TREATMENT PLANNING AND DELIVERY OF INTENSITY-MODULATED RADIATION THERAPY FOR PRIMARY NASOPHARYNX CANCER

PII S0360-3016(00)01389-4 Int. J. Radiation Oncology Biol. Phys., Vol. 49, No. 3, pp. 623 632, 2001 Copyright 2001 Elsevier Science Inc. Printed in the USA. All rights reserved 0360-3016/01/$ see front matter CLINICAL INVESTIGATION Head and Neck TREATMENT PLANNING AND DELIVERY OF INTENSITY-MODULATED RADIATION THERAPY FOR PRIMARY NASOPHARYNX CANCER MARGIE A. HUNT, M.S.,* MICHAEL J. ZELEFSKY, M.D., SUZANNE WOLDEN, M.D., CHEN-SHOU CHUI, PH.D.,* THOMAS LOSASSO, PH.D.,* KENNETH ROSENZWEIG, M.D., LANCEFORD CHONG, M.D., SPIRIDON V. SPIROU, PH.D.,* LISA FROMME, M.S.,* MOIRA LUMLEY, M.S.,* HOWARD A. AMOLS, PH.D.,* CLIFTON C. LING, PH.D.,* AND STEVEN A. LEIBEL, M.D. Departments of *Medical Physics and Radiation Oncology, Memorial Sloan-Kettering Cancer Center, New York, NY Purpose: To implement intensity-modulated radiation therapy (IMRT) for primary nasopharynx cancer and to compare this technique with conventional treatment methods. Methods and Materials: Between May 1998 and June 2000, 23 patients with primary nasopharynx cancer were treated with IMRT delivered with dynamic multileaf collimation. Treatments were designed using an inverse planning algorithm, which accepts dose and dose volume constraints for targets and normal structures. The IMRT plan was compared with a traditional plan consisting of phased lateral fields and a three-dimensional (3D) plan consisting of a combination of lateral fields and a 3D conformal plan. Results: Mean planning target volume (PTV) dose increased from 67.9 Gy with the traditional plan, to 74.6 Gy and 77.3 Gy with the 3D and IMRT plans, respectively. PTV coverage improved in the parapharyngeal region, the skull base, and the medial aspects of the nodal volumes using IMRT and doses to all normal structures decreased compared to the other treatment approaches. Average maximum cord dose decreased from 49 Gy with the traditional plan, to 44 Gy with the 3D plan and 34.5 Gy with IMRT. With the IMRT plan, the volume of mandible and temporal lobes receiving more than 60 Gy decreased by 10 15% compared to the traditional and 3D plans. The mean parotid gland dose decreased with IMRT, although it was not low enough to preserve salivary function. Conclusion: Lower normal tissue doses and improved target coverage, primarily in the retropharynx, skull base, and nodal regions, were achieved using IMRT. IMRT could potentially improve locoregional control and toxicity at current dose levels or facilitate dose escalation to further enhance locoregional control. 2001 Elsevier Science Inc. Nasopharynx cancer, Intensity modulation, Treatment planning. INTRODUCTION The current standard treatment for advanced nasopharyngeal carcinoma in the United States consists of definitive radiotherapy plus cisplatin-based chemotherapy. While the recent addition of chemotherapy has been shown to improve disease-free survival, the specific impact of chemotherapy on local control has not been well established (1, 2). Although conventional external-beam radiation techniques can achieve clinical local control rates for T1 and T2 tumors in the range of 70 90%, approximately 50% of patients with advanced disease will fail locally (3 7). Prognostic factors associated with decreased local control include age, stage, extent of parapharyngeal disease, and base of skull or cranial nerve involvement (3 6, 8). Although locoregional control improves as a function of dose, advanced stage Reprint requests to: M. A. Hunt, Department of Medical Physics, Memorial Sloan-Kettering Cancer Center, Box 84, 1275 York Avenue, New York, NY 10021. E-mail: huntm@mskcc.org Supported in part by Grant CA 59017 from the National Cancer tumors fare poorly even with doses in excess of 70 Gy (3, 7, 9 11). Radiotherapy of nasopharynx cancer traditionally employs parallel opposed lateral photon fields for a significant portion of the treatment. The spinal cord is blocked after 40 45 Gy and the dose to the cervical lymph nodes is supplemented by low-energy electrons. In some institutions, opposed lateral photon fields are used for the entire treatment course, whereas in others, a 3-field beam arrangement (12), brachytherapy (13 15), or stereotactic irradiation (16) is used for the final boost. Although each of these techniques has advantages, treatment of large volume disease particularly near the skull base remains a challenge. Kutcher et al. (17) demonstrated that three-dimensional conformal therapy (3D-CRT), consisting of CT-assisted 3D-planning Institute, Department of Health and Human Services, Bethesda, Maryland. Accepted for publication 9 August 2000. 623

