The potential for sparing of parotids and escalation of biologically effective dose with...

PII S0360-3016(99)00304-1

PHYSICS CONTRIBUTION

THE POTENTIAL FOR SPARING OF PAROTIDS AND ESCALATION OFBIOLOGICALLY EFFECTIVE DOSE WITH INTENSITY-MODULATED

RADIATION TREATMENTS OF HEAD AND NECK CANCERS:A TREATMENT DESIGN STUDY

QIUWEN WU, PH.D., MATTHEW MANNING, M.D., RUPERT SCHMIDT-ULLRICH, M.D.,AND RADHE MOHAN, PH.D.

Department of Radiation Oncology, Medical College of Virginia, Virginia Commonwealth University and McGuire VA Hospital,Richmond, VA

Purpose: Conventional radiotherapy for cancers of the head and neck (HN) can yield acceptable locoregionaltumor control rates, but toxicity of many normal tissues limits our ability to escalate dose. Xerostomia representsone of the most common complications. The purpose of this study is to investigate the potential of intensity-modulated radiotherapy (IMRT) to achieve adequate sparing of parotids and to escalate nominal and/orbiologically-effective dose to achieve higher tumor control without exceeding normal tissue tolerances.Methods and Materials: An IMRT optimization system, developed at our institution for research and clinicalpurposes, and coupled to a commercial radiation treatment planning system, has been applied to a number ofcases of HN carcinomas. IMRT plans were designed using dose- and dose-volume-based criteria for 4 and 6 MVcoplanar but non-collinear beams ranging in number from 5 to 15 placed at equi-angular steps. Detailed analysisof one of the cases is presented, while the results of the other cases are summarized. For the first case, the IMRTplans are compared with the standard 3D conformal radiation treatment (3DCRT) plan actually used to treat thepatient, and with each other. The aim of the 3DCRT plan for this particular case was to deliver 73 Gy to thetumor volume in 5 fractions of 2 Gy and 28 fractions of 2.25 Gy/fx; and 46 Gy to the nodes in 2 Gy/fx whilemaintaining critical normal tissues to below specified tolerances. The IMRT plans were designed to be deliveredas a “simultaneous integrated boost” (SIB) using the “sweeping window” technique with a dynamic MLC. Thesimultaneous integrated boost strategy was chosen, partly for reasons of efficiency in planning and delivery ofIMRT treatments, and partly with the assumption that dose distributions in such treatments are more conformaland spare normal tissues to a greater extent than those with sequential boost strategy. Biologically equivalentdose normalized to 2 Gy/fx, termed here as normalized total dose (NTD), for this strategy was calculated usingpublished head and neck fractionation data.Results: IMRT plans were more conformal than the 3DCRT plans. For equivalent coverage of the tumor and thenodes, and for the dose to the spinal cord and the brainstem maintained within tolerance limits, the dose to parotidswas greatly reduced. For the detailed example presented, it was shown that the tumor and the nodes in the 3DCRTplan receive NTDs of 78 and 46 Gy, respectively. For the IMRT plan, a nominal dose of 70 Gy could be delivered tothe tumor in 28 fractions of 2.5 Gy each, simultaneously with 50.4 Gy to nodes with 1.8 Gy/fx. The two are biologicallyequivalent to 82 and 46 Gy, respectively, if delivered in 2 Gy/fx. Similar computations were carried out for other casesas well. The quality of IMRT plans was found to improve with increasing number of beams, up to 9 beams.Dose-volume-based criteria led to a modest improvement in IMRT plans and required less trial and error.Conclusion: IMRT has the potential to significantly improve radiotherapy of HN cancers by reducing normaltissue dose and simultaneously allowing escalation of dose. SIB strategy is not only more efficient and yields betterdose distributions, but may also be biologically more effective. Dose-volume-based criteria is better than purelydose-based criteria. The quality of plans improves with number of beams, reaching a saturation level for a certainnumber of beams, which for the plans studied was found to be 9. © 2000 Elsevier Science Inc.

IMRT, Head and neck cancers, Parotid sparing, Dose escalation.

INTRODUCTION

For reasons that may be obvious, intensity-modulated ra-diotherapy (IMRT) has generated intense and widespread

interest since its introduction into clinical use only a fewyears ago (1–3). IMRT optimally assigns weights to indi-vidual rays of a beam as opposed to a single weight to thebeam as a whole. Such control of individual rays allows

Reprint requests to: Radhe Mohan, Department of RadiationOncology, Medical College of Virginia, 401 College Street, Rich-mond, VA 23298. Tel: (804) 828-8451; Fax: (804) 828-6042;E-mail: [email protected]

Acknowledgments—This work is supported by the grant CA74043from the National Cancer Institute.

Accepted for publication 19 July 1999.

