P

Tpta

mM4r

R

Int. J. Radiation Oncology Biol. Phys., Vol. 62, No. 2, pp. 571–578, 2005Copyright © 2005 Elsevier Inc.

Printed in the USA. All rights reserved0360-3016/05/$–see front matter

doi:10.1016/j.ijrobp.2005.02.033

HYSICS CONTRIBUTION

BENEFIT OF USING BIOLOGIC PARAMETERS (EUD AND NTCP) IN IMRTOPTIMIZATION FOR TREATMENT OF INTRAHEPATIC TUMORS

EMMA THOMAS, M.D., OLIVIER CHAPET, M.D., MARC L. KESSLER, PH.D.,THEODORE S. LAWRENCE, M.D., PH.D., AND RANDALL K. TEN HAKEN, PH.D.

Department of Radiation Oncology, University of Michigan, Ann Arbor, MI

Purpose: To investigate whether intensity-modulated radiotherapy (IMRT), optimized using the generalizedequivalent uniform dose (gEUD) and normal tissue complication probability (NTCP) models, can increase thesafe dose to intrahepatic tumors compared with three-dimensional conformal RT (3D-CRT). A secondaryobjective was to investigate the optimal beam arrangement for liver IMRT plans.Methods and Materials: Planning CT data of 15 patients with intrahepatic tumors, previously treated with3D-CRT, were used as input. The dose delivered using 3D-CRT had been limited either by tolerance of adjacentorgans, which were close to, or overlapped with, the planning target volume (PTV; overlap cases, n � 8), or livertoxicity (nonoverlap, n � 7). IMRT plans were created using the gEUD to maximize the dose across the PTV andthe NTCP to maintain the organ-at-risk toxicity to that of the conformal plan. Increased heterogeneity wasallowed across the PTV in overlap cases, without compromising the minimal PTV dose of the conformal plan andrestricting the maximal dose to within 110% of the mean. Three different beam arrangements were used for eachcase: seven-field equidistant axial, six-field noncoplanar (predominantly right-sided beams), and a reproductionof the conformal gantry angles. gEUDs were also computed and used for evaluation of the plans (regardless ofplanning technique) to reflect the response of both high- and low-grade tumors. The IMRT plan that allowed thegreatest gEUD across the PTV was used in the comparison with the 3D-CRT plan.Results: The use of IMRT significantly increased the maximal gEUD achievable across the PTV compared withthe 3D-CRT plans. This was the case for the assumptions of both high- and low-grade tumors, irrespective of thetumor position within the liver. The mean gEUD increase was 11 Gy (high grade) and 18.0 Gy (low grade) foroverlap cases (p � 0.001 and p � 0.003, respectively) and 10 Gy for nonoverlap cases (p � 0.020). Whencomparing the IMRT beam arrangements, gEUDs were considered equivalent if they differed by less than onefraction (1.5 Gy). In overlap cases (n � 8), an equivalent “best” gEUD value was obtained in 3, 5, and 7 casesfor the original conformal angle, seven-field axial, and six-field noncoplanar plan, respectively. The correspond-ing results were 5, 2, and 3 in the cases without an overlap (n � 7).Conclusion: We have successfully used mathematical/biologic models directly as cost functions within theoptimizing process to produce IMRT plans that maximize the gEUD while maintaining compliance with awell-defined protocol for the treatment of intrahepatic cancer. For both PTV–organ-at-risk overlap andnonoverlap situations, IMRT has the capacity to improve the maximal dose achievable across the PTV, expressedin terms of the gEUD. The use of multiple noncoplanar beams appears to confer an advantage over fewer beamsin cases with PTV–organ-at-risk overlap. When liver toxicity is the dose-limiting factor, high gEUD values areobtained most frequently when the field arrangement is chosen to provide the shortest possible transhepatic pathlength. © 2005 Elsevier Inc.

Intensity-modulated radiotherapy, Optimization, Equivalent uniform dose, Radiotherapy, Hepatic neoplasms.

illw(d

OATs

INTRODUCTION

he use of radiotherapy (RT) has been shown to benefitatients with both primary and secondary intrahepaticumors (1). A dose–response relationship exists, with anssociation between the delivery of a higher dose and

Reprint requests to: Randall K. Ten Haken, Ph.D., Depart-ent of Radiation Oncology, University of Michigan, 1500 E.edical Center Dr., UH-B2C432, Box 0010, Ann Arbor, MI

8109-0010. Tel: (734) 936-8695; Fax: (734) 935-7859; E-mail:th@umich.edu

Supported in part by NIH Grants P01CA59872 and

01CA85684. A

571

mproved clinical outcome (2, 3). The treatment of theiver with RT does, however, present a number of chal-enges. The liver is known to be a radiosensitive organith increasing risks of radiation-induced liver disease

RILD) when whole liver doses reach �30 Gy (4, 5). Theose that is safely deliverable to part of the liver may also

Presented in part at the ASTRO Annual Meeting, Atlanta, GA,ctober 3–7, 2004.cknowledgments—We are thankful to Robin Marsh, Danielatro, and Karen Vineberg for their assistance with the planningystem and generation of the figures.

