- Email: [email protected]
Adaptive Planning in Intensity-Modulated Radiation Therapy for Head and Neck Cancers: Single-Institution Experience and Clinical Implications
Int. J. Radiation Oncology Biol. Phys., Vol. 80, No. 3, pp. 677–685, 2011 Copyright Ó 2011 Elsevier Inc. Printed in the USA. All rights reserved 0360-3016/$–see front matter
doi:10.1016/j.ijrobp.2010.03.014
CLINICAL INVESTIGATION
Head and Neck
ADAPTIVE PLANNING IN INTENSITY-MODULATED RADIATION THERAPY FOR HEAD AND NECK CANCERS: SINGLE-INSTITUTION EXPERIENCE AND CLINICAL IMPLICATIONS PETER H. AHN, M.D.,* CHIN-CHENG CHEN, PH.D.,* ANDREW I. AHN, M.ENG.,y LINDA HONG, PH.D.,* PAOLA G. SCRIPES, M.S,* JIN SHEN, B.S.,* CHEN-CHIAO LEE, M.S.,* EKENI MILLER, A.S.,* SHALOM KALNICKI, M.D.,* AND MADHUR K. GARG, M.D.* )
Department of Radiation Oncology, Montefiore Medical Center and Albert Einstein College of Medicine of Yeshiva University, Bronx, New York and yAlbert Einstein College of Medicine of Yeshiva University, Bronx, New York Purpose: Anatomic changes and positional variability during intensity-modulated radiation therapy (IMRT) for head and neck cancer can lead to clinically significant dosimetric changes. We report our single-institution experience using an adaptive protocol and correlate these changes with anatomic and positional changes during treatment. Methods and Materials: Twenty-three sequential head and neck IMRT patients underwent serial computed tomography (CT) scans during their radiation course. After undergoing the planning CT scan, patients underwent planned rescans at 11, 22, and 33 fractions; a total of 89 scans with 129 unique CT plan combinations were thus analyzed. Positional variability and anatomic changes during treatment were correlated with changes in dosimetric parameters to target and avoidance structures between planning CT and subsequent scans. Results: A total of 15/23 patients (65%) benefited from adaptive planning, either due to inadequate dose to gross disease or to increased dose to organs at risk. Significant differences in primary and nodal targets (planning target volume, gross tumor volume, and clinical tumor volume), parotid, and spinal cord dosimetric parameters were noted throughout the treatment. Correlations were established between these dosimetric changes and weight loss, fraction number, multiple skin separations, and change in position of the skull, mandible, and cervical spine. Conclusions: Variations in patient positioning and anatomy changes during IMRT for head and neck cancer can affect dosimetric parameters and have wide-ranging clinical implications. The interplay between random positional variability and gradual anatomic changes requires careful clinical monitoring and frequent use of CT- based image-guided radiation therapy, which should determine variations necessitating new plans. Ó 2011 Elsevier Inc. Radiotherapy, Head-and-neck tumor, Positional change, Anatomic change, Dosimetry.
Precise radiotherapy delivery depends upon the predictability of planning parameters: patient external contour and skin separation, internal target motion, tumor size, and patient position. A change in any of these parameters may change the delivered dose to tumor, nodal areas at risk, and normal structures in ways that require further investigation (6). We report our results of a prospective, adaptive head and neck image-guided radiation therapy (IGRT) program for patients who underwent scheduled rescanning and were replanned if deemed necessary based on comparative dosimetric analyses. Changes in patient anatomy (weight loss and tumor shrinkage), as well as positioning variability, were correlated with dosimetric endpoints. We previously
INTRODUCTION The utilization of intensity-modulated radiation therapy (IMRT) in the treatment of head and neck cancer has increased dramatically in the last decade (1), driven by studies that demonstrate that controlling the dose to parotids (2), superior pharyngeal constrictors (3), cochlea (4), and other structures affect quality of life. IMRT’s dose conformality, which permits organ sparing, also allows for dose escalation to gross tumor, selectively using areas of radioresistant hypoxia as the target (5). The hallmark of IMRT plans is the creation of sharp dose gradients with high target conformality; IMRT and related technologies represent a major advance in improving the therapeutic ratio of head and neck radiotherapy. Reprint requests to: Madhur K. Garg, M.D., Department of Radiation Oncology and Otolaryngology, Montefiore Medical Center/Albert Einstein College of Medicine, 111 East 210th St., Bronx, NY 10467. Tel: (718) 920-4140; Fax: (718) 231-5064; E-mail: [email protected]
This work was presented in part at the 7th International Conference on Head and Neck Cancer, San Francisco, CA, July 2008. Conflict of interest: Dr. Ahn’s spouse is a shareholder in Pyronia Medical Technologies. Received Dec 14, 2009, and in revised form Feb 10, 2010. Accepted for publication March 12, 2010. 677
678
I. J. Radiation Oncology d Biology d Physics
reported results with the same patient set, quantifying the degree of variability in various portions of the head and neck, noting that the 95% confidence interval of uncertainty in the region of the mandible and lower neck is upwards of 7 mm in an unpredictable manner (7). METHODS AND MATERIALS Twenty-three consecutive patients with locoregionally advanced head and neck cancer underwent a course of intensity-modulated IGRT; 33 fractions with simultaneous integrated boost (SIB) differential dose painting, at Montefiore Medical Center, Bronx, NY. Patients were immobilized with a custom-fitted thermoplastic face mask (Bionix, Toledo, OH) and shoulder pulls (WFR/Aquaplast, Wyckoff, NJ). Computed tomography (CT) simulation was then performed with 2.5-mm-thick slices (GE Lightspeed; GE Healthcare, Chalfont St. Giles UK), with planned rescans at fraction numbers of 11, 22, and 33. A positron emission tomography (PET) CT scan in the treatment position was obtained from all patients and fused with the planning CT scan. A single radiation oncologist planned treatment for all 23 patients; volumes were outlined utilizing CT and PET-CT findings, as well as findings on endoscopic and physical examination at the time of initial consult. Planning was performed on an Eclipse planning system (Varian Medical Systems, Palo Alto, CA). The IMRT plans utilized seven fields, encompassing the entire neck, so no anterior lower-neck split field was used. In 22 patients, gross tumor volume (GTV) or high-risk areas received 66 to 69.96 Gy in 33 fractions with a 1-cm margin planning target volume (PTV-primary; uninvolved nodal areas at risk were treated to 54.12 Gy in 33 fractions, with a margin of 0.5-cm (PTV-nodal). The remaining patient was treated to a dose of 50.4 Gy in 1.8 Gy fractions to PTV-nodal and PTV-primary, followed by a boost to a total of 72 Gy in 1.8-Gy fractions to PTV-primary. For normal tissue constraints, the dose to a 5-mm expansion around the spinal cord was <45 Gy, the maximum brainstem dose was 54 Gy to 20%, the median dose to one parotid was <26 Gy, and 95% of the PTV-nodal and PTV-tumor volumes received 95% of the prescription dose. All patients were rescanned on the original CT simulator, for a total of 89 individual CT scans. Radioopaque fiducial markers were placed on the patient’s tattoos at the time of rescan. The patient’s treatment was divided into quartiles for purposes of comparison; S1 for original planning CT; S2 for rescan at around fraction no. 11 (range, 1–19 fractions); S3 for rescan at fraction no. 22 (range, 20–30 scans); and S4 for rescan at fraction no. 33 (range, 31–33 fractions). Avoidance and target structures from the original planning scan, including the spinal cord, brainstem, mandible, parotids, CTV (including individual right, left and retropharyngeal lymphatic contours as applicable), GTV and PTV were reconstituted on the rescan images; the contours were then modified by the original radiation oncologist to account for any changes in patient anatomy and positioning from the original pretreatment plan. GTV was changed only if there was clear anatomical displacement; at the primary site, this occurred most commonly with tumor shrinkage leading to a portion of the rescan GTV clearly overlapping with air in the laryngeal or pharyngeal space; in nodal volumes, tumor shrinkage outside the skin contour on the rescan CT required modification of the nodal GTV. The original field arrangement with which the patient was being treated was then fused onto the rescan CT based on carefully matched isocenters, soft tissue, and bony anatomy. The dose distribution was recalculated and dose–volume histograms were generated for the structures recontoured on the rescan CT. The pattern
Volume 80, Number 3, 2011
of dose distribution was also examined on axial slices. After analysis of dose distribution, patients were replanned if constraints to spinal cord and other normal structures were not met or there was inadequate coverage of the target structures. Subsequent CT scans during treatment had both the original field arrangement and any replanned field arrangements reconstituted on the images and examined for the dosimetric consequences, yielding 129 unique CT-field combinations and dosimetric plans. After patients completed treatment, patient anatomical parameters such as skin separation (at isocenter) and shifts in the isocenter, defined by radioopaque fiducial markers versus bony anatomy, and individual patient positioning parameters were determined on the original and rescan CTs.
Statistical analysis Statistical analysis was performed using multiple linear regression analysis, Pearson correlation calculations, and principal component’s multivariate analysis using Minitab statistical software. Strong correlation was considered to consist of R2-adjusted values of 0.7 to 1.0, while moderate correlation was considered to consist of R2-adjusted values of 0.4 to 0.7. Differences in dosimetric outcome for the population over time were calculated using a oneway unstacked analysis of variance comparison, and multivariate discriminant function analysis.
