dosimetric effects of the prone and supine positions on image guided localized prostate cancer...

Radiotherapy and Oncology 88 (2008) 67–76www.thegreenjournal.com

Prostate IGRT

Dosimetric effects of the prone and supine positionson image guided localized prostate cancer radiotherapy

Bei Liu, Fritz A. Lerma*, Shilpen Patel, Pradip Amin, Yuanming Feng,Byong Yong Yi, Cedric Yu

Department of Radiation Oncology, University of Maryland, Baltimore, MD, USA

Abstract

Purpose: To compare target coverage and doses to rectum and bladder in IMRT of localized prostate cancer in thesupine versus prone position, with the inclusion of image guidance.Materials and methods: Twenty patients with early stage localized prostate carcinoma who received external beam

radiotherapy in the supine and prone positions underwent approximately 10 serial CT examinations in their respectivetreatment position in non-consecutive days, except for one patient who was treated prone but serially imaged supine.The prostate, bladder and rectum were contoured on all CT scans. A PTV was generated on the first scan of each patient’sCT series by expanding the prostate with a 5 mm margin and an IMRT plan was created. The resultant IMRT plan was thenapplied to that patient’s remaining serial CT scans by aligning the initial CT image set with the subsequent serial CTimage sets using (1) skin marks, (2) bony anatomy and (3) center of mass of the prostate. The dosimetric results fromthese three alignments were compared between the supine and prone groups. To account for the uncertaintiesassociated with prostate delineation and intra-fractional geometric changes, a fictional ‘‘daily PTV’’ was generated byexpanding the prostate with a 3 mm margin on each serial CT scan. Thus, a more realistic target coverage index, V95,was quantified as the fraction of the daily PTV receiving at least 95% of the prescription dose. Dose–volume measures ofthe organs at risk were also compared. The fraction of the daily PTV contained by the initial PTV after each alignmentmethod was quantified on each patient’s serial CT scan, and is defined as PTV overlap index.Results: As expected, alignment based on skin marks yielded unacceptable dose coverage for both groups of patients.

Under bony alignment, the target coverage index, V95, was 97.3% and 93.6% for prone and supine patients (p < 0.0001),respectively. The mean PTV overlap indices were 90.7% and 84.7% for prone and supine patients (p < 0.0002),respectively. In the supine position 36% of cases showed a V95 < 95% after bony alignment, while only 12.5% of pronepatients with V95 < 95% following bony alignment. Under soft-tissue alignment matching the center of mass of theprostate, the mean V95 was 99.3% and 98.6% (p < 0.03) and the PTV overlap index was 97.7% and 94.8% (p < 0.0002) forprone and supine groups, respectively.Conclusions: Soft-tissue alignment combined with 5 mm planning margins is appropriate in minimizing treatment

planning and delivery uncertainties in both the supine and prone positions. Alignment based on bony structures showedimproved results over the use of skin marks for both supine and prone setups. Under bony alignment, the dose coverageand PTV overlap index for prone setup were statistically better than for supine setup, illustrating a more consistentgeometric relationship between the prostate and the pelvic bony structures when patients were treated in the proneposition.Published by Elsevier Ireland Ltd. Radiotherapy and Oncology 88 (2008) 67–76.

Keywords: Prostate cancer; IGRT IMRT; Supine; Prone

External beam radiation therapy (EBRT) has been widelyused in treating localized early stage prostate cancer. Dur-ing the full fractionated treatment course, internal organmovement and deformation as well as patient positioningor setup errors bring various uncertainties to the treatment.To insure appropriate target coverage, a planning targetvolume (PTV) is generated by adding a safety margin aroundthe clinical target volume (CTV) [1]. The CTV in CT based

0167-8140/$ - see front matter Published by Elsevier Ireland Ltd. doi:10.

planning of localized prostate cancer is manually drawn bya physician to include the entire prostate from its apex toits base. The PTV is then used to make a plan for the pa-tient. However, since the prostate is located between therectum and bladder, reducing incidental dosage to the rec-tum and bladder is also a significant physical challenge inimproving this modality: in an effort of reducing incidentaldoses to these organs at risk (OAR), target coverage may

1016/j.radonc.2007.11.034

68 Effects of prone and supine positioning on prostate cancer IGRT

often get compromised. Image guided radiotherapy (IGRT)allows daily shifts that may correct daily setup errors andpossibly organ shape changes, consequently permittingreductions of the PTV margin during the initial planning pro-cess, so that target coverage can be improved and dosing tothe rectum and bladder may be minimized.

In the absence of image-guidance, the prostate may belocalized daily by matching the patient’s skin marks, andthis method is hereby identified as, skin mark alignment.Under image guidance, the prostate may be localized bymatching the patient’s bony anatomy, which can beachieved by matching online radiographical 2D imaging toa digitally reconstructed radiograph (DRR) from the planningCT, or by matching online 3D CT imaging to the planning CT.This localization method is referred to as bony alignment. Insuch case that implanted radio-opaque prostate markers arepresent during treatment, the same matching process canlocalize the prostate nearly directly. Also, if an online CTimage is acquired and the prostate is contoured, then theprostate contour can be matched to the planning prostatecontour. These latter two approaches may be referred toas soft-tissue matching. The approach using skin marks isa non-IGRT strategy normally associated with a relativelylarge planning margin of 10 mm, with oftentimes, a 6–7 mm margin in the posterior direction [2–9] to reduce rec-tal dosage. Bony alignment based IGRT permits a reductionof the PTV margins, as it significantly reduces the effects ofsetup error. PTV margins are further minimized by aligningto prostate markers or by aligning to daily prostate con-tours, thus correcting setup error and inter-fractional organmotion [10–16]. In treating localized prostate carcinomaunder IGRT, the ideal CTV to PTV margin expansion has beenexamined by several investigators, who recommend marginvalues between 3 and 5 mm [2,10,17]. Therefore, underIGRT with soft-tissue alignment, the PTV can be createdwith a 3–5 mmmargin over the CTV when planning an inten-sity modulated radiation therapy (IMRT) treatment using theinitial planning CT. The margin selection in such image-guided IMRT planning process is independent of whetherthe patient is positioned supine or prone.

