Dosimetric intercomparison for two Australasian clinical trials using an anthropomorphic phantom

Int. J. Radiation Oncology Biol. Phys., Vol. 52, No. 2, pp. 566 –579, 2002 Copyright © 2002 Elsevier Science Inc. Printed in the USA. All rights reserved 0360-3016/02/$–see front matter

PII S0360-3016(01)02682-7

PHYSICS CONTRIBUTION

DOSIMETRIC INTERCOMPARISON FOR TWO AUSTRALASIAN CLINICAL TRIALS USING AN ANTHROPOMORPHIC PHANTOM TOMAS KRON, PH.D.,*† CHRIS HAMILTON, M.D.,*‡ MARIANNE ROFF,*

AND

JIM DENHAM, M.D.*‡

*Centre for Clinical Radiation Research and TROG Central Office, Newcastle Mater Misericordiae Hospital, Waratah, NSW, Australia; † School of Mathematical and Physical Sciences and ‡Faculty of Medicine, University of Newcastle, Callaghan, NSW, Australia Introduction: Many different factors can affect the accurate delivery of dose to the clinical target volume in radiotherapy. This is particularly important in the context of multicenter clinical trials where different equipment and techniques may be used for supposedly identical treatments. A dosimetric intercomparison employing an anthropomorphic phantom (level III dosimetric intercomparison) can be used to check many of the factors that could affect treatment by mimicking the radiotherapy pathway of a patient as closely as possible. Methods and Materials: An anthropomorphic phantom (ART) was taken to 18 radiotherapy centers in Australia and New Zealand and treated for two different treatment scenarios based on current clinical trials of the Trans-Tasman Radiation Oncology Group (TROG): a two-field treatment of a carcinoma of the tonsil (TROG 91.01), and a four-field prostate treatment (TROG 96.01). The dose distribution was assessed in two consecutive treatments using thermoluminescence dosimeters (TLDs) placed throughout the target volume and in “critical” structures such as the lens of the eye or the rectum. The study also included a check of absolute dose calibration in a slab phantom (level I dosimetric intercomparison). The influence of a variety of treatment parameters on the dose homogeneity in the target and the measured dose in the target and the critical organs was evaluated. Results: The dose measurements confirmed that in all participating centers the correct dose was delivered to the ICRU reference point (tonsil: 99.8 ⴞ 2.3%; prostate: 100.9 ⴞ 1.9% [1 SD]). Also the absolute dose calibration and the mean dose in the target volume were within the specified action levels of ⴞ 5% for all participating centers. No influence of shielding, beam modifiers, beam weighting, treatment planning approach (CT, 2D, 3D), and type of equipment used on the dose in the target and its homogeneity could be demonstrated. However, treatment technique and energy used influenced the dose to the critical organs. It was shown that the interpretation of results could be improved by including two complementary treatment scenarios and a level I intercomparison with the level III dosimetric intercomparison. Conclusion: The study demonstrated the feasibility of a level III dosimetric intercomparison service at a cost of approximately $US 1000 per center in Australasia. It confirmed that the dose delivered by all participating centers was as intended in the two treatment scenarios chosen. While this provides reassurance to the oncology community and the general public, the service must be extended to all centers and other potentially more complex treatment scenarios. The present study has built the foundation for this by establishing a baseline and action levels and suggesting improvements in phantom design which will be included in future TROG quality assurance exercises. © 2002 Elsevier Science Inc. Quality assurance, Multicenter clinical trials, Thermoluminescent dosimetry.

INTRODUCTION As a localized cancer treatment modality, radiotherapy relies on the accurate delivery of radiation dose to the target volume. While there may be differences in clinical practice, it must be ensured that once the target dose is specified it must be delivered as precisely as possible (1–3). This need

is highlighted in multicenter clinical trials where the assumption is made that all patients entered into the trial are treated according to the agreed protocol. However, the delivery of radiation is a complex procedure, which relies on the individual patient, the radiation sources available in different centers and a complex computerized treatment planning process. The introduction of more sophisticated

Reprint requests to: A/Prof. Tomas Kron, London Regional Cancer Centre, Department of Clinical Physics, 790 Commissioners Road East, London, Ontario N6A 4L6, Canada. Tel: (519) 685-8600 ext. 53144; Fax: (519) 685-8658; E-mail: tomas.kron@ lrcc.on.ca Presented in part at the 41st Annual Meeting of the American Association of Physicists in Medicine, Nashville, July 1999, abstract in Med Phys 1999;26:1162. The financial support of the following organizations is acknowl-

edged: the Newcastle Mater Misericordiae Hospital, the TransTasman Radiation Oncology Group, the Royal Australian and New Zealand College of Radiologists, and the University of Newcastle Research Management Council. Acknowledgment—We would like to thank S. Bazley and C. Hood for reading of TLDs and all participating radiotherapy centers for their collaboration. Received Apr 11, 2001, and in revised form Sep 19, 2001. Accepted for publication Sep 20, 2001. 566

Dosimetric intercomparison using an anthropomorphic phantom

● T. KRON et al.

567

Fig. 1. The role of different levels of dosimetric intercomparison within the treatment chain of radiotherapy.

treatment techniques such as conformal and intensity-modulated radiotherapy (IMRT) has highlighted this potential problem (4, 5). In general, the areas of uncertainty in dose delivery can be divided into the following categories: (1) absolute calibration at the reference point; (2) relative dosimetry (e.g., depth dose, off axis ratio, wedge factors); (3) treatment planning system; (4) treatment unit variations; (5) patient setup There are many quality assurance (QA) protocols which recommend procedures to verify that none of these uncertainties affect treatment delivery at an unacceptable level (6 – 8), and every radiotherapy center employs a set of QA activities to this effect. However, it is good practice to also conduct independent checks of certain parameters, e.g., by dosimetric intercomparisons. The need for independent checks has recently been highlighted in Safety Report N17 of the International Atomic Energy Agency (IAEA), which details more than 100 radiation accidents in radiotherapy (9). In general, three different levels of dosimetric intercomparisons can be defined: Level I: This constitutes an independent check of beam calibration under reference conditions in a physical phantom, usually water. It typically consists of only one measurement point. Such services are offered, e.g., by the Radiologic Physics Center in Houston or the IAEA (10, 11). Level II: A level II intercomparison verifies not only the dose under reference conditions but also the accuracy of some other factors required for treatment planning. As electron dosimetry intrinsically requires a verification of depth dose to determine the appropriate point of measurement and the electron energy, electron intercomparisons are necessarily a level II dosimetric exercise (12). The measurements in level II dosimetric intercomparisons are done in a physical phantom; however, this may include inhomogeneities or surface contour changes (13). Level III: A level III intercomparison requires the use of an anthropomorphic phantom that is planned and treated as similar to a patient as possible. This can be a full anthropomorphic phantom (14) or a semianatomic phantom (15– 17). The advantage of a level III dosimetric intercomparison is that the entire treatment chain from the acquisition of diagnostic images to the treatment setup and delivery can be verified. In addition to points 1– 4 in the list above, a level

