Regional Normal Lung Tissue Density Changes in Patients Treated With Stereotactic Body Radiation Therapy for Lung Tumors

International Journal of

Radiation Oncology biology

physics

www.redjournal.org

Clinical Investigation: Thoracic Cancer

Regional Normal Lung Tissue Density Changes in Patients Treated With Stereotactic Body Radiation Therapy for Lung Tumors Quentin Diot, Ph.D., Brian Kavanagh, M.D., Tracey Schefter, M.D., Laurie Gaspar, M.D., Kelly Stuhr, M.Sc., and Moyed Miften, Ph.D. Department of Radiation Oncology, University of Colorado School of Medicine, Aurora, Colorado Received Nov 9, 2010, and in revised form Nov 9, 2011. Accepted for publication Nov 13, 2011

Summary Sixty-two lung cancer patients treated with stereotactic body radiation therapy were retrospectively studied to obtain curves of normal lung tissue density response to the delivered dose at various times post-treatment. Response increased with dose and was mostly timeindependent after 6 months. Although patients treated with 3 fractions had a response similar to that of patients treated with >3 fractions after 6 months, the group treated with >3 fractions had twice the response of the other group at 6 months, demonstrating more pronounced lung normal tissue selflimited acute effects compared to late effects.

Purpose: To describe regional lung tissue density changes in normal lung tissue of patients with primary and metastatic lung tumors who received stereotactic body radiation therapy (SBRT). Methods and Materials: A total of 179 post-SBRT follow-up computed tomography (CT) scans of 62 patients who received SBRT between 2003 and 2009 were studied. Median prescription dose was 54 Gy (range, 30-60 Gy) in 3 to 5 fractions. SBRT-induced lung density changes on post-SBRT follow-up CT were evaluated at approximately 3, 6, 12, 18, 24, and 30 months after treatment. Dose-response curves (DRC) were generated for SBRT-induced lung damage by averaging CT number (HU) changes for regions of the lungs receiving the same dose at 5-Gy intervals. Results: For all follow-up interval periods, CT numbers linearly increased with dose until 35 Gy and were constant thereafter. For 3, 18, 24, and 30 months, the rate of relative electron density increase with dose was approximately 0.24% per Gy. At 6 months, the rate was also similar below 20 Gy but then rose to 0.6% per Gy above this threshold. After 6 months, DRCs were mostly time-independent. When split between patients treated with 3 fractions of 12 to 20 Gy (median, 20 Gy; average tumor volume, 12  16 cm3) and with >3 fractions of 6 to 12.5 Gy (median, 9 Gy; average tumor volume, 30  40 cm3), DRCs differed significantly. In both cases, CT changes at 3, 18, 24, and 30 months were identical to those of the population DRC; however, patients who received >3 fractions showed 6-month CT changes that were more than twice those for the group that received 3 fractions. Conclusions: This analysis of SBRT-induced normal lung density changes indicates that lung normal tissue has more pronounced self-limited acute effects than late effects. Differences in acute CT changes following treatments in 3 fractions were considerably less than for treatments in >3 fractions. Ó 2012 Elsevier Inc. Keywords: Computed tomography, Density changes, Fibrosis, Lung cancer, Stereotactic body radiation therapy

Reprint requests to: Quentin Diot, Ph.D., Department of Radiation Oncology, University of Colorado, MS F706, 1665 Aurora Court, Suite

Int J Radiation Oncol Biol Phys, Vol. 84, No. 4, pp. 1024e1030, 2012 0360-3016/$ - see front matter Ó 2012 Elsevier Inc. All rights reserved. doi:10.1016/j.ijrobp.2011.11.080

1032, Aurora, CO 80045. Tel: (720) 848-0140; Fax: (720) 848-0222; E-mail: [email protected] Conflict of interest: none.