Which patient setup method, supine or prone, is betterfor prostate EBRT has been a subject of debate for years.Multiple studies have been conducted by different investiga-tors to compare the dosimetric advantage and disadvantageof these two setups [3,8,18–20]. However, there is no con-sensus regarding the optimal setup. For example, Zelefskyet al. found significant reduction of dose to rectum andsmall bowel in prone setup [3], while Bayley et al. foundhigher dosages to rectum wall and bladder wall in prone set-up [20]. All these dosimetric comparisons were conductedunder conventional laser and skin mark alignment with noimage guidance, where a PTV margin as large as 10 mm isrequired to account for daily setup error, inter- and intra-fractional organ motion and deformation. The large PTVmargin may make the benefits of either setup position diffi-cult to demonstrate.

This study compares the dosimetric characteristics of su-pine and prone positions using IMRT for localized prostatecancer under image guidance where small treatment plan-ning margins are involved. Assessing such dosimetric differ-ence between the two setup methods under different image

guidance techniques has not been conducted before. AsIMRT and IGRT become the typical treatment techniquesfor prostate cancer, it is important to understand the differ-ence, if any, in dose coverage and critical organ sparingwhen using different IGRT techniques under different pa-tient setup positions. The two major differences betweenthis study and previous studies are the use of multiple serialimages per patient in a randomly assigned treatment posi-tion, which enables us to evaluate dosimetric endpointsfrom the recomputed dose distributions on the serial scans,and under different image guidance techniques using a re-duced, 5 mm PTV margin. The specific aims of this studyare to compare: (1) dose coverage to the target, (2) doseindices of rectum and bladder and (3) PTV overlap indices,between supine and prone patient setups under differentIGRT methods.

Materials and methodsTwenty patients with early stage localized prostate carci-

noma were randomized prospectively to receive externalbeam radiotherapy in supine and prone positions. As partof this study, all patients underwent serial CT imaging intheir respective treatment position: prone or supine; exceptfor one patient who was treated prone, but serially imagedwith CT in the supine position. During the delivery of theirprescribed EBRT course of 39 fractions, with 1.8 Gy per frac-tion, each patient was scheduled to receive about 10–11 se-rial CT scans in non-consecutive days, in addition to theinitial CT scan used for planning. Incidental doses arisingfrom the additional imaging were not considered to be ad-versely significant when added to the total treatment dosesdelivered to these patients. One of the patients quit theimaging study during the treatment course and receivedonly three serial CT scans, but completed his course ofRT. A summary of the acquired scans is given in Table 1. Atotal of 200 CT serial scans were collected. All scans wereacquired using the same CT scanning protocol, based on120 kVp, 3 mm slice thickness, and 0.94 mm resolution inthe axial plane. The CT scans were accessed with a virtualsimulation workstation (Voxel Q – Philips Medical, Milpitas,CA) to perform the contouring of the prostate, rectum andbladder.

Patients scanned in the supine position were placed onthe flat surface of the treatment couch and were positionedusing foot rings and a support cushion under their knees. Pa-tients selected for prone CT scanning were placed layingface down with a ‘‘belly’’ board under the pelvic region,which displaces the small bowel away from the irradiatedregion. Fig. 1 shows midline sagittal slices from two CTscans of two different patients, one in supine position andthe other in prone position. The red, blue and green colorsare used in Fig. 1 to visualize the prostate, rectum and blad-der, respectively. The couches for both setups and the bellyboard for prone setup can also be seen clearly in the scanimage.

All scans and contours were transferred to a computer-ized treatment planning system (Pinnacle 7.4f, Philips Med-ical, Milpitas, CA) to simulate an IMRT IGRT protocol. In this

Fig. 1. Sagittal, midline slices of the CT scans for (a) a supine patient and (b) a prone patient enrolled in this study. The prone patients in thisstudy were positioned using a foam belly board, beneath the patient’s thighs and pelvis, as can be seen in (b). The manual segmentations of theprostate, rectum and bladder are also shown.

Table 1Number of serial scans acquired for each patient and the associated mean and standard deviation r of prostate volumes over his serial scans

Supine Prone

Patient # Serial scans hCTVi ±r (cm3) Patient # Serial scans hCTVi ±r (cm3)

1 10 30.2 ± 2.0 12 10 79.0 ± 6.92 11 40.2 ± 2.3 13 9 39.7 ± 2.03 3 30.0 ± 4.8 14 10 40.7 ± 4.24 10 25.4 ± 2.3 15 10 37.3 ± 3.15 11 20.4 ± 1.7 16 10 87.6 ± 5.56 11 41.0 ± 2.4 17 8 30.3 ± 3.47 11 29.5 ± 4.9 18 12 33.9 ± 9.38 11 46.3 ± 3.9 19 10 48.3 ± 3.09 11 39.1 ± 1.9 20 10 102.1 ± 13.7

10 11 36.7 ± 5.811 11 76.2 ± 6.3

Patient #3 received a full course of EBRT treatment but he quit the imaging study during the course of treatment.

