UvA-DARE is a service provided by the library of the University of Amsterdam (https://dare.uva.nl)
UvA-DARE (Digital Academic Repository)
Radiation induced lung damage
Seppenwoolde, Y.
Publication date
2002
Document Version
Final published version
Link to publication
Citation for published version (APA):
Seppenwoolde, Y. (2002). Radiation induced lung damage.
General rights
It is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), other than for strictly personal, individual use, unless the work is under an open content license (like Creative Commons).
Disclaimer/Complaints regulations
If you believe that digital publication of certain material infringes any of your rights or (privacy) interests, please let the Library know, stating your reasons. In case of a legitimate complaint, the Library will make the material inaccessible and/or remove it from the website. Please Ask the Library: https://uba.uva.nl/en/contact, or a letter to: Library of the University of Amsterdam, Secretariat, Singel 425, 1012 WP Amsterdam, The Netherlands. You will be contacted as soon as possible.
Radiationn induced
lungg damage
Radiationn induced
lungg damage
Radiationn induced lung damage
ACADEMISCHH PROEFSCHRIFT
terr verkrijging van de graad van doctor aann de Universiteit van Amsterdam opp gezag van de Rector Magnificus prof.mr. P.F. van der Heijden tenn overstaan van een door het college voor promoties ingestelde commissie,, in het openbaar te verdedigen in de Aula der Universiteit
opp dinsdag 28 mei 2002, te 12:00 uur
Promotorr prof.dr. G.M.M. Bartelink Copromotor:: dr. J.V. Lebesque Faculteit:: Geneeskunde
Thee study described in this thesis was performed at the Radiotherapy Department of The Netherlandss Cancer Institute (Antoni van Leeuwenhoek Hospital) in Amsterdam and was financiallyy supported by the Dutch Cancer Society (grant 99-2043 and a gift for printing this thesis). .
©© 2002 Yvette Seppenwoolde
Coverr design and lay-out: Yvette Seppenwoolde. Fractal of lungs generated using Fractasketchh v1.8 (C) Peter Van Roy
Contents s
Chapterr 1 Introduction 9 Chapterr 2 Irradiation of the lung: dose-volume effects 15
Chapterr 3 Radiation dose-effect relations and local recovery 31 inn perfusion for patients with non-email cell lung
cancer r
Chapterr 4 Pulmonary function following high dose radiotherapy 45 off non-small cell lung cancer
Chapterr 5 Comparing différent NTCP models that predict the 59 incidencee of radiation pneumonitis
Chapterr 6 Optimizing radiation treatment plans for lung cancer 77 usingg lung perfusion information
Chapterr 7 Portal imaging to assess setup errors, tumor motion 97 andd tumor shrinkage during conformal radiotherapy of
non-smalll cell lung cancer
Chapterr 8 Precise and real-time measurement of 3D tumor motion 113 inn lung due to breathing and heartbeat, measured
duringg radiotherapy Chapterr 9 Discussion 131 Appendicess 143 Referencess 159 Summaryy 175 Samenvattingg 177 Dankwoordd 181 Curriculumm Vftae 185
Chapterr 1
Introduction n
Introduction n
Althoughh lung cancer is the most preventable malignancy in humans (trends in lung cancerr incidence are related to trends in past smoking behavior), its tnddenoB and mortaftyy is very high. In the period 1989-1997 me incidence among men has declined graduallyy from 109 to 93 per 100,000 person years (Janssen-Heijnen 2001). However, this beneficiall trend was partly offset by an increase of the incidence among women: from 18 toto 23 per 100.000 person years. In me Netherlands the incidence of lung cancer was high inn comparison to other European countries. Although the incidence and mortality decreased,, m 1997 still almost 20 men died each day of lung cancer in the Netherlands. Amongg women the end of the increase has not been reached and in 1997 over 5 women diedd each day of lung cancer (Janssen-Heijnen 2001). m the year 2000,8800 new cases andd 8600 deaths due to rung cancer were reported (Dutch Cancer Registration 2000).
Radiotherapyy for lung cancer patients
Non-smalll cell lung cancer (NSCLC) is diagnosed in eighty percent of the lung cancer patients. Off these patients 20-30% can be treated with surgery, while 70-80%, is inoperable due to locallyy advanced disease, a poor performance index, distant metastases or co-morbkJity. For 30%% of the inoperable patients, the only potentially curative therapy is radical irradiation of the tumorr (Dutch Cancer Registration 2000). Patients treated with radiotherapy alone have a mediann survival of merely 8 to 10 months. The 2-year survival is about 10 to 20% and the 5-yearr survival is only 3 to 7% (Perez 1986,1987, Schaake-Koning 1992).
Locall failure of radiotherapy with conventional doses is a major cause for the disappointing curee rate (Arriagada 1991). To improve clinical outcome new treatment approaches were designed,, consisting of combined radiotherapy and chemotherapy. Another approach is to use highh dose radiotherapy Hi order to improve focal control. A study based on tumor control probabilityy model calculations (Martel 1999) estimated that the dose to achieve a 50% probabilityy of tumor control might be in the order of 84 Gy. it is obvious that with an increasing radiotherapyy dose, the incidence of radiation-induced damage to surrounding healthy tissue willl increase as well.
Dosee escalation
Inn several RTOG trials (Perez 1986,1987, Cox 1990,1993, Sause 1995), patients with higher radiotherapyy doses showed a greater proportion of complete responses and subsequently improvedd survival. Total doses higher than the conventional 60 Gy are associated with improvedd focal tumor control (Byhardt 1995). Preliminary results of prospective trials using high-dosee 30 conforms! radiotherapy for NSCLC patients are suggesting that improvement of locall control can be reached by delivering higher radiation doses (Cox 1990, Armstrong 1993, Robertsonn 1997, Befderbos 2000, Mehta 2001, Hayman 2001).
Normall tissue damage
Increasingg the prescribed dose to the tumor will inherently increase dose to the surrounding tissuetissue in the thorax, causing extra radiation induced damage to the surrounding healthy tissue. inn addition to radiation pneumonitis, which is a binary type of complication, it is clinically relevantt to evaluate and predict the graded response of lung tissue by measuring changes in overalll pulmonary function. Graded response is also reflected in local changes in lung tissue densityy and lung perfusion.
ChapterChapter 1
Locall dose effects
Locall damage to lung tissue can be measured with 30 imaging techniques. Changes in lung perfusionn (and thus the distribution of the lung capillaries) can be measured by comparison of follow-upp 3D Single Photon Emission Computed Tomography (SPECT) scans with the pre-treatmentt scan. To obtain an image of the 3D local lung perfusion, ""Technetium-laoefled (T1/2
== 6 h) albumin micro aggregates are injected intravenously and the relatively targe albumin getss trapped in the small capillaries in the lungs. A dual head gamma camera rotates around thee patient and obtains 30 images per camera head at 6 intervals. Using a filtered back projectionn algorithm, trans-axial slices of the 3D perfusion can be made. These images can be matchedd to the CT scan of the patient (Kwa 1998b) and correlated with the 3D dose distribution.. With a pre-treatment and post-treatment SPECT-scan the local changes in lung perfusionn can be calculated. In previous studies (Boersma 1994, Marks 1997b, Theuws 1998a,, Garipagaoglu 1999, Theuws 2000) dose effect relations for radiation induced perfusion damagee were obtained for patients without intrathoracic tumor or pre-existent pulmonary diseasee ('healthy' lungs). For lung cancer patients the presence of reperfusion (Marks 1995) andd pre-existent pulmonary disease complicates the determination of a dose-effect relation. Thee SPECT lung perfusion information on local functionality of lung tissue provides additional informationn to design a treatment plan that minimizes the complication risk for perfusion damagee (Marks 1995) because the high dose regions can be directed through non-functioning lungg tissue (Marks 1993). The group of Marks (1999) suggested that the perfusion weighted lungg dose-volume histogram (where the volume receiving a certain dose is weighted with the averagee perfusion in that dose-region) could be a valuable tool in designing the optimal RT plan. .