Volume 84  Number 4  2012

Introduction With technological advances in online imaging, intensity modulation, radiation delivery including motion management that enables accurate delivery of highly conformal plans, stereotactic body radiotherapy (SBRT) is becoming a standard of care for patients with inoperable early-stage non-small cell peripheral lung cancer or with limited demarcated metastases. Delivery of higher doses to smaller lung planning target volumes (PTV) limits potential high-dose-induced toxicity to very limited normal tissue lung areas, and local control of up to 95% at 2 years has been reported (1-4). Although occurrences of severe toxicity are rare, radiation-induced density changes, which could relate to local lung injuries, are often observed on follow-up computed tomography (CT) scans. Typically, the response of lung normal tissue to radiation is divided into early pneumonitis complications and late fibrosis (5). Those 2 mechanisms are relatively well understood independently in a conventional fractionation context (5, 6); however, their individual contributions to global response for a given SBRT treatment is less well known and relates to lung function evolution. Symptomatic lung toxicity with SBRT is typically less than 10% (7); however, occurrences up to 25% have been reported (8), stressing the importance of a better understanding of SBRT treatment parameters to ensure consistent low toxicity levels. In addition, radiographically detectable normal tissue complications are common for SBRT, and a better understanding of their correlations with dosimetric data would help the design of treatment plans that limit those complications, as well as to discriminate radiation-induced lung injury from progression or recurrence. To that purpose, this retrospective study used followup CT scans to examine the dynamic response of normal lung tissue density to high doses delivered in SBRT fractionation to non-small cell peripheral lung cancer primary tumors and pulmonary metastases.

Methods and Materials Patients and treatments Between 2003 and 2009, 62 patients (29 men and 33 women), representing a total number of 81 lesions, were treated at our institution for lung tumors, using SBRT. Thirty-six of these lesions were primary lung cancers, and 45 lesions were metastatic lung tumors whose details are shown in Table. The patient median age was 66 years (range, 29-96 years). Most patient treatment regimens (94%) were planned using a BrainScan pencil beam algorithm, and patients were treated with dynamic conformal arcs on our Novalis linear accelerator (BrianLab, Munich, Germany). Treatment localization was performed using the ExacTrac imaging system (Elekta, Crawley, UK). The remaining patients (6%) were treated with three-dimensional conformal RT (3DCRT) on a Synergy linear accelerator, and treatments were planned using the XiO convolution superposition (Elekta, Crawley, UK). Localization was performed using the on-board XVI kV-cone beam CT system. All patients received a hypofractionated treatment with 3 to 5 fractions (fx) for a total prescribed dose of 30 to 60 Gy in 6 to 20 Gy per fx. Thirty-eight patients (61%), with an average  SD tumor size of 12  16 cm3 were treated with 3 fx of 12 to 20 Gy (median, 20 Gy) and included 71% of primary lesions. The average  SD tumor size for the remaining 24

Normal tissue density change in SBRT lung patients 1025 Table

Lesion breakdown Parameter

No. of patients

Sex F M Tumor location LLL LUL RUL RML RLL Fractionation 3 fx (10, 12, 14, 15, 16, 18, 20 Gy) 4 fx (12.5 Gy) 5 fx (6, 8, 9, 10, 12 Gy)

33 (53%) 29 (47%) 11 32 19 5 14 38 2 22

Site Primary lung cancer Metastasis (primary) Lung Tongue Penis Esophagus Colon Stomach Larynx Kidney Vulva Breast Neck Uterus Thymus Mouth Thigh Liver Ovary

No. of lesions 36

10 1 2 3 7 1 2 3 3 4 1 1 2 2 1 1 1

Abbreviations: LLL Z Left Lower Lobe; LUL Z Left Upper Lobe; RLL Z Right Lower Lobe; RML Z Right Medial Lobe; RUL Z Right Upper Lobe.

patients treated with 4 or 5 fx of 6 to 12.5 Gy (median, 9 Gy) was 30  40 cm3 with 44% of primary cancers.

Simulation and follow-up scans All patient treatments were planned using a CT scan acquired in our department on a HiSpeed (General Electric) flatbed scanner. For most patients, Elekta active breathing coordinator was used to manage the respiratory motion during simulation and treatment (9). Only scans scheduled for the purpose of following the tumor evolution and demonstrating local control at the time of scan acquisition were used in follow-up assessments. Most follow-up scans were acquired on a curved bed scanner at our institution’s radiology department, but occasionally, outside scans were used if they met previous requirements. Follow-up intervals, defined as the time between the last treatment date and follow-up CT scan, varied in length, frequency, and number, and data were binned into groups corresponding to standardized time intervals. The following intervals were selected to distribute patients uniformly: 3 months (3-18 weeks), 6 months (19-34 weeks), 12 months (35-61 weeks), 18 months (62-87 weeks), 24 months (88-121 weeks), and 30 months (122-140 weeks). When multiple followup scans belonged to one interval, the scan closest to the interval center was used, resulting in 179 CT scans.

