- Email: [email protected]
Late regional density changes of the lung after radiotherapy for breast cancer
Radiotherapy and Oncology 90 (2009) 148–152
Contents lists available at ScienceDirect
Radiotherapy and Oncology journal homepage: www.thegreenjournal.com
Pulmonary morbidity
Late regional density changes of the lung after radiotherapy for breast cancer Randi Vågane a,d, Turi Danielsen b, Sophie Dorothea Fosså c,e, Erik Løkkevik c, Dag Rune Olsen a,d,* a
Department of Radiation Biology, Institute for Cancer Research, Rikshospitalet University Hospital, Oslo, Norway Department of Medical Physics, The Norwegian Radium Hospital, Rikshospitalet University Hospital, Oslo, Norway Department of Clinical Cancer Research, The Norwegian Radium Hospital, Rikshospitalet University Hospital, Oslo, Norway d Department of Physics, University of Oslo, Norway e Faculty of Medicine, University of Oslo, Norway b c
a r t i c l e
i n f o
Article history: Received 10 April 2007 Received in revised form 28 December 2007 Accepted 28 December 2007 Available online 11 February 2008 Keywords: Breast cancer Late effects Pulmonary complications Dose–response relationship Computed tomography
a b s t r a c t Background and purpose: To investigate density changes in lung tissue, 3–4 years after postoperative adjuvant radiotherapy for breast cancer, based on dose dependence and regional differences. Material and methods: Sixty-one breast cancer patients, who had received computed tomography (CT) based postoperative radiotherapy, were included. CT scans were performed 35–51 months after start of radiotherapy. Dose information and CT scans from before and after radiotherapy were geometrically aligned in order to analyse changes in air-filled fraction (derived from CT density) as a function of dose for different regions of the lung. Results: Dose-dependent reduction of the air-filled fraction was shown to vary between the different regions of the lung. For lung tissue receiving about 50 Gy, the largest reduction in air-filled fraction was found in the cranial part of the lung. An increased air-filled fraction was observed for lung tissue irradiated to doses below 20 Gy, indicating compensatory response. Conclusions: The treatment-induced change in whole-lung density is a weighted response, involving the different regions, the irradiated volumes, and dose levels to these volumes. Simplistic models may therefore not be appropriate for describing the whole-lung dose–volume–response relationship following inhomogeneous irradiation. Ó 2008 Elsevier Ireland Ltd. All rights reserved. Radiotherapy and Oncology 90 (2009) 148-152
Radiotherapy (RT) of tumours in the thoracic region may be associated with radiation-induced acute and chronic pulmonary symptoms. In addition to lung, the heart and the spinal cord also are at risk. The probability of developing adverse effects in these organs is dose limiting for thoracic radiotherapy. Irradiation of the breast and the regional lymph nodes usually includes irradiation of lung tissue. It is therefore important to identify and describe various late effects to be able to improve and optimise treatment for future patients. In patients irradiated for thoracic tumours, a frequent acute effect is radiation pneumonitis (RP), with symptoms such as fever, dyspnoea, and cough. In a later phase, pulmonary radiation fibrosis may develop. This process depends on size, location, and absorbed dose of the irradiated volume. Pulmonary radiation effects after cancer treatment can be evaluated by computed tomography (CT), as reduced air content due to pulmonary radiation fibrosis can be detected in CT images as areas with increased density. In plain chest radiographs or CT, lung fibrosis is often scored by radiologists [1,2], while some groups have included computer assisted methods for detection of density changes
* Corresponding author. Department of Radiation Biology, Institute for Cancer Research, Radiumhospitalet, Montebello, N-0310 Oslo, Norway E-mail address: [email protected] (D.R. Olsen). 0167-8140/$34.00 Ó 2008 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.radonc.2007.12.031
[3–5]. Several studies have reported regional differences in pulmonary radiation response. Most of these are animal studies, where it is possible to conduct controlled experiments. Results from controlled irradiation of mouse and rat [6–9] have shown that there might be a different radiation response in different regions of the lung. Analyses of data from patients who have undergone radiotherapy for lung cancer and breast cancer have also shown regional differences in radiation response in the lung [10,11]. The purpose of this study was to address regional differences in radiation response, by analysing dose-dependent density changes for different parts of the lung. Materials and methods Patients and treatment In 2003/2004, women with breast cancer were invited to take part in a follow-up study assessing long-term toxicity after treatment for stage II/III breast cancer. The follow-up study consisted of a mailed questionnaire to be completed by the patient, followed by an outpatient visit at the Norwegian Radium Hospital. 318 patients both completed the questionnaire and attended the outpatient follow-up at the clinic, which among several examinations
R. Vågane et al. / Radiotherapy and Oncology 90 (2009) 148–152
included a thoracic CT scanning. All patients provided a written consent form to participate in the study, which was approved by the Ethical Committee of the Health Region South and the Data Inspectorate of Norway. In order to analyse late radiation-induced effects, only patients with a follow-up time of about 3 years or more from start of RT were eligible for the present sub-study. In addition, accessible CT scans (both preRT and postRT) and CT based dose planning data (introduced at the clinic in 2000/2001) were required in order to relate the late effects to radiation dose. In June 2005, 61 women were eligible. They had a median age at followup of 54 (39–70) years, 40 (35–51) months after start of radiotherapy. There were 31 right sided and 30 left sided breast cancers. Surgical treatment consisted of mastectomy/lumpectomy and axillary lymph dissection. Patients with primarily inoperable breast cancer received induction chemotherapy (paclitaxel and epirubicin, or paclitaxel only). Post-operative adjuvant chemotherapy (CMF (cyclophosphamide, methotrexate and 5-fluorouracil)) or (FEC (5-fluorouracil, epirubicin and cyclophosphamide)) and tamoxifen were administered according to the patient’s age at presentation and the tumour’s hormone receptor status. For radiotherapy, the target volume included the breast (after lumpectomy), the chest wall, the ipsilateral axilla, infra and supraclavicular fossa, and the lymph nodes along the ipsilateral internal mammary artery. The treatment planning system Helax-TMS (version 6.0 or higher), utilizing Pencil Beam algorithm, was used for CT based dose calculations. Irradiation was given by a beam arrangement consisting of 4 half-beams (mainly 6 MV), and with 25 fractions of 2 Gy. Two tangential beams were covering the caudal part of the target volume, and one 0° field and one oblique field (110–115°) were covering the cranial part, as indicated in Fig. 1. For the 31 last patients, the two last fractions were delivered by two tangential fields to the breast and/or chest wall only. Fourteen of the patients that underwent lumpectomy also received an electron boost of 5 fractions of 2 Gy (9 or 12 MeV), using a circular field with diameter 5–9 cm, which was not included in the CT-based treatment planning, to a total dose of 60 Gy to the tumour bed. Reconstruction of physical 3D dose distributions (without the boost dose) was performed using the treatment planning system. The physical dose distributions were exported and converted to the normalized total dose, NTD [12], using the linear-quadratic model and a a/b ratio of 3 Gy, to obtain the total biological equivalent dose given in fractions of 2 Gy.