B. Liu et al. / Radiotherapy and Oncology 88 (2008) 67–76 69

protocol, the initial CT scan for each patient is used as a ref-erence scan for treatment planning, and the remainingscans are used as images representing serial scans as wouldbe obtained on different treatment fractions. Hence, copy-ing a treatment plan from the reference scan to the serialscans and recalculating the dose distributions enable theevaluation of dosimetric impact of each IGRT strategy ontreatment delivery.

The CT scan that was acquired earliest for each patient isreferred to, as the initial scan (or the reference scan) andall other scans as serial scans, and we choose the initialCT scan to make a single IMRT plan for each patient. Eachpatient’s IMRT plan is designed to treat the prostate onlyat 1.8 Gy per fraction, leading to a total dose of 70.2 Gyfor 39 fractions. The IMRT plan is optimized based on aPTV expanded from prostate with 5 mm isotropic margin.This 5 mm expansion is used to account for contouringuncertainties [21] as well as intra- and inter-fractional mo-tion of the prostate [22], aiming to be suitable under soft-tissue IGRT-localization of the prostate. The IMRT plan isconstructed from seven 6 MV beams, using a clinically com-missioned model of a 21EX Varian linac with a 120-leaf MLC.

The beams are coplanar in the axial plane, at 0�, 50�, 100�,150�, 210�, 260� and 310� for the supine cases, and 30�, 80�,130�, 180�, 230�, 280� and 330� for the prone cases (usingIEC-1217 conventions). Fig. 2 illustrates the process of theIGRT simulations in this study. For each patient, the IMRTplan derived from his initial CT scan is copied to each ofthe serial CT scans for that patient, using three registrationstrategies – (a) skin mark alignment, (b) bony alignmentand (c) prostate center of mass (CM) alignment. The dosedistributions of the copied plans are recalculated on eachserial scan, followed by isodose and DVH analyses. Detailsof the analyses are as stated below.

Alignment strategiesImage-guided delivery of IMRT consists of two strategies

to align the treatment isocenter between the initial imageand the serial images. One strategy is based on matchingbony-anatomy landmark and the other one is based onsoft-tissue alignment. In this study, we evaluate the bony-anatomy alignment and soft-tissue alignment of the pros-tate, and compare them to the non-IGRT approach thataligns the patient using skin marks only. Thus, we compare

CT #1

CT #2 CT #3 CT #N

Make an IMRT plan

Register the IMRT plan by: (a) skin mark, (b) bony anatomy (c) prostate CM

……

Dose recalculation and DVH analysis

Fig. 2. Flowchart of our simulation study: a 5 mm margin expansionof the prostate contour on the initial CT scan, CT #1, yields the PTVused to create the single IMRT plan for that patient. The planparameters are registered onto each of the serial scans via skinmark, bony anatomy and prostate CM alignment. The registeredplan is recalculated (not re-optimized) for DVH and isodoseanalyses.


three strategies: (1) skin mark alignment, (2) bony align-ment and (3) CM of prostate alignment. In the skin markalignment approach, radio-opaque marker beads are placedon the patient’s skin prior to scanning, and are then identi-fied on the CT images to record the achievable position. Tosimulate bony-anatomy alignment, a surrogate approach istaken by aligning the serial CT scan to the initial CT scanusing a cross-correlation 3D-image registration package(Syntegra, Pinnacle 7.4f, Philips Medical, Milpitas, CA). Inthe soft-tissue alignment, or CM approach [23], the CM ofthe prostate volume is derived numerically by taking theaverage of the volume contained within the contour of theprostate in the serial scan and registered onto the CM ofthe prostate of the initial planning CT [24].

In this study, skin mark alignment, a non-IGRT alignmentstrategy, is used for comparisons to illustrate the improve-ment achieved by IGRT under prone and supine setups. IGRTbony-anatomy alignment is expected to reduce the PTVmargin by reducing patient setup error, but it does not ad-dress internal organ movement relative to bony anatomy.Notably, the prostate is subject to significant intra- and in-ter-fractional internal movement and deformation, partiallyby inter-fractional volume changes of bladder and rectum,such as filling and voiding [24–27]. Soft-tissue alignment isapplied to correct the internal organ motion, thus it furtherreduces the chances of target miss. Clinically, a gold seedmarker-based localization is an accepted surrogate tosoft-tissue alignment, since this process allows daily estima-tion of shifts in the CM of the prostate volume from 2D and3D radiographical imaging. However, marker migration,prostate deformation and intra-fractional prostate move-

ment are not corrected in the IGRT alignment of the im-planted markers alone. A number of authors suggest thatintra-fractional motion alone leads to a 2 or 3 mm uncer-tainty [28–35]. The soft-tissue alignment strategy in thisstudy matched the CM of prostate contours with no im-planted marker involved, thus the errors due to markermigration are not present. However, this approach doesnot account for intra-fractional prostate deformation androtation. In prostate IGRT, a 5–7 mm uniform treatmentplanning margin was recommended when bony alignmentis used for daily treatment localization, and a 3–5 mm mar-gin was recommended when soft-tissue alignment was ap-plied [2,10,17]. As a result, a margin of 5 mm is used fortreatment planning in this study.