RESULTS The characteristics of the studied patients are outlined in Table 1. A majority of the 23 patients had locoregionally advanced squamous cell carcinoma, divided largely between laryngeal/hypopharyngeal primary and oropharyngeal/oral cavity primary. Nineteen patients received cisplatin or carboplatin-based concurrent chemoradiotherapy, while 2 patients received concurrent cetuximab therapy. Changes in patient anatomy The average weight loss over the entire course of treatment was13 pounds (8.3%), with the most significant change occurring between scans S2 and S4. Patients did not have any significant change in skin separation between scans S1 and S2; there were significant changes in skin contour and separation between S1 and S3 to S4, with an average decrease in skin separation at the level of the isocenter of 10% by the end of treatment (Table 2). The average decrease in the size of GTV (primary and nodal) was 17.2%. Parotid volume decreased progressively by 24% during early treatment (scans S1–S2), stabilizing around scan S3 (Table 2). As would be expected with progressing treatment, lost weight correlated moderately to strongly with reduction in skin separation at C1, C4, and C7 and mid-neck level (Fig. 1). Dosimetric findings There is a progressive increase in dose inhomogeneity as one proceeds from S1 to S4. PTV coverage (both nodal and primary) was in S2 but improved in S3 and S4. With reduction in skin separation at mid-tumor level, the hot spot increased. Hot spot and PTV coverage also correlated both with anatomic and positional variations. In 14 of 23 patients (61%), PTV dose homogeneity and coverage were improved by replanning at S2 or S3.
Adoptive planning in IMRT for head and neck cancer d P. H. AHN et al.
Table 1. Patient characteristics by histology, primary site, operative status, and local and nodal stage Histology Primary site
Operative status Stage
Local stage
Nodal stage
Squamous cell
Patients (n = 22)
Adenoid cystic Nasopharynx Oral cavity Oropharynx Hypopharynx Larynx Unknown primary Definitive Postoperative 1 2 3 4a 4b T1 T2 T3 T4 N0 N1 N2a N2b N2c N3
1 3 6 5 2 6 1 19 4 1 2 1 17 2 5 8 4 6 5 1 5 5 5 2
Spinal cord dose initially increased in S2 but came down as treatment progressed. In 7/23 patients (30%), the maximum cord point dose exceeded our 45-Gy constraint at either S2 or S3 and needed correction. This increase correlated moderately with positional variability but did not correlate with anatomic changes. Parotid dose increased as treatment progressed, often but not always in accordance with a decrease in parotid volume (Table 2). This increase correlated with both positional and anatomic changes. In 5/23 patients (22%), the replan yielded improved parotid sparing. There was no significant difference in dose to the mandible or to the brainstem in our population. Correlation between dosimetric parameters and anatomic changes Weight loss and fraction number moderately correlated with skin separation, parotid volume, at various levels in the neck, and at the midpoint of involved neck nodes (Table 3). Coverage of GTV and PTV improved with decreases in skin separation. Dose to the spinal cord was affected adversely by decreased skin separation throughout the neck, which also occurred with shrinkage of tumor in the neck. Parotid dose increased with decreases in skin separation at the upper-mid neck or with shrinkage in parotid volume. Dose to the mandible was adversely affected as GTV volume decreased. Predicting dosimetric parameters based on positional variations Positional variation adversely affected dosimetric values (Table 4), especially cord dose, GTV, CTV, and PTV cover-
679
age. Variation in position of the cochlea (as a proxy for skull position) adversely affected dose to the parotid and led to changes in dose to GTV and PTV. Variation in position of elements of the cervical spine including lordosis and scoliosis influenced dose to the spinal cord, GTV, CTV, and PTV coverage, with changes in lordosis adversely affecting coverage of the retropharyngeal nodes. Positional variation of the mandible adversely influenced dose to the cord, parotids, CTV, and nodal PTV. As demonstrated by the correlation coefficients, most associations were weak to moderate without any single dominating factor. Overall, 15 out of 23 patients (65%) benefited from improved dose distributions second to replanning. The reasons varied from exceeding spinal cord, brainstem or parotid dose constraints, or inadequate (better dose homogeneity) PTV coverage (Fig. 2). External predictors of replan necessity We determined whether externally measurable parameters including anatomic changes in skin separation and/or positional variation could predict for the need for replanning patients for either overdose of normal avoidance structures or underdose of target structures. Need for replan was usually based on dosimetric parameters to spinal cord (maximal dose [Dmax] >45 Gy), parotid (D50% >26 Gy in one side); GTV-V95% of <95%; CTV-V95% of <95%; and PTVD95% of < 95% (Fig. 3). Comparisons also included correlations with weight loss, changes in skin separation at isocenter, and anatomic isocenter (Table 5). No single positional or anatomic variable predicted for the need for a replan; instead, multiple independent and statistically significant trends were evident for various normal and target structures (Table 6). Replan due to higher dose to spinal cord was driven largely by positional variation, including the cervical spine but also the cochlea and mandible. Higher dose to the parotids was driven by variations in isocenter, mandible position, and decreased skin separation. Underdose to GTV was driven by variations in isocenter, skull, mandible, and spine positions, and changes in separation in the lower spine. Decreases in skin separation throughout the neck and positional variation in isocenter, skull, and mandible and mid-neck predicted for the need to replan due to underdose to CTV and PTV. Large changes in the hotspot were due primarily to positional variation and decreases in skin separation with tumor shrinkage and weight loss. Clinically significant overdose to the mandible was primarily due to positional factors. DISCUSSION Several studies have demonstrated variability in patient positioning for head and neck radiotherapy. Zhang et al. (8) demonstrated somewhat independent movement between the maxilla and the lower cervical spine with respect to the variable of rotation of the skull with respect to the lower neck. Kaiser et al. (9), assuming rigidity between the skull and neck, demonstrated small variations in patient position
680
I. J. Radiation Oncology d Biology d Physics
Volume 80, Number 3, 2011
Table 2. Dosimetric parameters and skin separation with treatment progression Dosimetric endpoint PTV (nodal) D99% D95% V95% PTV (tumor) D99% D95% V95% CTV (nodal) D100% V100% V95% D95% GTV D100% V100% D95% Right parotid Dmean D50% D10% V26Gy V20Gy Volume Left parotid Dmean D50% D10% V26Gy V20Gy Volume Spinal cord Dmax D1cc Brainstem Dmax D20% Mandible V60 Separation @ isocenter
S1
S2
S3
S4
p value
49.19 Gy 53.08 Gy 97.48%
43.62 Gy 50.73 Gy 93.59%
44.90 Gy 51.64 Gy 94.9%
48.05 Gy 52.92 Gy 96.41%
<0.01 <0.01 <0.01
62.73 Gy 65.54 Gy 96.28%
58.51 Gy 63.82 Gy 91.93%
58.82 Gy 63.94 Gy 93.60%
60.55 Gy 64.68 Gy 94.21%
NS NS NS
40.14 Gy 92.25% 98.54% 54.09 Gy
36.10 Gy 90.87% 97.20% 53.20 Gy
37.18 Gy 91.56% 97.66% 53.69 Gy
36.45 Gy 93.78% 98.21% 53.62 Gy
<0.053 NS <0.05 NS
57.11 Gy 90.31% 69.33 Gy
57.85 Gy 85.80% 68.26 Gy
55.48 Gy 86.48% 66.74 Gy
55.04 Gy 90.00% 67.45 Gy
NS NS NS
30.81 Gy 28.92 Gy 52.94 Gy 49.61% 65.82% 26.98 cm3
33.02 Gy 30.13 Gy 54.46 Gy 55.08% 72.57% 24.67 cm3
32.63 Gy 30.86 Gy 52.90 Gy 52.62% 68.80% 20.16 cm3
33.39 Gy 30.63 Gy 56.78 Gy 57.44% 73.26% 20.61 cm3
<0.02 <0.05 <0.05 <0.02 <0.006 <0.001
28.55 Gy 24.97 Gy 49.29 Gy 44.71% 60.93% 24.9 cm3
31.92 Gy 27.81 Gy 50.54 Gy 49.39% 66.23% 22.26 cm3
29.37 Gy 26.41 Gy 49.38 Gy 48.62% 64.26% 18.29 cm3
31.03 Gy 28.61 Gy 50.32 Gy 51.52% 67.47% 18.09 cm3
<0.03 <0.04 NS <0.05 <0.04 <0.05
39.40 Gy 36.55 Gy
42.76 Gy 39.16 Gy
41.60 Gy 37.59 Gy
40.73 Gy 37.02 Gy
<0.02 <0.03
42.55 Gy 28.92 Gy
44.31 Gy 28.47 Gy
45.63 Gy 30.7 Gy
41.06 Gy 26.33 Gy
NS NS
14.35% 11.43 cm
14.18% 11.11 cm
14.68% 10.68 cm
11.05% 10.01 cm
NS <0.001
Abbreviations: S1 = initial planning CT Scan; S2 = fraction 1-19 (primarily fraction 11); S3 = fraction 20-30 (primarily fraction 22); S4 = fraction 31-33 (primarily fraction 33); NS = not statistically significant; Dxx% = dose to xx% of the target or organ volume; Vxx% = % of the target or organ volume receiving xx% of target dose; CTV = clinical target volume; GTV = gross tumor volume; PTV = planning target volume. Dosimetric endpoints for PTV, CTV, GTV, parotid glands, spinal cord, brainstem, mandible, as well as changes in horizontal separation in skin with treatment from planning CT scan (S1), around fraction 11 (S2), around fraction 22 (S3), to completion of treatment around fraction 33 (S4). In this set of patients, there is significant dosimetric variability in parotid and spinal cord dose as treatment proceeds, with differences in CTV and PTV coverage. Changes in GTV coverage and dose to the mandible and brainstem are not statistically significant in this cohort. All data are with the original fields transposed on rescan CT, and prior to any replan.