III dosimetric intercomparison allows one to study the influences of differences in treatment technique and equipment available in participating centers. The disadvantage of verifying many steps in one procedure is that it is often difficult to identify which step has contributed to a particular outcome. As such, it is required to repeat measurements and check smaller segments of the treatment chain if the overall level III check identifies a problem. Figure 1 illustrates the role of a level I and level III dosimetric intercomparison within the treatment QA chain of radiotherapy. There has been a good record of level I dosimetric intercomparisons in Australia and New Zealand which found that the beam calibration in nearly all participating centers was within ⫾ 1.5% of the IAEA reference dose (18). Also a level II intercomparison has recently been performed for electron dosimetry. However, neither survey checked the treatment planning process, which has become an increasingly complex and important part of the radiotherapy chain. A level III dosimetric intercomparison was reported more than 10 years ago for head-and-neck treatment in Europe by Johansson et al. (19). However, since then treatment techniques have developed and new technology has been introduced to radiotherapy. Recently, there have been studies conducted in Australia on treatments for Hodgkin’s disease (15), an intercomparison of dose at the junction of head and neck treatments (20), and a survey of dose distribution in tangential breast irradiation in 10 radiotherapy centers (14). However, none of these Australasian studies has specifically aimed to verify the dose delivered in clinical trials. The Trans-Tasman Radiation Oncology Group (TROG), which is Australia and New Zealand’s multicenter trials group, therefore decided to implement a level III intercomparison with several specific goals: 1. To establish the feasibility and costs of such intercomparison in two countries where many centers are separated by vast distances. 2. To compare absolute dose under reference conditions. 3. To compare doses at the ICRU point in simulated treatment conditions in a phantom. 4. To compare dose heterogeneities in the simulated treatment conditions. 5. To generate an inventory of equipment and techniques

568

I. J. Radiation Oncology

● Biology ● Physics

Volume 52, Number 2, 2002

Fig. 2. Treatment setup and number of TLDs employed in the head-and-neck radiotherapy scenario. The figure shows a schematic representation of the phantom.

used and establish their potential impact on the dosimetric “chain”. 6. To establish whether doses reported in the randomized clinical trial situation by each center can be confirmed and, if not, to assist in the identification of problems and their rectification. Another important feature of the present study was the inclusion of two different treatment scenarios in the study. If chosen appropriately, they can check different radiation qualities and different parts of the treatment planning process. In connection with the level I dosimetric intercomparison included in the study, this facilitates the interpretation of results if a problem in the overall outcome was detected. The choice of a head and neck and a prostate treatment in the present study was aimed at verifying different radiation qualities as well as computed tomography (CT) and simulator based planning. In addition, it was anticipated that the scenarios chosen would allow testing of open and blocked fields, wedges, and different patient geometries. METHODS AND MATERIALS Treatment scenarios The study investigated the dose distribution in two treatment scenarios based on two current clinical trials: 1. Treatment for carcinoma of the tonsil and neck nodes using two parallel opposed lateral fields of 10 ⫻ 8 cm2 (l ⫻ w) size. The treatment scenario was chosen to fit TROG protocol 91.01 (21). X-rays from a linear accelerator with an energy of 4 or 6 MV were to be used in an

isocentric setup. The field center was specified on the anterior inferior corner of the fourth cervical vertebrae, a landmark that could be easily determined during fluoroscopic screening. Treatment setup, wedges, beam weighting, and immobilization devices were to be used in accordance with local practice; however, no shielding was to be used. 2. A four orthogonal field prostate treatment treating a volume of 8 ⫻ 8 ⫻ 8 cm3. The treatment scenario was chosen to fit TROG protocol 96.01 (22). The treatment was specified using megavoltage equipment with a minimum focus to skin distance (FSD) of 80 cm. The phantom was to be set up in a prone position directly on the couch top as this position ensured a stable positioning of the phantom. The target isocenter was to be located midline and centered on a curvilinear line joining the superior borders of each superior pubic ramus. The isocenter location was not specified in anterior/posterior direction to reflect typical clinical variations in some measure. Rectal shielding was at the centers’ discretion. Beam weighting, wedges, and multileaf collimation (MLC) could be employed according to local practice. The treatment scenarios are illustrated in Figs. 2 and 3. All treatments were to be performed twice with a prescribed dose to the ICRU reference point (23) of 1 Gy each to avoid problems with thermoluminescence supralinearity (24). Participating centers Eighteen radiotherapy centers in Australia and New Zealand were invited to participate in the study. The centers,

Dosimetric intercomparison using an anthropomorphic phantom

● T. KRON et al.

569

Fig. 3. Treatment setup and number of TLDs employed in the prostate treatment scenario. The figure shows a schematic representation of the phantom.

which constitute about half of all radiotherapy facilities in Australasia, are listed in Table 1. They represent 17 public and one private health care service from an area spanning 5000 ⫻ 3000 km2 from Auckland to Perth (East/West) and Townsville to Christchurch (North/South). Table 1 also shows planning systems, treatment units, and radiation qualities used in the respective centers for this phantom study. The centers using a different planning system for the two treatment scenarios were doing so either because of compensator design or CT capability of their systems. Dose measurements Thermoluminescence dosimeters (TLDs, lithium fluoride doped with magnesium and titanium [LiF]: Mg, Ti, 3.1 ⫻ 3.1 ⫻ 0.9 mm3) were used in a commercial anthropomorphic phantom (“ART,” Radiologic Support Devices, Long Beach, CA). The phantom, which consists of 2.5-cm-thick slices, was assembled in two separate stacks for the two treatment scenarios as illustrated in Figs. 4 and 5. The number and location of the TLDs in the two phantoms are indicated in the figures. The locations labeled with bold letters in Figs. 4 and 5 were assessed using two or three TLDs in a purpose built holder. The TLD locations were chosen to allow the dose assessment at ICRU reference point as well as the determination of mean target dose and dose homogeneity. The phantom was transported by one member of the organizing team (T. Kron) to participating centers between June 1998 and March 2000. The phantom treatment was performed exactly in the same way a patient would be