Pulmonaryy function
Radiationn induced lung damage on the overall lung function was measured by repeated pulmonaryy function tests (PFTs). The vital capacity, the forced expiratory volume in one second,, the alveolar volume and diffusion capacity for carbon monoxide, corrected for hemoglobinn (VC, FEV1t VA and T^coc respectively) were determined with spirometry, before
andd after radiotherapy for patients with healthy lungs (Theuws 1998b, 1999). The presence of reperfusionn and pre-existent pulmonary disease in NSCLC patients can obscure the outcome off the measurements. Until now, the radiation-induced reduction in pulmonary function was correlatedd with 'pure' dose parameters. For patients with 'healthy* lungs this was considered to bee a good strategy because their lung tissue was functioning homogeneously (Theuws 1998b).. This is different for patients with lung cancer because of the inhomogeneity of their lungg tissue: none of the dose-volume parameters that correlated with the risk of pneumonitis weree successful in predicting changes in pulmonary function in NSCLC patiënte treated to high dosee (Allen 2001).
Radiationn pneumonitis
Radiationn pneumonitis is a more serious complication that may occur after radiotherapy. It becomess manifest between one and six months after treatment The clinical symptoms of radiationn pneumonitis range from mild symptoms such as fever, dyspnea and cough to death duee to respiratory failure. In patients who survive this pneumonitis phase, the type and severity
Introduction Introduction
off the early response may have bearing on the type and severity of the ensuing late response. Thee amount of irradiated volume and radiation dose have an effect on radiation-induced lung injury.. The severity and probability of developing radiation pneumonitis for an individual patient beforee treatment can be estimated using simple parameters as the mean lung dose or the volumee of lung receiving more tiran a threshold dose (Marks 1997a, Kwa 1998a, Graham 19919).. Although all the studied parameters showed correlationn with the incidence of radiation pneumonitis,, until now no consensus has been reached on the best model to predict radiation pneumonitis. .
Treatmentt optimization
Reducingg the volume of irradiated surrounding tissue, whilst keeping the dose to the tumor constantt decreases the incidence of normal tissue complications. There are several ways to reducee the irradiated lung volume: for example by using 3D conformal fields instead of conventionall radiation fields. This reduces the volume of irradiated normal tissue by 35% (Armstrongg 1993). Excluding irradiation of clinical and radiological uninvolved lymph nodes (electivee nodal irradiation) decreases lung toxicity (Belderbos 1997, McGibney 1999) because off the resulting smaller field sizes. Prospective studies (Robertson 1997, Hayman 2001) suggestt that it is safe to exclude the clinical and radiological uninvolved nodal areas.
Marginn reduction
Anotherr way to decrease field sizes is to reduce the safety margins around the clinical target volumee (CTV), by controlling the factors that determine the size of the margin between CTV andd planning target volume (PTV): systematic (setup) errors and tumor motion.
Setupp errors
Thee introduction of electronic portal imaging devices has enabled images to be acquired of the patientt in the treatment position. These images can be compared to digitally reconstructed radiographss (DRR) derived from the planning CT scan, to correct errors In setup. Systematic errorss occur mainly during the preparation of the treatment. These can be due to several factorss such as target delineation errors, position of the patient and tumor when the planning CTT scan is acquired or the use of simulator images rather than ORRs for setup reference imagess (Bel 1994, de Boer 2001). Setup errors are mainly due to day-to-day variations in patientt and organ position. To identify setup inaccuracies using thoracic portal images, comparisonn must be made to areas of relative stability. Samson et al. (1999) demonstrated thatt thoracic wall, trachea and clavicle are the most stable structures and the comparison of thee position of two of these landmarks in the reference image and the portal image, produces thee most reliable comparison. The extent of systematic and random (setup) errors can differ perr institute as it depends on imaging, immobilization and verification techniques.
Tumorr motion
Tumorss in lung tissue move due to breathing and/or heartbeat. When a part of the tumor movess out of the irradiation field, this part gets less dose than planned and the surrounding normall tissue receives a higher dose. Irradiation of lung cancer is done with the patient breathingg freely and therefore a targe margin is drawn around the tumor to ensure accurate targett coverage (ICRU Report 50 1993). In the absence of precise data, the size of the marginss is estimated rather roughly during fluoroscopy measurements (Ekberg 1998). Strategiess to compensate for breathing motion include the use of real-time tumor tracking (Shrratoo 2000a), gated radiotherapy (Ohara 1989), abdominal pressure (Blomgren 1995),
ChapterChapter 1
voluntaryy breath hold at variable phases of the respiratory cycle (Rosenzweig 2000, Kim 2001) andd active breathing control (ABC), (Wong 1999). In CT-scans, significant under- and overestimationn of the tumor size and distortions of the shape of the tumor can be due to breathingg (Baiter 1996). To construct a representative pfanning-CT scan it is important that the magnitudee and the direction of tumor motion can be modelled.
Aimm of the study
Thee aim of this study is two-fold:
1.. Factors associated with radiation induced pulmonary damage will be identified and quantifiedd by measuring local and overall dose-effects.
2.. The results of the first objective will be used to optimize the treatment, in combination with modellingg of factors that hinder accurate dose delivery such as patient setup inaccuracy and tumorr motion.
Chapterr 2
Irradiationn of the lung:
dose-volumee effects
Irradiationn of the lung: dose-volume effects
Manyy factors like fractionation, overall treatment time and patient specific aspects are importantt when studying and quantifying the effects of partial king irradiation. The tocat reactionss of lung tissue to irradiation are ascribed with reg^ to the dcee-volume effect Differentt models that are used to predict the incidence of radiation pneumonitis and the influencee of irradiation on the overal king function are discussed. The easy-to-cafcuiate meann king dose (MLD) and the volume irradiated with more than 20 Gy (V20) can both hee used to predict the incidence of radiation pneumonitis. These parameters represent twoo extremes m underfying local dose-effect r e t a t i ^
dinteaflyy applied treatment plane show a high «jrrelatton between the V20 and trie MLD, soo that the decision for the 'bes? underlying ioeaf dose-effect relation should be based on tiletile analysis of additional patient data. Dose-escalation studies and multi-center co-operationn wffl create more possibflities to investigate all confounding factors concerning lungg irradiation.
Introduction n
Locall failure is a major cause for thé disappointing cure rate of patiënte with inoperable noh-smal!! cell lung cancer (NSCLC) treated with radiotherapy. In a study of Martel et al. (1999), basedd on Tumor control probability (TCP) model calculations, ft was found that the dose to achievee a significant probability of tumor control rnay be targ^ (in the ordered 84 Gy) for longer (>> 30 months) local progression-free survival. However, the maximum safe radiation dose that cann be delivered to most thoracic tumors is limited by the tolerance of lung tissue.
Generally,, the radiation response in lung is divided into two distinct phases; early radiation effectss (radiation pneumonitis) become manifest between 1 and 8 months after radiotherapy. Thee clinical symptoms of radiation pneumonitis range from mild symptoms such as fever, dyspneaa and cough to death from respiratory failure. Late effects develop in some patients fromm 6 months after treatment onwards (Marks 1994). Consequently, both a recovery from the earlyy inflammatory response and a development of fibrosis may be observed between 3 and 188 months. In the early phase, inflammatory cells and interstitial and alveolar exudates are absorbedd and the epithelium and the microvasculature reverts to normal (Roswit 1977). The processs of fibrosis is characterized by a progressive thickening of the alveolar septa and vessell walls with bundles of elastic fibers and collagen, and consequently decreases local and overalll lung function. The fibrosis phase remains constant from a little over a year to 10 years followw up in a study of Skoczytas et al. (2000). Clinically, radiation pneumonitis mostly precedess fibrosis, but both processes can develop independently and may show variation in degreee and extent of the radiation response.