The shifts from the three alignment strategies can becompared and their differences can thus be used to quantifythe setup error and internal prostate movement accord-ingly. Setup error is defined as the difference betweenlocalizations achieved by skin mark-based alignment andbony-anatomy-based alignment, and is quantified by thevector connecting the isocenter derived from skin markeralignment to the isocenter derived from bony-anatomyalignment for each serial scan. The internal prostate move-ment in each serial scan is calculated by the vector connect-ing the isocenter derived from bony-anatomy alignment tothat derived from soft-tissue alignment.

Initial PTV, daily PTV and PTV overlap indexThe prostate, rectum and bladder were contoured in all

CT scans manually by one physician in order to minimizethe uncertainties in structure delineation among physicians.The rectum is identified from the level of ischial tuberosityto the rectosigmoid flexure. Initial PTV was created byexpansion from the prostate contour in the initial scan usingan isotropic 5 mm margin to make an IMRT plan for eachpatient. Rather than evaluating dose coverage to the CTV,we created a ‘daily PTV’ by expanding the prostate with a3 mm margin. The rationale of creating the ‘daily PTV’ isbased on the studies that a 3 mm margin is required, evenIGRT based on soft-tissue alignment without consideringdeformation is used [17]. Fig. 3a shows an example of the5 mm margin of the initial PTV on the initial scan and theIMRT plan optimized on it; and Fig. 3b shows the 3 mm mar-gin of the daily PTV and the isodose curves of the registered(aligned by prostate CM) and recalculated IMRT plan withoutre-optimization on a serial scan. The daily PTVs on theseserial scans are not used to design a new plan but toevaluate its dose coverage achieved by the differentalignment strategies. As mentioned above, the 3 mm marginused for generating the daily PTV accounts for uncertaintiesin prostate contouring on the serial scans, prostatedeformation and intra-fractional prostate motion thatwould possibly lead to missed prostate volume.

As part of the analysis, the initial plan was applied toeach of the serial CT scans and the dose distribution wasrecalculated after image alignment using each of the threealignment methods: skin marker alignment, bony-anatomyalignment and soft-tissue alignment. To quantify the abso-lute geometrical agreement between the initial plan PTVand the daily PTV, a geometrical overlap is estimated bycalculating the volume of the daily PTV contained by the

Fig. 3. (a) The 5 mmmargin of planning PTV on one patient’s initial scan and (b) the 3 mmmargin of daily PTV on a serial scan. The prostate, itsmargin and the isodose lines are also shown for the original IMRT plan on the initial scan, and for the registered (aligned by prostate CM) andrecalculated (no re-optimization) IMRT plan on the serial scan.


aligned initial plan PTV. The overlap volume is then dividedby the volume of the daily PTV, yielding a quantity hereaf-ter, defined as PTV overlap index. Thus,

PTV overlap index ¼ ðDaily PTV \ registered initial PTVÞ=Daily PTV:

In general, the PTV overlap index ranges in value from 0 to1.0, where 1.0 is the best geometric agreement.

Dose indices and statistical testDose–volumehistogram (DVH) of all daily PTV, bladder and

rectum volumes yields dose indices to support the compari-sons of this study. V95 is used as the target DVH index todenote the fractional target volume of daily PTV receiving95%of theprescriptiondose. Thedose indices used toquantifythe incidental dosage to the rectum and bladder are, respec-tively, the fractional volumes that receive 85.5% and 92.6% ofthe prescription dose, which correspond to total doses of60 Gy for rectum and 65 Gy for bladder, assuming the treat-ment is prescribed to 70.2 Gy on target in 39 fractions.

Two-tailed Student t-test [36] is used to test, underIGRT, if the differences of target coverage and the differ-ences of OAR (rectum and bladder) incidental dosagesbetween the supine and prone positions are statisticallysignificant or not. An a value of 0.05 is used [36], that isto say, if the p value calculated from the t-test is smallerthan 0.05, then the difference is considered to be statisti-cally significant; otherwise, it is not. Two-tailed Studentt-test is also used to test, under IGRT, if the differencesof PTV overlap index between supine and prone positionsare statistically significant.

ResultsAs shown by Table 1, the standard deviation of the delin-

eated prostate volumes is in line with an expected variation

due to contouring errors and prostate deformation betweendifferent imaging dates. The setup error, which is the shiftof the skin markers relative to bony anatomy, and the pros-tate movement relative to the bony anatomy are summa-rized in Table 2. The average setup error and prostatemovement are expected to be close to zero if the patientsample size is large enough. However, for the present data-set a small residual average shift less than 3 mm in eachdirection is observed for both supine and prone positions.This is understood to be the systematic shift arising fromthe reference position, which is defined by the single initialCT scan for each patient. The standard deviation reported inTable 2 is interpreted to arise from the random variationamong the scans for each patient. Assuming setup errorand prostate movement follow Gaussian distributions, themean magnitude of the random component for setup errorand organ movements is estimated by

ffiffiffi

2p

=ffiffiffi

pp

r � 0:8r.As Table 2 shows, patients in the prone position have lar-

ger random component of setup error (the magnitude�0.8r) than supine patients, mainly in the anterior–poster-ior (AP) direction, with the r values of setup error being5.1 mm for prone position and 2.3 mm for supine position.However, the random component in prostate movement isessentially the same for these two groups of patients in allthree directions: lateral, AP and superior–inferior (SI).One can also see that variability of setup error and that ofprostate movement are comparable, indicating that bonyalignment alone, which can only correct setup error, isnot enough and that prostate alignment should be used tocorrect internal organ movement in order to further reduceplanning PTV margins.