Dose-response curves To obtain the change in CT number corresponding to the delivered dose, the dose distribution computed from the planning CT scan was rigidly registered to each follow-up scan by overlapping the region of the plan that received the maximum dose with the

1026 Diot et al. corresponding region on the follow-up scans. Because lung volume and shape can vary depending on breathing phase, body mass change, evolution of the disease, or patient positioning at the time of the scan, the previous process was based on the identification of common anatomical landmarks on the two scans. Relative positions of these landmarks allowed the localization of the maximum dose region and permitted comparisons of pre- and post-treatment lung densities in medium- and high-dose regions. In addition, the registration was adjusted to enclose any residual tumor tissue inside the PTV. For each dose bin, a calculation mask was created that included lung tissue receiving the corresponding dose but excluded the PTV. Masks were then adjusted for lung contour variations, and they were used to extract the relevant voxels from CT scans and to compute the mean CT number (in Hounsfield units [HU]) change with respect to the pretreatment value. For each patient, dose-response curves (DRC) for normal lung tissue were obtained by reporting the mean CT number change as a function of the dose bins. Because of the above-mentioned variations between follow-up conditions, each scan required specific masks based on its own lung contours, which were obtained by deforming the planning CT contours using a Demons deformable registration algorithm (10). Each deformed contour was then visually inspected and adjusted when necessary. Fusions and manipulations of the dose and structures were all implemented in MATLAB (MathWorks, Natick, MA) after 256  256 pixel downsampled image data were imported in CERR software (Washington University, St-Louis MO) (11). Although patients held their breath during the inhalation phase for simulation and follow-up scans, inhalation was usually much deeper during follow-up, and these changes in lung volume were addressed by rescaling the follow-up scan density (CT number) such that the mean density of nonirradiated lung tissue was equal to the mean lung density before treatment (12). For each follow-up interval, patient DRCs were combined into a population DRC by weighting the contribution of each patient DRC by their relative lung volume in the population (12, 13). A population DRC was computed. To investigate the influence of different fractionation schedules on the population DRCs, all dose data were converted to their biologically equivalent doses (BED) before the patient DRCs were combined. As the dose per fraction exceeded 6 Gy for all patients, the linear-quadratic linear (LQ-L) model (14) was preferred to the LQ model for the conversion. The LQ-L model was designed as an extension of the LQ model for higher dose, and although its exact form is controversial, the need for a better fit to the high-dose data is well accepted and has been discussed by Fowler (15) and Kavanagh and Newman (16).

International Journal of Radiation Oncology  Biology  Physics

Fig. 1. interval.

Number of patients with follow-ups for each time

of the DRCs were time-independent. Above 20 Gy, the slope for the DRC at 6 months sharply increased until it reached 35 Gy, whereas the slope for all other follow-up times progressively flattened over the same interval. At 6 months, a different response was observed between patients treated with 3 or more fx and those treated with 3 fx. The group treated with more than 3 fx (Fig. 5b) followed the trends observed in the DRC combining all patients, with an even steeper response at 6 months and a slope similar to that of the 3-fx group below the physical dose of 20 Gy. In contrast, the 3-fx group (Fig. 5a) did not show any slope change or threshold at 6 months or at any other time, and all DRCs were quite similar up to 30 Gy. The weighted average of maximum density changes at 6 months with 60% of patients in the 3-fx group and 40% in the other group provided a value close to the whole cohort maximum (Fig. 2), suggesting that most of the 6-month increase visible on the population DRC was actually restricted to the group receiving more than 3 fx. Changes at 3 months were smaller than at other times, probably because the density increase was still transitory and had not yet peaked.

Discussion Similar normal lung tissue DRCs have been obtained for conventional fractionation RT (CFRT) treatment (17-20). However, the small, intense beams with sharp falloff that characterize SBRT treatments required more accurate and intensive data processing to investigate SBRT-specific correlations between dosimetric parameters and normal lung tissue density variations. The

Results Figure 1 shows the follow-up number for each time interval with at most one follow-up per patient per interval and less follow-ups available as time increased. Figure 2 shows population DRCs for 3- , 6- , 12- , 18- , 24- , and 30-month intervals. CT number changes and lung tissue relative electron densities are shown. Overall, the density response for a given dose bin increased during the acute phase, lasting approximately 6 months, and was followed by a less severe late response, beyond 1 year, which might indicate possible tissue repair (Fig. 3). For all follow-up intervals, CT numbers increased with doses below 35 Gy and either leveled off or decreased slightly above it (Fig. 2). Below 20 Gy, the slopes

Fig. 2. Population DRC for various follow-up intervals. The reported dose is the physical dose delivered to the patients. Error bars represent the standard error (68% confidence interval).