Besidess radiation pneumonitis and radiation fibrosis, reduction in pulmonary function is reportedd following radiotherapy for breast cancer, malignant lymphoma and lung cancer as earlyy as 4-8 weeks after radiotherapy. Generally, a decrease of lung volumes in ail compartmentss is measured, which is maximal 6-9 months after irradiation. At about the same time,, gas exchange abnormalities occur which are more or (ess proportional to the decrease inn lung volumes. Thereafter, some recovery in pulmonary function is reported, although this is onlyy minimal for the diffusion capacity, indicating true diffusion problems.
ChapterChapter 2
Duee to these pulmonary side effects, the total dose that can safely be delivered to the thorax iss limited and therefore tumor control is hampered. The majority of patiënte with NSCLC have marginall lung function already before radiotherapy due to pre-existent pulmonary disease. Patientss with chronic obstructive pulmonary disease (COPD) are more likely to have emphysemaa with a lower CT-density throughout the lungs, and their lung perfusion is inhorhogeneous.. Some patients suffer from bullous disease and have lung regions where locallyy the density is very low. These bullous regions are neither well ventilated nor perfused. Furthermore,, the ventilation and perfusion can be obstructed by atelectasis, infiltrates or tumor.. For these patients it is important to take into account every single factor that could limit radiationn damage to the remaining (healthy) lung tissue, while in the mean time the tumor is adequatelyy irradiated. During the last years, the use of 3D conformal radiotherapy has increased,, resulting in smaller field sizes and better tumor coverage. However, to optimize the treatmentt plans and escalate the dose, It is essential to know the relation between the 3D treatmentt plan and complications.
Pulmonaryy damage, general aspects
Thee probability and the severity of radiation-induced lung damage mainly depend on the radiationn dose, the amount of the irradiated volume and the fractionation schedule. Van Dyk ett al. (1981) made the first attempt to relate the incidence of radiation pneumonitis (as defined byy clinical symptoms and characteristic radiographic changes) to the absolute dose in the lung afterr total lung irradiation. They observed a steep dose-response relation that rose from 5% to 50%% complication probability with an increase of the (single) dose from 8.2 to 9.3 Gy.
Fractionationn and overall treatment time
Whenn multiple fractions are given, an impressive sparing effect of fractionation is seen; no casess of pneumonitis were reported among 110 patients who received 15 to 25 Gy in fractions off 1.5 and 1.75 Gy to both lungs (Newton 1969, Rab 1976, Breur 1978, Van Dyk 1981). To enablee comparison of clinical studies with different fractionation schedules, several methods cann be used that convert the different radiotherapy schedules to a single equivalent dose or specifiedd level of effect Currently the Linear-Quadratic (LQ) cell survival model is used. Fractionationn sensitivity is quantified by the ot/B ratio of the parameters of the LQ model. Bentzenn et al. (2000) concluded from a literature review that the cc/B ratio for normal lung tissue iss low, possibly 2-3 Gy. The physical dose-distribution of a radiation treatment plan is such that locallyy different doses per fraction occur. To take this dose per fraction into account, the physicall dose distribution can be converted into the biologically equivalent dose distribution givenn in fractions of 2 Gy, resulting in the Normalized Total Dose (NTD) distribution, by using thee LQ model (Mactejewskt 1986). A linear time component can be incorporated when the overalll treatment time of the fractionation schedules differs from conventional treatment scheduless (Van Dyk 1990). Two studies (Van Dyk 1989, Dubray 1995) tested for an effect of overalll treatment time and found that tilts was not statistically significant. Therefore, the use of thee LQ model with no time factor at all is probably the best strategy in clinical practice (Bentzen 2000). .
Chemo-- and hormonal therapy
Whenn cytotoxic drugs are combined with other potentially lung-toxic agents, such as radiation, ann enhancement of pulmonary side effects may be observed. In animal studies, cyclophosphamide,, bleomycin and adriamycin were found to enhance radiation-induced lung
IrradiationIrradiation of the kmg: dose-vobme effects
damagee (Collis 1983). In general, the radiation-modifying effect is more pronounced when chemotherapyy is administered concurrently with radiation than if time is allowed to elapse betweenn exposure to each modality (von der Maase 1986). No interaction was observed when thee time-interval between chemotherapy and radiotherapy was longer than 28 days.
AA number of clinical reports (Hrafhkelsson 1987) have indicated an increase in radiation pneumonitiss and fibrosis when chemotherapy and radiotherapy were given concomitantly to patientss with Hodgkin's disease, kmg cancer and breast cancer (Lagrange 1988). No consistentt differences were observed between irradiated patients receiving and not receiving additionall chemotherapy (MOPP and MOPP/ABVD) (Allavena 1992). Morgan et al. (1985) reportedd a lower diffusion capacity for patients treated with MOPP chemotherapy and irradiationn compared to those with radiotherapy atone. Excessive pulmonary toxicity was describedd in series with ABVD chemotherapy and radiotherapy (Brice 1991).
Thee most adverse chemotherapy-radiation interactions appear to arise from the ongoing clinicall breast cancer trials using high dose chemotherapy with autologous bone marrow rescue,, followed by radiotherapy (Marks 1992). For breast cancer patients treated with radiotherapyy combined with less aggressive chemotherapy such as CMF, severe pulmonary toxicityy is rarety noticed (Hardman 1994). For hormonal regiments the results are contradicting.. Theuws et al. (1998b) did not find an influence of tamoxifen on lung function. Bentzen,, however, found an increase of the incidence of (ate radiological lung changes in the irradiatedd apex of the lungs (Bentzen 1996).
Smoking g
Cigarettee smoke containing nicotine and carbon monoxide, has been reported to decrease the diffusionn capacity in smokers due to pulmonary vasoconstriction (Knudson 1989, Sansores 1992).. There are also some indications that smoking may "protect" both tumor (Siemann 1978, Bjermerr 1990) and normal lung tissue (Franzen 1989, Bjermer 1992, 1993, Theuws 1998b) fromm acute radiation effects. The exact mechanism of this effect is not known, but it is suggestedd mat smoking may suppress the local inflammatory reaction in the lung. In a Danish studiee (Jensen 1990), reduced lung function was reported in smokers with respect to non-smokers,, 8 years after radiotherapy.
Age e
Variouss results have been published concerning the impact of age on radiation-induced pulmonaryy changes. Jensen et al. (1990) reported that young patients treated for Hodgkin's diseasee had a more pronounced restrictive lung function 8 years after treatment than older patients.. Hardman et al. (1994), found no association between age and pulmonary damage. Forr radiation pneumonitis however, the risk increases considerably with age (Gagliardt 2000).
Cytokines s
Apartt from the factors mentioned above, individual variability in sensitivity to cytotoxic therapy suchh as radiotherapy and chemotherapy may play a rote as welt. It has been suggested that transformingg growth factor beta (TGF-&), known to pfay an important role in the development off excessive fibrosis, may be useful in predicting an individual patient's risk for developing normall tissue injury (Anscher 1998). This was first reported in breast cancer patients treated withh high-dose chemotherapy and autologous bone marrow transplantation. Anscher et al. (1998)) reported an association between the changes in TGF-R levels and radiation pneumonitis.. Normalisation of the plasma TGF-G level during thoracic radiotherapy for lung cancerr appeared to be useful for identifying patients not at risk for thee development of radiation pneumonitis.. Results from a pilot study with a limited number of patients from Vujaskovic and
ChapterChapter 2
Groenn (2000) suggest that elevated TGF-ft levels during radiotherapy may not only identify patientss with a higher risk of developing pulmonary toxicity but also patients with a higher risk off treatment failure.
Locall effects of lung irradiation
Too determine the effect of lung irradiation, the local damage to lung tissue can be measured withh 3D imaging techniques. Changes in lung density can be measured by comparison of follow-upp CT scans with the CT scan made before treatment. Density changes visible on CT aree caused by an inflammatory reaction of the lung tissue (early response) and fibrosis that generallyy develops later.
