Early clinical experience with anatomy-based inverse planning dose optimization for high-dose-rate boost of the prostate

Int. J. Radiation Oncology Biol. Phys., Vol. 54, No. 1, pp. 86 –100, 2002 Copyright © 2002 Elsevier Science Inc. Printed in the USA. All rights reserved 0360-3016/02/$–see front matter

PII S0360-3016(02)02897-3

CLINICAL INVESTIGATION

Prostate

EARLY CLINICAL EXPERIENCE WITH ANATOMY-BASED INVERSE PLANNING DOSE OPTIMIZATION FOR HIGH-DOSE-RATE BOOST OF THE PROSTATE BERNARD LACHANCE, M.SC.,* DOMINIC BE´ LIVEAU-NADEAU, B.SC.,* E´TIENNE LESSARD, M.SC.,*† MARIO CHRE´ TIEN, M.SC.,* I. CHOW JOE HSU, M.D.,† JEAN POULIOT, PH.D.,† LUC BEAULIEU, PH.D.,* ´ RIC VIGNEAULT, M.D., M.SC.* AND E *Service de Radio-Oncologie, Centre Hospitalier Universitaire de Que´bec, Que´bec, QC, Canada; †Department of Radiation Oncology, University of California, San Francisco, School of Medicine, San Francisco, CA Purpose: To present an exhaustive dosimetric comparison between three geometric optimization methods and our inverse-planning simulated annealing (IPSA) algorithm, with two different prescriptions for high-dose-rate (HDR) boost of the prostate. The objective of this analysis was to quantify the dosimetric advantages of the IPSA algorithm compared with more standard geometric optimizations. Methods and Materials: Between September 1999 and June 2001, 34 patients were treated to a dose of 40 – 44 Gy by external pelvic fields, followed by an HDR boost of 18 Gy in 3 fractions. The first 4 patients were treated with HDR using geometric optimization, and anatomy-based inverse-planning dose optimization was used for the remaining 30 patients. We retrospectively used the data from these 30 patients to create HDR dose distributions according to five different dose optimization protocols, including our IPSA algorithm. The various geometric optimization procedures differed in the way the dwell positions were activated and plan normalization was performed. Dose–volume histograms from all these plans were analyzed and multiple implant quality indexes extracted. Results: The IPSA algorithm provided better clinical tumor volume prescription dose coverage than did the geometric optimizations. The average prostate volume receiving 100% of the prescribed dose (V100) was 96.3% and 94.5% for IPSA with two different prescriptions compared with 92.1%, 92.6%, and 88.8% for the three geometric optimization schemes. The average urethra V150 value was 0.0% and 0.7% for IPSA with two different prescriptions, and the three geometric optimization protocols generated average values of 22.9%, 33.9%, and 38.8%. The bladder and rectal dose–volume histograms were similar, although the latest version of the IPSA algorithm slightly decreases the dose to these organs at risk because of organ-specific dose constraints included in the objective function. Conclusion: We found that planning an HDR prostate boost could be performed in a fast, secure, and effective manner with the IPSA algorithm. We demonstrated that our inverse-planning algorithm produces superior HDR plans than more conventional geometric optimizations for adenocarcinoma of the prostate. The organs at risk protection included in the objective function is a major feature of the algorithm and should allow us to escalate the HDR dose to the prostate without increasing undesirable side effects. © 2002 Elsevier Science Inc. Brachytherapy, Inverse planning, HDR, Prostate, Dosimetric study.

Interest in interstitial brachytherapy (BT) for prostate cancer was renewed in the 1980s, soon after Holm et al. (1) demonstrated that transrectal ultrasonography (TRUS) could be used to implant the prostate more accurately. This transperineal percutaneous approach was later refined and popularized by Blasko et al. (2) in North America. The success of TRUS for permanent 125I and 103Pa implants has

encouraged many centers to apply the technique to remote afterloading with high-dose-rate (HDR) 192Ir (3–10). TRUS is used to guide the insertion of the implant needles accurately into the gland; some centers also use the ultrasound images for target definition and treatment planning. The results of pioneering clinical studies have been encouraging (6, 11–14), and HDR BT using temporary implants has become an acceptable treatment option for conformal treatment of localized prostate cancer combined with three-

Reprint requests to: Bernard Lachance, M.Sc., Service de Radio-Oncologie, Centre Hospitalier Universitaire de Que´bec, CHUQ-HDQ, 11 Coˆte du Palais, Que´bec, QC G1R 2J6 Canada. Tel: 418-691-5264; Fax: 418-691-5268; E-mail: bernard.lachance@ chuq.qc.ca Supported by the National Cancer Institute of Canada with

funds from the Canadian Cancer Society. Selected for poster presentation at the 2001 American Brachytherapy Society, Vancouver, and for oral presentation at the 2001 American Association of Physicists in Medicine, Salt Lake City. Received Jan 9, 2002, and in revised form Apr 12, 2002. Accepted for publication Apr 26, 2002.

INTRODUCTION

86

Inverse planning for HDR boost of the prostate

dimensional (3D) external beam radiotherapy (EBRT) (8, 15, 16). This approach has even been used with some success for patients with unfavorable prostate cancer (17). There is a radiobiologic rationale in using EBRT in combination with an HDR BT boost (18). The recent use of posttreatment prostate-specific antigen (PSA) profile as a sensitive end point indicator and the use of posttreatment biopsies have revealed that permanent eradication of prostate cancer by either surgery or RT is not achieved as often as previously believed (19, 20). Before PSA testing, most studies of EBRT for locally advanced prostate cancer reported clinical local control rates in the range of 75% at 10 years (21). However, PSA monitoring, ultrasound-guided follow-up biopsies, and longer follow-up have made it clear that only 10 –20% of patients treated with EBRT to radiation doses of 65–70 Gy are free of recurrence at 5 years (22, 23). On the other hand, there is a clear dose– effect relation for prostate cancer, because increasing radiation dose to tumor improves local control and survival (18, 24). In recent years, dose escalation studies using either conformal EBRT (25) or prostate implants (26) have demonstrated the improved outcome provided by higher radiation doses. However, the potential gains provided by dose escalation are possibly limited because, for prostate cancer, the tumor and surrounding late-responding normal tissues are likely to have the same ␣/␤ values (18). This means that the tumor tissue and normal tissues have similar sensitivities to changes in fractionation and that the therapeutic ratio between tumor control and late sequelae cannot be significantly affected by fractionation. Thus, only by aiming toward a highly conformal dose distribution around the prostate volume can we solve the conflict between the need for a higher dose to achieve tumor control and the necessity to prevent severe late side effects. Benefits provided by higher tumor doses and conformity have been demonstrated by EBRT dose-escalating trials that took advantage of the tools provided by modern 3D conformal therapy and intensity-modulated RT (24, 25, 27). However, when striving for conformity, BT has an evident advantage over EBRT. First, BT has the ability to deliver a high dose of radiation within a well-defined volume, but with rapid falloff of the dose outside the implanted area. This rapid dose falloff is ideal for the treatment of prostate cancer because the target volume lies very close to critical normal tissues, in particular the anterior rectal wall. Second, a major problem with EBRT relates to the difficulty in measuring and correcting for daily internal organ motion and setup inaccuracies (28 – 30). To date, the only way to address this problem has been by adding safety margins around the clinical tumor volume (CTV) to create a planning tumor volume. Thus, when using EBRT, the dose is conformed to the planning tumor volume, with a consequent increase in field size and volume of tissue irradiated, thus potentially limiting the therapeutic gain (31, 32). HDR BT is not subject to this drawback, because high precision in prostate gland localization and needle placement is achievable with TRUS. Moreover, the dwell times are determined after needle implantation, thus providing

● B. LACHANCE et al.