Received Nov 29, 2004, and in revised form Feb 20, 2005.

ccepted for publication Feb 22, 2005.

bd

tCbgwppqPtpapmnapip

dptPddt(ctngbmt

wimwpapowop

P

3MloFal(sre(

T

tdcas

e

572 I. J. Radiation Oncology ● Biology ● Physics Volume 62, Number 2, 2005

e limited by the close proximity to the stomach anduodenum.Patients at our institution with intrahepatic tumors are

reated with three-dimensional (3D) conformal RT (3D-RT) according to a series of protocols (2, 6, 7) approvedy our institutional review board that seek to deliver thereatest possible dose to the planning target volume (PTV)hile subjecting the patient to a normal tissue complicationrobability (NTCP) for RILD equal to, or less than, arotocol-defined limit (15% in the current protocol). Fre-uently, the dose to other organs at risk (OARs) limits theTV dose to below that which could be safely administered

o the tumor as determined by the liver NTCP alone. This isarticularly pertinent when an overlap between the PTV andn extrahepatic OAR (primarily stomach or duodenum) isresent. To increase the mean PTV dose to more than theaximal tolerated dose of one of these OARs, it would be

ecessary to relax the PTV homogeneity constraints andllow a lower dose in the overlap region. The impact ofartial tumor boosting in liver tumors has never been clin-cally tested, but theoretical data exist to support such aractice (8, 9).Useful comparison of treatment plans becomes more

ifficult when the PTV homogeneity varies. For example,lans that achieve a greater mean dose cannot be assumedo be clinically superior, because cold spots within theTV may compromise the probability of tumor control,espite the presence of hot spots driving up the meanose. Some of these difficulties might be overcome withhe use of the concept of an equivalent uniform doseEUD) (10, 11). This may be described as the biologi-ally equivalent dose that, if given uniformly, would leado the same cell kill in the tumor volume as the actualonuniform dose distribution. Optimization using theeneralized version of the EUD (gEUD) has previouslyeen shown to be superior to dose–volume-based opti-ization in terms of OAR sparing, with equal or better

arget coverage (12–14).

Table 1. Pat

Pt. no. Tumor CTV (cm3)

1 Metastasis 3702 Cholangio 2483 Cholangio 1634 Cholangio 12515 Cholangio 2026 Cholangio 2457 Metastasis 2058 Cholangio 10039 Metastasis 366

10 Hepatoma 30511 Metastasis 52712 Metastasis 296913 Metastasis 148514 Cholangio 21615 Hepatoma 589

Abbreviation: CTV � clinical target volume.

The primary purpose of this study was to investigatehether the use of intensity-modulated RT (IMRT) could

mprove the dose safely deliverable to intrahepatic tu-ors compared with 3D-CRT. The gEUD for the targetas used as the evaluation endpoint so that treatmentlans with nonuniform dose distributions could be reli-bly compared. NTCP was also used in the optimizationrocess to ensure that all plans were comparable in termsf toxicity and that protocol limits to adjacent OARsere adhered to. A secondary objective was to study the

ptimal arrangement of external beams when treating intrahe-atic tumors with IMRT.

METHODS AND MATERIALS

atientsFifteen patients were identified (Table 1) who had undergone

D-CRT to intrahepatic tumors according to the University ofichigan dose-escalation protocols (2, 7). In 8 of these cases, the

ocation of the tumors was such that the PTVs were adjacent to, orverlapped with, the stomach or duodenum (overlap cases; Fig. 1A).or these cases, the PTV doses were limited by the tolerance ofn adjacent OAR (instead of the normal liver NTCP) to dosesower than what would have been permitted by the protocoldesigned such that a normal liver NTCP for RILD of 15% wasupposed to define or limit the PTV dose [2, 6, 7]). In theemaining 7 cases, the dose to the PTV had been successfullyscalated until a normal liver NTCP of 15% was achievednonoverlap cases; Fig. 1B).

reatment planningThe treatment planning CT scans of these patients were used for

he purposes of this study (patients had been asked not to eat orrink 6 h before CT). Information regarding gross tumor volumes,linical target volumes, and planning target volumes (PTVs) waslready available. All significant OARs were contoured (liver,tomach, duodenum, spinal cord, and kidneys).