Target coverage under the three alignmentstrategies

PTV geometrical overlap, expressed as a PTV overlap in-dex (Eq. (1)), is used to quantify the achievable accuracy incopying and aligning the initial EBRT plan to the serial CT

Table 2The mean and standard deviation (r) of setup error and prostate movement, for patients simulated in supine and prone positions

Setup error and prostate movement (mm) Mean r

Lateral AP SI Lateral AP SI

Setup errorSupine �2.81 0.94 0.22 3.54 2.31 2.33Prone �1.57 1.32 �1.23 3.90 5.12 2.57

Prostate movementSupine 0.31 �2.01 0.03 1.05 2.67 2.58Prone 0.21 �0.02 1.36 1.18 2.78 2.82

Setup error is measured by skin mark shift relative to bony anatomy, and prostate movement is measured by the shift of prostate center ofmass relative to bony anatomy. In lateral direction, ‘�’ means the shifts are to the left for supine position and to the right for proneposition; in AP direction, it means the shifts are to the anterior for supine position and posterior for prone position; in SI direction, it meansthe shifts are to the inferior directions for both positions. The ‘+’ means the opposite.


scans. Dosimetric coverage and geometrical overlap withthe registered initial PTV, which is derived from a 5 mmexpansion of the prostate contour from the initial CT scanof each patient, are evaluated using the daily PTV insteadof the prostate. Fig. 4 compares the PTV geometrical over-laps of a serial scan with the corresponding reference scanfor a supine case and those for a prone case under the threealignment strategies, as well as the corresponding dose–volume histograms (DVHs) of the serial fractions. In both su-pine and prone patient setups, one can see that bony-anat-omy alignment improves PTV geometrical overlap andtarget coverage over skin marker alignment, and that align-ing by the CM point of prostate has the largest PTV geomet-rical overlap and the best dosimetric coverage.

Fig. 4. Axial slices of CT studies in (a) the supine and (b) prone positions areby skin mark, bony-anatomy and soft tissue. DVH results for these casesstrategy.

Fig. 5 shows target coverage, V95 of daily PTV, plottedagainst PTV overlap index for supine and prone patients un-der each alignment strategy. One can see, as expected, thatthe target coverage and PTV overlap index are positivelycorrelated. Additionally, skin mark alignment leads to un-der-dosages (i.e., V95 < 95%) of more than 50% of all thecases for both supine and prone positions, agreeing withthe low PTV overlap indices ranging roughly from 50% to100% for supine patients and from 20% to 100% for prone pa-tients. Under bony alignment, the PTV overlap indices im-proved for both positions; accordingly, coverage is alsoimproved, with 36.0% and 12.5% of under-dosed treatmentfractions for supine and prone setups, respectively. Usingprostate CM alignment, the mean PTV overlap indices are

shown, containing the contours of daily PTV, the initial PTV alignedare also shown to illustrate dosimetric coverage in each alignment

Fig. 5. Target coverage vs. PTV overlap index for skin mark, bony anatomy and CM alignment for patients in the supine and prone positions.Under image guidance, the PTV overlap index of the initial PTV and daily PTVs is larger in the prone position than in the supine position onaverage.


greater than 90% for both positions; and the under-dosedtreatment fractions are reduced to only 7.0% for supine set-up and to only 3.8% for prone setup. It should be noted thatthe target coverage V95 is positively correlated with PTVoverlap index, but the correlation strength is not verystrong, especially for soft-tissue alignment. For example,some fractions with PTV overlap indices less than 0.8 haveadequate target coverage with V95 > 95%, showing thatalthough the daily PTV may not be completely containedwithin the aligned initial PTV, dosimetric coverage may stillbe acceptable, i.e., more than 95% of the daily PTV may becovered by 95% of the full prescription dose.

Table 3 summarizes the mean target coverage V95 andthe mean PTV overlap index for the supine and prone posi-tions. As Table 3 also shows, image guidance providesimportant improvements in PTV overlap index and targetcoverage over non-IGRT approach for both supine and pronepositions. Notably, statistical comparisons using a two-tailed Student t-test confirm that target coverage and PTV

overlap index are higher in the prone position than thoseachieved in the supine position under bony-anatomy alignedand soft-tissue aligned IMRT.

Incidental dosage to OARs: rectum and bladderThe incidental dosage to the rectum and bladder is also

evaluated in this study. The dose indices of interest arethe fraction of the rectum volume receiving 85.5% of theprescribed tumor dose, and the fraction of the bladder vol-ume receiving 92.6% of the prescribed tumor dose. Fig. 6plots these dose indices for (i) rectum and (ii) bladder inboth the supine and prone positions under the three align-ment strategies. The dose indices are all within the clini-cally accepted range at our institution. Comparisonsbetween the supine and prone positions under each align-ment strategy further show that there is no statistically sig-nificant difference between the incidental dosages of thetwo setup positions for rectum and bladder. The p valuesof the rectum incidental dose indices are 0.15 and 0.07

Table 3Average target coverage and PTV geometrical agreement (PTV overlap index) of treatment localization simulations for patients imaged insupine and prone positions, under skin marker alignment, bony-anatomy alignment and soft-tissue alignment

Target coverage V95 PTV overlap index

Skin mark, non-IGRT Bony-anatomyIGRT

Soft-tissueIGRT(CM)

Skin mark,non-IGRT

Bony-anatomyIGRT

Soft-tissueIGRT(CM)

Supine 0.9088 0.9363 0.9859 0.7889 0.8466 0.9477Prone 0.8855 0.9731 0.9934 0.7743 0.9068 0.9771p Value 0.00004 0.022 0.00014 0.00018

Statistical differences between the supine and prone positions for target coverage and PTV overlap index are evaluated using Studentt-test.