149
Fig. 1. Contours of the body and lung, fused with the dose distribution. Slices are from the cranial, mid and caudal regions of a patient who received 25 fractions of 2 Gy. Arrows indicate the beam arrangement. In the mid slice, both the two cranial and the two caudal beams contribute to the dose distribution.
Density analysis For comparison of the pre-treatment CT scans (preRT CT), follow-up examinations (postRT CT), and the dose data, it was necessary to transform the different sets of data into the same coordinate system and geometrically align features in images of the same tissue taken at different times, in a process referred to as image registration. All handling of CT images and dose data and analysis were performed using routines implemented in IDL (Interactive Data Language 6.2, Research Systems Inc., Boulder, CO). For the dose-CT registration process, correlation of the body contours of CT and dose was used. Matching of postRT and preRT CT was based on correlation of lung contours. Only simple rigid alignment methods were applied. Finally, lung segmentation was performed on the registered volumes, giving three 3D datasets with dose or density information of lung voxels only. Changes in lung density were expressed as changes in air-filled fraction [13], fair = 0.001 NCT, where NCT is the measured CT number in Hounsfield units. An increased density corresponds to a reduced fair. For each lung-voxel, the postRT fair was expressed relative to the preRT fair, and then normalized to the median relative change of the low-dose region 0–12 Gy [14,5], to eliminate some of the influence of breathing level and possible scanner dif-
ferences. The median reduction in fair was calculated for dose intervals of 4 Gy. Dose intervals with less than 100 voxels were excluded. Dose intervals with a median reduction of fair below 50% or above 50% were also excluded, as they mainly would be caused by non-lung voxels in dose intervals with few voxels. Regional differences were analysed after dividing the lung into several parts. For each region, the mean response of the patients was calculated only for dose intervals where at least 40% of the patients were represented. The two lung halves were defined as the ipsilateral or contralateral lung, based on the site of the tumour. In the cranial–caudal direction, the division was based on the treatment technique, with two beams for the caudal part and two beams for the cranial part (Fig. 1). Since the CT slice thickness was 10 mm, only the isocentric slice achieved a substantial dose fraction from all four fields. This mid slice was defined as one region. The lung volumes were also split into one anterior and one posterior region, by locating the position of the most anterior and posterior lung voxels in each slice and dividing the lung at the mean of these positions. In addition, the distal and central regions were based on the positions of the most right and left lung voxels. The central region was defined to cover the mid 50% of
Radiation-induced pulmonary late effects
the lung in each slice, and the distal region covered the remaining lung areas on both sides of the central region. For a more detailed study, the lungs were divided in six compartments: the ipsilateral and contralateral halves of the cranial, mid, and caudal parts. Using the responses and the number of voxels from the dose–volume histograms for these compartments, the weighted response (W) was calculated for each dose interval (d), based on the regional responses (ri) and the number of voxels (ni) in each region (i), according to
, X X WðdÞ ¼ ni ðdÞ ri ðdÞ ni ðdÞ
3500 60000 40000
2500
20000 0 0
2
4
6
2000
1500
ð1Þ
1000
Dose–effect relationships were described using linear regression. Slopes and intercepts were given with 95% confidence intervals, and differences between patient groups were based on the 95% confidence intervals of the regression parameters. The Pearson correlation coefficient was used for evaluation of linear relationships, and p < 0.05 was regarded as statistically significant.
500
i
cranial region mid slice caudal region
80000
3000
Number of voxels
150
i
0 0
10
20
30
40
50
60
NTD [Gy] Fig. 3. Dose–volume histograms for the cranial region, mid slice and caudal region. Mean of all 61 patients, dose intervals of 1 Gy.