treated in the participating centers. This included the acquisition of patient data using CT or simulation, treatment planning, and setup. Marks on the phantom were made on tape which was removed after the treatment to avoid any one center’s influence on the treatment from the other centers. In most cases the phantom was treated in the early evening after normal patient treatment was completed for the day. All tasks were performed by local staff with the member of the organizing team as an observer. All TLDs were individually calibrated and kept in batches of 100. The average reproducibility of individual TLDs was 2% (single standard deviation). The TLDs were read using a Harshaw 5500 reader and annealed in a dedicated annealing oven. More details on the TLD measurements can be found in Kron 1994 and 1999 (20, 24). Twenty-six TLDs were used in the tonsil treatment and 29 for the prostate phantom as indicated in Figs. 2 and 3. The locations of the TLDs in the prostate treatment were adjusted in slice 33 to match the location of the isocenter in anterior/posterior direction, which was not specified in the treatment scenario. Either locations 1 to 9 was chosen with 3 TLDs in the central location 5 or locations 4 to 12 with 3 TLDs in location 8 (compare Fig. 5). Each of the treatments was performed twice with the TLDs changed between treatments by the member of the organizing team. Two batches of 100 TLDs were used per center. This allowed setting aside at least 20 TLDs from each batch for irradiation under reference conditions in the participating centers and in parallel at the TROG Quality Assurance Center associated with the Newcastle Mater Hospital (center 10 in Table 1) using a

570

Table 1. List of participating centers Energy for tonsil (MV)

Energy for prostate (MV)

Target location for prostate evaluation

Alfred Hospital, Melbourne Auckland Hospital Charles Gairdener Hospital, Perth Christchurch Hospital East Coast Radiotherapy Center Geelong Hospital

1 2 3

Varian 2100C Varian 2100C/D Varian 2100C

Plato ADAC Pinnacle CMS Focus

6 6 6

18 18 6

Mid Anterior Mid

4 5

Varian 2100C Varian 2100C

CMS Focus Cadplan

6 6

18 10

Anterior Mid

6

Varian 2100C

Theraplan V

6

Anterior

Illawarra Cancer Care Centre, Wollongong Liverpool Cancer Care Centre, NSW Nepean Hospital, NSW Newcastle Mater Hospital

7

Varian 2100C

GE Target II

6

6AP/PA, 18lat 6

8

Mevatron MXE

Cadplan

6

6

Anterior

9 10

Varian 2100C Varian 600C (4MV), Varian 1800

RadPlan Theraplan V, ADAC Pinnacle

6 4

18 6 and 18

Posterior Posterior

Peter MacCallum Cancer Institute, Melbourne Prince of Wales, Sydney

11

Varian 2100C/D

6

18

Mid

12

4

18

Posterior

Queensland Radium Institute, Mater Hospital Royal Adelaide Hospital Royal Brisbane Hospital Townsville General Hospital Wellington Hospital

13

Varian 1800, Varian 4 Varian 600C

Theraplan V, CadPlan GE Target II RadPlan

6

6

Mid

14 15 16

Varian 6 Philips SL20 Philips SL18

GE Target Plato Plato

6 6 6

6 6 10

Anterior Anterior Anterior

17

Varian 2100C/D

Cadplan

6

6

Anterior

Westmead Hospital, NSW Total number of centers:

18 18

Philips SL 75

RadPlan Total number of treatments:

6 19

6 21

Anterior

Mid

Comment

The prostate was treated using two lateral arcs

In total 4 prostate and 2 tonsil treatments were performed using different energies and shielding arrangements

The prostate was treated using two lateral arcs

Volume 52, Number 2, 2002

Number

● Biology ● Physics

Center

I. J. Radiation Oncology

Treatment planning system

Treatment unit (s)

Dosimetric intercomparison using an anthropomorphic phantom

● T. KRON et al.

571

Fig. 4. TLD locations in the head-and-neck phantom treatment: S1– 6 superior target, M1: mid target (2 TLDs), T1– 8 target slice, I1–3 inferior target. Additional TLDs were placed in the lens of the left eye. The spacing between adjacent TLD locations was 3 cm. The TLD locations indicated in bold contain more than one TLD in a single holder per irradiation.

radiation quality matched to the one used in the center. The beams available in Newcastle were 4 MV (D20/D10 ⫽ 0.542), 6 MV (0.583), and 18 MV (0.667). A fading correction was employed if there was a time difference between the reference irradiation at the center and Newcastle. The dose readings in the phantom were evaluated twice, relative to the dose given under reference conditions in the respective center and relative to the dose delivered in the TROG QA Center. The uncertainty of the TLD determination depends on the number of TLDs evaluated for a particular measurement, the reproducibility of the TLD response, and systematic sources of error such as fading, energy dependence of the TLDs, and the uncertainty of the dose delivered under reference conditions. Evaluation The readings of the two TLDs in the same location in the two consecutive treatments were averaged. In addition to the dose at the location of the “critical” organs (tonsil: lens; prostate: rectum, femoral heads), the following dose measurements were recorded: 1. Comparison of absolute dose under reference condition in the center and at the TROG QA Center.

2. The dose at the ICRU reference point in the phantom. For the head-and-neck treatment, the two TLDs from location M1 in Fig. 3 were averaged. In the prostate treatment, the 3 TLDs from location 5 or 8 were chosen depending on the position of the isocenter in the participating radiotherapy unit. Due to the averaging of the TLDs in the two consecutive identical runs, the readings of at least 4 TLDs were evaluated to obtain the ICRU reference dose. The dose at the ICRU reference point was evaluated in two ways: first, in comparison to the TLD calibration at the respective center, illustrating the internal consistency of dose delivery at the center; and second, in comparison to the TLD calibration at the TROG QA center. The latter illustrates the overall variation of dose delivered in all participating centers at the reference point compared to a single reference calibration. 3. The mean dose in the target volume. This is the average of all TLDs in the target volume as illustrated in Figs. 4 and 5. In the head-and-neck treatment, the target volume consisted of 12 locations (S3 to S6, M and T4 to 8). The more anterior locations S1, S2, T1 to T3, and the inferior target locations I1 to I3 proved to be already too close to the field edge to give reliable results (compare Fig. 4). In