Volume 84  Number 4  2012

Normal tissue density change in SBRT lung patients 1027

Fig. 3. Follow-ups of a patient treated with 3 fx of 20 Gy to the PTV. The first column shows the dose distribution in the transverse and coronal plane (inset). The isodose lines are spaced by 10 Gy: the blue line is 10 Gy and the pink line is 70 Gy. Other columns show the lung density evolution over time. After 4 months, the tumor had disappeared, and the normal tissue density in the high dose region appeared to peak. After 4 months, the density decreased until it became invisible after 24 months. irradiated area was much smaller than that for CFRT, and the observed mean values were more prone to delineation or fusion errors as well as observer variability. Specifically, whereas CFRT delivers uniform dose to large areas, SBRT delivers radiation to small areas with sharp dose gradients. Because the overlap between the dose and scans was smaller due to steep dose gradients, DRCs for SBRT were more sensitive to fusion errors than those for CFRT. In addition, the overlap changed over time as the tumor shrank and normal tissue expanded into the vacant space, modifying tissue location with respect to the dose distribution and lowering its apparent density. The average volume of the region encompassed by the dose plateau (above 35 Gy) was 53 cm3 at 6 months (less than 2% of lung volume) and was comparable to the average PTV volume of 50 cm3, possibly showing that the redistribution of tissue around the border of the PTV volume is a probable

Fig. 4. Population DRC for various follow-up intervals. The reported dose is the BED calculated using the LQ-L model. Error bars represent the standard error (68% confidence interval).

mechanism. Considering that the GTV size ranged from less than 1 cm3 to as much as 150 cm3 across the patient cohort, the effect on the density around the high-dose region outside the PTV was likely to be important for large tumors and indicative of a correlation between the DRC plateau for SBRT (Fig. 2) and the migration of the tissue, rather than a biological saturation effect due to high-dose radiation. Because the high-dose volume-totumor size ratio is larger for CFRT than for SBRT, such effects are not expected to be as dramatic and do not appear on conventional treatment DRCs. Similar processes have been reported with fibrotic tissues that contracted and changed the density of the surrounding tissues (19). Consequently, it was difficult to assess the lung damage severity from the density changes that occurred above 35 Gy because each patient was affected differently by tissue redistribution at high doses. Despite these additional difficulties, compared to CFRT follow-up processing, their effects on our data are probably very limited as each CT scan fusion was manually adjusted to compensate for the changing anatomy and tissue redistribution. Moreover, the number of scans in our study should average out the remaining uncertainties. Finally, Palma et al (21) reported similar results from a lung SBRT study in which manual fusion was replaced by deformable registration. Rigid registration was preferred to deformable registration as the delivered dose was highly localized around the tumor, and rigid registration allowed observers to make manual adjustments over high dose regions not available with the automated process of deformable registration. This choice assumed that large deformations of the lungs should not sensibly affect the final result as long as the region around the tumor was accurately, rigidly registered. Although studies using vascular and airway bifurcations as landmarks to assess registration errors have shown that deformable registration could lead to reduced errors, these studies

1028 Diot et al.

International Journal of Radiation Oncology  Biology  Physics 0.06

Deformable − Rigid (%)

(a)

0.04

0.02

0

−0.02

(b)

3 mo 6 mo 12 mo 18 mo 24 mo

0

10

20

30

40

50

60

Dose (Gy)

Fig. 6. CT number difference between deformable registration (Vista software) and rigid registration for the population DRC curves.

