Impact of IMRT and leaf width on stereotactic body radiotherapy of liver and lung lesions

Int. J. Radiation Oncology Biol. Phys., Vol. 61, No. 5, pp. 1572–1581, 2005 Copyright © 2005 Elsevier Inc. Printed in the USA. All rights reserved 0360-3016/05/$–see front matter

doi:10.1016/j.ijrobp.2004.12.075

PHYSICS CONTRIBUTION

IMPACT OF IMRT AND LEAF WIDTH ON STEREOTACTIC BODY RADIOTHERAPY OF LIVER AND LUNG LESIONS PAVEL DVORAK, M.SC., DIETMAR GEORG, PH.D., JOACHIM BOGNER, PH.D., BERNHARD KROUPA, M.SC., KARIN DIECKMANN, M.D., AND RICHARD PÖTTER, M.D. Department of Radiotherapy and Radiobiology, AKH Vienna, Medical University of Vienna, Vienna, Austria Purpose: The present study explored the impact of intensity-modulated radiotherapy (IMRT) on stereotactic body RT (SBRT) of liver and lung lesions. Additionally, because target dose conformity can be affected by the leaf width of a multileaf collimator (MLC), especially for small targets and stereotactic applications, the use of a micro-MLC on “uniform intensity” conformal and intensity-modulated SBRT was evaluated. Methods and Materials: The present study included 10 patients treated previously with SBRT in our institution (seven lung and three liver lesions). All patients were treated with 3 ⴛ 12 Gy prescribed to the 65% isodose level. The actual MLC-based conformal treatment plan served as the standard for additional comparison. In total, seven alternative treatment plans were made for each patient: a standard (actual) plan and an IMRT plan, both calculated with Helax TMS (Nucletron) using a pencil beam model; and a recalculated standard and a recalculated IMRT plan on Helax TMS using a point dose kernel approach. These four treatment plans were based on a standard MLC with 1-cm leaf width. Additionally, the following micro-MLC (central leaf width 3 mm)– based treatment plans were calculated with the BrainSCAN (BrainLAB) system: standard, IMRT, and dynamic arc treatments. For each treatment plan, various target parameters (conformity, coverage, mean, maximal, and minimal target dose, equivalent uniform doses, and dose–volume histogram), as well as organs at risk parameters (3 Gy and 6 Gy volume, mean dose, dose–volume histogram) were evaluated. Finally, treatment efficiency was estimated from monitor units and the number of segments for IMRT solutions. Results: For both treatment planning systems, no significant difference could be observed in terms of target conformity between the standard and IMRT dose distributions. All dose distributions obtained with the micro-MLC showed significantly better conformity values compared with the standard and IMRT plans using a regular MLC. Dynamic arc plans were characterized by the steepest dose gradient and thus the smallest V6 Gy values, which were on average 7% smaller than the standard plans and 20% lower than the IMRT plans. Although the Helax TMS IMRT plans show about 18% more monitor units than the standard plan, BrainSCAN IMRT plans require approximately twice the number of monitor units relative to the standard plan. All treatment plans optimized with a pencil beam model but recalculated with a superposition method showed significant qualitative, as well as quantitative, differences, especially with respect to conformity and the dose to organs at risk. Conclusion: Standard conformal treatment techniques for SBRT could not be improved with inversely planned IMRT approaches. Dose calculation algorithms applied in optimization modules for IMRT applications in the thoracic region need to be based on the most accurate dose calculation algorithms, especially when using higher energy photon beams. © 2005 Elsevier Inc. Stereotactic body radiotherapy, Intensity-modulated radiotherapy, Inverse planning, Liver, Lung.

INTRODUCTION Stereotactic radiotherapy (RT) and radiosurgery of cranial targets have become well-established treatment techniques since the early 1990s (1). Improvements in patient positioning and the development of dedicated immobilization devices have allowed the extension of stereotactic applications to targets outside the cranial region (i.e., regions such as the head and neck, pelvis, and thorax). As with single-fraction

radiosurgery and hypofractionated stereotactic RT to cranial tumors, stereotactic body RT (SBRT) can be characterized by hypofractionation, multimodality imaging technologies for treatment preparation and dose delivery, and the use of immobilization and positioning devices correlated with a stereotactic coordinate system. In general, dose prescription is performed to a certain isodose level, and a nonuniform target dose is accepted.

Reprint requests to: Dietmar Georg, Ph.D., Division of Medical Radiation and Physics, Department of Radiotherapy and Radiobiology, Medical University of Vienna, Waehringer Guertel 18-20, Vienna A-1090 Austria. Tel: (⫹43) 1-40-400-2695; Fax: (⫹43) 1-40-400-2696; E-mail: [email protected]

P. Dvorak was supported by the EC project HCMPT-2001-0318. Acknowledgments—The authors thank BrainLAB for providing various software modules for data and image transfer. Received Jul 13, 2004, and in revised form Dec 8, 2004. Accepted for publication Dec 16, 2004. 1572