87

good control on the resultant dose distribution. These two aspects of HDR BT make the realization of a highly conformal 3D dose region on the CTV or the prostate itself possible. This constitutes a step forward in the quest for conformity in the treatment of prostate cancer, and it can be anticipated that it will allow additional tumor dose escalation without increases in severe late side effects (17, 33). The accurate localized dose observed with BT is an appreciable advantage for patients with localized disease, but it is also a potential disadvantage for patients who have microscopic spread outside the implantation volume. EBRT can encompass a safety margin outside the prostate gland to include the seminal vesicles or even treat the pelvic nodes for high-risk patients. Also, patients with bulky local disease are very likely to benefit from the dose escalation provided by BT. Therefore, for intermediate- or high-risk patients, combination therapy is very interesting, because it allows an adequate dose for microscopic disease to be delivered with EBRT, followed by a boost using HDR afterloading to the site of macroscopic disease within the prostate. The recent preliminary analysis of Radiation Therapy Oncology Group (RTOG) study 94-13 (34) presented at the 2001 American Society for Therapeutic Radiology and Oncology meeting showed a significant advantage in terms of biochemical PSA control when treating the whole pelvis with EBRT in patients with a risk of lymph node involvement ⬎15%. In our clinic, most of the patients considered as candidates for HDR prostate treatments have a ⬎15% pelvic lymph node invasion risk and would therefore benefit from pelvic RT. Using a BT boost is appealing because of the potential reduction in complexity it provides compared with dose-escalation strategies using only EBRT and because it decreases the overall treatment time by 2–3 weeks compared with intensity-modulated RT or 3D conformal RT. On this basis, a program of combined EBRT and HDR BT boost was initiated at our institution, and the preliminary dosimetric results have already been presented (35, 36). Because we have been offering permanent 125I or 103Pa implants to our patient with localized prostate cancer since 1994 (37–39), we chose to offer the HDR BT boost to patients with intermediate- or high-risk disease (PSA ⱖ10 ng/mL, Gleason score ⱖ7, Stage T3 or higher). These patients may be best served by including an external beam component to enable full coverage of the microscopic disease. When starting this program, we adapted our computerplanning program based on a fast-simulated annealing algorithm that was developed for permanent implant optimization (37). We applied the concept of inverse planning to obtain an anatomy-based optimization of the dose distribution without any manual modification to deliver a highly conformal HDR prostate treatment. The details of an early version of this algorithm have been previously published (40). The dwell times are determined under control of the algorithm such that the resulting dose distribution will closely respect the clinical criteria. Those criteria are included in a dedicated objective function. The objective of

88

I. J. Radiation Oncology

● Biology ● Physics

this paper was to report our early clinical experience with anatomy-based inverse planned HDR boost for adenocarcinoma of the prostate and to present an exhaustive dosimetric comparison between the plans obtained with our new optimization algorithm and those produced by conventional geometric optimization methods. METHODS AND MATERIALS Since September 1999, patients not eligible for permanent implant alone (PSA ⱖ10 ng/mL, Gleason score ⱖ7, Stage T3 or higher) were offered EBRT at 40 – 44 Gy followed by an 18 or 19.5-Gy HDR boost delivered in 3 fractions during 24 h. Between September 1999 and June 2001, 34 patients were treated to an HDR dose of 18 Gy. The first 4 patients were treated using geometric optimization, and anatomy-based inverse-planning dose optimization was used for the remaining 30 patients. These 30 patients are included in the present retrospective analysis. EBRT was delivered with 4 pelvic fields at the rate of 2 Gy/fraction daily. The treatment sequence was always EBRT first, followed by the HDR boost. The interval between EBRT and BT was limited to 1–2 weeks to maintain important time and dose radiobiologic relationships. Patients received hormonal neoadjuvant and adjuvant therapy at the discretion of the referring physician. The implant procedure has been described elsewhere (6, 7), and only the main outlines are presented here. Our current method calls for a single interstitial implant and 3 subsequent HDR fractions. Needle placement was performed in the operation room under general anesthesia. The patient was placed in the lithotomy position to expose the perineum and rotate the pubic arch anteriorly. Flexible needles (25-cm Flexi-Needles) were inserted in the prostate gland through the perineum under TRUS guidance. A total of 18 catheters were implanted with the help of a template (disposable prostate template) that was afterward sutured to the perineum to maintain the needles in place. Care was taken to avoid implantation of the urethra and rectum. The proper placement of the needles was verified using fluoroscopy and cystoscopy. Postimplant cystoscopy helped to ensure that the needle tips were positioned well into the prostate base, just under the bladder wall. Four radiopaque gold markers were always inserted into the prostatic tissue at the time of the implant. These markers were carefully implanted at the apex and at the base of the prostate gland in each prostate lobe. The gold markers were used to align the anatomy and dwell positions for the first 22 patients. Anatomy definition and catheter reconstruction A CT scan with a slice spacing of 2.5 mm was acquired postoperatively for each patient. The superior and inferior scan limits were chosen such that the entire bladder and rectal volumes would be included in the examination. We chose a field of view that would provide maximal zoom on the prostate volume and at the same time encompass the bladder and rectum on all slices. The radiation oncologist

Volume 54, Number 1, 2002

drew the target and organs at risk (OAR) contours from the CT scan on a commercial treatment planning system. For every patient, the following contours were obtained: target volume (CTV), urethra, bladder, and rectum. The contour coordinates were exported in text format to be transferred to the anatomy-based inverse-planning and geometric optimization algorithms. Two methods have been used to reconstruct the dwell position coordinates: variable angle X-ray simulator films or in-house CT reconstruction. When using variable angle films, the alignment of the dwell position coordinates with the anatomy was performed using the four radiopaque gold markers, which are visible under both imaging modalities (simulator films and CT slices). The positions of these gold markers in both coordinate systems (simulator films and CT) were used to derive a coordinate transform that was afterward applied to the dwell position coordinates. This method was used to reconstruct the catheters for our first 22 patients. Starting with Patient 23, we used in-house developed CT reconstruction to obtain the dwell position coordinates. This allowed us to save about 2 hours in the treatment planning procedure, because it was no longer necessary to go to the simulator room. CT reconstruction greatly improved patient comfort and the procedure logistic. The dwell position coordinates were exported in text format to be transferred to the various optimization algorithms. Treatment planning and treatment delivery overview The entire treatment procedure, consisting of treatment planning and delivery, was performed during 2 days. The first day began with the implantation procedure, performed early in the morning. The patient usually left the operating room just before 12 pm and was then transferred to the CT room. After the CT examination was completed, the patient was directed to the simulator room, if CT reconstruction was not used. He was then kept under observation while the treatment plan was prepared. The CT slices were imported into the treatment planning computer, and the radiation oncologist drew the contours. Once the patient anatomy and dwell positions were obtained, the planning process involved the use of this information to define the dwell time values to produce a clinically acceptable dose distribution. In the case of the present 30 patients, the actual delivered treatment plans were obtained with our anatomy-based inverse-planning algorithm (40), which automatically finds the dwell times once the dose constraints and prescription is specified. These dwell times were always imported into our commercial treatment planning system to perform the final illustration of the isodose plan. The isodose distributions were calculated on multiple slices to allow 3D dose visualization and evaluation. Once the plan was accepted, the patient went to the treatment room, and the flexible needles were attached to the 18 channels of the afterloading unit (Nucletron MicroSelectron High Dose Rate Remote Afterloader) for treatment delivery. A rigorous and consistent connection pattern