Mahh et al. (1987, 1994) conducted a study to obtain a dose-response curve for CT density changes.. An increase of more than 5% in lung density in the high dose region was scored as damage.. They found a probit-like dose-incidence curve with a D ^ of 35.7 Gy (NTD). When the fulll 3D dose-distribution of the applied treatment plan is known, the average local damage in eachh dose-interval can be calculated, leading to a local dose-effect relation for lung density changes.. These effect relations cannot in a simple way be compared with dose-incidencee curves. Dose-incidence curves represent the probability to develop a certain amount off damage and the steepness of the dose-incidence curve represents the inter-patient variance.. Dose-effect relations quantify the average amount of damage in each dose-interval. Forr malignant lymphoma, breast- and lung cancer patients, the dose-effect relation for lung densityy changes is shown in Figure 1 (dotted line).
400 60 Dosee (Gy)
FigureFigure 1. The dose-effect data for perfusion changes for 81 breast cancer and lymphoma patients (triangles) and the wellwell perfused regions of 25 lung cancer patients (squares) in the Netherlands Cancer Institute (NKI, Chapter 3) and the datadata of Duke University (Garipagaoglu 1999) (solid circles), consisting of fifty patients (32 lung cancer, 7 lymphoma, 9 breastbreast cancer and 2 other thoracic tumors). The data could well be described with a logistic function with a Dso of 63 GyGy and a steepness parameter kof 1.7 (soTid line). The dashed line represents the dose^ffect relatbn for CT density changeschanges in the patients of the Netherlands Cancer Institute.
IrradiationIrradiation of the lung: dose-volume effects
AA similar analysis was performed for changes in lung perfusion and ventilation due to radiotherapy.. Three-dimensional perfusion and ventilation scans are obtained by single photon emissionn computed tomography (SPECT) lung perfusion/ventilation imaging. The images can bee matched to the pre-treatment CT scan by means of chamfer matching (Kwa 1998b) or by hand.. To correct for attenuation, a post-reconstruction correction method incorporating a variable-effectivee linear-attenuation coefficient calculated from the spatially correlated CT data wass developed (Damen 1994a). The resolution of the gamma camera used for SPECT is not ideal,, causing blurring of the perfusion/ventilation patterns. This means that perfusion defects thatt are small and sharply outlined cannot be quantified correctly, leading to an underestimationn of the dose-effect in the higher dose-bins (Garipagaoglu 1999). The same appliess for scatter in the patient. However, as long as dose-gradients are shallow and irradiatedd fields are rather large, scatter and blurring will not influence dose-effect relations in thee lower dose bins and only minimally affect the higher dose-bins.
Forr each volume element in the lung the effect of the treatment can be calculated by comparingg the pre-and post-treatment normalized SPECT-values. The dose-effect relations thuss obtained were found to be similar for patients with healthy lungs before treatment (malignantt lymphoma and breast cancer patients) and for the well-perfused lung regions of patientss with NSCLC (Boersma 1994, Marks 1997b). The dose-effect curve for perfusion is almostt linear but can also be described by a logistic function with a Dso of 63 Gy and a steepnesss parameter k of 1.7. The data of the Netherlands Cancer Institute are in accordance withh the data of Duke University (Figure 1). The dose-effect relations for perfusion and ventilationn (the latter not shown) demonstrate a similar, almost linear dose-dependent increase off early pulmonary damage, whereas the dose-effect relation for lung density changes is different,, suggesting the existence of two different biological endpoints, i.e. functional and structurall damage.
Forr lung cancer patients the presence of tumor can result in locally reduced perfusion before treatment.. Tumor regression during treatment can cause reperfusion in some cases. Reperfusionn will have an effect on the measured dose-effect relation. As an example, local reperfusionn can be visualized by correcting the post-treatment perfusion pattern for the dose-effectt (Figure 2).
A.. Pre-RT B. Post-RT C. Post-RT perfusion D.
Dose-perfusionn perfusion corrected for irradiation distribution
FigureFigure 2. Perfusion patterns of a patient with lung cancer. In black the GTVand the king contours are indicated. A. Pre-treatment pattern.pattern. B. Post- treatment pattern. C. Post-treatment pattern corrected for the dose-effect relation as shown in Figure 1. D. Dose-distribution.Dose-distribution. The region posterior of the tumor is better perfused after treatment than before, despite the fact that a high dosedose is applied in this region.
ChapterChapter 2
Overalll effects of lung irradiation
Radiationn pneumonitis, data and models
Thee amount of irradiated volume and radiation dose have an effect on radiation-induced lung injury.. It is important to be able to estimate the severity and probability of developing radiation pneumonitiss for an individual patient. Severe radiation pneumonitis during the first 6 months afterr irradiation may be life threatening. In patients who survive this pneumonitis phase, the typee and severity of the early response may have bearing on the type and severity of the ensuingg late response. The severity of radiation pneumonitis can be scored according to prescribedd criteria, for example according to the SWOG criteria where a grade 1 radiation pneumonitiss is scored when radiological changes are observed but the symptoms do not requiree steroids. Grade 2 is scored when steroids are required. For grade 3 oxygen is needed andd for grade 4 assisted ventilation is required.
Severall theoretical models have been developed to estimate the risk of radiation pneumonitis basedd on the 3D dose distribution. These models should describe the clinical data with sufficientt accuracy. Clinical data on the relation between the 3D dose distribution and the probabilityy of developing radiation pneumonitis were sparse when these models and parameterr values were published. In the following section the clinical data collected since then willl be discussed with respect to the different models.
Thee Lyman mode)
Inn 1991, Emami et al. (1991) estimated that a homogeneous dose of 17.5 Gy given to the wholee lung, or 30 Gy given to 2/3 Of the lung, or 40 Gy given to 1/3 of the lung would result in 5%% probability on a complication within 5 years. For a 50% complication probability these figuress were 24.5 Gy, 40 Gy, and 65 Gy respectively. These doses were not corrected for tissuee inhomogeneities. Subsequently, Burman et al. (1991) applied the model of Lyman (1985)) to these estimations, resulting in parameter values for TD» of 24.5 Gy, a steepness parameterr m of 0.18, and a volume exponent n of 0.87. For an inhomogeneous dose distribution,, the Dose Volume Histogram (DVH) has to be reduced into a single step DVH. This iss usually done according to Kutcher and Burman (1989) using a power-law relationship betweenn tolerance dose and irradiated volume and the same volume exponent n. The group off Michigan (Mattel 1994) discussed that the TO» value should be corrected to 28 Gy. to accountt for tissue heterogeneity. Using this value, they observed a good correlation between thee calculated Normal Tissue Complication Probability (NTCP) and the observed incidence of radiationn pneumonitis in lymphoma patients treated with mantie field irradiation. However, for lungg cancer patients, thé correlation was weak (Martel 1994). Marks et al. (1997a) reported similarr findings: a reasonable correlation was observed between the observed incidence of pneumonitiss and the NTCP (calculated with a TD» of 29.5 Gy) in a study of 100 patients. However,, in the subgroup of the 67 lung cancer patients the correlation was weak, but when patientss with a poor pulmonary function prior to radiotherapy were excluded, the correlation improvedd considerably. In the 60 lung cancer patients studied by Graham et al. (Graham 1994) alsoo a weak correlation between the calculated NTCP (using a TDM of 26 Gy) and the
observedd incidence of radiation pneumonitis was observed. Oetzel et al. (1995), found a good correlationn in a patient group irradiated for lung and esophageal cancer, using the original valuee for TD» of 24.5 Gy. However, apart from differences in the value of TDso there are other differencess between these studies. In the study of Martel et al. (1994) the best correlation was observedd when the lungs were considered to be a paired organ, whereas Oetzel et al. (1995)
IrradiationIrradiation of the lung: dose-volume effects
foundfound the best results when the lungs were considered as independent separate organs. Furthermoree Martel ét al. and Oetzel et al. scored any grade of radiation pneumonitis as a complication,, whereas Graham et at. only considered severe complications (2: grade 3) as a complication.. Finally, Oetzel et at. converted the dose in each voxel to the Normalized Total Dose,, whereas the other groups only used the physical dose.