Impact of IMRT on SBRT

Several authors have reported on the good clinical results of SBRT for primary and/or metastatic tumors (2– 6). However, large variations exist among the reported SBRT treatment protocols, with differences in total target doses, number of fractions used, and normalization or prescription isodose levels. Most treatment protocols have consisted of one to four fractions with the doses per fraction varying between 10 and 26 Gy, and mostly prescribed to the 65% or 80% isodose level (2, 3, 7, 8). Various SBRT techniques have been described for tumors in the head and neck, lung, liver, abdomen, and pelvis. However, both primary and metastatic tumors in the liver and/or lung are currently the primary sites for SBRT (2, 3, 7–13). In contrast to cranial stereotactic RT, SBRT is mostly associated with target movement for various physiologic reasons (e.g., respiration, heartbeat, differences in organ filling). One consequence of the presence of organs at risk (OARs) in close proximity to tumors and the high doses delivered is the need for highly conformal dose distributions when applying SBRT. Although small liver metastases are usually of a rather regular shape, lung lesions can be complex. Organ sparing by applying steep dose gradients and highly conformal dose distributions are typical features of intensity-modulated RT (IMRT) based on computerized treatment plan optimization. Moreover, inverse planning can replace traditional “trial and error” forward optimization methods and has the potential to reduce the treatment planning workload for complex cases. The present study explored the impact of IMRT on SBRT for liver and lung tumors. Some authors have already reported on IMRT applications for lung treatment (14 –16). However, all these studies were based on a nonstereotactic “conformal approach” in which they sought to encompass the planning target volume (PTV) with the 95% isodose surface. IMRT applications in regions in which large inhomogeneities are present can be influenced by the dose calculation algorithm used during optimization (17). This aspect was also addressed in the present study, although the experimental results of a related dosimetric study will be presented in a separate communication. Furthermore, because target dose conformity can be affected by the leaf width of a multileaf collimator (MLC), especially for small lesions and stereotactic applications (18 –20), the impact of leaf width on a “uniform intensity” conformal and intensity-modulated approach for SBRT was investigated. METHODS AND MATERIALS Patients and treatment protocol The present study was based on the anatomic information from 10 patients treated previously with SBRT in our institution. Seven patients had either primary or metastatic lung lesions and three had liver metastases. The targets were classified as small if the PTV was ⬍50 cm3 or large if the PTV was ⬎100 cm3. Using these criteria, three targets were scored as small and seven as large. All parameters described below (see “Evaluation parameters”) were also evaluated with respect to this classification (i.e., lung vs. liver and small vs. large).

● P. DVORAK et al.

1573

Patient immobilization was performed using a stereotactic body frame (Elekta, Crawley, UK). For treatment planning, all patients underwent CT; if needed, MRI was used for tumor localization. A so-called diaphragm-control device, connected to the body frame, was used to reduce breathing effects as much as possible. After clinical target volume (CTV) definition, a 7-mm margin was added in the lateral and AP directions and a 10-mm margin in the craniocaudal direction to determine the PTV. All patients were treated according to the same SBRT treatment protocol, in which 3 ⫻ 12 Gy prescribed to the 65% isodose level was delivered within 6 – 8 days. SBRT treatment planning was performed with Helax TMS, version 6.1 or 6.1A (Nucletron, Veenendaal, The Netherlands). Typically, five to seven noncoplanar and individually weighted irregular MLC shaped beams were used on the basis of a beam’s-eye-view planning technique. Treatments were performed on a linear accelerator providing 6-, 10-, and 25-MV photon beams (Sli Precise, Elekta Oncology Systems, Crawley, UK). The integrated standard MLC has a projected leaf width at the isocenter of 1 cm. In general, two of the three available photon beam energies were combined to achieve the optimal dose distribution. Before each fraction, a verification CT was done, in which patients were repositioned in the stereotactic body frame using skin tattoos and laser markers. Verification CT images and planning CT images were registered and corrections performed if necessary. Finally, patients in the body frame were transferred from the CT to the linear accelerator at which a final electronic portal imaging device (EPID)– based position verification was done. More details on the treatment procedure using a stereotactic body frame can be found in the literature (5, 11, 21).

Treatment planning systems For the treatment planning comparison, the following treatment planning systems (TPS) were used: Helax TMS, version 6.1(A) (Nucletron) and BrainSCAN, version 5.2 (BrainLAB, Munich, Germany). All treatment techniques available with this stereotactic TPS (i.e., conformal RT, IMRT, and dynamic arc therapy) were restricted to a 6-MV photon beam but were based on the use of an external micro-MLC (m3, BrainLAB) with a variable leaf width for field shaping (22). This beam energy and the use of a microMLC for IMRT, dynamic arc, and conformal RT mimic the situation of the dedicated stereotactic Novalis treatment unit, which is also designed for SBRT applications. The dose calculation algorithm in the TPS is based on a pencil beam model and an equivalent path length inhomogeneity correction. The calculation grid for BrainSCAN was 2 mm. As mentioned earlier, Helax TMS can be used only in conjunction with a standard MLC, either for conformal RT or IMRT, and offers multiple photon beam energies (6, 10, and 25 MV). This TPS allows the selection of either a pencil beam model or a so-called collapsed cone algorithm, which is basically a point kernel dose calculation algorithm (23), for dose calculation, although during optimization for IMRT, a pencil beam model is applied. Therefore, each standard plan, as well as each IMRT treatment plan made on Helax TMS, was recalculated with the collapsed cone algorithm. For all recalculated plans, the same normalization was performed as for the pencil beam plans. The calculation grid for HELAX TMS was 3 mm, a compromise between calculation speed and accuracy. To ensure maximal compatibility of anatomic information on both TPSs, CT images, as well as volumes-of-interest, were transferred from Helax TMS to BrainSCAN via a DICOM transfer tool.

1574

I. J. Radiation Oncology

● Biology ● Physics

Treatment planning categories To compare various treatment options all 10 patients underwent repeated treatment planning, as described below, but the treatment plan for the actual treatment (made on Helax TMS) was used as the standard for further comparison. Then, by using the same beam incidences as in the “standard” conformal plan, another conformal plan was made on the BrainSCAN system. The overall treatment planning strategy was to achieve maximal target coverage and best dose distribution conformity while keeping at the OAR doses below the general toxicity criteria and/or equal to the corresponding parameters of the original reference treatment plan. Therefore, if the resulting dose distribution with the original beam setup was not optimal for the BrainSCAN plans (based on 6 MV plus a micro-MLC), the beam incidences were modified and/or beams were added. Special attention was paid to keep the beam geometries as similar as possible for all treatment planning categories. For 6 patients, identical beam geometries were used for all treatment plan categories, with the exception of dynamic arc therapy. For 3 patients, beam geometries differed in 1 of 7 beams used in total in one treatment plan category only, and for 1 patient, the number of beams was different (7 for all IMRT plans, 8 for Helax, and 10 for BrainSCAN for standard conformal plans). It has been our policy to constrain the normal tissue by limiting the volume of liver receiving ⬎6 Gy per fraction to less than one-third of the liver. For lung lesions, dose–volume constraints for SBRT were derived from dose–volume histogram (DVH) parameters used in the department (20 Gy to 50% of the homolateral side of the lung, 20 Gy to 30% of the contralateral side of the lung, and 20 Gy to 40% of the total lung) for 2 Gy/fraction by applying the linear-quadratic model (␣/␤ ⫽ 3). To evaluate the potential benefit of IMRT on SBRT and to compare IMRT solutions of both TPSs for SBRT, IMRT plans were made for each patient. On either TPS, the beam geometry of the respective standard plan was kept for the IMRT technique. Although the inverse planning modules and/or algorithms of the TPS have been described, the most important differences are summarized briefly. The IMRT option on the Helax TMS is based on a segmental MLC delivery and allows one to limit the maximal number of segments per field, to quantify a minimal number of leaf pairs opened, and to define the minimal segment area (24). BrainSCAN offers segmental MLC and dynamic MLC delivery as