Int. J. Radiation Oncology Biol. Phys., Vol. 46, No. 1, pp. 195–205, 2000Copyright © 2000 Elsevier Science Inc.Printed in the USA. All rights reserved

0360-3016/00/$–see front matter

195

IMRT to produce dose distribution patterns that are muchcloser to the desired patterns than previously possible. Thepotential for improved radiation dose distributions withIMRT is so dramatic that even in its current, relativelyrudimentary form, it is rapidly emerging as an importanttool in clinical radiation therapy. In this paper, we describeour experience in designing IMRT plans for patients withhead and neck cancers in preparation for initiating intensity-modulated treatments. Because of the complex anatomy andnumerous tissue types encountered, intensity-modulatedtreatment design for head and neck (HN) carcinomas isparticularly challenging. We use an IMRT system designedand developed for research and clinical use. One of thepurposes of the paper was to determine the strengths andflexibility as well as limitations of IMRT, in particular, ourIMRT optimization methodology, to produce superior andefficiently deliverable dose distributions for such cases ascompared to the standard 3D approach. In our approach, thetreatments to the entire head and neck region, including theregions of microscopic and suspected disease, are deliveredusing multiple intensity-modulated fields incident fromfixed directions. Intensity modulation is produced with adynamic multileaf collimator (DMLC) employing the socalled “sweeping window” technique (4–8). This is incontrast with a commercial IMRT system (NOMOS Pea-cock system) which delivers IMRT using a temporally-modulating slit multileaf, a rotating gantry, and a translatingcouch. Our technique is also different from the techniqueadopted by other investigators (Boyeret al. [9], Ma et al.[10], and Xiaet al. [11]) who have adopted the “stop-and-shoot” (also called “step-and-shoot”) method for creatingintensity modulation, or Yuet al. (12), who have suggestedusing intensity-modulated arc therapy (IMAT).

While xerostomia is not life threatening, it may be quitedebilitating, in particular after radiotherapy of naso- ororopharyngeal carcinomas. The effective sparing of parotidglands will significantly improve quality of life and reducemorbidity. Attempts to design plans to spare parotids havebeen made in the past. A recent noteworthy example is thework of Eisbruchet al. (13), who used a beam’s eye viewtool of 3D conformal therapy to spare from direct radiationmost of one parotid gland. The main objective of the studypresented in this paper is to determine whether the use ofIMRT can lead to significant sparing of the parotids while atthe same time allowing escalation of radiation dose to thetumor.

Another important area that we addressed was the choiceof fractionation strategy for delivering IMRT. Conventionaltreatments are often divided into two (or sometimes evenmore) phases: an initial large-field treatment followed by aboost to a reduced volume. In the large-field treatment,radiation is delivered not only to the gross disease but alsoto the regions of microscopic extensions and suspecteddisease. Toward the end of the treatment course, treatmentvolumes are reduced, partly with the assumption that mi-croscopic disease requires a lower dose to control, andpartly to spare an overlapping or adjacent normal structure.

On the other hand, IMRT dose distributions are most con-formal when designed to be delivered in the form of a“simultaneous integrated boost” or SIB. In SIB strategy,volumes of regional, suspected, and gross disease are alltreated simultaneously. For each of the intensity-modulatedfields, radiation is delivered with a single sweep of MLCleaves across the entire treatment region. However, dose perfraction to each treatment volume may be quite differentfrom conventional regimens and may have potential conse-quences with regard to local control, as well as normaltissue complications, which must be considered.

METHODS AND MATERIALS

We designed IMRT plans for several patients with char-acteristic presentations of HN carcinomas. In this paper, wepresent details of IMRT plan design considerations for onecase and summarize important treatment plan data for threeothers. For the first case, we also show a comparison withthe 3D conformal treatment plan used to actually treat thepatient. Each of the patients presented unique challenges fornormal tissue sparing using conventional radiotherapy dueto the tumor extent and location. The sites selected includedsoft palate, maxillary sinus, base of tongue, and nasophar-ynx with and without lymphadenopathy for studies on theextent of conformality achievable with IMRT and on thepotential of parotid gland sparing for typical presentation ofHN carcinomas.

Patient 1 was a 33-year-old female with a 53 5 3 6-cmmuco-epidermoid carcinoma of the right soft palate, withextension into the soft tissues of right pharyngeal wall andthe nasopharynx. Based on clinical and CT evaluation, thetumor was staged T3N0M0. Patient 2 was a 61-year-oldmale with an advanced squamous cell carcinoma (SCC) ofthe left maxillary sinus. The tumor extended into the nasalcavity, the bilateral ethmoid and sphenoid sinuses, and theipsilateral orbit with involvement of extraocular muscles. Inthe absence of clinical or radiographic evidence of lymph-adenopathy, the tumor was staged T4N0M0 SCC of themaxillary sinus. Patient 3 was a 61-year-old female with a3-cm exophytic SCC of the left base of tongue. There wasalso a 2.53 2.5-cm level II lymph node in the left neck. Thetumor was staged T2N1M0 SCC of the left base of tongue.Patient 4 was a 67-year-old female with a 3.5-cm exophyticlymphoepithelioma involving the roof and posterior wall ofthe nasopharynx. A 63 9-cm mass of matted lymph nodeswas found in the left and a 43 6-cm group of nodes in theright neck. Both nodal groups extended from the subdigas-tric regions into the posterior neck. The primary tumorcaused a significant obstruction of the nasopharynx. Thetumor was staged T2N3M0 lymphoepithelioma of the na-sopharynx.

For the purpose of this study, the primary tumor and theinvolved lymph nodes were included in the gross tumorvolume (GTV). Where appropriate, all lymph node-bearingregions, levels I to VI, including the bilateral retropharyn-geal nodes, were mapped and specified as electively irradi-

196 I. J. Radiation Oncology● Biology ● Physics Volume 46, Number 1, 2000

ated regions. It may be relevant to note that the posteriorneck and supraclavicular fossae were included in IMRTplans, thus eliminating field junctions and obviating theneed for electron beams.