624 I. J. Radiation Oncology Biology Physics Volume 49, Number 3, 2001 techniques and multiple-shaped fields can concomitantly improve tumor coverage in this site and decrease normal tissue doses. As a result, a 7-field conformal technique was designed in the early 1990s at Memorial Sloan-Kettering Cancer Center to improve coverage in the retropharynx and skull base (18, 19). Although this technique has been implemented at MSKCC for approximately 10 years, its use is still confined primarily to boost treatment because of substantial dose inhomogeneity and the complexity involved in treatment planning and delivery. Intensity-modulated radiotherapy (IMRT) is an advanced form of 3D-CRT with two key enhancements: (1) computerized iterative treatment plan optimization using the inverse technique, and (2) the use of intensity-modulated radiation beams. IMRT was implemented at MSKCC in October 1995 for the treatment of cancer of the prostate (20). Beginning in May 1998, we extended the use of IMRT to the treatment of the nasopharynx using the same 7-field beam arrangement as in our previous 3D-CRT experience. This approach yielded the dual benefit of improving the dose distribution and simplifying the planning and treatment delivery processes. The purpose of this report is to discuss the technical aspects of this treatment, i.e., treatment planning, quality assurance, and delivery, for the first 23 consecutive patients treated in our department, and to compare the IMRT dose distributions with those of conventional treatments. METHODS AND MATERIALS IMRT planning and delivery Between May 1998 and June 2000, 23 patients with newly diagnosed primary nasopharynx cancer were treated with IMRT delivered with dynamic multileaf collimation (DMLC). All patients were immobilized in the supine position with custom Aquaplast (Aquaplast, Wycoff Heights, NJ) masks and target localization was accomplished using CT Simulation (AcQSim, Marconi Medical Systems, Cleveland, OH). CT images indexed every 3 mm were obtained, extending from the vertex to 5-cm inferior to the clavicular heads. Planned treatment for each patient consisted of 54 Gy to sites of presumed microscopic disease and 70 Gy to sites of gross disease prescribed to the maximum isodose level encompassing the PTV. Treatment was delivered using a concomitant boost schedule consisting of 1.8 Gy fractions once a day to 36 Gy followed by twice-a-day treatment using 1.8 Gy and 1.6 Gy fractions separated by at least 6 h. The 1.6 Gy fractions treated sites of gross disease only. All treatments were delivered using 6-MV x-rays. Separate target volumes in the nasopharynx and nodal regions were defined according to the definitions of ICRU 50 (21). The nasopharynx clinical target volume, CTV NX, encompassed the entire nasopharynx and extension of the disease into the parapharyngeal and retropharyngeal tissues, whereas the uninvolved cervical lymph nodes were included in the elective nodal target volume, CTV END. The gross tumor volume, GTV, consisting of gross disease in the nasopharynx Fig. 1. The 7-field beam arrangement used for intensity-modulated radiation therapy (IMRT) treatment of the nasopharynx. The same beam arrangement, planned conventionally, was also used for a segment of the three-dimensional (3D) conformal comparison plan. For both the IMRT and 3D conformal plans, the supraclavicular nodes are treated with a separate lower neck field. and neck, was also defined. The planning target volumes (PTV NX, PTV END, PTV GR ) corresponded to the two CTVs and the GTV with an additional one centimeter margin in all directions except posteriorly near the spinal cord and brainstem where the margin was decreased to 0.6 cm. PTV NX and PTV END were used for planning the initial phase of treatment to 54 Gy, whereas PTV GR was used to plan the 16 Gy boost. The supraclavicular nodes were treated conventionally with either a single anterior or parallel opposed anterior and posterior fields matched to the IMRT beams at the level of the thyroid notch. No graphical planning was performed for the supraclavicular treatment and its effect is not included in this analysis. The normal structures evaluated included the spinal cord, brainstem, mandible, temporal lobes, and parotid glands. For the first 5 patients, treatment was started with standard parallel opposed lateral fields for the first 18 Gy to allow sufficient time for IMRT planning and quality assurance procedures. Thereafter, the procedures were streamlined sufficiently to shorten the IMRT planning process to 5 working days, the same as other complex plans at our institution, allowing IMRT to be used for the full 70 Gy treatment course. The IMRT beam arrangement consisted of 7 equi-spaced coplanar beams that entered laterally, posterio-laterally, and posteriorly (Fig. 1). This is the same beam arrangement used at MSKCC for conventional 3D conformal boost treatment of the nasopharynx as described by Leibel et al. (18). Treatment planning was performed using the inverse planning algorithm of Spirou and Chui (22), integrated with the MSKCC treatment planning system (23). The inverse algorithm uses a least squares objective function and conjugate