Results The dose–effect for the ipsilateral and contralateral lung, as well as for the total lung volume, is shown in Fig. 2. For both the ipsilateral lung and the whole lung, the slope of the linear regression line was significantly positive, with 0.09 (± 0.03)% reduction of fair per Gy. For the contralateral lung, all doses were below 20 Gy. There were large inter-patient variations in the contralateral response, and the reduction of fair had a non-significant negative slope of 0.46 (± 0.53)% per Gy. When analysing the dose–volume histograms for the two lungs (data not shown), only doses below 2 Gy were dominated by contralateral lung voxels. Hence, the negative responses for the whole lung, i.e., the increased fair for the 12– 20 Gy intervals, were mainly based on voxels from the ipsilateral lung, even though the contralateral lung showed a larger increase of fair for these dose intervals. In Fig. 1 examples of the dose distributions for one patient, for slices in the cranial, mid, and caudal regions are presented. Differential dose–volume histograms (mean of all 61 patients) for these three lung regions are shown in Fig. 3. Global maxima for doses be-
5 4
low 5 Gy are mainly from contralateral lung, while the local maxima correspond well to the ipsilateral dose seen in Fig. 1. Dose– effect relations for the three regions are shown in Fig. 4. The cranial part demonstrated the most dose-dependent change of the airfilled fraction, with a significant reduction of 0.16 (± 0.04)% per Gy. For the caudal part, the reduction was also significant but lower, 0.04 (± 0.02)% per Gy. Although the slope of the regression line was low, the reduction in fair for the caudal region was evident in all dose intervals but the lowest. There was a significant difference in dose dependence (slope) between the cranial and caudal regions (p < 0.05). For the mid slice, the reduction in fair was 0.08 (± 0.05)% per Gy. Increased fair occurred for doses up to about 20 Gy in the cranial part, and for doses up to 10 Gy in the mid slice. Dose–effects for the ipsilateral and contralateral parts of the cranial, mid, and caudal regions are shown in Fig. 5. Linear regression showed significant positive slopes for the ipsilateral regions only, with 0.18 (±0.04), 0.10(±0.01), and 0.05(±0.01)% reduction per Gy for the cranial, mid and caudal parts, respectively. The slopes of the cranial- and the mid-contralateral lung were both negative and related to doses below 16 Gy (cranial) and below 8 Gy (mid). However, only the cranial–contralateral slope was significantly negative (0.50 (± 0.36)% per Gy). In the caudal–contralateral region, only the dose interval 0–4 Gy had enough patients to be included, hence no slope was obtained. The weighted response
3 2 8
6
0 0
10
20
30
40
50
60
-1 4
NTD [Gy] -2 -3 -4 -5
ipsilateral lung contralateral lung whole lung
-6 -7
Reduction in fair [%]
Reduction in fair [%]
1
2
0 0 -2
-4
-8
10
20
30
40
50
60
NTD [Gy] cranial region mid slice caudal region
-6
Fig. 2. Reduction in air-filled fraction (fair) as a function of dose for the ipsilateral and contralateral lung, and for the whole lung. The values are mean response of all 61 patients. Dose intervals are of 4 Gy, and error bars indicate the standard error of the mean.
Fig. 4. Mean reduction in air-filled fraction (fair) as a function of dose for all 61 patients for the cranial region, mid slice and caudal region of the lung, in dose intervals of 4 Gy. Error bars indicate the standard error of the mean.
R. Vågane et al. / Radiotherapy and Oncology 90 (2009) 148–152 8 6
Reduction in f air [%]
4 2 0 0
10
20
30
40
50
60
NTD [Gy]
-2
cranial ipsilateral mid ipsilateral caudal ipsilateral cranial contralateral mid contralateral caudal contralateral
-4 -6 -8
Fig. 5. Reduction in air-filled fraction (fair) as a function of dose for the cranial, mid slice and caudal part of the ipsilateral and contralateral lung, in dose intervals of 4 Gy. Error bars indicate the standard error of the mean.
5
4
Reduction in fair [%]
3
2
1
0 0 -1
-2
-3
10
20
30
40
50
60
NTD [Gy] weighted whole lung dominant: cranial ipsilateral mid ipsilateral caudal ipsilateral
Fig. 6. Reduction in air-filled fraction (fair) as a function of dose, for dominant regions (symbols and lines), weighted response and whole-lung response (lines).
(Eq. (1)) of the six regions is presented in Fig. 6 together with the whole-lung response. Symbols were added to indicate the response of the dominating lung region (largest number of voxels) for each dose interval. For doses from 12 to 48 Gy, the response was dominated by the cranial ipsilateral response with values almost monotonously increasing from negative to positive. For the anterior–posterior difference (data not shown), the results indicated that the posterior part had the largest slope, and negative values were seen for doses below 20 Gy. The anterior half had a slightly less steep dose–effect and positive values only, and for most doses, the anterior reduction of fair was larger than in the posterior region. Analysis of the differences between the distal and central parts of the lung (data not shown) indicated a slightly (non-significant) steeper regression line for the central part than for the distal part. However, the reduction in fair was largest for the distal region for most dose intervals, and the central part demonstrated increased fair at doses below 20 Gy. Discussion The results of this study indicate that the dose-dependent reduction of the air-filled fraction (i.e., increased density) depends
151
on the location of the lung tissue. The cranial part of the lung seemed to be the region with the most dose-dependent changes (i.e., largest slope). The density decrease was dominated by the cranial–ipsilateral–posterior parts of the lung, irradiated to doses below 20 Gy. As indicated by Fig. 6 the dose–effect for the whole lung was composed of different response patterns from different regions, based on the amount of voxels for each lung region in the different dose intervals. It is therefore not possible to apply a simple linear model, or a logistic model [13], to describe the dose–effect of the lung. The weighted response follows the whole-lung response for most dose intervals, except for doses where regions were excluded due to too few patients. In most dose intervals, both the weighted and whole-lung responses were close to the values of the cranial–ipsilateral region, due to the inhomogeneous dose distribution. Our data indicated that cranial part of the lung expresses the most pronounced dose–response relationship. However, for lymphoma and breast cancer patients, Theuws et al. [5] found no regional differences in radiosensitivity, except for gravity-related effects in the measuring procedure. For breast cancer patients, early radiological changes in the central part of the lung appeared to be more important for the development of radiation pneumonitis (RP) than changes in the apex [11]. Similarly, for NSCLC patients, a greater risk of