PII S0360-3016(01)02599-8

PHYSICS CONTRIBUTION

THE INFLUENCE OF INTERPATIENT AND INTRAPATIENT RECTUMVARIATION ON EXTERNAL BEAM TREATMENT OF PROSTATE CANCER

DI YAN, D.SC., BENHUA XU, M.D., DAVID LOCKMAN, D.SC., KAMAL KOTA, M.S.,DONALD S. BRABBINS, M.D., JOHN WONG, PH.D., AND ALUARO A. MARTINEZ, M.D., F.A.C.R.

Department of Radiation Oncology, William Beaumont Hospital, Royal Oak, MI

Purpose: The rectal dose/volume relationship and inherent variations thereof are fundamental parameters toguide dose escalation in prostate cancer treatment. This study evaluates the effect of rectal dose/volume variationon the risk of rectal complication for different planning target volume (PTV) constructions.Methods and Materials: Thirty prostate patients with multiple daily CT scans obtained during the treatmentcourse were included in this retrospective study. The dose distribution was calculated based on the pretreatmentCT image alone. Treatment plans were generated by applying the four-field-box beam arrangement to each ofthree different PTVs: PTVs with 0.5-cm and 1.0-cm uniform margins, and a patient-specific PTV constructedusing treatment imaging feedback. For each of the 30 patients, the rectal wall as manifested on each of multipleCT images was delineated after image bony registration to the pretreatment CT image, and applied to thecorresponding treatment plan to obtain the rectal wall dose–volume histogram (DVH). Interpatient and intrapa-tient rectal dose/volume variations were quantified accordingly. The corresponding uncertainty and sensitivity ofthe risk of rectal complication to the variations were evaluated for each of the three PTVs. Finally, the efficacyof using multiple CT images to reduce uncertainty in planning evaluation was examined.Results: Sensitivity of the risk of rectal complication to rectal dose/volume variation strongly depends on theclinical target volume (CTV)-to-PTV margin or prescription dose, or both. Compared to the conventionaltwo-dimensional (2D) prostate cancer treatment, the sensitivity for a conformal treatment can be 3 times higheror more. Due to the interpatient rectal dose/volume variation, the individual normal tissue complicationprobability (NTCP) was distributed from 10% to 37% when a common prescription dose was applied for allpatients. The intrapatient rectal dose/volume variation introduces at least � 25% uncertainty in the NTCPcalculation for at least 10% or 25% of the patients treated with the PTV of 1.0- or 0.5-cm margin, respectively.These uncertainties were larger for the smaller PTV, with the standard deviation up to 20%. By applyingmultiple CT image feedback, the NTCP uncertainty could be reduced by a factor of 2.Conclusions: Shape and position variation of rectum has less influence on treatment planning in the conventional2D treatment of prostate cancer. However, this influence is quickly growing with high treatment dose or smallCTV-to-PTV margins. To reduce the variation and uncertainties in the treatment planning evaluation associatedwith the inter- and intrapatient rectal dose/volume variation, the iso-NTCP model and treatment image feedbacktechnique can be applied in dose escalation trials of prostate cancer treatment. © 2001 Elsevier Science Inc.

Rectal dose, Volume variation, Sensitivity of rectal NTCP variation, Uncertainty of rectal NTCP evaluation.

INTRODUCTION

Rectal toxicity is the major limiting factor for dose escala-tion in external beam treatment of prostate cancer. In mostclinical dose escalation trials, a single prescription dose hasbeen applied for patients who have similar clinicopathologiccriteria (1–3). This prescription dose is generally deter-mined though a clinical trial and error process, based on theobserved clinical dose response of critical normal tissues(4–6). Because treatment techniques in these clinical trials

are quite different, one can hardly apply the same prescrip-tion dose from one special treatment technique to others.Correlation between rectal toxicity and dose/volume param-eters has been performed recently (6–9). These studiesprovide very useful guidance for future dose escalation,specifically when different treatment techniques will beapplied. However, current correlation studies rely on arectal dose/volume relationship calculated based on singlestatic planning computed tomography (CT) image alone.Therefore, the rectal dose/volume variation during the entire

Reprint requests to: Di Yan, D.Sc., Department of RadiationOncology, William Beaumont Hospital, 3601 W. Thirteen MileRd., Royal Oak, MI 48073-6769. Tel: (248) 551-6589; Fax: (248)551-3784; E-mail: dyan@beaumont.edu

Presented in part at the 42nd ASTRO meeting, Boston, October2000.