with respect to roll, pitch, and yaw that were unchanged between weeks 1 and 4 of fractionated radiotherapy. Hong et al. (10) reported small systematic but large random variations in rotational and translational position, as well as vector displacement. Using theoretical IMRT plans, an adverse and unpredictable variation in dose deposition when taking into account these movements was demonstrated, with instances of geographic miss and normal tissue overdosing (10). Barker et al. (11) found that in patients with nodal disease greater than 4 cm, the tumor in the neck decreased by 69.5%, with medial displacement of the center of mass of the parotid in a manner that was highly correlative
with patient weight loss. Medial shift in the center of mass of the tumor was also noted. Barker et al. (11) retrospectively examined dosimetric aspects of anatomic deformation in a retrospective study of patients selected for replanning at the midpoint of chemoradiotherapy. In those patients selected due to significant weight loss or tumor shrinkage, replanning led to improvement in coverage of the PTV, with decreased dose to the spinal cord, brainstem, and right parotid compared to that with no replanning. In a repeat CT scan study of 8 patients, Ballivy et al. (12) examined weekly CT scans from head and neck patients and found higher doses to the contralateral parotid and spinal cord than originally planned, with
Adoptive planning in IMRT for head and neck cancer d P. H. AHN et al.
681
Fig. 1. In this patient with an unknown primary squamous cell carcinoma, the 45-Gy isodose line approaches a 5-mm avoidance structure around the spinal cord on the original plan (a). At the 24th fraction, there have been significant changes in skin separation due to weight loss and tumor shrinkage, leading to the 45-Gy isodose line encroaching on the spinal cord (b). The patient was replanned in order to avoid excess dose to the spinal cord (c).
dose to the spinal cord greater than 45 Gy in 57% of repeat CT scans. Robar et al. (13) also found that positional uncertainty was greatest in the lower neck and that while mean maximal dose variation to craniospinal structures was generally on the order of 2.2%, the range in maximal dose could reach 39% in the lower neck. O’Daniel et al. (14) examined the effect of anatomic change with radiotherapy and positioning variability at C2 on dose to the parotid gland, and found that almost half of the patients had parotid gland doses 5 to 7 Gy above that expected from the planning CT. There was a contribution from both inconsistent patient immobilization and anatomic changes as treatment proceeded (14). That study did not find any detriment to CTV coverage with anatomic change and variations in patient positioning. Another study looked at the dosimetric effect of tumor shrinkage on target and organs at risk (OARs), using a deformation algorithm. Those authors found that although the dose to target did not change significantly, the parotid dose increases by
Table 3. Statistically significant correlations between anatomic parameters, fraction number, and weight loss External parameter
Anatomic correlation
Weight loss
C1 lateral separation C4 lateral separation Parotid shrinkage
Fraction number
C1 lateral separation C4 lateral separation C7 lateral separation Separation at midneck tumor GTV volume
Pearson correlation coefficient 0.44 0.49 0.52 (right), 0.67 (left) 0.71 0.69 0.40 0.75 0.67
Abbreviation: GTV = gross tumor volume External parameters such as progressive weight loss and fraction number have the highest correlation with parotid and GTV shrinkage, respectively.
I. J. Radiation Oncology d Biology d Physics
682
Volume 80, Number 3, 2011
Table 4. Correlations between dosimetric parameters and positional variation
Dosimetric parameter Cord Dmax increase
Parotid D50% increase GTV-D95% decrease
CTV-D100% decrease CTV-V95% decrease RP-CTV-D100% decrease RP-CTV-V100% decrease PTV nodal - V95% decrease PTV tumor - V95% decrease
Positional correlation Anatomic isocenter vector increase Lordosis/Scoliosis change C1-C7 vector increase Mandible vector increase Cochlea vector increase
Pearson correlation coefficient 0.43 0.37–0.43 0.27–0.49; 0.27 0.27 0.41
Mandible vector increase Anatomic isocenter vector increase C1-C7 vector increase Cochlea vector; incisive vector increase Anatomic isocenter vector increase Mandible vector increase C2-C7 vector increase C2-C7 vector increase Anatomic isocenter vector increase Lordosis increase
0.42 0.32
Mandible vector increase
0.39
Cochlea vector; C1-C7 vector increase Anatomic isocenter vector increase Cochlea vector; C1-C7 vector increase
0.31–0.43; 0.24–0.44 0.30
0.31–0.37 0.29; 0.37 0.33 0.22 0.38–0.44 .27–.32 0.19 0.19
0.25–0.38; 0.16–0.31
Abbreviations: Dmax = maximum dose; D50% = dose to 50% of volume; D95% = dose to 95% of volume; D99% = dose to 99% of volume; D100% = dose to 100% of volume; V95% = volume receiving 95% of target dose; V100% = volume receiving 100% of target dose; RP-CTV = retropharyngeal node clinical target volume. There are consistent and statistically significant associations between variation in mandible, cochlea and C1-C7 position and decreased dosimetric coverage of the GTV, nodal CTV and retropharyngeal nodes, as well as increased dose to the spinal cord and parotid glands.
almost 10% during treatment as a result of tumor shrinkage. Dose changes due to positional variability, and its potential interplay with clinical changes such as weight loss was not addressed in the study. The original impetus behind the present study was the generation of an algorithm to predict which external anatomic or clinical variables such as changes in skin separation, fraction number or weight loss would predict for the need for replanning patients. We initially attempted to correlate changes in these anatomic and clinical parameters with decrements in PTV, CTV, and GTV coverage and increases in dose to normal organs that mandated replanning the patient’s treatment. Single anatomic-clinical correlations were not sufficiently robust, but different anatomic elements and positional variation significantly affected dose parameters. The combination of
Fig. 2. Skin separation was measured by taking a horizontal line through the midplane at the bottom of the two foramina of a given C-spine level (C4 in this case). The skin separation measurement was taken from skin to skin.
independent random positioning events and more gradual and predictable anatomy changes drive the need for replanning; in consequence, during adoptive therapy, the decision to replan a patient could hypothetically be driven exclusively by dosimetric findings due to random positional variation during the particular CT session. Conversely, the random positional variation could mask a dosimetric consequence that occurs on other days due to anatomic change as treatment
PTV-GTV=8 cord=7 PTV-nodes=6 parotid=5 mandible =1 brainstem=1
Fig. 3. A total of 15 of 23 patients required a replan at some point, based on dosimetric parameters including Dmax exceeding 45 Gy for cord, PTV-tumor, or PTV-node D95% below 95% of prescription, parotid D50% increased significantly above 26 Gy, mandible V60 Gy above 10%, and brainstem V54 Gy above 20%. Two patients each required two replans each for PTV-GTV underdosage; 2 patients each required two replans each for cord overdosage; 1 patient required two replans for parotid overdosage.
Adoptive planning in IMRT for head and neck cancer d P. H. AHN et al.
Table 5. Correlations between dosimetric and anatomic changes Dosimetric parameter Cord Dmax increase
Parotid D50% increase
Mandible V60 increase GTV-D100% increase GTV-D95% increase CTV PTV nodal – D95% incr PTV tumor – D95% incr
Anatomic variable Mandibular joint separation decrease C4-C6 lateral separation decrease neck midtumor separation decrease Mandibular joint separation decrease C1-C5 lateral separation decrease Parotid volume decrease Weight loss GTV volume decrease C7 lateral separation decrease GTV volume decrease C7 lateral separation decrease None C1-C5 lateral separation decrease PTV nodal volume change decrease C1-C3 lateral separation decrease PTV nodal and tumor volume decrease
Pearson correlation coefficient
Table 6. Need for replan and reason correlated with positional and anatomic changes* Dosimetric parameter Cord Dmax overdose
0.30 0.17–0.27 0.18 0.22–0.39
Parotid D50 overdose
0.17–0.28 GTV underdose 0.22 0.30–0.35 0.26 0.18 0.47 0.23
CTV nodal underdose
0.19–0.25 0.27 0.23–0.24 0.30–0.31
All correlations are statistically significant. Cord, parotid, and mandible doses were increased with decreases in skin separation or volume change of select structures. CTV coverage was not significantly affected by changes in anatomy, although there was increase in PTV nodal-D95% coverage with decreased separation. Coverage of GTV-D100% was improved with decreases in GTV volumes as treatment proceeded.