RP after irradiation of caudally located lung tumours than irradiation of tumours located in other parts of the lungs has been reported by Seppenwoolde et al. [10]. Their results also indicated that peripheral tissue might be slightly more sensitive than tissue of the central region. This is in agreement with our study, where more pronounced increase in lung density was seen in the distal compared to the central part. Also in experiments on rodents, regional differences in radiation sensitivity are observed. Novakova-Jiresova et al. [15] found that the irradiation of the apical and the left regions of the rat lung had the most severe impact on respiratory function (breathing rate). On the other hand, different studies of RP in mice [6,7] and early DNA damage in rats [8,9] have indicated that the caudal part of the lung is more sensitive to radiation than the cranial region. In our analysis, only variations in dose over the lung have been addressed. However, patients with primarily inoperable breast cancer were also given induction chemotherapy (paclitaxel, epirubicin). In addition, post-operative adjuvant chemotherapy using CMF or FEC and tamoxifen were given according to the patient’s age and the hormonal status. It is more than likely that the chemotherapy might have contributed to the development of lung fibrosis in these patients, and that the observed dose–response is not solely due to the irradiation. On the other hand, there is no reason to assume regional variations in exposure to chemotherapy throughout the lung; the analysis demonstrating regional variation in response to radiation is therefore probably valid although possible impact of chemotherapy has been omitted. The differences seen between the dose–volume histograms of the cranial, mid, and caudal regions Fig. 3 are related to the arrangement of the beams (Fig. 1). Due to the 0° beam in the cranial region, almost all (99.9%) lung voxels in the cranial ipsilateral region received doses above 10 Gy. Cranial voxels receiving high doses (>40 Gy), had the largest density increase and were adjacent to regions receiving medium doses (10–25 Gy). In the caudal ipsilateral lung, covered by two tangential beams, most voxels (81.8%) received less than 10 Gy, and the caudal high-dose voxels were adjacent to voxels receiving negligible radiation doses. The main difference between the response of the high dose regions of the cranial and the caudal regions is probably caused by a large difference in the amount of voxels in those regions irradiated to the different dose levels. Hence, lung tissue irradiated to a high dose might be more radiosensitive if the nearby lung tissue also is irradiated to moderate doses. In this study, the observed volume
152
Radiation-induced pulmonary late effects
effects due to the beam arrangement might overshadow a possible regional difference in radiosensitivity due to physiological heterogeneities of different lung regions. When using reconstructed dose information and CT images to calculate dose dependent density changes, there are several sources of inaccuracies contributing to the final result. During the automatic registration process, mismatch and segmentation errors could not be eliminated, but were assumed to be negligible when averaging over 61 patients. However, with modifications of the registration and calculation processes, the slopes of the linear regression lines and the correlation coefficients will vary. Still, the present data strongly indicate a dose dependent increase in lung density for breast cancer patients treated with radiotherapy. In conclusion, regional variations in dose-dependent reduction of the air-filled fraction were observed. The cranial part of the lung seemed to express the most predominant dose-dependent changes. An increased air-filled fraction was observed in parts of the upper and mid regions where the received dose was below 20 Gy. Lung voxels with doses of about 12–48 Gy were mainly in the cranial–ipsilateral lung region, and hence, the whole-lung response was dominated by the cranial–ipsilateral response. Due to the beam arrangement, the whole-lung dose–response is a weighted response of different regions with different dose distributions and response patterns. Acknowledgement This work was supported by the Research Council of Norway. References [1] Järvenpää R, Holli K, Pitkänen M, et al. Radiological pulmonary findings after breast cancer irradiation: a prospective study. Acta Oncol 2006;45:16–22.
[2] Lind PA, Bylund H, Wennberg B, Svensson C, Svane G. Abnormalities on chest radiographs following radiation therapy for breast cancer. Eur Radiol 2000;10:484–9. [3] Levinson B, Marks LB, Munley MT, et al. Regional dose response to pulmonary irradiation using a manual method. Radiother Oncol 1998; 48:53–60. [4] Rosen II, Fischer TA, Antolak JA, et al. Correlation between lung fibrosis and radiation therapy dose after concurrent radiation therapy and chemotherapy for limited small cell lung cancer. Radiology 2001;221: 614–22. [5] Theuws JCM, Seppenwoolde Y, Kwa SL, 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;47:1201–8. [6] Travis EL, Liao ZX, Tucker SL. Spatial heterogeneity of the volume effect for radiation pneumonitis in mouse lung. Int J Radiat Oncol Biol Phys 1997;38:1045–54. [7] Tucker SL, Liao ZX, Travis EL. Estimation of the spatial distribution of target cells for radiation pneumonitis in mouse lung. Int J Radiat Oncol Biol Phys 1997;38:1055–66. [8] Khan MA, Hill RP, Van Dyk J. Partial volume rat lung irradiation: an evaluation of early DNA damage. Int J Radiat Oncol Biol Phys 1998;40:467–76. [9] Khan MA, Van Dyk J, Yeung IW, Hill RP. Partial volume rat lung irradiation; assessment of early DNA damage in different lung regions and effect of radical scavengers. Radiother Oncol 2003;66:95–102. [10] Seppenwoolde Y, De Jaeger K, Boersma LJ, Belderbos JS, Lebesque JV. Regional differences in lung radiosensitivity after radiotherapy for non-small-cell lung cancer. Int J Radiat Oncol Biol Phys 2004;60:748–58. [11] Wennberg B, Gagliardi G, Sundbom L, Svane G, Lind P. Early response of lung in breast cancer irradiation: radiologic density changes measured by CT and symptomatic radiation pneumonitis. Int J Radiat Oncol Biol Phys 2002;52:1196–206. [12] Lebesque JV, Keus RB. The simultaneous boost technique: the concept of relative normalized total dose. Radiother Oncol 1991;22:45–55. [13] Theuws JCM, Kwa SL, Wagenaar AC, et al. Dose–effect relations for early local pulmonary injury after irradiation for malignant lymphoma and breast cancer. Radiother Oncol 1998;48:33–43. [14] Boersma LJ, Damen EM, de Boer RW, et al. Dose–effect relations for local functional and structural changes of the lung after irradiation for malignant lymphoma. Radiother Oncol 1994;32:201–9. [15] Novakova Jiresova A, Van Luijk P, Van Goor H, Kampinga HH, Coppes RP. Pulmonary radiation injury: identification of risk factors associated with regional hypersensitivity. Cancer Res 2005;65:3568–76.