This work is supported in part by NCI Grant CA71785.Received March 9, 2001, and in revised form June 5, 2001.

Accepted for publication August 15, 2001.

Int. J. Radiation Oncology Biol. Phys., Vol. 51, No. 4, pp. 1111–1119, 2001Copyright © 2001 Elsevier Science Inc.Printed in the USA. All rights reserved

0360-3016/01/$–see front matter

1111

course of treatment due to inherent rectum motion and beamplacement error has not been considered in these studies.

Shape and position variation of the rectum during thetreatment course has been observed and discussed in previ-ous studies (10–12). Utilizing biweekly CT imaging duringthe course of prostate cancer treatment for 11 patients,Lebesque et al. (10) have accomplished a detailed analysison the interpatient and intrapatient rectal wall volume anddose variation. The corresponding normal tissue complica-tion probability (NTCP) variation has also been evaluatedfor their specific treatment technique and two levels ofprescribed doses.

In the following study, similar data analysis will berepeated by using the standard four-field-box treatmenttechnique applied to three different planning target volume(PTV) constructions. In addition, sensitivity of the risk ofrectal complication to inter- and intrapatient rectal dose/volume variation will be carefully investigated with respectto each PTV and the corresponding prescription dose. Fi-nally, we will prove that using an isotoxicity guided doseprescription strategy and treatment image feedback, uncer-tainty in the calculated risk of rectal complication, inducedby interpatient and intrapatient rectal dose/volume varia-tion, can be considerably reduced.

METHODS AND MATERIALS

Data of rectal wall volumesMultiple CT images for each of 30 patients with prostate

cancer were included in this retrospective study. Each pa-tient had one pretreatment CT image and 12 to 18 repeat CTimages (average 17 per patient) obtained either before orafter daily treatment. All CT scans were acquired with thesame patient setup procedure and immobilization (supinewith a “w” -shaped universal leg pillow), and 3-mm sliceintervals. The pretreatment CT scan was obtained 1 to 2weeks before the treatment with 20 cc urethra contrast. Thetreatment CT images were acquired 4 times in the first week oftreatment, and twice per week for subsequent treatment weeks.

Treatment CT images of each patient were registered tothe pretreatment CT image with respect to pelvic bonystructure. After the registration, the rectal wall manifestedon the CT image was contoured by a single observer.Previously defined rules of rectal wall definition (10) havebeen adapted in this study. The cranial border of the rectumwas defined as the location where the rectum turned hori-zontally into the sigmoid colon if it was at a level inferior tothe caudal border of the sacroiliac joint. Otherwise, it wasdefined at the level of the caudal border of the sacroiliacjoint. The caudal border of the rectum was defined at thebottom of the ischial tuberosities. After completing thecontour of outer rectal wall, the inner rectal wall on each CTslice was delineated by the observer based on the rule thatthe cross-sectional area of any fixed slice of the rectal wallis the same among the multiple CTs of each patient.

In this study, we assumed that the rectal wall volume ofeach patient was constant during the whole treatment

course. Therefore, we intended to contour the rectal walls ofeach patient such that they had a similar volume. The meanrectal wall volume of each patient was calculated and usedas this standard volume. For each patient, deviation betweenindividual rectal wall volume and the mean volume wasconsidered to represent delineation error. The caudal andcranial borders of rectal contours were carefully inspectedand modified independently by at least one additional ob-server based on the same rules of rectal wall definition. How-ever, we paid less attention to delineation errors due to thedefinition of wall thickness, because it had little influence onthe calculation of dose in % rectal wall volume, as long as theerror was uniformly distributed throughout the rectal wall.