Thee relative serially model
Gagliardii et at. (2000) characterized the dose-response curve for radiation pneumonitis (grade 11 and higher) following radiotherapy for 68 breast cancer patients. The NTCP model in this studyy was the relative seriality model (Kallman 1992) with parameter values Dso, a steepness parameterr y and a seriality parameter s, which represents the functional seriality of the organ: forr a parallel organs like kidney or lungs s will be dose to zero, for a serial organ like the spinal cordd s is equal to one. The seriality model assumes Poisson statistics to describe cell survival andd the organization of normal tissue in substructures responsible for the organ function. They foundd a D& of 30.1 Gy, a 7 of 0.97 and a relative seriality factor of 0.01, which indicates a strongg parallel behavior of lung tissue. In this study the lungs were considered as separate organs. .
Thee critical volume model
Ass in the relative seriality model, this model is based on the concept of functional subunits (FSUs)) defined either structurally or functionally, and an assumption that normal tissue complicationn probability is fully determined by the number or fraction of surviving FSUs in that organn or tissue (Jackson 1993, Yorke 1993, Niemierko 1993). One can use a sigmoidal function,, characterized by a D50 and a steepness parameter k as the underlying local dose-effectt relation. This sigmoid function represents the FSU-kHI due to locally applied dose. If the dose-effectt relation for perfusion represents the dose-effect relation for FSU-kMI, the average reductionn of local perfusion over the whole lung can be calculated (analogue to the fraction of FSUss killed). A strong correlation was observed between the average reduction in perfusion andd the incidence of radiation pneumonitis (Kwa 1998c). However, the limited number of patientss and the low incidence of pneumonitis in this study did hot allow a reliable comparison withh the other NTCP models.
WinCimm |MiBiinun«
Apartt from fitting the data to the theoretical models, Graham et al. (Graham 1999) investigated whetherr a straightforward parameter as the percentage of the total lung volume that received moree titan 20 Gy was related to the incidence of radiation pneumonitis, for 99 patients irradiatedd for lung cancer (Figure 3A). Univariate analysis revealed the percent of the total lungg volume exceeding 20 Gy (V20), the effective volume (V^, which is the volume that, irradiatedd homogeneously with the prescribed dose, gives the same NTCP as the inhomogeneouss dose-distribution (Ten Haken 1993)) and the mean King dose, and location of thee primary tumor (upper versus tower lobes) to be correlated with the development of a Grade 22 pneumonitis. Multivariate analysis revealed the V20 to be the most significant predictor of radiationn pneumonitis. Simitar findings were reported by Marks et al. (1997a) for the lung volumee receiving more than 30 Gy. Armstrong et al. (1997) found a higher incidence of grade 33 radiation pneumonitis for patients with more than 30% of their lungs irradiated with more than 255 Gy compared to patients with a smaller V25.
Anotherr rather straightforward parameter as the mean dose seems to be correlated with the incidencee of pneumonitis as well. In the study of Martel et al. (1994), the mean lung dose in thee 5 lymphoma patients with complications was 26.1 Gy, whereas the mean dose in the 16
ChapterChapter 2
lymphomaa patients without complications was 21 Gy. For the 9 lung cancer patients with complicationss and the 33 patients without complications these figures were 24 Gy and 18 Gy, respectively.. These differences were even larger when the lungs were considered to be independentt separate organs (30 Gy vs 21.3 Gy, and 34.2 Gy vs 18.2 Gy, respectively). In the studyy of Oetzel et al. (1995) these differences were somewhat smaller: the mean physical lung dosee in the 37 lung cancer patients without complications was 20.1 Gy, whereas this figure wass 23.8 Gy for the 9 patients with pneumonitis.
100 20 30 A.. V20 (%)
100 20 B.. Mean lung dose (Gy)
FigureFigure 3. A. The incidence of radiation pneumonitis (grade >2) as a function of the volume receiving more than 2020 Gy for 99 lung cancer patients (Graham 1999). B. The incidence of radiation pneumonitis (grade >2)asa function ofof mean (bbtogical) lung dose. The data of 540 patients from 5 institutions were pooled (Kwa 1998a). The error bars representrepresent the 68% confidence intervals. The fit parameters were TDsr30.5 Gy, and m=0.30.
Inn a large multi-center study (Kwa 1998a), 540 patients were pooled to study the dose-incidencee curve for radiation pneumonitis. Three-dimensional dose calculations were performedd with the same type of inhomogeneity correction (equivalent-path-length method in 44 of the 5 centers). Dose-volume histograms were calculated for the lungs as one organ, from whichh the mean (biological) lung dose, MLD, was obtained. Radiation pneumonitis was scored whenn a complication grade 2 (SWOG) or higher occurred. The mean lung dose, MLD, ranged fromm 0 Gy to 34 Gy and 73 of the 540 patients experienced pneumonitis. In all centers, an increasingg pneumonitis rate was observed with increasing MLD. Between the lung groups of differentt institutes, significant differences were present. Accounting for these differences by addingg center dependent offset values for lung cancer patients, the data was fitted with an TD50=30.5+1.44 Gy and 2 1 SE) for all patients, and an offset of 0-11% for the lungg group, depending on the center. In Figure 3B the dose-incidence curve for radiation pneumonitiss is shown, corrected for the center-dependent offset. The MLD, which is relatively easyy to calculate, can be used to predict the risk of radiation pneumonitis, grade > 2. About halff of the complications grade > 2 were complications grade > 3 in all dose-bins.
ComparisonComparison ofDVH reduction techniques and simple parameters using the critical volume model Theree are differences and similarities of the different parameters that are predictive for radiationn pneumonitis. In Figure 4A these parameters are shown, based on a DVH of a lung cancerr patient: the cumulative DVH is plotted with the V20 and the V30, (which are easily derivedd from the DVH), the mean lung dose and the equivalent uniform doses. The equivalent
IrradiationIrradiation of the lung: dose-volume effects
uniformm dose (EUD) has a more general definition than the definition used by Niemierko (1997). .
Wee defined the EUD as the equivalent dose which, according to the underlying theoretical model,, when given as a uniform dose to the entire organ would produce the same NTCP as thee original inhomogeneous dose distribution. Therefore, this EUD is not yet uniquely defined ass it still depends on the underlying theoretical model. All the parameters in Figure 4A can be obtainedd by reducing the DVH to a single step-DVH using an underlying local dose-effect relationn representing dose-dependent local damage or FSU kill. When the MLD is used as a
100% % 80% % ffi ffi ii 60%
I I
40% % 20% % 0% % -- Cumulalive DVH MLDD = 236y V200 = 44% __ _ V30 = 30% EUDD <n=0.25) = 44 Gy EUD(D500 = 45Gy.k=2.5) = 27GyV V
: : ii ^ i i i i A A 100% % '' Linear (Power-taw n=1) Stepp (V20) -- — Step(V30) Power-toww (n=0.25) Sigmoidd (D50=45 Gy, k=2.5> 200 40 60 Dosee (Gy) 80 0 40 0 Dosee (Gy)FigureFigure 4. A. The cumulative DVH of a king cancer patient with the V20 and the V30, (which are easily derived from the DVH),DVH), the mean king dose, the equivalent uniform dose based on the KutcherDVH reduction scheme (EUD^,^ and basedbased on a local sigmoid dose-effect relation (EUD3^. B. Different dose^ffect relations that are used to derive the singlesingle parameters in Figure 4A, which are predictive for radiation pneumonitis.
predictorr for radiation pneumonitis, a linear dose-response for the underlying local damage is assumed,, while for the V20 and V30 a step-function is taken as the local dose-response curve. Inn the Kutcher DVH reduction scheme, a power-law relation, (characterized by n) between the locall dose and the effect is assumed (Kwa 1998c), resulting in an EUDpo^ertaw or EUDLKB (Appendixx I, equation 7), which is the same as the D ^ defined by Mohan (1992), D ^ accordingg to Damen (1994b) and the EUD defined by Niemierko (1999). The underlying dose-effectt relations are shown in Figure 4B. When n=1 in the Kutcher model, the E U D , » ^ ^ will bee equal to the MLD (Kwa 1998c). If a sigmoid logistic expression (characterized with a D^ andd a k) for the local dose-effect relation is used, an EUDsigri,0|d can be calculated as well (Appendixx I, equation 12). When the dose-effect relation for FSU-kill is linear (which is almost thee case when perfusion data is used, Figure 1), the EUD-ag^c will be equal to the mean lung dose. .