Inverse planning for HDR boost of the prostate

was always observed to minimize the risks of misadministration. We also used an extra dummy cable check run to ensure that all the channels were operational before initiating the treatment sequence. Usually, everything was ready to deliver the first fraction at the end of the afternoon. The average total dwell time for a 6-Gy prescription was 2260 s*Ci, and the total treatment time (including control cable displacements) was around 20 min. The second fraction was delivered the next morning. A waiting period of at least 6 h was then respected before delivering the third and last HDR fraction. The implant was removed in the treatment room, immediately after this last fraction. Optimization methods used for retrospective analysis This study was designed to compare quantitatively the plans obtained with our IPSA algorithm and the plans generated with more conventional geometric optimization. We retrospectively used the data from 30 patients (anatomy and dwell positions) to create dose distributions according to five different dose optimization protocols, including the IPSA algorithm. For most patients, we used three different geometric optimization schemes and two IPSA optimizations. The various geometric optimization procedures differed in the way the dwell positions were activated and the plan normalization was performed. We refer to these optimization schemes as Geo1, Geo2, and Geo3. Geo1 and Geo2 were anatomy-based optimizations, and Geo3 plans were based on the information provided only by simulator films (no CT information or contours). Geo1. For Geo1, the active dwell positions were those included between specific superior and inferior limits, on the basis of the CTV contours. Thus, the active positions were those within the first and last CTV contour, plus a 5-mm margin. After the geometric optimization was performed on these dwell positions, the absolute dwell times (plan normalization) were adjusted such that the prescription isodose would adequately cover the prostate contours. It has been our experience that an harmonic mean of 720 cGy on the prostate contour points is necessary to provide a plan that will satisfy the clinical criteria for a 600-cGy prescription. Geo2. In the second geometric method, an attempt was made to create a more conformal plan. Therefore, the active dwell positions were carefully selected using 3D visualization of the prostate volume, OAR contours, and catheter positions. The creation of these plans required much more time and effort from the treatment planner than did the Geo1 optimization scheme. Good clinical experience was also necessary to eliminate the dwell positions not necessary to cover the CTV without creating cold spots in the periphery of the prostate. As with Geo1, geometric optimization was performed on the selected dwell positions, and plan normalization was adjusted on the prostate contour points. Geo3. Geo3 reproduced the treatment planning performed without the anatomic information provided by CT. We used the X-ray simulator films of the 18 patients for whom we did not used CT-based reconstruction. With only

● B. LACHANCE et al.

89

these films showing the gold markers at the apex and base of the prostate gland, the radiation oncologist determined the superior and inferior limits for the HDR treatment. Because this procedure was done ⬎1 year after those patients were actually treated, it was impossible for the radiation oncologist to remember their CT-derived contours. Thus, the Geo3 method was not slanted by previous knowledge of anatomic information. In this method, the active dwell positions were those included between the limits selected from the X-ray films. After the geometric optimization was performed on these dwell positions, the plans were presented to the radiation oncologist for selection of a prescription isodose. This step was performed without CT information, and the isodose selection was based solely on the visualization of the catheters. Only after the plan was completed, did we illustrate it on the anatomy provided by the CT examination for dosimetric analysis. These 3 dwell activation and normalization procedures were meant to illustrate a gradation in complexity (with Geo3 the simplest and Geo2 the more complex) and time involvement necessary to create plans with standard methods. Together, they should be representative of the practice of most centers using HDR prostate boosts. Although never used clinically at our institution, the Geo3 method was presented to demonstrate clearly the need for anatomic information in the planning of HDR treatment. In this study, we used a basic geometric optimization on distance, in which the relative dwell time was inversely proportional to the contribution from all other active dwell positions: Ti ⫽

再冘

n j⫽1 j⫽i

冎

⫺1

Tj[(xj ⫺ xi)2 ⫹ (yj ⫺ yi)2 ⫹ (zj ⫺ zi)2]⫺1

(1) where Ti is the relative time for dwell position i; Tj is the relative time for all other active dwell positions; xi, yi, and zi are the coordinates of dwell position i; and xj, yj, and zj are the coordinates of all other active dwell positions. IPSA algorithm. Our IPSA algorithm is based on the fast-simulated annealing concept. Details of the algorithm have been recently published (40), and we only present an overview of its structure and the principal steps leading to dose optimization. IPSA minimizes an objective function that is based on dose constraints, known as dose potentials. These dose potentials are designed to mathematically describe the clinical criteria specified by the radiation oncologist. The potentials actually translate a given dose distribution into a penalty value that measures the degree of fulfillment of the clinical criteria. The objective function includes constraints on the prescribed dose to the CTV, dose uniformity, maximal dose to the urethra, and falloff dose to the surrounding OAR and healthy tissues. Because the optimization process is governed by fast-simulated annealing, IPSA avoids the local minimum inherent to multiparameter optimization problems, ensuring consistent and clinically acceptable solutions.

90

I. J. Radiation Oncology

● Biology ● Physics

Volume 54, Number 1, 2002

where the parameters Dmin, mmin, Dmax, and mmax are the minimal dose constraint, slope of the minimal dose constraint, maximal dose constraint, and slope of the maximal dose constraint, respectively (Fig. 1). Each clinical criterion is controlled by its own potential active on its respective dose calculation points. The actual dose potentials that were used to treat our patients to a 6-Gy prescription are presented in Fig. 2. The sum of the penalty values wi over m dose points was performed to obtain the global penalty En of a given clinical criterion n:

冘 wm m

En ⫽

Fig. 1. Dose potential defined by Eq. 2.

i

(3)

i

The dose potentials used in the IPSA algorithm have the general form defined by the following relation, giving the penalty wi for the dose Di at dose point i: wi ⫽

再

mmin共Di ⫺ Dmin兲 if Di Ɱ Dmin mmax共Di ⫺ Dmax兲 if Di Ɑ Dmax 0 if Dmin ⱕ D i ⱕ D max

Finally, the penalty value for the entire objective function E(k) of iteration k is given by the sum of all terms associated with the different clinical criteria:

冘 E (k) 4

E(k) ⫽ (2)

n

n⫽1

Fig. 2. Dose potentials used to treat our patients with a prescription of 6.0 Gy to the prostate contours. (A) Prostate contours dose potential. (B) Prostate uniformity dose potential. (C) Urethral dose potential. (D) External zone dose potential.