Let Vp,k be the rectal wall volume manifested on the k-thCT image for the patient p. Therefore, a data set of rectalwall volumes, �Vp,k� k � 0, 1, . . ., N � 13 � 19�, foreach of the 30 patients (p � 1, . . ., 30) is available for rectaldose/volume evaluation.

Evaluation of rectal dose/volume variationThe rectal dose/volume relation was evaluated with re-

spect to each of three different PTV constructions andconformal four-field-box treatment. The first two PTVs,denoted PTV-1.0 and PTV-0.5, were defined as the clinicaltarget volume (CTV) (prostate � seminal vesicles) plus 1.0and 0.5 cm uniform margin, respectively, generated in 3D.The third was the cl-PTV, which has been constructed byutilizing multiple CT and portal image feedback in ourclinical adaptive treatment process for prostate cancer (13).18-MV beams were applied and pointed to the geometriccenter of each PTV. The beam aperture was designed to be1.1 cm from cranial and caudal borders of the PTV, and 0.7cm elsewhere. Dose distribution in each treatment plan wascalculated using the pretreatment CT image with 3-mmgrid, with no tissue density correction. The dose distributionwas normalized to the beams’ isocenter (100%), and appliedfor each rectal dose/volume calculation.

We adopted the concept of effective volume (Veff) (14)and used it as a single factor to represent the rectal dose/volume relation. By normalizing to the correspondingwhole rectal wall volume, the effective volume of eachrectal wall was calculated by assigning the reference dose tobe the isocenter dose, and applying the volume effectivefactor n � 0.12 (15). Therefore, for each of the threetreatment plans, a data set of effective volumes, denoted�Veffp,k� k � 0, 1, . . ., N � 13 � 19�, which corre-sponded to the rectal wall volumes �Vp,k� k � 0, 1, . . ., N� 13 � 19� for each p of the 30 patients, was used toquantify interpatient and intrapatient rectal dose/volumevariations. In this study, we only consider the variation thatwas caused as a result of internal organ motion, withoutincluding daily patient setup errors. There is, however, notheoretical impediment to including the effects of setuperror in the analysis.

The interpatient rectal dose/volume variation was definedas the variation of each patient’s effective volume from themean effective volume, as calculated from the pretreatment

1112 I. J. Radiation Oncology ● Biology ● Physics Volume 51, Number 4, 2001

CT images of all patients. It can be denoted �Veffp,0

� Veffmean,0� p � 1, . . ., 30�, where Veffp,0 is the effectivevolume calculated using the rectal wall volume on thepretreatment CT image and Veffmean,0 is the sample mean of�Veffp,0� p � 1, . . ., 30�, representing the average effectivevolume of all patients treated using a specific treatmenttechnique. The intrapatient rectal dose/volume variationwas defined as the variation between the mean effectivevolume of each patient, calculated from his multiple CTimages, and the effective volume from his pretreatment CTalone. This can be denoted �Veffp,mean � Veffp,0� p� 1, . . ., 30�, where Veffp,mean is the sample mean of�Veffp,k� k � 0, 1, . . ., N � 13 � 19�, representing theaverage effective volume of the patient p treated using aspecific treatment technique. Observe that interpatient rectaldose/volume variation introduces potential variations on therectal complication among patients, when a common pre-scription dose is used for all patients. The intrapatient rectaldose/volume variation introduces uncertainty in the rectalcomplication evaluation for the individual patient, when asingle pretreatment CT image is applied for the patient’splanning and treatment.

Sensitivity analysis of rectal complication variationSensitivity of the risk of rectal complication to the inter-

and intrapatient rectal dose/volume variation was estimatedusing the Lyman-Kutcher-Burman NTCP model (14, 16).Two technical considerations were applied in the followinganalysis. First, we applied the model of iso-NTCP proposedby Ten Haken (17) to guide the dose escalation according tothe calculated risk of rectal complication. In this case, apotential prescription dose, Dprop, can be decided using thepower law relation based on a calculated effective volume,Veffprop, in a proposed treatment technique,

Dprop � Dref � � Veffref

Veffprop� n

, (1)

with respect to �Veffref, Dref�, the effective volume and pre-scription dose in a reference treatment technique, and n, thevolume effective factor used in the effective volume calcu-lation. In this study, Veffref � 100% of the whole rectal wallvolume and Dref � 70.2 Gy (1.8 Gy � 39) to treatmentisocenter were assumed from a reference treatment tech-nique of 2D planning and 4-field delivery. The correspond-ing rectal NTCP reference, NTCPref � 20%, was calculatedfrom the Lyman-Kutcher-Burman NTCP model.