IMRT for nasopharynx cancer M. A. HUNT et al. 625 Table 1. Clinical criteria and inverse planning algorithm constraint template Structure Clinical plan criteria IMRT constraint template PTV Minimum dose; prescription Prescription dose: 70 Gy (or 54 Gy) Maximum dose: 120% of prescription Minimum dose: 66.5 Gy (or 51.3 Gy) Penalty: 25 Maximum dose: 73.5 Gy (or 56.7 Gy), Penalty: 25 Spinal cord Maximum dose: 40 Gy Maximum dose: 35 Gy Penalty: 100 Brain stem Maximum dose: 45 Gy Maximum dose: 35 Gy Penalty: 100 PTV Refers to all PTVs including PTV NX, PTV END, PTV GR. For each PTV, the appropriate prescription dose is entered into the algorithm constraint template. gradient minimization method to find an optimum solution based on a set of user-defined constraints including maximum and minimum dose constraints for targets and dose and dose volume constraints for normal structures. All of these constraints are so-called soft constraints indicating that they may be violated with a cost or penalty. The penalties are specified by the user during the planning process and define the relative importance of each constraint. Our clinical criteria defining an acceptable plan are listed in Table 1 together with the optimization algorithm constraint template. Because structure doses are penalized during optimization only if they exceed the limits set by the user, the algorithm constraints are always more stringent than the clinical criteria. The optimization constraint template listed in Table 1 is a result of trial and error using our planning system and is not necessarily applicable to other inverse planning systems. In practice, this template serves as the starting point for planning with refinement of the constraints done for each patient to produce an optimal plan. For patients with bilateral positive neck nodes, one IMRT plan is used for the entire course, treating all involved sites. For patients with unilaterally positive or an uninvolved neck, two plans are created. The first plan treats the nasopharynx (PTV NX ) and the elective neck (PTV END )to54gy. The second plan, used for the 16 Gy boost, targets sites of gross disease only (PTV GR ) and considers the uninvolved neck as a normal structure during optimization. Treatment is delivered using dynamic multileaf collimation (DMLC) on a Varian accelerator equipped with a 52-leaf MLC (Varian Medical Systems, Palo Alto, CA). During the inverse planning process, an intensity profile is created for each beam that is subsequently translated into leaf motion using the algorithm designed by Spirou and Chui (24). For the first 7 patients, relative and absolute dose distributions were verified using the film dosimetry technique described by Wang et al. (25) and Losasso et al. (26). Briefly, the dose distribution of each IM field was recalculated in a flat phantom at a depth of 10 cm and compared with film measurements made in the same geometry. This comparison between measurement and calculation served as verification of three major aspects of the treatment: the planning process, the calculation algorithm, and the treatment delivery. Beginning in May 1999, a monitor unit (MU) verification program for DMLC fields became available as an additional tool to verify the dose calculations and the planning process. This program converts leaf motion for a given beam into its corresponding intensity profile and then calculates the dose at a user specified check point using a pencil beam convolution-based algorithm. A digitally reconstructed radiograph is created for each field and overlaid with a projection of the DMLC aperture, which indicates the initial and final leaf positions (Fig. 2). DMLC leaf motion files and a file containing the DMLC apertures for all treatment fields are transferred to the treatment machine via floppy diskette. A localization film of each field with the DMLC aperture is compared with the DRR before the first Fig. 2. Digitally reconstructed radiograph (DRR) for a left posterior oblique beam showing wire frame outlines of the spinal cord and brainstem as well as the dynamic multileaf collimator (DMLC) aperture, which indicates the initial and final positions of the leaf pairs used during dynamic treatment.