Contents lists available at ScienceDirect
Radiotherapy and Oncology journal homepage: www.thegreenjournal.com
Pulmonary morbidity
Late regional density changes of the lung after radiotherapy for breast cancer Randi Vågane a,d, Turi Danielsen b, Sophie Dorothea Fosså c,e, Erik Løkkevik c, Dag Rune Olsen a,d,* a
Department of Radiation Biology, Institute for Cancer Research, Rikshospitalet University Hospital, Oslo, Norway Department of Medical Physics, The Norwegian Radium Hospital, Rikshospitalet University Hospital, Oslo, Norway Department of Clinical Cancer Research, The Norwegian Radium Hospital, Rikshospitalet University Hospital, Oslo, Norway d Department of Physics, University of Oslo, Norway e Faculty of Medicine, University of Oslo, Norway b c
a r t i c l e
i n f o
Article history: Received 10 April 2007 Received in revised form 28 December 2007 Accepted 28 December 2007 Available online 11 February 2008 Keywords: Breast cancer Late effects Pulmonary complications Dose–response relationship Computed tomography
a b s t r a c t Background and purpose: To investigate density changes in lung tissue, 3–4 years after postoperative adjuvant radiotherapy for breast cancer, based on dose dependence and regional differences. Material and methods: Sixty-one breast cancer patients, who had received computed tomography (CT) based postoperative radiotherapy, were included. CT scans were performed 35–51 months after start of radiotherapy. Dose information and CT scans from before and after radiotherapy were geometrically aligned in order to analyse changes in air-filled fraction (derived from CT density) as a function of dose for different regions of the lung. Results: Dose-dependent reduction of the air-filled fraction was shown to vary between the different regions of the lung. For lung tissue receiving about 50 Gy, the largest reduction in air-filled fraction was found in the cranial part of the lung. An increased air-filled fraction was observed for lung tissue irradiated to doses below 20 Gy, indicating compensatory response. Conclusions: The treatment-induced change in whole-lung density is a weighted response, involving the different regions, the irradiated volumes, and dose levels to these volumes. Simplistic models may therefore not be appropriate for describing the whole-lung dose–volume–response relationship following inhomogeneous irradiation. Ó 2008 Elsevier Ireland Ltd. All rights reserved. Radiotherapy and Oncology 90 (2009) 148-152
Radiotherapy (RT) of tumours in the thoracic region may be associated with radiation-induced acute and chronic pulmonary symptoms. In addition to lung, the heart and the spinal cord also are at risk. The probability of developing adverse effects in these organs is dose limiting for thoracic radiotherapy. Irradiation of the breast and the regional lymph nodes usually includes irradiation of lung tissue. It is therefore important to identify and describe various late effects to be able to improve and optimise treatment for future patients. In patients irradiated for thoracic tumours, a frequent acute effect is radiation pneumonitis (RP), with symptoms such as fever, dyspnoea, and cough. In a later phase, pulmonary radiation fibrosis may develop. This process depends on size, location, and absorbed dose of the irradiated volume. Pulmonary radiation effects after cancer treatment can be evaluated by computed tomography (CT), as reduced air content due to pulmonary radiation fibrosis can be detected in CT images as areas with increased density. In plain chest radiographs or CT, lung fibrosis is often scored by radiologists [1,2], while some groups have included computer assisted methods for detection of density changes
* Corresponding author. Department of Radiation Biology, Institute for Cancer Research, Radiumhospitalet, Montebello, N-0310 Oslo, Norway E-mail address: [email protected] (D.R. Olsen). 0167-8140/$34.00 Ó 2008 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.radonc.2007.12.031
[3–5]. Several studies have reported regional differences in pulmonary radiation response. Most of these are animal studies, where it is possible to conduct controlled experiments. Results from controlled irradiation of mouse and rat [6–9] have shown that there might be a different radiation response in different regions of the lung. Analyses of data from patients who have undergone radiotherapy for lung cancer and breast cancer have also shown regional differences in radiation response in the lung [10,11]. The purpose of this study was to address regional differences in radiation response, by analysing dose-dependent density changes for different parts of the lung. Materials and methods Patients and treatment In 2003/2004, women with breast cancer were invited to take part in a follow-up study assessing long-term toxicity after treatment for stage II/III breast cancer. The follow-up study consisted of a mailed questionnaire to be completed by the patient, followed by an outpatient visit at the Norwegian Radium Hospital. 318 patients both completed the questionnaire and attended the outpatient follow-up at the clinic, which among several examinations
R. Vågane et al. / Radiotherapy and Oncology 90 (2009) 148–152
included a thoracic CT scanning. All patients provided a written consent form to participate in the study, which was approved by the Ethical Committee of the Health Region South and the Data Inspectorate of Norway. In order to analyse late radiation-induced effects, only patients with a follow-up time of about 3 years or more from start of RT were eligible for the present sub-study. In addition, accessible CT scans (both preRT and postRT) and CT based dose planning data (introduced at the clinic in 2000/2001) were required in order to relate the late effects to radiation dose. In June 2005, 61 women were eligible. They had a median age at followup of 54 (39–70) years, 40 (35–51) months after start of radiotherapy. There were 31 right sided and 30 left sided breast cancers. Surgical treatment consisted of mastectomy/lumpectomy and axillary lymph dissection. Patients with primarily inoperable breast cancer received induction chemotherapy (paclitaxel and epirubicin, or paclitaxel only). Post-operative adjuvant chemotherapy (CMF (cyclophosphamide, methotrexate and 5-fluorouracil)) or (FEC (5-fluorouracil, epirubicin and cyclophosphamide)) and tamoxifen were administered according to the patient’s age at presentation and the tumour’s hormone receptor status. For radiotherapy, the target volume included the breast (after lumpectomy), the chest wall, the ipsilateral axilla, infra and supraclavicular fossa, and the lymph nodes along the ipsilateral internal mammary artery. The treatment planning system Helax-TMS (version 6.0 or higher), utilizing Pencil Beam algorithm, was used for CT based dose calculations. Irradiation was given by a beam arrangement consisting of 4 half-beams (mainly 6 MV), and with 25 fractions of 2 Gy. Two tangential beams were covering the caudal part of the target volume, and one 0° field and one oblique field (110–115°) were covering the cranial part, as indicated in Fig. 1. For the 31 last patients, the two last fractions were delivered by two tangential fields to the breast and/or chest wall only. Fourteen of the patients that underwent lumpectomy also received an electron boost of 5 fractions of 2 Gy (9 or 12 MeV), using a circular field with diameter 5–9 cm, which was not included in the CT-based treatment planning, to a total dose of 60 Gy to the tumour bed. Reconstruction of physical 3D dose distributions (without the boost dose) was performed using the treatment planning system. The physical dose distributions were exported and converted to the normalized total dose, NTD [12], using the linear-quadratic model and a a/b ratio of 3 Gy, to obtain the total biological equivalent dose given in fractions of 2 Gy.