Second, the sensitivity of the risk of rectal complicationto the rectal dose/volume variation was evaluated using thefollowing relation, derived directly from the Lyman-Kutcher-Burman NTCP model by taking the derivative withrespect to the effective volume:

�NTCP

� � 1

�2�� exp ��

1

2� �Dprop � Veff prop

n � TD50

m � TD50� 2�

�n � Dprop � Veff prop

n � 1

m � TD50� � �Veff (2)

where the previously published parameter values for rectaldose response (15) were used (m � 0.15, n � 0.12, andTD50 � 80 Gy). TD50 is the dose to the whole rectum thatwould lead to a complication probability of 50 percent.

Variation of rectal NTCP due to interpatient rectal dose/volume variation. Here, we assumed that all 30 patients hadsimilar clinicopathologic criteria, and that a single prescrip-tion dose based on average risk of rectal complication wasapplied to all the patients. This prescription dose Dprop wasobtained for each of the three treatment plans by specifyingVeffprop � Veffmean,0 in Eq. 1. The interpatient NTCP vari-ation, �NTCPinter, was then calculated using Eq. 2 with theinterpatient rectal dose/volume variation, �Veff � Veffp,0

� Veffmean,0 for each p � 1, . . ., 30, respectively. Thesevariations indicate the potential discrepancy between theaverage risk of rectal complication of all the patients and theindividual risk, when a common prescription dose is applied.

To eliminate these variations and maximize the potentialprescription dose for each individual, the model of iso-NTCP can be applied. In this case, each patient receives themaximum dose according to his risk of rectal NTCP calcu-lated using Veffp,0, the effective volume determined from hispretreatment CT.

Uncertainty of rectal NTCP due to intrapatient rectaldose/volume variation. The effective volume calculatedbased on the pretreatment CT can be used to determine asuitable prescription dose for each patient in a clinical doseescalation trial for prostate cancer treatment. However,uncertainties in the evaluation of rectal complicationexist due to the intrapatient rectal dose/volume variationcaused by rectum motion and patient setup. As mentionedbefore, only the uncertainty caused by rectum motion wasconsidered in this study. To evaluate this uncertainty, weassumed that the prescription dose Dprop of each patientp was determined based on his effective rectal volume,calculated from the pretreatment planning CT by substi-tuting Veffprop � Veffp,0 in Eq. 1. The intrapatient NTCPvariation, �NTCPintra, was then calculated using Eq. 2with the intrapatient rectal dose/volume variation, �Veff� Veffp,mean � Veffp,0 for p � 1, . . . , 30, respectively.Uncertainty in the NTCP calculated based on the singlepretreatment CT image was quantified by normalizing theintrapatient NTCP variation to the reference NTCP thus:�NTCPintra

NTCPref. Therefore, a positive or a negative uncer-

tainty implies an underestimate or overestimate, respec-tively, of the risk of rectal complication if the rectal wallon the pretreatment planning CT is the only one used forindividual NTCP evaluation.

1113Rectal dose/volume variation and its effect ● D. YAN et al.