572

I. J. Radiation Oncology

● Biology ● Physics

Volume 52, Number 2, 2002

Table 1 lists which prostate target location was evaluated in each center. 4. The dose homogeneity was recorded as the single standard deviation of all TLDs in the target region defined as above for both treatment scenarios. It must be noted that in particular in the prostate treatment the homogeneity reflects primarily the dose distribution in the central plane. Action levels An “action” level of ⫾ 5% of the specified dose was chosen at the onset of the study. This action level was to be applied to two measurement results: 1. The absolute dose comparison: In this case, the combined uncertainty allows for the TLD measurement uncertainty (⫾ 3%, 2 SD), the uncertainty of dose delivered in the participating centers (3%) and the reference center 10 (3%). Assuming normal distribution and independence of all these uncertainties of each other, this adds up to approximately ⫾ 5% combined uncertainty on a 95% confidence level. 2. The dose at the ICRU reference point compared to the calibration performed in the respective center: The action level here includes random influences, such as uncertainty of the dose delivery, and the TLD measurement and systematic components, such as treatment planning and transfer of patient from planning to treatment. As the combined uncertainty is difficult to assess, the action level was chosen to reflect the often quoted aim to deliver radiation in radiotherapy with a combined uncertainty of less than ⫾ 5% (2, 3).

Fig. 5. TLD locations in the prostate phantom treatment: S1,2 superior target; L1,2 left lateral beam; A1 anterior beam; R1,2 right lateral beam; F1, 2 femoral heads (2 TLDs each); 1–12 center slice; I1,2 inferior target; R rectum (3 TLDs). The spacing between adjacent TLD locations was 3 cm. The TLD locations indicated in bold contain more than one TLD in a single holder per irradiation.

the prostate treatment, the target volume evaluation depended on the positioning of the isocenter in anterior/ posterior direction. Three different averages were used to account for this: ● ●

●

“posterior”—locations: S1,2; M4 to 9; I1, 2; in this case 12 TLDs were averaged (n ⫽ 6); “mid”—locations: S1, M1 to 6, M8; I1; in this case the number of TLDs evaluated varied between 8 and 12 depending on the loading (n ⫽ 6); “anterior”—locations: S1, M1 to M6, I1; in this case the number of TLDs evaluated varied between 5 and 10 depending on the loading (n ⫽ 9).

A protocol was in place if either of the action levels defined above was crossed. If a calibration difference exceeding 5% was found (action level 1), the respective center was to be informed immediately by phone and the difference discussed. If it could not be resolved clearly, the measurement was to be repeated first by mailed TLDs and then using a site visit with a calibrated ionization chamber. If the dose in the anthropomorphic phantom was found to be out of the specified limits (action level 2 above), the reason for the discrepancy was to be established by first carefully reexamining all the treatment documentation and discussion of the results with the radiotherapy facility in question. As the study included two treatment scenarios, important information can be gathered from the difference (or otherwise) in outcome. If the reason for the discrepancy could not be resolved and rectified, additional measurements using physical phantoms were to be performed to identify the source of the discrepancy. RESULTS The level III dosimetric intercomparison A typical site visit took 1 to 2 days. This includes travel, patient data acquisition (CT, simulation), treatment plan-

Dosimetric intercomparison using an anthropomorphic phantom

● T. KRON et al.

573

Table 2. Results of the two-field head-and-neck treatment

Absolute dose compared to the TROG QA Centre in Newcastle (center 10) (Gy) ICRU reference point dose (Gy) for 1 Gy prescribed compared to calibration in the respective center Mean target dose compared to calibration in the respective center (Gy) Variation of dose in the target volume (%, ISD) ICRU reference point dose (Gy) for 1 Gy prescribed evaluated using the calibration at the TROG QA Centre “Lens dose” (Gy) for 1 Gy tumor dose

ning, and two treatments for each scenario (2 h machine time). At least one member of local radiotherapy staff (preferably more and a physicist) needed to be present. Including the preparation of TLDs and the treatment evaluation, the total time requirements were approximately 4 person-days for staff of the QA center and 1 person-day for staff at the participating center. Therefore, the cost per center was approximately $US 1000 including travel, accommodation, and TLD evaluation. In the 18 centers a variety of different treatment approaches were taken. In the head-and-neck treatment this includes photon energies from 4 MV to 6 MV and the use of wedges, dynamic wedges, compensators, or a combination thereof. In the prostate scenario, energies from 6 MV to 10 MV and 18 MV were used. Some centers employed blocking using lead blocks, customized shielding, or MLC. Two centers used two dynamic arc treatments over 120° to treat the prostate. These findings are summarized in Table 1. Absolute dose comparison The results of the absolute dosimetric intercomparisons are shown in Table 2 (head-and-neck treatment) and Table 3 (prostate treatment). Listed in row 1 of both tables is the ratio of doses delivered under reference conditions between participating centers and the TROG QA Center in Newcastle. This constitutes a level I dosimetric intercomparison similar to the one offered by the IAEA (10). All centers participating in the study used the IAEA TecDoc 277 protocol (25) for absolute beam calibration of their linear

Mean

Standard deviation (%)

Range

0.998

2.0

0.972–1.049

0.998 0.994 3.0

2.3 1.9

0.955–1.027 0.959–1.028 2.2–4.0

1.001 0.007

3.5

0.937–1.050 0.005–0.014

accelerators. However, four different ways of defining reference conditions were found in participating centers. The choice of reference conditions under which 100 monitor units deliver 1 Gy was influenced by the setup of the treatment planning system used in the center. The four definitions of reference conditions were: ● ● ● ●

100 cm focus surface distance (FSD), reference point at depth of maximum dose, dmax (n ⫽ 9); 100 cm focus reference point distance, build-up to dmax (n ⫽ 4); 100 cm focus reference point distance, 5 or 10 cm build-up depending on energy (n ⫽ 4); 100 cm focus reference point distance, measurement in air with build-up cap (n ⫽ 1).