626 I. J. Radiation Oncology Biology Physics Volume 49, Number 3, 2001 treatment. During daily treatment, all dynamic leaf positions are verified at 55-ms intervals by the MLC controller. In addition, initial and final leaf positions for each dynamic field are verified by our in-house record and verify system. During the initial development of this technique, two challenges related to treatment delivery were encountered. Due to the mechanical design of the MLC, DMLC delivery on Varian accelerators is limited to fields no wider than 14.5 cm. For most nasopharynx patients, beams from the posterior and posterior oblique directions exceed this limit. Each wide field is therefore divided into two narrower fields separated by a 0.5 cm gap at the level of the spinal cord and brainstem where the beam intensity is minimal. This increases the average number of treatment fields to 10, each of which is delivered in approximately one minute. The second delivery issue is related to the need to extend treatment beyond the patient skin surface in the area of the neck to ensure adequate margin to account for setup uncertainty, patient motion and breathing. Under normal circumstances, the intensity profile created by our inverse planning algorithm ends at the patient skin surface. A modification to the algorithm was made to allow the intensity profile to be extended in the nodal regions by a user-defined amount, typically at least 2 cm beyond the skin surface. This issue is not unique to the treatment of head and neck tumors and was recently discussed by Hong et al. (27) in their implementation of IMRT for breast cancer. Plan comparison IMRT treatment plans were compared with a traditional plan and a 3D conformal plan for 6 patients. Two patients each had no nodal involvement, unilateral, or bilateral neck disease. The traditional plan consisted of parallel opposed lateral 6-MV photon fields and represents a common method of treatment. The initial lateral fields encompassed all sites of gross and presumed microscopic disease. The blocks were modified twice: once at 45 Gy to exclude the spinal cord and again at 54 Gy to exclude all but the gross disease. For each set of fields, the prescription dose was delivered to a point midplane on the central axis or in the center of the treatment area if the central axis was near the aperture edge. After spinal cord blocking, lateral electron fields (9 MeV prescribed to the 90% isodose line) were used to boost the dose to the posterior cervical lymph nodes to 54 Gy for presumed microscopic involvement or to 70 Gy for gross tumor involvement. At our institution, treatment with this method would have been delivered based on a manual treatment time calculation excluding inhomogeneity corrections. Therefore, the beam weighting was first determined without inhomogeneity corrections and the dose distributions were then recalculated to include the tissue homogeneity effects. The 3D plan was the standard treatment at MSKCC before the implementation of IMRT. Parallel opposed lateral 6-MV photon fields identical to those described for the Traditional plan were used to deliver 36 Gy followed by a 7-field 3D plan for an additional 34 Gy. The 7-field beam directions were identical to those used for IMRT (Fig. 1). Dose was prescribed to midplane on the central axis for the first phase and to an isodose level encompassing generally at least 95% of the PTV for the 7-field portion of the treatment. The 7-field 3D plan uses full thickness blocks over the spinal cord and brainstem to create a concave dose distribution encompassing as much of the PTV as possible while sparing the spinal cord and brainstem. Partial transmission blocks (70 80%) placed over the entire width of the field and extending inferiorly from approximately 2-cm below the mandibular ramus were used to improve target dose homogeneity in the neck. Forty-five degree wedge filters were used for all but the posterior beams. A summary of the beam arrangements and treatment regimens for the three plan types is given in Table 2. Plans were compared through visual inspection of dose distributions and analysis of dose volume histogram statistics. To evaluate target coverage, the dose received by 95% of the PTV (D 95 ), dose received by 5% of the PTV (D 05 ), and mean dose were recorded. D 05 was used as the measure of maximum dose for the spinal cord, brainstem, mandible, and temporal lobes. V 66 Gy and V 60 Gy (volumes receiving at least 66 and 60 Gy) were chosen as indicators of the high-dose volumes for the mandible and temporal lobes, respectively. For the parotid glands, the mean dose and the volume receiving at least 50 Gy were evaluated. RESULTS IMRT planning and delivery All patients completed the planned treatment with intensity modulation. The average IMRT planning time including contouring, beam definition, optimization, evaluation and documentation was approximately 8hincomparison to an estimated 12 h for the 3D plan and 1 h for the traditional plan. Daily patient setup and treatment times (including weekly verification films) averaged 20 25 min for IMRT, compared with an estimated 40 min for the 7-field 3D plan and 15 20 min for parallel opposed treatments. Actual IMRT treatment delivery time, including leaf-motion file download and beam-on time, is approximately 1 min per field, similar to that for conventional treatments. A typical intensity profile for a left posterior oblique IMRT beam is shown in Fig. 3. The area of diminished intensity corresponds to the spinal cord and brainstem. Immediately adjacent, the intensity increases to improve coverage in the skull base. The intensity pattern is quite irregular and could not be easily reproduced using simple beam modulators such as partial transmission blocks or wedges. A comparison of the calculated and measured dose distributions in a flat phantom at a depth of 10 cm for a left posterior oblique beam is presented in Fig. 4. Excellent agreement is observed throughout the field, within 2% in the central region and 2 mm in regions of high-dose gradient.