To reduce this uncertainty, image feedback techniquescan be applied during the course of treatment to monitor andquantify the rectum variation. This variation can then beincorporated into the treatment planning to evaluate theongoing treatment and, if necessary, modify the treatmentand prescription dose accordingly (13). This adaptive pro-cess was retrospectively tested by simply applying the meaneffective volume calculated from multiple CT scans of eachpatient and determining the relationship between the num-ber of images used in the treatment CT image feedback andthe potential reduction of the uncertainty. Variation betweenthe mean effective volume from all CT images of eachpatient and the i-mean effective volume calculated as themean from the first i CT images (i � 3 to 7) of the patient,denoted {Veffp,mean � Veffp,i�mean� p � 1, . . ., 30}, wasused in our analysis. The prescription dose Dprop for eachpatient p in the adaptive process was determined by speci-fying Veffprop � Veffp,i�mean in Eq. 1, as if using the i-meaneffective volume in the treatment planning. The correspond-

ing uncertainty,�NTCPi�mean

NTCPref, in the NTCP evaluation was

then calculated using Eq. 2 with �Veff � Veffp,mean �Veffp,i�mean for p � 1, . . ., 30, respectively.

RESULTS

Figure 1 shows the distribution of interpatient rectaldose/volume variation �Veffinter � Veffp,0 � Veffmean,0 foreach of three PTVs. The average effective volumes,Veffmean,0, calculated using the rectal wall manifested onthe pretreatment CT image of all patients, are 42.2%,

34.4%, and 30.6% for PTV-1.0, cl-PTV, and PTV-0.5,respectively. For each of the PTVs, a prescription dose,designed for all patients, was calculated based on the iso-NTCP model, by specifying the corresponding average ef-fective volume in Eq. 1. The prescription doses are 78 Gyfor the PTV-1.0, 80 Gy for the cl-PTV, and 81 Gy for thePTV-0.5, each to the treatment isocenter. Figure 2 shows theinterpatient NTCP variation �NTCPinter calculated usingthese prescription doses and the interpatient rectal dose/volumevariation in Figure 1. Although NTCPref � 20% was expectedbased on the average effective volume of rectal wall for theplanning reference, the individual NTCP was distributed from10% to 37%, with standard deviations 5.0%, 5.9%, and 5.4%for PTV-1.0, cl-PTV, and PTV-0.5, respectively.

Figure 3 shows the distribution of intrapatient rectaldose/volume variation, calculated as �Veffintra � Veffp,mean

� Veffp,0. Similar distributions appear for all three PTVs,indicating a weak correlation between the intrapatient rectaldose/volume variation and PTV design. Therefore, thesmaller the PTV margin is, the larger relative variation willbe, because a small PTV induces a lower value of meanVeffp,mean. Figure 4 shows the uncertainties in the NTCPevaluation when calculated using the pretreatment CTalone. The likelihood of a treatment with at least � 25% ofuncertainty in the calculated NTCP is 10% for both PTV-1.0and cl-PTV, and 25% for PTV-0.5. The maximum uncer-tainty could be bigger than 50%. Furthermore, the distribu-tion of uncertainties is relatively broader for the PTV witha smaller margin. To wit, the standard deviation for PTV-1.0 is 14%, 16% for cl-PTV, and 20% for PTV-0.5.

Fig. 1. Distribution of interpatient rectal dose/volume (Veff) variation. The standard deviations are 10.4% for PTV-1.0,10.1% for cl-PTV, and 8.2% for PTV-0.5.

1114 I. J. Radiation Oncology ● Biology ● Physics Volume 51, Number 4, 2001

Finally, Figure 5 shows the potential reduction of theuncertainties when multiple CT images were used for treat-ment planning evaluation. The standard deviation of theuncertainty distribution can be reduced from 14% to 8% forPTV-1.0, 16% to 8.7% for cl-PTV, and 20% to 10.9% for

PTV-0.5, when two additional daily CT images are appliedas treatment imaging feedback for treatment planning eval-uation. Further reduction to 6% for PTV-1.0, 7.5% forcl-PTV, and 8% for PTV-0.5 can be expected when sixadditional daily CT images are applied.

Fig. 2. Distribution of interpatient rectal NTCP variation. The standard deviations are 5% for PTV-1.0, 5.9% for cl-PTV,and 5.4% for PTV-0.5.

Fig. 3. Distribution of intrapatient rectal dose/volume (Veff) variation. The standard deviations are 5.4% for PTV-1.0,5.3% for cl-PTV, and 5.6% for PTV-0.5.