In one center the reference dose measurement was repeated because the initial reading from the center was different by more than 5% from the dose delivered in Newcastle. This apparent discrepancy could be traced to the reference conditions interpreted wrongly by the visiting member of the organizing team. Head-and-neck treatment As center 10 treated the phantom twice, a total of 19 head-and-neck treatments were assessed. Four of these were performed using physical compensators (3 also combined with a wedge), 8 treatments were delivered with wedges (3 physical, 5 dynamic or virtual), and 10 treatments did not employ any beam modification. Nine of the 19 treatments

Table 3. Results of the four-field prostate treatment

Absolute dose compared to the TROG QA Centre in Newcastle (center 10) (Gy) ICRU reference point dose (Gy) for 1 Gy prescribed compared to calibration in the respective center Mean target dose compared to calibration in the respective center (Gy) Variation of dose in the target volume (%, 1 SD) ICRU reference point dose (Gy) for 1 Gy prescribed evaluated using the calibration at the TROG QA Centre “Femoral head dose”* (Gy) for 1 Gy tumor dose “Rectal dose”* (Gy) for 1 Gy tumor dose * Excluding two centers where two lateral arcs were treated.

Mean

Standard deviation (%)

Range

0.987

2.0

0.95–1.022

1.009 0.995 2.5

1.9 2.0

0.976–1.046 0.955–1.041 1.07–4.80

0.996 0.534 0.545

3.3

0.948–1.060 0.452–0.614 0.401–0.734

574

I. J. Radiation Oncology

● Biology ● Physics

Volume 52, Number 2, 2002

Fig. 6. Histogram of the measured dose at the ICRU reference point in participating centers. Two methods of evaluation are shown: the evaluation of the TLD results using standards of the respective center (black bars) and the evaluation using standards irradiated at the TROG QA center (gray bars). (a) Head-and-neck treatment. (b) Prostrate treatment.

were planned from CT datasets, either acquired using a diagnostic CT scanner (n ⫽ 7) or a Simulator CT (n ⫽ 2). The rest of the centers used manual outlines. The phantom was simulated using a simulator in 17 centers. Rows 2 and 3 in Table 2 list the dose at the ICRU reference point and the mean target dose given to the phantom. They are evaluated using the TLD standard irradiation at the respective center. It should be noted that the dose delivered can only be expected to be within ⫾ 1% of the requested dose due to the discrete nature of the monitor units delivered on the treatment units. All values in participating centers are within the expected range of ⫾ 5% of the reference value. There is no significant difference in the standard deviation and range between calibration under reference conditions, dose at ICRU reference point, and mean target dose in all participating centers. This can also be seen in Fig. 6a. In all but one center, the planned target volume could be evaluated as outlined in the previous section. This indicates that the field placement has been adequate within the relatively large uncertainty resulting from the relatively large distance between adjacent TLD positions. In only one center the target was placed more anteriorly and the 13 TLDs S1 to S5, M and T1 to T5 were evaluated for the mean target dose. This indicates that the field was misplaced by approximately 5 mm. In this center, no simulator could be used before the irradiation on the day of the study, which is likely to have contributed to the problem. Patients who are simulated would not be affected in the same way. Table 2 also shows the mean homogeneity of the dose in the target volume. It is expressed as single standard deviation of the dose variations (1 SD) in the target volume. Given is the mean of this homogeneity index in all centers. Within the limited number of experiments and the variations of treatments in different centers, no variation of dose homogeneity could be found with the use of beam-modify-

ing devices. Similarly, the use of CT scanners for planning or a three-dimensional treatment planning approach could not be shown to influence the dose homogeneity in the target. The fifth row in Table 2 shows the dose at the ICRU reference point in participating centers using a TLD calibration in Newcastle. This figure combines both the variation in absolute calibration between the center and Newcastle, and the difference in dose at ICRU reference point in the anthropomorphic phantom and the standard geometry used for TLD calibration. This results in the larger spread of data, as could be expected. The magnitude of the spread is compatible with the assumption that both the difference in dose in reference conditions and the variation of dose delivery at the ICRU reference point are normally distributed and independent of each other. As expected, the lens dose in the anthropomorphic phantom was found to be small in all participating centers. However, it could be demonstrated that the dose depends on the radiation quality. When using 4-MV photons the average scattered dose was 0.011 Gy (1.1% of the target dose) whereas the choice of a higher energy of 6 MV resulted in less scatter and an average lens dose of 0.006 Gy (0.6%).

Prostate treatment In total, 21 treatments were evaluated for the prostate cancer case because the phantom was treated 4 times in center 10 with and without shielding and using different radiation energies. All centers were using either a CT scanner (n ⫽ 19) or a Simulator CT (n ⫽ 2) for the acquisition of the phantom data for prostate treatment planning. All but two centers were treating the phantom with a four-field box technique; however, two centers were using two lateral dynamic arcs to treat the same target volume. Rectal shielding was used in the lateral fields in six treatments, two of

Dosimetric intercomparison using an anthropomorphic phantom

them employing MLC. In 10 of the 21 treatments a threedimensional dose planning approach was taken. Table 3 shows the same results for the prostate treatment as Table 2 for the head-and-neck treatment. However, the lack of specifying the target location in anterior/posterior direction caused some problems in the evaluation. As the prostate cannot be identified in the anthropomorphic phantom using radiologic means, a large variation in target positioning by more than 3 cm in anterior/posterior direction was found in different centers. This is largely a result of different protocols used to center the CT scan before planning (e.g., midplane or a specified distance from the couch top) and prompted the need to evaluate three different target volumes as indicated in “Methods and Materials”. There was no influence of target position on ICRU reference dose, mean target dose, and the homogeneity of dose in the target. As can be seen in Table 3, all parameters were within the expected range. Only when evaluated using the TLD calibration performed in Newcastle the spread of results exceeds ⫾ 5%. The ratio between dose at ICRU reference point and mean target dose was higher using 6-MV photons (1.016 ⫾ 0.005, mean ⫾ 1 standard error of the mean, n ⫽ 10) than 10 or 18 MV (1.003 ⫾ 0.006, n ⫽ 10), which indicates that more dose must be given to the reference point when using a lower photon energy to achieve the same target coverage. The center where 6 MV was used for the anterior and posterior fields and 18 MV for the laterals was not included in this evaluation. Although the choice of target location in anterior/posterior position did not significantly affect the dose distribution in the target, it had a noticeable effect on the measured rectal dose. The mean rectum dose in the centers where the target was placed posteriorly was 61% of the target dose (n ⫽ 6) while the mid and anterior target position resulted in 50% (n ⫽ 6) and 47% (n ⫽ 11), respectively. As the rectal dose was measured in all cases in the same location in the phantom, this finding reflects a placement problem rather than a true difference in rectal dose. The choice of treatment technique also affected the rectal dose, with the two centers using two lateral arcs delivering some 30% less dose to the rectum (mean 23% vs. 55% of the target dose). However, this was achieved by increasing the femoral head dose from 53% to 73% of the target dose. Figure 7 shows the distribution of rectal and femoral head doses in participating centers. Other influences on the results A large number of parameters were recorded in the course of the study. The influence of the following parameters on the ICRU reference dose, the mean target dose, the target dose homogeneity, and the dose to critical organs was evaluated as follows: 1. Reference conditions for calibration to 1 Gy. This parameter was not found to influence the results. 2. Beam energy. Except for the increased scatter to the lens due to the use of 4 MV in the tonsil treatment and

● T. KRON et al.