3D-CRT plans. The clinical cases had been planned by experi-nced dosimetrists with the number and angle of fields chosen to

aracteristics

Location Liver volume (cm3)

lobe adjacent to stomach 1168date lobe adjacent to duodenum 1738lobe adjacent to duodenum 1751lobe adjacent to stomach 1993

date lobe adjacent to duodenum 1625date lobe adjacent to duodenum 2027lobe adjacent to duodenum 2136t lobe adjacent to duodenum 2677t lobe 1560lobe central 2414t lobe 1329

lacing right lobe of liver 3662t lobe 3156t lobe 1314t lobe central 2182

ient ch

LeftCauLeftLeftCauCauLeftRighRighLeftRighRepRighRighRigh

m(fntwlsdtaTtIttsc

w

1

2

3

APsrd

wdt

fat�twbc

upvwtoou

tnAoieptph

SK

S

DL

Ftit

573EUD-based optimization of liver IMRT ● E. THOMAS et al.

inimize the path length through the liver and avoid local OARs7, 15). The number of fields used varied from two to six (medianour). Treatment planning objectives had been defined by ourormal liver toxicity–based dose-escalation protocols to producehe lowest possible effective volume for normal liver (2, 6, 7)hile maintaining PTV dose homogeneity of �7%. Dose calcu-

ations had originally been performed using our 3D-CRT planningystem (UM-plan) using 16-MV photon beams. Our in-house–eveloped IMRT inverse planning system (UM-Opt) is equippedo use the same beam energy but uses a different dose calculationlgorithm (convolution/superposition based [16]) for the beamlets.hus, all clinical 3D-CRT reference plans were recalculated using

his system to allow meaningful comparison with the IMRT plans.n some cases, small adjustments in the beam weights were madeo ensure that the maximal dose to the PTV had been delivered andhat no violations of the clinical protocol had occurred. The re-ultant conformal plans were used for all subsequent analyses andomparisons.

IMRT plans. For each case, three IMRT plans were generatedith different beam arrangements:

. Reproduction of the beam angles used in the clinical 3D-CRTplan (original angle)

. Seven-field equidistant axial, with gantry angles of 0°, 51°,

ig. 1. (a) Computed tomography slice of overlap case. (b) Computedomography slice of nonoverlap case. Gross tumor volume denoted bynner contour; Planning target volume by outer encompassing con-our; stomach denoted by outer intersecting contour.

103°, 154°, 206°, 257°, and 309° (seven field) c

. Six-field noncoplanar with predominantly right-sided beams,with gantry angles of 10°, 220°, 270°, and 320° with the couchat 0°, 20°, and 330°, with a 90° couch rotation (standard angle)

grid of 1-cm beamlets was applied to each beam to cover theTV. In 2 cases in which the tumors were very large, the beamletize had to be increased (1.5 cm and 2.0 cm) owing to limitselated to the maximal number of beamlets and calculation pointensity.Optimization of treatment plans was performed using UM-Opt,

hich allows the use of multiple different parameters in theevelopment of costlet functions (17, 18). One such costlet func-ion has been developed using the gEUD equation (11):

gEUD � � 1

N�

i�1

N

Dia�1 � a

or N dose points, Di, and tumor-specific parameter a. All tumorsre represented by negative a values, with variation depending onhe grade or aggressiveness of the individual tumor (11–14). If a ��, the gEUD is equal to the minimal tumor dose. This represents

he situation of a very aggressive tumor in which tumor controlill be adversely affected by even the smallest cold spot. When theiologic behavior of tumors is more benign, the a values can beorrespondingly less negative.

During our early work of optimizing using the gEUD, we tested these of multiple different a values to determine how best to produce alan applicable to all tumor grades (19). We found that the use of aery negative a value (�50) was able to maximize the lowest doseithin the PTV and, therefore, define the position of the shoulder of

he cumulative dose–volume histogram (DVH) for the PTV. The usef less negative a values in optimization maximized the dose through-ut the rest of the PTV. In this study, a values of �50 and �5 weresed sequentially in the optimization process.