Volume 61, Number 5, 2005

IMRT options. However, to allow for a maximal comparability between both TPSs, the segmental MLC delivery technique was chosen. The mean number of segments per beam (the only “segment” parameter to be defined in BrainSCAN system) was set to be equal to the maximal number of segments in the IMRT option of Helax TMS. The spatial resolution of the IMRT fluence map for Helax TMS was 1 cm in direction perpendicular to leaf movement and 0.5 cm in the direction of leaf movement. The respective value for BrainSCAN was 4 mm in the direction of leaf movement; in the perpendicular direction, the resolution was defined by the variable leaf thickness. The BrainSCAN IMRT solution automatically calculates four treatment plans giving zero (target only), low, medium, and high priority to the predefined OARs. For most SBRT patients, the treatment plan with medium or low priority to OARs best fulfilled the clinical demands of the present study. Inverse treatment planning, as well as the segmentation-related parameters used in both TPSs, are summarized in Table 1. For IMRT planning, the same strategy as described above was applied. It was aimed at achieving maximal target coverage and dose distribution conformity while minimizing the doses to OARs. Therefore, basically, PTV dose–volume (DVH) constraints and OARs dose constraints were used. Additionally, a CTV dose– volume constraint needed to be introduced to obtain approximately the same PTV DVH curve for IMRT plans as in the standard plans, which was much more significant in the case of the Helax TMS system. However, DVH values of OARs derived from original standard treatment plans were always used in addition as a standard for the IMRT planning process. In addition to uniform intensity conformal therapy and IMRT, the BrainSCAN TPS offers another treatment option, the so-called dynamic arc technique, in which leaf shapes are dynamically adapted during arc therapy according to the beam’s-eye-view projection. The highest priorities during treatment planning were again given to target coverage and dose distribution conformity while simultaneously trying to minimize the doses to the OARs. Typical dynamic arc plans consisted of three to five arcs with manually optimized field shapes. In summary, the following seven alternative treatment plans were available for each patient: (1) standard and (2) IMRT plans calculated on Helax TMS with a pencil beam model and recalculated (3) standard and (4) IMRT

Table 1. Specifications of two different inverse treatment planning systems for IMRT approach Inverse planning parameters IMRT delivery Maximal/mean segment No./field Optimization algorithm Dose calculation algorithm Spatial resolution intensity map (mm) Direction of leaf movement Perpendicular to leaf direction Inhomogeneity correction Typical %D/%V constraints* PTV CTV OAR

Helax TMS

BrainSCAN

Segmental MLC delivery 15

Segmental MLC delivery 15

Quadratic difference Pencil beam

Dynamic penalized likelihood Pencil beam

5 10 EPL

4 3–5.5 EPL

(55–59)/0.98, 90/0.5, 105/0.02 90/(0.5–0.8), 105/0.02 16/x, 32/x, 50/x

71/1 90/(0.7–0.9) 16/x, 32/x, 50/x

Abbreviations: IMRT ⫽ intensity-modulated radiotherapy; MLC ⫽ multileaf collimator; EPL ⫽ equivalent path length; %D ⫽ percentage of dose; %V ⫽ percentage of volume; PTV ⫽ planning target volume; CTV ⫽ clinical target volume; OAR ⫽ organ at risk; x ⫽ value taken from standard Helax TMS plan. * Summary of DVH constraints applied in the respective treatment planning system.

Impact of IMRT on SBRT

plans applying a collapsed cone algorithm; and (5) standard, (6) IMRT, and (7) dynamic arc plans calculated with BrainSCAN. All BrainSCAN plans, as well as all standard Helax TMS plans, were normalized to 100% at the isocenter, corresponding to a single dose of 18.5 Gy. However, in IMRT, the common method to normalize the dose distributions to a predefined normalization point (e.g., the isocenter) is no longer recommended for Helax TMS. The prescribed dose to the target is usually given by a certain dose interval. Therefore, an arbitrary point in the target will result in a dose value inside the required interval (if the optimization can fulfill the constraints). As a consequence, demanding a fixed dose to the normalization point (e.g., 100%) may necessitate rescaling the whole optimized plan. To avoid this, it is preferable to use a volume-based normalization method. The Helax TMS system allows the user to normalize the dose distributions to the minimal, maximal, mean, or median doses of a predefined volume of interest. Because it is commonly used in clinical practice for IMRT planning, the normalization to the mean CTV dose was chosen, in which 100% again corresponded to a single dose of 18.5 Gy. For recalculated IMRT plans on Helax TMS (applying the collapsed cone algorithm), this implies that the segment shapes and relative segment weights originally derived from an optimization process, which was based on a pencil beam model, are used for dose calculation but their absolute weight (number of monitor units [MUs]) is adapted to deliver the prescribed dose to the CTV.