575

Fig. 7. Dose at the rectum and the femoral heads determined in the anthropomorphic phantom during the prostate treatment. The dose delivered in the two centers using two lateral arcs is indicated.

3.

4.

5. 6. 7. 8.

9.

10.

the ratio between ICRU dose and target dose as discussed above in the prostate treatment, no influence of beam energy on the results could be found. Shielding. Due to the variation in target position in anterior/posterior location in the prostate treatment, no influence of the use of shielding on the rectal dose could be found. Beam modifiers. Within the uncertainty of the study, no influence of beam modifiers on the results could be established. Beam weighting in the prostate treatment was not found to influence the results in the present study. The use of CT in planning of the head-and-neck treatment was not found to influence the results. Planning approach (2D vs. 3D) in the planning of the prostate treatment was not found to influence the results Manufacturer. No influence of make, model, or manufacturer of the treatment unit or the planning system on the results of participating centers could be demonstrated. Single manufacturer. It was tested if treatment results would improve if all equipment, and particularly simulator and treatment unit, were from the same manufacturer. This could not be demonstrated in the present study. Age of the treatment unit was not found to influence the results of the study.

It must be noted that the study was not designed to test

576

I. J. Radiation Oncology

● Biology ● Physics

Volume 52, Number 2, 2002

Fig. 8. Correlation of results obtained in the same center in the two different treatment scenarios. Each point represents a single center. (a) Calibration of the machines used for head-and-neck and prostate treatment. The best fit regression line to the data is shown as well as the 95% confidence intervals for the regression line. (b) The dose at the ICRU reference point and the mean target dose.

the influence of any of the parameters shown above. As shown in Fig. 1, the results of the level III intercomparison make it difficult to draw conclusions about individual factors. However, the satisfactory overall outcome makes it likely that all centers kept the influence of individual uncertainties on the result within acceptable limits.

Correlation between the two treatment scenarios Figure 8 shows the correlation of results for the two treatment scenarios in participating centers. As can be expected, there is a correlation between the calibration of the beams used in the two treatments as shown in Fig. 8a. This indicates that a center where the absolute dose calibration of the treatment unit used for head-and-neck treatment was high compared to the TROG QA center is also likely to deliver a higher absolute dose on the unit used for the prostate treatment. This reflects the facts that all centers are compared to the TROG QA Center, that the same calibration equipment and procedures are used in any individual center, and that in several centers the same treatment unit was used for both treatments (compare Table 1). Figure 8b shows the correlation between the ICRU reference points and the mean target dose in the two treatment scenarios. The fact that no clear correlation could be established for either of them indicates that both treatment scenarios can be regarded as independent. Therefore, it is unlikely that a common systematic error, e.g., in the treat-

ment planning system, affects the outcomes in participating centers.

DISCUSSION The level III dosimetric intercomparison The study achieved its main goal in verifying the dose delivery in participating centers in the two treatment scenarios given. This, of course, provides assurance that the doses achieved in the two trials are very close to those reported by the participating centers. As a level III dosimetric intercomparison it was focused on the global outcome. This has three important consequences: First, it does not allow conclusions to be drawn regarding individual steps in the treatment chain. If this is required, the level III study must be complemented with other checks or information provided by the participating centers. In the present study, the level I absolute calibration was aimed to provide some of the required information. In addition, the fact that two different and independent treatment scenarios were tested provided additional information to identify major problems if a discrepancy had been found. Second, the level III dosimetric intercomparison is relatively insensitive to small variations in individual parameters. As only the end result is verified, small deviations can be masked by other counteracting influences. However, as the global outcome is verified, a level III intercomparison

Dosimetric intercomparison using an anthropomorphic phantom

offers a reasonable reassurance that individual steps are within tolerable limits. Third, if highly complex treatment approaches such as IMRT (4) are to be verified, a level III dosimetric intercomparison should be complemented by additional measurements. As there are even more influences on the global outcome, the additional measurements, which will probably take the form of a level II intercomparison, can be designed to target particular factors that influence the overall outcome. A different consideration is that IMRT treatments vary considerably more from patient to patient than the conventional approaches tested in the present study. As such, a single phantom study will be insufficient to test all scenarios. A potential complement to a level III intercomparison could be in vivo dosimetric measurements on a selected number of patients (26, 27). Despite these potential shortcomings, only a level III intercomparison verifies every step in the treatment chain in a single measurement. This offers significant time savings if one can expect most participating centers to conform with the required specifications, as would be the case in a randomized controlled trial. It also ensures that no step (including the less obvious ones like data transcription from planning into a treatment record) is left out, a feature which becomes increasingly important when verifying more complex radiotherapy approaches. As a level III dosimetric intercomparison only checks the overall outcome, it can be used to compare centers using different treatment techniques for the same disease. An example for this is the difference in approach to improve a dose distribution in complex geometries. No lower level intercomparison can check the use of different energies, wedges (physical and dynamic), compensators, and dynamic or step and shoot IMRT in a single measurement. As such, a level III intercomparison is perfectly suited to audits where one expects good conformity and the end result is more important than the individual steps to achieve it. This would be the case in multicenter clinical trials. The study also succeeded in collected data on equipment and protocols in participating centers. This is complemented by discussions with staff and observations made by the visiting member of the study team. This renders a level III dosimetric intercomparison more than a quality assurance exercise. The data of many centers can be combined and potential influences of equipment or techniques on treatment outcomes can be assessed. This information is useful as feedback for all participating centers and the scientific community at large. Within its scope and uncertainty, the present study did not identify any influences of the parameters listed at the end of the “Results” section on the study outcome. This in itself is an interesting and reassuring result. Furthermore, the site visit during the intercomparison can be used as an opportunity for valuable information exchange. This would enhance the value of such a study to participating centers and may help with the introduction and adherence to protocols in new clinical trials.

● T. KRON et al.