Alll the models can thus be compared using the EUDmodel. From the step-dose-effect relations theoreticallyy no EUD can be calculated. The V20 and V30 are then the analogies of the EUDVx.
ChapterChapter 2
Reductionn in overall lung function; dose-volume parameters
Sincee the mean lung dose can be used to predict the average reduction in perfusion over the completee lung, Theuws et al. (1998b) investigated whether this parameter could be used to predictt the reduction in overall lung function. For 81 patients (41 lymphoma patients and 40 breastt cancer patients) the correlation coefficients were calculated between the mean lung dosee and the change in overall lung function parameters, measured prior to radiotherapy and 3-44 months after therapy (Figure 5). The correlation coefficients between the mean lung dose andd the reduction of the Alveolar Volume (VA), Vital Capacity (VC), Forced Expiratory Volume inn 1 second (FEV,) and Transfer Factor for Carbon Monoxide (TLiCOc), were 0.73, 0.70, 0.69 andd 0.58 respectively. 50 0 40 0
>'1 1
££ 5II I
30 0 20 022 1
TOTO u _22 2
S S II 10 -10 0 -20 0 3 months 18 months Is? ? I I ^ è è""^lll
10 0 15 5 20 0 25 5 30 0 Meann lung dose (Gy)FigureFigure 5. Comparison of the mean relative changes in VA SEM, 3 and 18 months after radbtherapy for 110 breast
cancercancer and and malignant lymphoma patients as a function of the mean lung dose. The plotted lines represent the dose-dependentdependent reduction in VA at 3 months (0.9% per Gy) and 18 months (0.4% per Gy).
Thee relation between the mean lung dose and the reduction in overall lung function parameterss could be described with one regression line through the origin and a slope of % reductionn in overall lung function for each increase of 1 Gy in mean lung dose. For patients treatedd with MOPP/ABV chemotherapy prior to radiotherapy, the overall lung function was
7-12%% lower prior to the start of radiotherapy than for patients without treatment before radiotherapy.. CMF chemotherapy given after radiotherapy caused an additional decrease in TL.COCC of 6% which should be added to the estimated radiation-induced reduction of TLCOc- The groupp from the Duke University (Marks 1997a), investigated whether the reduction in FEV-, andd TL C O c was correlated with the lung volume irradiated to > 30 Gy or with the estimated NTCPP values for lung cancer patients. In a study of 100 patients only a poor correlation was foundd between these parameters. However, when patients with a poor pulmonary function priorr to therapy were excluded, the correlation improved.
Irradiator}Irradiator} of the lung: dose-volume effects
Recoveryy of radiation induced pulmonary damage
Severall studies have been performed to evaluate the effects of radiotherapy on lung tissue overr time. Pulmonary function tests (VA, VC, FEV^ and TL,Coc). SPECT perfusion and ventilationn scans and CT scans were made before, 3,18, and 48 months after radiotherapy for 1100 patients treated for breast cancer or malignant lymphoma (Boersma 1996). Dose-effect relationss for changes in local perfusion, ventilation, and lung tissue density were determined forr each individual patient for each follow-up period. Average dose-effect relations for both subgroupss were determined, as well as dose-effect relations for different regions. In general, partiall improvement of local pulmonary injury was observed between 3 and 18 months for eachh of the three endpoints. After 18 months, no further improvement was seen.
Patientss with breast cancer and malignant lymphoma showed a similar improvement, which wass attributed to a recovery from the early radiation response and could not be explained by contractionn effects of fibrosis of lung parenchyma. Also in the pulmonary function tests, a partiall recovery from the radiation induced lung damage was observed between 3 and 18 monthss after treatment. Early and late reduction in VA, VC, FEV^ could be estimated before radiotherapyy for breast cancer and lymphoma patients based on the mean radiation dose (Figuree 5), whereas for the reduction in TL C O c also the chemotherapy regimen should be taken intoo account (Theuws 1999). After 18 months no further recovery could be observed.
Discussion n
Forr dose-escalation studies it is important to be able to estimate the radiation pneumonitis risk forr each patient individually, based on the 3D treatment plan. Lung tissue shows a strong parallell behavior and a strong volume effect when irradiated. This is reflected in the value of thee parameters in the models predicting the risk of radiation pneumonitis. The volume
1 1 0.8 8 0.6 6 Q--z Q--z 0.4 4 0.2 2 0 0 ( ( Dosee (Gy)
FigureFigure 6. Probability of radiation pneumonitis (grade > 2) versus dose for irradiation of the whole lung, 213 and 1/3 of thethe lung according to the Lyman model (solid line) with the parameter values D50 = 30.5 Gy, m = 0.3 andn=1 (Kwa
1998c).1998c). For the seriality model (dashed line) the seriality parameter sis 0.01 (Gagliardi 2000) whereas the Dso and the steepnesssteepness parameter were taken as in the Lyman model to facilitate comparison between the models.
ChapterChapter 2
parameterr in the Lyman model n was equal to one, whereas in the seriality model the seriality parameterr was very small (s = 0.01). The dose-effect relations resulting from those two models forr the whole lung, 2/3 and 1/3 of the lung are shown in Figure 6. Both models show a comparablee volume effect. Parallel behavior with a volume parameter n equal to one has also beenn demonstrated for other parallel organized tissues like the parotid gland (Eisbruch 1999, Moerlandd 2000) and recently also for liver (Ten Haken 2000).
70% %
100 15 20
Meann Lung Dose (Gy)
25 5 30 0
FigureFigure 7. Scatter-plot of the mean lung dose (MLD) and the volume irradiated with more than 20 Gy (V20) of clinical treatmenttreatment plans of 3 different patient groups of the NKI: breast cancer patients (circles), malignant lymphoma patients (squares),(squares), NSCLC patients (triangles). The correlation between the two parameters is high, r2
= 0.75 and is even higher
whenwhen the lymphoma patients are excluded (r1
= 0.9).
Becausee the models to predict the incidence of radiation pneumonitis give similar results, it is difficultt to make a decision about which model is best suited for clinical use. Of the simple parameters,, the V20 (or V30) and the mean lung dose represent the two extremes in the underlyingg dose-effect relations while the more complicated models use underlying dose-effect relationss that are more or less in between these extremes (Figure 4). The ultimate decision of thee 'best' underlying local dose-effect relation should be based on the analysis of patient data. However,, clinically applied treatment plans in the Netherlands Cancer Institute show a high correlationn (r = 0.94) between the V20 and the MLD (Figure 7). Because the two parameters aree not independent in our data set, it is not possible to test the models independently. For example,, because of the correlation between the V20 and the MLD, a certain treatment plan willl result in the same incidence of radiation pneumonitis, regardless of the use of V20 (Figure 4A)) or the MLD (Figure 4B). The shape of the mantle field irradiation of the lymphoma patients withh a low dose, but a large field-size, yields some variation in the distribution of MLD and V20, butt still more data is needed both in the region where a large part of the lung is irradiated with moree than 20 Gy and a low MLD, and in the region where a high mean lung dose coincides withh a small V20 to be able to distinguish between the models. It is quite possible that when moree clinical data becomes available, the true dose-effect relation will appear to be somewheree in between the MLD and V20, like a sigmoid local dose-effect relation. Hopefully
IrradiationIrradiation af the king: doss-volume effects
withh the ongoing dose-escalation studies additional data will be collected in the 'empty* regions off the MLD/V20 scatter-ptot
Pre-treatmentlungPre-treatmentlung function
Markss et al. (1097a) proposed to take into account both the pre-treatment lung function and thee 3D dose distribution in the design of the optimal treatment plan, by calculating Functional Dosee Volume Histograms (fDVH). These fDVHs were based on pre-treatment SPECT perfusionn data. This could result in beam setups that minimize the incidental irradiation of functioningg tissue. Although this approach certainly may have theoretical advantages, ft is basedd on the assumption that nonfunctioning areas do not and will not contribute to tine overalll lung function. However, in many patients reperfuston is observed adjacent to the tumor (Markss 1995) due to tumor regression. This probably explainss why in another paper of Marks ett al. (1997a) the prediction of radiation pneumonitis was not improved using the fDVHs comparedd to the conventional DVHs.
