PURPOSE:To determine the association between radiation doses delivered to the mandible and the occurrence of osteoradionecrosis (ORN). METHODS AND MATERIALS:We reviewed the records of 402 oropharyngeal cancer patients with stage T1 or T2 disease treated with definitive radiation between January 2000 and October 2008 for the occurrence of ORN. Demographic and treatment variables were compared between patients with ORN and those without. To examine the dosimetric relationship further, a nested case-control comparison was performed. One to 2 ORN-free patients were selected to match each ORN patient by age, sex, radiation type, treatment year, and cancer subsite. Detailed radiation treatment plans for the ORN cases and matched controls were reviewed. Mann-Whitney test and conditional logistic regression were used to compare relative volumes of the mandible exposed to doses ranging from 10 Gy-60 Gy in 10-Gy increments. RESULTS:In 30 patients (7.5%), ORN developed during a median follow-up time of 31 months, including 6 patients with grade 4 ORN that required major surgery. The median time to develop ORN was 8 months (range, 0-71 months). Detailed radiation treatment plans were available for 25 of the 30 ORN patients and 40 matched ORN-free patients. In the matched case-control analysis, there was a statistically significant difference between the volumes of mandible in the 2 groups receiving doses between 50 Gy (V50) and 60 Gy (V60). The most notable difference was seen at V50, with a P value of .02 in the multivariate model after adjustment for the matching variables and dental status (dentate or with extraction). CONCLUSIONS:V50 and V60 saw the most significant differences between the ORN group and the comparison group. Minimizing the percent mandibular volume exposed to 50 Gy may reduce ORN risk.

PURPOSE:Consistency in defining and contouring target structures in radiation therapy (RT) is critical for highly conformal RT, for evaluating treatment plans, and for quality assurance in multi-institutional RT trials. The Radiation Therapy Oncology Group (RTOG) has published consensus guidelines for contouring targets for postmastectomy RT. To aid in contouring such structures, we evaluated the potential use of an automated contouring technique, known as deformable image registration-based breast segmentation (DEF-SEG). METHODS AND MATERIALS:The RTOG definitions were used to contour the chest wall (CW); levels I, II, and III axillary nodes (Ax1, Ax2, Ax3); supraclavicular (SCV) nodes; internal mammary (IM) nodes; and the heart. Left-sided and right-sided templates were created. The DEF-SEG was then used to generate auto-segmented contours from the appropriate template to computed tomographic scans of 20 test cases (10 left, 10 right). To assess the accuracy of this method, those contours were manually modified as necessary to match the RTOG definitions, and the extent of the overlap was compared. The dosimetric impact of the difference in contours was then evaluated by comparing dose-volume histograms for modified and unmodified contours. RESULTS:Mean volume-overlap ratios between the unmodified DEF-SEG-generated contours and modified contours were as follows: CW, 0.91; Ax1, 0.68; Ax2, 0.64; Ax3, 0.68; SCV node, 0.66; IM node, 0.32, and the heart, 0.93. Mean differences in volume receiving 45 Gy (V45) for the modified versus unmodified contours were as follows: CW, 2.1%; SCV node, 4.8%; Ax1, 5.1%; Ax2, 5.6%; Ax3, 3.0%; and IM node, 10.1%. Mean differences in V10 between the modified heart and the unmodified heart were 0.4% for right-sided treatment and 0.5% for left-sided treatment. CONCLUSIONS:The DEF-SEG can be helpful for delineating structures according to the RTOG consensus guidelines, particularly for the CW and the heart. No clinically significant dosimetric differences were found between the modified and unmodified contours. The DEF-SEG may be useful for evaluating treatment plans for postmastectomy RT in multi-institutional trials.