Evaluation parameters Each treatment plan was evaluated with respect to target criteria, OAR criteria, and treatment efficiency. Target coverage was defined as the percentage fraction of the PTV covered by the prescription isodose volume. The conformity parameter was defined as the ratio of the total volume encompassed by the 65% isodose (V65%) and the PTV fraction covered by the 65% isodose (V65%-PTV). On the basis of this definition and assuming complete target coverage, the ideal value of the conformity parameter is 1.0. Additionally, the mean CTV and mean PTV doses, as well as the minimal (Dmin) and maximal doses to the PTV, were recorded. Finally, as an alternative to the mean dose, equivalent uniform doses were calculated for the CTV (EUDCTV) and PTV from DVH information (25, 26). Depending on the target site, either the liver or lung was considered as the primary OAR. In the case of lung lesions, the left and right lung were treated as one paired organ. Although various biologic models are available to quantify toxicity for these organs for hypofractionated treatments, their application is not straightforward. Hence, the mean liver and lung doses were chosen as an OAR parameter, because they can be easily calculated and these parameters have been demonstrated to be reliable estimators for organ toxicity (27, 28). Additionally, the total volume encompassed by the 3- and 6-Gy isodoses (V6 Gy and V3 Gy, respectively) were derived for each treatment plan. To quantify treatment efficiency, the total number of MUs necessary to deliver the prescribed dose and the total number of segments in the case of IMRT treatment plans were evaluated.

RESULTS Target coverage and target conformity Table 2 summarizes the mean values and standard deviations, as well as the observed range for the coverage and conformity parameters, for all treatment plan categories.

● P. DVORAK et al.

1575

Target coverage was given the highest priority during treatment planning for all techniques (i.e., a treatment plan was accepted only if the target coverage was better than the minimally acceptable value of 97%). No significant difference could be observed between the standard and IMRT dose distributions for both TPSs. For lung lesion patients, target coverage of the standard Helax TMS plans recalculated with the collapsed cone algorithm was, on average, 4% lower than the original plans calculated with the pencil beam algorithm. For liver patients, the target coverage decreased by only 2%. The dose distributions of the IMRT plans using the pencil beam model but recalculated with the collapsed cone algorithm showed a “blurred” dose distribution, with larger isodose volumes compared with the original IMRT plan. Consequently, the recalculated Helax TMS IMRT (collapsed cone) plans showed comparable or even slightly greater target coverage values. All dose distributions obtained with the micro-MLC on the BrainSCAN system, including dynamic arc therapy, showed significantly smaller values of the conformity parameter than the standard and IMRT Helax TMS plans using a regular MLC. For the collapsed cone recalculated standard plans (on the Helax TMS system), the conformity parameter decreased by approximately 7%. The original conformity of the IMRT plans using the pencil beam algorithm was completely lost when recalculating these plans with the collapsed cone algorithm. The conformity parameter increased by about 40% and even more in case of small targets. Dmin and maximal PTV dose Table 3 summarizes the minimal and maximal PTV doses, and Fig. 1 shows the average cumulative DVHs for the PTV of the 10 patients considered in this study. In general, the dose distributions for the standard Helax TMS plans showed, on average, 10% larger Dmin values compared with the standard BrainSCAN plans. This situation was reversed for IMRT plans, in which the Dmin values for the Helax TMS IMRT plans were about 15% smaller than those for BrainSCAN, especially for large PTVs. When comparing the Dmin of the original pencil beam– based Helax TMS plans and recalculated plans, the observed tendency correlated with the results obtained for the conformity parameter. For standard plans, the collapsed cone algorithm decreased the Dmin by about 10%, but it increased the Dmin values by about 9% and more for the IMRT plans. This effect correlated with the reduced target conformity (increased conformity parameter) for recalculated Helax TMS IMRT plans that resulted in enlarged volumes encompassed by a certain isodose level. Dynamic arc plans showed on average about 5% greater Dmin values compared with standard conformal plans on BrainSCAN, and IMRT plans presented with about 15% greater values. Equivalent uniform dose The average values of EUDCTV and EUDPTV for all patients and respective treatment plan category are indicated

1576

I. J. Radiation Oncology

● Biology ● Physics

Volume 61, Number 5, 2005

Table 2. Mean target coverage and conformity parameter for different plan categories Coverage Treatment plan category Standard Helax TMS IMRT Helax TMS Standard BrainSCAN IMRT BrainSCAN Dynamic arc Standard Helax, collapsed cone IMRT Helax, collapsed cone

Conformity

Mean ⫾ SD

Range

Mean ⫾ SD

Range

99.0 ⫾ 0.8 98.4 ⫾ 0.8 98.8 ⫾ 0.4 99.2 ⫾ 1.0 99.0 ⫾ 0.5 95.5 ⫾ 3.1 98.7 ⫾ 1.2

98.0–100.0 97.0–99.0 98.0–99.0 97.0–100.0 98.0–100.0 89.0–99.0 97.0–100.0

1.41 ⫾ 0.11 1.40 ⫾ 0.16 1.23 ⫾ 0.05 1.26 ⫾ 0.06 1.26 ⫾ 0.05 1.31 ⫾ 0.14 2.02 ⫾ 0.50

1.18–1.61 1.21–1.66 1.15–1.30 1.19–1.35 1.20–1.34 1.14–1.63 1.29–2.86

Abbreviation: IMRT ⫽ intensity-modulated radiotherapy.

in Table 4. Regarding EUDCTV, lower values for both the BrainSCAN standard and dynamic arc plans were obtained, consistent with the results of the mean CTV dose. Two plan categories, the standard Helax TMS plans and recalculated IMRT collapsed cone plans, showed slightly greater average EUD values for PTV than the other categories. Mean OAR dose The average values for the mean OAR dose for all patients and the respective treatment plan category are indicated in Table 5. The only treatment plans differing significantly from the others were the recalculated IMRT (collapsed cone) plans, in which the mean values to the OARs were, on average, 30% greater. As an example, Fig. 2 shows a cumulative DVH for the lung (whole organ) for a patient with lung metastasis. V3 Gy and V6 Gy No significant difference was found for the V6 Gy when comparing the standard and IMRT Helax TMS plans. Standard plans calculated with the BrainSCAN system showed slightly smaller V6 Gy values than standard Helax TMS plans, but the V6 Gy values of the IMRT BrainSCAN plans were 10 –20% greater than those for the IMRT Helax TMS plans, especially for large targets. Dynamic arc plans were characterized by the steepest dose gradient and thus the smallest V6 Gy values, which were, on average, 7% smaller than for the BrainSCAN