For patient 1, for whom the detailed analysis and com-parison with the 3DCRT is being presented, the objectivesof the 3D conformal treatment were to deliver 73 Gy to theplanning target volume (PTV) enclosing the gross diseaseand 46 Gy to the lymph node-bearing areas. The plan usedto treat the patient delivered 46 Gy in 23 fractions of 2 Gyto the gross tumor and all lymph node-bearing tissues, 16Gy of which were delivered using off-cord lateral photonportals and bilateral electron portals for posterior neck irra-diation. An integrated boost of an additional daily 0.25 Gywas initiated after the initial 10 Gy. Boost irradiation ofgross tumor with 2.25 Gy/fx was continued after 46 Gyusing multiple entry portals and shrinking fields to conformthe dose to the highest extent possible with the targetvolume. The total tumor dose of 73 Gy was delivered in 33fractions over 45 days. The doses to the spinal cord and thebrainstem were limited to 45 and 55 Gy, respectively. Theparotid glands were allowed to receive high doses to ensurethe delivery of adequate tumor dose. The 3DCRT planemployed 4-MV photon fields and 9-MeV electron fieldswhich included blocks, wedges, and compensators.

The IMRT plans were designed using an IMRT optimi-zation system developed by us as part of our research effortand coupled to a commercial 3D treatment planning system.The optimization process considers both the gross diseaseand the larger volume of occult, microscopic, or suspecteddisease simultaneously. Therefore, in contrast with standard3DCRT, regions of microscopic extensions of the disease,and nodal volumes, were also meticulously delineated. Ad-ditional contours were defined to provide margins for pen-umbra (due to lateral transport of radiation) around the PTVand for those critical structures which required additionalsafety margins to account for uncertainty in position. Thiswas necessary to allow the optimization process to automat-ically set the boundaries of beam apertures. It should benoted that typically the margins for penumbra for IMRT canbe substantially smaller than the corresponding margins forconventional treatments due to the ability of the former tosharpen beam boundaries (14).

Plans from 5 to 15 coplanar beams with equi-angularspacing were used to design IMRT plans with the objectiveof determining the minimum acceptable number. Followinga suggestion by Steinet al. (15), each configuration ofbeams included one incident beam from the posterior direc-tion (i.e., the direction with shortest distance to the cord).Steinet al.have demonstrated that such configurations leadto the most homogeneous target dose distributions for thesame normal structure dose. In contrast with standard ra-diotherapy, the fine-tuning of beam angles is not importantfor IMRT as long as a sufficient number of beams is used.Beam angle optimization is complex and time consuming,and is probably unnecessary.

Next, intensity distributions within the radiation fields of

each of these beams were optimized. In its current state, ouroptimization system employs dose- and/or dose-volume ob-jective function to iteratively adjust the intensity of each rayof each beam based on the effect of each such adjustment onthe value of the objective function. A somewhat simplifiedmathematical representation of the criteria we employed is:

S5 Oi

~DT,0 2 DTi!2 1 O

n

Pn 3 Oj

H~Dn,0 2 Dnj!

3 ~Dn,0 2 Dn, j!2 (1)

The aim is to minimize the treatment plan scoreS repre-sented by this sum in whichDT,0 is the desired dose to thetarget volume andDn,0 is the tolerance dose of thenthcritical normal structure.DT,i is the computed dose in theithvoxel of the target andDn,j is the computed dose at thejthvoxel of thenth normal structure. For normal organs, thefunction H(Dn,0 2 Dn,j) is defined as follows:

H~Dn,0 2 Dn, j! 5 0 for Dn, j # Dn,0 and

5 1 for Dn, j . Dn,0 (2)

In other words, as long as the dose in a normal structurevoxel does not exceed the tolerance limit, the voxel does notcontribute to the score. The quantityPn is the “relativepenalty” for exceeding the tolerance dose.

For dose-based criteria, we specified the prescriptiondose and upper and lower limits for the target dose. Inprinciple, all three may be the same. In addition, we spec-ified a penalty for dose values above the upper limit and adifferent penalty for dose values below the lower limit. Wespecified these data for the volume of gross disease as wellas regions of microscopic and suspected disease. For eachorgan at risk, as well as nonspecific tissues outside the targetvolume, we specified the upper dose limit and the penaltyfor exceeding it.

For dose-volume based criteria, in addition to dose limits,we also specified limits to the volumes as well. For the PTVof gross disease, for example, we specified that at least 98%of the volume should receive 70 Gy or more, and that nomore than 5% of the volume should exceed 81 Gy. Forparotid glands, as another example, we specified that nomore than 25% of the volume should receive dose greaterthan 30 Gy. Of course, as discussed later, what we actuallyachieve is a trade-off among various requirements, and isoften different from the specified limits. The objective func-tion of the type represented by Eqs. 1 and 2 is used for notonly dose-based criteria but, with appropriate modificationsof the algorithm following a suggestion by Bortfeld (16),also for dose-volume based criteria. For the latter, only asubset of voxels receiving higher than the specified dose isincluded in the computation of the score.

The optimization of ray intensities is carried out itera-tively using the gradient technique. The dose in every voxelis calculated for an initial set of weights of rays using the

197Potential for sparing of parotids with IMRT● Q. WU et al.

dose computation engine of the host treatment planningsystem. The dose computation system we use employs thesuperposition of dose spread kernels (17, 18). A significantadvantage of this method is that it is able to correct forinternal inhomogeneities and surface irregularities more ac-curately than other conventional methods, a feature that isimportant for high precision treatments of HN cancers.

The optimized intensity distributions within each of theset of beams thus obtained are used to predict the radiationdose distributions that would be deposited in the patient atthe time of treatment. Dose distributions are evaluated in thecustomary fashion. Typically, they are displayed in threeorthogonal planes superimposed on the patient’s CT im-ages. The displays are supplemented with dose-volume his-tograms (DVHs) and additional quantitative information.