IMRT for nasopharynx cancer M. A. HUNT et al. 627 Table 2. Summary of treatment plans used for plan comparison Plan name Field arrangement PTVs included Delivered dose Cumulative dose Traditional Opposed laterals PTV NX 45 Gy 45 Gy 6 MVX PTV END PTV GR Opposed laterals with cord block PTV NX 9Gy 54Gy bilateral 9 MeVC E strips PTV END Opposed lateral gone down, PTV GR 16 Gy 70 Gy involved neck 9 MeV E strips PTV GR PTV NX 3D Conformal Opposed lateral 6 MVX PTV END 36 Gy 36 Gy PTV GR PTV NX 7-field conformal paln PTV END 18 Gy 54 Gy PTV GR 7-field conformal plan PTV GR 16 Gy 70 Gy IMRT 7-field IMRT plan PTV NX 54 Gy 54 Gy PTV END PTB GR 7-field IMRT plan PTV GR 16 Gy 70 Gy Plan comparison Dose distributions for the IMRT, traditional, and 3D plans through the center of the nasopharynx and neck for a patient with bilateral involved neck nodes are shown in Fig. 5. PTV coverage at the 70 Gy level is poor for the traditional plan as a result of the midplane prescription method and the inability of opposed fields to adequately cover the skull base. Coverage of the posterior cervical neck nodes is also inadequate with this plan due to the inability of the electrons to treat the medial aspects of the nodes while sparing the spinal cord. Improved target coverage is evident with both the 3D and IMRT plans. Maximum doses of 135 140% of the prescription are unavoidable with the seven field conformal phase of the 3D plan. Because this beam arrangement is used for approximately half of the treatment, the maximum dose for the entire course is generally about 125% of the prescription. With IMRT and the same 7 beams, the target dose homogeneity is improved to within 15 20% of the prescription. Dose volume histograms for PTV GR, the brainstem, parotid, mandible and temporal lobes for this patient are presented in Fig. 6. Inadequate coverage of the PTV with the traditional plan is again clearly evident. Coverage improves with either the 3D plan or IMRT but the dose homogeneity is superior with IMRT. The brainstem dose is highest with the traditional plan and lowest with IMRT. The maximum doses received by the mandible, temporal lobes, and parotid glands are similar for all three plans, but less volume is treated to the highest dose levels using IMRT. A summary of DVH statistics for the six patients is presented in Table 3. IMRT delivered the highest minimum PTV dose, 69.4 Gy, which was 5.5% higher than the 3D plan and nearly 30% higher than the traditional plan. IMRT was unable to achieve complete coverage of PTV GR at the Fig. 3. Intensity profile for a left posterior oblique dynamic multileaf collimator (DMLC) field. On the left, the intensity profile for the entire field is shown. The profile for a single leaf pair near the center of the field is shown on the right.

628 I. J. Radiation Oncology Biology Physics Volume 49, Number 3, 2001 Fig. 4. A comparison of the calculated and measured dose distributions for a left posterior oblique dynamic multileaf collimator (DMLC) field. The central low intensity area corresponds to the position of the spinal cord and brainstem. 70 Gy level for only 1 patient, who had disease extending into the sphenoid and ethmoid sinuses. For the other 5 patients, the IMRT plan was the only one that delivered the prescription dose or more to all three PTVs. Dose uniformity across the target, as measured by the ratio of D 05 and D 95, also improved with the IMRT technique, decreasing from 36% with the traditional plan to 22% with 3D and 18% with IMRT. Inadequate target coverage with both the 3D and traditional plans was observed primarily in the retropharynx, the skull base, and medial aspects of the nodal volumes in the neck. The magnitude of the underdosage was much higher with the traditional plan with only 46% of the PTV GR receiving 70 Gy or more, compared to 87% with 3D and 95% with IMRT. The average maximum spinal cord dose (D 05 ) decreased from 49 Gy with the traditional plan to 35 Gy with IMRT, whereas brainstem doses decreased from nearly 56 Gy to 33 Gy. During IMRT optimization, a deliberate attempt was made to constrain the spinal cord and brainstem to very low doses because of the tight conformality of the dose distribution around these structures (Fig. 5a). No attempt to limit the dose to the parotid glands was made for any of the patients considered in this study. The mean parotid dose with IMRT was 60.5 Gy, 10% less than the 3D and the traditional plans. This slight improvement is a result of a tighter conformality of the IMRT dose distribution which leads to a decrease in the parotid volume irradiated to the highest dose levels. DISCUSSION Local control remains a significant problem for many patients with nasopharynx cancer, particularly those with advanced locoregional disease. Chua et al. (28) observed large variations in tumor volume within a given stage, particularly for advanced tumors, and poorer local control and disease-specific survival for patients with volumes larger than 60 cc. Other studies have demonstrated the independent prognostic value of extensive parapharyngeal disease, base of skull or cranial nerve involvement in local control (8, 29, 30). These studies highlight the importance of identifying and delivering high doses to all sites of disease. Unfortunately, because of the close proximity of the spinal cord and brainstem as well as less critical structures, such as the parotid glands, adequate treatment of these regions with conventional approaches is either impossible or associated with significant morbidity. Our data demonstrate that significant improvements in target and normal structure doses are possible using IMRT. The minimum dose to sites of gross disease increased from an average of 54.6 Gy with traditional parallel opposed fields to 69.4 Gy with IMRT. As shown in Fig. 5, this increase of nearly 30% is due principally to improved coverage in the retropharynx, base of skull and medial aspects of the nodal volumes. On average, 95% of the PTV received doses of 70 Gy or more using IMRT compared with only 46% using parallel opposed fields for the entire treatment course. In addition to