standard plans and 20% lower than for the BrainSCAN IMRT plans. For large targets, no significant difference was found in the V6 Gy between the original standard Helax TMS plans and those recalculated with the collapsed cone algorithm. However, for small targets, the original standard Helax TMS plans presented, on average, with 18% greater V6 Gy values compared with the recalculated plans. The recalculated Helax TMS IMRT plans, in contrast, showed a very different dose distribution compared with the one optimized with a pencil beam algorithm, with maximal differences of ⬎70% for individual patients. In general, the differences were larger for lung lesion patients than for patients with liver lesions. Table 6 summarizes the results related to the V6 Gy. For treatment plans on the Helax TMS system, the volume covered by the V3 Gy was, on average, 10% greater for the IMRT dose distributions than for the standard plans, especially for large targets. This difference between IMRT and standard plans was only 4% on the BrainSCAN system. Of all the micro-MLC– based treatment techniques, the dynamic arc plans showed the lowest V3 Gy values. These differences reached 20% and more for small targets. Although only small differences in V3 Gy were observed between standard conformal plans calculated with the pencil beam algorithm and recalculated ones applying the collapsed cone algorithm, the recalculated IMRT plans differed significantly. Again, the recalculated IMRT plans showed a very different dose distribution compared with the one opti-

Table 3. Minimal and maximal doses of normalization value (18.5 Gy) to PTV as function of treatment plan category Dmin (%) Treatment plan category Standard Helax TMS IMRT Helax TMS Standard BrainSCAN IMRT BrainSCAN Dynamic arc Standard Helax, collapsed cone IMRT Helax, collapsed cone

Dmax (%)

Mean ⫾ SD

Range

Mean ⫾ SD

Range

59.5 ⫾ 7.9 53.3 ⫾ 7.0 51.6 ⫾ 6.1 60.1 ⫾ 4.1 54.3 ⫾ 4.3 52.5 ⫾ 5.0 58.6 ⫾ 8.3

48.5–73.8 42.5–62.7 40.0–60.0 53.0–65.0 46.0–60.0 46.5–62.4 42.4–67.3

104.1 ⫾ 2.6 104.6 ⫾ 3.7 106.7 ⫾ 4.4 107.3 ⫾ 2.9 103.1 ⫾ 4.3 102.7 ⫾ 2.4 106.6 ⫾ 5.3

100.9–108.6 95.3–108.3 103.0–116.0 101.0–111.0 96.0–112.0 99.2–107.3 95.1–114.0

Abbreviations: Dmin ⫽ minimal dose; Dmax ⫽ maximal dose; other abbreviations as in Table 1.

Impact of IMRT on SBRT

Fig. 1. Averaged (10 patients) cumulative dose volume histograms for planning target volume for different treatment plan categories.

mized with a pencil beam algorithm and resulted in greater V3 Gy values. Table 6 also summarizes the results related to the V3 Gy. Figure 3 illustrates the isodose distributions in an axial plane for the various treatment plan categories. Note, two lung lesions were treated with a single isocenter. Treatment efficiency The average values of the total number of MUs necessary to deliver the prescribed dose per fraction for all patients and respective treatment plan category are presented in Table 7. No significant MU variation among the standard plans of both TPSs and the dynamic arc plans could be observed. However, a large discrepancy was noted between the IMRT solutions. Although the Helax TMS IMRT plans showed about 18% more MUs than the standard plan, the BrainSCAN IMRT plans required approximately twice the number of MUs relative to the standard plan. The total number of segments for all beams in the “step-and-shoot” IMRT delivery was significantly larger for IMRT BrainSCAN (mean 92, range 82–100) than for the IMRT solution offered by the Helax TMS system (mean 36, range 8 – 62). DISCUSSION Organ sparing by applying steep dose gradients and highly conformal dose distributions are typical features of inverse planning and IMRT delivery techniques. Moreover,

● P. DVORAK et al.

1577

computerized treatment plan optimization has the potential to replace tedious treatment planning based on human intelligence by a more efficient process (29). These arguments have been verified for many tumor entities (e.g., prostate, head and neck, and tumor recurrence). In some centers, IMRT has become the standard treatment option for such cases. The present study explored the impact of IMRT on a SBRT technique using a hypofractionated dose prescription for relatively small targets surrounded by a much larger OAR region. Concerning dose distribution conformity and target coverage, comparable results were obtained for traditionally forward planned conformal uniform intensity techniques and inversely planned IMRT techniques. A variety of target dose parameters was considered (mean CTV and PTV doses, EUD for CTV and PTV, Dmin, and maximal dose). With the exception of lower mean PTV dose values for IMRT BrainSCAN plans relative to the other treatment plans, which was most probably related to the BrainSCAN IMRT solution (see arguments below), comparable target dose characteristics were observed for IMRT and standard conformal plans. The same held true for normal tissue sparing. The differences between the mean OAR doses of a standard and an IMRT treatment plan, made on the same TPS, were only small and most probably not of any clinical importance (i.e., inverse planning and computerized treatment plan optimization did not improve the conformal dose distributions). Volumes covered by low and medium isodose levels, V3 Gy and V6 Gy, were even larger for the IMRT plans, especially for the BrainSCAN IMRT solution. Such large fractional doses can cause normal tissue complications (e.g., fibrosis). Hence, the known IMRT feature of larger lower isodose volumes is considered a significant disadvantage for a hypofractionated dose prescription as applied in SBRT. These lower isodose levels in IMRT are significantly influenced by the IMRT solution, including leaf sequencing, and to some extent by photon beam energy. For most patients, the V6 Gy values of the BrainSCAN IMRT plans, with the large number of segments and the high number of MUs, were significantly larger compared with the standard plans using the same nominal beam energies. Because the standard BrainSCAN plans (6 MV only) showed slightly lower V6 Gy