Purpose: The Radiation Therapy Oncology Group (RTOG) has published consensus guidelines for contouring relevant anatomy for postmastectomy radiation therapy (RT). How these contours relate to current treatment practices is unknown. We analyzed the dose-volume histograms (DVHs) for these contours using current clinical practice at University of Texas MD Anderson Cancer Center and compared them with the proposed treatment plans to treat RTOG-defined targets to full dose.Methods and Materials: We retrospectively analyzed treatment plans for 20 consecutive women treated with postmastectomy RT for which the treatment targets were the chest wall (CW), level III axilla (Ax3), supraclavicular (SCV), and internal mammary (IM) nodes. The RTOG consensus definitions were used to contour the following anatomic structures: CW; level I, II, and III axillary nodes (Ax1, Ax2, Ax3); SCV; IM; and heart (H). DVHs for these contours and the ipsilateral lung were generated from clinically designed treatment that had actually been delivered to each patient. For comparison regarding dose to normal tissue, new treatment plans were generated with the goal of covering 95% of the anatomic contours to 45 Gy.Results: The prescribed dose was 50 Gy in each case. The mean percent of volumes that received 45 Gy (V45) for the RTOG guideline-based contours were CW 74%, Ax1 84%, Ax2 88%, Ax3 96%, SCV 84%, and IM 80%. Mean heart V10 values were 11% for treatment of left-sided tumors and 6% for right-sided tumors. Mean ipsilateral lung V20 values were 28% for left-sided tumors and 34% for right-sided tumors. For the contour-based plans, mean V45 values were CW 94%, Ax1 95%, Ax2 97%, Ax3 98%, SCV 98%, and IM 85%. Mean heart V-10 values were 14% for treatment of left-sided tumors and 12% for right-sided tumors. Mean ipsilateral lung V-20 values were 32% for left-sided tumors and 45% for right-sided tumors.Conclusions: Clinically derived treatment plans, which have proven efficacy and are the current standard, cover 74% to 96% of the anatomy-based RTOG consensus volumes to the prescription dose. This discrepancy should be considered if treatment planning protocol guidelines are designed to incorporate these new definitions. (C) 2012 American Society for Radiation Oncology. Published by Elsevier Inc. All rights reserved.

The Radiation Therapy Oncology Group (RTOG) has published consensus guidelines for contouring targets for postmastectomy radiation therapy (PMRT). These contours are not typically used with conventional, clinically-derived radiation therapy (RT). To evaluate the dosimetric impact of these target definitions, we compared the dosimetric differences between a contour-based treatment design and conventional RT which is the current institutional practice for PMRT. The RTOG consensus definitions were used to contour the following targets on planning CTs for 10 patients consecutively treated for right-sided tumors: chest wall (CW), level III axilla (Ax3), supraclavicular (SCV), and internal mammary (IM) nodes. Ipsilateral lung (IL), and heart (H) were also delineated. During planning, the superficial 5 mm of CW volume in dose buildup region was subtracted. Clinically-derived treatment plan: Actual delivered treatment plans for each patient were used. The dose distribution from each RT plan was mapped to the CT dataset for a single patient using deformable image registration method (DIR). DIR allows for voxel-by-voxel mapping of anatomical correspondence between two patients. The resulting dose distribution is “proportionally” equivalent to the original plan but linked to the RTOG defined contours for direct comparison. DVHs were collected to determine the dose-volume statistics for target structures. Contour-based treatment plan: Treatment was planned with goals of 95% coverage of RTOG defined target volumes to 45 Gy, referencing the publication by Li et al, Int. J. Radiat. Onc. Biol. Phys, V73 (2009). Planning was on same patient dataset used for dose mapping described above. DVH analysis for RTOG defined target volumes based on the clinically-derived treatment plan showed median V45 as follows: CW 75% (range, 55-87), SCV 89% (68-97), Ax3 98% (75-100), and IM 74% (20-95). The DVH for the contour-based plan showed V45 values as follows: CW 95.6%, Ax3 100%, SCV 97.7%, and IM 100%. The contoured, but untreated, volume of CW in the clinically-derived plan was largely in the posterior and deep border of the caudal aspect of the field. For SCV, the contoured, but untreated, portion in the clinically-derived plan was in the medial and deep border of the field. Normal tissue doses in clinically-derived treatment were: median V20 for IL 38.5 (29-47%), median V25 for H 1.5 (0-6%). Normal tissue doses in the contour-based plan were: V20 for IL 51.1%, the V25 for H 4.7%. When compared to the existing practice, the RTOG contour-based plan delivered greater coverage to the contoured targets at the expense of increased normal tissue doses. Clinicians should be aware of these differences if the RTOG consensus definitions for PMRT are to be adopted in treatment planning.