(4)

Inverse planning for HDR boost of the prostate

For the IPSA algorithm, two sets of dose potentials were used in this study, corresponding to prescriptions of 6 Gy and 6.5 Gy on the prostate contours. The 30 patients included in this study were actually treated with the potentials corresponding to a prescription of 6 Gy, but we later used a 6.5-Gy prescription for treatment. These 6.5-Gy potentials were used in conjunction with the latest version of the IPSA algorithm. This new version of the algorithm features specific dose potentials for more than one target and multiple OARs (here, the bladder and rectum were included as OARs). The IPSA computation time for a typical implant (250 possible dwell positions) is about 1–2 min on a Pentium III 700 Mhz. No manual tweaking of the plans was done after optimization was achieved with the IPSA algorithm. Dosimetric analysis Once the various optimization schemes have been performed for a given patient, the target and OARs dose– volume histograms (DVHs) were calculated to perform a quantitative comparison of the plans. In this study, the OARs were the urethra, bladder, and rectum. For the prostate (CTV), bladder, and rectum, the volume analyzed was the entire organ volume, as contoured by the radiation oncologist on the CT examination. For the rectum and bladder, we did not contour the organ walls exclusively, even though they constituted the actual radiosensitive portion. In this study, we were interested in comparing the rectal and bladder DVHs for different plans more than in the DVHs for themselves. Consequently, we believe that using the whole organ volume instead of the wall volume was justified in our analysis. For the urethra, the analysis was limited to the volume included within the first and last CTV contour (urethral volume within the prostate). This approach should provide the most solid basis for comparison of different plans for a given patient and for comparison of plans from one patient to another. Multiple indexes were extracted from the DVHs, including D90, D5, and values for the organ volume receiving 200% and 100% of the prescribed dose (V200 and V100), and so forth. These various implant quality indexes are presented. RESULTS Typical dose distributions for 2 different patients are presented in Fig. 3. The results of the IPSA 6.0-Gy optimizations (Fig. 3b,d) were compared with the Geo1 optimizations (Fig. 3a,c). The data for these 2 patients illustrate the somewhat extreme cases of a small prostate volume (29.38 cm3, Fig. 3a,b) and a very large CTV (102.84 cm3, Fig. 3c,d). The IPSA plans show that the prescription isodose conformed quite well to the CTV contour and that every portion of the target was covered. Although small portions of the large prostate contour were missed by the prescription isodose, the Geo1 distributions also constituted good plans with respect to target coverage and conformity. However, when looking at the Geo1 plans, we can see that the urethra

● B. LACHANCE et al.

91

is completely encompassed by the 800-cGy isodose (red). This is true for both the small and the large prostate volume plans. The isodose distributions presented in Fig. 3 clearly demonstrate the urethra protection provided by the urethra dose constraint included in the IPSA objective function. The improved uniformity provided by the IPSA optimizations is also visible as a reduction in the high-dose regions (800cGy isodose). The prostate DVHs associated with the small prostate volume case (Fig. 3a,b) are plotted in Fig. 4. Figure 5 presents the urethral DVHs for the same patient. Three optimization schemes are compared in these two figures. The IPSA 6.0 Gy and Geo1 plans presented in Figure 1 are compared with the IPSA 6.5-Gy optimization. We can see in Fig. 4 that both IPSA optimizations provide a much sharper dose falloff around the prescription dose, giving a clear indication that the IPSA plans are more homogeneous and provide better target coverage. The improved CTV coverage is reflected in the higher V100 values obtained with the IPSA (IPSA 6.0 Gy, 98.2%; IPSA 6.5 Gy, 97.7%; and Geo1, 95.3%). The tails in the high-dose region that are characteristic of the IPSA plans are the result of the nonuniform dwell times produced by inverse planning. Indeed, geometric optimizations always provide somewhat uniform dwell times, slowly increasing toward the periphery of the implant volume. On the other hand, to spare the urethra and provide good CTV coverage at the same time, IPSA has to turn off some dwell positions and increase the dwell time significantly at other positions. Those higher dwell times, although necessary to improve the overall plan, are responsible for the tails seen in Fig. 4, as the dose rises very close to these positions. We observed that, for a given patient, the total irradiation time was similar with the IPSA and geometric optimization methods, but that IPSA distributed this irradiation time more adequately. The DVHs presented in Fig. 5 clearly demonstrate the reduction in the dose received by the urethra when using IPSA. For this particular patient, the D5 value was 722 cGy, 831 cGy, and 943 cGy for IPSA 6.0 Gy, IPSA 6.5 Gy, and Geo1, respectively. Comparing both 6-Gy prescriptions, we see that D5 is 221 cGy higher with the geometric optimization, indicating that the prescription could be raised significantly with inverse planning and still have an equivalent urethral dose than with a 6.0-Gy Geo1 plan. Looking at the increase in the D5 values between the IPSA 6.0-Gy and IPSA 6.5-Gy plans, a simple ratio calculation leads to the conclusion that a 7.0-Gy prescription with IPSA would still be equivalent to the Geo1 6.0 Gy with respect to urethral dose. This is one of the major gains provided by inverse planning for HDR prostate treatment. The urethral protection provided by the IPSA algorithm is further illustrated by the 3D view presented in Fig. 6. The results of our retrospective analysis for all cases are presented in Figs. 7 to 9 and Tables 1 to 3. The most important implant quality indexes extracted from the prostate and urethral DVHs are presented in Table 1. The average prostate V100 values were about 2% higher than

92

I. J. Radiation Oncology

● Biology ● Physics

Volume 54, Number 1, 2002

Fig. 3. Comparative isodose curves for 2 different patients illustrating (a,b) a small prostate volume (see Figs. 4 and 5 for corresponding DVHs) and (c,d) a large prostate. CT slices presented were located in the middle of the prostate gland. (b,d) IPSA dose distribution (prescription dose 6 Gy) and (a,c) geometric optimization (Geo1, prescription dose 6 Gy). Green indicates CTV, purple, the rectum; the isodose values were 800 cGy (red), 600 cGy (orange), 400 cGy (light blue), and 200 cGy (dark blue).