149
Fig. 1. Contours of the body and lung, fused with the dose distribution. Slices are from the cranial, mid and caudal regions of a patient who received 25 fractions of 2 Gy. Arrows indicate the beam arrangement. In the mid slice, both the two cranial and the two caudal beams contribute to the dose distribution.
Density analysis For comparison of the pre-treatment CT scans (preRT CT), follow-up examinations (postRT CT), and the dose data, it was necessary to transform the different sets of data into the same coordinate system and geometrically align features in images of the same tissue taken at different times, in a process referred to as image registration. All handling of CT images and dose data and analysis were performed using routines implemented in IDL (Interactive Data Language 6.2, Research Systems Inc., Boulder, CO). For the dose-CT registration process, correlation of the body contours of CT and dose was used. Matching of postRT and preRT CT was based on correlation of lung contours. Only simple rigid alignment methods were applied. Finally, lung segmentation was performed on the registered volumes, giving three 3D datasets with dose or density information of lung voxels only. Changes in lung density were expressed as changes in air-filled fraction [13], fair = 0.001 NCT, where NCT is the measured CT number in Hounsfield units. An increased density corresponds to a reduced fair. For each lung-voxel, the postRT fair was expressed relative to the preRT fair, and then normalized to the median relative change of the low-dose region 0–12 Gy [14,5], to eliminate some of the influence of breathing level and possible scanner dif-
ferences. The median reduction in fair was calculated for dose intervals of 4 Gy. Dose intervals with less than 100 voxels were excluded. Dose intervals with a median reduction of fair below 50% or above 50% were also excluded, as they mainly would be caused by non-lung voxels in dose intervals with few voxels. Regional differences were analysed after dividing the lung into several parts. For each region, the mean response of the patients was calculated only for dose intervals where at least 40% of the patients were represented. The two lung halves were defined as the ipsilateral or contralateral lung, based on the site of the tumour. In the cranial–caudal direction, the division was based on the treatment technique, with two beams for the caudal part and two beams for the cranial part (Fig. 1). Since the CT slice thickness was 10 mm, only the isocentric slice achieved a substantial dose fraction from all four fields. This mid slice was defined as one region. The lung volumes were also split into one anterior and one posterior region, by locating the position of the most anterior and posterior lung voxels in each slice and dividing the lung at the mean of these positions. In addition, the distal and central regions were based on the positions of the most right and left lung voxels. The central region was defined to cover the mid 50% of
Radiation-induced pulmonary late effects
the lung in each slice, and the distal region covered the remaining lung areas on both sides of the central region. For a more detailed study, the lungs were divided in six compartments: the ipsilateral and contralateral halves of the cranial, mid, and caudal parts. Using the responses and the number of voxels from the dose–volume histograms for these compartments, the weighted response (W) was calculated for each dose interval (d), based on the regional responses (ri) and the number of voxels (ni) in each region (i), according to
, X X WðdÞ ¼ ni ðdÞ ri ðdÞ ni ðdÞ
3500 60000 40000
2500
20000 0 0
2
4
6
2000
1500
ð1Þ
1000
Dose–effect relationships were described using linear regression. Slopes and intercepts were given with 95% confidence intervals, and differences between patient groups were based on the 95% confidence intervals of the regression parameters. The Pearson correlation coefficient was used for evaluation of linear relationships, and p < 0.05 was regarded as statistically significant.
500
i
cranial region mid slice caudal region
80000
3000
Number of voxels
150
i
0 0
10
20
30
40
50
60
NTD [Gy] Fig. 3. Dose–volume histograms for the cranial region, mid slice and caudal region. Mean of all 61 patients, dose intervals of 1 Gy.