Conclusions s
Inn the analysis of the effect of irradiation on the lungs, many known and yet unknown factors playy a role. Because óf the complexity of the problem, the data are still inconclusive with regardd to an appropriate description of the dose-volume effect Therefore additional patient dataa is needed to adequately perform multi-variate analyses. Dose-escalation studies and multi-centerr cooperation will create more possibilities to investigate all confounding factors concerningg lung irradiation.
Chapterr 3
Radiationn dose-effect relations and
locall recovery in perfusion for patients
withh non-small-cell lung cancer
Yvettee Seppenwookte, Sara H. Mutter, Jacqueline C M . Theuws, Paul Baas, Joséé SA. Belderbos, Uesbeth J. Boersma, Joos V. Lebesque
Radiationn dose-effect relations and local recovery in
perfusionn for patients with non-small cell lung cancer
Too determine local dose-effect relations for king perfusion and density changes due to irradiationn for patients with noo-smal cell lung cancer (NSCLC) and to quantity the effect off reperfusion, registered Single Photon Emission Computed Tomography (SPECT) lung perfusionn scans and CT-scans were made, before and alter radiotherapy lor 25 NSCLC-patientss and a reference group of 81 patients with healthy lungs. Average dose-effect relationsrelations for perfusion and CT-density changes were calculated and compared with the dose-effectt relation of the reference group. On the basis of these dose-effect relations, tf» post-RTT perfusion was predicted for each patient and compared to the measured post-RT perfusion.. WeH-perfused rung regions of the NSCLC-patients showed the same dose-effectt relation as the reference patients. By comparing predicted and measured post-treatmentt perfusion scans, regions of reperfusion could be determined for 18 of 25 NSCLC-patientss but for none of the reference patients. Well-perfused lung tissue of patiëntee with NSCLC behaves like healthy lung tissue with respect to radiation. The dose-effectt relation for perfusion and CT-density was extended for doses up to 80 Gy. Radiation damagee in poorly perfused lung regions was less than predicted as a consequence of locall reperfusion.
Introduction n
Whenn optimizing treatment plans for lung cancer patients, lung tissue is one of the dose-limitingg structures, besides the spinal cord, the esophagus and the heart Although the mean lungg dose can be used to predict Normal Tissue Complication Probabilities (NTCP's) (Kwa 1998a),, it might be better to include functional information of lung tissue to design the plan that minimizess the complication risk (Marks 1993, 199$). Single Photon Emission Computed Tomographyy ( S P E C T ) lung ventiiation/perfusion scans provide information in three dimensionss about local functionality of lung tissue.
Thee effect of inhomogeneous dose distributions on lung perfusion for patients with healthy lungss (malignant lymphoma and breast cancer patiënte) can be predicted using an average dose-effectt relation for the whole group (Theuws 1998a). in that patient group it was assumed thatt all observed damage was due to irradiation. However, for patients with intra-thoracic tumorr or for patients with pre-existent lung disease, prediction of post-radiotherapy <RT) lung functionn can be inadequate because not all changes will be caused by the irradiation, i.e. tumorr progression can also induce lung damage.
Patientss with non-small cell lung cancer (NSCLC) can have inhomogeneous perfusion already beforee RT due to several causes. Patients with chronic obstructive pulmonary disease (COPD) aree more likely to have a tower CT-density throughout the lungs, and their lung perfusion is inhomogeneous.. Some patients suffer from bullous disease and have lung regions where locallyy the density is very tow. These bullous regions are neither well ventilated nor perfused. Furthermore,, the perfusion can be obstructed by atelectasis, infiltrates and, most importantly, tumor. .
Iff reduced perfusion is caused by compression of an artery due to a (large) tumor, tumor regressionn can induce re-opening of these btood vessels which may cause a part of the lung, distall from the tumor, to re-perfuse. Therefore, in the high dose regions two opposite effects
ChapterChapter 3
cann be induced by the treatment: on one hand the radiation causes damage resulting in a decreasee in perfusion and an increase of tissue density; on the other hand, tumor-regression mayy cause a 'recovery* in perfusion. When mis local recovery takes place in a lung region mat iss also damaged by the irradiation, the observed damage will be less than expected on the bastss of the applied radiation dose.
inn this paper we investigated whether the local dose-effect relation for NSCLC-patients is the samee as for breast cancer and malignant lymphoma patients, we determined the dose-effect relationn for higher doses and found a way to visualize and quantify local recovery effects.
Methodss and Materials
Patients s
Thee 25 patients in this study were irradiated for non-small cell lung cancer. All patients had a follow-upp of three months to assess changes in lung perfusion and tissue density. Six patients weree classified as stage I NSCLC, two patients as stage II, seven as stage IIIA and ten as stagee NIB. All patients were irradiated with conformal AP-PA fields and a boost up to 70 Gy in fractionss of 2 Gy, except for two patients who got 51 Gy in fractions of 3 Gy. All patients were scannedd and treated in supine position with the arms raised above their head in a forearm supportt No further immobilization was used but all possible effort was made to reproduce the treatmentt position during the different scans. Eighty-one malignant lymphoma and breast cancerr patients from a previous study (Theuws 1998b) were used to provide a reference set forr evaluating the current results. These patients had healthy lungs and good pulmonary functionn test results. The presence of intra pulmonary tumor in the lymphoma group did not affectt the dose-effect relation.
Thee local hospital ethics committee approved the study, and before the patients were included, writtenn informed consent was obtained.
Dosee calculation
CT-basedd dose calculations were performed as described previously (Boersma 1994), using aa 3D treatment planning system (U-MPIan, University of Michigan) with tissue inhomogeneity correction.. For inhomogeneous dose distributions, the dose per fraction largely differs for differentt regions of the lungs. To take this dose per fraction effect into account, the physical dosee distribution was converted into the normalized total dose (NTD) distribution, using the linearr quadratic model with an ct/p* ratio of 3.0 Gy (Van Dyk 1989, Newcomb 1993). The NTD iss defined as the total biological equivalent dose given in fractions of 2 Gy (Maciejewski 1986). Alll radiation doses in the data presented in this paper (local dose, mean dose, etc.) are biologicall equivalent doses.
Dataa acquisition
Beforee radiotherapy, and three to four months after radiotherapy, a SPECT lung perfusion scan andd a GT thorax scan were obtained. For SPECT, a dual-head gamma camera (ADAC Genesyss or ADAC Vertex) was used equipped with medium-energy general-purpose collimators.. After administration of about 4 mCi of 99mTc-macroaggregated albumin to the patientt in supine position, SPECT lung perfusion scans were made (scan time: approximately 155 minutes). For three patients ventilation was measured simultaneously using the dual-isotopee acquisition mode and 81fnKr. The SPECT scans were reconstructed using filtered back projectionn with software provided by the manufacturer (ADAC). The number of voxels was
RadiationRadiation dose-effect relations and heal recovery
64x64x644 and the voxel size was approximately 6x6x6 mm3. The resolution of the reconstructedd SPECT images was 20-25 mm (full width at half maximum), as was measured withh a line source filled with " T c . The CT scan was made within one week of the accompanyingg SPECT scans, with the patient in the same position (scanner Siemens, Somatomm Pius). Both CT and SPECT acquisition were performed under normal breathing conditionss (no "breath-artd-hold" procedure) and included the entire lung volume. Five external skinn positions were marked with 57Co point sources during SPECT scanning, and with crossed radio-opaquee catheters during CT scanning.