577

Feasibility The present study demonstrates the feasibility of a multicenter dosimetric intercomparison using an anthropomorphic phantom. The costs per center were approximately $US 1000, including travel, accommodation, and TLD evaluation. While this seems to be not too excessive for an independent audit of the whole treatment chain, sending an unaccompanied phantom to participating centers (28) could reduce the costs. The advantage of the present study design, in which the phantom was accompanied, was that the site visit also allowed observing the treatment independently and establishing a database of equipment in the participating centers. In general, all participating centers performed well. This is reassuring for the centers involved as well as the clinical trials group and the general public. However, the participating centers constitute only a subset of radiotherapy units in Australia and New Zealand. These centers are also actively participating in multicenter clinical trials and are typically better-equipped and more focused on quality assurance than others. The present study was initiated by a clinical trials group (TROG) using resources and expertise from one participating clinical center. It would be no problem, however, to extend the level III intercomparison to centers that do not participate in trials for a more general audit of radiotherapy practice. It would make logistic and economic sense to utilize the same infrastructure for clinical trials QA and general radiotherapy audits. However, the latter would require an additional level of administration and governance. This could be in the form of an independent board including representatives from government, consumers, and the professional organizations involved. In addition, the center organizing the level III dosimetric intercomparison would require an external physical audit. Such a development is presently being actively considered in Australia. It must also be noted that the two treatment scenarios tested here are relatively simple. It is likely that intercomparisons looking at more complex treatment approaches may show larger deviations. This has been shown in mantle treatments (15) as well as at the junction of fields during head-and-neck radiotherapy (20). It will therefore be very interesting to use a similar approach for advanced treatment techniques that are now being introduced in Australia and New Zealand. The present study illustrates that clinical trials groups may have a role to play in the controlled introduction of new technology. Their QA network and intercomparison tools can be instrumental in ensuring that new technology is introduced in a comprehensive and safe manner. The planned follow-on trial to TROG 96.01, involving a planned dose escalation using conformal techniques, may provide a good example of this principle. Absolute dose comparison The level I dosimetric intercomparison performed as part of the present study is less accurate than similar services provided e.g., by the IAEA (10). This is a result of the local

578

I. J. Radiation Oncology

● Biology ● Physics

slab phantom used in each center instead of a uniform reference depth in water and the fact that Newcastle is not a standard laboratory. In future dosimetric intercomparisons the use of an ionization chamber at a standard depth of 5 cm for the absolute dose comparison under reference conditions should be considered for all centers. This would allow reducing the action level and increasing the level of confidence in the absolute calibration. Parameters influencing the outcome of the study The present study was not designed to specifically target variations of dose distribution with particular parameters. As such it is not surprising that only few obvious influences of physical parameters could be detected. The increased scatter to critical organs outside of the primary beam using 4 MV rather than 6 MV found in the head-and-neck treatment is such an example. However, when choosing the radiation energy for a particular treatment this is only one parameter to consider, whereas others such as dose distribution and dose build-up may be more important. Similarly, the use of laterals arcs in the prostate treatment to reduce rectal dose must be balanced against the increase in dose in other organs and the greater difficulty to verify dynamic treatments. Action levels The present choice of an action level of ⫾ 5% appears to be appropriate. It is compatible with general requirements of accuracy for radiotherapy delivery and reflects the uncertainties inherent in a complex level III dosimetric intercomparison. As shown in Fig. 6, the results are broadly normally distributed with a standard deviation of the order of 2%. To detect the 5% of centers that would be outside this distribution, the action level would need to be set around ⫾ 4%. This and the fact that none of the 18 participating centers was outside the action level of ⫾ 5% is compatible with the aim to detect all gross deviations. The action level for the absolute dosimetric intercomparison could probably be reduced if an ionization chamber measurement was chosen for this measurement or if the reference center was associated with a standard laboratory. The fact that the distribution of results shown in Fig. 6 is centered around 1 provides reassurance to the TROG QA center in Newcastle that its absolute dosimetry is adequate. Correlation between the two treatment scenarios The present study is the only level III intercomparison testing two treatment scenarios in one site visit. This approach was chosen as it assists with the identification of problems in the QA chain if the overall result is not satisfactory. It also constitutes an internal consistency check for the dose measurement as illustrated in Fig. 8. As the comparison of calibration includes a common component (the calibration at the reference center), there should be a correlation between the two treatment scenarios as can be seen in Fig. 8a. The lack of correlation for ICRU reference point measurements and the mean target dose, and the fact that the

Volume 52, Number 2, 2002

range of results is identical for both treatment scenarios, indicates that the results are due to random variations and the action level is set appropriately. Potential improvements The study identified some shortcomings that should be addressed in future dosimetric intercomparisons. The main point is the poor definition of target volume using isolated TLD detectors. Future phantom design should also incorporate a two-dimensional dosimetry system such as radiographic or radiochromic film. In addition, the ART phantom, which is divided into 2.5-cm-thick axial slices, proved to be difficult to reassemble without altering the geometry by up to 2 mm. A more rigid phantom such as the one described by Paliwal et al. (28) would be better suited for dosimetric intercomparisons. Another potential improvement is the number of repeat measurements. The two measurements chosen for the present study did not prove to be sufficient to identify repeatability of the results. This is acceptable if no discrepancy is found between the two measurements; however, it would fail if there is a difference. As such, future measurements should consist of at least three repeat measurements. The additional time required for changing of the detectors and additional treatment can be minimized by making the phantom modular and exchanging only preloaded modules instead of individual TLDs. A phantom to address these concerns is currently under construction in the TROG QA Center in Newcastle. It is planned to make level III dosimetric intercomparisons an integral part of multicenter clinical trials that involve advanced radiotherapy techniques. The intercomparison would again involve two treatment scenarios at one time and be complemented by other QA activities as appropriate for the respective trial. In addition, the level III intercomparison in the context of clinical trials may form the basis for a wider quality assurance exercise in radiotherapy in Australia and New Zealand.

CONCLUSION The results of the study allow the following conclusions: All participating centers completed the dosimetric intercomparison successfully. In particular, the study confirmed good absolute dose calibration in all participating centers and an overall good agreement between treatment planning and delivered dose. The level III dosimetric intercomparison complements other QA activities as it is the only means to verify the entire treatment chain. However, as this includes many different steps, it is difficult to draw exact conclusions about the influence of individual parameters on the treatment outcome. The level III dosimetric intercomparison presented here is not sensitive to small variations in any parameter as these could be counteracted by other influences. However, the

Dosimetric intercomparison using an anthropomorphic phantom

global outcome is verified, which offers reasonable reassurance that individual steps are within tolerable limits. A level III dosimetric intercomparison is feasible at a reasonable cost in an Australasian setting spanning some 5000 km. The results are of importance for the oncology community as well as the general public and serve as valuable quality assurance for multicenter clinical trials. In

● T. KRON et al.