Geo1 and Geo2 for IPSA 6.5 Gy and about 4% with IPSA 6.0 Gy. Geo1 and Geo2 are anatomy-based optimizations and are quite good at providing adequate CTV coverage. The average V100 values were slightly lower and the standard deviations were similar to those of the IPSA plans. On the other hand, Geo3, which is not an anatomy-based optimization, is less efficient at covering the CTV, because the average V100 value is only 88.76%. The variation in the quality of the target coverage is also much more with Geo3, reflected by the higher standard deviation (7.51%). V100 values as low as 73% have been observed with some Geo3 plans. The superiority of IPSA may be best appreciated when comparing the V150 and V200 values for the CTV. The average V150 values of 29.4% and 25.2% for IPSA plans appear as a significant improvement compared with the 50 –58% values characteristic of geometric plans. IPSA plans do provide more homogeneous dose distributions. This is clearly demonstrated by Fig. 7, in which the average V100, V150, and V200 values are plotted for the 5 different optimization schemes. Three elements depicted in Fig. 7 are obvious. First, IPSA provides superior target coverage than Geo1 and Geo2, and Geo3 is not a reliable planning method.

Second, IPSA plans are more homogeneous than geometric plans. Finally, our inverse planning provides much more consistent results from one patient to another. The boxes (1 standard deviation) and bars (minimal and maximal values) of Fig. 7 reveal the improved consistency of the IPSA plans. The large variations in V150 and V200 values characteristic of geometric plans are correlated with the prostate volume. Smaller prostate volumes will be hotter than large ones when using geometric optimization. This relationship between volume and homogeneity is absent in IPSA plans, because the uniformity is taken into account in the objective function. Figure 8 presents the average urethral V100, V150, and V200 values plotted for the five different optimization schemes. The superiority of inverse planning is visible on this plot. The V150 and V200 values are almost nonexistent for IPSA plans, demonstrating the urethral protection provided by the algorithm. On the other hand, the average V150 values range from 22.9% for Geo1 to 38.8% for Geo3. Furthermore, the geometric optimizations present wide variations in the V150 and V200 values, as depicted by the boxes (1 standard deviation) and bars (minimal and maxi-

Inverse planning for HDR boost of the prostate

● B. LACHANCE et al.

93

Fig. 4. DVH of the prostate for a given patient. Three optimization schemes are presented: IPSA with dose potentials for a 6.0-Gy prescription (solid curve), IPSA with dose potentials for a 6.5-Gy prescription (dashed curve), and geometric optimization (Geo1) for a 6.0-Gy prescription (dotted curve).

mal values) in Fig. 8. This can be expected, because no constraints on the dose to the urethra are provided when performing geometric optimizations. Smaller prostates will tend to have a higher urethral dose for a given prescription

dose to the prostate contours because of the increased concentration of catheters in the center of the CTV. The average D10 values for the urethra are also presented in Table 1. The D10 was 715 cGy and 770 cGy for IPSA 6.0 Gy and IPSA

Fig. 5. DVH of the urethra for a given patient. Three optimization schemes are presented: IPSA with dose potentials for a 6.0-Gy prescription (solid curve), IPSA with dose potentials for a 6.5-Gy prescription (dashed curve), and geometric optimization (Geo1) for a 6.0-Gy prescription (dotted curve).

94

I. J. Radiation Oncology

● Biology ● Physics

Volume 54, Number 1, 2002

Fig. 6. (a– d) Comparative isosurfaces for a given patient. (b,d) IPSA plan (prescription dose 6 Gy), (a,c) geometric optimization (Geo1, prescription dose 6 Gy). (a,b) Prescription isosurface (600 cGy) (c,d) 800 cGy isosurface. Green indicates CTV; purple the rectum; yellow, the bladder.

6.5 Gy compared with 924 cGy, 979 cGy, and 1010 cGy for Geo1, Geo2, and Geo3, respectively. The data related to our analysis of the rectal and bladder DVHs are presented in Table 2. The various rectal indexes are almost the same for the IPSA 6.0-Gy plan and the 3 geometric methods. This is not surprising, because our radiation oncologist took great care not to place the catheters too close to the rectum. When the catheters are implanted with optimum positions within the prostate, anatomy-based geometric methods will do a good job of delivering conformal dose to the CTV, thus protecting the rectum. The only method that stands out is the IPSA 6.5-Gy optimization. Even though the prescription dose was higher, the D90 and D10 values were appreciably lower compared with the other optimization schemes. This is the result of organ-specific constraints for the rectum and bladder that were later included in the objective function. With finetuning of this improved version of the algorithm, it is possible to reduce the dose to the rectum. The gain is less obvious when considering the bladder data. The D10 value was actually higher for both IPSA methods than with Geo1 and Geo2. Two arguments can help understand this behavior. First, the bladder-specific constraints in IPSA 6.5 Gy are less effective than they are for the rectum because of the relative positions of these organs with respect to the prostate gland. The rectum is close to the prostate all along its length, and the bladder is in close proximity to the prostate

only in the base region. The higher average bladder D10 is also due to the better dose coverage provided by IPSA in this region. The base of the prostate, just under the bladder mucosa, is always the most difficult portion of the gland to cover adequately with the catheters. It is difficult for the radiation oncologist to assess the position of the needle tips precisely and the anterior catheters may retract slightly during patient displacements. Geometric optimizations cannot compensate for this geographic miss, and the prostate base is often under-dosed. Because IPSA has the ability of raising some dwell times to provide adequate coverage in the base region, it obviously increases the dose to the bladder at the same time. The improved CTV coverage provided by IPSA can be seen by comparing prescription isosurface for the IPSA plan in Fig. 6b and the Geo1 plan in Fig. 6a. The IPSA plan shows better coverage in the base region than the Geo1 plan. A good HDR plan must provide adequate CTV coverage (high V100 value), but to be clinically satisfying, it must also be as conformal as possible. To evaluate the degree of conformity of our different plans, we computed the volume of the prescription isodose (the 100% isodose) and defined a conformity index (CI) as follows:

CI ⫽

Volume of prescription isodose Volume of CTV

(5)

Inverse planning for HDR boost of the prostate

● B. LACHANCE et al.

Fig. 7. Mean values of prostate volume receiving V100, V150, and V200. The 5 optimization schemes used in our retrospective analysis are presented. Triangles indicate the mean values, boxes represent 1 standard deviation, and bars the minimal and maximal values.

Fig. 8. Mean values of urethral volume V100, V150, and V200. The five optimization schemes used in our retrospective analysis are presented. Triangles indicate mean values, boxes represent 1 standard deviation, and bars the minimal and maximal values.

95

96

I. J. Radiation Oncology

● Biology ● Physics

Volume 54, Number 1, 2002

Fig. 9. (a) Prostate V100 value as a function of the CI. (b) Prostate V100 value as a function of the urethral V150 value. All optimization schemes for the 30 patients included in this study presented.

The CI will tend toward 1 when a plan is made more conformal, with an ideal plan having a CI equal to unity. Thus, provided a given plan has a high V100 value (good CTV coverage), the CI becomes an indicator of the plan quality. Table 3 shows the average, minimal, and maximal CI values for our various optimization schemes. The average, minimal, and maximal prostate volume is also presented in Table 3. With an average CI of 1.27, the IPSA 6.5 Gy plan is the most conformal optimization method. Geo2 is next, with an average CI of 1.39. The IPSA 6.0 Gy and Geo1 plans have similar average CI values, but the standard deviation shows that our inverse planning is more consistent. The clinical uncertainties associated with non–anato-

my-based planning are revealed in Table 3. Geo3 has the highest average CI value (1.61), and the standard deviation of 0.27 is significantly higher than observed with the other optimization methods. DISCUSSION The primary objective of this study was to compare dose optimization methods for HDR BT boost of the prostate. In particular, we wanted to assess quantitatively the dosimetric superiority, if any, of the IPSA algorithm for the determination of dwell times. Our data indicate that a significant dosimetric advantage exists in using IPSA for treatment

Inverse planning for HDR boost of the prostate

● B. LACHANCE et al.