We debated two different fractionation strategies whileattempting to design head and neck IMRT plans. As men-tioned earlier, IMRT dose distributions are best when de-signed to be delivered in the form of an SIB. In such afractionation strategy, the intensity-modulated dose distri-bution remains the same for all fractions of the treatmentcourse. Every fraction contains a higher dose region sur-rounded by a lower dose region, and radiation is deliveredwith a single sweep of MLC leaves across the volume ofregional disease and gross disease for each of the intensity-modulated fields. Alternatively, one could consider dividingthe treatment into two phases in the same manner as forconventional treatments. A major portion of the dose wouldbe delivered in the primary phase using uniform fieldsdesigned with standard 3D conformal methods. This wouldbe followed by an IMRT boost treatment to shape the dosedistributions so that the total dose at every point is as nearlyas possible the same as for SIB treatment. We can refer tothis strategy as “IMRT-boost-only” treatment. The totaldose delivered with the IMRT-boost-only treatment may beapportioned into the large-field phase and the IMRT-boostphase in any number of ways.

One can give arguments in favor of both strategies (seeResults and Discussion section). We have chosen the SIBstrategy for all HN carcinomas due primarily to the superi-ority of its dose distributions, but we must consider radio-biological aspects of altered fractionation for tumors as wellas normal tissues. Following Witherset al. (19), we definethe terms normalized total dose (NTD) as the biologicallyequivalent total dose, normalized to 2 Gy/fx. We also usethe term “nominal dose” to define the actual physical dose.We used the linear-quadratic iso-effect model, includingaccelerated repopulation of the tumor, anda/b values andclonogen doubling times, derived from the analyses of Ma-ciejewskiet al.and Witherset al.(19–21), to convert NTDsinto nominal doses, and vice versa. For a given set ofprescribed nominal doses, we calculated nominal doses, orvice versa, for the gross tumor volume, regional disease, andnodes for the 3DCRT and various IMRT fractionation op-tions. For example, we considered the option of continuingto give the standard 2 Gy/fx to the gross tumor volume. Inmost cases, this would lead to plans in which the volume of

suspected microscopic disease would receive an unaccept-ably low dose (,1.8 Gy) for every fraction but over a longerperiod of time. Radiobiologically, this may be ineffective incontrolling the nodal disease. As an alternative, we consid-ered delivering higher than 2 Gy/fx to the gross disease inorder to give 2 Gy/fx to nodes. This option should lead tohigher probability of tumor control for the same nominaldose based on radiobiological principles. There are manyin-between options also. For instance, dose of 1.8 Gy/fx tothe nodes may be quite acceptable and may not requireexcessive increase in dose per fraction to the gross disease.

Different fractionation strategies would result in differentdoses per fraction to normal critical structures as well. In allIMRT plans designed, normal tissues outside the targetvolume received not only lower nominal doses but alsolower doses per fraction compared to 3DCRT plans, pre-sumably reducing the probability of injury. On the otherhand, the risk of a larger fraction size to the normal tissuesembedded within the primary tumor volume, for instancemucosa, must be considered. The radiobiological parame-ters for predicting the effect of the response of normaltissues are not well established. Nevertheless, as reported inthe next section, we have made some assumptions and usedtypical values of the parameters for late-responding tissuesto estimate the potential effect of SIB fractionation and doseescalation on the embedded tissues.

RESULTS AND DISCUSSION

In this section, we present and discuss results of applyingthe methods described above to four cases of head and neckcarcinoma. We describe detailed data for patient 1, includ-ing comparison with the 3D conformal treatment plan. Forthe three other cases, we only summarize the IMRT data.

Figure 1 shows transverse and sagittal views of the 3Dconformal treatment plan for patient 1 and two 9-fieldIMRT plans, and Figs. 2 and 3 show the correspondingDVHs. The nine coplanar beams of the IMRT plan werearranged in equi-angular steps around the target volume.The IMRT plan was obtained employing dose-volume op-timization criteria. Ninety-eight percent of the PTV receives70 Gy in the IMRT plan and 73 Gy in the 3DCRT plan. Ourcalculations of biologically equivalent NTD showed that theformer is effectively greater because it is intended to bedelivered in a smaller number of fractions. (See Table 2 formore details.) However, both provide the requisite cover-age. In the 3DCRT plan, the electively treated nodes receive30 Gy from photon beams plus 16 Gy from electron beams,for a total of 46 Gy. For the IMRT plan, they receive 46 Gyto 95% of the volume and 50 Gy to 90% of the volume. Thiswas considered clinically satisfactory. DVHs for the spinalcord and brainstem are different in the two plans, but bothmet the specified criteria. The major differences, as seen inboth the dose distribution displays as well as the DVHs, arefor parotids. In the 3DCRT plan, almost 95% of the parotidvolume receives in excess of 46 Gy. According to datapublished by Emamiet al. (22), such doses would almost


certainly lead to xerostomia. On the other hand, accordingto the same data, if 50% of the parotid volume could bemaintained below 32 Gy, there would be minimal probabil-ity of injury. Figure 2 shows that this dose-volume objectivewas indeed met for the IMRT plan.

Although the IMRT plan shown above was consideredexcellent, there was concern that the 45-Gy line was tooclose to the spinal cord. Therefore, the original set ofcontours around the cord was distended by 0.5 cm and the

Fig. 1. Transverse and sagittal views of dose distributions comparing (a) the standard 3DCRT plan with (b) an IMRTplan and (c) another IMRT plan, in which a margin of 0.5 cm around the cord has been added. All plans are for 4 MV.The IMRT plans employed nine coplanar, noncollinear beams placed at equiangular steps. The IMRT optimizationcriteria was dose-volume based.