Fig. 5. DRCs for patients treated with 3 fx (top) and patients treated with more than 3 fx (bottom). The reported dose is the physical dose received by the patient. Error bars represent the standard error (68% confidence interval). often considered deformations between different 4DCT phases (22). In the context of follow-up examinations separated by long time periods, allowing large tissue changes, Gu et al (23) showed that the use of rigid or deformable registration of pulmonary nodules did not significantly change resulting errors, suggesting that rigid registration was a simpler, yet equally valid approach. We also deformed the dose for 49 of our patients, using Vista version 5.1 software (MIM Software, Cleveland OH), before running the same analysis used with the rigid registration. The two different methods resulted in differences in the DRC curve of less than 5% for a dose less than 60 Gy (Fig. 6), indicating that both methods are practically equivalent. Despite the limitations of the technology currently available, deformable registration research is quite active, and new technologies could soon become available that could manage local deformations not addressed by rigid registration and potentially refine the results of this study. At 6 months, the slope of the DRC varied significantly around a 20-Gy threshold; relative density increased by 0.24% per Gy below 20 Gy and rose to 0.6% per Gy above. This large increase occurred only for the 6-month follow-ups, and the slopes of the DRCs for all other follow-up times were identical to the 6-month low-dose slope in the 0 to 30 Gy range and flat above it. Conventionally, the dominant response mechanism around 3 to 4 months is pneumonitis, with damage to vascular tissue and epithelium cells, and it is fibrosis for later times (greater than 9-12 months), which is characterized by thickening of alveolar septa by bundles of elastic fibers and collagen (24). Consequently, the similar responses of the DRCs at 3 months and those at 12 months and over is surprising. Ma et al (17) published a study of lung patients treated with CFRT, and, although no DRC fitting parameters were provided, the DRC slope seemed to change after

receiving 40 Gy. For CFRT treatments, this threshold corresponds to a lung BED of 67 Gy and is biologically comparable to a 20-Gy threshold for the 3  20 fx schedule. In their data, relative density increased by 0.2% per Gy below 40 Gy and 0.6% per Gy above, which is very similar to our observations. For our study, however, the density peaked at 6 months and then decreased, whereas no decrease was observed in this CFRT study. In our data, a threshold was present only at 6 months instead of every follow-up time longer than 3 months for the CFRT study. Although patients in our study received their treatment in 3, 4, or 5 fx of different sizes, we used the physical dose to sum the effects of different fractionation schedules. To ensure that the previous trends and thresholds did not depend on this choice, the same data were converted to BED before summation. Figure 4 shows that results were identical to the physical dose plot. Two other CFRT studies including breast patients (19, 20) showed a time evolution similar to that in our study, with an initial density increase peaking at 6 months, followed by a decrease and stabilization after 12 months. Those two studies used cruder dose and follow-up interval bins or slightly different endpoints, limiting quantitative comparisons among our study and that by Ma et al (17), particularly for the determination of slope and thresholds. Those studies used mostly lower dose to the tumor and consequently fewer fractions (20, 21, 24e27) compared to those in the study by Ma et al (more than 30). To explain difference in response between the 3-fx group and the other group, we examined multiple parameters. Age has been cited as a predisposing factor to in the development of pulmonary injuries from RT (6, 20, 25); however, the median age for the 3-fx group was 67 years compared to 72 years for the other group, with a nonsignificant P value of .82. Similarly, although most of the differences observed on the DRC occurred above 35 Gy, the V35 (volume receiving more than 35 Gy) was not a good predictor as it was 56 cm3 for 3-fx and 48 cm3 for more than 3-fx (P Z .62). Kyas et al (26) studied volume effects for single-fraction treatment with predictors converted from CFRT, using the LQ model (26) and found the converted V20 (27) in 2-Gy fractions to be a good predictor as well as mean lung dose. Their data, however, were for single-fraction treatments, and the dose or volume constraints they found were quite different from those with CFRT. In our study, the mean lung dose received by the 3-fx group was 4.0 Gy compared to 5.4 Gy for the other group (P Z .013), and according to the analysis by Kyas et al (26), most of our patients