Constraints were placed (Table 2) to ensure that the maximalolerated doses to the spinal cord, kidneys, stomach, and duode-um, as defined by the protocol, were not exceeded in all cases.dditional constraints were used to prevent any hotspots of �10%f the mean PTV dose (prescription dose) within either the PTVtself or uninvolved tissue, because the toxic consequences of thosextremes (especially in the uninvolved tissues) are uncertain, andreliminary studies for the PTV (not shown) indicated no addi-ional gain in the gEUD for hotspots beyond that point (a trend alsoointed out by others [9]). For nonoverlap cases, for which loss ofomogeneity was not necessary to escalate the dose to the PTV,

Table 2. Liver protocol organ-at-risk dose constraints

OAR Dose constraints

pinal cord �38.4 Gy in 1.5-Gy fractionsidneys If one kidney receives �20 Gy, 90% of

contralateral kidney constrained to �18 Gytomach �50 Gy (up to 20% of stomach volume allowed

to receive up to 60 Gy in cases in whichtumor abuts stomach)

uodenum �68 Gyiver Normal liver NTCP �15%

Abbreviation: OAR � organ at risk; NTCP � normal tissue

omplication probability.

tt

t3(ToptpwvdIc3O

E

aTrv�ttcf

vmfifc

O

2levtNNn

tttao

mgCwrdgva�TCG�idtf

g“s(3st

N

ccgpt(wNcGImTh�a

d7

LLSD

F

574 I. J. Radiation Oncology ● Biology ● Physics Volume 62, Number 2, 2005

he dose homogeneity range of the 3D-CRT plan was replicated inhe IMRT plans.

For overlap cases, the hepatic, duodenal, and gastric toxicities ofhe IMRT plans were matched as closely as possible to those of theD-CRT plans as determined using the NTCP model of Lyman20) and the DVH reduction scheme of Kutcher and Burman (21).his was accomplished via another costlet that can be used in theptimization process within UM-Opt (18) to create comparablelans that differ only with respect to the dose distribution acrosshe PTV. NTCP values (using parameters associated with therotocol-defined 1.5 Gy/fraction; Table 3) of the 3D-CRT plansere first calculated for uninvolved liver (liver minus gross tumorolume or whole liver if the gross tumor volume could not beefined), stomach, and duodenum and were used as the standard.MRT plans were then optimized using NTCP costlets, with a highost assigned for any value greater than that calculated for theD-CRT plan. In this way, we ensured that the NTCPs of theseARs in the IMRT plans did not exceed 0.5% of the standard.

valuationThe plans were also compared using the gEUD values computed

cross the PTV, regardless of the method used to generate the plan.o reflect the existence of wide range of tumor grades, two

esulting gEUD values were calculated for each plan using aalues representing different tumor aggressiveness. An a value of5 was used to demonstrate the gEUD in the case of a low-grade

umor and �20 to demonstrate the situation of a very aggressiveumor. The IMRT plan with the greatest gEUD in each case wasompared with the 3D-CRT plan. Statistical analysis was per-ormed using a paired t test.

The effects of the beam arrangement on the resulting targetolume gEUDs for the IMRT plans in each case were also docu-ented and compared. Clinically significant differences were de-ned as those in which the gEUD varied by at least one treatmentraction (1.5 Gy for this protocol); plans exhibiting smallerhanges were judged to be clinically equivalent.

RESULTS

verlap casesThe DVHs of a representative case are presented in Fig.

. The NTCP values of the IMRT plans for uninvolvediver, stomach, and duodenum (DVHs in Fig. 2B–D) did notxceed those of the 3D-CRT plans by �0.5% on an indi-idual plan basis and were, on average, slightly lower thanhe 3D-CRT plans for stomach and duodenum (mean liverTCP 1.3% for CRT and 1.5% for IMRT; mean stomachTCP 16.0% for CRT and 14.1% for IMRT; mean duode-

Table 3. NTCP parameter values (based on 1.5 Gy/Fx)

n m TD50 Endpoint

iver (primary) 0.97 0.12 39.8 Any RILDiver (metastasis) 0.97 0.12 45.8 Any RILDtomach 0.07 0.3 62 Bleedinguodenum 0.12 0.49 180 Bleeding

Abbreviations: NTCP � normal tissue complication probability;x � fraction; RILD � radiation-induced liver disease.