Fig. 2. DVHs corresponding to plans (a) and (b) of Fig. 1 (i.e., the3DCRT plan and the IMRT plan with no margin around the cord).The most significant difference is in the sparing of parotids in theIMRT plan.

Fig. 3. DVHs corresponding to plans (b) and (c) of Fig. 1 (i.e., theIMRT plans without and with a margin of 0.5 cm around the cord).There is a slight improvement in the sparing of the cord. Note thatall calculations were carried out assuming no uncertainty in thepositioning of the patient. A greater gain would have been ob-served if the positioning uncertainty had been incorporated.


plan was re-optimized. Figure 1c shows the new plan. Thecorresponding DVHs are shown in Fig. 3. This change hadvirtually no apparent impact on the plan other than a slightimprovement in cord sparing. If positioning uncertainty hadbeen incorporated in the computation of dose distributionsand DVHs, we might have observed a significant improve-ment.

Figure 4 compares the dose-volume-based IMRT planshown in Figure 1b with the one produced with dose-basedobjectives. Figure 5 shows the corresponding DVHs indi-cating modest improvement in parotid and cord sparing. Wehave observed that, in general, the use of dose-volumebased criteria not only produces better dose distributions butalso requires less trial-and-error than the purely dose-basedcriteria. This may be explained as follows. When the criteriaare specified in terms of dose or dose-volume limits, thelimits specified are almost always different from the valuesactually desired or achieved. The reason for this is that theoptimization process attempts to counterbalance the desiredtumor dose against desired doses to normal structures, and

desired normal structure doses against each other. The spec-ified dose limits and penalty values are adjusted in a trial-and-error fashion to obtain desired results. If the criteria arespecified in terms of dose-volume limits, the number ofpossible solutions increases infinitely. This is due to the factthat there is no restriction as to which volume can receivethe specified dose, no upper limit on the dose to high-dosevolume or no lower limit on the dose to the low-dosevolume. Therefore, the use of dose-volume based criteriamakes the solution to the optimization problem much moreachievable, giving the optimization process more flexibility.

As indicated earlier, we designed plans with a range ofnumber of beams in an attempt to find the minimum suitablenumber. Figure 6 compares 5-, 7-, 9-, and 15-beam IMRTplans for patient 1. All plans were developed using dose-volume criteria and coplanar, non-collinear beams placed atequi-angular steps. All plans employed the same optimiza-tion criteria. From the examination of dose distributiondisplays and the corresponding DVHs (two of which areshown in Fig. 7), it is apparent that there is no perceptibleimprovement in the quality of the plan when the number ofbeams is increased from 9 to 15. However, there is somedegradation when the number is reduced from 9 to 7. Thedegradation is quite significant when the number is reducedto 5. Because normal tissues were constrained strongly(high values of penalties) so as not to exceed dose limits tospecified volumes, the degradation occurs mainly in thetarget dose distribution. In general, we have found that, dueto their complexity, most head and neck cases require 9equispaced beams. In rare, simple cases 7 beams are suffi-cient, and in some unusually complex cases a higher num-ber, perhaps as many as 15, may be required. In all of theHN IMRT plans we have designed thus far, including the 4

Fig. 4. Transverse views of dose distributions comparing IMRT plans designed using dose-based and dose-volume-based optimization criteria. Dose-volume based criteria yields a somewhat better dose distribution and was found torequire less trial-and-error.

Fig. 5. DVHs comparing IMRT plans designed with dose- anddose-volume-based criteria and shown in Fig. 4.


presented here, in spite of their highly variable complexity,we have found that 9 beams are quite sufficient. Therefore,as a matter of practice, we start the HN IMRT planningprocess with 9 beams and increase the number only if asatisfactory plan cannot be achieved.

Beam configuration can have a significant impact on thequality of an optimized intensity-modulated treatment plan.It may be intuitively apparent that, in principle, a highernumber of beams produces a better plan. Further, because ofthe greater control over dose distributions afforded by op-timized intensity modulation, fewer intensity-modulatedbeams may be needed than uniform beams to achieve thesame or even higher and more homogeneous tumor dose forthe same or lower level of normal tissue exposure. More-over, it is likely that even fewer intensity-modulated beamsmay be needed if beam angles are optimized than if beamsare placed at equi-angular steps. Unfortunately, beam angle

optimization is a nontrivial task. Fortuitously, it is notimportant as long as there are sufficient number of beams.We also observed that when small adjustments to beamangles are made to avoid obstructions such as couch bars,plans with larger number of beams (9 or more) are notmaterially affected.

From a practical standpoint, it is not essential to reducethe number of beams to too few. Because IMRT treatmentsare delivered automatically or under remote control using adynamic multileaf collimator, there is no need to enter theroom between treatments. Therefore, a 5-beam treatment todeliver 200 cGy to the tumor can be delivered in almost thesame time as a 15-beam treatment to deliver the same dose.However, a larger number of beams means additional effortfor dosimetric verification of the IMRT plan. Therefore, asuitable compromise is desirable.

IMRT inherently possesses the capability to sharpen

Fig. 6. Dose distributions demonstrating the improvement in the quality of IMRT plans with increasing number of beamsfrom 5 to 15. In all cases, dose-volume-based criteria were employed, beams were coplanar and placed at equiangularsteps.


beam boundaries and reduce the margins assigned for thepenumbra. The optimization process accomplishes this byappropriately increasing the intensity of rays just inside thebeam boundaries to compensate for the dose deficit causedby lateral transport of radiation. The beam boundary sharp-ening and margin reduction feature of IMRT can be takenadvantage of only if the dose computation method is able toadequately take into account the lateral transport of radia-tion. All plans shown here were planned in this manner.