97

Table 1. Results of retrospective analysis of 30 patients treated with IPSA algorithm: Implant quality indexes for prostate and urethra

IPSA 6.0 Gy Mean SE (n ⫽ 30) SD IPSA 6.5 Gy Mean SE (n ⫽ 26) SD Geo1 Mean SE (n ⫽ 30) SD Geo2 Mean SE (n ⫽ 30) SD Geo3 Mean SE (n ⫽ 18) SD

Prostate D90 (cGy)

Prostate D10 (cGy)

Prostate V200 (%CTV)

Prostate V150 (%CTV)

Prostate V100 (%CTV)

Urethra D10 (cGy)

Urethra V200 (%volume)

Urethra V150 (%volume)

Urethra V100 (%volume)

681.16 5.78 31.64

1135.88 8.47 46.41

7.96 0.27 1.49

29.44 0.67 3.69

96.33 0.50 2.73

715.49 1.43 7.81

0.00 0.00 0.00

0.03 0.03 0.18

93.12 0.80 4.40

687.24 3.36 17.13

1184.63 13.09 66.75

7.04 0.35 1.78

25.18 0.89 4.55

94.45 0.47 2.40

769.50 8.55 43.59

0.09 0.05 0.26

0.72 0.24 1.21

94.82 0.55 2.79

634.19 10.59 57.98

1175.47 31.62 173.17

10.11 2.58 14.11

49.77 3.16 17.33

92.05 0.60 3.30

924.25 26.08 142.84

2.69 1.90 10.43

22.90 4.93 26.98

88.75 1.30 7.13

644.24 8.84 48.45

1253.94 34.74 190.30

15.49 2.79 15.27

54.77 2.95 16.18

92.64 0.52 2.83

978.60 29.87 163.62

4.75 2.21 12.10

33.93 5.36 29.33

90.36 1.34 7.34

592.88 35.61 151.06

1331.85 53.39 226.52

22.93 4.50 19.09

58.23 4.32 18.31

88.76 1.77 7.51

1010.35 39.31 166.80

4.04 1.58 6.70

38.78 7.33 31.10

86.54 2.39 10.15

Abbreviations: IPSA ⫽ inverse-planning simulated annealing; D90 ⫽ minimum dose to 90% of prostate (or urethra, or bladder, . . .) volume; D10 ⫽ minimum dose to 10% of prostate (or urethra, or bladder, . . .) volume; V200, V150, V100 ⫽ prostate volume receiving 200%, 150%, or 100% of prescribed dose; CTV ⫽ clinical target volume; SE ⫽ standard error; SD ⫽ standard deviation.

planning. The two scatter plots presented in Fig. 9 summarize most of the conclusions that can be drawn from the data presented in this study. Figure 9a is a plot of the prostate V100 value as a function of CI for all 30 patients included in this study. An ideal optimization method would lead to

clustering of the dots in the top left corner of this graph. The IPSA methods (especially IPSA 6.5 Gy) tend to occupy a much smaller zone on the graph than do the Geo1 and Geo3 methods. This illustrates that inverse planning provides more consistent results than these geometric methods, and

Table 2. Results of retrospective analysis of 30 patients treated with IPSA algorithm: Implant quality indexes for rectum and bladder

IPSA 6.0 Gy Mean SE (n ⫽ 30) SD IPSA 6.5 Gy Mean SE (n ⫽ 26) SD Geo1 Mean SE (n ⫽ 30) SD Geo2 Mean SE (n ⫽ 30) SD Geo3 Mean SE (n ⫽ 18) SD

Rectum D90 (cGy)

Rectum D10 (cGy)

Rectum V100 (%volume)

Rectum V20 (%volume)

Bladder D90 (cGy)

Bladder D10 (cGy)

Bladder V100 (%volume)

Bladder V20 (%volume)

64.87 3.45 18.92

324.76 12.06 66.07

1.16 0.16 0.88

54.10 2.93 16.07

74.59 4.18 22.87

296.36 12.42 68.01

0.77 0.19 1.05

57.90 3.58 19.59

57.43 2.89 14.75

285.88 10.09 51.47

0.09 0.02 0.10

43.22 2.55 13.01

69.59 3.98 20.30

276.78 10.40 53.04

0.20 0.05 0.28

48.46 3.40 17.32

66.88 3.57 19.53

346.10 13.86 75.89

1.57 0.26 1.43

56.70 2.83 15.49

70.96 4.04 22.11

259.76 10.26 56.21

0.32 0.08 0.46

52.83 3.51 19.24

62.66 3.21 17.60

325.54 13.47 73.79

1.42 0.22 1.21

52.16 2.80 15.31

68.81 3.92 21.48

249.32 9.22 50.50

0.17 0.05 0.28

50.57 3.47 19.01

66.38 5.57 23.62

325.82 21.92 92.98

1.70 0.43 1.82

51.88 4.47 18.95

71.00 5.42 23.00

297.60 20.60 87.40

1.15 0.35 1.50

54.98 4.42 18.76

Abbreviations as in Table 1.

98

I. J. Radiation Oncology

● Biology ● Physics

Volume 54, Number 1, 2002

Table 3. Conformity indexes for various optimization schemes Conformity index

Mean SE (n ⫽ 30; 30; 26; 30; 30; 18) SD Minimum Maximum

Prostate volume (cm3)

IPSA 6.0 Gy

IPSA 6.5 Gy

Geo1

Geo2

Geo3

45.75 3.54 19.41 19.60 102.84

1.51 0.03 0.14 1.09 1.78

1.27 0.01 0.06 1.14 1.41

1.57 0.03 0.19 1.15 2.15

1.39 0.02 0.12 1.10 1.70

1.61 0.06 0.27 1.17 2.10

Abbreviations as in Table 1.