um NTCP 7.4% for CRT and 6.5% for IMRT). Although fi

he maximal liver doses were greater for the IMRT plans,he mean doses for the normal liver were comparable;herefore, given a Lyman model parameter n having a valuepproaching unity, the NTCPs were equivalent (22). Allther protocol constraints were adhered to.The minimal dose received by the PTV was not compro-ised by the use of IMRT in any case. The IMRT plans had

reater mean PTV doses (mean 84.0 Gy) compared with theRT plans (mean 64.8 Gy). Heterogeneity across the PTVas wider in some of the IMRT plans, but the aim of

estraining the maximal dose to within 10% of the meanose was observed. The IMRT plan resulting in the greatestEUD value for each case was compared with the gEUDalue of the 3D-CRT plan (Table 4). In each case, the gEUDcross the PTV was evaluated using a values of �5 and20 to reflect different degrees of tumor aggressiveness.he IMRT plans produced greater gEUD values than theRT plans in all cases. The mean increase in gEUD was 11y if it was presumed that the tumor was high grade (a �20) and 18 Gy for a low-grade tumor. Statistical compar-

son of the gEUDs obtained using the two RT techniquesemonstrated a highly statistically significant difference be-ween them (p � 0.001 for an a value of �5 and p � 0.003or an a value of �20).

The beam arrangements that led to the best plans (greatestEUDs) are summarized in Table 5. As mentioned inMethods and Materials,” the resulting gEUDs were con-idered equivalent if they differed by less than one fraction1.5 Gy). An equivalent “best” gEUD value was obtained in, 5, and 7 of 8 cases for the original conformal angle,even-field axial, and six-field noncoplanar plan, respec-ively.

onoverlap casesDose volume histograms of a representative nonoverlap

ase are presented in Fig. 3. As in the overlap cases, noompromise occurred with the minimal PTV dose. Homo-eneity across the PTV was as good as with the 3D-CRTlan in all cases. In these cases, the dose-limiting organ washe liver, so all plans had a normal liver NTCP of 15%normal liver DVHs, Fig. 3B). All other protocol constraintsere also adhered to (with the stomach and duodenumTCPs, in particular, not a factor in these nonoverlap

ases). The IMRT plans produced greater mean doses (84.7y) compared with the 3D-CRT plans (74.5 Gy). Use of

MRT resulted in increased gEUDs in 5 of 7 cases andatched the conformal gEUDs in the remaining 2 cases.he increase in the gEUD with the use of IMRT for bothigh- and low-grade tumors was statistically significant (p

0.018 and p � 0.022, respectively), with a mean oflmost 10 Gy in both cases (Table 6).

The proportion of the different beam arrangements pro-ucing at least equal “best” gEUD plans is shown in Tableand was obtained in 5 of 7 original angle, 2 of 7 seven-

eld, and 3 of 7 standard angle plans.

tddhpmcgtm

onpwtnmt(

t

IMRT

575EUD-based optimization of liver IMRT ● E. THOMAS et al.

DISCUSSION

Intensity-modulated radiotherapy gives us the unique abilityo shape isodose curves, thus allowing avoidance of excessiveose to OARs. It also provides a straightforward way to pro-uce plans with a variation of the dose across the PTV. Weave used the gEUD as a cost function in the IMRT planningrocess to escalate the dose to intrahepatic tumors, whileaintaining the doses to adjacent OARs within defined proto-

ol limits. For a plan with increased heterogeneity and a greaterEUD to be accepted by all clinicians as superior, two condi-ions must be fulfilled: no possibility of inferior tumor kill (i.e.,

Fig. 2. Comparison of dose volume histograms for diffevolume. (b) Normal liver. (c) Stomach. (d) Duodenum.

inimal dose not compromised) and no increased toxicity to o

ther OARs. By optimizing using a gEUD costlet with a veryegative value of a � �50, we were able to achieve IMRTlans that did not compromise the minimal tumor dose. Weere also able to fulfill the second condition by using NTCP in

he cost function to match the plans in terms of toxicity toearby OARs. The design of this study, therefore, allowedeaningful comparison of IMRT and 3D-CRT plans for in-

rahepatic tumors, with confidence that greater gEUD valuesobserved during the evaluation phase) represent better plans.

Almost all our IMRT plans were considerably superioro the 3D-CRT plans in terms of gEUD evaluation. The

anning techniques for overlap case. (a) Planning target� intensity-modulated radiotherapy.

rent pl

nly 2 cases in which IMRT failed to gain an advantage

wtswsp3gtTsstsplcwpshwh

w

maaCwhnstenf

MP

Pm

T

Fn(

576 I. J. Radiation Oncology ● Biology ● Physics Volume 62, Number 2, 2005

ere also the cases in which beamlet sizes of �1 cm hado be used owing to the large tumor size and the con-traints of our planning system. Because liver toxicityas the dose-limiting factor in both cases, it was not