Volume 84  Number 4  2012 should have a probability close to 100% of developing grade 1 pneumonitis symptoms characterized by newly developed hyperdense areas in the treated area, and the response for the two groups should be identical. Many of our patients did not show any grade 1 symptoms, and the discrepancy should be explained by a fraction number greater than 1. Nevertheless, the value of the mean dose to the whole lung for the 3-fx group was lower than that for the other group, indicating that it could potentially account for significant differences between the groups. The difference in mean dose can be explained primarily by the size of the tumor as the mean  SD gross tumor volume (GTV) and PTV volume for the 3-fx group were 12  16 cm3 and 51  39 cm3, respectively, compared with GTV and PTV volumes of 30  40 cm3 (P < .02) and 91  75 cm3 (P < .02), respectively, for the group receiving more than 3 fx. Tumor size could influence the steepness of dose falloff as suggested by the V3, which was 36% for 3-fx and 44% for the other group (P < .05), indicating a larger lung volume exposition to low-level radiation for the second group. Nonetheless, the observation of dosimetric differences between the two groups were in the low-dose region, and the reasons why a larger volume exposed to low dose would influence the high-dose response are unknown. Logistic regression multivariate analysis (28) combining many lung and target dosimetric parameters such as V3, V20, V35, mean dose, Rx50 (volume of the 50% prescription dose divided by the PTV volume), PTV or GTV volume, mean prescription dose, and number of lesions provided no significant predictor for individual patient density increase. Different mechanisms could underlie the weak and strong responses, and the volume irradiated might influence which is predominant. For instance, the observation of vascular lesions or inflammations of lung parenchyma during the pneumonitis phase have been reported for different dose levels (5, 29). Those results, however, were for single-fraction irradiation to large uniform volumes compared to our SBRT treatments, making comparisons difficult. In the context of cytokine-mediated inflammation, the absolute PTV volume receiving high dose might also be a discriminant parameter as a larger PTV size should produce more mediators and trigger a stronger tissue response (30). Given the improved local control and toxicity results reported by recent lung SBRT studies (1-4), future directions for this technique would include larger lesion treatment, and despite the reported low toxicity, our work as well as that by Palma et al (21) suggest that toxicity might increase with lesion size. A better understanding of the relationship between the delivered dose and the density increase is therefore essential to effectively limit toxicity. Despite some large density increase, only 1 of 62 patients included in the study had documented clinical pneumonitis beyond the radiographic stage (grade 2), requiring more data to appropriately model correlations between radiographic and clinical data. Such correlations have been observed in animals (31), which suggests that reducing radiation-induced lung injury is likely to at least increase the quality of life of the many patients experiencing shortness of breath.

Conclusions This analysis of SBRT-induced normal lung density changes indicates that normal lung tissue self-limited acute effects are more pronounced than late effects. Differences in acute CT changes for patients treated with 3 fx were considerably less than

Normal tissue density change in SBRT lung patients 1029 those for patients treated with 4 or 5 fx, but the changes seemed to be explained by either increased low-dose exposure to normal lung or differences in tumor volume. As SBRT treatments will potentially become more aggressive in terms of tumor size, a better understanding of the relation between tumor volume and toxicity will become essential to limit the latter.

References 1. Nagata Y, Takayama K, Matsuo Y, et al. Clinical outcomes of a phase I/II study of 48 Gy of stereotactic body radiotherapy in 4 fractions for primary lung cancer using a stereotactic body frame. Int J Radiat Oncol Biol Phys 2005;63:1427e1431. 2. Rusthoven KE, Kavanagh BD, Burri SH, et al. Multi-institutional phase I/II trial of stereotactic body radiation therapy for lung metastases. J Clin Oncol 2009;27:1579e1584. 3. Timmerman R, McGarry R, Yiannoutsos C, et al. Excessive toxicity when treating central tumors in a phase II study of stereotactic body radiation therapy for medically inoperable early-stage lung cancer. J Clin Oncol 2006;24:4833e4839. 4. Xia TY, Li HQ, Sun QX, et al. Promising clinical outcome of stereotactic body radiation therapy for patients with inoperable stage I/II non-small-cell lung cancer. Int J Radiat Oncol Biol Phys 2006;66: 117e125. 5. Gross N. Pulmonary effects of radiation therapy. Ann Intern Med 1977;86:81e92. 6. Movsas B, Raffin TA, Epstein AH, et al. Pulmonary radiation injury. Chest 1997;111:1061e1076. 7. Marks LB, Bentzen SM, Deasy JO, et al. Radiation dose-volume effect in the lung. Int J Radiat Oncol Biol Phys 2010;76(Suppl):S70eS76. 8. Yamashita H, Nakagawa K, Nakamura N, et al. Exceptionally high incidence of symptomatic grade 2-5 radiation pneumonitis after stereotactic radiation therapy for lung tumors. Radiat Oncol 2007;2:11. 9. Wong JW, Sharpe MB, Jaffray DA, et al. The use of active breathing control (ABC) to reduce margin for breathing motion. Int J Radiat Oncol Biol Phys 1999;44:911e919. 10. Thirion JP. Image matching as a diffusion process: An analogy with Maxwell’s demons. Med Image Anal 1998;2:243e260. 11. Deasy JO, Blanco AI, Clark VH. CERR: A computational environment for radiotherapy research. Med Phys 2003;30:979e985. 12. Woel RT, Munley MT, Hollis D, et al. The time course of radiation therapy-induced reductions in regional perfusion: A prospective study with >5 years of follow-up. Int J Radiat Oncol Biol Phys 2002;52: 58e67. 13. Fan M, Marks LB, Lind P, et al. Relating radiation-induced regional lung injury to changes in pulmonary function tests. Int J Radiat Oncol Biol Phys 2001;51:311e317. 14. Astrahan M. Some implications of linear-quadratic-linear radiation dose-response with regard to hypofractionation. Med Phys 2008;35: 4161e4172. 15. Fowler JF. Linear quadratics is alive and well: In response to Park et al. (Int J Radiat Oncol Biol Phys 2008;70:847e852) [Letter]. Int J Radiat Oncol Biol Phys 2008;72. 957e957. 16. Kavanagh BD, Newman F. Toward a unified survival curve: In response to Park et al. (Int J Radiat Oncol Biol Phys 2008;70:847e852) and Krueger et al. (Int J Radiat Oncol Biol Phys 2007;69:1262e1271) [Letter]. Int J Radiat Oncol Biol Phys 2008;71:958e959. 17. Ma J, Zhang J, Zhou S, et al. Regional lung density changes after radiation therapy for tumors in and around thorax. Int J Radiat Oncol Biol Phys 2010;76:116e122. 18. Theuws JCM, Seppenwoolde Y, Kwa SLS, et al. Changes in local pulmonary injury up to 48 months after irradiation for lymphoma and breast cancer. Int J Radiat Oncol Biol Phys 2000;47:1201e1208. 19. Skoczylas JZ, Bentzen SM, Overgaard M, et al. Time course of radiological lung density changes after postmastectomy radiotherapy. Acta Oncologica 2000;39:181e187.