579

addition, it may be an essential tool for the widespread introduction of modern complex treatment techniques, such as IMRT, in a controlled manner. The study identified a number of potential improvements that will be incorporated in future dosimetric intercomparisons. Of particular note is the inclusion of a dosimeter with better spatial resolution and an improved phantom design.

REFERENCES 1. Bentzen SM, Overgaard J. Patient-to-patient variability in the expression of radiation-induced normal tissue injury. Semin Radiat Oncol 1994;4:68 – 80. 2. Brahme A. Dosimetric precision requirements in radiation therapy. Acta Radiol Oncol 1984;23:379 –391. 3. Dutreix A. When and how can we improve precision in radiotherapy? Radiother Oncol 1984;2:275–292. 4. Burman C, Chui CS, Kutcher G, et al. Planning, delivery, and quality assurance of intensity-modulated radiotherapy using dynamic multileaf collimator: A strategy for large-scale implementation for the treatment of carcinoma of the prostate. Int J Radiat Oncol Biol Phys 1997;39:863– 873. 5. Hanks GE, Schultheiss TE, Hanlon AL, et al. Optimisation of conformal radiation treatment of prostate cancer: Report of a dose escalation study. Int J Radiat Oncol Biol Phys 1997;37: 543–550. 6. Kutcher GJ, Coia L, Gillin M, Hanson W, Leibel S, Morton RJ, et al. Comprehensive QA for radiation oncology: Report of AAPM therapy committee task group 40. Med Phys 1994; 21:581– 618. 7. Millar M, Cramb J, Das R, Ackerly T, Brown G, Webb D. ACPSEM Position Paper. Recommendations for the safe use of external beams and sealed brachytherapy sources in radiation oncology. Aust Phys Eng Sci Med 1997;20(Suppl.):1–35. 8. World Health Organisation. Quality assurance in radiotherapy. Geneva: WHO; 1988. 9. International Atomic Energy Agency. Lessons learned from accidental exposures in radiotherapy. Safety Report Series; N17. Vienna: IAEA; 2000. 10. Izewska J, Andreo P. The IAEA/WHO postal programme for radiotherapy hospitals. Radiother Oncol 2000;54:65–72. 11. Pfalzner PM, Alvarez S. Inter-comparison of absorbed dose in Cobalt 60 teletherapy using mailed LiF dosimeters. Acta Radiol Ther Phys Oncol 1968;7:379 –387. 12. Nisbet A, Thwaites DI, Sheridan ME. A dosimetric intercomparison of kilovoltage x-rays, megavoltage photons and electrons in the Republic of Ireland. Radiother Oncol 1998; 47:95–102. 13. Bridier A. A comparative description of three multipurpose phantoms (MPP) for external audits of photon beams in radiotherapy: The water MPP, the Umeå MPP and the EC MPP. Radiother Oncol 2000;55:285–293. 14. Delaney G, Beckham W, Veness M, Ahern V, Back M, Boyages J, et al. Three-dimensional dose distribution of tangential breast irradiation: Results of a multicentre phantom dosimetry study. Radiother Oncol 2000;57:61– 68. 15. Amies C, Rose A, Metcalfe P, Barton M. Multicentre dosimetry study of mantle treatment in Australia and New Zealand. Radiother Oncol 1996;40:171–180.

16. Danoff BF, Goodman RL, Glick JH, Haller DG, Pajak TF. The effect of adjuvant chemotherapy on cosmesis and complications in patients with breast cancer treated by definitive irradiation. Int J Radiat Oncol Biol Phys 1983;9:1625–1630. 17. Davis JB, Miltchev V. Tangential breast irradiation: A multicentric inter-comparison of dose using a mailed phantom and thermoluminescent dosimetry. Radiother Oncol 1999;52:65– 68. 18. Huntley R, Izewska J. The 1998 Australian external beam radiotherapy survey and IAEA/WHO TLD postal dose quality audit. Aust Phys Eng Sci Med 2000;23:21–29. 19. Johansson K, Horiot J-C, Van der Scheuren E. Quality assurance control in the EORTC cooperative group of radiotherapy: 3 - Inter-comparison in an anthropomorphic phantom. Radiother Oncol 1987;9:289 –298. 20. Kron T, Barnes K, Bazley S, O’Brien P. TLD for multicentre dose inter-comparison in complicated radiotherapy treatments: The field junction in head-and-neck radiotherapy. Phys Med Dosimetry 1999;85:357–360. 21. Poulsen M, Denham JW, Peters L, Lamb D, Spry N, et al. A randomised trial of accelerated and conventional radiotherapy for stage III and IV squamous carcinoma of the head-andneck: A Trans-Tasman Radiation Oncology Group Study. Radiother Oncol 2001;60:113–122. 22. Steigler A, Mameghan H, Lamb D, Joseph D, Matthews J, Franklin I, et al. A quality assurance audit: Phase III trial of maximal androgen deprivation in prostate cancer (TROG 96.01). Australas Radiol 2000;44:65–71. 23. International Commission on Radiation Units, and Measurements. Prescribing, recording, and reporting photon beam therapy. ICRU Report 50. Bethesda, MD: ICRU; 1993. 24. Kron T. Invited review: Thermoluminescence dosimetry and its applications in medicine - Part 1: Physics, materials and equipment. Australas Phys Eng Sci Med 1994;17:175–199. 25. International Atomic Energy Agency. Absorbed dose determination in photon and electron beams. Technical Report Series; N277. Vienna: IAEA; 1987. 26. Kron T, Schneider M, Murray A, Mameghan H. Clinical thermoluminescence dosimetry: How do expectations and results compare? Radiother Oncol 1993;26:151–161. 27. Noel A, Aletti P, Bey P, Malissard L. Detection of errors in individual patients in radiotherapy by systematic in vivo dosimetry. Radiother Oncol 1995;34:144 –151. 28. Paliwal BR, Ritter MA, McNutt T, Mackie TR, Thomadsen B, Purdy JA, Kinsella, TJ. A solid water pelvic and prostate phantom for imaging, volume rendering, treatment planning and dosimetry for an RTOG multi-institutional, 3-D dose escalation study. Int J Radiat Oncol Biol Phys 1998;42:205– 211.