proceeds and falsely indicate adequate dosimetric coverage or acceptable normal tissue exposure when the anatomy would dictate that a labor-intensive replanning process is indicated. Reduced skin separation, weight loss, and loss in parotid volumes correlated weakly with increase in parotid D50%. Variations in these anatomic or clinical surrogates, however, were nonsignificant and by themselves did not predict for the need for a replan. Dose to normal and target structures are also adversely affected by variations in positioning of the skull, mandible, and elements of the cervical spine. In our study, 15/23 patients would have a better dose distribution on replanning either because of better dose homogeneity to the target or better sparing of OAR. This has implications for not only better tumor control but also better quality of life by reducing dose to OAR and therefore toxicity. This is the first dosimetric study that we are aware of in which patients were prospectively studied and scanned at regular intervals, correlating multiple aspects of positioning
683
PTV gross underdose PTV nodal underdose
Hotspot overdose
Mandible overdose
Need for replan Cochlea/incisive roll, pitch, RL, AP Mandible roll, yaw C3-C7 vector C2-6 pitch C2-C5 AP Lordosis Anatomic isocenter AP
p value 0.024–0.054 0.034 0.002–0.043 0.001–0.005 0.001 0.001 0.002
Mandible vector, AP, SI Skin separation (C2-C3) Anatomic isocenter vector, AP Weight Loss Mid-cochlea/incisive vector Mandible vector, roll C1-C7 midpoint vector Scoliosis Skin separation (C6-C7) Anatomic isocenter AP, RL
0.001–0.006 0.07–0.08 0.001–0.022
Cochlea SI Weight Loss Mandible roll C3 roll C4-5 SI Skin separation (C2-C7) Anatomic isocenter RL
0.009 0.065 0.075 0.033 0.001–0.020 0.001–0.037 0.015
Cochlea midpoint SI Weight loss Cochlea/incisive vector, roll, RL Mandible vector/yaw/roll C2-C4 AP, pitch C3 roll C2-C5 vector Skin separation (C2-C7) Anatomic isocenter vector Incisive foramen vector; cochlea vector/pitch/AP Mandible vector Lordosis C4 AP C1-C7 vector Skin separation at mid-tumor Anatomic isocenter vector, RL Cochlea/incisive AP, RL, roll Mandible vector/yaw/roll/pitch
0.063 0.099 0.001–0.026
0.003 0.002–0.052 0.001–0.064 0.001–0.028 0.035 0.009 0.002–0.008
0.001–0.012 0.003–0.034 0.033 0.002–0.050 0.002–0.017 0.001 0.001–0.043 0.017 0.001 0.021 0.001 0.036 0.001–0.002 0.013–0.039 0.004–0.028
Abbreviations: AP = anterior-posterior; SI = superior-inferior; RL = right-left. * Reasons for patient replan correlated with anatomic and positional variables. Statistically significant correlations exist between positional and anatomic variation, and the need for replan, usually based on dosimetric parameters to spinal cord (Dmax >45Gy), parotid (D50% >26Gy), GTV-V95% <95%, CTV-V95% <95%, and PTV-D95% <95%. Potential externally measurable measures would include weight loss, skin separation at various portions of the head and neck, and fraction number. Eighteen replans were performed.
684
I. J. Radiation Oncology d Biology d Physics
and anatomic variability with dosimetric parameters in an attempt to determine which of these external factors might predict for the need for a rescan and replan. No single positional or anatomic variable predicted for the need for a replan; instead, multiple independent and statistically significant trends were evident for various normal and target structures. Therefore, it is not only important to monitor changes in anatomy but also look at the various positional variability that can happen during IMRT for head and neck cancers. Isocenter and three-point triangulation verification using orthogonal images during treatment may not be adequate for patients undergoing IMRT with tighter margins. Anatomic changes such as skin separation, target, and parotid volumes along with positional alignment using multiple points of interest such as skull position, spine curvature, and mandible position should be carefully monitored during IMRT for head and neck cancers. Weight loss and decreased skin separation (due either to tumor shrinkage or weight loss) during the course of radiation therapy can lead to head movement inside the mask, giving rise to positional variability. In addition, weight loss and reduced skin separation by itself has been associated with higher parotid dose and target dose inhomogeneity. IGRT using CT scans should be used more frequently for these patients to assess anatomic and positional variability. Although resolution of presently available CT-based OBI is suboptimal, it can be used as a preliminary tool to assess a patient for resimulation and a replan. In our clinic, we perform planned rescans in the midcourse of treatment regardless of weight loss or tumor shrinkage on the diagnostic-quality CT scanner, and volumes are reconstituted on the new images, using deformable registration and manual correction. Original treatment fields are then reconstituted on this new CT scan with the deformed volumes, and dose is calculated. The fidelity of nonrigid patient positioning between the original CT and the rescan CT is examined, since spurious dosimetric findings can be a result of a random positional event specific to that day. Examining the resulting dosimetry, we do not replan if dose to the subclinical PTV is compromised because we often find that dose to subclinical CTV is improved, but we will replan if there is excess dose to normal structures, including the mandible (with significant GTV shrinkage). Any underdose of GTV will prompt a replan. If there is
Volume 80, Number 3, 2011
excess dose to the spinal cord, and variation in positioning of the spinal canal between the two CT scans is noted, we will examine the cone beam images to determine if any variation in cord position is consistent (i.e., systematic) or random. If there is indeed a systematic variation in cord position, we will replan the patient’s treatment. Similarly, if there is no variation in translational or rotational position of the skull or upper neck and there is an increase in parotid dose, we will replan the patient. We recommend that patients receiving IMRT for head and neck cancers undergo a rescan with a new mask if either weight loss or tumor shrinkage results in the mask being loose. They should also undergo CT-based IGRT in addition to three-point triangulation and orthogonal kV images every day for positional accuracy. Facilities using head and neck IMRT without CT-based IGRT capabilities should carefully monitor patients for weight loss and decreased skin separation due to tumor shrinkage. In addition, translational and rotational set up variability should be looked at carefully in addition to verifying the isocenter position on orthogonal images. Additional margins should also be used around OARs to constrain dose to these structures for optimization. CONCLUSIONS Both positional and anatomic variability can have wideranging effect on dosimetric outcome in patients undergoing IMRT for head and neck cancers. No single external variable such as weight loss, fraction number, or changes in skin separation in isolation can reliably predict for the need to initiate the resource-intensive replanning process, due to the confounding effect of positional variability. Isocenter and three-point triangulation verification using orthogonal images, during treatment, may not be adequate for patients undergoing IMRT with tighter margins. In addition to carefully monitoring clinical/anatomic parameters such as weight loss and skin separation, IGRT using CT scans should be used more frequently in these patients to assess both anatomic and positional variability. Research that correlates dose actually received by organs with clinical outcomes such as xerostomia, dysphagia, and tumor control is warranted.
REFERENCES 1. Mell LK, Mehrotra AK, Mundt AJ. Intensity-modulated radiation therapy use in the U.S., 2004. Cancer 2005;104:1296– 1303. 2. Chao KS, Deasy JO, Markman J, et al. A prospective study of salivary function sparing in patients with head-and-neck cancers receiving intensity-modulated or three-dimensional radiation therapy: Initial results. Int J Radiat Oncol Biol Phys 2001;49: 907–916. 3. Eisbruch A, Schwartz M, Rasch C, et al. Dysphagia and aspiration after chemoradiotherapy for head-and-neck cancer: Which anatomic structures are affected and can they be spared by IMRT? Int J Radiat Oncol Biol Phys 2004;60:1425–1439.
4. Tsien C, Eisbruch A, McShan D, et al. Intensity-modulated radiation therapy (IMRT) for locally advanced paranasal sinus tumors: Incorporating clinical decisions in the optimization process. Int J Radiat Oncol Biol Phys 2003;55:776–784. 5. Lee NY, Mechalakos JG, Nehmeh S, et al. Fluorine-18-labeled fluoromisonidazole positron emission and computed tomography-guided intensity-modulated radiotherayp for head and neck cancer: A feasibility study. Int J Radiat Oncol Biol Phys 2008;70:2–13. 6. Feng M, Eisbruch A. Future issues in highly conformal radiotherapy for head and neck cancer. J Clin Oncol 2007;25: 1009–1013.
Adoptive planning in IMRT for head and neck cancer d P. H. AHN et al.
7. Ahn PH, Ahn AI, Lee CJ, et al. Random positional variation amongst the skull, mandible and cervical spine with treatment progression during head and neck radiotherapy. Int J Radiat Oncol Biol Phys 2009;73:626–633. 8. Zhang L, Garden AS, Lo J, et al. Multiple regions-of-interest analysis of setup uncertainties for head-and-neck radiotherapy. Int J Radiat Oncol Biol Phys 2006;64:1559–1569. 9. Kaiser A, Schultheis TE, Wong JYC, et al. Pitch, roll and yaw variations in patient positioning. Int J Radiat Oncol Biol Phys 2006;66:949–955. 10. Hong TS, Tome WA, Chappell RJ, et al. The impact of daily setup variations on head-and-neck intensity-modulated radiation therapy. Int J Radiat Oncol Biol Phys 2005;61:779–788. 11. Barker JL, Garden AS, Ang KK, et al. Quantification of volumetric and geometric changes occurring during fractionated ra-
685
diotherapy for head-and-neck cancer using an integrated CT/ linear accelerator system. Int J Radiat Oncol Biol Phys 2004; 59:960–970. 12. Ballivy O, Parker W, Vuong T, et al. Impact of geometric uncertainties on dose distribution during intensity modulated radiotherapy of head-and-neck cancer: The need for a planning target volume and a planning organ-at-risk volume. Curr Oncol 2006;13:108–115. 13. Robar JL, Day A, Clancey J, et al. Spatial and dosimetric variability of organs at risk in head-and-neck intensity-modulated radiotherapy. Int J Radiat Oncol Biol Phys 2007;68:1121–1130. 14. O’Daniel JC, Garden AS, Schwartz DL, et al. Parotid dose in intensity-modulated radiotherapy for head and neck cancer: Is what you plan what you get? Int J Radiat Oncol Biol Phys 2007;69:1290–1296.