this dosimetric advantage should translate to a better predictability of treatment outcome. High CI values were observed for a few patients with both Geo1 and Geo3 planning. This is not surprising for Geo3 in which no CT data were used for plan normalization. In the case of Geo1, these extreme cases correspond to small prostate volumes, in which the 0.5-cm margin taken on each side of the CTV includes normal tissues that represent a substantial fraction of the target volume. Very poor target coverage (V100 values ⬍80%) was observed in 4 patients with the Geo3 method. One patient had a V100 value as low as 72.71% and still had a relatively high CI (1.62). In that particular case, the inferior limit defined on the X-ray simulator films on the basis of the radiopaque marker location was inappropriate and did not entirely encompass the CTV, leading to as much as 20% of the prostate untreated (Fig. 10). In the Geo2 data, although the average V100 value was lower than with the IPSA plans, the method is nevertheless quite conformal (low CI) and highly reproducible (data clustering). Therefore, by carefully selecting the active dwell positions, an experienced planner can actually create

very good dose distributions with respect to conformity and CTV dose coverage. However, in addition to being less dependent on the accuracy of catheter positioning, the IPSA algorithm has a dramatic clinical advantage over the Geo2 method (Fig. 9b). Figure 9b is a plot of the prostate V100 value as a function of the urethral V150 value for all 30 patients in this study. When the urethral dose is considered, even Geo2 is no longer a match with the IPSA algorithm. The average urethral V150 value is even higher in the Geo2 plan than in the Geo1 plan (Table 1). The Geo2 method behaves in this way because some of the dwell positions used for Geo1 were manually turned off (especially in the extremities of the prostate) to increase dose conformity. By doing so, fewer dwell positions were available in the periphery of the CTV to deliver the same prescription dose to the prostate contours. In these conditions, the geometric optimization process increases the hot region in the middle of the prostate, thereby raising the urethral V150 values. Only the IPSA algorithm can generate a conformal dose distribution and at the same time limit the dose to the urethra because of the dose constraint included in the ob-

Fig. 10. Example of a poor treatment plan, which can sometimes be obtained when working without the anatomic information provided by CT. 3D view of a Geo3 plan showing the prescription isodose (wireframe) and contoured organs. Green indicates CTV; purple, the rectum; and yellow, the bladder.

Inverse planning for HDR boost of the prostate

jective function. Manually tweaking the dwell times to create dose distributions that would provide the same urethral protection characteristic of IPSA plans is practically impossible. The process of trial and error necessary to achieve such distributions would be much too long to be clinically useful. Figure 9b illustrates further that Geo3 sometimes generates plans that are clinically undesirable. For 1 patient (the square in the top right corner), the urethral V150 value reached as high as 95.37%. For this particular patient, both the superior and inferior limits included too much tissue, leading to a CI value of 2.01. Although this plan presented an excellent prostate V100 value (98.9%), the urethra would have been severely overdosed and the surrounding tissues would have received an unnecessary dose, diminishing the potential advantage of BT. Those kinds of treatment are a lot less than optimal and the use of anatomy-based treatment planning should be thought of as mandatory. Considering the unreliable nature of the Geo3 method, we believe it is clinically unacceptable to treat without the full anatomic information provided by CT. Once all the steps of the procedure had been carefully examined and optimized, we found that planning an HDR prostate boost could be performed in a fast, secure, and effective manner. CT reconstruction together with the IPSA algorithm now allow us to perform the entire planning procedure routinely, including image acquisition, anatomy contouring, catheter reconstruction, dosimetry, and quality assurance procedures in ⬍2 hours. Our inverse-planning algorithm is one of the key elements making the procedure effective. Indeed, with ⬍2 min of computation time, the

● B. LACHANCE et al.

99

dose optimization process is now a negligible fraction of the time necessary to create a plan. The algorithm ensures that the highest quality of planning is used for every patient and that the final result will be independent of the dosimetrist’s experience. It will also produce a plan that is consistent from patient to patient, providing patients with the best plans and allowing comparisons between them. This dosimetric study reported the first clinical experience with anatomy-based inverse-planning dose optimization for HDR prostate boost. It is also an effort to demonstrate the need for a quantitative understanding of the dosimetric quality of HDR implants (41). CONCLUSIONS The feasibility and safety of anatomy-based inverse planning for the creation of HDR plans has been clearly established. We demonstrated that our inverse-planning algorithm produces superior HDR plans from a dosimetric point of view than do more conventional geometric optimizations for adenocarcinoma of the prostate. The algorithm generates highly conformal dose distributions that provide good CTV coverage (high V100 values). The urethral protection included in the objective function is a major feature of the algorithm. With IPSA, raising the prescription by about 1.0 Gy/fraction would still provide a similar urethral dose as that obtained with geometric optimizations. These results suggest a possible advantage in using IPSA for prostate HDR boost over standard geometric optimization and should allow us to escalate the HDR dose to the prostate without increasing undesirable side effects.

REFERENCES 1. Holm HH, Pedersen JF, Hansen H, et al. Transperineal I-125 iodine seed implantation in prostatic cancer guided by transrectal ultrasonography. J Urol 1983;130:283–286. 2. Blasko JC, Radge H, Schumacher D. Transperineal percutaneous iodine-125 implantation for prostatic carcinoma using transrectal ultrasound and template guidance. Hypertherm Oncol 1987;3:131–139. 3. Bertermann H. The European experience: Use of transrectal ultrasound in diagnosis and management of prostate cancer. In: Lee F, McLeaey RD, editors. The use of transrectal ultrasound in diagnosis and management of prostate cancer. Prog Clin Biol Res 1987;237:177–194. 4. Schwarzer JU, Hofmann R, Kneschaurek P, et al. Munich experience with 192Ir brachytherapy. In: Bruggmoser G, Sommerkamp H, Mould RF, editors. Brachytherapy of prostatic cancer. Veenendaal, The Netherlands: Nucletron; 1991. p. 199 –204. 5. Bertermann H, Wirth B, Hansen H, et al. Ultrasound guided HDR brachytherapy: The Kiel experience. In: Prostate cancer in the 1990s: Advances in diagnosis and treatment. Symposium abstract. White Plains, New York: 1991. 6. Mate TP, Gottesman JE, Hatton J, et al. High dose-rate afterloading 192Iridium prostate brachytherapy: Feasibility report. Int J Radiat Oncol Biol Phys 1998;41:525–533. 7. Demanes DJ, Rodriguez RR, Altieri GA. High dose rate prostate brachytherapy: The California endocurietherapy (CET) method. Radiother Oncol 2000;57:289 –296.

8. Borghede G, Hedelin H, Holmang S, et al. Irradiation of localized prostatic carcinoma with a combination of high dose rate iridium-192 brachytherapy and external beam radiotherapy with three target definitions and dose levels inside the prostate gland. Radiother Oncol 1997;44:245–250. 9. Stromberg J, Martinez A, Gonzalez J, et al. Ultrasound-guided high dose rate conformal brachytherapy boost in prostate cancer: Treatment description and preliminary results of a phase I/II clinical trial. Int J Radiat Oncol Biol Phys 1995;33: 161–171. 10. Hoskin PJ. High dose rate brachytherapy boost treatment in radical radiotherapy for prostate cancer. Radiother Oncol 2000;57:285–288. 11. Bertermann H, Brix F. Ultrasonically guided interstitial high dose rate brachytherapy with Ir-192: Technique and preliminary results in locally confined prostate cancer. In: Martinez AA, Orton CG, Mould R, editors. Brachytherapy HDR and LDR: Remote afterloading state of the art. Leerson, The Netherlands: Nucletron International BV; 1990. p. 281–303. 12. Stromberg JS, Martinez AA, Horwitz EM, et al. Conformal high-dose rate iridium-192 boost brachytherapy in locally advanced prostate cancer: Superior prostate-specific antigen response compared with external beam treatment. Cancer J Sci Am 1997;3:346 –352. 13. Borghede G, Hedelin H, Holmang S, et al. Combined treatment with temporary short-term high dose rate iridium-192 brachytherapy and external beam radiotherapy for irradiation

100

14. 15.

