; Osborne, E.;
2.
3. 4.
5.
6.
Ryugo, D.; Seddon J.; Urie, M.; Verhey, L.; Suit, H. Progress in low-LET heavy particle therapy: intracranial and paracranial tumors and uveal melanomas. Rad. Res. 104(2): S219-S226; 1985. Austin-Seymour, M.; Munzenrider, J.; Goitein, M.; Verhey, L.; Urie, M.; Gentry; R.; Bimbaum, S.; Ruotolo, D.; McMannus, P.; Skates, S.; Ojemann, R.; Rosenberg, A.; Schiller, A.; Koehler, A.; Suit, H. Fractionated proton radiation therapy for chordomas and low grade chondrosarcomas of the base of skull. J. Neurosurgery 70:13-17; 1989. Brahme, A. Dosimetric precision requirements in radiation therapy. Acta Radiol. Oncol. 23 (Fasc.5):379-391; 1984. Castro, J.; Collier, J.; Petti, P.; Nowakowski, V.; Chen G., Lyman, J.; Lindstadt, D.; Gauger, G.; Gutin, P.; Decker, M.; Phillips, T.; Baken, K. Charged particle radiotherapy for lesions encircling the brain stem or spinal cord. Int. J. Radiat. Oncol. Biol. Phys. 17:477-484; 1989. Goitein, M. Compensation for inhomogeneities in charged particle radiotherapy using computed tomography. Int. J. Radiat. Oncol. Biol. Phys. 4:499-508; 1978. Goitein, M. Calculation of the uncertainty in dose delivered
during radiation therapy. Med. Phys. 12:608-612; 1985. 7. Goitein, M Tumor control probability for an inhomogeneously irradiated target volume. NC1 report, section 5.8, S. Zink, project officer; 1987. 8. Goitein, M.; Abrams, M. Multi-dimensional treatment planning: I. Delineation of anatomy. Int. J. Radiat. Oncol. Biol. Phys. 9:777-788; 1983. 9. Goitein, M.; Abrams, M.; Rowell, D.; Pollari, H.; Wiles, J. Multi-dimensional treatment planning: II. Beam’s eye view, back projection and projection through CT sections. Int. J. Radiat. Oncol. Biol. Phys. 9:789-797; 1983. 10. Kutcher, G.; Burman, C. Calculations of complication probability factors for non-uniform normal tissue irradiation: the effective volume method. Int. J. Radiat. Oncol. Biol. Phys. 16:1623-1630; 1989. 11. Lyman, J. Complication probability as assessed from dosevolume histograms. Rad Res. 104:S13-S19; 1985. 12. Lyman, J.; Wolbarst, A. Optimization of radiation therapy III: a method of assessing complication probabilities from dose-volume histograms. Int. J. Radiat. Oncol. Biol. Phys. 13(1):103-109; 1987. 13. Rabinowitz, I.; Broomberg, J.; Goitein, M.; McCarthy, K.; Leong, J. Accuracy of Radiation Field Alignment in Clinical
Importance of precise positioning in proton therapy 0 H. Practice. Int. J. Radiat. One. Biol. Phys. 11(10):1857-1867; 1985. 14. Urano, M.; Verhey, L.; Goitein, M.; Tepper, J.; Suit, H.; Phil, D.; Mendiondo, 0.; Gragoudas, E.; Koehler, A. Relative biological effectiveness of modulated proton beams in various murine tissues. Int. J. Radiat. Oncol. Biol. Phys. 10(4):509-514; 1984. 15. Urie, M.; Goitein, M.; Wagner, M. Compensating for hetero-
TATSUZAKIAND M.
M. URIE
765
geneities in proton radiation therapy. Phys. Med. Biol. 29: 553-566; 1984. 16. Urie, M.; Goitein, M.; Holley , W.; Chen, G. Degradation of the Bragg peak due to inhomogeneities. Phys. Med. Bio. 31(1):1-15; 1986. 17. Verhey, L.; Guotein, M.; McNulty, P.; Munzenrider, J.; Suit, H. Precise positioning of patients for radiation therapy. Int. J. Radiat. Oncol. Biol. Phys. 8:289-294, 1982.