Results The dose–effect for the ipsilateral and contralateral lung, as well as for the total lung volume, is shown in Fig. 2. For both the ipsilateral lung and the whole lung, the slope of the linear regression line was significantly positive, with 0.09 (± 0.03)% reduction of fair per Gy. For the contralateral lung, all doses were below 20 Gy. There were large inter-patient variations in the contralateral response, and the reduction of fair had a non-significant negative slope of 0.46 (± 0.53)% per Gy. When analysing the dose–volume histograms for the two lungs (data not shown), only doses below 2 Gy were dominated by contralateral lung voxels. Hence, the negative responses for the whole lung, i.e., the increased fair for the 12– 20 Gy intervals, were mainly based on voxels from the ipsilateral lung, even though the contralateral lung showed a larger increase of fair for these dose intervals. In Fig. 1 examples of the dose distributions for one patient, for slices in the cranial, mid, and caudal regions are presented. Differential dose–volume histograms (mean of all 61 patients) for these three lung regions are shown in Fig. 3. Global maxima for doses be-
5 4
low 5 Gy are mainly from contralateral lung, while the local maxima correspond well to the ipsilateral dose seen in Fig. 1. Dose– effect relations for the three regions are shown in Fig. 4. The cranial part demonstrated the most dose-dependent change of the airfilled fraction, with a significant reduction of 0.16 (± 0.04)% per Gy. For the caudal part, the reduction was also significant but lower, 0.04 (± 0.02)% per Gy. Although the slope of the regression line was low, the reduction in fair for the caudal region was evident in all dose intervals but the lowest. There was a significant difference in dose dependence (slope) between the cranial and caudal regions (p < 0.05). For the mid slice, the reduction in fair was 0.08 (± 0.05)% per Gy. Increased fair occurred for doses up to about 20 Gy in the cranial part, and for doses up to 10 Gy in the mid slice. Dose–effects for the ipsilateral and contralateral parts of the cranial, mid, and caudal regions are shown in Fig. 5. Linear regression showed significant positive slopes for the ipsilateral regions only, with 0.18 (±0.04), 0.10(±0.01), and 0.05(±0.01)% reduction per Gy for the cranial, mid and caudal parts, respectively. The slopes of the cranial- and the mid-contralateral lung were both negative and related to doses below 16 Gy (cranial) and below 8 Gy (mid). However, only the cranial–contralateral slope was significantly negative (0.50 (± 0.36)% per Gy). In the caudal–contralateral region, only the dose interval 0–4 Gy had enough patients to be included, hence no slope was obtained. The weighted response
3 2 8
6
0 0
10
20
30
40
50
60
-1 4
NTD [Gy] -2 -3 -4 -5
ipsilateral lung contralateral lung whole lung
-6 -7
Reduction in fair [%]
Reduction in fair [%]
1
2
0 0 -2
-4
-8
10
20
30
40
50
60
NTD [Gy] cranial region mid slice caudal region
-6
Fig. 2. Reduction in air-filled fraction (fair) as a function of dose for the ipsilateral and contralateral lung, and for the whole lung. The values are mean response of all 61 patients. Dose intervals are of 4 Gy, and error bars indicate the standard error of the mean.
Fig. 4. Mean reduction in air-filled fraction (fair) as a function of dose for all 61 patients for the cranial region, mid slice and caudal region of the lung, in dose intervals of 4 Gy. Error bars indicate the standard error of the mean.
R. Vågane et al. / Radiotherapy and Oncology 90 (2009) 148–152 8 6
Reduction in f air [%]
4 2 0 0
10
20
30
40
50
60
NTD [Gy]
-2
cranial ipsilateral mid ipsilateral caudal ipsilateral cranial contralateral mid contralateral caudal contralateral
-4 -6 -8
Fig. 5. Reduction in air-filled fraction (fair) as a function of dose for the cranial, mid slice and caudal part of the ipsilateral and contralateral lung, in dose intervals of 4 Gy. Error bars indicate the standard error of the mean.
5
4
Reduction in fair [%]
3
2
1
0 0 -1
-2
-3
10
20
30
40
50
60
NTD [Gy] weighted whole lung dominant: cranial ipsilateral mid ipsilateral caudal ipsilateral
Fig. 6. Reduction in air-filled fraction (fair) as a function of dose, for dominant regions (symbols and lines), weighted response and whole-lung response (lines).
(Eq. (1)) of the six regions is presented in Fig. 6 together with the whole-lung response. Symbols were added to indicate the response of the dominating lung region (largest number of voxels) for each dose interval. For doses from 12 to 48 Gy, the response was dominated by the cranial ipsilateral response with values almost monotonously increasing from negative to positive. For the anterior–posterior difference (data not shown), the results indicated that the posterior part had the largest slope, and negative values were seen for doses below 20 Gy. The anterior half had a slightly less steep dose–effect and positive values only, and for most doses, the anterior reduction of fair was larger than in the posterior region. Analysis of the differences between the distal and central parts of the lung (data not shown) indicated a slightly (non-significant) steeper regression line for the central part than for the distal part. However, the reduction in fair was largest for the distal region for most dose intervals, and the central part demonstrated increased fair at doses below 20 Gy. Discussion The results of this study indicate that the dose-dependent reduction of the air-filled fraction (i.e., increased density) depends
151
on the location of the lung tissue. The cranial part of the lung seemed to be the region with the most dose-dependent changes (i.e., largest slope). The density decrease was dominated by the cranial–ipsilateral–posterior parts of the lung, irradiated to doses below 20 Gy. As indicated by Fig. 6 the dose–effect for the whole lung was composed of different response patterns from different regions, based on the amount of voxels for each lung region in the different dose intervals. It is therefore not possible to apply a simple linear model, or a logistic model [13], to describe the dose–effect of the lung. The weighted response follows the whole-lung response for most dose intervals, except for doses where regions were excluded due to too few patients. In most dose intervals, both the weighted and whole-lung responses were close to the values of the cranial–ipsilateral region, due to the inhomogeneous dose distribution. Our data indicated that cranial part of the lung expresses the most pronounced dose–response relationship. However, for lymphoma and breast cancer patients, Theuws et al. [5] found no regional differences in radiosensitivity, except for gravity-related effects in the measuring procedure. For breast cancer patients, early radiological changes in the central part of the lung appeared to be more important for the development of radiation pneumonitis (RP) than changes in the apex [11]. Similarly, for NSCLC patients, a greater risk of