Table 4. Mean equivalent uniform dose for CTV and PTV for different treatment plan categories EUDCTV (Gy) Treatment plan category Standard Helax TMS IMRT Helax TMS Standard BrainSCAN IMRT BrainSCAN Dynamic arc Standard Helax, collapsed cone IMRT Helax, collapsed cone

EUDPTV (Gy)

Mean ⫾ SD

Range

Mean ⫾ SD

Range

52.1 ⫾ 2.1 52.3 ⫾ 2.1 50.6 ⫾ 2.8 52.8 ⫾ 0.9 47.4 ⫾ 3.5 49.2 ⫾ 2.9 51.7 ⫾ 2.4

48.9–55.1 48.5–55.0 46.8–55.0 51.3–54.0 43.3–54.1 44.2–52.2 46.8–54.1

43.1 ⫾ 2.0 41.0 ⫾ 2.0 41.1 ⫾ 2.0 41.6 ⫾ 1.5 41.2 ⫾ 1.2 39.7 ⫾ 2.0 42.9 ⫾ 2.5

39.5–46.8 38.1–44.8 38.3–44.7 39.4–44.0 39.3–42.8 36.9–43.1 37.4–46.1

Abbreviations: EUD ⫽ equivalent uniform dose; other abbreviations as in Table 1.

1578

I. J. Radiation Oncology

● Biology ● Physics

Volume 61, Number 5, 2005

Table 5. Mean organ at risk dose of normalization values (18.5 Gy) for different treatment plan categories Mean OAR dose Treatment plan category

Mean ⫾ SD

Range

Standard Helax TMS IMRT Helax TMS Standard BrainSCAN IMRT BrainSCAN Dynamic arc Standard Helax, collapsed cone IMRT Helax, collapsed cone

12.6 ⫾ 5.8 12.5 ⫾ 5.9 12.9 ⫾ 6.0 13.0 ⫾ 5.8 12.5 ⫾ 5.8 12.7 ⫾ 5.5 16.4 ⫾ 7.2

5.1–21.9 5.0–21.4 5.1–21.6 5.3–22.2 4.9–23.4 5.3–21.8 6.9–27.1

Abbreviations: OAR ⫽ organ at risk; IMRT ⫽ intensity-modulated radiotherapy.

Fig. 2. Cumulative dose volume histogram (DVH) for lung lesion patient. Whole-lung DVH for different treatment plan categories presented.

values relative to the standard Helax TMS plans (varying nominal energies), which was probably an effect of the conformity gain using the micro-MLC, the reason for the significantly greater value of V6 Gy for IMRT BrainSCAN plans was most probably not related to the differences in nominal energy. These V6 Gy differences between the IMRT solutions were likely due to the differences in segment shape, segment number, and the large number of MUs for the IMRT plans. Although the V3 Gy was greatest for the IMRT plans in general, the interpatient variation in V3 Gy values was considerable for a specific treatment technique and planning system. Dose distribution conformity was best for all micro-MLC– based treatment techniques. However, the improved conformity resulted in only negligible improvements of normal tissue sparing. The negligible influence of leaf width on the mean OAR dose for this SBRT application can be explained by the relatively large-volume ratio between PTV and OAR. IMRT treatment planning for a simultaneous integrated boost technique was not straightforward. These findings are in agreement with a recent publication reporting on a simultaneous integrated boost technique for head-and-neck cancer (30). Although in both TPSs, the IMRT or optimization module allows one to set DVH constraints for target and/or OARs, different DVH constraints needed to be applied (i.e., using the same DVH constraint set on both TPSs did not result in the same or acceptable IMRT solutions). This was because of the dif-

ferent optimization algorithms applied in either TPS. The Helax TMS system uses an objective function in the optimization module, which is based on the quadratic difference between the desired and actual dose, and the BrainSCAN system uses a so-called likelihood method during optimization (18). The definition of the constraints was not straightforward and, therefore, the class solution concept for DVH constraints was much easier to implement on the BrainSCAN system. Moreover, additional help structures (“fictive” volumes of interest) needed to be defined and constraints needed to be specified to achieve the best possible IMRT plan on the Helax TMS system. The number of MUs increased slightly for the Helax TMS IMRT plans but doubled for IMRT plans obtained with the BrainSCAN system. This was also accompanied by a relatively high number of IMRT segments for BrainSCAN; these, however, were not shown to correlate with the MU number. The delivery time for the segmental MLC delivery technique cannot be directly derived from the number of MUs and the dose rate of the linear accelerator. In between the delivery of segments, the leafs need to move and the beam on and off time may play a role in the overall treatment delivery time. IMRT techniques resulting in large delivery times might hinder the application of respiration control techniques, such as deep inspiration breath hold. Quality assurance for IMRT (i.e., verification of treatment plans) is often based on experimental techniques. Hence, the

Table 6. Mean 6-Gy and 3-Gy volumes for different treatment plan categories 6-Gy Isodose volume (cm3)

3-Gy Isodose volume (cm3)

Treatment plan category

Mean ⫾ SD

Range

Mean ⫾ SD

Range

Standard Helax TMS IMRT Helax TMS Standard BrainSCAN IMRT BrainSCAN Dynamic arc Standard Helax, collapsed cone IMRT Helax, collapsed cone

416 ⫾ 226 424 ⫾ 231 395 ⫾ 217 473 ⫾ 260 373 ⫾ 233 408 ⫾ 198 659 ⫾ 292

111–840 96–840 92–698 102–924 75–878 131–786 222–1190

1093 ⫾ 453 1205 ⫾ 487 1133 ⫾ 414 1176 ⫾ 415 1022 ⫾ 450 1131 ⫾ 447 1648 ⫾ 546

429–1855 388–1901 420–1703 488–1627 256–1635 495–1855 721–2391

Abbreviation: IMRT ⫽ intensity-modulated radiotherapy.