We should note here that the property of IMRT thatallows sharpening of the penumbra would also lead to amore homogeneous dose distribution within the target vol-ume. There seems to be an erroneous impression amongsome that IMRT inherently produces inhomogeneous dosedistribution within the target volume. This is exactly theopposite of what IMRT is capable of in principle. Inhomo-geneity observed may be due to one of two reasons: theinability of the IMRT optimization system to take adequateaccounting of lateral transport of radiation, or due to theoverriding need to partially or wholly protect one or morecritical organs. If all things were equal, the IMRT planshould produce more homogeneous dose distribution than aplan designed with uniform beams.

Table 1 summarizes the IMRT planning results for allfour patients. It contains the data we typically use for theevaluation of treatment plans. Columns 6 and 7 indicate theextent of tumor coverage and the last four columns show the

adequacy of parotid gland sparing. Column 6 contains thenominal prescribed dose and column 7 contains the valuesof the corresponding equivalent uniform dose (or EUD).EUD is the dose which, if given uniformly to the targetvolume, leads to the same cell kill as the nonuniform actualdose distribution. EUD is calculated using the linear-qua-dratic model and has been shown to correlate well withtumor response (23). Columns 8 and 9 give the fractions ofleft and right parotids receiving dose greater than 32 Gy,and column 10 gives the fraction of the combined volume ofparotids receiving greater than 32 Gy. Column 11 showsparotid NTCP values computed using Lyman’s model (24)and the parameters deduced by Burmanet al. (25) from thedata of Emamiet al. (22). It is interesting that, regardless ofthe extent and location of the disease, parotid glands seemto be spared equally well in all cases. Target coverageappears to be adequate for all cases except one; patient 2,who had received prior chemotherapy. The prescriptiondose for this patient was 60 Gy in contrast with 70 Gy forthe other three patients. The target volume was very largeand extended to the skin surface. For both the 3DCRT andIMRT plans, even with a 1-cm bolus, it was not possible toraise the dose to the portion of the tumor near the surface tothe desired level for the 6-MV beams used.

As emphasized in the Methods and Materials section,IMRT dose distributions are best when designed to bedelivered in the form of simultaneous integrated boosts.

Fig. 7. DVHs comparing (a) a 15-beam IMRT plan with a 9-beam plan, and (b) a 5-beam plan with a 9-beam plan.

Table 1. Summary of IMRT planning data for four cases of head and neck carcinomas of different subsites and tumor sizes

1 2 3 4 5 6 7 8 9 10 11

Pat. # Site StageTumorvol. (cc) Rx (Gy)

Dose to98% vol.

(Gy)EUD(Gy)

% Lt.parotid vol. 32 Gy

% Rt.parotid vol. 32 Gy

% Parotidsvol. . 32 Gy

ParotidsNTCP (%)

1 Oropharynx T3N0M0 124 70 70.4 74 24 33 29 22 Lt. maxil. sinus T4N0M0 488.5 60 55.5 58.4 52 0 28 33 Base of tongue T2N1M0 15.61 70 72.2 74.3 39 39 39 64 Nasopharynx T2N3M0 110.6 70 70.3 73.3 32 34 33 5


However, radiobiological aspects of such a strategy must beevaluated. As an illustration of how one can estimate theprescribed nominal doses for a given NTD for the tumor,and vice versa, and to evaluate potential effects on normaltissues, we show in Table 2 the results of our calculationsusing the linear-quadratic iso-effect model. We chose pa-tient 1 for the purposes of the sample results shown. Theparameters of the model (a/b 5 15 and clonogen doublingtime 5 3–5 days) were obtained from the analysis of headand neck data by Maciejewskiet al. and Witherset al.(19–21). We adjusted the doubling time to 3.7 days to make46 Gy in 23 fractions equivalent to 50.4 Gy in 28 fractions.The first data group in the table corresponds to the 3DCRTplan. The total dose to the tumor volume is 73 Gy in 5fractions of 2 Gy and 28 fractions of 2.25 Gy. This resultedin a biologically equivalent NTD (delivered in 2 Gy/fx) of78 Gy. The electively-treated nodes receive 46 Gy in 23fractions. For SIB-IMRT, four different options were exam-ined. In the first option (IMRT 1), the required dose is to bedelivered in 23 fractions. According to the iso-effect model,

to deliver biologically equivalent of the same NTD of 78Gy, we would need to deliver a nominal dose of 63.3 Gy in23 fractions of 2.75 Gy. The nodes would still receive thesame 46 Gy with 23 fractions. In the second option, therequired dose would be delivered in 28 fractions. Thisoption would require 68.2 Gy at a rate of 2.44 Gy/fx toachieve the NTD of 78 Gy to the primary. However, thenodes would receive dose at 1.8 Gy/fx, and therefore wouldrequire 50.4 Gy to reach the required biologically equivalentNTD of 46 Gy. Similarly, in IMRT 3 option, the nominalequivalent of NTD of 78 Gy to the primary disease in 30fractions is 69.7, whereas the equivalent of 46 Gy to theelectively-treated nodes is 51.7 Gy, with fraction size of1.72. In the fourth option, 70 Gy would be delivered to thetumor volume in 28 fractions of 2.5 Gy each, correspondingto the NTD of 82 Gy. This would represent an escalation ofthe gross tumor dose. The nodes would receive 50.4 Gy in28 fractions in this option also. Thus, there are many op-tions. If the intent is not to escalate dose, one might beinclined to accept the IMRT 2 option simply because theelectively-treated nodes do not receive less than 1.8 Gy/fx,for which there is precedence and experience; or, to useoption IMRT 4, if the intent is to escalate the dose.