1115Rectal dose/volume variation and its effect ● D. YAN et al.

DISCUSSION

The current study focuses on understanding the sensitiv-ity of the risk of rectal complication to rectal dose/volume

variation, instead of fully quantifying the variation itself.The interpatient and intrapatient rectal dose/volume varia-tions quantified in this study only include an average effectof internal rectum motion. The daily patient setup error, as

Fig. 4. Distribution of uncertainties in the rectal NTCP calculation. The corresponding mean and standard deviation are–5 � 14% for PTV-1.0, �2 � 16% for cl-PTV, and �7 � 20% for PTV-0.5.

Fig. 5. Distribution of uncertainties in the rectal NTCP calculation using multiple CT images and cl-PTV. Thecorresponding mean and standard deviation are 2.1 � 8.7% for 2 extra CT images, 2.5 � 8.6% for 4 extra CT images,and 1.6 � 7.5% for 6 extra CT images.

1116 I. J. Radiation Oncology ● Biology ● Physics Volume 51, Number 4, 2001

an additional source of rectal dose/volume variation, wasnot included. Compared to rectum motion, daily setup errorcould cause even more rectal dose/volume variation. This isbecause setup error moves the entire dose matrix, relative tothe rectal wall, in each of three dimensions. Concurrently,the geometric expansion of rectum filling could inducepartial rectal wall volume in and partial volume out of thehigh-dose region simultaneously. Therefore, we expect alarger rectal dose/volume variation, compared to the onequantified in this study, during the external beam treatmentwhen daily patient setup error is also considered. It is alsoworth mentioning that rectal dose/volume variation is de-pendent on both daily rectal wall motion and actual dosedistribution. The latter could have more effect on the vari-ation when the rectal wall is located at a higher dosegradient region as demonstrated by Figure 6. Figure 6aindicates the daily positions of a rectal wall (green and redcurves) and the 95% isodose curve (yellow) superimposedon the sagittal CT slice at the treatment central axis for apatient in the course of prostate cancer treatment. Becausethe rectal wall located in the dose gradient region, the dailyrectal dose–volume histograms (DVHs), as shown in Figure6b, varied significantly due to the rectum motion.

The sensitivity of the risk of rectal complication to rectaldose/volume variation depends on both the treatment tech-nique and the corresponding prescription dose. Compared toconventional 2D treatment technique, a higher variation inrectal NTCP can be expected in a treatment with a smaller

CTV-to-PTV margin or higher prescription dose, even if theinter- or intrapatient rectal dose/volume variation is un-changed. This relationship was evaluated using Eq. 2 andplotted as solid curves in Figure 7. Each solid curve inFigure 7 represents the sensitivity of rectal NTCP variation,�NTCP

�Veff, calculated as a function of the prescription dose for

a given value of rectal effective volume indicated at theright-hand side of each curve for Veff � 10% to 100%.Generally, the sensitivity will increase as the prescriptiondose increases, or effective volume decreases when theprescription dose is higher than 70 Gy. The cross curve inFigure 7 shows the sensitivities with respect to differenttreatment techniques. Each point on this curve correspondsto a pair of effective volume and prescription dose obtainedfrom Eq. 1, based on the iso-NTCP model. As an example,the intersection between the cross curve and the solid curvewith Veff � 20% indicates a specific treatment plan with therectal effective volume of 20% and the prescription dose of85 Gy calculated from Eq. 1 to remain the equal risk ofrectal complication. In this case, the sensitivity of rectalNTCP variation to intrapatient rectal dose/volume variationequals 1.0. It implies that 10% variation on the rectal Veffevaluation introduces 10% variation or 50% uncertainty onthe rectal NTCP evaluation.

From the curves in Figure 7, one can conclude that in theconventional 2D treatment of prostate cancer (reference toVeff � 100% and the prescription dose � 70.2 Gy), the

Fig. 6. (a) Daily rectal wall positions (green and red curves) of a patient relative to the prescribed isodose contour(yellow) shown on the patient’s sagittal section through the treatment central axis. (b) The corresponding daily rectalwall DVHs. The red curves indicate the anterior rectal wall position in a lower dose region resulting in a lower DVH;the green curves indicate the anterior rectal wall in a higher dose region.