0

0. 05

0.1

0. 15

skin mark bony CM

supine

prone

0

0. 05

0. 1

0. 15

skin mark bony CM

supine

prone

(i) Incidental dosage of rectum

(ii) Incidental dosage of bladder

Frac

tiona

l vol

ume

Frac

tiona

l vol

ume

Fig. 6. Average fractional volume of the (i) rectum receiving 85.5%of the target dose and (ii) bladder receiving 92.6% of the target doseover all patients’ scans in the supine and prone position groups arecompared for each of the three alignment strategies.


for bony-anatomy alignment and soft-tissue (CM) alignment,respectively; and the p values for bladder incidental doseindices are 0.67 and 0.26, respectively; indicating that nei-ther setup position shows a significant advantage in reducingOAR incidental dosage over the other position.

DiscussionA number of studies have quantified reduction in setup

error and organ movements of prostate patients under var-ious image-guidance techniques for EBRT [12,14,17]. How-ever, there is not a study comparing the supine vs. pronepositions from a large serial imaging study. This study com-pares the achievable target coverage and OAR incidentaldosages between the supine and prone positions in image-guided IMRT treatments of prostate cancer, from a large

serial imaging study. Additionally, this report comparesthe geometrical agreement between the daily PTV of the se-rial scans and the registered initial PTV for two image-guided alignment strategies: bony alignment and soft-tissuealignment. Unlike the measurement of setup error and or-gan movements, which measure only the translational com-ponents, the PTV overlap index in this study also measuresthe deformation component.

In previous works of comparison between supine andprone positions [3,8,18–20], DVH analyses were conductedon a single CT scan, which corresponds to the initial scanin our study. These studies only showed the dosimetric dif-ference resulting from the difference of static pelvic anat-omy between supine and prone positions, instead of thedosimetric difference caused by localization, inter-frac-tional prostate-, rectum- and bladder-shape changes inthe two setup positions. In some previous studies wheremultiple CT scans per patient were utilized to comparethe difference of the two setup positions, only the differ-ence of the prostate movement and deformation was com-pared, not the dosimetric impact [8,18]. While in this study,DVH analyses of the target coverage, incidental dosages tothe bladder and rectum are conducted on about 10 serialCT scans for each patient, which therefore illustrate the im-pact of day-to-day organ shape and position changes on thetarget coverage and OAR incidental dosage. By prospectiverandom assignment to supine and prone setup positions fortheir treatment and scanning the patients in their respec-tive treatment position during the course of treatment,rather than flip the patient for a supine and a prone CT scan,we eliminated the uncertainty for the setup position thatthe patient is not treated with and thus not used to. Thisstudy also compares the PTV overlap index in correlationwith the target coverage index V95. The results show targetcoverage is positively correlated with PTV overlap index.However, in some cases with PTV overlap indices below0.8 may show V95 values larger than 95%. This can be under-stood from the fact that the volume receiving 95% of theprescription dose is larger than the initial PTV in the initialCT scan, which is used for treatment plan.

The statistically demonstrated improvement in coverageand PTV overlap afforded by the prone position over the su-pine position is illustrated in Fig. 5, where most prone casesfollowing soft-tissue alignment fall with 0.95 < V95 < 1.00and 0.9 < PTV overlap index < 1.0 interval. These resultsmay indicate that the day-to-day prostate shape variationin prone position is less than that of supine position, and


thus the prostate may hold a more consistent shape in theprone position. Notably, the largest difference in geometricuncertainty and associated under-dosage between supineand prone setups was found in bony alignment, where thesupine position shows marginal improvements over the skinmark alignment results. This may be due to the fact that theprostate rests on the pubic symphysis in the prone position,and thus has a more consistent geometrical relationshipwith bony structures. Therefore, it may follow that whenthe images are aligned based on the pelvic bony anatomyin prone setup, the prostate is indirectly brought to closerregistration to that achieved by soft-tissue alignment. Incontrast, the prostate shape and position relative to bonyanatomy in supine position vary more among different frac-tions than those in prone position, due to the force exertedby the table pressure [37] in conjunction with changes in thefilling and voiding of the bladder and rectum, whereas thetable or belly board pressure is blocked by the pubic sym-physis in prone position.

From Table 2, we can see that the setup errors for bothsupine and prone positions are large without image guid-ance. It is thus not surprising that the comparisons of thesetwo setup positions conducted by others before the IGRTera could not show any advantage of one vs. the other.The data from our study showed that when there is no imageguidance, both setup positions performed equally poorly.When three-dimensional alignment based on soft tissue isused, both setup positions are sufficiently accurate, withvery small difference. Only under bony alignment, the twosetup positions illustrate significant difference in targetingaccuracy.

Ideally, all prostate setups should be under soft-tissuealignment with three-dimensional on-line imaging. In real-ity, however, such three-dimensional imaging may not beavailable in some institutions. With current on-boardcone-beam CT scans, not only the image acquisition takeslonger to acquire than orthogonal image pairs, the align-ment also takes longer time in general. The images acquiredwith these on-board imagers are often lack soft-tissue con-trast, making the identification of the prostate more diffi-cult. These practical limitations make bony alignmentmuch more practical. Our results point out that if bonyalignment based on orthogonal image pairs is chosen asthe IGRT strategy, prone position provides more accurateand reproducible patient setup.