IMRT for nasopharynx cancer M. A. HUNT et al. 629 Fig. 5. Axial dose distributions through the center of the nasopharynx and neck for the intensity-modulated radiation therapy (IMRT) [(a), (b)], three-dimensional (3D) conformal [(c), (d)], and traditional [(e), (f)] treatment plans. Note the relatively poor coverage of the skull base and medial nodal regions using the traditional plan and the improved dose conformality of the IMRT plan. improved target coverage, dose uniformity within the target improved with IMRT as demonstrated by a decrease in the dose uniformity index from 34% with parallel opposed fields to 18% with IMRT. Some of the improvement between traditional and IMRT treatment results from using a multi-field beam arrangement. However, as our results show, the dose distribution achieved with a 7-field 3D conformal plan can be further improved with IMRT. Minimum PTV dose increased from 65.7 Gy with 3D to 69.4 Gy with IMRT, an improvement of 5%, and the dose uniformity index decreased from 23% to 18%. One of the main reasons we combine the conventional seven-field beam arrangement with parallel opposed fields in the conformal 3D plan is poor dose uniformity, typically 35 40% from the 3D plan alone. IMRT, using the same 7-field beam arrangement, yields an inherently superior dose distribution, reducing the dose heterogeneity by onehalf. These improvements in the target dose distribution are possible while maintaining or decreasing normal tissue doses. Because of the tight conformality of the dose distributions, we currently require that the spinal cord and brainstem doses not exceed the conservative limits of 40 and 45 Gy, respectively. Our results show that IMRT was capable of further reducing the doses to these structures to approximately 35 Gy without sacrificing PTV coverage. We are currently investigating the effects of random and systematic setup uncertainties on the dose distributions from all three types of plans. Better understanding of the effects of treatment uncertainty may allow us to relax the dose constraints for these structures at current dose levels or dose escalate if desired. No specific attempts were made during this initial development of nasopharyngeal IMRT to constrain the dose to the parotid glands. Although the mean parotid dose decreased by 10% to 60 Gy and the parotid volume exceeding 50 Gy decreased nearly 25% with IMRT, this improvement is not nearly sufficient to preserve salivary function. Eisbruch et al. (31) have used multifield conformal radiotherapy to decrease the irradiated volume of at least one parotid gland for head and neck sites other than nasopharynx to levels which impact morbidity. The average dose delivered to the spared parotid was 21 Gy in 15 patients with oropharynx, oral cavity, hypopharynx, or supraglottic larynx cancer. More recently, the same authors (32) demonstrated a threshold relationship between dose and salivary function on the same patients. Meaningful function was preserved when the mean parotid gland dose was kept below approximately 25 Gy. As shown by Nowak et al. (33), Level

630 I. J. Radiation Oncology Biology Physics Volume 49, Number 3, 2001 Fig. 6. Dose volume histograms comparing the intensity-modulated radiation therapy (IMRT), three-dimensional (3D) conformal, and traditional plans for the patient in Fig. 5. II nodes lie just adjacent to the parotid, implying that at least part of the gland must be included in the PTV of nasopharynx patients. In a preliminary investigation on the feasibility of parotid sparing using IMRT in nasopharynx patients, we have observed that a mean dose of 25 Gy or less to at least one gland may be feasible in patients with N 0 or N 1 disease. For patients with N 2 disease, we have been unsuccessful in achieving a mean dose less than 25 Gy due to the close proximity of the 70 Gy target volume to the parotid glands. These observations are based on a small number of patients, however, and further evaluation will be needed to draw definitive conclusions. Quality assurance (QA) of the treatment planning and delivery processes are significant issues when implementing IMRT treatment for any new site. Initially, we required 2 weeks for planning and QA for each patient. During this time, film dosimetry was routinely performed for all patients to verify DMLC delivery, the accuracy of the calculated MU settings and the integrity of the leaf motion data being transferred to the treatment floor. The comparison between film and calculation consistently demonstrated agreement to within 2% in the primary portion of the field and 2 mm in areas of steep dose gradients. Based on these results and our long term observations of DMLC reproduc- Table 3. Dose volume histogram statistics comparing the intensity-modulated radiation therapy (IMRT), traditional, and three-dimension conformal (3D Conf.) plans Structure Statistic IMRT 3D Conf. Traditional PTV GR Max. dose (D 05 ) 81.8 Gy (3.3) 80.2 Gy (1.0) 74.2 (Gy (2.5) Min. Dose (D 95 ) 69.4 Gy (6.2) 65.7 Gy (5.0) 54.6 Gy (1.7) Mean dose 77.3 Gy (2.4) 74.6 Gy (2.2) 67.9 Gy (1.3) Spinal cord Max. dose (D 05 ) 34.5 Gy (5.5) 44.2 Gy (1.7) 49.1 Gy (0.9) Brain stem Max. dose (D 05 ) 33.1 Gy (5.0) 43.3 Gy (2.7) 56.2 Gy (7.0) Mandible Max. dose(d 05 ) 69.3 Gy (7.4) 73.9 Gy (5.3) 74.6 Gy (0.9) V 66 Gy (%) 9.7% (5.9) 18.6% (11.7) 26.8% (13.9) Temporal lobes Max. dose(d 05 ) 58.7 Gy (12.5) 59.4 Gy (11.1) 67.0 Gy (3.5) V 60 Gy (%) 6.3% (7.1) 9.2% 913.1) 17.3% (8.8) Parotid gland Mean dose 60.5 (8.9) 67.1 (7.0) 67.0 (4.7) V 50 Gy (%) 78.4% (21.2) 97.5% (2.9) 99.9% (0.1) Values represent average and (standard deviations) for 6 patients.