Purpose: To evaluate the change in parotid gland density over the course of treatment for patients who have undergone adaptive radiation therapy to the head‐and‐neck. Method and Materials: Computed tomography (CT) images from sixteen patients were analyzed to assess the change in mean parotid density and parotid volume over the course of treatment. Daily CT images were acquired for a group of patients treated with adaptive radiation therapy using a CT‐on‐rails unit. A total of 529 CT scans (∼33 fractions per patient) were analyzed. Parotid contours were deformed from the planning CT to each treatment day CT image using an in‐house deformable image registration tool. The parotid volume, and average and standard deviation in CT number (HU) were obtained for each treatment fraction. Linear regression analysis was performed with a 95% confidence interval to determine the rate and trend of parotid density change. The Pearson correlation coefficient was used to evaluate possible correlations with parotid volume and function measurements. Results: Eleven of the sixteen patients showed a steady decrease in density for both parotids over the course of treatment. Two patients showed a steady decrease only in their ipsilateral parotid. The linear regression analysis for this subset of patients (p‐value <0.01) revealed an average rate of decrease of 0.30 HU/fraction (range 0.13–0.70 HU/fraction). The density reduction correlated well with parotid volume change (in 24/32 instances), and was moderately correlated with patient follow‐up saliva‐flow measurements (9 patients, correlation coefficient range 0.27–0.63) for the first two follow‐up appointments. Conclusion: The mean parotid density decreased steadily and correlated well with the volume shrinkage in most patients observed in this study. In the limited available data, the density change also correlated with saliva‐flow, which warrants future studies with a larger patient population and additional treatment parameters.

Purpose/Objective(s)Increasingly, conformal radiation treatment fields are being used in treatment of cervical cancer (CCa), making accurate delineation of regions at risk critical. Currently, the nodal basins at risk in CCa are defined by the location of the normal pelvic and paraaortic lymph nodes (LNs) as well as surgical data defining the distribution of LN metastasis. To define the spatial distribution of LN metastasis on CT imaging, we generated a map of 2-deoxy-2-[F-18] fluoro-d-glucose (FDG) avid LNs on PET/CT scans in patients with CCa.Materials/MethodsWe identified 50 patients with stage I-IV (I-16, II-20, III-11, IV-3) CCa treated in our department with definitive chemoradiation from 2006-2009 who had FDG avid LNs on pretreatment PET/CT. Forty-one patients were consecutively identified. Nine additional patients with positive paraaortic LNs were added to the analysis. PET/CT images were transferred to treatment planning system and involved LNs were contoured on individual scans. The nodal contours were mapped to a single CT image set using deformable image registration. The mapped contours were edited as necessary by a diagnostic radiologist and radiation oncologist specializing in gynecologic malignancies. A volume overlapping probability map was created to indicate the spatial likelihood distribution of positive LNs throughout the pelvic and paraaortic lymphatics. Percentage of positive LNs along major blood vessels (defined as within a 3 cm radial margin of vessel) and in presacral and perirectal regions were collected for the initial 41 patients.ResultsWe identified 190 PET positive LNs from 50 patients (122 LNs from initial 41 patients, 68 from the additional 9 patients). Deformable image registration was found to be reasonably accurate to map LNs to a template patient, in spite of drastically different body mass indices. The highest density of nodes was found along bilateral external iliac (EI) vessels, followed by the left common iliac (CI) vessels. A lower density of nodes was identified along presacral (4), perirectal (2), and medial inguinal (2) regions outside of conventionally described nodal drainage regions. The highest density of CI lymph nodes was between the CI vein and psoas muscle, extending posteriorly between the muscle and sacrum. The distribution for initial 122 LNs was as follows: left (L) and right (R), EI 43 (35%) and 38 (31%), respectively; L and R internal iliac 3 (2.5%), and 3 (2.5%); L and R CI 18 (15%,) and 