ConclusionBy prospectively randomizing 20 prostate cancer patients

into supine and prone groups, we systematically analyzedthe differences between these two setups under non-IGRTskin marker alignment, bony-anatomy IGRT alignment andsoft-tissue IGRT alignment. Without image guidance, thereis no statistically significant advantage in target coverageor in incidental dosage on rectum or bladder for either setupposition. Under image guidance using either bony-anatomyalignment or soft-tissue alignment, no statistically signifi-cant difference is found in the incidental dosage on rectumor bladder. However, statistically significant improvement

on target coverage is observed in the prone position overthe supine position. Such improvement is very small withsoft-tissue alignment, because both the supine and pronepositions have excellent target coverage even with justtranslation. Therefore, soft-tissue alignment in conjunctionwith reduced (�5 mm) PTV margins is appropriate in mini-mizing treatment planning and delivery uncertainties inboth the supine and prone position. Large and statisticallysignificant differences in both V95 and PTV overlap indexwere observed between prone and supine setups with bonyalignment favoring the prone positions. In situations wheresoft-tissue alignment is not available and bony alignmentIGRT is available, prone positioning should be used.

* Corresponding author. Fritz A. Lerma, Department of Radia-tion Oncology, University of Maryland, 22 South Greene Street,Baltimore, MD 21201, USA. E-mail address: [email protected]

Received 4 August 2007; received in revised form 15 November2007; accepted 25 November 2007; Available online 18 January2008

References[1] ICRU-62. Prescribing, Recording and Reporting Photon Beam

Therapy (Supplement to ICRU Report 50). Bethesda, MD;International Commission on Radiation Unit and Measure-ments; 1999.

[2] Song W, Schaly B, Bauman G, Battista J, Van Dyk J. Image-guided adaptive radiation therapy (IGART): radiobiological anddose escalation considerations for localized carcinoma of theprostate. Med Phys 2005;32:2193–203.

[3] Zelefsky MJ, Leibel SA, Kutcher GJ, Kelson S, Ling CC, Fuks Z.The effect of treatment positioning on normal tissue dose inpatients with prostate cancer treated with three-dimensionalconformal radiotherapy. Int J Radiat Oncol Biol Phys1997;37:13–9.

[4] Fiorino C, Reni M, Bolognesi A, Bonini A, Cattaneo GM,Calandrino R. Set-up error in supine-positioned patientsimmobilized with two different modalities during conformalradiotherapy of prostate cancer. Radiother Oncol1998;49:133–41.

[5] Hanks GE, Hanlon AL, Schulthesis TE, Pinover WH, Movsas B,Epstein BE, et al. Dose escalation with 3D conformal treat-ment: five year outcomes, treatment optimization, and futuredirections. Int J Radiat Oncol Biol Phys 1998;41(3):501–10.

[6] Boersma LJ, van de Brink M, Bruce AM, Shouman T, Gras L, teVelde A, et al. Estimation of the incidence of late bladder andrectum complications after high-dose (70–78 GY) conformalradiotherapy for prostate cancer, using dose–volume histo-grams. Int J Radiat Oncol Biol Phys 1998;41:83–92.

[7] Zelefsky MJ, Leibel SA, Gaudin PB, Kutcher GJ, Fleshner NE,Venkatramen ES, et al. Dose escalation with three-dimen-sional conformal radiation therapy affects the outcome inprostate cancer. Int J Radiat Oncol Biol Phys 1998;41:491–500.

[8] Stroom J, Koper P, Korevaar G, van Os M. Internal organmotion in prostate cancer patients treated in prone and supinetreatment positions. Radiother Oncol 1999;51:237–48.

[9] Happersett L, Mageras GS, Zelefsky MJ, Burman CM, Leibel SA,Chui C, et al. A study of the effects of internal organ motionon dose escalation in conformal prostate treatments. Radio-ther Oncol 2003;66:263–70.

[10] Schaly B, Bauman GS, Song W, Battista JJ, Van Dyk J.Dosimetric impact of image-guided 3D conformal radiationtherapy of prostate cancer. Phys Med Biol 2005;50:3083–101.

mailto:[email protected]


[11] Yan D, Wong J, Vicini F, Michalski J, Pan C, Frazier A, et al.Adaptive modification of treatment planning to minimize thedeleterious effects of treatment setup errors. Int J RadiatOncol Biol Phys 1997;38:197–206.

[12] Wu J, Haycocks T, Alasti H, Ottewell G, Middlemiss N, AbdolellM, et al. Position errors and prostate motion during conformalprostate radiotherapy using on-line isocenter setup verifica-tion and implanted prostate markers. Radiother Oncol2001;61:127–33.

[13] Mackie TR, Kapatoes J, Ruchala K, Lu W, Wu C, Olivera D,et al. Image guidance for precise conformal radiotherapy. IntJ Radiat Oncol Biol Phys 2003;56:89–105.

[14] Nederveen AJ, Dehnad H, van der Heide UA, van MoorselaarRJ, Hofman P, Lagendijk JJ. Comparison of megavoltageposition verification for prostate irradiation based on bonyanatomy and implanted fiducials. Radiother Oncol2003;68:81–8.