IMRT for nasopharynx cancer M. A. HUNT et al. 631 ibility, we believe that the DMLC treatment delivery process is extremely reliable. Therefore, we have eliminated routine film dosimetry for pretreatment delivery verification for individual patients but continue to use the accelerator MLC position readout and our record and verify system for verifying treatment delivery on a daily basis. For pretreatment, patient-specific QA, we rely on a thorough plan check and an independent calculation of MU settings, just as for conventional treatments. Since this calculation accesses the leaf motion files that will be used for patient treatment as input, it serves two purposes: verification of the dose calculations and assurance that the DMLC data transfer to the treatment floor is correct. Film dosimetry is now reserved for new treatment sites, unusual intensity profiles or independent MU verification checks which yield an unresolved discrepancy in excess of 3%. As a result of QA streamlining and our increasing experience with the optimization routines, we have decreased the required time for a nasopharynx IMRT plan to 5 working days, the same as other complex plans at our institution. Several changes and improvements to our current technique are planned for the following year. To overcome the technical limit of 14.5 cm in treatable field width, modifications are being made to the planning software to automatically divide a DMLC field of excessive width into two deliverable fields with different collimator settings. The division will be done in a way that minimizes potential dosimetric problems resulting from field junctioning. Work will also be done to better understand the impact of setup uncertainty on the exquisitely conformal dose distributions delivered with intensity-modulated beams. In addition, we will continue to evaluate the potential for parotid sparing in this group of patients. In conclusion, IMRT using multiple static DMLC beams improves target coverage and dose uniformity compared to both traditional and 3D conformal methods of treatment. The implementation of this technique at our institution has increased the minimum dose delivered to previously untreated nasopharynx patients to 69.4 Gy, a 5% increase from our previous 3D conformal treatment method. This improvement is possible while simultaneously decreasing normal tissue doses, implying that further dose escalation may be possible if warranted by clinical follow-up. Although the initial commitment to QA and planning procedures is significant, the implementation of an independent method of MU verification and experience with the DMLC delivery systems eventually allows IMRT planning to be completed with similar or less effort than other complex plans. REFERENCES 1. Al-Sarraf M, LeBlanc M, Giri PG, et al. Chemoradiotherapy versus radiotherapy in patients with advanced nasopharyngeal cancer: Phase III randomized Intergroup study 0099. J Clin Oncol 1998;16:1310 1317. 2. Cooper JS, Lee H, Torrey M, et al. Improved outcome secondary to concurrent chemoradiotherapy for advanced carcinoma of the nasopharynx: Preliminary corroboration of the intergroup experience [in process citation]. Int J Radiat Oncol Biol Phys 2000;47:861 866. 3. Perez CA, Devineni VR, Marcial-Vega V, et al. Carcinoma of the nasopharynx: factors affecting prognosis. Int J Radiat Oncol Biol Phys 1992;23:271 280. 4. Frezza G, Barbieri E, Emiliani E, et al. Patterns of failure in nasopharyngeal cancer treated with megavoltage irradiation. Radiother Oncol 1986;5:287 294. 5. Fu KK. Prognostic factors of carcinoma of the nasopharynx. Int J Radiat Oncol Biol Phys 1980;6:523 526. 6. Sanguineti G, Geara FB, Garden AS, et al. Carcinoma of the nasopharynx treated by radiotherapy alone: Determinants of local and regional control [see comments]. Int J Radiat Oncol Biol Phys 1997;37:985 996. 7. Vikram B, Mishra UB, Strong EW, et al. Patterns of failure in carcinoma of the nasopharynx: I. Failure at the primary site. Int J Radiat Oncol Biol Phys 1985;11:1455 1459. 8. Teo P, Yu P, Lee WY, et al. Significant prognosticators after primary radiotherapy in 903 nondisseminated nasopharyngeal carcinoma evaluated by computer tomography. Int J Radiat Oncol Biol Phys 1996;36:291 304. 9. Levendag PC, Nowak PJ, van der Sangen MJ, et al. Local tumor control in radiation therapy of cancers in the head and neck. Am J Clin Oncol 1996;19:469 477. 