1117Rectal dose/volume variation and its effect ● D. YAN et al.

sensitivity of the risk of rectal complication to rectal dose/

volume variation is relatively small (�NTCP

�Veff� 0.2). That is,

10% variation in the effective volume causes no more than2% variation on the calculated risk of rectal complication. Inthis case, the interpatient rectal dose/volume variation,which is at most � 25% for PTV-1.0 (see Fig. 1), inducesa maximum NTCP variation of � 5% (with one standarddeviation � 2%) among patients treated using a commonprescription dose. Therefore, the potential risk of rectalcomplication had relatively small deviation among patientswith a common prescription dose. Moreover, the intrapa-tient rectal dose/volume variation, which is at most � 14%(see Fig. 3), induces a maximum uncertainty of � 14%(with one standard deviation � 5.4%) in the individualNTCP calculation. Therefore, the NTCP evaluated using therectal dose/volume relationship based on a single pretreat-ment CT could give a fair representation to the one evalu-ated using the actual rectal dose/volume relationship.

However, the situation is decidedly different for a treat-ment technique employing a smaller CTV-to-PTV margin

and high prescription dose. For a 3D conformal treatment,the sensitivity of rectal complications could be 3 (or more)times higher than the one in the conventional 2D treatment.For PTV-1.0, cl-PTV, and PTV-0.5, the sensitivities in-crease to 0.5, 0.6, and 0.66, respectively. These sensitivities,of course, will be further exacerbated when treatments aremore conformal with high dose gradient around the edge ofthe PTV than the 4-field-box treatment. In each of thesecases, the interpatient rectal dose/volume variation couldintroduce large NTCP variation (Fig. 2), up to � 17% (withone standard deviation up to 5.9%), among patients treatedusing a common prescription dose. Meanwhile, the intrapa-tient rectal dose/volume variation could introduce uncer-tainties (Fig. 4) in the NTCP evaluation from �60% to 40%(with one standard deviation up to 20%). These uncertain-ties will be even larger when the CTV is defined as theprostate alone, or when a higher conformal treatment isapplied. Furthermore, the magnitude of both the uncertaintyand the variation could be considerably increased after includ-ing the daily setup error in the rectal dose/volume evaluation.

To reduce the influence of interpatient rectal dose/volume

Fig. 7. Solid curve: the sensitivity of rectal NTCP variation calculated using Eq. 2 as the function of the prescriptiondose for each fixed rectal effective volume. Cross curve: the sensitivity of rectal NTCP variation calculated using Eq.2 for each pair of rectal effective volume and prescription dose obtained from Eq. 1.

1118 I. J. Radiation Oncology ● Biology ● Physics Volume 51, Number 4, 2001

variation on a clinical dose escalation trial of prostate cancertreatment, the iso-NTCP model can be applied. Using thismodel, the prescription dose for each patient can be selectedbased not only on clinicopathologic criteria, but also onindividual risk of rectal complication. Furthermore, sensi-tivity evaluation, as proposed in Eq. 2, should also beincluded in the treatment planning to estimate the effect ofpotential rectal dose/volume variation. To better assess therisk of rectal complications, we should also apply the feed-back information of treatment CT and portal images in theplanning evaluation. Our results demonstrate that applyingmultiple CT images can significantly reduce the uncertain-ties of NTCP evaluation. The uncertainty can be reduced byapproximately a factor of 2 if two additional CT images

obtained in the first week of treatment are applied for NTCPevaluation.

In summary, geometric variation of rectum has less influ-ence on the dose prescription and treatment planning in theconventional 2D treatment of prostate cancer. However, thisinfluence is more pronounced for higher doses or smallerCTV-to-PTV margin. To reduce the variation and uncertaintiesin the treatment planning evaluation induced by interpatientand intrapatient rectal dose/volume variation, the iso-NTCPmodel and treatment image feedback technique can be appliedin dose escalation trials of prostate cancer treatment. Our studyshows that the CT images acquired in the first week of treat-ment could significantly improve the quality of treatment plan-ning evaluation and dose prescription.