doi:10.1016/j.ijrobp.2010.03.014
CLINICAL INVESTIGATION
Head and Neck
ADAPTIVE PLANNING IN INTENSITY-MODULATED RADIATION THERAPY FOR HEAD AND NECK CANCERS: SINGLE-INSTITUTION EXPERIENCE AND CLINICAL IMPLICATIONS PETER H. AHN, M.D.,* CHIN-CHENG CHEN, PH.D.,* ANDREW I. AHN, M.ENG.,y LINDA HONG, PH.D.,* PAOLA G. SCRIPES, M.S,* JIN SHEN, B.S.,* CHEN-CHIAO LEE, M.S.,* EKENI MILLER, A.S.,* SHALOM KALNICKI, M.D.,* AND MADHUR K. GARG, M.D.* )
Department of Radiation Oncology, Montefiore Medical Center and Albert Einstein College of Medicine of Yeshiva University, Bronx, New York and yAlbert Einstein College of Medicine of Yeshiva University, Bronx, New York Purpose: Anatomic changes and positional variability during intensity-modulated radiation therapy (IMRT) for head and neck cancer can lead to clinically significant dosimetric changes. We report our single-institution experience using an adaptive protocol and correlate these changes with anatomic and positional changes during treatment. Methods and Materials: Twenty-three sequential head and neck IMRT patients underwent serial computed tomography (CT) scans during their radiation course. After undergoing the planning CT scan, patients underwent planned rescans at 11, 22, and 33 fractions; a total of 89 scans with 129 unique CT plan combinations were thus analyzed. Positional variability and anatomic changes during treatment were correlated with changes in dosimetric parameters to target and avoidance structures between planning CT and subsequent scans. Results: A total of 15/23 patients (65%) benefited from adaptive planning, either due to inadequate dose to gross disease or to increased dose to organs at risk. Significant differences in primary and nodal targets (planning target volume, gross tumor volume, and clinical tumor volume), parotid, and spinal cord dosimetric parameters were noted throughout the treatment. Correlations were established between these dosimetric changes and weight loss, fraction number, multiple skin separations, and change in position of the skull, mandible, and cervical spine. Conclusions: Variations in patient positioning and anatomy changes during IMRT for head and neck cancer can affect dosimetric parameters and have wide-ranging clinical implications. The interplay between random positional variability and gradual anatomic changes requires careful clinical monitoring and frequent use of CT- based image-guided radiation therapy, which should determine variations necessitating new plans. Ó 2011 Elsevier Inc. Radiotherapy, Head-and-neck tumor, Positional change, Anatomic change, Dosimetry.
Precise radiotherapy delivery depends upon the predictability of planning parameters: patient external contour and skin separation, internal target motion, tumor size, and patient position. A change in any of these parameters may change the delivered dose to tumor, nodal areas at risk, and normal structures in ways that require further investigation (6). We report our results of a prospective, adaptive head and neck image-guided radiation therapy (IGRT) program for patients who underwent scheduled rescanning and were replanned if deemed necessary based on comparative dosimetric analyses. Changes in patient anatomy (weight loss and tumor shrinkage), as well as positioning variability, were correlated with dosimetric endpoints. We previously
INTRODUCTION The utilization of intensity-modulated radiation therapy (IMRT) in the treatment of head and neck cancer has increased dramatically in the last decade (1), driven by studies that demonstrate that controlling the dose to parotids (2), superior pharyngeal constrictors (3), cochlea (4), and other structures affect quality of life. IMRT’s dose conformality, which permits organ sparing, also allows for dose escalation to gross tumor, selectively using areas of radioresistant hypoxia as the target (5). The hallmark of IMRT plans is the creation of sharp dose gradients with high target conformality; IMRT and related technologies represent a major advance in improving the therapeutic ratio of head and neck radiotherapy. Reprint requests to: Madhur K. Garg, M.D., Department of Radiation Oncology and Otolaryngology, Montefiore Medical Center/Albert Einstein College of Medicine, 111 East 210th St., Bronx, NY 10467. Tel: (718) 920-4140; Fax: (718) 231-5064; E-mail: [email protected]
This work was presented in part at the 7th International Conference on Head and Neck Cancer, San Francisco, CA, July 2008. Conflict of interest: Dr. Ahn’s spouse is a shareholder in Pyronia Medical Technologies. Received Dec 14, 2009, and in revised form Feb 10, 2010. Accepted for publication March 12, 2010. 677
678
I. J. Radiation Oncology d Biology d Physics
reported results with the same patient set, quantifying the degree of variability in various portions of the head and neck, noting that the 95% confidence interval of uncertainty in the region of the mandible and lower neck is upwards of 7 mm in an unpredictable manner (7). METHODS AND MATERIALS Twenty-three consecutive patients with locoregionally advanced head and neck cancer underwent a course of intensity-modulated IGRT; 33 fractions with simultaneous integrated boost (SIB) differential dose painting, at Montefiore Medical Center, Bronx, NY. Patients were immobilized with a custom-fitted thermoplastic face mask (Bionix, Toledo, OH) and shoulder pulls (WFR/Aquaplast, Wyckoff, NJ). Computed tomography (CT) simulation was then performed with 2.5-mm-thick slices (GE Lightspeed; GE Healthcare, Chalfont St. Giles UK), with planned rescans at fraction numbers of 11, 22, and 33. A positron emission tomography (PET) CT scan in the treatment position was obtained from all patients and fused with the planning CT scan. A single radiation oncologist planned treatment for all 23 patients; volumes were outlined utilizing CT and PET-CT findings, as well as findings on endoscopic and physical examination at the time of initial consult. Planning was performed on an Eclipse planning system (Varian Medical Systems, Palo Alto, CA). The IMRT plans utilized seven fields, encompassing the entire neck, so no anterior lower-neck split field was used. In 22 patients, gross tumor volume (GTV) or high-risk areas received 66 to 69.96 Gy in 33 fractions with a 1-cm margin planning target volume (PTV-primary; uninvolved nodal areas at risk were treated to 54.12 Gy in 33 fractions, with a margin of 0.5-cm (PTV-nodal). The remaining patient was treated to a dose of 50.4 Gy in 1.8 Gy fractions to PTV-nodal and PTV-primary, followed by a boost to a total of 72 Gy in 1.8-Gy fractions to PTV-primary. For normal tissue constraints, the dose to a 5-mm expansion around the spinal cord was <45 Gy, the maximum brainstem dose was 54 Gy to 20%, the median dose to one parotid was <26 Gy, and 95% of the PTV-nodal and PTV-tumor volumes received 95% of the prescription dose. All patients were rescanned on the original CT simulator, for a total of 89 individual CT scans. Radioopaque fiducial markers were placed on the patient’s tattoos at the time of rescan. The patient’s treatment was divided into quartiles for purposes of comparison; S1 for original planning CT; S2 for rescan at around fraction no. 11 (range, 1–19 fractions); S3 for rescan at fraction no. 22 (range, 20–30 scans); and S4 for rescan at fraction no. 33 (range, 31–33 fractions). Avoidance and target structures from the original planning scan, including the spinal cord, brainstem, mandible, parotids, CTV (including individual right, left and retropharyngeal lymphatic contours as applicable), GTV and PTV were reconstituted on the rescan images; the contours were then modified by the original radiation oncologist to account for any changes in patient anatomy and positioning from the original pretreatment plan. GTV was changed only if there was clear anatomical displacement; at the primary site, this occurred most commonly with tumor shrinkage leading to a portion of the rescan GTV clearly overlapping with air in the laryngeal or pharyngeal space; in nodal volumes, tumor shrinkage outside the skin contour on the rescan CT required modification of the nodal GTV. The original field arrangement with which the patient was being treated was then fused onto the rescan CT based on carefully matched isocenters, soft tissue, and bony anatomy. The dose distribution was recalculated and dose–volume histograms were generated for the structures recontoured on the rescan CT. The pattern
Volume 80, Number 3, 2011
of dose distribution was also examined on axial slices. After analysis of dose distribution, patients were replanned if constraints to spinal cord and other normal structures were not met or there was inadequate coverage of the target structures. Subsequent CT scans during treatment had both the original field arrangement and any replanned field arrangements reconstituted on the images and examined for the dosimetric consequences, yielding 129 unique CT-field combinations and dosimetric plans. After patients completed treatment, patient anatomical parameters such as skin separation (at isocenter) and shifts in the isocenter, defined by radioopaque fiducial markers versus bony anatomy, and individual patient positioning parameters were determined on the original and rescan CTs.
Statistical analysis Statistical analysis was performed using multiple linear regression analysis, Pearson correlation calculations, and principal component’s multivariate analysis using Minitab statistical software. Strong correlation was considered to consist of R2-adjusted values of 0.7 to 1.0, while moderate correlation was considered to consist of R2-adjusted values of 0.4 to 0.7. Differences in dosimetric outcome for the population over time were calculated using a oneway unstacked analysis of variance comparison, and multivariate discriminant function analysis.