urprising that the large-beamlet IMRT plans failed torotect the normal liver more than was possible usingD-CRT. Despite these 2 cases, the difference in theEUD using the two treatment techniques was still sta-istically significant for the nonoverlap group as a whole.his confirmed the ability of IMRT to allow relativeparing of an organ in which a tumor is embedded. Ithould also be noted that these increases in the gEUD (forhe nonoverlap cases) occurred despite maintaining theame target volume dose homogeneity as in the 3D-CRTlans. The scale of increase in the gEUD values (almost0 Gy) confirmed the potential of IMRT to make a reallinical difference in terms of tumor control in patientshose radiation dose is limited by liver toxicity. Theotential gain (in terms of the resulting gEUD) washown to be even greater in the cases in which an overlapas occurred between the PTV and a nonhepatic OAR,ith mean gEUD increases ranging from 11 Gy forigh-grade tumors to 18 Gy for low-grade tumors.Our findings are reflected in a recent Phase II study in

hich IMRT was used to treat pancreatic and bile duct

Table 4. gEUD comparison for overlap cases

Pt. no.

gEUD a � �20 (Gy) gEUD a � �5 (Gy)

3D-CRT IMRT 3D-CRT IMRT

1 59 64 61 692 66 76 67 823 56 69 57 714 55 64 57 745 56 67 58 696 67 73 67 787 74 97 75 1188 61 73 67 93

ean 62 73 64 82(t test) 0.001 0.003

Abbreviations: gEUD � generalized equivalent uniform dose;t. No. � patient number; 3D-CRT � three-dimensional confor-al radiotherapy; IMRT � intensity-modulated RT.

Table 5. Optimal field arrangement for overlap cases

Pt. no. Original angle Seven field Standard angle

1 X X2 X X X3 X X X4 X X X5 X6 X7 X8 X

otal 3 5 7

vAbbreviation: Pt. No. � patient number.

alignancies (23). They found that IMRT was well toler-ted, with a reduced mean dose to the liver, kidneys, stom-ch, and small bowel, without compromise of local control.heng et al. (24) redid the plans of 12 patients using IMRTho had previously experienced RILD after 3D-CRT forepatocellular carcinoma. Contrary to our findings, theyoted a reduction in the liver NTCP with IMRT, despite aignificant increase in the mean liver dose. As recognized byhe authors, their choice of the Lyman NTCP model volumeffect parameter (n � 0.32), based on clinical estimates, didot reflect mounting data that a much greater volume effector RILD is likely to exist (22). We believe that with an n

ig. 3. Comparison of dose volume histograms for different plan-ing techniques for nonoverlap case. (a) Planning target volume.b) Normal liver. IMRT � intensity-modulated radiotherapy.

alue approaching 1, the increase in the mean dose observed

wI

t(tptnsmgbsPihaIpcst

itcuepseehaeda

itc

iomdTtwalotnsi

osswwIrrsmmtlrntmt

ctOtte

T

Mp

577EUD-based optimization of liver IMRT ● E. THOMAS et al.

ould produce an increased probability of RILD using theirMRT planning technique.

We acknowledge that some of the greater IMRT advan-age gained in terms of the gEUD for the overlap casescompared with the nonoverlap cases) may have been due tohe relaxation of the homogeneity constraints for thoselans. However, we have also demonstrated that much ofhe benefit resulted from the technique itself, given that theonoverlap cases with matched homogeneity also had atatistically significant IMRT advantage. Furthermore, itay have been possible to narrow the differences in the

EUD between the IMRT and 3D-CRT in our overlap casesy redoing the 3D-CRT plans with the use of multipleegments and thus allowing dose escalation to part of theTV. In fact, in 4 of the 8 cases, methods to produce an

nhomogeneous dose distribution (tolerance outside �7%)ad already been used. In all these cases, (Patients 1, 3, 4,nd 5), the gEUD values remained inferior to those of theMRT plans. The production of 3D-CRT plans with multi-le segments is a labor-intensive and time-consuming pro-ess that cannot always be justified within the clinicaletting. It is, therefore, reasonable to compare new treatmentechniques with the current standard.

The choice of beam angles in the conformal planning ofntrahepatic tumors has traditionally focused on minimizinghe path length through the liver. We were, therefore, con-erned that the use of multiple IMRT beams would result innacceptable liver toxicity. Although we did not plannough cases to perform a formal statistical analysis, clearatterns were observed. When the liver is the dose-limitingtructure, as expected, it appears preferable to use the short-st possible transhepatic path lengths, as chosen by experi-nced dosimetrists. When another OAR is dose limiting,owever, the use of multiple beams incident from presetngles seems to confer an advantage. This finding can bexplained by the superior ability of multiple beamlets inci-ent from multiple angles to generate sharp dose gradientsnd shield adjacent OARs.