10. Lee AW, Chan DK, Fowler JF, et al. Effect of time, dose and fractionation on local control of nasopharyngeal carcinoma. Radiother Oncol 1995;36:24 31. 11. Marks JE, Bedwinek JM, Lee F, et al. Dose-response analysis for nasopharyngeal carcinoma. An historical perspective. Cancer 1982;50:1042 1050. 12. Fletcher G, Million RR. Radiation therapy for head and neck neoplasms: Indications, techniques, and results. In: Fletcher G, ed. Textbook of radiotherapy. Philadelphia: Lea & Febiger, 1980. p. 371 377. 13. Goffinet DR, Martinez A, Fee WE Jr. 125 I vicryl suture implants as a surgical adjuvant in cancer of the head and neck. Int J Radiat Oncol Biol Phys 1985;11:399 402. 14. Wang CC, Busse J, Gitterman M. A simple afterloading applicator for intracavitary irradiation of carcinoma of the nasopharynx. Radiology 1975;115:737 738. 15. Harrison LB, Sessions RB, Fass DE, et al. Nasopharyngeal brachytherapy with access via a transpalatal flap. Am J Surg 1992;164:173 175. 16. Cmelak AJ, Cox RS, Adler JR, et al. Radiosurgery for skull base malignancies and nasopharyngeal carcinoma [see comments]. Int J Radiat Oncol Biol Phys 1997;37:997 1003. 17. Kutcher GJ, Fuks Z, Brenner H, et al. Three-dimensional photon treatment planning for carcinoma of the nasopharynx. Int J Radiat Oncol Biol Phys 1991;21:169 182. 18. Leibel SA, Kutcher GJ, Harrison LB, et al. Improved dose distributions for 3D conformal boost treatments in carcinoma of the nasopharynx. Int J Radiat Oncol Biol Phys 1991;20: 823 833. 19. Hunt MA, Kutcher GJ, Burman C, et al. The effect of setup uncertainties on the treatment of nasopharynx cancer. Int J Radiat Oncol Biol Phys 1993;27:437 447. 20. Ling CC, Burman C, Chui CS, et al. Conformal radiation treatment of prostate cancer using inversely- planned intensity-modulated photon beams produced with dynamic multileaf collimation [see comments]. Int J Radiat Oncol Biol Phys 1996;35:721 730. 21. International Commission on Radiation Units and Measurements. Prescribing, recording, and reporting photon beam

632 I. J. Radiation Oncology Biology Physics Volume 49, Number 3, 2001 therapy. Bethesda, MD. International Commission on Radiation Units and Measurements, 1993. 22. Spirou SV, Chui CS. A gradient inverse planning algorithm with dose volume constraints. Med Phys 1998;25:321 333. 23. Mohan R, Barest G, Brewster LJ, et al. A comprehensive three-dimensional radiation treatment planning system. Int J Radiat Oncol Biol Phys 1988;15:481 495. 24. Spirou SV, Chui CS. Generation of arbitrary intensity profiles by dynamic jaws or multileaf collimators. Med Phys 1994;21: 1031 1041. 25. Wang X, Spirou S, LoSasso T, et al. Dosimetric verification of intensity-modulated fields. Med Phys 1996;23(3):317 327. 26. LoSasso T, Chui C, Ling C. Physical and dosimetric aspects of a multileaf collimation system used in the dynamic mode for implementing intensity modulated radiotherapy. Med Phys 1998;25:1919 1927. 27. Hong L, Hunt M, Chui C, et al. Intensity-modulated tangential beam irradiation of the intact breast. Int J Radiat Oncol Biol Phys 1999;44:1155 1164. 28. Chua DT, Sham JS, Kwong DL, et al. Volumetric analysis of tumor extent in nasopharyngeal carcinoma and correlation with treatment outcome. Int J Radiat Oncol Biol Phys 1997; 39:711 719. 29. Chua DT, Sham JS, Kwong DL, et al. Prognostic value of paranasopharyngeal extension of nasopharyngeal carcinoma. A significant factor in local control and distant metastasis. Cancer 1996;78:202 210. 30. Sham JS, Choy D. Prognostic value of paranasopharyngeal extension of nasopharyngeal carcinoma on local control and short-term survival. Head Neck 1991;13:298 310. 31. Eisbruch A, Ship JA, Martel MK, et al. Parotid gland sparing in patients undergoing bilateral head and neck irradiation: Techniques and early results. Int J Radiat Oncol Biol Phys 1996;36:469 480. 32. Eisbruch A, Ten Haken RK, Kim HM, et al. Dose, volume, and function relationships in parotid salivary glands following conformal, and intensity-modulated irradiation of head, and neck cancer [In Process Citation]. Int J Radiat Oncol Biol Phys 1999;45:577 587. 33. Nowak PJ, Wijers OB, Lagerwaard FJ, et al. A three-dimensional CT-based target definition for elective irradiation of the neck. Int J Radiat Oncol Biol Phys 1999;45:33 39.