RESULTS The characteristics of the studied patients are outlined in Table 1. A majority of the 23 patients had locoregionally advanced squamous cell carcinoma, divided largely between laryngeal/hypopharyngeal primary and oropharyngeal/oral cavity primary. Nineteen patients received cisplatin or carboplatin-based concurrent chemoradiotherapy, while 2 patients received concurrent cetuximab therapy. Changes in patient anatomy The average weight loss over the entire course of treatment was13 pounds (8.3%), with the most significant change occurring between scans S2 and S4. Patients did not have any significant change in skin separation between scans S1 and S2; there were significant changes in skin contour and separation between S1 and S3 to S4, with an average decrease in skin separation at the level of the isocenter of 10% by the end of treatment (Table 2). The average decrease in the size of GTV (primary and nodal) was 17.2%. Parotid volume decreased progressively by 24% during early treatment (scans S1–S2), stabilizing around scan S3 (Table 2). As would be expected with progressing treatment, lost weight correlated moderately to strongly with reduction in skin separation at C1, C4, and C7 and mid-neck level (Fig. 1). Dosimetric findings There is a progressive increase in dose inhomogeneity as one proceeds from S1 to S4. PTV coverage (both nodal and primary) was in S2 but improved in S3 and S4. With reduction in skin separation at mid-tumor level, the hot spot increased. Hot spot and PTV coverage also correlated both with anatomic and positional variations. In 14 of 23 patients (61%), PTV dose homogeneity and coverage were improved by replanning at S2 or S3.
Adoptive planning in IMRT for head and neck cancer d P. H. AHN et al.
Table 1. Patient characteristics by histology, primary site, operative status, and local and nodal stage Histology Primary site
Operative status Stage
Local stage
Nodal stage
Squamous cell
Patients (n = 22)
Adenoid cystic Nasopharynx Oral cavity Oropharynx Hypopharynx Larynx Unknown primary Definitive Postoperative 1 2 3 4a 4b T1 T2 T3 T4 N0 N1 N2a N2b N2c N3
1 3 6 5 2 6 1 19 4 1 2 1 17 2 5 8 4 6 5 1 5 5 5 2
Spinal cord dose initially increased in S2 but came down as treatment progressed. In 7/23 patients (30%), the maximum cord point dose exceeded our 45-Gy constraint at either S2 or S3 and needed correction. This increase correlated moderately with positional variability but did not correlate with anatomic changes. Parotid dose increased as treatment progressed, often but not always in accordance with a decrease in parotid volume (Table 2). This increase correlated with both positional and anatomic changes. In 5/23 patients (22%), the replan yielded improved parotid sparing. There was no significant difference in dose to the mandible or to the brainstem in our population. Correlation between dosimetric parameters and anatomic changes Weight loss and fraction number moderately correlated with skin separation, parotid volume, at various levels in the neck, and at the midpoint of involved neck nodes (Table 3). Coverage of GTV and PTV improved with decreases in skin separation. Dose to the spinal cord was affected adversely by decreased skin separation throughout the neck, which also occurred with shrinkage of tumor in the neck. Parotid dose increased with decreases in skin separation at the upper-mid neck or with shrinkage in parotid volume. Dose to the mandible was adversely affected as GTV volume decreased. Predicting dosimetric parameters based on positional variations Positional variation adversely affected dosimetric values (Table 4), especially cord dose, GTV, CTV, and PTV cover-
679
age. Variation in position of the cochlea (as a proxy for skull position) adversely affected dose to the parotid and led to changes in dose to GTV and PTV. Variation in position of elements of the cervical spine including lordosis and scoliosis influenced dose to the spinal cord, GTV, CTV, and PTV coverage, with changes in lordosis adversely affecting coverage of the retropharyngeal nodes. Positional variation of the mandible adversely influenced dose to the cord, parotids, CTV, and nodal PTV. As demonstrated by the correlation coefficients, most associations were weak to moderate without any single dominating factor. Overall, 15 out of 23 patients (65%) benefited from improved dose distributions second to replanning. The reasons varied from exceeding spinal cord, brainstem or parotid dose constraints, or inadequate (better dose homogeneity) PTV coverage (Fig. 2). External predictors of replan necessity We determined whether externally measurable parameters including anatomic changes in skin separation and/or positional variation could predict for the need for replanning patients for either overdose of normal avoidance structures or underdose of target structures. Need for replan was usually based on dosimetric parameters to spinal cord (maximal dose [Dmax] >45 Gy), parotid (D50% >26 Gy in one side); GTV-V95% of <95%; CTV-V95% of <95%; and PTVD95% of < 95% (Fig. 3). Comparisons also included correlations with weight loss, changes in skin separation at isocenter, and anatomic isocenter (Table 5). No single positional or anatomic variable predicted for the need for a replan; instead, multiple independent and statistically significant trends were evident for various normal and target structures (Table 6). Replan due to higher dose to spinal cord was driven largely by positional variation, including the cervical spine but also the cochlea and mandible. Higher dose to the parotids was driven by variations in isocenter, mandible position, and decreased skin separation. Underdose to GTV was driven by variations in isocenter, skull, mandible, and spine positions, and changes in separation in the lower spine. Decreases in skin separation throughout the neck and positional variation in isocenter, skull, and mandible and mid-neck predicted for the need to replan due to underdose to CTV and PTV. Large changes in the hotspot were due primarily to positional variation and decreases in skin separation with tumor shrinkage and weight loss. Clinically significant overdose to the mandible was primarily due to positional factors. DISCUSSION Several studies have demonstrated variability in patient positioning for head and neck radiotherapy. Zhang et al. (8) demonstrated somewhat independent movement between the maxilla and the lower cervical spine with respect to the variable of rotation of the skull with respect to the lower neck. Kaiser et al. (9), assuming rigidity between the skull and neck, demonstrated small variations in patient position
680
I. J. Radiation Oncology d Biology d Physics
Volume 80, Number 3, 2011
Table 2. Dosimetric parameters and skin separation with treatment progression Dosimetric endpoint PTV (nodal) D99% D95% V95% PTV (tumor) D99% D95% V95% CTV (nodal) D100% V100% V95% D95% GTV D100% V100% D95% Right parotid Dmean D50% D10% V26Gy V20Gy Volume Left parotid Dmean D50% D10% V26Gy V20Gy Volume Spinal cord Dmax D1cc Brainstem Dmax D20% Mandible V60 Separation @ isocenter
S1
S2
S3
S4
p value
49.19 Gy 53.08 Gy 97.48%
43.62 Gy 50.73 Gy 93.59%
44.90 Gy 51.64 Gy 94.9%
48.05 Gy 52.92 Gy 96.41%
<0.01 <0.01 <0.01
62.73 Gy 65.54 Gy 96.28%
58.51 Gy 63.82 Gy 91.93%
58.82 Gy 63.94 Gy 93.60%
60.55 Gy 64.68 Gy 94.21%
NS NS NS
40.14 Gy 92.25% 98.54% 54.09 Gy
36.10 Gy 90.87% 97.20% 53.20 Gy
37.18 Gy 91.56% 97.66% 53.69 Gy
36.45 Gy 93.78% 98.21% 53.62 Gy
<0.053 NS <0.05 NS
57.11 Gy 90.31% 69.33 Gy
57.85 Gy 85.80% 68.26 Gy
55.48 Gy 86.48% 66.74 Gy
55.04 Gy 90.00% 67.45 Gy
NS NS NS
30.81 Gy 28.92 Gy 52.94 Gy 49.61% 65.82% 26.98 cm3
33.02 Gy 30.13 Gy 54.46 Gy 55.08% 72.57% 24.67 cm3
32.63 Gy 30.86 Gy 52.90 Gy 52.62% 68.80% 20.16 cm3
33.39 Gy 30.63 Gy 56.78 Gy 57.44% 73.26% 20.61 cm3
<0.02 <0.05 <0.05 <0.02 <0.006 <0.001
28.55 Gy 24.97 Gy 49.29 Gy 44.71% 60.93% 24.9 cm3
31.92 Gy 27.81 Gy 50.54 Gy 49.39% 66.23% 22.26 cm3
29.37 Gy 26.41 Gy 49.38 Gy 48.62% 64.26% 18.29 cm3
31.03 Gy 28.61 Gy 50.32 Gy 51.52% 67.47% 18.09 cm3
<0.03 <0.04 NS <0.05 <0.04 <0.05
39.40 Gy 36.55 Gy
42.76 Gy 39.16 Gy
41.60 Gy 37.59 Gy
40.73 Gy 37.02 Gy
<0.02 <0.03
42.55 Gy 28.92 Gy
44.31 Gy 28.47 Gy
45.63 Gy 30.7 Gy
41.06 Gy 26.33 Gy
NS NS
14.35% 11.43 cm
14.18% 11.11 cm
14.68% 10.68 cm
11.05% 10.01 cm
NS <0.001
Abbreviations: S1 = initial planning CT Scan; S2 = fraction 1-19 (primarily fraction 11); S3 = fraction 20-30 (primarily fraction 22); S4 = fraction 31-33 (primarily fraction 33); NS = not statistically significant; Dxx% = dose to xx% of the target or organ volume; Vxx% = % of the target or organ volume receiving xx% of target dose; CTV = clinical target volume; GTV = gross tumor volume; PTV = planning target volume. Dosimetric endpoints for PTV, CTV, GTV, parotid glands, spinal cord, brainstem, mandible, as well as changes in horizontal separation in skin with treatment from planning CT scan (S1), around fraction 11 (S2), around fraction 22 (S3), to completion of treatment around fraction 33 (S4). In this set of patients, there is significant dosimetric variability in parotid and spinal cord dose as treatment proceeds, with differences in CTV and PTV coverage. Changes in GTV coverage and dose to the mandible and brainstem are not statistically significant in this cohort. All data are with the original fields transposed on rescan CT, and prior to any replan.