The main clinical limitation of IMRT in the treatment ofntrahepatic tumors relates to difficulties with immobiliza-ion. Because of its position in relation to the diaphragm,

Table 6. gEUD comparison for nonoverlap cases

Pt. no.

gEUD a � �20 gEUD a � �5

3D-CRT IMRT 3D-CRT IMRT

9 71 81 72 8110 94 108 95 10811 58 67 59 6712 69 70 69 7013 69 71 69 7114 89 113 89 11415 69 79 69 79

ean 74 84 75 84(t test) 0.018 0.022

Abbreviations as in Table 4.

onsiderable organ motion occurs during respiration. Good

ntrafraction reproducibility during liver RT has been dem-nstrated using active breathing control (25). In addition,ethods of on-line imaging and setup adjustment have been

escribed to improve targeting between fractions (26, 27).hus, for purposes of this study, we assumed that much of

he systematic differences between treatment sessionsould be accounted for using the methods mentioned

bove. Those findings suggest that accurate localization ofiver tumors will not be an insurmountable barrier to the usef IMRT, although careful consideration of organ localiza-ion is necessary for clinical implementation of these tech-iques. Additional ongoing studies seek to include the re-idual effects of setup uncertainty and organ motion directlynto the dose planning (28–30).

The treatment of patients with 3D-CRT at the Universityf Michigan according to a standard protocol has not re-ulted in unacceptable toxicity. Typically, the dose-limitingtructures have been the liver, stomach, and duodenum,ith the doses to other OARs, such as the kidneys, fallingell below the protocol maximal constraints. The method of

MRT optimization in this study allowed the renal dose toise with minimal cost incurred until the protocol limit waseached. If this approach were to be used clinically, theafety of the protocol-defined limits would be tested to theiraximum, and vigilance would be required to ensure thatodification of these limits was not required. In addition,

he inevitable consequence of the use of multiple fields is aarger volume of uninvolved tissue exposed to a low level ofadiation and the potential for induction of second malig-ancies this may cause. This is not important in the pallia-ive setting such as the treatment of metastases, but may beore of an issue in patients with better prognoses, such as

hose receiving adjuvant therapy.

CONCLUSION

We have successfully used biologic models directly asost functions within the optimizing process to maximizehe gEUD for the PTV while constraining the dose to localARs using NTCP. Compared with the use of 3D-CRT for

he treatment of intrahepatic tumors, IMRT has the capacityo improve the functional dose achievable across the PTVxpressed in terms of the gEUD. The use of multiple non-

Table 7. Optimal field arrangement for nonoverlap cases

Pt. no. Original angle Seven field Standard angle

9 X10 X11 X12 X X13 X14 X X15 X X

otal 5 2 3

Abbreviation: Pt. No. � patient number.

cbW

va

1

1

1

1

1

1

1

578 I. J. Radiation Oncology ● Biology ● Physics Volume 62, Number 2, 2005

oplanar beams appears to confer an advantage over fewereams in cases in which the PTV and an OAR overlap.

hen liver toxicity is the dose-limiting factor, high gEUD l

REFEREN

188–196.

1

1

1

2

2

2

2

2

2

2

2

2

2

3

alues can be obtained most frequently when the fieldrrangement is chosen to provide the shortest possible path

ength.

CES

1. Greco C, Catalano G, Grazia A, et al. Radiotherapy of livermalignancies: From whole liver irradiation to stereotactic hy-pofractionated radiotherapy. Tumori 2004;90:73–79.

2. Dawson LA, McGinn CJ, Normolle D, et al. Escalated focalliver radiation and concurrent hepatic artery fluorodeoxyuri-dine for unresectable intrahepatic malignancies. J Clin Oncol2000;18:2210–2218.

3. Park HC, Seong J, Han KH. Dose-response relationship inlocal radiotherapy for hepatocellular carcinoma. Int J RadiatOncol Biol Phys 2002;54:150–155.

4. Ingold J, Reed G, Kaplan H. Radiation hepatitis. AJR Am JRoentgenol 1965;93:200–208.

5. Wharton JT, Delclos L, Gallager S, et al. Radiation hepatitisinduced by abdominal irradiation with the cobalt 60 movingstrip technique. AJR Am J Roentgenol Radium Ther Nucl Med1973;117:73–80.

6. Ten Haken RK, Martel MK, Kessler ML, et al. Use of Veffand iso-NTCP in the implementation of dose escalation pro-tocols. Int J Radiat Oncol Biol Phys 1993;27:689–695.