3 (2.5%); presacral 2 (1.5%); perirectal 3 (2.5%); and paraaortic 9 (7%).ConclusionsOur study is the first to provide probabilistic distribution of involved LNs in CCa. This analysis provides a template to define LNs at risk, which can be used to improve target definition for conformal therapy. We identified regions anterior and posterior to the psoas muscle, which require careful attention during treatment planning. Purpose/Objective(s)Increasingly, conformal radiation treatment fields are being used in treatment of cervical cancer (CCa), making accurate delineation of regions at risk critical. Currently, the nodal basins at risk in CCa are defined by the location of the normal pelvic and paraaortic lymph nodes (LNs) as well as surgical data defining the distribution of LN metastasis. To define the spatial distribution of LN metastasis on CT imaging, we generated a map of 2-deoxy-2-[F-18] fluoro-d-glucose (FDG) avid LNs on PET/CT scans in patients with CCa. Increasingly, conformal radiation treatment fields are being used in treatment of cervical cancer (CCa), making accurate delineation of regions at risk critical. Currently, the nodal basins at risk in CCa are defined by the location of the normal pelvic and paraaortic lymph nodes (LNs) as well as surgical data defining the distribution of LN metastasis. To define the spatial distribution of LN metastasis on CT imaging, we generated a map of 2-deoxy-2-[F-18] fluoro-d-glucose (FDG) avid LNs on PET/CT scans in patients with CCa. Materials/MethodsWe identified 50 patients with stage I-IV (I-16, II-20, III-11, IV-3) CCa treated in our department with definitive chemoradiation from 2006-2009 who had FDG avid LNs on pretreatment PET/CT. Forty-one patients were consecutively identified. Nine additional patients with positive paraaortic LNs were added to the analysis. PET/CT images were transferred to treatment planning system and involved LNs were contoured on individual scans. The nodal contours were mapped to a single CT image set using deformable image registration. The mapped contours were edited as necessary by a diagnostic radiologist and radiation oncologist specializing in gynecologic malignancies. A volume overlapping probability map was created to indicate the spatial likelihood distribution of positive LNs throughout the pelvic and paraaortic lymphatics. Percentage of positive LNs along major blood vessels (defined as within a 3 cm radial margin of vessel) and in presacral and perirectal regions were collected for the initial 41 patients. We identified 50 patients with stage I-IV (I-16, II-20, III-11, IV-3) CCa treated in our department with definitive chemoradiation from 2006-2009 who had FDG avid LNs on pretreatment PET/CT. Forty-one patients were consecutively identified. Nine additional patients with positive paraaortic LNs were added to the analysis. PET/CT images were transferred to treatment planning system and involved LNs were contoured on individual scans. The nodal contours were mapped to a single CT image set using deformable image registration. The mapped contours were edited as necessary by a diagnostic radiologist and radiation oncologist specializing in gynecologic malignancies. A volume overlapping probability map was created to indicate the spatial likelihood distribution of positive LNs throughout the pelvic and paraaortic lymphatics. Percentage of positive LNs along major blood vessels (defined as within a 3 cm radial margin of vessel) and in presacral and perirectal regions were collected for the initial 41 patients. ResultsWe identified 190 PET positive LNs from 50 patients (122 LNs from initial 41 patients, 68 from the additional 9 patients). Deformable image registration was found to be reasonably accurate to map LNs to a template patient, in spite of drastically different body mass indices. The highest density of nodes was found along bilateral external iliac (EI) vessels, followed by the left common iliac (CI) vessels. A lower density of nodes was identified along presacral (4), perirectal (2), and medial inguinal (2) regions outside of conventionally described nodal drainage regions. The highest density of CI lymph nodes was between the CI vein and psoas muscle, extending posteriorly between the muscle and sacrum. The distribution for initial 122 LNs was as follows: left (L) and right (R), EI 43 (35%) and 38 (31%), respectively; L and R internal iliac 3 (2.5%), and 3 (2.5%); L and R CI 18 (15%,) and 3 (2.5%); presacral 2 (1.5%); perirectal 3 (2.5%); and paraaortic 9 (7%). We identified 190 PET positive LNs from 50 patients (122 LNs from initial 41 patients, 68 from the additional 9 patients). Deformable image registration was found to be reasonably accurate to map LNs to a template patient, in spite of drastically different body mass indices. The highest density of nodes was found along bilateral external iliac (EI) vessels, followed by the left common iliac (CI) vessels. A lower density of nodes was identified along presacral (4), perirectal (2), and medial inguinal (2) regions outside of conventionally described nodal drainage regions. The highest density of CI lymph nodes was between the CI vein and psoas muscle, extending posteriorly between the muscle and sacrum. The distribution for initial 122 LNs was as follows: left (L) and right (R), EI 43 (35%) and 38 (31%), respectively; L and R internal iliac 3 (2.5%), and 3 (2.5%); L and R CI 18 (15%,) and 3 (2.5%); presacral 2 (1.5%); perirectal 3 (2.5%); and paraaortic 9 (7%). ConclusionsOur study is the first to provide probabilistic distribution of involved LNs in CCa. This analysis provides a template to define LNs at risk, which can be used to improve target definition for conformal therapy. We identified regions anterior and posterior to the psoas muscle, which require careful attention during treatment planning. Our study is the first to provide probabilistic distribution of involved LNs in CCa. This analysis provides a template to define LNs at risk, which can be used to improve target definition for conformal therapy. We identified regions anterior and posterior to the psoas muscle, which require careful attention during treatment planning.

Characterization of nonperiodic head and neck soft tissue movement would assist the development of nonuniform PTV expansion or management of intrafractional soft tissue motion during treatment delivery. Since a simulation CT provides only a snapshot of baseline tissue anatomy and is prone to internal positional errors caused by swallowing, we intend to assess daily positional changes of the thyroid cartilage (as a landmark for larynx) relative to hyoid bone (landmark for lower oropharynx) in oropharyngeal patients prospectively enrolled onto an institutional image-guided adaptive radiotherapy protocol. Seventeen patients were enrolled and received daily in-room imaging with a CT-on-Rails unit. The hyoid bone, thyroid cartilage, and C2 vertebral body were contoured onto the planning CT. Using 3D rigid image registration techniques, daily positional variations of the hyoid bone and thyroid cartilage were determined relative to C2. Daily separation between the hyoid and thyroid cartilage was tabulated. Both systematic and random variations were analyzed as a group and for each patient. To identify time trends, positional changes were analyzed fraction-by-fraction using a linear regression method. A total of 555 daily CT image sets were analyzed (approximately 33 CTs/patient). Hyoid and thyroid cartilage moved significantly (up to 16 mm) relative to C2, most notably in the superior-inferior (SI) direction. Systematic errors relative to baseline position from simulation for individual patients were large (up to 12 mm), with a SD for SI error 4.5 mm. Median absolute random shifts (SD) for hyoid were 3.3 (2.9) mm, 1.4 (1.4) mm, and 1.3 (1.1) mm in the SI, anterior-posterior, and left-right directions, respectively. A statistical trend was identified (consisting typically of superior directional shifts during the later portions of treatment) in 13/17 hyoid and 11/17 thyroid structures (ANOVA; p = 0.05), potentially due to tumor response or acute soft tissue treatment effects. Thyroid cartilage was highly correlated with hyoid bone movement (Spearman's correlation = 0.839; p < 0.001) with similar systematic, median shifts, and random variations. However, in 15% of treatment fractions (1 in 7), separation between the hyoid and thyroid changed more than 4 mm. Such separation changes typically resulted from an active swallow. This is the first study to document relative positional shifts of oropharyngeal (hyoid bone) and laryngeal (thyroid cartilage) landmarks during oropharyngeal IMRT. Active swallow can occur in 1/7 CT scans and may be incidentally captured by a significant proportion of simulations, distorting oropharyngeal, and laryngeal positional relationships, and producing a large systematic setup error during the course of treatment.