REFERENCES

1. Radiation Therapy Oncology Group. A phase I/II dose esca-lation study using three-dimensional conformal radiation ther-apy for adenocarcinoma of the prostate. 3D-CRT OncologyGroup, 3D/OG 94-06, 1998.

2. Hanks GE, Hanlon AL, Schultheiss TE, et al. Dose escalationwith 3D conformal treatment: Five year outcomes, treatmentoptimization, and future directions. Int J Radiat Oncol BiolPhys 1998;41:501–510.

3. Zelefsky MJ, Leibel SA, Gaudin PB, et al. Dose escalationwith three-dimensional conformal radiation therapy affects theoutcome in prostate cancer. Int J Radiat Oncol Biol Phys1998;41:491–500.

4. Schultheiss TE, Lee WR, Hunt MA, et al. Late GI, and GUcomplications in the treatment of prostate cancer. Int J RadiatOncol Biol Phys1997;37:3–11.

5. Zelefsky MJ, Leibel SA, Kutcher GJ, et al. The feasibility ofdose escalation with 3D conformal radiotherapy in patientswith prostate carcinoma. Cancer J 1995;1:142–150.

6. Storey MR, Pollack A, Zagars G, et al. Complications fromradiotherapy dose escalation in prostate cancer: Preliminaryresults of a randomized trial. Int J Radiat Oncol Biol Phys2000;48:635–642.

7. Skwarchuk MW, Jackson A, Zelefsky MJ, et al. Late rectaltoxicity after conformal radiotherapy of prostate cancer (I):Multivariable analysis and dose response. Int J Radiat OncolBiol Phys 2000;47:103–113.

8. Jackson A, Skwarchuk MW, Zelefsky MJ, et al. Late rectalbleeding after conformal radiotherapy of prostate cancer (II):Volume effects and dose–volume histograms. Int J RadiatOncol Biol Phys 2000;49:685–698.

9. Boersma LJ, van den Brink M, Bruce AM, et al. Estimation ofthe incidence of late bladder and rectum complications afterhigh-dose (70–78 Gy) conformal radiotherapy for prostate

cancer, using DVH. Int J Radiat Oncol Biol Phys 1998;41:83–92.

10. Lebesque JV, Bruce AM, Kroes APG, et al. Variation involumes, dose–volume histograms, and estimated normal tis-sue complication probabilities of rectum and bladder duringconformal radiation therapy of T3 prostate cancer. Int J RadiatOncol Biol Phys 1995;33:1109–1119.

11. Ten Haken RK, Forman JD, Heimburger DK, et al. Treatmentplanning issues related to prostate movement in response todifferential filling of the rectum and bladder. Int J RadiatOncol Biol Phys 1991;20:1317–1324.

12. Roeske JC, Forman JD, Messina CF, et al. Evaluation ofchanges in the size and location of the prostate, seminal vesicle,bladder and rectum during a course of external beam radiationtherapy. Int J Radiat Oncol Biol Phys 1995;33:1321–1339.

13. Yan D, Lockman D, Brabbins D, Tyburski L, Martinez A. Anoff-line strategy for constructing a patient-specific planningtarget volume in adaptive treatment process of prostate cancer.Int J Radiat Oncol Biol Phys 2000;48:289–302.

14. 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.

15. Burman C, Kutcher GJ, Emmi B, et al. Fitting of normal tissuetolerance data to an analytic function. Int J Radiat Oncol BiolPhys 1991;21:123–135.

16. Lyman JT, Wolbarst AB. Optimization of radiation therapyIII: A method of assessing complication probabilities fromdose–volume histograms. Int J Radiat Oncol Biol Phys 1987;13:103–109.

17. Ten Haken, RK, Martel MK, Kessler M, et al. Use of Veff andiso-NTCP in the implementation of dose escalation protocols.Int J Radiat Oncol Biol Phys 1993;27:689–695.

1119Rectal dose/volume variation and its effect ● D. YAN et al.