1030 Diot et al.

International Journal of Radiation Oncology  Biology  Physics

20. Kaha´n Z, Csenki M, Varga Z, et al. The risk of early and late lung sequelae after conformal radiotherapy in breast cancer patients. Int J Radiat Oncol Biol Phys 2007;68:673e681. 21. Palma DA, van So¨rnsen de Koste J, Verbakel WFAR, et al. Lung density changes after stereotactic radiotherapy: A quantitative analysis in 50 patients. Int J Radiat Oncol Biol Phys 2011;81:974e978. 22. Brock KK. Results of a multi-institution deformable registration accuracy study (MIDRAS). Int J Radiat Oncol Biol Phys 2010;76: 583e596. 23. Gu SC, Wilson D, Tan J, et al. Pulmonary nodule registration: Rigid or nonrigid? Med Phys 2011;38:4406e4414. 24. McDonald S, Rubin P, Phillips TL, et al. Injury to the lung from cancer-therapy: Clinical syndromes, measurable end-points, and potential scoring systems. Int J Radiat Oncol Biol Phys 1995;31: 1187e1203. 25. Takeda T, Takeda A, Kunieda E, et al. Radiation injury after hypofractionated stereotactic radiotherapy for peripheral small lung tumors: Serial changes on CT. Am J Roentgenol 2004;182:1123e1128.

26. Kyas I, Hof H, Debus J, et al. Prediction of radiation-induced changes in the lung after stereotactic body radiation therapy of non-small-cell lung cancer. Int J Radiat Oncol Biol Phys 2007;67:768e774. 27. Graham MV, Purdy JA, Emami B, et al. Clinical dose-volume histogram analysis for pneumonitis after 3D treatment for non-small cell lung cancer (NSCLC). Int J Radiat Oncol Biol Phys 1999;45: 323e329. 28. Gayou O, Parda DS, Miften M. EUCLID: An outcome analysis tool for high-dimensional clinical studies. Phys Med Biol 2007;52: 1705e1719. 29. Novakova-Jiresova A, van Luijk P, van Goor H, et al. Changes in expression of injury after irradiation of increasing volumes in rat lung. Int J Radiat Oncol Biol Phys 2007;67:1510e1518. 30. Chen Y, Williams J, Ding I, et al. Radiation pneumonitis and early circulatory cytokine markers. Semin Radiat Oncol 2002;12:26e33. 31. Ghobadi G, Hogeweg LE, Faber H, et al. Quantifying local radiationinduced lung damage from computed tomography. Int J Radiat Oncol Biol Phys 2010;76:548e556.