16.

17.

18. 19. 20.

21.

22. 23.

24.

25.

26. 27. 28.

I. J. Radiation Oncology

● Biology ● Physics

of localized prostatic carcinoma. Radiother Oncol 1997;44: 237–244. Paul R, Hofmann R, Schwarzer JU, et al. Iridium-192 highdose-rate brachytherapy—A useful alternative therapy for localized prostate cancer? World J Urol 1997;15:252–256. Dinges S, Deger S, Koswig S, et al. High-dose rate interstitial with external beam irradiation for localized prostate cancer— Results of a prospective trial. Radiother Oncol 1998;48:197– 202. Kovacs G, Wirth B, Bertermann H, et al. Prostate preservation by combined external beam and HDR brachytherapy at node negative prostate cancer patients—An intermediate analysis after 10 years experience [Abstract]. Int J Radiat Oncol Biol Phys 1996;36(Suppl.1):198. Martinez AA, Kestin LL, Stromberg JS, et al. Interim report of image-guided conformal high-dose-rate brachytherapy for patients with unfavorable prostate cancer: The William Beaumont phase II dose-escalating trial. Int J Radiat Oncol Biol Phys 2000;47:343–352. Brenner DJ, Hall EJ. Fractionation and protraction for radiotherapy of prostate carcinoma. Int J Radiat Oncol Biol Phys 1999;43:1095–1101. Paulson DF, Moul JW, Walther PJ. Radical prostatectomy for clinical stage T1-2 N0 M0 prostatic carcinoma: Long term results. J Urol 1990;144:1180 –1184. Stamey TA, Ferrari MK, Schmid HP. The value of serial prostate specific antigen determinations 5 years after radiotherapy: Steeply increasing values characterize 80% of patients. J Urol 1993;150:1856 –1859. Hanks GE, Diamond JJ, Krall JM, et al. A 10-year follow-up of 682 patients treated for prostate cancer with radiation therapy in the United States. Int J Radiat Oncol Biol Phys 1987;13:499 –505. Forman JD, Oppenheim T, Liu H. The frequency of residual neoplasm in the prostate following three-dimensional conformal radiotherapy. Prostate 1993;23:235–243. Pilepich MV, Caplan R, Byhardt RW, et al. Phase III trial of androgen suppression using goserelin in unfavorable-prognosis carcinoma of the prostate treated with definitive radiotherapy: Report of Radiation Therapy Oncology Group protocol 85-31. J Clin Oncol 1997;15:1013–1021. Hanks GE, Hanlon AL, Pinover WH, et al. Survival advantage for prostate cancer patients treated with high dose threedimensional conformal radiotherapy. Cancer J Sci Am 1999; 5:152–158. Zelefsky MJ, Leibel SA, Gaudin PB, et al. Dose escalation with three-dimensional conformal radiation therapy affects the outcome in prostate cancer. Int J Radiat Oncol Biol Phys 1998;41:491–500. Stock RG, Stone NN, Tabert A, et al. A dose-response study for I-125 prostate implants. Int J Radiat Oncol Biol Phys 1998;41:101–108. Pollack A, Zagars GK. External beam radiotherapy dose response of prostate cancer. Int J Radiat Oncol Biol Phys 1997;39:1101–1018. Vigneault E, Pouliot J, Laverdie`re J, et al. Electronic portal imaging device detection of radiopaque markers for the eval-

Volume 54, Number 1, 2002

29.

30.

31.

32.

33.

34.

35.

36.

37. 38. 39. 40.

41.

uation of prostate position during megavoltage irradiation: A clinical study. Int J Radiat Oncol Biol Phys 1997;37:205–212. Vigneault E, Beaulieu L, Aubin S, et al. Online image guided external beam radiation therapy for the treatment of prostate cancer using electronic portal imaging: Prostate motion and setup evaluation. Int J Radiat Oncol Biol Phys 2001;51(Suppl. 1):317–318. Aubin S, Pouliot S, Girouard LM, et al. Performance of an automatic radio-opaque markers detection algorithm for online setup correction using a standard video-based Electronic Portal Imaging Device. Int J Radiat Oncol Biol Phys 2001; 51(Suppl. 1):386. Hsu IC, Pickett B, Shinohara K, et al. Normal tissue dosimetric comparison between HDR prostate implant boost and conformal external beam radiotherapy boost: Potential for dose escalation. Int J Radiat Oncol Biol Phys 2000;46:851– 858. Kestin LL, Martinez AA, Stromberg JS, et al. Matched-pair analysis of conformal high-dose-rate brachytherapy boost versus external-beam radiation therapy alone for locally advanced prostate cancer. J Clin Oncol 2000;18:2869 –2880. Martinez AA, Pataki I, Edmundson G, et al. Phase II prospective study of the use of conformal high-dose-rate brachytherapy as monotherapy for the treatment of favorable stage prostate cancer: A feasibility report. Int J Radiat Oncol Biol Phys 2001;49:61– 69. Roach M, Lu J, Lawton C, et al. A phase III trial comparing whole-pelvic (WP) versus prostate only (PO) radiotherapy and neoadjuvant versus adjuvant total androgen suppression (TAS): Preliminary analysis of Radiation Therapy Oncology Group (RTOG) 94-13. Int J Radiat Oncol Biol Phys 2001; 51:3. Lachance B, Be´ liveau-Nadeau D, Lessard E´ , et al. Feasibility and dosimetric evaluation of HDR prostate boost using inverse planning dose optimization: Early clinical experience. Presented at the American Brachytherapy Society Annual Meeting, Vancouver, 2001. Lachance B, Be´ liveau-Nadeau D, Lessard E´ , et al. Early clinical experience with anatomy-based inverse planning dose optimization for HDR boost of the prostate. Med Phys 2001; 28:1280. Pouliot J, Tremblay D, Roy J, et al. Optimization of permanent 125I prostate implants using fast simulated annealing. Int J Radiat Oncol Biol Phys 1996;36:711–720. Taschereau R, Roy J, Pouliot J. Monte Carlo simulation of prostate implants to improve dosimetry and compare planning methods. Med Phys 1999;26:1952–1959. Pouliot J, Taschereau R, Coˆ te´ C, et al. Dosimetric aspects of permanents radioactive implants for the treatment of prostate. Phys Can 1999;55:61– 68. Lessard E, Pouliot J. Inverse planning anatomy-based dose optimization for HDR-brachytherapy of the prostate using fast simulated annealing algorithm and dedicated objective function. Med Phys 2001;28:773–779. Kini VR, Edmundson GK, Vicini FA, et al. Use of threedimensional radiation therapy planning tools and intraoperative ultrasound to evaluate high dose rate prostate brachytherapy implants. Int J Radiat Oncol Biol Phys 1999;43:571–578.