Concerning the effects of SIB-IMRT on normal tissues,we should point out that for parotid glands, the biologicallyequivalent doses would be lower than the nominal doses dueto smaller dose per fraction locally. For the spinal cord andother normal tissues outside the target volume also, even ifthe total nominal dose is the same for all regimens, thelower doses per fraction in IMRT strategies may provide anadvantage. For this reason, calculation of biologicallyequivalent doses for these tissues was not considered nec-essary. On the other hand, for the tissues embedded withinthe primary target volume, it is important to know thepotential consequences of SIB-IMRT strategy. Unfortu-nately, the data to make such estimates are not too reliable.In any case, we applied the iso-effect model to compute theNTD for the 3DCRT plans and all the IMRT options shown,assuming that the embedded normal tissues do not regen-erate and do not change their sensitivity during the course oftreatment. We also assumed that thea/b value of 2.5 Gy,corresponding to the value for late responding tissues, isapplicable to embedded tissues. The last row (“late effecttissues”) in each data group of Table 2 shows the estimatesof NTD for each fractionation regimen and indicates that forthe same NTD (78 Gy), IMRT plans may lead to a slightlylower embedded normal tissue NTD than the 3DCRT plan,even though the number of fractions is smaller. This ispresumably due to the fact that the nominal dose is alsolower. However, when the dose is escalated from 78 to 82Gy (NTD), the NTD of the embedded normal tissues in-creases from 74.9 to 77.9 Gy.

Obviously, calculations of the type discussed above donot have a strong foundation; however, they are useful as ameans of designing fractionation strategies for use in clin-ical protocols and trials with SIB-IMRT. They also indicatethe caution one must observe with regard to embeddednormal tissues while attempting to escalate dose.

Table 2. Nominal doses and biologically-equivalent normalizeddoses (NTDs) for the primary tumor, nodal volume, and late-

effect tissues within the tumor volume

Nominaldose

Numberof fx Dose/fx NTD

Standard 3DCRTTumor volume 73 5, 28 2, 2.25 78Nodes 46 23 2 46Parotids 65 35 1.86 ;65Spinal cord 46 23 2 46Late effect tissues 73 33 2.21 76.5

IMRT 1Tumor volume 63.3 23 2.75 78Nodes 46 23 2 46Parotids 31 23 1.3 , 31Spinal cord 46 23 2 46Late effect tissues 63.3 23 2.75 74.1

IMRT 2Tumor volume 68.2 28 2.44 78Nodes 50.4 28 1.8 46Parotids 31 28 1.1 , 31Spinal cord 46 28 1.65 ;46Late effect tissues 68.2 28 2.44 74.9



Calculations were carried out using the linear-quadratic iso-effect model, including the effect of accelerated repopulation ofthe tumor, for the 3DCRT plan and for four different IMRTfractionation options. In options IMRT 1, IMRT 2, and IMRT 3,the NTD for the primary tumor is the same as for the 3DCRT plan.Option IMRT 4 represents an escalation of the tumor dose.


The simultaneous integrated boost strategy is a significantdeparture from conventional experience. Conventional ra-diotherapy experience is based on giving a high uniformdose to the target volume for every fraction and then reduc-ing the volume for the boost portion. In principle, one canalso design IMRT for the boost portion only, and therewould be an inclination to follow the convention unless aclear advantage of the SIB strategy can be demonstrated,perhaps in terms of significantly superior dose distributionsor a significant potential for improved therapeutic ratio.That, in fact, would usually be the case. If a large fractionof the dose has already been delivered using uniform beamsin the primary phase of the two-phase IMRT-boost-onlystrategy, it may be very difficult or impossible to achieve thedesired pattern of total dose with the remaining fractionswith IMRT. We should also keep in mind the fact that theSIB strategy is a more efficient way of planning and deliv-ering IMRT, because it involves the use of the same plan forthe entire course of treatment. Furthermore, there may be anadvantage in terms of higher biologically-effective tumordose and/or lower biologically-effective dose normal tissuesoutside the tumor volume. Potential consequences for tis-sues embedded within the tumor volume need to be inves-tigated more thoroughly. More detailed radiobiologicalanalysis of fractionation strategies will be presented inanother report.

CONCLUSION

We have presented IMRT plans for four patients with HNcarcinomas. Our ongoing work with a range of cases of HNcarcinomas with different sizes, shapes, and locations of

tumor volumes and proximity to critical normal structuresindicates that we can achieve similar results in almost allcases. We conclude that IMRT is an effective tool fordesigning highly conformal HN treatment plans in whichsignificant fractions of the parotid glands can be spared sothat the probability of xerostomia is reduced, and that thiscan be achieved without compromising tumor dose or doseto other normal tissues, such as cord or brainstem. Further-more, because IMRT dose distributions are most conformalwhen designed to be delivered as SIB, the same plan can beused for the entire course of treatment, and the need forelectron beams is obviated. There is no need to enter thetreatment room between fields to change the accessories,allowing the process to be automated. These factors makethe planning and delivery of these treatments more efficient.Because the large field and boost doses are delivered in thesame number of fractions, there could be biological advan-tages. For the same or lower doses as well as lower dose perfraction to normal tissues outside the target volume, higherdose per fraction could be delivered to the tumor, effectivelyescalating biologically equivalent dose to the tumor, andpotentially improving local control. However, dose to nor-mal tissues embedded within the target volume may presenta hindrance to dose escalation and will need to be furtherexamined. Using the methodology presented in this paper,we have designed clinical protocols to escalate the dose insmall steps as well as to quantitate the effects of improvedparotid sparing. Based on the analysis presented here, webelieve that, on the whole, IMRT has a significant potentialfor improving the outcome for patients with HN carcino-mas, and have already started applying these IMRT tech-niques for such cases at our institution.