Lungg contour matching
Too obtain lung contours, the CT images were segmented by binary thresholding. The threshold valuee was chosen at a density of 0.7 g/rhi. The Gross Tumor Volume (GTV) delineated by the radiationn oncologist was excluded. For correlation of the CT and SPECT scans, chamfer-matchingg (Kwa 1998b) was applied on the lung contours. To obtain lung contours from SPECT, thee lung perfusion scans were segmented by binary thresholding as well, using an initial thresholdd of 200 counts that was adjusted during matching with the lung contours in the CT-scann until the best fitting threshold was obtained. Because the patients were CT-scanned white continuouslyy breathing, contours in the slices near the diaphragm were quite different from scann to scan. These contours were manually omitted from the post-treatment scan before matching.. After correlating the SPECT lung perfusion scans with the CT images, a first order Chang-likee (Chang 1978) attenuation correction (Damen 1994a) was applied on the SPECT perfusionn scans, based on CT-densrty in homogeneities. The quality of the chamfer match was visuallyy evaluated and in case the lung contours did not correlate well, the five skin markers weree used to align the SPECT-scans with the accompanying CT-scan (in only 3 of the 50 SPECT-CTT matches we used the skin markers). These five skin positions were identified visuallyy in the SPECT and CT-scans, and the root-mean-square distance between the correspondingg markers was minimized by allowing translations and rotations.
Dose-effectt relations
Thee dose-effect relations for local changes in perfusion and lung density (which is quantified byy changes in air-filled fraction1) for each individual patient were determined by calculating the reductionn in function compared to the pre-RT function for each dose interval (Boersma 1993). Beforee the dose-effect relation of an individual patient could be determined, voxels with an inaccuratelyy measured effect should be excluded. In all data sets, voxels positioned near the diaphragmm were excluded because of the poor image correlation due to breaming movements. Voxelss with a high dose gradient (more than 10 Gy/cm), located at field and block edgess were alsoo excluded from the analysis because small spatial inaccuracy in matching of the data sets couidd cause a considerable uncertainty in dose in these voxels. To avoid further mismatch errors,, the peripheral king region (one voxel thickness) was excluded from the analysis as well.. On average 40% of the lung volume was excluded for the NSCLC-patients. To limit the effectt of this large excluded volume, a threshold for the minimum number of voxels in a dose-intervall could be estimated from variance analysis. For each patient a minimum of 30 voxels perr dose interval was needed to assure that the intra-patient variance was less than 20% of thee total variance.
1Theair-«edfcac8onisdefirieda$thevolun»of f
B»fotawinBrelaöorck=(1^)=^-0O^T/w w inn HouwSeld unite.
ChapterChapter 3
Normalization n
Whenn comparing multiple SPECT scans, internal normalization is required because the perfusionn pattern through the lungs is altered after treatment and the amount of injected activityy that reaches the lung capillaries is not known exactly. In general, normalization is appliedd on the average number of SPECT counts in the parts of the lungs receiving less than aa certain low dose (in our study 8 Gy), assuming that such a low radiation dose has no significantt influence on perfusion resistance in that region. Due to the parallel structure of lung tissue,tissue, damage in high dose regions leads to compensation effects in the rest of the lungs and normalizationn on the average number of counts in the tow dose area yields the correct representationn of the damage in other dose bins. However, due to possible local recovery of perfusionn capacity in the normalization area, damage in the higher dose bins may be overestimated.. Therefore, normalization was applied on welt-perfused low dose regions of the lung,, assuming not only that lung injury is not present at this tow dose level but also that well-perfusedd areas are less influenced by tumor-regression. Formulas used tor the normalization proceduree can be found in Appendix II (Equations 6-9). The difference between the two normalizationn methods was tested for the reference patient group. For changes in CT-densrty, normalizationn on well-perfused low dose regions was applied to correct for changes in breathingg level; although in the low-dose regions the tissue-density will not increase due to treatment,, it is possible mat the patient at the follow-up time breathes at a different level. Lung tissuee density changes with the amount of inhaled air. For the construction of an average dose-effectt relation, tile normalized effect in each dose interval was averaged over all patients or patientt subgroups. This averaging was performed logarithmically (geometric mean, because thee effect is a ratio of a pre- and post-RT value) and giving equal weights to each patient Dose-effectt (DE) relations were calculated for thé entire patient group and for patient subgroups: :
DEmm is the average dose-effect relation for all patients (n=25);
DEhomm is the average dose-effect relation for patients with homogeneous perfusion ; throughoutt the lungs (without decreased perfusion adjacent to the tumor before irradiation,, n=6);
DEtnhomm is the average dose-effect relation for patients having reduced perfusion likely due too obstruction of vessels by the tumor (n=19);
DEwpp is the average dose-effect relation for well-perfused (>60% of the maximum pre-RT perfusion)) areas of alt patiënte (n=25) and
DEppp is the average dose-effect relation for poorly perfused (<30% of the maximum pre-RTT perfusion) areas of all patients (n=25).
3-DD evaluation of local recovery
Too separate the effect of the irradiation and the opposite effect of local recovery, we constructedd a prediction of the post-RT SPECT lung perfusion scan, based on the pre-RT perfusion,, the 3-D dose distribution and the dose-effect relation for healthy lung tissue (see Figuree 1 and Equation II-23). The average dose response curve for patients with malignant lymphomaa and breast cancer (Theuws 1998a) could be combined with the dose-response for thee welt-perfused lung regions of the NSCLC-pattents, yielding DE^patavvprwdc By applying thiss extended average dose-effect relation to the pre-RT perfusion scans, the post-RT function off the lung could be predicted, based on the individual dose distribution and the individual functioningg of the lung before radiotherapy. The scans constructed in this manner could be comparedd with the measured post-RT perfusion SPECT-scans, resulting in a local recovery image;; for each voxel the ratio between the measured and the predicted post-RT perfusion
RadiationRadiation dose-effect relations and local recovery
wass calculated. This method gave us the possibility to differentiate (in three dimensions) betweenn regions of the lung which react as predicted on the radiation dose distribution and regionss which react stronger or less strong than predicted.
3DD dose distribution Perfusionn pre-RT Averagee dose-effect relation Predictedd post-RT perfusion n
FigureFigure 1. Construction of predicted post-RT perfusion, based on the individual 3D-dose distribution, the average dose-effecteffect relation of the reference patients and the well-perfused regions of the patients with NSCLC (0E«/:p8f4APnscfc) and
friee individual perfusion pre-RT for each patient. For each voxel, the reduction in perfusion based on the local dose in
thatthat voxel is calculated, using DE^^oiNPna*- The
resulting perfusion is represented in the predicted post-RT perfusion image. image.
Quantificationn of local recovery effects
Too evaluate whether NSCLC-patients show a clinically relevant functional increase in average perfusionn throughout their lungs due to local recovery (reperfusion), the average measured andd average predicted function loss (see Appendix II, Equations 21 and 30) over the lungs was calculatedd for each patient. The amount of reperfusion is given by the difference between the averagee measured and predicted function loss (Equation II-33).
Results s
Thee influence of the new normalization method on the dose-effect relation was tested for malignantt lymphoma and breast cancer patients. For the dose-effect relation normalized on well-perfusedd low dose regions, a difference of less than 1 % was present. The change in fit-parameterss was not statistically significant; consequently, the dose-effect relation can be fitted withh the same sigmoidal curve as before (Dso = (54.7 1.2) Gy and k = 2.2 0.2 (Theuws 1998a)). .
8 0 ,, , 80
FigureFigure 2. Doseeffect relations for perfusion (A) and the air-filled fraction (B) for 25 NSCLC-patients. The solid fine (extrapolation(extrapolation 'is dashed) represents the dose-effect relation (logistic fit) for the reference patients