with respect to roll, pitch, and yaw that were unchanged between weeks 1 and 4 of fractionated radiotherapy. Hong et al. (10) reported small systematic but large random variations in rotational and translational position, as well as vector displacement. Using theoretical IMRT plans, an adverse and unpredictable variation in dose deposition when taking into account these movements was demonstrated, with instances of geographic miss and normal tissue overdosing (10). Barker et al. (11) found that in patients with nodal disease greater than 4 cm, the tumor in the neck decreased by 69.5%, with medial displacement of the center of mass of the parotid in a manner that was highly correlative
with patient weight loss. Medial shift in the center of mass of the tumor was also noted. Barker et al. (11) retrospectively examined dosimetric aspects of anatomic deformation in a retrospective study of patients selected for replanning at the midpoint of chemoradiotherapy. In those patients selected due to significant weight loss or tumor shrinkage, replanning led to improvement in coverage of the PTV, with decreased dose to the spinal cord, brainstem, and right parotid compared to that with no replanning. In a repeat CT scan study of 8 patients, Ballivy et al. (12) examined weekly CT scans from head and neck patients and found higher doses to the contralateral parotid and spinal cord than originally planned, with
Adoptive planning in IMRT for head and neck cancer d P. H. AHN et al.
681
Fig. 1. In this patient with an unknown primary squamous cell carcinoma, the 45-Gy isodose line approaches a 5-mm avoidance structure around the spinal cord on the original plan (a). At the 24th fraction, there have been significant changes in skin separation due to weight loss and tumor shrinkage, leading to the 45-Gy isodose line encroaching on the spinal cord (b). The patient was replanned in order to avoid excess dose to the spinal cord (c).
dose to the spinal cord greater than 45 Gy in 57% of repeat CT scans. Robar et al. (13) also found that positional uncertainty was greatest in the lower neck and that while mean maximal dose variation to craniospinal structures was generally on the order of 2.2%, the range in maximal dose could reach 39% in the lower neck. O’Daniel et al. (14) examined the effect of anatomic change with radiotherapy and positioning variability at C2 on dose to the parotid gland, and found that almost half of the patients had parotid gland doses 5 to 7 Gy above that expected from the planning CT. There was a contribution from both inconsistent patient immobilization and anatomic changes as treatment proceeded (14). That study did not find any detriment to CTV coverage with anatomic change and variations in patient positioning. Another study looked at the dosimetric effect of tumor shrinkage on target and organs at risk (OARs), using a deformation algorithm. Those authors found that although the dose to target did not change significantly, the parotid dose increases by
Table 3. Statistically significant correlations between anatomic parameters, fraction number, and weight loss External parameter
Anatomic correlation
Weight loss
C1 lateral separation C4 lateral separation Parotid shrinkage
Fraction number
C1 lateral separation C4 lateral separation C7 lateral separation Separation at midneck tumor GTV volume
Pearson correlation coefficient 0.44 0.49 0.52 (right), 0.67 (left) 0.71 0.69 0.40 0.75 0.67
Abbreviation: GTV = gross tumor volume External parameters such as progressive weight loss and fraction number have the highest correlation with parotid and GTV shrinkage, respectively.
I. J. Radiation Oncology d Biology d Physics
682
Volume 80, Number 3, 2011
Table 4. Correlations between dosimetric parameters and positional variation
Dosimetric parameter Cord Dmax increase
Parotid D50% increase GTV-D95% decrease
CTV-D100% decrease CTV-V95% decrease RP-CTV-D100% decrease RP-CTV-V100% decrease PTV nodal - V95% decrease PTV tumor - V95% decrease
Positional correlation Anatomic isocenter vector increase Lordosis/Scoliosis change C1-C7 vector increase Mandible vector increase Cochlea vector increase
Pearson correlation coefficient 0.43 0.37–0.43 0.27–0.49; 0.27 0.27 0.41
Mandible vector increase Anatomic isocenter vector increase C1-C7 vector increase Cochlea vector; incisive vector increase Anatomic isocenter vector increase Mandible vector increase C2-C7 vector increase C2-C7 vector increase Anatomic isocenter vector increase Lordosis increase
0.42 0.32
Mandible vector increase
0.39
Cochlea vector; C1-C7 vector increase Anatomic isocenter vector increase Cochlea vector; C1-C7 vector increase
0.31–0.43; 0.24–0.44 0.30
0.31–0.37 0.29; 0.37 0.33 0.22 0.38–0.44 .27–.32 0.19 0.19
0.25–0.38; 0.16–0.31
Abbreviations: Dmax = maximum dose; D50% = dose to 50% of volume; D95% = dose to 95% of volume; D99% = dose to 99% of volume; D100% = dose to 100% of volume; V95% = volume receiving 95% of target dose; V100% = volume receiving 100% of target dose; RP-CTV = retropharyngeal node clinical target volume. There are consistent and statistically significant associations between variation in mandible, cochlea and C1-C7 position and decreased dosimetric coverage of the GTV, nodal CTV and retropharyngeal nodes, as well as increased dose to the spinal cord and parotid glands.
almost 10% during treatment as a result of tumor shrinkage. Dose changes due to positional variability, and its potential interplay with clinical changes such as weight loss was not addressed in the study. The original impetus behind the present study was the generation of an algorithm to predict which external anatomic or clinical variables such as changes in skin separation, fraction number or weight loss would predict for the need for replanning patients. We initially attempted to correlate changes in these anatomic and clinical parameters with decrements in PTV, CTV, and GTV coverage and increases in dose to normal organs that mandated replanning the patient’s treatment. Single anatomic-clinical correlations were not sufficiently robust, but different anatomic elements and positional variation significantly affected dose parameters. The combination of
Fig. 2. Skin separation was measured by taking a horizontal line through the midplane at the bottom of the two foramina of a given C-spine level (C4 in this case). The skin separation measurement was taken from skin to skin.
independent random positioning events and more gradual and predictable anatomy changes drive the need for replanning; in consequence, during adoptive therapy, the decision to replan a patient could hypothetically be driven exclusively by dosimetric findings due to random positional variation during the particular CT session. Conversely, the random positional variation could mask a dosimetric consequence that occurs on other days due to anatomic change as treatment
PTV-GTV=8 cord=7 PTV-nodes=6 parotid=5 mandible =1 brainstem=1
Fig. 3. A total of 15 of 23 patients required a replan at some point, based on dosimetric parameters including Dmax exceeding 45 Gy for cord, PTV-tumor, or PTV-node D95% below 95% of prescription, parotid D50% increased significantly above 26 Gy, mandible V60 Gy above 10%, and brainstem V54 Gy above 20%. Two patients each required two replans each for PTV-GTV underdosage; 2 patients each required two replans each for cord overdosage; 1 patient required two replans for parotid overdosage.
Adoptive planning in IMRT for head and neck cancer d P. H. AHN et al.
Table 5. Correlations between dosimetric and anatomic changes Dosimetric parameter Cord Dmax increase
Parotid D50% increase
Mandible V60 increase GTV-D100% increase GTV-D95% increase CTV PTV nodal – D95% incr PTV tumor – D95% incr
Anatomic variable Mandibular joint separation decrease C4-C6 lateral separation decrease neck midtumor separation decrease Mandibular joint separation decrease C1-C5 lateral separation decrease Parotid volume decrease Weight loss GTV volume decrease C7 lateral separation decrease GTV volume decrease C7 lateral separation decrease None C1-C5 lateral separation decrease PTV nodal volume change decrease C1-C3 lateral separation decrease PTV nodal and tumor volume decrease
Pearson correlation coefficient
Table 6. Need for replan and reason correlated with positional and anatomic changes* Dosimetric parameter Cord Dmax overdose
0.30 0.17–0.27 0.18 0.22–0.39
Parotid D50 overdose
0.17–0.28 GTV underdose 0.22 0.30–0.35 0.26 0.18 0.47 0.23
CTV nodal underdose
0.19–0.25 0.27 0.23–0.24 0.30–0.31
All correlations are statistically significant. Cord, parotid, and mandible doses were increased with decreases in skin separation or volume change of select structures. CTV coverage was not significantly affected by changes in anatomy, although there was increase in PTV nodal-D95% coverage with decreased separation. Coverage of GTV-D100% was improved with decreases in GTV volumes as treatment proceeded.