Impact of IMRT on SBRT

Fig. 3. Isodose distributions in axial plane obtained with various approaches for patient with two lung lesions but treated with single isocenter. (a) standard Helax TMS, (b) standard Helax TMS collapsed cone, (c) Intensity-modulated radiotherapy (IMRT) Helax TMS, (d) IMRT Helax TMS collapsed cone, (e) standard BrainSCAN, (f) IMRT BrainSCAN, (g) dynamic arc. Blue areas represent 32%, green 65%, and red 90% isodose levels.

total workload when applying IMRT to SBRT would be considerably increased, taking into account treatment planning, delivery, and quality assurance. A dose prescription to the 65% isodose surface with large dose inhomogeneities in the PTV is one characteristics of SBRT in the thoracic region. The evaluated IMRT optimization algorithms tended to aim at a sharp fall off in the DVH of the PTV beyond the prescription isodose. This is, however, not optimal for SBRT appli-

● P. DVORAK et al.

1579

cations. In both treatment planning systems, dose prescriptions for the CTV, as a boost region, had to be defined for the IMRT plans. Regarding target parameters, it was shown that the mean PTV doses of the IMRT BrainSCAN plans were lower than those of the standard plans using the same system. This could be explained by the lower doses in the volume between the PTV and CTV (Fig. 1). This behavior is in contrast to the Helax TMS IMRT solution, in which target parameters showed values similar to those for the standard plans. The BrainSCAN IMRT solution was significantly more sensitive to multiple dose specifications in the target (e.g., for the simultaneous integrated boost technique). Regarding normal tissue sparing, the dynamic arc technique resulted in the lowest 3-Gy and 6-Gy isodoses volume. The steep dose gradient being a characteristic feature of this delivery technique was, however, only slightly reflected in the mean OAR dose, again owing to the relatively large-volume ratio between the PTV and OARs. Considering dynamic arc results for EUDCTV, the mean CTV dose and, particularly, for the mean PTV dose, one can conclude that the dynamic arc may be disadvantageous when considering tumor control. However, for applications in which target dose uniformity is of concern and thus higher treatment isodose levels are used, the advantage of smaller, lower isodose volumes will dominate. With the Helax TMS system, standard plans, as well as IMRT plans, originally optimized with a pencil beam algorithm but recalculated with a superposition method, resulted in a different treatment plan compared with the original plan using a pencil beam model. For the recalculated standard conformal plans, most evaluation parameters considered in this study were generally comparable or slightly worse compared with the original pencil beam plan. However, these recalculated plans could be further modified by adapting the field size and shape in a trial-and-error process to achieve the desired treatment planning goal. Because of the normalization, which is necessary to achieve comparable IMRT plans with the Helax TMS, the Table 7. Mean number of monitor units necessary to deliver prescribed dose for different treatment plan categories MU/treatment plan Treatment plan category

Mean ⫾ SD

Range

Standard Helax TMS IMRT Helax TMS Standard BrainSCAN IMRT BrainSCAN Dynamic arc Standard Helax, collapsed cone IMRT Helax, collapsed cone

1940 ⫾ 119 2273 ⫾ 295 1878 ⫾ 117 4074 ⫾ 728 1938 ⫾ 182 2079 ⫾ 213 2512 ⫾ 287

1671–2096 1750–2668 1706–2071 3033–5247 1666–2299 1685–2448 1811–2836

Abbreviations: MU ⫽ monitor unit; IMRT ⫽ intensity-modulated radiotherapy.

1580

I. J. Radiation Oncology

● Biology ● Physics

mean dose to the CTV will be the same for both IMRT plans. This is not necessarily true for the OARs and the PTV, because the resulting “blurred” relative dose distribution after recalculation is very different from the original (Fig. 3). Consequently, the target dose and organ dose parameters were different for the recalculated IMRT plans. In general, for any treatment planning system, discrepancies in the final dose distribution of an IMRT plan might occur after recalculation with the most accurate dose calculation algorithm, because the dose calculation engine used during optimization can be based on a simplified model for the sake of calculation speed. Another limiting factor in IMRT might be that beam parameters or kernel parameters for almost all dose calculation algorithms are usually derived from fitting calculated dose distributions to measured ones for a variety of uniform intensity beams. In other words, they are optimized for “standard” applications and often no option to “tune” an algorithm for IMRT is available. For all these reasons, it is considered to be of utmost importance that in the presence of large inhomogeneities, computerized treatment optimization modules are based on the most accurate dose calculation algorithms, especially when using greater

Volume 61, Number 5, 2005

energy photon beams. Similar findings have been reported by others (17). CONCLUSION Standard conformal treatment techniques for SBRT of thoracic lesions, based on forward planning, could not be improved with inverse IMRT approaches. Inverse treatment planning and IMRT for SBRT of liver and lung lesions was not straightforward, because a relatively small target is surrounded by a much larger OAR (i.e., liver or lung). This geometry is very different from those for established IMRT applications (e.g., prostate or head and neck) in which, in general, the nearby OARs that need to be spared are smaller than the PTV or of about equal size. Stereotactic dose prescriptions and inhomogeneous target dose distributions can be simulated as an integrated simultaneous boost for inverse planning. Finally, dose calculation algorithms applied in optimization modules for IMRT applications in the thoracic region need to be based on the most accurate dose calculation algorithms (e.g., superposition algorithm), especially when using higher energy photon beams.