[15] Orton NP, Tome W. The impact of daily shifts on prostate IMRTdose distributions. Med Phys 2004;59:6–10.

[16] Poulsen PR, Muren LP, Heyer M. Residual set-up errors andmargins in on-line image-guided prostate localization inradiotherapy. Radiother Oncol 2007;85:201–6.

[17] Wu Q, Ivaldi G, Liang J, Lockman D, Yan D, Martinez A.Geometric and dosimetric evaluations of an online image-guidance strategy for 3D-CRT of prostate cancer. Int J RadiatOncol Biol Phys 2006;64:1596–609.

[18] Mclaughlin PW, Wygoda A, Sahijdak W, Sandler HM, Marsh L,Roberson P. The Haken RK. The effect of patient position andtreatment technique in conformal treatment of prostatecancer. Int J Radiat Oncol Biol Phys 1999;45:407–13.

[19] Weber DC, Nouet P, Rouzaud M, Miralbell R. Patient position-ing in prostate radiotherapy: is prone better than supine? Int JRadiat Oncol Biol Phys 2000;47:365–71.

[20] Bayley AJ, Catton CN, Haycocks T, Kelly V, Alasti H, Bristow R,et al. A randomized trial of supine vs. prone positioning inpatients undergoing escalated dose conformal radiotherapyfor prostate cancer. Radiother Oncol 2004;70:37–44.

[21] Fiorino C, Reni M, Bolognesi A, Cattaneo GM, Calandrino R.Intra- and inter-observer variability in contouring prostate andseminal vesicles: implications for conformal treatment plan-ning. Radiother Oncol 1998;47:285–92.

[22] Court LE, D’Amico AV, Kadam D, Cormack R. Motion and shapechange when using an endorectal balloon during prostatetherapy. Radiother Oncol 2006;81:184–9.

[23] Poulsen PR, Fokdal L, Petersen J, Heyer M. Accuracy of image-guided radiotherapy of prostate cancer based on the BeamCathurethral catheter technique. Radiother Oncol 2007;83:25–30.

[24] Zelefsky MJ, Crean D, Mageras GS, Lyass O, Happersett L, LingCC, et al. Quantification and predictors of prostate positionvariability in 50 patients evaluated with multiple CT scans

during conformal radiotherapy. Radiother Oncol1999;50:225–34.

[25] Wachter S, Gerstner N, Domer D, Goldner G, Colotto A,Wambersie A, et al. The influence of a rectal balloon tube asinternal immobilization device on variations of volumes anddose–volume histograms during treatment course of confor-mal radiotherapy for prostate cancer. Int J Radiat Oncol BiolPhys. 2002;52:91–100.

[26] Miralbell R, Ozsoy O, Pugliesi A, Carballo N, Amalte R, EscudeL, et al. Dosimetric implications of changes in patient repo-sitioning and organ motion in conformal radiotherapy forprostate cancer. Radiother Oncol 2003;66:197–202.

[27] Pinkawa M, Siluschek J, Gagel B, Demirel C, Asadpour B, HolyR, et al. Influence of the initial rectal distension on posteriormargins in primary and postoperative radiotherapy for pros-tate cancer. Radiother Oncol 2006;81:284–90.

[28] Antolak JA, Rosen II, Childress CH, Zagars GK, Pollack A.Prostate target volume variations during a course of radio-therapy. Int J Radiat Oncol Biol Phys 1998;42:661–72.

[29] Huang E, Dong L, Chandra A, Kuban DA, Rosen II, Evans A,et al. Intrafraction prostate motion during IMRT for prostatecancer. Int J Radiat Oncol Biol Phys 2002;53:261–8.

[30] Poggi MM, Gant DA, Sewchand W, Warlick WB. Marker seedmigration in prostate localization. Int J Radiat Oncol Biol Phys2003;56:1248–51.

[31] Pouliot J, Aubin M, Langen KM, Liu KM, Pickett B, Shinohara K,et al. (Non)-migration of radiopaque markers used for on-linelocalization of the prostate with an electronic portal imagingdevice. Int J Radiat Oncol Biol Phys 2003;56:862–6.

[32] Deurloo KE, Steenbakkers RJ, Zijp LJ, de Bois JA, Nowak PJ,Rasch CR, et al. Quantification of shape variation of prostateand seminal vesicles during external beam radiotherapy. Int JRadiat Oncol Biol Phys 2005;61:228–38.

[33] Schallenkamp JM, Herman MG, Kruse JJ, Pisansky TM. Prostateposition relative to pelvic bony anatomy based on intraprosticgold markers and electronic portal imaging. Int J Radiat OncolBiol Phys 2005;63:800–11.

[34] van der Heide UA, Kott AN, Dehnad H, Hofman P, Lagenijk JJ,van Vulpen M. Analysis of fiducial marker-based positionverification in the external beam radiotherapy of patientswith prostate cancer. Radiother Oncol 2007;82:38–45.

[35] Van den Heuvel F, Fugazzi Seppi E, Forman JD. Clinicalapplication of a repositioning scheme, using gold markers andelectronic portal imaging. Radiother Oncol 2006;79:94–100.

[36] Rice JA. Mathematical Statistics and Data Analysis, Duxbury,Belmont; 1995.

[37] Bentel G, Munley M, Marks L, Anscher M. The effect of pressurefrom the table top and patient position on pelvic organlocation in patients with prostate cancer. Int J Radiat OncolBiol Phys 2000;47:247–53.

dosimetric effects of the prone and supine positions on image guided localized prostate cancer...

Documents