7. McGinn CJ, Ten Haken RK, Ensminger WD, et al. Treatmentof intrahepatic cancers with radiation doses based on a normaltissue complication probability model. J Clin Oncol 1998;16:2246–2252.

8. Deasy JO. Partial tumor boosts: Even more attractive thantheory predicts [letter]. Int J Radiat Oncol Biol Phys 2001;51:279–280.

9. Tomé A, Fowler J. Selective boosting of tumor subvolumes.Int J Radiat Oncol Biol Phys 2000;48:593–599.

0. Niemierko A. Reporting and analyzing dose distributions: Aconcept of equivalent uniform dose. Med Phys 1997;24:103–110.

1. Niemierko A. A generalized concept of equivalent uniformdose (EUD) [Abstract]. Med Phys 1999;26:1100.

2. Choi B, Deasy DO. The generalized equivalent uniform dosefunction as a basis for intensity-modulated treatment planning.Phys Med Biol 2002;47:3579–3589.

3. Wu Q, Mohan R, Niemierko A, et al. Optimization of intensitymodulated radiotherapy plans based on the equivalent uniformdose. Int J Radiat Oncol Biol Phys 2002;52:224–235.

4. Thieke C, Bortfeld T, Niemierko A, et al. From physical doseconstraints to equivalent uniform dose constraints in inverseradiotherapy planning. Med Phys 2003;30:2332–2339.

5. Ten Haken RK, Lawrence TS, McShan DL, et al. Technicalconsiderations in the use of 3-D beams in the abdomen.Radiother Oncol 1991;22:19–28.

6. Mackie TR, Scrimger JW, Battista JJ. A convolution methodof calculating dose for 15-MV x-rays. Med Phys 1985;12:

7. Vineberg KA, Eisbruch A, Coselmon MM, et al. Is uniformtarget dose possible in IMRT plans in the head and neck? IntJ Radiat Oncol Biol Phys 2002;52:1159–1172.

8. Kessler ML, McShan DL, Vineberg K, et al. Costlets: Ageneralized approach to cost functions for automated optimi-zation. Optim Engrg 2005; in press.

9. Chapet O, Thomas E, Kessler ML, et al. Esophagus sparingwith IMRT in ling tumor irradiation: An EUD-based optimi-zation technique. Int J Radiat Oncol Biol Phys 2005; In press.

0. Lyman JT. Complication probability as assessed from dose-volume histograms. Radiat Res 1985;8:S13–S19.

1. Kutcher GJ, Burman C. Calculation of complication probabil-ity factors for non-uniform normal tissue irradiation: Theeffective volume method. Int J Radiat Oncol Biol Phys 1989;16:1623–1630.

2. Dawson LA, Normolle D, Balter JM, et al. Analysis of radi-ation-induced liver disease using the Lyman NTCP model. IntJ Radiat Oncol Biol Phys 2002;53:810–821.

3. Milano MT, Chmura SJ, Garofalo MC, et al. Intensity-mod-ulated radiotherapy in treatment of pancreatic and bile ductmalignancies: Toxicity and clinical outcome. Int J RadiatOncol Biol Phys 2004;59:445–453.

4. Cheng JC, Wu J, Huang C, et al. Dosimetric analysis andcomparison of three-dimensional conformal radiotherapy andintensity-modulated radiation therapy for patients with hepa-tocellular carcinoma and radiation-induced liver disease. Int JRadiat Oncol Biol Phys 2003;56:229–234.

5. Dawson LA, Brock KK, Kazanjian S, et al. The reproducibil-ity of organ position using active breathing control (ABC)during liver radiotherapy. Int J Radiat Oncol Biol Phys 2001;51:1410–1421.

6. Balter J, Brock KK, Litzenberg DW, et al. Daily targeting ofintrahepatic tumors for radiotherapy. Int J Radiat Oncol BiolPhys 2002;52:266–271.

7. Fuss G, Salter BJ, Cavanaugh SX, et al. Daily ultrasound-based image-guided targeting for radiotherapy of upper ab-dominal malignancies. Int J Radiat Oncol Biol Phys 2004;59:1245–1256.

8. Lujan AE, Ten Haken RK, Larsen EW, et al. Quantization ofsetup uncertainties in 3D dose calculations. Med Phys 1999;26:2397–2402.

9. Lujan AE, Balter JM, Ten Haken RK. A method for incorpo-rating organ motion due to breathing into 3D dose calcula-tions: Sensitivity to variations in motion. Med Phys 2003;30:2643–2649.

0. Rosu M, Dawson LA, Balter JM, et al. Alterations in normalliver doses due to organ motion. Int J Radiat Oncol Biol Phys

2003;57:1472–1479.