REFERENCES

1. Ling CC, Burman C, Chui CS,et al. Conformal radiationtreatment of prostate cancer using inversely-planned intensity-modulated photon beams produced with dynamic multileafcollimation. Int J Radiat Oncol Biol Phys1996;35:721–730.

2. Ling CC, Burman C, Chui CS,et al.Implementation of photonIMRT with dynamic MLC for the treatment of prostate cancer.In: Sternick ES, editor. The theory & practice of intensity-modulated radiation therapy. Madison, WI: Advanced Medi-cal Publishing; 1997. p. 219–228.

3. Carol MP, Peacock TM. A system for planning and rotationaldelivery of intensity-modulated fields.Int J Imaging SystTechnol1995;6:56–61.

4. Bortfeld T, Kahler DL, Waldron TJ,et al. X-ray field com-pensation with multileaf collimators.Int J Radiat Oncol BiolPhys1994;28:723–730.

5. Convery DJ, Rosenbloom ME. The generation of intensity-modulated fields for conformal radiotherapy by dynamic col-limation. Phys Med Biol1992;37:1359–1374.

6. Spirou SV, Chui CS. Generation of arbitrary fluence profilesby dynamic jaws or multileaf collimators.Med Phys1994;21:1031–1041.

7. Stein J, Bortfeld T, Doerschel B,et al. Dynamic x-ray com-pensation for conformal radiotherapy by means of multi-leafcollimation.Radiother Oncol1994;32:163–173.

8. Wang X, Spirou S, LoSasso T,et al.Dosimetric verification of

an intensity modulated treatment.Med Phys1996;23:317–328.

9. Boyer AL, Geis P, Grant W,et al.Modulated beam conformaltherapy for head and neck tumors.Int J Radiat Oncol BiolPhys1997;39:227–236.

10. Ma L, Boyer A, Xing L,et al. An optimized leaf-sequencingalgorithm for beam intensity modulation using dynamic mul-tileaf collimators.Phys Med Biol1998;43:1629–1643.

11. Xia P, Verhey LJ. Multileaf collimator leaf sequencing algo-rithm for intensity modulated beams with multiple static seg-ments.Med Phys1998;26:1424–1434.

12. Yu CX. Intensity-modulated arc therapy with dynamic mul-tileaf collimation: An alternative to tomotherapy.Phys MedBiol 1995;40:1435–1449.

13. Eisbruch A, Ship JA, Martel MK,et al. Parotid gland sparingin patients undergoing bilateral head and neck irradiation:Techniques and early results.Int J Radiat Oncol Biol Phys1996;36:469–480.

14. Mohan R, Wu Q, Wang X-H,et al. Intensity modulationoptimization, lateral transport of radiation and margins.MedPhys1996;23:2011–2022.

15. Stein J, Mohan R, Wang X-H,et al. Number and orientationsof beams in intensity-modulated radiation treatments.MedPhys1997;24:149–160.

16. Bortfeld T, Stein J, Preiser K. Clinically relevant intensitymodulation optimization using physical objectives. Leavitt


DD, editor. Proceedings of XII International Conference onthe Use of Computers in Radiotherapy, Salt Lake City, UT,1997. Madison, WI: Medical Physics Publishing. pp. 1–4.

17. Mackie TR, Scrimger JW, Battista JJ. A convolution method ofcalculating dose for 15-MV X-rays.Med Phys1985;12:188–196.

18. Mackie TR, Bielajew AF, Rogers DWO,et al. Generation ofphoton energy deposition kernels using the EGS Monte Carlocode.Phys Med Biol1988;33:1–20.

19. Withers HR, Taylor JMG, Maciejewski B. The hazard ofaccelerated tumor clogen repopulation during radiotherapy.Acta Oncol1988;27:131–146.

20. Maciejewski B, Withers HR, Taylor JMJ,et al.Dose fraction-ation and regeneration in radiotherapy for cancer of the oralcavity and oropharynx: Tumor dose-response and repopula-tion. Int J Radiat Oncol Biol Phys1989;16:831–843.

21. Ang KK, Thames HD, Peters LJ. Altered fractionation sched-ules. In: Perez CA, Brady LW, editors. Principles and practiceof radiation oncology. Philadelphia: Lippincott-Raven; 1997.p. 119–142.

22. Emami B, Lyman J, Brown A,et al. Tolerance of normaltissue to therapeutic radiation.Int J Radiat Oncol Biol Phys1991;21:109–122.

23. Niemierko A. Reporting and analyzing dose distributions: Aconcept of equivalent uniform dose.Med Phys1997;24:103–110.

24. Lyman JT. Complication probability as assessed from dosevolume histograms.Radiat Res1985;104:S13–S19.

25. Burman C, Kutcher GJ, Emami B,et al. Fitting of normaltissue tolerance data to an analytic function.Int J RadiatOncol Biol Phys1991;21:123–135.


The potential for sparing of parotids and escalation of biologically effective dose with...

Documents

Transcript of The potential for sparing of parotids and escalation of biologically effective dose with...