proceeds and falsely indicate adequate dosimetric coverage or acceptable normal tissue exposure when the anatomy would dictate that a labor-intensive replanning process is indicated. Reduced skin separation, weight loss, and loss in parotid volumes correlated weakly with increase in parotid D50%. Variations in these anatomic or clinical surrogates, however, were nonsignificant and by themselves did not predict for the need for a replan. Dose to normal and target structures are also adversely affected by variations in positioning of the skull, mandible, and elements of the cervical spine. In our study, 15/23 patients would have a better dose distribution on replanning either because of better dose homogeneity to the target or better sparing of OAR. This has implications for not only better tumor control but also better quality of life by reducing dose to OAR and therefore toxicity. This is the first dosimetric study that we are aware of in which patients were prospectively studied and scanned at regular intervals, correlating multiple aspects of positioning
683
PTV gross underdose PTV nodal underdose
Hotspot overdose
Mandible overdose
Need for replan Cochlea/incisive roll, pitch, RL, AP Mandible roll, yaw C3-C7 vector C2-6 pitch C2-C5 AP Lordosis Anatomic isocenter AP
p value 0.024–0.054 0.034 0.002–0.043 0.001–0.005 0.001 0.001 0.002
Mandible vector, AP, SI Skin separation (C2-C3) Anatomic isocenter vector, AP Weight Loss Mid-cochlea/incisive vector Mandible vector, roll C1-C7 midpoint vector Scoliosis Skin separation (C6-C7) Anatomic isocenter AP, RL
0.001–0.006 0.07–0.08 0.001–0.022
Cochlea SI Weight Loss Mandible roll C3 roll C4-5 SI Skin separation (C2-C7) Anatomic isocenter RL
0.009 0.065 0.075 0.033 0.001–0.020 0.001–0.037 0.015
Cochlea midpoint SI Weight loss Cochlea/incisive vector, roll, RL Mandible vector/yaw/roll C2-C4 AP, pitch C3 roll C2-C5 vector Skin separation (C2-C7) Anatomic isocenter vector Incisive foramen vector; cochlea vector/pitch/AP Mandible vector Lordosis C4 AP C1-C7 vector Skin separation at mid-tumor Anatomic isocenter vector, RL Cochlea/incisive AP, RL, roll Mandible vector/yaw/roll/pitch
0.063 0.099 0.001–0.026
0.003 0.002–0.052 0.001–0.064 0.001–0.028 0.035 0.009 0.002–0.008
0.001–0.012 0.003–0.034 0.033 0.002–0.050 0.002–0.017 0.001 0.001–0.043 0.017 0.001 0.021 0.001 0.036 0.001–0.002 0.013–0.039 0.004–0.028
Abbreviations: AP = anterior-posterior; SI = superior-inferior; RL = right-left. * Reasons for patient replan correlated with anatomic and positional variables. Statistically significant correlations exist between positional and anatomic variation, and the need for replan, usually based on dosimetric parameters to spinal cord (Dmax >45Gy), parotid (D50% >26Gy), GTV-V95% <95%, CTV-V95% <95%, and PTV-D95% <95%. Potential externally measurable measures would include weight loss, skin separation at various portions of the head and neck, and fraction number. Eighteen replans were performed.
684
I. J. Radiation Oncology d Biology d Physics
and anatomic variability with dosimetric parameters in an attempt to determine which of these external factors might predict for the need for a rescan and replan. No single positional or anatomic variable predicted for the need for a replan; instead, multiple independent and statistically significant trends were evident for various normal and target structures. Therefore, it is not only important to monitor changes in anatomy but also look at the various positional variability that can happen during IMRT for head and neck cancers. Isocenter and three-point triangulation verification using orthogonal images during treatment may not be adequate for patients undergoing IMRT with tighter margins. Anatomic changes such as skin separation, target, and parotid volumes along with positional alignment using multiple points of interest such as skull position, spine curvature, and mandible position should be carefully monitored during IMRT for head and neck cancers. Weight loss and decreased skin separation (due either to tumor shrinkage or weight loss) during the course of radiation therapy can lead to head movement inside the mask, giving rise to positional variability. In addition, weight loss and reduced skin separation by itself has been associated with higher parotid dose and target dose inhomogeneity. IGRT using CT scans should be used more frequently for these patients to assess anatomic and positional variability. Although resolution of presently available CT-based OBI is suboptimal, it can be used as a preliminary tool to assess a patient for resimulation and a replan. In our clinic, we perform planned rescans in the midcourse of treatment regardless of weight loss or tumor shrinkage on the diagnostic-quality CT scanner, and volumes are reconstituted on the new images, using deformable registration and manual correction. Original treatment fields are then reconstituted on this new CT scan with the deformed volumes, and dose is calculated. The fidelity of nonrigid patient positioning between the original CT and the rescan CT is examined, since spurious dosimetric findings can be a result of a random positional event specific to that day. Examining the resulting dosimetry, we do not replan if dose to the subclinical PTV is compromised because we often find that dose to subclinical CTV is improved, but we will replan if there is excess dose to normal structures, including the mandible (with significant GTV shrinkage). Any underdose of GTV will prompt a replan. If there is
Volume 80, Number 3, 2011
excess dose to the spinal cord, and variation in positioning of the spinal canal between the two CT scans is noted, we will examine the cone beam images to determine if any variation in cord position is consistent (i.e., systematic) or random. If there is indeed a systematic variation in cord position, we will replan the patient’s treatment. Similarly, if there is no variation in translational or rotational position of the skull or upper neck and there is an increase in parotid dose, we will replan the patient. We recommend that patients receiving IMRT for head and neck cancers undergo a rescan with a new mask if either weight loss or tumor shrinkage results in the mask being loose. They should also undergo CT-based IGRT in addition to three-point triangulation and orthogonal kV images every day for positional accuracy. Facilities using head and neck IMRT without CT-based IGRT capabilities should carefully monitor patients for weight loss and decreased skin separation due to tumor shrinkage. In addition, translational and rotational set up variability should be looked at carefully in addition to verifying the isocenter position on orthogonal images. Additional margins should also be used around OARs to constrain dose to these structures for optimization. CONCLUSIONS Both positional and anatomic variability can have wideranging effect on dosimetric outcome in patients undergoing IMRT for head and neck cancers. No single external variable such as weight loss, fraction number, or changes in skin separation in isolation can reliably predict for the need to initiate the resource-intensive replanning process, due to the confounding effect of positional variability. Isocenter and three-point triangulation verification using orthogonal images, during treatment, may not be adequate for patients undergoing IMRT with tighter margins. In addition to carefully monitoring clinical/anatomic parameters such as weight loss and skin separation, IGRT using CT scans should be used more frequently in these patients to assess both anatomic and positional variability. Research that correlates dose actually received by organs with clinical outcomes such as xerostomia, dysphagia, and tumor control is warranted.
REFERENCES 1. Mell LK, Mehrotra AK, Mundt AJ. Intensity-modulated radiation therapy use in the U.S., 2004. Cancer 2005;104:1296– 1303. 2. Chao KS, Deasy JO, Markman J, et al. A prospective study of salivary function sparing in patients with head-and-neck cancers receiving intensity-modulated or three-dimensional radiation therapy: Initial results. Int J Radiat Oncol Biol Phys 2001;49: 907–916. 3. Eisbruch A, Schwartz M, Rasch C, et al. Dysphagia and aspiration after chemoradiotherapy for head-and-neck cancer: Which anatomic structures are affected and can they be spared by IMRT? Int J Radiat Oncol Biol Phys 2004;60:1425–1439.
4. Tsien C, Eisbruch A, McShan D, et al. Intensity-modulated radiation therapy (IMRT) for locally advanced paranasal sinus tumors: Incorporating clinical decisions in the optimization process. Int J Radiat Oncol Biol Phys 2003;55:776–784. 5. Lee NY, Mechalakos JG, Nehmeh S, et al. Fluorine-18-labeled fluoromisonidazole positron emission and computed tomography-guided intensity-modulated radiotherayp for head and neck cancer: A feasibility study. Int J Radiat Oncol Biol Phys 2008;70:2–13. 6. Feng M, Eisbruch A. Future issues in highly conformal radiotherapy for head and neck cancer. J Clin Oncol 2007;25: 1009–1013.
Adoptive planning in IMRT for head and neck cancer d P. H. AHN et al.
7. Ahn PH, Ahn AI, Lee CJ, et al. Random positional variation amongst the skull, mandible and cervical spine with treatment progression during head and neck radiotherapy. Int J Radiat Oncol Biol Phys 2009;73:626–633. 8. Zhang L, Garden AS, Lo J, et al. Multiple regions-of-interest analysis of setup uncertainties for head-and-neck radiotherapy. Int J Radiat Oncol Biol Phys 2006;64:1559–1569. 9. Kaiser A, Schultheis TE, Wong JYC, et al. Pitch, roll and yaw variations in patient positioning. Int J Radiat Oncol Biol Phys 2006;66:949–955. 10. Hong TS, Tome WA, Chappell RJ, et al. The impact of daily setup variations on head-and-neck intensity-modulated radiation therapy. Int J Radiat Oncol Biol Phys 2005;61:779–788. 11. Barker JL, Garden AS, Ang KK, et al. Quantification of volumetric and geometric changes occurring during fractionated ra-
685
diotherapy for head-and-neck cancer using an integrated CT/ linear accelerator system. Int J Radiat Oncol Biol Phys 2004; 59:960–970. 12. Ballivy O, Parker W, Vuong T, et al. Impact of geometric uncertainties on dose distribution during intensity modulated radiotherapy of head-and-neck cancer: The need for a planning target volume and a planning organ-at-risk volume. Curr Oncol 2006;13:108–115. 13. Robar JL, Day A, Clancey J, et al. Spatial and dosimetric variability of organs at risk in head-and-neck intensity-modulated radiotherapy. Int J Radiat Oncol Biol Phys 2007;68:1121–1130. 14. O’Daniel JC, Garden AS, Schwartz DL, et al. Parotid dose in intensity-modulated radiotherapy for head and neck cancer: Is what you plan what you get? Int J Radiat Oncol Biol Phys 2007;69:1290–1296.