RP after irradiation of caudally located lung tumours than irradiation of tumours located in other parts of the lungs has been reported by Seppenwoolde et al. [10]. Their results also indicated that peripheral tissue might be slightly more sensitive than tissue of the central region. This is in agreement with our study, where more pronounced increase in lung density was seen in the distal compared to the central part. Also in experiments on rodents, regional differences in radiation sensitivity are observed. Novakova-Jiresova et al. [15] found that the irradiation of the apical and the left regions of the rat lung had the most severe impact on respiratory function (breathing rate). On the other hand, different studies of RP in mice [6,7] and early DNA damage in rats [8,9] have indicated that the caudal part of the lung is more sensitive to radiation than the cranial region. In our analysis, only variations in dose over the lung have been addressed. However, patients with primarily inoperable breast cancer were also given induction chemotherapy (paclitaxel, epirubicin). In addition, post-operative adjuvant chemotherapy using CMF or FEC and tamoxifen were given according to the patient’s age and the hormonal status. It is more than likely that the chemotherapy might have contributed to the development of lung fibrosis in these patients, and that the observed dose–response is not solely due to the irradiation. On the other hand, there is no reason to assume regional variations in exposure to chemotherapy throughout the lung; the analysis demonstrating regional variation in response to radiation is therefore probably valid although possible impact of chemotherapy has been omitted. The differences seen between the dose–volume histograms of the cranial, mid, and caudal regions Fig. 3 are related to the arrangement of the beams (Fig. 1). Due to the 0° beam in the cranial region, almost all (99.9%) lung voxels in the cranial ipsilateral region received doses above 10 Gy. Cranial voxels receiving high doses (>40 Gy), had the largest density increase and were adjacent to regions receiving medium doses (10–25 Gy). In the caudal ipsilateral lung, covered by two tangential beams, most voxels (81.8%) received less than 10 Gy, and the caudal high-dose voxels were adjacent to voxels receiving negligible radiation doses. The main difference between the response of the high dose regions of the cranial and the caudal regions is probably caused by a large difference in the amount of voxels in those regions irradiated to the different dose levels. Hence, lung tissue irradiated to a high dose might be more radiosensitive if the nearby lung tissue also is irradiated to moderate doses. In this study, the observed volume
152
Radiation-induced pulmonary late effects
effects due to the beam arrangement might overshadow a possible regional difference in radiosensitivity due to physiological heterogeneities of different lung regions. When using reconstructed dose information and CT images to calculate dose dependent density changes, there are several sources of inaccuracies contributing to the final result. During the automatic registration process, mismatch and segmentation errors could not be eliminated, but were assumed to be negligible when averaging over 61 patients. However, with modifications of the registration and calculation processes, the slopes of the linear regression lines and the correlation coefficients will vary. Still, the present data strongly indicate a dose dependent increase in lung density for breast cancer patients treated with radiotherapy. In conclusion, regional variations in dose-dependent reduction of the air-filled fraction were observed. The cranial part of the lung seemed to express the most predominant dose-dependent changes. An increased air-filled fraction was observed in parts of the upper and mid regions where the received dose was below 20 Gy. Lung voxels with doses of about 12–48 Gy were mainly in the cranial–ipsilateral lung region, and hence, the whole-lung response was dominated by the cranial–ipsilateral response. Due to the beam arrangement, the whole-lung dose–response is a weighted response of different regions with different dose distributions and response patterns. Acknowledgement This work was supported by the Research Council of Norway. References [1] Järvenpää R, Holli K, Pitkänen M, et al. Radiological pulmonary findings after breast cancer irradiation: a prospective study. Acta Oncol 2006;45:16–22.
[2] Lind PA, Bylund H, Wennberg B, Svensson C, Svane G. Abnormalities on chest radiographs following radiation therapy for breast cancer. Eur Radiol 2000;10:484–9. [3] Levinson B, Marks LB, Munley MT, et al. Regional dose response to pulmonary irradiation using a manual method. Radiother Oncol 1998; 48:53–60. [4] Rosen II, Fischer TA, Antolak JA, et al. Correlation between lung fibrosis and radiation therapy dose after concurrent radiation therapy and chemotherapy for limited small cell lung cancer. Radiology 2001;221: 614–22. [5] Theuws JCM, Seppenwoolde Y, Kwa SL, 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;47:1201–8. [6] Travis EL, Liao ZX, Tucker SL. Spatial heterogeneity of the volume effect for radiation pneumonitis in mouse lung. Int J Radiat Oncol Biol Phys 1997;38:1045–54. [7] Tucker SL, Liao ZX, Travis EL. Estimation of the spatial distribution of target cells for radiation pneumonitis in mouse lung. Int J Radiat Oncol Biol Phys 1997;38:1055–66. [8] Khan MA, Hill RP, Van Dyk J. Partial volume rat lung irradiation: an evaluation of early DNA damage. Int J Radiat Oncol Biol Phys 1998;40:467–76. [9] Khan MA, Van Dyk J, Yeung IW, Hill RP. Partial volume rat lung irradiation; assessment of early DNA damage in different lung regions and effect of radical scavengers. Radiother Oncol 2003;66:95–102. [10] Seppenwoolde Y, De Jaeger K, Boersma LJ, Belderbos JS, Lebesque JV. Regional differences in lung radiosensitivity after radiotherapy for non-small-cell lung cancer. Int J Radiat Oncol Biol Phys 2004;60:748–58. [11] Wennberg B, Gagliardi G, Sundbom L, Svane G, Lind P. Early response of lung in breast cancer irradiation: radiologic density changes measured by CT and symptomatic radiation pneumonitis. Int J Radiat Oncol Biol Phys 2002;52:1196–206. [12] Lebesque JV, Keus RB. The simultaneous boost technique: the concept of relative normalized total dose. Radiother Oncol 1991;22:45–55. [13] Theuws JCM, Kwa SL, Wagenaar AC, et al. Dose–effect relations for early local pulmonary injury after irradiation for malignant lymphoma and breast cancer. Radiother Oncol 1998;48:33–43. [14] Boersma LJ, Damen EM, de Boer RW, et al. Dose–effect relations for local functional and structural changes of the lung after irradiation for malignant lymphoma. Radiother Oncol 1994;32:201–9. [15] Novakova Jiresova A, Van Luijk P, Van Goor H, Kampinga HH, Coppes RP. Pulmonary radiation injury: identification of risk factors associated with regional hypersensitivity. Cancer Res 2005;65:3568–76.