REFERENCES 1. Podgorsak EB, Podgorsak MB. Stereotactic irradiation. In: Van Dyk J, editor: The modern technology of radiation oncology. Madison, WI: Medical Physics Publishing;1999. 2. Nagata Y, Negoro Y, Aoki T, et al.: Clinical outcomes of 3D conformal hypofractionated single high-dose radiotherapy for one or two lung tumors using a stereotactic body frame. Int J Radiat Oncol Biol Phys 2002;52:1041–1046. 3. Schefter T, Gaspar LE, Kavanagh B, et al. Hypofractionated extracranial stereotactic radiotherapy for liver tumors. Int J Radiat Oncol Biol Phys 2003;57(Suppl.):S282. 4. Qian G, Lowry J, Silverman P, et al. Stereotactic extra-cranial radiosurgery for renal cell carcinoma [Abstract]. Int J Radiat Oncol Biol Phys 2003; 57(Suppl):S283. 5. Blomgren H, Lax I, Naslund I, et al. Stereotactic high dose fraction radiation therapy of extracranial tumors using an accelerator: Clinical experience of the first thirty-one patients. Acta Oncol 1995;34:861– 870. 6. Uematsu M, Yamammoto F, Takai K, et al. Stereotactic radiation therapy for primary or metastatic lung cancer: Preliminary experience with a linear accelerator-based treatment unit [Abstract]. Int J Radiat Oncol Biol Phys 1996;36(Suppl. 1):352. 7. Wulf J, Haedinger U, Oppitz U, et al. Stereotactic radiotherapy of extracranial targets: CT-simulation and accuracy of treatment in the stereotactic body frame. Radiother Oncol 2000;57:225–236. 8. Hädinger U, Thiele W, Wulf J. Extracranial stereotactic radiotherapy: Evaluation of PTV coverage and dose conformity. Zeitschrift Med Phys 2002;12:221–229. 9. Negoro Y, Nagata Y, Aoki T, et al. The effectiveness of an immobilization device in conformal radiotherapy for lung tumor: Reduction of respiratory tumor movement and evaluation of the daily setup accuracy. Int J Radiat Oncol Biol Phys 2001;50:889 – 898. 10. Herfarth KK, Debus J, Lohr F, et al. Extracranial stereotactic radiation therapy: Set up accuracy of patients treated for liver metastases. Int J Radiat Oncol Biol Phys 2000;46:329 –335. 11. Wulf J, Hädinger U, Oppitz U, et al. Impact of target repro-

12.

13.

14.

15.

16.

17.

18.

19.

20.

21.

ducibility on tumor dose in stereotactic radiotherapy of targets in the lung and liver. Radiother Oncol 2003;66:141–150. Yenice KM, Lovelock DM, Hunt MA, et al. CT image-guided intensity-modulated therapy for paraspinal tumors using stereotactic immobilization. Int J Radiat Oncol Biol Phys 2003; 55:583–593. Murphy MJ, Martin D, Whyte R, et al. The effectiveness of breath-holding to stabilize lung and pancreas tumors during radiosurgery. Int J Radiat Oncol Biol Phys 2002;53:475– 482. Grills IS, Yan D, Martinez AA, et al. Potential for reduced toxicity and dose escalation in the treatment of inoperable non–small-cell lung cancer: A comparison of intensity-modulated radiation therapy (IMRT), 3D conformal irradiation, and elective nodal irradiation. Int J Radiat Oncol Biol Phys 2003;57:875– 890. Liu HH, Wang X, Dong L, et al. Feasibility of sparing lung and other thoracic structures with intensity-modulated radiotherapy for non–small-cell lung cancer. Int J Radiat Oncol Biol Phys 2004;58:1268 –1279. Murshed H, Liu HH, Liao Z, et al. Dose and volume reduction for normal lung using intensity-modulated radiotherapy for advanced-stage non–small-cell lung cancer. Int J Radiat Oncol Biol Phys 2004;58:1258 –1267. Scholz C, Nill S, Oelfke U. Comparison of IMRT optimization based on a pencil beam and a superposition algorithm. Med Phys 2003;30:1909 –1913. Georg D, Dieckmann K, Bogner J, et al. Impact of micromultileaf collimator on stereotactic radiotherapy of uveal melanoma. Int J Radiat Oncol Biol Phys 2003;55:881– 891. Monk JE, Perks JR, Doughty D, et al. Comparison of a micro-multileaf collimator with a 5-mm-leaf-width collimator for intracranial stereotactic radiotherapy. Int J Radiat Oncol Biol Phys 2003;57:1443–1449. Zhu XR, Klein EE, Low DA. Geometric and dosimetric analysis of multileaf collimation conformity. Radiother Oncol 1998;47:63– 68. Lax I, Blomgren H, Naslund I, et al. Stereotactic radiotherapy

Impact of IMRT on SBRT

22. 23. 24. 25. 26.

27.

of malignancies in the abdomen: Methodological aspects. Acta Oncol 1994;33:677– 683. Cosgrove VP, Jahn U, Pfaender M, et al. Commissioning of a micro multi-leaf collimator and planning system for stereotactic radiosurgery. Radiother Oncol 1999;50:325–336. Ahnesjö A, Aspradakis MM. Dose calculations for external photon beams in radiotherapy. Phys Med Biol 1999;44:R99 – R155. Georg D, Kroupa B. Pre-clinical evaluation of an inverse planning module for segmental MLC based IMRT delivery. Phys Med Biol 2002;47:N303–N314. Niemierko A. Reporting and analyzing dose distributions: A concept of equivalent uniform dose. Med Phys 1997;24:103–110. Kavanagh BD, Timmerman RD, Benedict SH, et al. How should we describe the radiobiologic effect of extracranial stereotactic radiosurgery: Equivalent uniform dose or tumor control probability? Med Phys 2003;30:321–324. Dawson LA, Normolle D, Balter JM, et al. Analysis of radi-

● P. DVORAK et al.

1581

ation-induced liver disease using the Lyman NTCP model. Int J Radiat Oncol Biol Phys 2002;53:810 – 821. 28. Cheng JC, Wu JK, Huang CM, et al. Radiation-induced liver disease after three-dimensional conformal radiotherapy for patients with hepato-cellular carcinoma: Dosimetric analysis and implication. Int J Radiat Oncol Biol Phys 2002;54:156 –162. 29. Galvin JM, Ezzell G, Eisbrauch A, et al., for the American Society for Therapeutic Radiology and Oncology and American Association of Physicists in Medicine. Implementing IMRT in clinical practice: A joint document of the American Society for Therapeutic Radiology and Oncology and the American Association of Physicists in Medicine. Int J Radiat Oncol Biol Phys 2004;58:1616 –1634. 30. Fogliata A, Bolsi A, Cozzi L, et al. Comparative dosimetric evaluation of the simultaneous integrated boost with photon intensity modulation in head and neck cancer patients. Radiother Oncol 2003;69:267–275.